BEARING CAPACITY OF ROADS, RAILWAYS AND AIRFIELDS

PROCEEDINGS OF THE 10TH INTERNATIONAL CONFERENCE ON THE BEARING CAPACITY

OF ROADS, RAILWAYS AND AIRFIELDS (BCRRA 2017), ATHENS, GREECE, 28–30 JUNE 2017

Bearing Capacity of Roads, Railways

and Airfields

Editors
Andreas Loizos
National Technical University of Athens, Greece

Imad L. Al-Qadi
University of Illinois at Urbana-Champaign, USA

A. (Tom) Scarpas
Delft University of Technology, The Netherlands
CRC Press/Balkema is an imprint of the Taylor & Francis Group, an informa business

© 2017 Taylor & Francis Group, London, UK

Typeset by V Publishing Solutions Pvt Ltd., Chennai, India

Printed and bound in Great Britain by CPI Group (UK) Ltd, Croydon, CR0 4YY

All rights reserved. No part of this publication or the information contained herein may be reproduced,
stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, by
photocopying, recording or otherwise, without written prior permission from the publisher.

Although all care is taken to ensure integrity and the quality of this publication and the information
herein, no responsibility is assumed by the publishers nor the author for any damage to the property or
persons as a result of operation or use of this publication and/or the information contained herein.

Published by: CRC Press/Balkema

Schipholweg 107C, 2316 XC Leiden, The Netherlands
e-mail: [email protected]
www.crcpress.com – www.taylorandfrancis.com

ISBN: 978-1-138-29595-7 (Hbk + USB-card)

ISBN: 978-1-315-10033-3 (eBook)
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Table of contents

Preface xxiii
Organization xxv
Supporting organizations xxix

Subgrade soils
Evaluation of a resilient modulus model for unsaturated soil conditions 3
P. Thirthar Palanivelu & C.E. Zapata
Investigating correlations between stiffness indexes for Brazilian soils 11
M. dos Reis & R. Moura Fortes
Analysis of compaction and California Bearing Ratio (CBR) test results of a mixture of iron
ore tailings with Brazilian tropical soil in different proportions for road construction purposes 21
A.A.N. Dantas, T.A. Mendes, A.J.A. Posse, L.F.M. Ribeiro & C.M.C. Gurjão
Suitability assessment of soils for pavement subgrade using gyratory compaction and
bearing capacity testing 27
A. Athanasopoulou & G. Kollaros
Experimental and statistical investigation of the mechanical properties of limestone rocks
in Lebanon 33
M. Mrad, G. Saad, G.R. Chehab & R. Mrad
Application of pattern classification techniques for anisotropic characterization
of pavement foundations 39
R.S. Ashtiani & M. Asadi
Plastic strains of a compacted clayey soil under repeated axial loading 49
Z.-S. Li & J.-M. Fleureau
Estimation of shakedown loads of flexible pavements 57
A.G. Stathas & K.V. Spiliopoulos
Bearing capacity on stockpiled fly ash 65
C.A. Lenngren
Tunnels interaction with surrounding alluvial soils 71
O. Naeemifar & R. Rahbari

Unbound granular materials

Structural characteristics of unbound aggregate materials in conventional flexible pavements 79
L. Wang & X.G. Xie
Investigation of resilient modulus values for base aggregates 85
B. Cliatt & A. Loizos
Unsaturated resilient strain behaviours of a granular material 93
P. Jing, C. Chazallon & H. Nowamooz
Laboratory investigation on unbound materials used in a highway with premature damage 101
D.M. Barbieri, I. Hoff & H. Mork

v
An investigation into relevance between fractal dimension and dynamic behavior
of pure granular materials 109
S. Altun, A. Sezer, A.B. Göktepe, T. Günay & P. Ahmedzade
Experimental characterization of unbound granular materials subjected to high
harmonic loads 117
G. Canon Falla, S. Leischner, F. Wellner & T. Spanier
Numerical simulation and laboratory testing of unbound base course materials considering
the effect of temperature 125
J. Patzak & F. Wellner
Some recent research on the hydraulic conductivity of road materials 135
P.J. Vardanega, S. Feng & C.J. Shepheard
Laboratory investigations of thermal properties of crushed rock materials 143
K. Rieksts, I. Hoff, E. Kuznetsova & J. Côté
Stiffness and permanent deformation characteristics of open-graded unbound granular materials 151
M.S. Rahman, S. Erlingsson & F. Hellman
Characterisation of permanent deformation behaviour of unbound granular materials
using repeated load triaxial testing 159
N. Zhalehjoo, A. Tolooiyan, R. Mackay & D. Bodin
Implementation of shakedown and packing theories for unbound granular materials 167
T.F. Yideti, A. Dawson & B. Birgisson

Asphalt mixes
Experimental study on the effects of angular rate on the torsional shear property
of asphalt mixture 179
Y. Li, Y. Li, G. Wang, S. Zhang, Y. Liu & J. Chen
Comparison of flexural strength and crack propagation test characteristics of SMA mixtures
including the ageing impact 185
P. Vacková, J. Valentin & P. Mondschein
Computational study of the influence of form and angularity of coarse aggregates
in the linear viscoelastic properties of asphalt mixtures 193
D. Castillo & S. Caro
Linear viscoelastic behaviour of bituminous mixtures with multi-Recycled Asphalt Pavement 201
A. Pedraza, H. Di Benedetto, C. Sauzéat & S. Pouget
Apps for modal analysis to characterize the complex modulus of asphalt concrete 209
A. Gudmarsson & N. Ryden
An implementation of the Illinois flexibility index testing protocol for balanced
asphalt mix designs 217
M.K. Barry, I.L. Al-Qadi, H. Ozer & F. Safi
Laboratory evaluation of stiffness and fatigue susceptibility of asphalt paving materials
incorporating environmental factors 223
A. Aljuboryl, G.D. Airey & J.R.A. Grenfell
Laboratory test and numerical simulation of microwave heating properties of asphalt mixture 231
H. Wang, P. Apostolidis, X. Liu, T. Scarpas, J. Yang & L. Xu
Influence of lateral confining pressure on flow number tests 237
E. Santagata, O. Baglieri, P.P. Riviera, M. Lanotte & M. Alam
Degradation of asphalt mixtures due to impact method of compaction 243
P. Singh & A.K. Swamy
Evaluation of mixing temperature impact on warm mix asphalt performance 251
A. Abed, N. Thom & J.R.A. Grenfell

vi
Effect of compaction temperatures on the warm mix asphalt volumetrics and stability 259
H.I. Ozturk & O.C. Pamuk
Comparison of DSR and BBR tests for determining the Performance Grade (PG) of asphalt
binder at low temperature 267
C. Riccardi, A. Cannone Falchetto, M.P. Wistuba & M. Losa
Aging of bituminous binders in asphalt pavements and laboratory tests 273
X. Lu, H. Soenen & O.-V. Laukkanen
Effects of mineral fillers and bitumen on ageing of asphalt mastics properties 281
R. Alfaqawi, G.D. Airey & J.R.A. Grenfell
Development of empirical models for the estimation of the rheological properties
of asphalt binders 289
F. Martinez, M. Cauhape Casaux & S. Angelone
An alternative method for determining thermal stress in asphalt binder based
on Laplace transform 297
A. Cannone Falchetto, C. Riccardi, D. Wang, M.P. Wistuba & K.H. Moon
Laboratory investigation of physical performance and FT-IR analysis of warm mix asphalt binders 305
J. Chen, W. Wang & B. Xu
Study of asphalt binder diffusion using atomic force microscopy 311
A. Savarnya, A. Kuity, S. Gupta, A. Chandra & A. Das

Performance of asphalt mixes with additives

Performance properties of asphalt mixture containing Linz-Donawitz (LD) steel slag 319
J. Groenniger & M.P. Wistuba
Feasibility study on a thermoset polymer-coated emulsified warm-mix asphalt mixture 327
P.J. Yoo, B.S. Ohm, K.S. Park & I.L. Al-Qadi
Effect of new types of synthetic waxes designed for low-temperature asphalt mixtures 335
L. Benešová, J. Mastný & J. Valentin
An investigation of the mechanical properties of rubber modified asphalt mixtures using
a modified dry process 343
A. Subhy, G.D. Airey & D. Lo Presti
Rutting performance of bituminous mixtures composed with red mud 349
M.S.S. Lima, L.P. Thives & V. Haritonovs
An experimental study on waxy bitumens 357
J. Oner, B. Sengoz & G. Malkoc
Comparative investigation on the reinforcing effects of polymer-based fiber
into bituminous mixtures 365
E. Toraldo, E. Mariani & M. Crispino
Fibre behaviour and influence on the properties of asphalt mortar 371
M. Mohammed, T. Parry & J.R.A. Grenfell
Effect of waste metallic particles on asphalt induction heating 379
B. Gómez-Meijide, H. Ajam, A. García & H.S. Al Mahely
Improving the raveling resistance of porous asphalt with kraft lignin modified bitumen 387
N.R.Z. Poeran, B.W. Sluer, M.F.C. van de Ven & W.F. Gard
Usage of ion-irradiated recycled polypropylene as modifier in bituminous binder 397
P. Ahmedzade, T. Günay, S. Altun, B. Kultayev, A. Fainleib, O. Grigoryeva & O. Starostenko
Chemical and physical properties of an asphalt binder modified by the sap of Euphorbia
Tirucalli plant: Application in bituminous prime coat 403
L.M. Gondim, S.A. Soares, S.H.A. Barroso & C.M.C. Alecrin

vii
Physical properties of sepiolite clay nanofibers modified bitumen 411
T. Günay, P. Ahmedzade & S. Altun
High temperatures performance investigation of geopolymer modified bitumen binders 417
S.I.A. Ali, H.A.M. Yahia, A.N.H. Ibrahim & R.A. Al Mansob

Effect of traffic loading

Effects of super heavy trucks on the condition of road infrastructure 425
P. Kolisoja, T. Saarenketo & P. Varin
Dynamic pavement response coefficient to estimate the impact of variation in dynamic
vehicle load 433
S.-F. Kazemi, P.E. Sebaaly, R.V. Siddharthan, E.Y. Hajj,
A.J.T. Hand & Md. Ahsanuzzaman
Dynamic behavior of flexible inelastic pavements under moving loads 443
N.D. Beskou, A.P. Chassiakos & D.D. Theodorakopoulos
Responses and performance of flexible pavements in cold climate due to heavy
vehicle loading 451
S. Erlingsson & J. Ullberg
Impact of overloaded vehicles on load equivalency factors and service period
of flexible pavements 459
D. Rys, J. Judycki & J. Jaskula
Implications of changing the maximum legal truck load for the pavement service life 467
C.A. Lenngren & R. Salini
Modeling of dynamic loads in pavement structural response models 475
J.A. Ramos-García, F. Sánchez-Domínguez & J.M. Sanz-García
Impact of longer and heavier vehicles on the performance of asphalt pavements:
A laboratory study 483
A.W. Ahmed, M.S. Rahman & S. Erlingsson
Investigation of influence of heavy traffic loads on asphalt pavement response by SAFEM 491
P. Liu, D. Wang & M. Oeser
The increase of pavement cost due to the traffic overloads 497
J.C. Pais & P. Pereira

Structural design methods

A comparison of mechanistic-empirical pavement design methods for Norwegian conditions 507
I.E. Trangen & H. Mork
Tool for enhancing the 1993 AASHTO pavement design method to incorporate the dynamic
modulus of asphalt mixture 515
Y.S. Hamdar & G.R. Chehab
Determination of AASHTO 1993 layer coefficients considering time- and
temperature-dependency of the asphalt mixture 523
M. Lanotte & M.E. Kutay
A mechanistic empirical design concept for low volume flexible pavement using unbound
granular materials with application of concentration factor in a layered system 531
P.P. Biswas, M.K. Sahis, G.C. Mandal & D. Majumder
Perpetual flexible pavement design life 537
G. Kollaros, A. Athanasopoulou & A. Kokkalis
Mechanistic-empirical pavement design for unpaved roads: Development of a damage function
for structural rutting 543
J.-P. Bilodeau, G. Doré, M. Le Vern & P.M. Thiam

viii
Reliability demystified, at last 549
A.M. Ioannides & D.D. Rodriguez
Effect of treated subgrade layer on mechanistic-empirical pavement design 557
S. Islam, A. Sufian & M. Hossain
Comparison of practice for aggregate use in road construction—results from
an international survey 563
M. Fladvad, J. Aurstad & B.J. Wigum

In-situ measurements techniques and monitoring

Influences of measurement conditions on structural indicators obtained
from FWD data 573
C. Van Geem
A comparison of TSD and FWD deflections at Norwegian roads with an interpretation
of bearing capacity from TSD measurements 579
V. Antonsen & H. Mork
Assessment of pavement structures at traffic speed 585
A. Zofka, J. Sudyka & D. Sybilski
Integration of traffic speed deflectometer and ground penetrating radar for network-level
roadway structure evaluation 589
K.R. Maser, A. Carmichael, P. Schmalzer & B. Shaw
Methods for calculating deflections from traffic speed deflectometer data and impacts
on pavement design 595
P. Schmalzer & N. Weitzel
Quality control tool for asphalt emulsion-based chip seal curing times 603
M.A. Montoya, J.E. Haddock & W.J. Weiss
F/HWD international round robin tests on the STAC’s test facility 609
M. Broutin & S. Belon
Reproducibility of decisions for rehabilitation of existing roads based on deflection
measurements with curviameter or FWD 615
C. Van Geem
Investigation of sampling strategy on estimating strains in flexible pavements through
Response Surface Method (RSM) 623
R. Aswathy & A.K. Swamy
Effectiveness of Spectral Analysis of Surface Waves (SASW) method for
pavement evaluation 631
A. Loizos, C. Plati, B. Cliatt & K. Gkyrtis
Continuous health monitoring of asphalt concrete pavements using surface-mounted
battery-free wireless sensors 637
H. Hasni, A.H. Alavi, K. Chatti & N. Lajnef
Instrumentation to evaluate the field performance of composite overlays using accelerated
pavement testing 645
H. Rizvi, A. Ali, Y. Mehta, A. Francoise, C. Purdy & A. Nolan
Assessment of the bearing capacity of pavements using fiber optic sensors 653
V. Papavasiliou & A. Loizos
Subgrade cracking monitoring using distributed optical fiber sensing technique 661
W. Liu, B. Wang, X. Chen & L. Li
Ensuring reliable robust FWD measurements on the English road network 669
S. Brittain, A. Wright, D. Gershkoff & R. Fairclough

ix
Inferring pavement layer properties from a moving measurement platform 675
S. Andersen, E. Levenberg & M.B. Andersen
Effect of embedment scheme on damage detection performance of self-powered pavement
monitoring sensor 683
A.H. Alavi, H. Hasni, K. Chatti & N. Lajnef

Backcalculation analyses of deflection measurements

Analysing FWD data—from the deflection bowl to material parameters 691
M. Čičković
A layered pavement structural and remaining capacity model by analysis of FWD and TSD data 701
J.D. Roberts
Determination of an optimum backcalculation cross-section for flexible pavements 711
K.A. Tutu & D.H. Timm
Backcalculation of asphalt concrete moduli using field-measured strain 719
M.M. Robbins, D.H. Timm & K.A. Tutu
Optimization of deflection bowl measurements 727
V. Le Boursicaud, J.-M. Simonin & P. Hornych
Evolution of asphalt modulus from falling weight deflectometer tests and challenges
associated with its interpretation and applications: A case study using LTPP data 735
M. Oshone, M. Elshaer, E. Dave & J.S. Daniel
Application of falling weight deflectometer for the estimation of in-situ shear strength
parameters of subgrade layer 743
H. Nabizadeh, E.Y. Hajj, R.V. Siddharthan, S. Elfass & M. Nimeri
A model to adjust the falling weight deflections due to temperature variations 751
J.C. Pais & P. Pereira
Development of a frequency temperature correction model for FWD back-calculated moduli
based on frequency-temperature superposition principle 757
J.M. Flores, P. Le Van, C.K. Park, W. Kim & H.J. Lee

Structural evaluation
Evaluation of soil bearing capacity by plate load test 767
Q. Al-Obaidi, A. Al-Shamoosi & A. Ahmed
Variability of Light Weight Deflectometer during laboratory and field applications for subgrade
soils under variable moisture conditions 773
M. Mazari, G. Garcia, I. Abdallah, J. Garibay & S. Nazarian
Comparing CBR values obtained from field with laboratory test results 779
S. Karahancer, E. Eriskin, O. Sarioglu, M. Saltan & S. Terzi
Contribution to lateritic soils calibration using the dynamic penetrometer PANDA 783
Y. Gansonré, C. Bacconnet, P. Breul, M. Benz, P. Moustan, R. Gourvès & S. De-Maistre
Using QC LWD data for improving pavement design and management in Chile and Peru 791
J.R. Marcobal, F. Cerrolaza, J. Arias & M.A. Moreno
E* prediction algorithm for pavement quality control assessment 799
K. Georgouli & A. Loizos
Implementation of modulus-based quality control testing for pavement construction
in Thailand 807
A. Sawangsuriya, P. Jitareekul, Y. Taesiri, S. Sirisak & W. Lawanwisut

x
A practical non-destructive testing based approach to improve the quality of the asphalt
compaction process 813
P. Georgiou & A. Loizos
Predicting roller-compacted concrete properties from mixture proportions 819
J. LaHucik & J. Roesler
Bearing capacity assessment of a flexible pavement subjected to seasonal effects 825
G. Bazi, S. Saboundjian, R. Briggs & P. Ullidtz
Evaluation of pavement load bearing capacity comprised of insulation layers during
thaw season 833
L. Hashemian, N. Tavafzadeh & A. Bayat
Evaluation of cement-bitumen treated material bearing capacity by in-situ tests 839
M. Meocci & F. La Torre
Quality assurance of traffic-speed structural condition surveys 847
A. Wright, S. Brittain, D. Gershkoff, P. Werro & R. Fairclough

Pavement serviceability condition

Non-destructive crack identification for concrete pavements: A case study 855
L.S. Salles, L. Khazanovich, J.T. Balbo & A. Cargnin
Review of pavement cracking data collection practices 861
B. Yang, H. Ceylan, O. Smadi, K. Gopalakrishnan, S. Kim, Y. Turkan,
A.A. Alhasan & O. Adarkwa
Structural Health Monitoring of pavement assets through acoustic signature 869
R. Fedele, F.G. Praticò, R. Carotenuto & F.G. Della Corte
Automated patch detection and quantification for pavement evaluation 877
G.M. Hadjidemetriou & S.E. Christodoulou
Challenges when modelling ravelling in porous friction courses 885
L. Manrique-Sanchez, S. Caro & S. Torres
Data driven frameworks for classifying pavement surface anomalies based on pattern
recognition techniques and smartphone technology 893
C. Kyriakou, S.E. Christodoulou & L. Dimitriou
Advanced analysis of pavement longitudinal profiles for rehabilitation diagnostic 901
J.-P. Bilodeau, G. Doré & L.-A. Grégoire
In-lab versus on-site measurement of surface performance of flexible pavements 907
F.G. Praticò, S. Noto & A. Astolfi
Comparative laboratory evaluation of macro texture depth of chip seal samples using
sand patch and outflow meter test methods 915
I. Gökalp, V.E. Uz & M. Saltan
Evaluation of highway pavement skid resistance performance made of Engineered
Cementitious Composite (ECC) 921
A.A. Bawono, N. Nguyen Dinh, E.H. Yang & B. Lechner
A laboratory evaluation on skid resistance performance of surface coatings manufactured
by both natural aggregates and by-products 929
I. Gökalp, V.E. Uz & M. Saltan
Estimating new road rolling resistance using neural networks 935
C.A. Lenngren & R. Salini
Pavement performance and durability through rational design 941
A. Mouratidis & G.P. Papageorgiou

xi
Sampling functional condition indices at traffic-speed 947
K.P. Drenth, F.H. Ju & J.Y. Tan
Assessment of tire-pavement noise by using On-Board Sound Intensity (OBSI) method
in the State of Qatar 953
M. Ohiduzzaman, O. Sirin & E. Kassem
Investigation of tire-pavement noise level using On-Board Sound Intensity (OBSI) system 961
A. Zofka, E. Zofka, T. Mechowski & A. Urbanik
Overview of California studies on noise reduction for asphalt and concrete surfaces 967
J. Harvey, R. Wu, I. Guada, Q. Lu, A. Ongel, A. Rezaei, E. Kohler & C. Reyes

Rehabilitation and maintenance issues

Assessment of the efficiency of pavement surface rehabilitation techniques 977
J. Neves & H. Simas
Assessment of shotblasting moving speed for pavement preservation—a pilot-case study 983
C. Plati, A. Loizos, T. Stergiou, M. Pomoni, F. Papadimitriou & K. Drimeris
Precast concrete pavements for rapid rehabilitation of high traffic volume highways—US state
of practice 989
S. Tayabji & S. Tyson
Road bridge decks sealing, joints and pavement practice, the Greek motorway experience 997
A. Kokkalis, A. Athanasopoulou, G. Kollaros & P. Panetsos
An efficiency based approach to multi-year network-level maintenance programming 1005
T. Lorino & P. Hankach
Road asset valuation system using long term pavement data analysis 1013
A. Heitor, J. Davis, P. Tobin & K. Bogie
Why PWL is a better quality measure for developing PRS 1021
S.W. Haider, G. Musunuru & K. Chatti
Effect of neglected maintenance for the Austrian State road network 1029
J. Litzka & A. Weninger-Vycudil
A low cost method of real time pavement condition data sharing to expedite maintenance
intervention 1035
W. Uddin, O.W. Uddin & J.V. Merighi
Macroeconomic and external costs caused by inefficient pavement management: Empirical
evidence from the A9 motorway in Germany 1045
W.H. Schulz, L. Heinrich & S. Scheler

Stabilization and reinforcement

The stabilization of a soft soil subgrade layer using a new sustainable binder produced from
free-cement blending of waste materials fly ashes 1053
H.M. Jafer, W. Atherton, F. Ruddock & E. Loffill
Geosynthetic impact on subgrade bearing capacity and their possible improvement 1061
K. Pospisil & P. Zednik
Swelling pressure and consolidation of soft clay stabilized with bagasse ash and lime 1069
H. Hasan, L. Dang, H. Khabbaz & B. Fatahi
US highway 65 emergency pavement subgrade improvement 1077
R.L. Boudreau, J.P. Donahue & R.W. Brown
Influences of curing conditions on strength and microstructure of lime-amended fly ash 1083
A. Pani & S.P. Singh

xii
Compressive and tensile behaviour of polymer-treated pavement foundation materials 1091
R.N. Georgees, R.A. Hassan, R.P. Evans & P. Jegatheesan
Pavement subgrade soil bearing capacity as influenced by stabilization process 1097
A. Athanasopoulou & G. Kollaros
Pavement design considerations for subgrades stabilized with recycled materials 1105
N. Bandara, E. Jensen & T. Binoy
Design of mortars for controlled modulus columns: From laboratory to field experiments 1115
G. Blanc, G. Escadeillas, A. Turatsinze, A.-C. Ariane & B. Quandalle
Implication of the mechanically stabilised granular layer for access road over saline soils
at uranium in situ leaching mine in South Kazakhstan 1121
A. Kuznetsova, C. Doulala-Rigby, G. Solovyev & E. Orlov
Sustainable field applications of quarry byproducts mixed with large size
unconventional aggregates 1127
I. Qamhia, E. Tutumluer, H. Ozer & H. Kazmee
Bearing capacity of a stabilised granular layer on clay subgrade 1135
A.S. Lees
Mechanical characteristics and insolubilization effect of incineration bottom ash from
MSW using aging method 1143
K. Sato & T. Fujikawa
Analytical and numerical design approaches for stabilized road pavement base layers 1147
P. Skels, A. Zarins, K. Bondars & V. Haritonovs
Geosynthetic sub-base stabilization of roads and artificial turf installations 1153
C. van Gurp, T. Messinella, M. Hazenkamp & F. Harmeling
Geogrid in paved and unpaved road systems: A review of mechanisms
and design methods 1161
M. Sharbaf, N. Ghafoori & N. Dumitru
Laboratory evaluation of triangular aperture geogrid reinforced flexible pavements 1169
M. Sharbaf & N. Ghafoori
In situ damage evaluation of geogrid used in asphalt concrete pavement 1177
C. Chazallon, T.C. Nguyen, M.L. Nguyen, P. Hornych, D. Doligez,
L. Brissaud & E. Godard
High-performance synthetic microfibers for the structural reinforcement of hot mix asphalts 1183
M. Pasetto, E. Pasquini, G. Giacomello, A. Baliello & N. Baldo
Bituminous pavements reinforcement with interlayer systems: Proposing a routine laboratory
cyclic flexural testing procedure 1191
P. Pezzano, C. Sangiorgi, P. Tataranni, N. Bonucchi & C. Lantieri
Assessment of fiber reinforced HMA and WMA mixes using viscoelastic continuum
damage model 1197
H.A. Kassem, G.R. Chehab, N.F. Saleh & A. Zalghout
Lifetime prediction of asphalt interlayer systems, and the correlation between a simplified
design method and real cases 1205
P. Pezzano, A. Simone, F. Mazzotta, C. Sangiorgi, V. Vignali & G. Dondi

Recycled materials and techniques

Performance evaluation of a 100% recycled asphalt pavement mixture using a polymer
binder: A pilot study 1217
E.Y. Hajj, M. Piratheepan & P.E. Sebaaly

xiii
Maximising asphalt recycling in road surface courses: The importance of a preliminary
binder design 1225
G.M. Pires, A. Jiménez del Barco Carrión, G.D. Airey & D. Lo Presti
Structural evaluation of cold recycling mixture with foamed asphalt 1235
F.V. Guatimosim, K.L. Vasconcelos & L.L.B. Bernucci
Effects of vibratory and static compaction in laboratory to strength properties of cold
recycled materials 1243
K. Mollenhauer
Towards high quality re-use of porous asphalt 1253
J. Qiu, M. Huurman, M.H.T. Frunt & R. Gelpke
Mechanical performance of recycled aggregates in cyclic triaxial testing 1261
C. Grégoire, A. Van der Wielen & B. Janssens
Performance of warm recycled mixtures in field trial sections 1267
A. Stimilli, F. Frigio, F. Cardone & F. Canestrari
Instrumented test section for analyzing the curing process of cold-recycled mixtures 1275
C. Godenzoni, A. Graziani, M. Bocci, A. Grilli & E. Bocci
Sustainability evaluation of pavements using recycled materials 1283
Y. Zhang, D. Goulias & A. Aydilek
Moisture effects on the properties of RAP-foamed bitumen mixtures 1293
D.B. Sánchez, G.D. Airey, J.R.A. Grenfell & S. Caro
Influence of bio-based additives on RAP clustering and asphalt binder rheology 1301
M. Orešković, S. Bressi, G. Di Mino & D. Lo Presti
Evaluation of bituminous binders miscibility for warm-mix recycling techniques 1307
S. Vassaux, V. Gaudefroy, L. Jean Soro, A. Pévère, V. Mouillet, L. Boulangé & V. Barragan-Montero

Performance modeling
3D-FE Modeling and dynamic response analysis of asphalt pavements subjected
to FWD impact loads 1317
W. Uddin, Y. Nanagiri & S. Garza
Road pavement responses estimated through finite element modeling analysis 1327
B. Cliatt, A. Loizos & K. Gkyrtis
Infinite element technology of adaptive FEMOL applied in mechanical analysis of layered
elastic systems for asphalt pavement 1335
Y. Dong, Q. Xing, S. Yuan & N. Fang
Structural pavement responses using nonlinear finite element analysis of unbound materials 1343
A. Loizos, K.V. Spiliopoulos, B. Cliatt & K. Gkyrtis
3D modeling and measuring of tire-pavement contact pressure 1351
E.Y. Manyo, I. Leandry, B. Picoux, P. Reynaud, F. Allou & C. Petit
Prediction of heavy vehicle impact on rut development using PEDRO model 1359
S.F. Said & A.W. Ahmed
Evaluation of rutting potential in cold bituminous emulsion mixture using finite element analysis 1367
H.K. Shanbara, F. Ruddock, W. Atherton & G. Rothwell
Pavement rutting prediction models for the coastal roads of southern Norway 1375
E. Taddesse
Analytical modeling of rutting for asphalt concrete pavement 1383
B.A. Feyissa
Permanent deformation measurements under flexible pavements using image correlation 1391
A. Bowman & S. Haigh

xiv
Prediction of HMA fatigue performances using different Visco-Elastic Continuum
Damage (VECD) models 1399
M. Lanotte & M.E. Kutay
Modelling of reflective cracking in composite pavements 1407
E. Manola, A.C. Collop & N. Thom
Research on cracking of asphalt pavement based on cohesive zone model and extended
finite element method 1415
Y. Sun, K. Li, X. Yuan, Z. Ma & J. Wang
Fatigue performance evaluation of high modulus asphalt binder using the LAS test
and S-VECD model 1423
J. Zhang, H. Zhang & C. Wang
Analysis of expansive reactions in cement concrete pavement structures
by microscopical methods 1429
J. Stryk, A. Frybort, J. Stulirova, K. Pospisil & M. Gregerova
Influence of layer thickness on the flow of asphalt under simulated compaction 1435
E. Ghafoori Roozbahany, M.N. Partl & A. Guarin
Thin overlay analysis using finite element method 1443
H. Dhasmana, H. Ozer & I.L. Al-Qadi
A three-dimensional finite element analysis of temperature distribution in hot mix asphalt
pothole repair 1449
J. Byzyka, M. Rahman & D.A. Chamberlain
Determination of asphalt layer thickness above which axle load-induced strains initiate
top-down Cracking 1457
A. Nikolaides & E. Manthos
Cold and hot asphalt pavements modelling 1465
H.K. Shanbara, A. Dulaimi, F. Ruddock, W. Atherton & G. Rothwell
Assessment of moisture sensitivity of warm mix asphalt using advanced 3-D imaging technique 1473
M.O. Hamzah, S.Y. Teh, B. Golchin & J. Voskuilen
Analysis of the relationship between asphalt mix design and functional performance 1481
J. Qiu, M.M.J. Jacobs, M.H.T. Frunt & A. Rering
Estimation of optimal pavement performance models for highways 1489
M. Khadka & A. Paz
Performance evaluation of perpetual pavement after ten years 1495
S. Islam, A. Sufian, M. Hossain & D. Gedafa

Mechanics of layers interfaces

Analysis of the behavior of pavement layers interfaces from in situ measurements 1503
N.S. Duong, J. Blanc & P. Hornych
Static and cyclic evaluation of interlayer bonding 1511
C. Raab, M.N. Partl, E. Fourquet & A.O. Abd El Halim
Characterization of the bond between asphalt layers and glass grid layer with help
of a wedge splitting test 1517
M. Gharbi, M.L. Nguyen, S. Trichet & A. Chabot
Interface shear fatigue performance of asphalt pavement structures in function of normal
stress and temperature 1525
I. Isailović, M.P. Wistuba, S. Büchler & A. Cannone Falchetto
Research on the crack resistance of HVASSAL based on the nonlinear contact
between layers 1531
Y. Sun, Z. Gu, J. Pan & K. Li

xv
Effect of glass fibre grids on the bonding strength between two asphalt layers 1537
L. Sagnol, C. Chazallon & M. Stöckner

Geophysical assessment
A comparative investigation of the pavement layer dielectrics by FDTD modelling
and reflection amplitude GPR data 1545
F. Tosti, A.M. Alani, A. Benedetto, L. Bianchini Ciampoli,
M.G. Brancadoro & L. Pajewski
Signal processing for optimisation of low-powered GPR data with application
in transportation engineering (roads and railways) 1553
L. Bianchini Ciampoli, F. D’Amico, A. Calvi, F. Benedetto & F. Tosti
Pavement thickness evaluation with air-coupled GPR systems 1559
V. Marecos, M. Solla, S. Fontul & M.L. Antunes
Imaging boulders using the Ground Penetrating Radar (GPR) method 1567
A. Vafidis, N. Economou, G. Kritikakis, M. Galetakis, A. Vasiliou,
G. Apostolopoulos & T. Michalakopoulos
Second-generation analysis approach for condition assessment of transportation
infrastructure using Step-Frequency (SF) Ground-Penetrating-Radar (GPR) array system 1573
N. Gagarin, J.R. Mekemson & D. Goulias
Soft soils boundaries detection by GPR (the Case of Port Antioquia Survey, Colombia) 1583
Y.A. Sukhobok, V.V. Pupatenko, G.M. Stoyanovich & E.V. Fedorenko
Study on the use of low field nuclear magnetic resonance for detecting asphalt aging 1589
I. Menapace, E. Masad, M. Nogueira d’Eurydice, P. Galvosas, M.W. Hunter & O. Sirin
Assessment of pavement structures using non-invasive imaging and geotechnical testing 1597
N. Khoury, C. Saad, Y. Maalouf & M. Nasr
How to create a full-wave GPR model of a 3D domain of railway track bed? 1603
M.G. Brancadoro, F. Tosti, L. Bianchini Ciampoli, L. Pajewski, D. Pirrone,
A. Benedetto & A.M. Alani

Airfield pavements structures

Development of new FAA design procedures for extended airport pavement life 1611
D.R. Brill & T.A. Parsons
Development of rational ACN/PCN system 1619
C. Fabre & G. Vaurs
Why can’t the PCN index be uniquely defined? 1629
A. Loizos, A. Armeni & B. Cliatt
Harmonizing PCN reporting on extended airport maneuver areas 1637
C. Busch & A. Rolim
Towards airfield pavement design using cold recycled bound materials 1647
H.I. Lacalle Jimenez, N. Thom & B. Hakim
Design of rigid and flexible airfield pavements on cement-treated base 1653
C. Rabaiotti, D. Tsirantonaki & M. Schnyder
Innovative construction techniques and functional verification on airfield
pavements—a Dutch case study 1663
F.R. Bijleveld, A.H. de Bondt, R. Khedoe & M. Stet
Effect of winter contaminants removal techniques on airfield pavement surfaces 1671
S. Anastasio, I. Hoff, K.R. Holm, L. Fensholt & L. Næsbø

xvi
Comparing near surface aircraft pavement responses calculated by finite element
and layered elastic methods 1677
G. White & W. Smith
Assessment of critical loading conditions on rigid airfield pavements by means
of numerical simulation 1687
D. Mounier, J.-M. Piau & O. Chupin
Domain analysis for airfield pavement: Moving forward from point responses 1693
J.A. Hernandez, A. Gamez & I.L. Al-Qadi
Airfield pavement management program of Hellenic air force 1701
E. Adamidou & P. Gomatou
Nondestructive detection of voids under airfield pavement 1707
M. Abbasghorbani, A. Bamdad & N. Tabatabaee
Effect of subbase quality on performance of rapid-setting replacement slabs for airfield pavements 1713
J.S. Tingle, H.P. Bell, L.P. Priddy & L. Edwards
Implementing the traffic speed deflectometer for airfield runway assessment 1721
D.D. Rodriguez & L. Edwards
Managing the impact of more demanding aircraft 1727
G. White
Expedient runway upgrade technologies 1735
G. White
Preliminary analysis of pavement life indicators of large- and medium-hub airport runways 1743
T.A. Parsons & D.R. Brill
Impact of joint filling materials on cracking in concrete gutters at large airfields 1751
R. Breitenbuecher
Asphalt Pavement Analyzer—laboratory tool for characterization of rutting performance
of airport pavement Hot Mix Asphalt (HMA) 1757
N. Garg, Q. Li & J. Stein
Neural networks prediction of critical responses related to top-down and bottom-up cracking
in airfield concrete pavement 1765
A. Rezaei-Tarahomi, O. Kaya, H. Ceylan, K. Gopalakrishnan, S. Kim & D.R. Brill
Influence of airfield pavement roughness and ponding on gear life and safety
of commercial aircraft 1773
J.V. Merighi & W. Uddin
Life cycle cost analysis of runway reconstruction alternatives for commercial service airports 1783
M.T. McNerney & G.P. Vittas
Perpetual pavement responses under single and dual wheel aircraft gear loading 1791
H. Wang, M. Li & N. Garg

Railway track structures

Experimental study on cyclic deformation and particle breakage of railway ballast 1801
X. Bian, D. Sun & W. Li
Ballast degradation analysis by Los Angeles Abrasion test and image analysis method 1811
Y.L. Guo & G.Q. Jing
Influence of parallel grading on hydro-mechanical characteristics of unsaturated fouled ballast 1817
T. Ishikawa, S. Matsutani, T. Tokoro, T. Nakamura & Y. Momoya
Analysis and improvement of the dynamic track behaviour in transition zone 1827
H. Wang & V.L. Markine

xvii
Influence of water content and triaxial size effects on determination of bearing capacity
of fouled ballast 1837
A.K. Rohrman, Z. Yang, H.F. Kashani & C.L. Ho
Roadbed improvement of an existing railway line located in cold region by reusing
crushed deteriorated ballast 1845
L.L. Fu, J.H. Xiao, S.H. Zhou, D. Zhang, Y.H. Wang, W.J. Liu & L.H. Jiang
Laboratory characterisation of rigid polyurethane foam reinforced ballast 1851
R.F. du Plooy, P.J. Gräbe & C. Turner
Engineering performance of steel slag as railway ballast 1859
C.-M. Kuo & C.-C. Lin
Effect of temperature and traffic on mix-design of bituminous asphalt for railway
sub-ballast layer 1865
F. Martinez Soto, G. Di Mino & F. Acuto
Some results on the properties and behavior of railway ballast 1877
E. Fortunato, A. Paixão, S. Fontul & J. Pires
Theoretical assessment of railway ballast degradation under cyclic loading 1885
P.K. Bajpai & A. Das
Performance testing of geogrid-stabilised ballast and sub-ballast subjected to
high speed rail forces 1891
Z. Yu, D.P. Connolly, P.K. Woodward, M. Horton & T.L.H. Oliver
Cumulative deformation characteristic and shakedown limit of railway ballast under
cyclic loading 1899
J.H. Xiao, D. Zhang, Y.H. Wang & Z. Luo
The impact of heavy freight train on the roadbed 1905
A.V. Petriaev
Track defects and the dynamic loads due to suspended (sprung) masses of railway vehicles 1911
K. Giannakos
Ensuring of railway ballast and subballast bearing capacity in terms of heavy axle load
train operation 1921
A.A. Konon
Performance of track transition remedies implemented in high tonnage heavy axle
load environment 1925
D. Li, J. Baillargeon & L. Maal
Design optimization of the modern tram track infrastructure in soft soil area 1931
Y. Shan, J. Zhang, S.H. Zhou, Q. Gong & B. Wang
Towards analytical asphalt trackbed design 1937
B. Hakim, C. Bonney, M. Brough & C. Govan
Study on the evaluation of track support stiffness of the ballasted track using the FWD 1945
H. Tanigawa, T. Nakamura & Y. Momoya
Use of ballast support condition back-calculator for quantification of ballast pressure
distribution under concrete sleepers 1955
Z. Gao, M.S. Dersch, Y. Qian, M.V. Csenge & J.R. Edwards
Effect of stiffness and bearing capacity of railway components on track behaviour:
A laboratory approach for design optimization of railway tracks 1961
M. Sol-Sánchez, L. Pirozzolo, F. Moreno-Navarro, G. Martínez-Montes & M.C. Rubio-Gámez
Laboratory evaluation of under-ballast mat effectiveness to mitigate differential movement
problem in railway transition zones 1969
A. de O. Lima, M.S. Dersch, Y. Qian, E. Tutumluer & J.R. Edwards

xviii
Bearing capacity evaluation of a subgrade in a heavy haul railway in Brazil 1977
R. Costa, R. Motta, L.L.B. Bernucci, E. Moura, J. Pires & L. Oliveira
Laser triangulation measurement system: Application on railway track 1985
S. Cafiso, C. D’Agostino, E. Delfino, A. Di Graziano, R. Fox-Ivey & J. Laurent
Introduction of new systems for evaluation of ballast bearing capacity 1993
S. Cafiso, B. Capace, C. D’Agostino, E. Delfino & A. Di Graziano
Evaluation of the ballast aggregates shape properties using digital image
processing techniques 2003
D.F. Diógenes, R.S. Maia & V.T.F. Castelo Branco
Finite element model on stress distribution of ballasted railway subgrade and its validation 2009
Y. Feng, C. Zhao & X. Zhang
Modeling the viscoplastic creep behavior of asphalt concrete for use in ballastless railway
tracks in tunnels 2015
O. Lopez Polanco, N. Calon, T. Gabet & P. Hornych
Effect of creep and destructuration under Sebou high speed railway line embankment 2023
A.H. Mridakh, F. Lahlou, A. Mridekh, H. Ejjaaouani & H. Labied
Effect of uneven subgrade settlement in high-speed railway on double-block ballastless
track regularity 2031
Y. Guo & W.M. Zhai
Railway track vibration and degradation under train operation 2037
M.J.M.M. Steenbergen, A. Zoeteman, H.J. de Graaf & R. van Dommelen
Impact hammer testing of a railway track 2041
E. Balic, C. McNally & D. Hester
Effects of model scale on lateral resistance characteristic of sleepers in railway
ballasted tracks 2047
H. Tomita, K. Hayano & P.T. Anh
Prediction of railway ballast service life 2055
R. Nålsund
Study on the mechanism and inspection method of railway pumping 2063
C.P. Kuo, C.H. Hsu, C.W. Wu, P.L. Liu & D.W. Chen
Next problems with the railway prestressed sleepers cracking in concrete as an effect
of delayed ettringite formation 2069
J. Jasiczak & Z. Owsiak

Environmental challenges
Evaluation of Norwegian gradation based regulation for frost susceptibility of crushed rock
aggregates in roads and railways 2077
B. Loranger, E. Kuznetsova, I. Hoff, J. Aksnes & K.A. Skoglund
Seasonal temperature distribution in rigid pavements 2087
K. Bayraktarova, L. Eberhardsteiner & R. Blab
Longitude profiling as a tool for evaluation of frost actions active pavement section 2095
T. Edeskär, V. Perez, J. Ullberg & P. Ekdal
A method to assess climate change induced damage on flexible pavements with
machine learning 2103
Y. Qiao, Y. Zhang, M. Elshaer & J.S. Daniel
Testing and modeling the behavior of water saturated asphalt concrete under freezing
conditions in relation to pavement damaging 2111
V.-T. Vu, O. Chupin, J.-M. Piau, F. Hammoum & P. Lecontee

xix
Predicting subgrade soil strength using FWD and meteorological time series data 2117
L. Gáspár, M. Karoliny & C. Tóth
Bearing capacity analysis of pavement structures for short term flooding events 2127
M. Elshaer, M. Ghayoomi & J.S. Daniel
The importance of maintaining pavement roughness to reduce carbon footprint 2135
A. Drainakis, M. Pomoni & C. Plati
Thermal energy harvesting from asphalt pavement roadways 2141
U. Datta, S. Dessouky & A.T. Papagiannakis
Water-heat response of embankments on permafrost to increasing precipitation 2149
Z. Wen, Z.H. Bao, P.G. Liao, M.L. Zhang, Z.Z. Sun & S.J. Zhang
A water preservation base for pervious pavement of traffic road 2155
S.-Y. Wu, H.-J. Liao & Y.-T. Huang
Applications of an innovative load bearing permeable concrete pavement 2159
Y.H. Lee, H.W. Ker, N.S. Chou & J.W. Chen
Accumulation managing pavement—design, construction and performance for confined areas 2169
A. Dodhiya, M. Dave, P. Modi & J. Shah
The effect of highway geometry on fuel consumption of heavy-duty vehicles operating
in eco-driving mode 2179
G.K. Booto, R.A. Bohne, H. Vignisdottir, K. Pitera, G. Marinelli, H. Brattebø,
H. Wallbaum & B. Ebrahimi
Detoxication of the heavy metal ions in water resources by means of mineral geoantidotes 2187
A.S. Sakharova, L.B. Svatovskaya, M.M. Baidarashvili & A.V. Petriaev

Life cycle assessment and sustainability

Sustainable pavement rehabilitation strategy using consequential life cycle assessment:
An example of interstate 95 2193
C. DeCarlo, W. Mo, E. Dave & J. Locore
Sustainable implementation of future smart road solutions: A case study on the electrified road 2201
F. Chen, R. Balieu & N. Kringos
Life cycle assessment of an optimized network arch highway bridge utilizing timber 2209
R. O’Born, K. Vertes, G. Pytten, L.O. Hortemo & A. Brændhagen
Economic benefit calculation of quality improvements in road construction 2215
M. Granden, J.M. Johansen & L. Bakløkk
Pay adjustment of asphalt pavement based on the Life-Cycle Cost Analysis (LCCA)
and a brief case study in China 2221
X. Zhao, A. Sha & Z. Liu

Full-scale testing
Deflections and dynamic responses of asphalt pavement with graded-broken-stone base:
An investigation based on Heavy Vehicle Simulator (HVS) 2229
A. Sha, Z. Liu, W. Zhao, L. Hu, J. Wang & Z. Han
Fast Falling Weight Deflectometer (FastFWD) for Accelerated Pavement Testing (APT) 2235
M. Manosalvas-Paredes, A. Navarro Comes, M. Francesconi, S. Khosravifar & P. Ullidtz
Pavement strengthening with respect to frost penetration: A laboratory study using
a full-scale heavy vehicle simulator 2243
J.-P. Bilodeau, J.-P. Cloutier, G. Doré & P.M. Thiam

xx
Use of accelerated pavement testing for degree of risk due to road cavity 2251
H.M. Park, Y.T. Kim, J.Y. Choi & B. Kim
Full-scale accelerated pavement testing of geogrid stabilized roads 2255
J.S. Tingle, G.J. Norwood, W.J. Robinson, M.H. Wayne & J. Kwon

Case histories
Case study of a performance based contract for asphalt pavement overlay in Korea 2265
J. Lee, B. Kim, J. Lim & Y. Kim
Design of Ultra-Thin Continuously Reinforced Concrete (UTCRC) and Enrobé à Module
Élevé (EME) implemented as alternative remedial actions 2271
S.J. Bredenhann, J.L. van Heerden, P.J. Strauss & P.J. Joubert
Resurfacing of the Queen Elizabeth II bridge: Binder selection and evaluation 2281
I. Artamendi, B. Allen, C. Allpress, P. Phillips & C. Wingrove
Geotechnical assessment strategy for bridge maintenance—case study 2289
O. Hamza

Author index 2299

xxi
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Preface

The series of international conferences on the Bearing Capacity of Roads, Railways and Airfields
(BCRRA) provides a unique opportunity for professionals and practitioners to interact, exchange
concepts, discuss best practices, innovative solutions and relevant experience for pavement materials,
laboratory and field testing, design, construction, maintenance and rehabilitation aspects related to trans-
portation infrastructure issues.
Traditionally, the selection and testing of materials intended for the design, construction and reha-
bilitation of transportation infrastructure has been one of the most critical challenges facing industry
professionals, researchers and practitioners involved in transportation infrastructure engineering. This
continues to be the case and enhanced bearing capacity design and analysis still remain challenging issues
as they aim at often interrelated goals towards constructing and maintaining sustainable and resilient
infrastructure systems that are resistant to even more demanding traffic and environmental loads.
The BCRRA 2017 is the 10th conference in the series initiated in Trondheim (Norway), back in 1982,
and over the years it has been held four more times in Trondheim as well as in Plymouth (UK), Minne-
apolis (USA), Lisbon (Portugal) and Urbana-Champaign (USA). The conference was initially organized
under the title “Bearing Capacity of Roads and Airfields—BCRA”. During the organization of the sixth
BCRA conference in Lisbon, Portugal, the third component (Railways) was added to the overall scope
of the conference as it constitutes a vital and critical element for transportation infrastructure world-
wide. The acronym BCRRA was first adopted during the organization of the 8th conference in Urbana-
Champaign, Illinois, USA to emphasize infrastructure problems that all three transportation modes have
concerning bearing capacity challenges. The current 10th BCRRA conference is jointly organized by
the National Technical University of Athens (NTUA), the University of Illinois at Urbana-Champaign
(UIUC) and the Delft University of Technology (TU DELFT).
Throughout the previous conferences the term “bearing capacity” has been notably broadened and was
further applied to describe the response and performance of a broader range of pavement materials based
on research progress, technological advances and increasing environmental awareness. Also, new materials
and methods have been and are continuing to be developed that demand an improved and/or smoother
transition towards the implementation of mechanistic concepts for the design of sustainable pavements
and railroad track structures. In setting up the BCRRA 2017 conference, a decision was made to adopt
this wider definition and invite contributions spanning the whole range of pavement and railway materials
and engineering, including advanced mechanical testing and characterization and serviceability condition
assessment. The significant number of contributions from around the world with original content and
high scientific level covering the overall spectrum of roads, railways and airfields infrastructure justifies
this BCRRA 2017 initiative.
All submitted contributions were subject to a rigorous refereed peer review procedure by a minimum
of two qualified reviewers, often including members of the conference Scientific Committee. Based on the
reviewers’ recommendations, those contributions that best suited the conference goals, scientific level and
objectives were chosen. In total, more than 300 contributions were presented and discussed during the
conference sessions.
Although not included in the proceedings book, it should be noted that in the framework of the organi-
zation of the BCRRA 2017, multiple pre-conference workshops have also been organized with subjects
related to life cycle assessment and sustainability, vehicle-tire-pavement interaction, airport pavement
design procedures, railway track monitoring and maintenance and infrastructure engineering in cold cli-
matic conditions.

xxiii
 It is the editors’ ambition that the proceedings book will offer valuable updated information to the
worldwide scientific community with identifiable advances in the relevant knowledge for professionals,
researchers and practitioners. The editors would like to take the time to thank the Scientific Committee
members and individual reviewers for their dedication and contributions of their time and efforts to
ensure the noted scientific and technical quality of the accepted papers. In addition, the guidance and
continued input from the International Advisory Committee members was essential for the planning of
the conference, and is deeply appreciated. Finally, we would like to gratefully acknowledge the Organizing
Committee members for their help in local logistics of the conference, suggestions and contributions to
the management of the conference affairs.

Andreas Loizos, Imad L. Al-Qadi, Tom Scarpas

Athens, Greece, June 2017

xxiv
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Organization

CHAIRMEN

Andreas Loizos, National Technical University of Athens, Greece

Imad Al-Qadi, University of Illinois at Urbana-Champaign, USA
A. (Tom) Scarpas, Delft University of Technology, The Netherlands

INTERNATIONAL ADVISORY COMMITTEE

Leif Bakløkk (Chair), Norwegian Public Roads Administration, Norway

Andreas Loizos (Co-chair), National Technical University of Athens, Greece
Christina Plati (Secretary), National Technical University of Athens, Greece

Imad L. Al-Qadi, University of Illinois at Urbana-Champaign, USA

Alessandra Bianchini, U.S. Army Engineer Research and Development Center, USA
David R. Brill, Federal Aviation Administration, USA
Antonio Gomes Correia, University of Minho/DEC, Portugal
David D. Davis, Association of American Railroads, USA
Andrew Dawson, The University of Nottingham, United Kingdom
Rolf Dollevoet, Delft University of Technology, The Netherlands
Guy Doré, Laval University, Canada
Sigurdur Erlingsson, University of Iceland and Swedish Road and Transport Research Institute,
Iceland and Sweden
Brian Ferne, Transport Research Laboratory (TRL), United Kingdom
Ralph Fischer, Deutsche Bahn AG - DB Systemtechnik, Germany
Rita Moura Fortes, Mackenzie Presbyterian University, Brasil
Gregers Hildebrand, COWI, Denmark
Inge Hoff, Norwegian University of Science and Technology, Norway
Buddhima Indraratna, University of Wollongong, Australia
Geoff Jameson, ARRB Transport Research Ltd., Australia
Elias Kassa, Norwegian University of Science and Technology, Norway
Pauli Kolisoja, Tampere University of Technology, Finland
Jens Melsom, Norwegian National Rail Administration, Norway
Helge Mork, Norwegian University of Science and Technology, Norway
Jean Michel Piau, French Institute of Science and Technology for Transport, Development and Networks, France
Tom Scarpas, Delft University of Technology, The Netherlands
Ramesh Sinhal, Highways Agency, United Kingdom
Ted Sussmann, University of Hartford/ Volpe Transportation Center, USA
Susana Svetlicic, DRI Investment Management, Slovenia
Mark Swanlund, Federal Highway Administration, USA
Dariusz Sybilski, Road and Bridge Research Institute, Poland
Gabriele Tebaldi, University of Parma, Italy
Erol Tutumluer, University of Illinois at Urbana-Champaign, USA
Xinglong Wang, Heilongjiang Institute of Highway and Transport Research, P.R. China
Tomas Winnerholt, Swedish Transport Administration, Sweden
Benjamin Worel, Minnesota Department of Transportation, USA

xxv
SCIENTIFIC COMMITTEE

Gordon Airey, The University of Nottingham, United Kingdom

Jostein Aksnes, Norwegian Public Roads Administration, Norway
Sara Anastasio, Norwegian University of Science and Technology, Norway
George Apostolopoulos, National Technical University of Athens, Greece
Joralf Aurstad, Norwegian Public Roads Administration, Norway
Jose Tadeu Balbo, University of Sao Paulo, Brazil
Hartmut Johannes Beckedahl, University of Wuppertal, Germany
Andrea Benedetto, Roma Tre University, Italy
Geir Berntsen, Norwegian Public Roads Administration, Norway
Amit Bhasin, The University of Texas at Austin, USA
Ronald Blab, Technische Universität Wien, Austria
Francesco Canestrari, Università Politecnica delle Marche, Italy
Silvia Caro, Universidad de Ios Andes, Colombia
Armelle Chabot, IFSTTAR, France
Athanasios Chassiakos, University of Patras, Greece
Karim Chatti, Michigan State University, USA
Chassan Chehab, American University of Beirut, Lebanon
Andrew Collop, De Montfort University, United Kingdom
Jo Daniel, University of New Hampshire, USA
Eshan Dave, University of New Hampshire, USA
Herve Di Benedetto, University of Lyon, ENTPE, France
Nikos Economou, Technical University of Crete, Greece
Nikolaos Eliou, University of Thessaly, Greece
Sandra Erkens, TU Delft, Netherlands
Marit Fladvad, Norwegian Public Roads Administration, Norway
Simona Fontul, National Laboratory for Civil Engineering, LNEC, Portugal
Jeffrey S. Gagnon, Airport Pavement R & D Section, USA
Navneet Garg, Federal Aviation Administration (FAA), USA
Kenneth Gavin, Delft University of Technology, The Netherlands
Konstantina Georgouli, National Technical University of Athens, Greece
Konstantinos Giannakos, Railways R.C., Greece
Dimitris Goulias, University of Maryland, USA
Elie Hajj, University of Nevada Reno, USA
Kevin D. Hall, University of Arkansas, USA
Meor Othman Hamzah, Universiti Sains Malaysia, Malaysia
John Harvey, University of California, USA
Jake Hiller, Michigan Technological University, USA
Pierre Hornych, LUNAM Université, IFSTTAR, France
Ivar Horvli, Norwegian Public Roads Administration, Norway
Wei-Hsing Huang, National Central University, Taiwan
Tatsuya Ishikawa, Hokkaido University, Japan
Kim Jenkins, Stellenbosch University, South Africa
Torbjørn Jørgensen, Norwegian Public Roads Administration, Norway
Kamil Elias Kaloush, Arizona State University, USA
Malal Kane, LUNAM Université, IFSTTAR, France
Lev Khazanovich, University of Minnesota, USA
Alexandros Kokkalis, Democretus University of Thace, Greece
Niki Kringos, KTH Royal Institute of Technology, Sweden
Murali J. Krishnan, Indian Institute of Technology Madras, India
Emin Kutay, Michigan State University, USA
Elena Kuznetsova, Norwegian University of Science and Technology, Norway
Sabine Leischner, Technische Universität Dresden, Germany
Zhen Leng, The Hong Kong Polytechnic University, China
Bjørn Ove Lerfald, Norwegian University of Science and Technology, Norway

xxvi
Eyal Levenberg, Technical University of Denmark, Denmark
Jianming Ling, Tongji University, China
Dallas Little, Texas A&M University, USA
Luis Guillermo Loria-Salazar, University of Costa Rica, Costa Rica
Massimo Losa, University of Pisa, Italy
Nikos Malakatas, CLPW, Greece
Gülay Malkoç, ASMUD, Turkey
Rajib Mallick, Worcester Polytechnic Institute, USA
Evangelos Manthos, Aristotle University of Thessaloniki, Greece
Alessandro Marradi, University of Pisa, Italy
Eyad Masad, Texas A&M University at Qatar, Qatar
Lillian Uthus Mathisen, Veidekke AS, Norway
George Mintsis, Aristotle University of Thessaloniki, Greece
Goran Mladenovic, University of Belgrade, Serbia
Anastatios Mouratidis, Aristotle University of Thessaloniki, Greece
Roar Nålsund, The Norwegian National Rail Administration, Norway
Soheil Nazarian, University of Texas at El Paso, USA
Athanassios Nikolaides, Aristotle University of Thessaloniki, Greece
Markus Oeser, RWTH Aachen University, Germany
David P. Orr, Cornell University, USA
Hasan Ozer, University of Illinois at Urbana-Champaign, USA
Jorge Carvalho Pais, University of Minho, Portugal
Tom Papagiannakis, University of Texas San Antonio, USA
Manfred Partl, EMPA, Switzerland
Marco Pasetto, University of Padova, Italy
Christophe Petit, Université de Limoges, France
Andrei Petriaev, Emperor Alexander I St. Petersburg State Transport University, Russia
Filippo Pratico, Mediterranea University of Reggio Calabria, Italy
Christiane Raab, EMPA, Switzerland
Jeffery Roesler, University of Illinois, USA
Rabbira Garba Saba, Norwegian Public Roads Administration, Norway
Mofreh Saleh, University of Canterbury, New Zealand
Ezio Santagata, Politecnico di Torino, Italy
Charles Schwartz, University of Maryland, USA
Mohamed Shahin, Curtin University, Australia
Kjell Arne Skoglund, Sintef, Norway
Brynhild Snilsberg, Norwegian Public Roads Administration, Norway
Konstantinos Spiliopoulos, National Technical University of Athens, Greece
Wynand JvdM Steyn, University of Pretoria, South Africa
Even Sund, Norwegian Public Roads Administration, Norway
Nader Tabatabaee, Sharif University of Technology, Iran
Ephrem Taddesse, University of Agder, Norway
Serdal Terzi, Suleyman Demirel University, Turkey
Carl Thodesen, Norwegian University of Science and Technology, Norway
Susan Tighe, University of Waterloo, Canada
David Timm, Auburn University, USA
Fabio Tosti, University of West London, United Kingdom
Yiannis Tsompanakis, Technical University of Crete, Greece
Waheed Uddin, University of Mississippi, USA
Nils Sigurd Uthus, Norwegian Public Roads Administration, Norway
Carl Van Geem, Belgian Road Research Centre, Belgium
Julie Marie Vandenbossche, University of Pittsburg, USA
Christ Van Gurp, KOAC-NPC, The Netherlands
Konstantinos Vogiatzis, University of Thessaly, Greece
Hainian Wang, Chang’an University, China
Kelvin Wang, Oklahoma State University, USA

xxvii
Linbing Wang, Virginia Tech, USA
Yuhong Wang, The Hong Kong Polytechnic University, China
Greg White, University of the Sunshine Coast, Australia
Michael Wistuba, TU Braunschweig, Germany
Pyeong Jun Yoo, Korea Institute of Construction Technology, S. Korea
Zhanping You, Michigan Technological University, USA
Adam Zofka, Road and Bridge Research Institute, Poland

xxviii
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Supporting organizations

xxix
 ROMA
TRE
UNIVERSITÀ DEGLI STUDI

xxx
Subgrade soils
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Evaluation of a resilient modulus model for unsaturated soil conditions

P. Thirthar Palanivelu & C.E. Zapata

Arizona State University, Tempe, Arizona, USA

ABSTRACT: The strength and the durability of pavement systems are directly related to the stiffness
of the unbound and subgrade soil materials. The 1993 AASHTO flexible pavement design guide and
more recently, the AASHTO ware software for the mechanistic-empirical design of pavements relies on
the resilient modulus as the primary input parameter to characterize the stiffness of the subgrade and
unbound base materials. Models that describe the modulus as a function of externally imposed stresses
has been adopted in design procedures, but models that consider the internal stress state of the material
at all levels of saturation have not been evaluated due to the lack of data available. This paper presents
an evaluation of the Cary and Zapata (2010) model parameters, which includes the matric suction stress
state of the material as a driver of the moisture soil retention capability. The evaluation was performed
with data collected from resilient modulus testing on six different soil types at different moisture contents.
Laboratory test performed include the soil-water characteristic curve, filter paper method and a compre-
hensive set of index properties. The resilient modulus tests were performed on soil specimens in accord-
ance with the NCHRP 1-28 A protocol at different compaction conditions.

1 INTRODUCTION approach. Incorporation of moisture content into

the design approach emphasizes the role of envi-
In mechanistic pavement design procedures, the ronmental factors into the pavement design guide
stiffness of pavement layers is characterized by procedures (Lekarp 2000). In the early to mid-2000
the elastic modulus and the Poisson’s Ratio. For period, major advances allowed for the incorpora-
unbound layer materials, this modulus is typi- tion of unsaturated soil principles into a pavement
cally referred to as the “resilient modulus”, which design methodology in United States in a more
reflects only the true elastic or recoverable portion mechanistic fashion. This was accomplished with
of the deformation phase. The variability of the the development of the American Association
resilient modulus depends on different factors such of State Highway and Transportation Officials
as moisture content, dry density, number of load (AASHTO) Mechanistic-Empirical Pavement
repetitions, the compaction energy involved, the Design Guide (M-EPDG).
compaction method used, and the state of stress Recently, several experimental studies have
in the field (Andrei 2003). Historically, pavement been performed to analyze the influence of mois-
design has been based, for the most part, on the ture content on the resilient response of the pave-
selection of material design properties that are at ment unbound material (Doucet 2004, Cary 2010,
the “worst possible environmental conditions”, Andrei 2003, Rahman 2012). Various models have
regardless of where the site is located. Thus, all been developed using these studies to predict the
unbound materials (base, subbase and subgrades) moisture content parameter and its variation due
are classically selected at fully saturated moisture to seasonal effects. Results from previous studies
conditions. However, in the past 30 years, there has shown that the values of MR in soils decrease
have been significant advancements in the field with an increasing in moisture content and increase
of unsaturated soil mechanics, that has allowed with a decrease in moisture content; which in turns
assessing pavement performance predictions using means that the seasonal variation of post compac-
real time environmental, seasonal and aging effects tion moisture content in base and subgrade soil
to predict actual changes in the unbound mate- materials has a determinant influence in the overall
rial strength and resilient behavior of the materi- structural performance of pavement systems.
als. One of the major factors playing a significant Unsaturated soil mechanics allow us to explain
role in the mechanical response of the unbound that the variations in the moisture content of
pavement materials is the moisture content of the fine-grained subgrades due to seasonal variations
soil. Hence, proper consideration should be given can result in changes in the stress state of the soil.
to this factor in any sustainable pavement design This change can be explained in terms of changes

3
in the soil matric suction, which is the difference Cary and Zapata (2010) proposed a model that
between the pore-air and the pore-water pres- incorporates the influence of seasonal environ-
sures in the soil matrix. Therefore, a more rational mental variations by including the matric suction
approach to incorporate seasonal effects in the as stress state variable into the model developed by
resilient modulus predictive models would be to Witczak and Uzan. They studied the effects of the
consider matric suction as a fundamental vari- moisture content in the soil in terms of suction and
able in the stress state of the subgrade soil, rather obtained sets of regression constants from triaxial
than considering moisture content as a driving experiments on granular base and subgrade mate-
mechanism. rials. The proposed model is shown in Eqn (1):
The overall objective of this study is to inves-
tigate the influence of moisture content on the ⎛ ⎞ 2
k k3
⎜ θ nnet − 3Δuw − sat ⎟⎟ ⎛
τ ooc ⎞

pa + 1⎟⎠
mechanical response of the unbound pavement MR k1 pa ⎜⎜ ⎜
⎜⎝
octt ⎟
⎟
materials using the model developed by Cary and ⎜⎝ pa ⎟⎠

Zapata (2010) and the evaluation of the Cary and ⎛ (ψ

k
m0 − Δψ m )
⎞ 4
Zapata parameters, which includes the suction ⎜
⎜ p + 1⎟⎟ (1)
⎜⎝ a ⎟⎠
stress of the material that represents the saturation
state.
In this study, data from six different subgrade where pa = atmospheric pressure; θ net θ − 3ua = the
soil were used in the analysis. The six soil materi- net bulk stress (θ = bulk stress = σ 1 + σ 2 + σ 3 and ua
als were tested at the Arizona State University = pore-air pressure); Δuw sat = pore-water pressure
(ASU) Geotechnical Laboratory using a custom build up under saturated condition; τoct = octahedral
built advanced triaxial cells and control units. The shear stress = 13 ( 1 − 2 )2 + ( 1 − 3 )2 + ( 2 − 3 )2 ;
system enables full control/measurement of pore- ψ m0 = initial matric soil suction; Δψm = relative
water and pore-air pressure of the cylindrical test change in soil matric suction with respect to ψ m0
specimen during the Repeated Load Triaxial (RLT) due to pore-water pressure build up under unsatu-
test. The load sequences used for this purpose were rated condition; k1 k2 0, k3 ≤ 0 and k4 ≥ 0
in accordance to the NCHRP 1-28 A (2004) pro- are regression constants.
tocol “Harmonized Test Methods for Laboratory The Cary and Zapata model was validated and
Determination of Resilient Modulus for Flex- the regression constants were determined and cali-
ible Pavement design” after some modifications to brated for all six soils.
the procedure that would allow for testing soil in
unsaturated conditions with matric suction control.
More details on the modification to the NCHRP 3 SUBGRADE SOIL PROPERTIES
1-28 A loading procedure and triaxial cell control
unit can be found in Cary and Zapata (2010). 3.1 Soil index properties
Six soils were used in this study. Data for two of
the soils used for this study was collected and ana-
2 MODELING OF SOIL RESILIENT lyzed by Cary (2010). These soils were named IOSG
MODULUS and GESG, after the location of the soils found in
Arizona, USA. The other two soils were obtained
Resilient modulus of the soil is widely used to from two different sites in Sweden named Torpsbruk
characterize the mechanical stiffness of the pave- and Luleå. Data for these two soils was obtained and
ment unbound materials and stiffness of unbound analyzed by Salour (Salour 2014). The remaining two
pavement materials. The resilient modulus of the soils (County and Dupont clays) were obtained from
soil is measured under the conditions representa- the Federal Aviation Administration (FAA) National
tive of the external stress state and the environment Airfield Pavement Test Facility located in New Jersey,
conditions experienced by the unbound materials USA. All the six soils were analyzed in ASU labora-
in the pavement system. Various researches have tory. Wet sieve analysis, specific gravity, maximum dry
been studied the influence of different parameters density, optimum moisture content and Soil-Water
and to develop mathematical model to predict the Characteristic Curve (SWCC) tests were performed
stress dependence of the resilient modulus of the on the selected materials. A summary of the soil index
soil using a total stress approach. A model devel- properties is presented in Table 1. The test results of
oped by Witczak and Uzan (1992), also known as Proctor Compaction tests are presented in Table 2.
the Universal model, clearly explain the influence
of the externally applied stress parameters on the
3.2 Soil Water Characteristic Curve (SWCC)
resilient modulus, but it does not directly account
for the influence of the environmental factors into The SWCC for the six subgrade materials were
the stress sate conditions. determined using the Fredlund SWCC device, an

4
Table 1. Summary of the subgrade soil index properties.

Subgrade USCS Passing Specific Plasticity

soil class* no.200 (%) gravity index

Luleå SM 42.2 2.68 NP

Torpsbruk SM 27.4 2.67 NP ...0
IOSG SC 37.5 2.72 7.2 ., 40
GESG SM 47.7 2.72 3.3 .,~
c
Cl
20
County ML 58.0 2.84 17
Dupont MH 80.0 2.79 29
0.001 0.1 10 1000 100000
*USCS—Unified Soil Classification System. Matric Suction [kPa}

Table 2. Proctor compaction test results. Figure 2. SWCC for County and Dupont soils.

Max. dry Opt.

Subgrade unit weight moisture 100
soil (kN/m3)* content. (%)
~ 80
Luleå 19.6 10.1
-~
Torpsbruk 20.3 7.6 60
IOSG
GESG
18.7
19.0
12.1
12.5
~ 40
"
County 16.3 21.0 1l'
0
20
Dupont 14.8 26.3

*Maximum dry unit weight is based on standard Proctor 10 100 1,000 10,000 100,000 1,000,000
compaction test (ASTM D698).
Matric Suction (kPa)

100 I I II Ill I I II
*90 Lulea (North)
'CBO
,g 70 - - Torpsbruk (South)
Figure 3. SWCC for IOSB soils (Cary 2010).

~60
iii 50 4 RESILIENT MODULUS TESTING
"'
(540 ~
~ 30
.,
!»20 ~ 4.1 Equipment and system set-up
010 !!!!!
Cl) 0 The triaxial system used to perform the resilient
0.01 0.1 10 1 00 1 000 10000 1 000001 000000 ""' modulus testing of the soil for this testing is a
Matric Suction [kPa]
custom-built system, which is capable of applying
Figure 1. SWCC for Lulea and Torpsbruk soils (Salour, repeated cyclic load of a haversine-shaped load
2014). pulse. The unsaturated soil testing system is an
electro-hydraulic system with closed-loop digital
oedometer-type apparatus for applying matric suc- servo control. A direct measurement or control
tions from near zero values up to 1500 kPa under of the pore water pressure (uw) and pore air pres-
various stress paths. Saturated ceramic stones sure (ua) can be achieved in this system. The pore
with variable High Air Entry Values (HAEV), air pressure (ua) is applied through the top platen
were used for this test. The soils specimens were though a porous stone. The bottom plate consists
compacted in the brass ring of the apparatus to of a HAEV ceramic disc similar to the one used
reach the target dry density based on the standard in the SWCC device. The pore water pressure (uw)
Proctor test results, and then they were completely is applied through the bottom platen. The HAEV
saturated. The drying branch of the SWCCs was disc restricts the air and only allows the water to
obtained from this test as air pressure was applied pass through the disc, provided the water pressure
to a saturated specimen. For each soil, the equilib- does not exceeds the HAEV.
rium moisture contents at different matric suctions All the components mentioned above can be
were measured; and the SWCC for the whole range seen in Figure 4. GCTS CATS v1.6 is the software
of suction values was predicted. The results can be available for the operation of the testing systems
seen in Figures 1 to 3 for the Lulea and Torpsbruk used in this study. The software allows a testing
soils, the County and Dupont clays, and the IOSB load sequence according to the NCHRP 1–28 A
soil, respectively. protocol, which was previously modified Cary and

5
 Table 3. Target degrees of saturation, moisture con-
tents and matric suctions for resilient modulus testing of
the subgrade soils.

Degree of Moisture Matric

Subgrade saturation, content, suction,
soil S (%) w (%) m0 (kPa)

Luleå 30.0 3.8 443.9

50.0 6.4 81.0
79.3 10.1 35.0
94.3 12.0 6.7
@!8- Pofe-air ptHIWt Torpsbruk 30.0 3.6 316.0
@33--- POfe-w<~let' ptHSUfe 50.0 5.4 51.6
B Volumecl'langedevk:t

~ Cl'll pfHlUft
70.1 7.6 23.3
El 97.1 10.4 0.0
!ill GESG 71.0 10.5 236.0
84.0 12.2 52.0
98.0 14.2 1.0
IOSB 78.0 12.1 157
Figure 4. Equipment used to conduct resilient modulus 80.0 12.4 30
test. 83.0 13.0 0.0
County 79.1 22.3 343.3
Zapata (2010) to allow the testing of unsaturated 82.3 25.3 163.3
soil conditions. A double-layered glass chamber 83.6 26.5 70.0
was used for confinement and air was used as Dupont 82.3 26.0 333.3
the confining material in this study. The vertical 85.5 30.9 236.6
88.4 32.3 160.0
deformations on the test specimen were measured
using two Linear Variable Differential Transform-
ers (LVDTs). The LVDTs are mounted directly to retain the homogeneity in moisture distribution.
onto studs buried during the preparation of the For the County and Dupont soils, the soil was
soil specimen. A schematic overview of the triaxial compacted to the target moisture content and dry
system is depicted in Figure 4. density directly and the preconditioning process
was not performed.
After the specimens were at equilibrium, they
4.2 Resilient modulus triaxial testing
were mounted on the triaxial cell pedestal. The air
The resilient modulus soil specimens were pre- entrapped under the ceramic stone and connec-
pared in accordance to the NCHRP1-28 A test tion lines was first flushed out. The confinement
protocol. The soil samples were thoroughly mixed chamber was assembled and a confining pressure
at the required moisture content and stored in a was applied. Then the target matric suction of the
sealed container in a controlled temperature room soils was imposed. Under the applied matric suc-
for approximately 48 hours in order to obtain a tion, the specimen absorbed or released water until
homogenous mix. The materials were then com- moisture equilibrium was reached. This was the
pacted to the targeted dry density based upon the last phase of the moisture equilibration process.
standard Proctor test results. These specimens were Once the equilibration process was completed, the
cylindrical with 102 mm in diameter and 203 mm resilient modulus test was started.
in height (4 × 8 inches). The soils were tested at dif-
ferent moisture contents as shown in Table 3.
The target moisture content of the soil to run 5 TEST RESULTS AND MODEL
the resilient modulus tests was achieved in the PARAMETER ANALYSIS
initial preconditioning of the soil. The specimens
were either saturated or dehydrated to the target The Resilient modulus tests on the soil samples were
moisture content. This was controlled by regular performed according to the NCHRP 1-28 A test
measurements of the weight of the specimens. protocol. The test results showed that an increase
Once the specimen was near to the target moisture in the bulk stress resulted in increase in the resilient
content, the latex membrane was assembled and modulus. Also, an increase in the moisture content,
the specimen was stored in a moisture-sealed bag or decrease in suction, resulted in decrease in the
for about 24 hours. During this period, the speci- resilient modulus. It was also observed that at lower
men was intermittently turned over upside down degrees of saturation, the resilient modulus of the

6
subgrade increased at higher rate with increase in the Table 4. Resilient modulus model regression
bulk stress. The subgrade materials generally exhib- parameters for the subgrade soils.
ited a softening behavior with respect to increase
in the deviator stress. For a constant confinement Model Goodness
parameters of fit
stress, an increase in the deviator stress resulted in Subgrade
decrease in the resilient modulus. However, the sof- soil k1 k2 k3 k4 R2
tening behavior of the materials diminished as the
moisture content was increased. At high degrees Luleå 1489 0.746 −2.633 0.434 0.761
of saturation, the magnitude of the deviator stress Torpsbruk 1524 0.774 −1.470 0.475 0.784
had no significant influence on the resilient modulus IOSG 1480.3 0.420 −2.932 1.650 0.534
results. GESG 993.8 0.288 −1.384 12.711 0.454
The regression parameters for the Cary and County 2402.7 0.136 −5.074 0.665 0.714
Zapata model presented in Eq. (1) were calculated by Dupont 332.5 0.220 −3.140 2.044 0.798
the least square curve fitting on the resilient modu-
lus test data. In total, 123 data points for the Luleå,
150 data points for the Torpsbruk subgrade, 162 data
120,000
points for GESG soil, 108 data points for ISOB soil,
115 data points for County soil and 128 data points 100,000

for Dupont soil, that correspond to different combi- ~ 80,000

nations of stress invariants and matric suctions, were .,::1.. 60,000
used in the statistical analysis. The Solver function in
:a
Microsoft Excel was used for the parametric optimi- 0: 40,000

zation. The coefficient of determination (R2) was cal- 20,000

culated to evaluate the goodness of fit of the model
used. The calibrated model parameters and the 20,000 40,000 60,000 80,000 100,000 120,000

goodness of fit statistics are summarized in Table 4. Measured MR (psi)

Figure 5 to 9 shows the measured versus predicted

results for Lulea, Torpsbruk, IOSG, GESG, County
Figure 5. Measured and predicted resilient modulus for
and Dupont clay respectively. GESG soil (after Cary 2010).
In Table 4, it can be seen that the parameters of
k1, k2, k3 and k4 varies by soil type. It was noted
that the k4 parameter increases with an increase in
the plasticty of the soil with Dupont clay having
500
-~ - -Line of Equality I- .. /

..;; /. .
• Torpsbruk Subgro de1
/
the highest value of 2.0445. The k4 parameter is <0
a..
influenced by the suction of the soil which can be ~400 /

.. :..../ .... .
II>
in turn related to the PI or P200 of the soil. Also, it :::>
can be noticed that the value of k2, which is related
to the influence of the bulk stress on the magnitude
"5
'8
~
c
300 ..
.. -.; .... ..· /
7
of the resilient modulus is lower for the soils with
higher plasticity. Both County and Dupont clay are ~
Cii
- .·:·v·Z~ :-. . .
•• • ..
~ !.

. .
G)
plastic soils which has a low k2 parameter. In fine 200
grained soils, the octahedral stress state parameter .
Q::
"0

.:.
·:~ ~ ~··•.

tV /......••
0 ~ -
-

plays a dominant role on the resilient modulus of '0

!
the soil and in a granular base soil the influence a.. 100
of the bulk stress more. Considering that the both
County and Dupont are fine grained soils, the /
/ t I
value of k2 parameter was found to be low when 0 -V-
0 100 200 300 400
compared with the other soils. Figure 10 shows the
Measured Resilient Modulus [MPa]
influence of P200% on the k2 and k4 parameters of
the model. Figure 6. Measured and predicted resilient modulus for
The R2 value obtained from the regression analy- Torpsbruk soil (Salour, 2014).
sis are 0.76 and 0.78 for Luleå and Torpsbruk sub-
grades, respectively. This is considered to be good
for unbound subgrade materials which generally the proposed Cary and Zapata model that includes
show high data scatters in experimental tests such matric suction as a fundamental stress variable
as the resilient modulus. From the graphs plotted performs well for all kinds of soil. The model pro-
in Figure 6 and given the goodness of fit param- vides a very good fit for both granular and fine
eters presented in Table 4, it can be concluded that grained soils.

7
 500

1 ~ r

0.6
... +

···· ...
·· ...
I.
··· ...
··...
0 .4
100 200 300 400
Measured Resilient Modulus (MPa] 0 .2

Figure 7. Measured and predicted resilient modulus for

10 20 30 40 50 60 70 80 90
Luleå soil (Salour, 2014). P2 00 %

Figure 10. Influence of P200 on k2 and k4 parameters.

140000

~120000

]: 6 CONCLUSIONS
::9100000
"5
"8 Changes in suction of unbound materials is identi-
:!: 80000
fied as one of the major driving factors for sea-
..
c
sonal changes in the resilient modulus. The matric
-~ 60000
a:
.,.. suction, that can be considered an independent
t
'6
40000 stress state variable, directly affects the soil mois-
£ 20000 ture content, which in turn affects the resilient
modulus prediction of subgrade soil materials.
Results from several resilient modulus tests, con-
20000 40000 60000 80000 100000 120000 140000
ducted on six different subgrade soils and at various
Actual Resilient Modulus - (psi)
degrees of saturation, were presented. The test data
Figure 8. Measured and predicted resilient modulus for
were used to determine regression parameters of the
Dupont soil. Cary and Zapata model that takes into account mat-
ric suction as a stress state variable. The prediction
model showed a relatively good correlation with the
experimental data, for soils with low plasticity but
140000
deteriorated for soils with high plasticity. The bulk
~120000 stress state and suction stress state parameters were
~ found to be related to the P200 value of the soil.
~100000
Given the goodness of fit of the prediction model, it
~0
:!: 80000
can be concluded that considering matric suction as
.~ a stress state variable might be a rational approach,
~ 60000 but a more extensive database that includes high
.,a:.. plastic soils is necessary to validate the approach
't 40000
:;; and improve the model presented in this study.
£ 20000

20000 40000 60000 80000 100000 120000 140000

ACKNOWLEDGMENTS
Actual Resilient Modulus - (psi)
This study was partially supported by Federal Avi-
Figure 9. Measured and predicted resilient modulus for ation Administration (FAA). The authors would
County Soil. like to acknowledge the contribution of the data

8
obtained by Dr. Salour and Dr. Cary to the conclu- Tests. Transportation Research Record: Journal of
sions of this study. the Transportation Research Board, No. 2335, Trans-
portation Research Board of the National Academies,
pp. 121–129.
Salour, F., Erlingsson. S. & Zapata, C. 2014. Resilient
REFERENCES Modulus Modeling of Unsaturated Subgrade Soils
with Matric Suction 12th International Conference on
Andrei, D. 2003. Development of a Predictive Model for Asphalt Pavements, ISAP 2014 - Raleigh, NC, United
the Resilient Modulus of Unbound Materials, Doc- States, volume number 2, 1145–1154.
toral Dissertation, Arizona State University, Tempe, Santha, B. L. 1994. Resilient Modulus of Subgrade Soils:
Arizona. Comparison of Two Constitutive Equations, Trans-
ARA, Inc., 2004. ERES Consultants Division. Guide portation Research Record: Journal of the Trans-
for Mechanistic–Empirical Design of New and Reha- portation Research Board, Vol. 1462, Transportation
bilitated Pavement Structures. Final report, NCHRP Research Board of the National Academies, pp. 79–90.
Project 1-37 A. Transportation Research Board of the SWC-150 Fredlund Soil Water Characteristic Device,
National Academies, Washington, D.C. (Available at: 2007. User’s Guide and Reference Manual, Version
http:// www.trb.org/mepdg/) 1.1, GCTS Testing Systems. (Available at: www.gcts.
Cary, C. E. & Zapata, C. E. 2010. Enhancement of the com).
Model for Resilient Response of Soils due to Sea- Uzan. J, Witczak M.W., Scullion. T, & Lytton R. L. 1992.
sonal Environmental Changes Implemented in the Development and validation of realistic pavement
M-EPDG. Transportation Research Record: Journal response models. Proceedings of the 7th International
of the Transportation Research Board. Vol. 2170, Conference on Asphalt Pavements, Nottingham.
pp. 36–44. International Society for Asphalt Pavements (ISAP),
Cary, C. E. & Zapata, C. E. 2011. Resilient Modulus Philadelphia, PA, USA, vol. 1, pp. 334–350.
Testing for Unsaturated Unbound Materials, Inter- Witczak, M.W., & C. E. Zapata. 2000. Development
national Journal of Road Materials and Pavement of Bulk and Octahedral Shear Stress Relationships
Design, Vol. 12, No. 3, pp. 615–638. between Equivalent Pavement Depths. Development
Doucet, F. & Doré, G. 2004. Resilient Modulus and of the 2002 Guide for the Development of New and
Resilient Poisson Coefficient of the C-LTTP Granu- Rehabilitated Pavement Structures, NCHRP 1-37 A,
lar Materials. Proceedings of the Annual Conference Inter Team Technical Report, Arizona State Univer-
of the Canadian Geotechnical Society, Canadian sity, Tempe, Arizona.
Geotechnical Society, Quebec City, Canada. Witczak M.W., X. Qi, & Mirza. W.M. 1995. Use of Non-
Khoury, N. N. & Zaman, M. 2004. Correlation among linear Subgrade Modulus in AASHTO Design Pro-
Resilient Modulus, Moisture Variation and Soil Suc- cedure. ASCE Journal of Transportation Engineering,
tion for Subgrade Soils, Transportation Research Vol. 123, No. 3, 1995, pp. 273–282.
Record: Journal of the Transportation Research Witczak M.W., W.N. Houston, & Andrei, D. 2000. Resil-
Board, Vol. 1874, Transportation Research Board of ient Modulus as Function of Soil Moisture—A Study
the National Academies, pp. 99–107. of the Expected Changes in Resilient Modulus of the
Lekarp, F., Isacsson, U. & Dawson, A. 2000. State of Unbound Layers with Changes in Moisture for 10 LTPP
the Art. I: Resilient Response of Unbound Aggre- Sites. Development of the 2002 Guide for the Devel-
gates. Journal of Transportation Engineering, 126(1), opment of New and Rehabilitated Pavement Struc-
pp. 66–75. tures, NCHRP 1-37 A, Inter Team Technical Report
National Cooperative Highway Research Program, 2004. (Seasonal 2). Arizona State University, Tempe, Arizona.
Laboratory Determination of Resilient Modulus Yang, S. R., Huang, W. H. & Tai, Y. T. 2005. Variation of
for Flexible Pavement Design, In Research Results Resilient Modulus with Soil Suction for Compacted
Digest, No. 285, Transportation Research Board of Subgrade Soils. Transportation Research Record:
the National Academies, Washington, D.C., 52 p. Journal of the Transportation Research Board, Vol.
Rahman, M. S. & Erlingsson, S. 2012. Moisture Sensitiv- 1913, Transportation Research Board of the National
ity of Unbound Granular Materials. Proceedings of Academies, pp. 99–106.
the 4th European Pavement and Asset Management Zapata, C. E., Andrei, D., Witczak, M. W. & Houston,
Conference (EPAM 4). CD-ROM., Malmö, Sweden. W. N. 2007. Incorporation of Environmental Effects in
Salour, F. & Erlingsson, S. 2013. Moisture Sensitive and Pavement Design, International Journal of Road Mate-
Stress Dependent Behavior of Pavement Unbound rials and Pavement Design, Vol. 8, No 4, pp. 667–693.
Materials from Insitu Falling Weight Deflectometer

9
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Investigating correlations between stiffness indexes for Brazilian soils

Marcus dos Reis

Coordinator and Professor of Universidade Paulista (Paulista University)—UNIP—Chácara Santo Antonio.
Technical Director of JS Global Construção Civil e Assessoria em Obras and CAPT—Centro de Apoio a
Pesquisa Tecnológicas, SP, Brazil

Rita Moura Fortes

Professor of Civil Engineering, Universidade Anhembi Morumbi (Anhembi Morumbi University) and Research
Director of ERI— “Engineering and Research Institute” Pesquisas Ltda., São Paulo, Brazil

ABSTRACT: This paper presents a research developed using samples collected in four (04) different
regions in two (02) cities in the São Paulo state for different types of tropical and non-tropical soils
through DCP (Dynamic Cone Penetration) testing. The samples were analyzed in the laboratory by means
of characterization and molding of test specimens for spindling of the DCP in the laboratory and field
to obtain a correlation between these results. In the research, ten penetrations were performed using the
DCP and also ten drilled HILF cylinder for the laboratory determination of the CBR (California Bearing
Ratio). Seven points were collected in each of the four regions of these two cities. With the analysis of the
results, it can be observed that the results obtained in the tests agree with the correlations stipulated by
the standard. That is, the equations presented are applicable, considering an acceptable margin of error.

1 INTRODUCTION was used to estimate in situ CBR or shear strength

using the appropriate correlation, for example, the
The methodology used to pavement design in equation below recommended by the United States
Brazil is based on the resistance of the materials Army Corps of Engineers:
obtained through the CBR (California Bearing
Ratio) results. CBR = 292/(DCP1.12) (1)
To establish the in situ CBR value was used
Dynamic Cone Penetration (DCP) Testing. This equation is used for all soils except for CL
In Brazil, there are no standards for the test soils below CBR 10 and CH soils. For these soils.
running and how to correlate the DCP with the following equations are recommended by the
the CBR, therefore, we use to consider the US Army Corps of Engineers:
American Standard ASTM D6951/D6951M-09:
“Standard Test Method for Use of the Dynamic CL soils CBR < 10: CBR 5 1/(0.017019*DCP)2 (2)
Cone Penetrometer in Shallow Pavement
CH soils: CBR 5 1/0.002871*DCP (3)
Applications”.
Because of its practicality and efficiency, this
The DCP index unit is the penetration in mm
test has been widely used to determine the soil sup-
per blow (mm/blows).
port capacity.

2.2 Place of test run

2 METHODOLOGY
To analyze the correlation between the value of the
DCP and in situ CBR or shear strength, accord-
2.1 DCP executive process
ing to the soil type, it was decided to carry out the
Following the procedure recommended in ASTM tests in three regions in the area of the CAPT—
D6951/D6951M-09, the test was performed. The Technological Research Support Center located in
estimated in situ CBR was computed using the São Roque city (Figures 1, 2 and 3) and in Cotiac-
DCP index for each set of readings. ity, both located in state of São Paulo, near São
The penetration per blow was plotted on a scale Paulo city (Figure 4), for different types of soils
reading or total depth. The penetration per blow (lateritic and not lateritic).

11
 In Figures 1, 2 and 3 are presented the schematic
H~ 7 Ht
sketches of the points where the in situ tests were
performed and the sample collected.
c: c.
)(
H~
H, H2 ~X c. Ho X
X )(
X
Ca
c, 2: Co

light post~ G 2.3 Characterization of materials used

Hi= Hilf drilling
Sampling, sample collection, characterization tests.
C= DCP testing
South African Cone—DCP and drilled Hilftest.
Figure 1. Schematic diagram of the location of the Following the samples up to 2 m deep in each
points tested in São Roque—SP (CAPT)—Region 1. region, they were sent to the laboratory for charac-
terization and tests performance.
Figure 5 presents the preparation of the sam-
ples in general lines. Each sample was air-dried,
H,
~ H3 H4 H5 H6 H7 Ha Hs H,o de-routed, homogenized, quenched and later the
~ ~ ®
fractions were defined for the performance of the
X X X ® ® 0 ®
c, X X X X
~ c3 c4 c5 41 ~ Ca Cs C,o
tests.
<I-- The samples were classified according to the
ASTM D2487 – 11. AASHTO Soil Classification
System (ASTM. 2015) and MCT Methodology for
~
Rapid Classification of Tropical Soils (FORTES &
CAPT

light post~ MERIGHI. 2003).

Figure 6 shows the test specimens (disk method)
Figure 2. Schematic diagram of the location of the used in the study developed by Fortes; Merighi
points tested in São Roque—SP (CAPT)—Region 2. (2003) to the MCT methodology, classifying the
soil preliminarily and the feasibility using the soil
dosage study with additive (Figueiredo and Fortes.
2008).
The proctor test was performed with the normal
energy, according to the ASTM D698–12e2 (2012).
H, H.
The results obtained are also presented in Table 1
c, c,
(Figure 7).
The CBR results according to the ASTMD1883–
c, 16 (2016) are presented in Table 3 (Figures 7, 8 and 9).
Figure 9 presents the rupture of the specimens.
It was analyzed (a) the molded in laboratory CBR
49 km of Raposo Tavares Highway Cylinder (b) the milled in the field specimen CBR
Figure 3. Schematic diagram of the location of thep-
oints tested in São Roque—SP (CAPT)—Region 3.

Figure 5. Preparation of each sample (drying in air)

and separation fractions to be tested.

Hi= Hilf drilling

C= DCP testing

Figure 4. Schematic diagram of the location of the Figure 6. Molding of specimens for the disk method
points tested in Cotia—SP (CAPT)—Region 1. (FORTES; MERIGHI. 2003).

12
Table 1. Result of soil characterization.

Sample ST – 01 ST – 02 ST – 03 ST – 04

Depth (m) 0.10/2.00 0.15/1.8 0.95/1.70 0.20/1.10

Place São Roque São Roque São Roque Cotia
Region 1 Region 2 Region 3 Region 1
Description and Identifi-cation Light brown- Clay Silt Sandy. Silt. clayed Sand Clayed.
of Soils (ASTM D2499/09) claysand light brown sandy. brown pink
Maximum dry unit weight (γd.max). 1.908 1.580 1.559 1.578
(kg/m3) (ASTM D698/12)
Optimum water content (wopt) (%) 12.1 20.5 19.3 18.4
CBR (%) (ASTM D1883-16) 16.0 5.2 8.0 4.5
Soil expansibility (%) 0.3 3.1 0.78 1.9
LL (%) 22.1 38.0 58.0 40.0
PL (%) 14.2 21.0 31.06 22.33
PI (%) 7.9 17.0 26.94 17.67
Group index 2.0 10.8 12.0 9.0
FORTES, MERIGHI (2003) NA-NS′ NS′ LG′ NS′
CLASSIFICATION
AASHTO Soil Classification System A-4 A-6 A-7-5 A-6

Figure 7. Sample separation. overview from start Sieve

Analysis. sample preparation for Proctor and CBR tests
in the laboratory.

Figure 9. Determination of CBR value: (a) specimen

molded in the normal energy in the laboratory and (b)
specimen obtained by Hilfdrilling. (c) determination of
the DCP in the specimen molded in the normal energy
in the laboratory.

Table 2. Results of particle-size distribution (Grada-

Figure 8. Test specimens after molding for the CBR tion) of soils using Sieve analysis.
determination DCP test molded in the laboratory in the
compaction same energy of the field. The specimens were (ASTM D6913-04(2009)e1)
inserted into the tank.
Sieve (mm) Passing

test (with the undisturbed sample collected after Sample ST – 01 ST – 02 ST – 03 ST – 04

the Hilfcylinder was drilled) and (c) DCP test per-
formed on specimen molded in the normal energy 19.1 — — — —
9.52 100.00 100.00 100.00 100.00
in the laboratory.
4.75 100.00 100.00 100.00 99.44
In Tables 1 and 2 are presented the results
2.00 99.23 95.36 97.61 98.86
obtained in the characterization test of the soils
1.19 98.20 95.16 95.24 98.55
used in this research.
0.6 96.00 94.02 81.02 93.82
In Figure 10 it can be noticed in the right side
0.42 89.90 93.52 74.62 88.89
the schematic sketch of the South African Cone— 0.25 88.30 92.68 66.37 82.23
DCP, the center of the field test and the right the 0.149 86.80 91.47 60.29 76.06
detail of the DCP and Hilf Cylinder drilled in the 0.074 44.30 85.42 53.80 65.19
ground, spaced 50 cm each other.

13
Table 3. Penetration obtained in the soil of São Roque—SP (CAPT)—region 1.

Sample 1 Sample 2 Sample 3 Sample 4 Sample 5 Sample 6 Sample 7

Place (cm) (cm) (cm) (cm) (cm) (cm) (cm)

L0 4.3 5.2 5.3 6.5 4.8 4.5 5.9

L1 6.0 8.1 8.0 11.4 7.1 6.5 9.7
L2 7.3 10.6 10.1 14.0 9.0 8.1 12.4
L3 9.1 13.0 12.5 16.4 11.1 10.1 14.8
L4 11.2 15.0 14.5 18.0 13.1 12.2 17.0
L5 14.1 17.0 16.4 – 15.6 14.8 18.9
L6 17.5 18.6 18.0 – 18.1 17.8 20.7
L7 21.1 – – – 20.3 20.2 21.2
L8 23.2 – – – 23.3 23.3 –
L9 25.7 – – – 25.5 25.6 –
L10 27.8 – – – 27.7 27.8 –
L11 31.8 – – – 31.6 31.7 –

Figure 12. Equipment adapted for testing and to deter-

mination of CBR values.

Figure 10. (a) the dynamic penetrometer DCPsche-

matic, (b) DCP test being carried out in situ and (c) the 0.5 mm piston and the 150 mm cylinder with the
DCP realization de-filing right and Hilf Cylinder drilled extensometer to measure the expansion, the Hilf
in the ground. cylinder with the 100 mm collar, the dynamometric
ring, the extensometer, the 0.33 mm piston, (C) in
the press for the determination of the CBR in the
laboratory.

2.4 Efficiency of the test

To prove the efficiency of the test specimens, these
tropical soils were analyzed in the laboratory to
Figure 11. Specimens prepared for the tests. verify the discrepancy of the results.

In Figure 11 are presented from left to right, (A) 3 RESULTS

the test specimens (CP1 and CP2) used to perform
the tests with the soils collected in the 4 regions, The DCP test was performed in 10 points and the
(C) determination of the DCP in the experimental Hilf cylinder was also drilled for undisturbed sam-
section (field) by means of the drilling of the Hilf ple collection at these points. Tables 3 to 6 show the
cylinder to collect the undisturbed sample, (D) results obtained during the execution of the DCP
determination of the expansion and the CBR in test in this section.
the deformed and undisturbed samples. In Tables 7 to 10 are presented the data obtained
In Figure 12, it can be observed the equipment’s in the section tests in the two locations (São Roque
adapted for the tests until the moment of the rup- and Cotia cities) in the four regions (3 in São
ture, being that from left to right we illustrate, Roque City and 1 in CotiaCity). It was used the
(A) the Hilf specimen immersed in water with the correlation between DCP and CBR (Equation 1)
overload and adapted extensometer for the Hilf and the results to show that the CBR are close to
cylinder with a diameter of 100 mm, (B) as well those obtained in the laboratory tests, as presented
as the dynamometric ring, the extensometer, the below.

14
Table 4. Penetration obtained in the soil of São Roque—SP (CAPT)—region 2.

Sample 1 Sample 2 Sample 3 Sample 4 Sample 5 Sample 6 Sample 7

Place (cm) (cm) (cm) (cm) (cm) (cm) (cm)

L0 17.6 17.7 17.7 17.6 17.6 17.7 19.7

L1 30.3 30.0 30.2 30.2 30.2 30.1 26.9
L2 35.4 35.6 35.5 37.5 36.5 36.5 31.2
L3 39.8 39.5 39.7 39.7 39.7 39.6 38.5
L4 43.1 43.2 43.2 43.1 43.1 43.2 41.4
L5 47.0 47.1 47.1 43.0 45.0 45.1 44.1
L6 50.5 50.6 50.6 50.5 50.5 50.6 46.8
L7 54.3 54.4 54.4 54.3 54,3. 54.4 52.4
L8 58.7 58.9 58.8 58.8 58.8 58.8 56.6
L9 62.4 62.4 62.4 62.4 62.4 62.4 60.6
L10 65.2 65.0 65.2 65.2 65.1 65.1 63.8
L11 68.1 68.3 67.2 67.2 67.7 67.7 66.4

Table 5. Penetration obtained in the soil of São Roque—SP (CAPT)—region 3.

Sample 1 Sample 2 Sample 3 Sample 4 Sample 5 Sample 6 Sample 7

Place (cm) (cm) (cm) (cm) (cm) (cm) (cm)

L0 4.3 4.4 4.9 4.7 4.8 4.6 19.7

L1 6.0 6.2 5.3 4.8 5.0 5.3 26.9
L2 7.3 7.5 6.8 6.4 6.6 6.8 31.2
L3 9.1 9.6 8.5 7.9 8.2 8.5 7.5
L4 11.2 11.1 10.1 9.6 9.9 10.2 9.0
L5 14.1 13.9 12.6 11.9 12.2 12.7 11.1
L6 17.5 17.2 15.7 14.9 15.3 15.8 13.8
L7 20.1 20.8 19.2 18.3 18.7 19.3 17.1
L8 23.2 23.1 21.6 20.9 21.2 21.7 20.0
L9 25.7 25.2 24.2 23.7 24.0 24.3 22.6
L10 27.8 26.9 26.3 26.0 26.2 26.4 25.0
L11 29.9 28.2 28.5 28.2 27.2 27.9 28.9

Table 6. Penetration obtained in Cotiasoil—SP—region 1.

Sample 1 Sample 2 Sample 3 Sample 4 Sample 5 Sample 6 Sample 7

Place (cm) (cm) (cm) (cm) (cm) (cm) (cm)

L0 6.0 5.9 5.5 5.5 5.6 5.1 4.5

L1 7.2 7.0 6.9 8.6 9.7 9.3 6.2
L2 8.3 7.6 7.4 11.0 12.0 12.4 7.2
L3 9.5 8.3 8.1 14.1 14.2 14.7 7.9
L4 10.9 9.0 9.0 15.7 17.1 16.6 8.5
L5 12.3 10.1 10.0 16.6 20.0 20.0 9.1
L6 13.1 11.0 10.9 17.5 22.4 23.8 9.4
L7 15.4 12.4 11.8 18.4 24.5 26.4 10.0
L8 – – – – – – –
L9 – – – – – – –
L10 – – – – – – –
L11 – – – – – – –

(deformed sample) using the correlation of

In Tables 7 to 10 were used the following
ASTM. D6951/D6951M-09;
symbols:
**CBRV: DCP obtained in one of the topside of
*CBR: value of DCP in the optimum moisture the CBR test in the optimum moisture content
content value laboratory molded specimen value laboratory molded specimen (deformed

15
Table 7. Results obtained in Region 01 of São Roque in the deformed and undisturbed samples.

SÃO ROQUE—SP (CAPT)

Region 1

Deformed sample Undeformed sample

Number of DCP *CBR DCP **CBRV CBRomc DCP ***CBRH CBRC

specimens (mm/blows) (%) (mm/blows) (%) (%) (mm/blows) (%) (%)

1 122.22 15.0 96.43 15.9 14.0 28.91 7,0 6.7

2 204.67 15.8 161.48 16.3 13.8 31.00 6.5 6.2
3 364.28 15.2 287.42 15.7 15.0 30.00 6.9 6.5
4 316.22 15.5 249.50 16,0 12.7 30.00 7,0 6.5
5 263.35 15.1 207.79 15.9 13.4 28.73 7.1 6.8
6 193.03 15.9 152.30 16.3 14.9 28.82 7.6 6.8
7 209.03 16.0 164.93 16.0 14.0 30.29 6.8 6.4
Mean – 15.5 – 16.0 14.0 – 7.0 6.5
SD – 0.41 – 0.22 0.81 – 0.33 0.22
CV – 2.63 – 1.37 5.77 – 4.78 3.34

Table 8. Results obtained in Region 02 of São Roque in the deformed and undisturbed samples.

SÃO ROQUE—SP (CAPT)

Region 2

Deformed sample Undeformed sample

Number of DCP *CBR DCP **CBRV CBRomc DCP ***CBRH CBRC

specimens (mm/blows) (%) (mm/blows) (%) (%) (mm/blows) (%) (%)

1 177.0 3.1 139.65 4.3 2.0 61.9 3.5 2.9

2 215.5 4.0 170.03 5.0 1.8 62.1 3.9 2.9
3 199.0 3.9 157.01 5.1 2.0 62.0 2.8 2.9
4 312.1 4.3 246.25 4.9 2.5 61.0 3.5 2.9
5 252.5 4.3 199.22 5.2 1.9 61.5 3.8 2.9
6 235.5 4.4 185.81 5.3 2.3 61.6 3.8 2.9
7 215.8 4.5 170.23 4.9 2.5 60.4 3.7 3.0
Mean – 4.1 – 5.0 2.1 – 3.6 2.9
SD – 0.48 – 0.33 0.3 – 0.37 0.05
CV – 11.76 – 6.57 13.4 – 10.44 1.57

sample), using the correlation of ASTM. D6951/ SD = standard deviation;

D6951M-09; CV = coefficient of variation;
CBRomc: Value of the CBR, obtained after 4 days
of immersion, in in the optimum moisture It is noteworthy that since points (1, 2, 3 and 7)
content value laboratory molded specimen have (mm/blows) and the CBRC value correspond-
(deformed sample), using the CBR press on ing CBR 9.2% to 14.9% are distant from the other
the other topside of the specimen used to DCP points, this should be discarded for generation of
determination (**CBR); the correlation that best corresponds to reality. In
***CBRH obtained in the undeformed sample of order to use it, it would be necessary to fill in these
the HILF test specimen (undeformed sample) in empty intervals with new results in the experimen-
the field energy; tal section.
CBRC is the calculated value of CBR, using It can be noted that in the specimen (CBRomc)
the correlation presented in ASTM. D6951/ that due to the DCP crimping at the top of
D6951M-09, by DCP obtained in field; the other side to that used in the CBR piston

16
Table 9. Results obtained in Region 03 of São Roque in the deformed and undisturbed samples.

SÃO ROQUE—SP (CAPT)

Region 3

Deformed sample Undeformed sample

Number of DCP *CBR DCP **CBRV CBRomc DCP ***CBRH CBRC

specimens (mm/blows) (%) (mm/blows) (%) (%) (mm/blows) (%) (%)

1 455 8.3 359 8.0 7.7 26.5 7.7 7.4

2 705 8.2 556 8.0 7.5 25.6 7.5 7.7
3 560 8.0 442 7.9 7.5 25.9 7.3 7.6
4 1390 7.7 1097 7.8 7.4 25.6 7.4 7.7
5 1005 7.9 793 7.8 7.5 24.7 7.7 8.0
6 1040 8.0 821 8.0 7.7 25.3 7.2 7.8
7 783 8.4 617 8.1 6.9 26.3 7.2 7.5
Mean – 8.1 – 7.9 7.4 – 7.4 7.7
SD – 0.24 – 0.11 0.27 – 0.21 0.20
CV – 3.0 – 1.43 3.62 – 2.87 2.58

Table 10. Results obtained in Region 01 of Cotia in the deformed and undisturbed samples.

COTIA—SP

Region 1

Defprmed sample Undeformed sample

Number of DCP *CBR DCP **CBRV CBRomc DCP ***CBRH CBRC

specimens (mm/blows) (%) (mm/blows) (%) (%) (mm/blows) (%) (%)

1 475 4.0 375 4.5 3.3 22.0 5.3 9.2

2 615 3.8 485 4.4 3.6 17.7 4.4 11.7
3 545 3.9 430 4.4 3.7 16.9 3.8 12.3
4 655 3.9 517 4.4 3.8 26.3 4.8 7.5
5 770 3.7 608 4.6 2.9 35.0 5.3 5.4
6 500 4.0 394 4.5 3.7 37.7 4.9 5.0
7 830 4.2 655 4.5 3.3 14.3 3.7 14.9
Mean – 3.9 – 4.5 3.5 – 4.6 9.43
SD – 0.16 – 0.08 0.32 – 0.66 3.72
CV – 4.1 – 1.69 9.22 – 14.3 39.44

penetration, the expansion results showed great Due the high dispersion, the results highlighted
variations, based on the expansion obtained in the in red in the tables were not considered in the
characterization of the sample. graphs and in Table 11, presented below.
The CBRomc presented a lower CBR value than It was molded (CP1) in the normal energy and
the *CBR. This was due to the perforation of the the DCP (mm/blows) was determined to deter-
specimen because of the DCP test, which was car- mine the CBR in the optimum moisture content
ried out before placing the specimens for 4 days of perD6951/ASTM D6951M-09.
immersion and before the rupture. ** It was molded (CP2) in the normal energy and
Among the several of mathematical models used at the top of the test body, on the opposite side to
in Microsoft Excel software, it was verified that the be carried out the penetration of the piston for the
“power” model was the one that best fit the experi- determination of the CBR, the DCP (mm/blows)
mental points. The equations of the CBR value was determined, before being immersed for deter-
and R2 corresponding to the model are indicated mination of Expansion value and the CBR in the
in Table 11. The equations 4 to 15, obtained by optimum moisture content V) per ASTMD6951/
correlations are presented. D6951M-09 e ASTM D1883-16.

17
Table 11. Summary of the equations obtained in the correlations.

Equation City Region Sample Specimen Correlation R2

4 SÃO ROQUE 1 Deformed CP 1 CBR = 57.002 DCP (−0.261)* 0.8526

5 CP 2 CBR = 63.796 DCP (−0.293)** 0.9434
6 Undeformed – CBR = 7088.3 DCP (−2.037)*** 0.965
7 2 Deformed CP 1 CBR = 0.2203 DCP (0.422)* 0.9617
8 CP 2 CBR = 0.239 DCP (0.4317)** 0.9307
9 Undeformed – CBR = 4E-16 DCP (8.9426)*** 0.976
10 3 Deformed CP 1 CBR = 9.3698 DCP (−0.033)* 0.8876
11 CP 2 CBR = 9.2975 DCP (−0.033)** 0.8876
12 Undeformed – CBR = 275.83 DCP (−1.116)*** 0.9989
13 COTIA 34 Deformed CP 1 CBR = 16.636 DCP (−0.24)* 0.9607
14 CP 2 CBR = 140.45 DCP (−0.604)** 0.9644
15 Undeformed – CBR = 1.166 DCP (0.4277)*** 0.9852

***The HILF cylinder was pinned in the field in

the same region as the DCP (mm / blows) to deter- *CBRC (%) X CBRH
mine the (CBR C) per ASTM D6951/D6951M-09
e ASTM D1883-16.
...
.-.-~· ··
.......•...
...····
4 ANALYSIS OF RESULTS
...
··········· CBRc = 1,2388/CBRH1•7349
It should be noted that the undisturbed specimens ~······ R2 = 0,9799
obtained by the Hilf cylinder drilled, after extrac-
CBRH
tion, were wrapped with paraffin to maintain the e *CBR C(%) ASTM 0695 I M-09 (DCP)
structure and moisture field content of the, the
compaction degree, moisture condition and CBR Figure 13. Correlation between the CBR in the speci-
value to perform the correlations between sample men obtained by the HILF specimen obtained in the field
molded in the laboratory, collected in the field, (undeformed sample) and DCP, São Roque, Region 2.
ruptured in the press and tested in situ with the
DCP.
It should be noted that the 10 specimens (undis- the standard ASTM D6951/D6951M-09, validat-
turbed sample) obtained with a Hilf cylinder drill ing their use for these soils.
were placed in a wet chamber until the laboratory The best correlation obtained for undeformed
tests, to maintain the it’s moisture content. specimens is presented in Figure 13.
The values obtained in the *CBR were close to In this figure the values obtained in the deter-
those of the **CBRv. minations with the Cotia soil were not considered,
The values of **CBRv showed high dispersion since they had a high dispersion.
because the specimen was shaken by the spike of In the case of the comparison between the
the DCP in one of its top and then placed 4 days results obtained in the sample molded in the labora-
in immersion. To avoid this problem, it is recom- tory and the value of the DCP obtained in the field,
mended that two specimens be molded in the the values did not present a good correlation since
optimum moisture content, so that one of them is the soil in situ presented a degree of compaction
penetrated by the DCP and in the other specimen smaller than the one obtained in laboratory, since
will be done the expansion and CBR determina- in the field the soil was in the natural state. In this
tion test. way, the correlation obtained presented a low R2.
For this reason, it was decided to disregard some The authors intend to continue the research,
values in the correlations presented in Table 11. increasing the collection of results in order to bet-
ter adjust the correlations for Brazilian soils.
5 CONCLUSION
ACKNOWLEDGMENTS
Analyzing the results obtained in the test it can be
noted that their performance was close to those The authors are grateful to the JS GLOBAL
estimated by the correlation equation presented in Civil Construction Consultancy laboratories and

18
Paulista University—UNIP for collaborating/ ASTM D3282-15. Standard Practice for Classification of
executing tests of this research, as well to the team Soils and Soil-Aggregate Mixtures for Highway Con-
that contributed a lot to the development of this struction Purposes. ASTM International. West Consho-
research, Almir Donizette Vicente Gouvêa, Fran- hocken. PA, 2015. www.astm.org
ASTM D4318-10e1. Standard Test Methods for Liq-
cisco Danilo Silvério Caracas, Dhiego de Oliveira uid Limit. Plastic Limit. and Plasticity Index of Soils.
Melo, Marcos Tiharu Takeda and Roberta Mércia ASTM International. West Conshohocken. PA, 2010.
F. dos Santos. www.astm.org
ASTM D6913-04(2009)e1. Standard Test Methods for
Particle-Size Distribution (Gradation) of Soils Using
REFERENCES Sieve Analysis. ASTM International. West Consho-
hocken. PA, 2009. www.astm.org
ASTM. D6951/D6951M-09: Standard Test Method for ASTM D698-12e2. Standard Test Methods for Laboratory
Use of the Dynamic Cone Penetrometer in Shallow Compaction Characteristics of Soil Using Standard
Pavement Applications. Pensilvânia/EUA. 2009, 7p. Effort (12 400 ft-lbf/ft3 (600 kN-m/m3)). ASTM Inter-
ASTM D1883-16. Standard Test Method for California national. West Conshohocken. PA. 2012. www.astm.org.
Bearing Ratio (CBR) of Laboratory-Compacted Soils. Figueredo. Alessandra C. Andrade; Fortes. Rita Moura;
ASTM International. West Conshohocken. PA, 2016. “The Use of The Disk Method of MCT Methodol-
www.astm.org ogy. in The Dosage of Soil Cement for The Execution
ASTM D2487-11. Standard Practice for Classification of Cement Bases”. First International Conference on
of Soils for Engineering Purposes (Unified Soil Transport Infrastructure (ICTI 2008). Beijing, China.
Classification System). ASTM International. West April. 23–26. 2008.
Conshohocken. PA, 2011. www.astm.org Fortes. Rita Moura & Merighi. João Virgilio. The use of
ASTM D2488-09a. Standard Practice for Description MCT methodology for rapid classification of tropical
and Identification of Soils (Visual-Manual Procedure). soils in Brazil. IJP - International Journal of Pavements.
ASTM International. West Conshohocken. PA, ISSN 1676–2797. Vol. 2. No. 3. September 2003.
2009. www.astm.org pp. 1–13.

19
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Analysis of compaction and California Bearing Ratio (CBR) test results

of a mixture of iron ore tailings with Brazilian tropical soil in different
proportions for road construction purposes

A.A.N. Dantas & T.A. Mendes

University of Brasília (UnB), Brasília – DF and Federal Institute of Education, Science and Technology
of Goiás (IFG), Brazil

A.J.A. Posse, L.F.M. Ribeiro & C.M.C. Gurjão

University of Brasília (UnB), Brasília – DF, Brazil

ABSTRACT: Mining in Brazil plays a very important role in economic and social development, and it
contributes directly to a better quality of life for people. However, mining activities, although performed
in a responsible and sustainable manner, can cause harmful damages to the environment. In this paper,
different mixtures of iron ore tailings with Brazilian tropical soil were evaluated in order to obtain a sig-
nificant increase in the compaction test and California Bearing Ratio (CBR) results. The materials studied
correspond to iron ore tailings and a Brazilian tropical soil, both from the Northeast region of Brazil.
For the Compaction tests, three different compaction energies were used: normal, intermediate and modi-
fied energy. Thus, the values of optimal moisture content and dry density obtained ranged from 8.07%
to 11.53%, and 2.11 g/cm3 to 2.58 g/cm3, respectively. It was verified that in all the compaction tests the
optimal moisture content decreased with the increase of the dry bulk and compaction energy. Regarding
the CBR tests, it was found that the mixture can be used up to the sub-base layer of pavements, since,
according to Brazilian standards established by the National Department of Infrastructure and Transport
(DNIT), materials with a CBR greater than or equal to 60.0% are required for the base layer in pavements.
Therefore, the use of iron ore tailing as a constituent material in the sub-base layers in pavements offers a
significant contribution in the improvement of the structural behavior of tropical soils, as well as a better
environmental use of the iron ore tailings.

1 INTRODUCTION which is essentially magnetic, may have 60% to 70%

iron content. In addition to the high iron contents,
The current work is research developed in the state they present low levels of impurities, with phospho-
of Rio Grande do Norte in the municipality of rus ranging from 0.03% to 0.05%, silica from 1.5%
Jucurutu, located in the Northeast region of Brazil to 2.0%, and a fire loss of less than 1.5%. Despite
about 233 km from the state capital Natal. The region the high ore rates, there is a lot of waste.
is formed by tropical soils, and the territory área being The great concern with the environment today,
934 km2 making it the ninth municipality with greater the debate over reused materials is increasing, des-
extension of Rio Grande do Norte. Also, the región tination and quantity of materials that are released
has a large reserve of high quality iron ore, making the into the environment has become something of con-
state of Rio Grande do Norte an iron ore exporter. cern. Measures need to be taken to minimize the
Mhag Serviços e Mineração S/A, a mining com- exploitation of new deposits and to conserve nature.
pany, operating in Jucurutu—RN, in 2012 pro- The environmental concern is due to the reduction
duced one million tons of sinter feed. The sinter of natural resources that have become increasingly
feed corresponds to the fine fraction of iron ore. scarce and to the environment that does not have
Brazil, specifically the region of Carajás in the state the capacity to absorb the current amount of waste
of Pará, is known for the quality of it’s sinter feed. without there being an environmental imbalance.
Transportion are by trucks, to Juazeirinho—PB, Thus, the objective of this paper is to evaluate
and by train, Companhia Ferroviária do Nordeste different proportions (mixtures) of iron ore resi-
(CFN), to the Port of Suape in Pernambuco (PE). due and Brazilian tropical soil in order to obtain
The results of the research carried out in several significant gains in compaction tests and Cali-
blocks of the Bonito Mine, inserted within a 150 km fornia Bearing Ratio (CBR) for roads and to be
radius, indicated the existence of Itabirito type ore, use on base layers, sub-base and reinforcement of

21
sub-bed, thus being an alternative to the explora-
tion of new deposits, contributing in a positive way
to a less degradation of the environment.

2 MATERIALS AND METHODS

The compaction tests were performed using

three compaction energies recommended by the
Brazilian standards. California Bearing Ratio
(CBR) tests were performed on intermediate and
modified energies. The tests were carried out at
the Soil Mechanics Laboratory at the Federal Uni-
versity of Rio Grande do Norte (UFRN). Table 1
shows the steps, tests involved, Brazilian standards
used and quantitative of tests. Figure 1. Waste deposit of pure iron ore.
For the compaction test, forty-five tests were
performed: fifteen tests in normal energy, fifteen
tests of intermediate energy and fifteen tests in of this material for the different compositions of
modified energy. In each energy three tests were the mixtures. Soil samples were collected on June
performed for the pure iron ore residue, pure gran- 10, 2014 at a specific point located on the banks of
ular soil, the mixture of 15% of iron ore residue the road itself. The exact Geographic location of
and 85% of granular soil, 25% of iron ore residue the deposit is 5°53′00″ S and 35°28′06″ W.
and 75% of granular soil, 50% iron ore residue and
50% granular soil, referred to as FP, SP, F15S85,
2.1 Tests applied to paving
F25S75, F50S50, respectively.
For the CBR test, thirty tests were carried out: Tests applied to paving. At this stage, Proctor
fifteen tests on the intermediate energy and fifteen compaction and California Bearing Ratio (CBR)
tests on the modified energy. In each energy, three assays were performed.
tests were performed for the pure iron ore (FP), The compaction and CBR tests are responsible
pure granular soil (SP), the mixture of 15% of iron for determining the parameters used in the design
ore residue and 85% of granular soil (F15S85), of pavements, as well as directly interfering with the
25% of Iron ore residue and 75% granular soil acceptability of the material for paving applications.
(F25S75), 50% iron ore residue and 50% granular Approximately four kilograms of material was
soil (F50S50). used for each compaction test. The materials con-
The group MHAG Serviços e Mineração S/A, tained in the hermetically sealed bags contained
responsible for the exploration of the Bonito mine, an initial moisture content of less than 1.0%, cor-
was the supplier of the residue of the iron ore for responding to the location where it was stored in the
the purposes of this work. The deposit is located mine and on the bank of the BR-304. Water was
in the municipality of Jucuturu, in the microregion added until a moisture content of 6.0% was reached.
of the Açu Valley, with geographical location of This was the initial moisture value for the tests. At
6°02′02″ S and 37°01′12″ W. Samples of the iron each new point, an amount of water was added,
ore beneficiation residue were collected at their which corresponded to 2.0% by weight of the mate-
disposal site, near the mine (Fig. 1). rial. The mixture of the material to be compacted
Regarding the granular soil used, the company was given inside a plastic tray in order to avoid loss
SBS Engenharia, responsible for the duplication of of water. Further, while the test was performed, the
the BR-304, federal highway in the municipality of remainder of the mixture remained inside the tray,
Macaíba in Rio Grande do Norte, was the supplier sealed with plastic wrap so that the water contained
in the mixture did not evaporate. Between five and
Table 1. Summary of tests performed. seven points were obtained for each test performed.
Based on the results, a simple arithmetic mean
Standard Numbers of of the compaction tests results of the same energy
Steps Tests (ABNT) tests was used, so that the mean represented the criteria
of apparent specific mass and optimum moisture
Paving tests Compaction NBR 45 content of the specimens to be cast in the CBR tests.
7.182/1986
For each CBR test, approximately 6 kilos of
CBR NBR 30
9.895/1987
material were required, either pure or in the differ-
ent mixtures. Water has been added to the materials

22
to achieve optimum moisture. Prior to commencing 3.2 Tests applied to paving
compaction, filter paper was placed on the bottom
The results of the Proctor compaction tests on the
of the cylinder in order to decrease the excessive
normal, intermediate and modified energy and the
water flow during the test. As the layers were com-
respective dry densities (ρs) and optimal moisture
pacted, the waiting material stored in the tray was
contents (wot) for the FP, F50S50, F25S75, F15S85
covered with plastic wrap in order to avoid excessive
and SP samples are shown in Figures 3 to 7.
evaporation of the water present in the sample and
From Figures 3 to 7 it can be seen that as the
thereby prevent the compaction of the material with
energy applied in the compaction increases, the spe-
moisture contents other than the optimum moisture.
cific dry mass increases, although in a small magni-
During the test, a sample of the second layer was
tude, the optimum moisture content decreases. This
collected to identify the actual moisture content of
is behavior compatible with soils in a general way.
the material during the test. In the period of four
Pereira (2012) using compacted soil-cement mix-
days the cylinder with the compacted material was
tures with 3.0%, 6.0% and 9.0% content, it obtained
submerged in water, and the expansion reading was
dry densities of 1.75 g/cm3, 1.77 g/cm3 and 1.78 g/
performed daily. At the end of the period the test
cm3, respectively. Castro (2008) found dry density of
specimen was broken, determining the results of the
1.96 g/cm3 for the pure soil and 1.93 g/cm3, 1,905 g/
CBR tests.
cm3 and 1.876 g/cm3 respectively for the contents
of 10.0%, 20.0% and 30.0% of the kaolin residue.
Nóbrega & Melo (2009), using construction waste
3 RESULTS AND DISCUSSIONS
instead of paving aggregates, obtained a specific dry
density of 1.818 g/cm3 and wot of 14.3%. Already
In this item the presentation and discussion of
(Pereira, 2012) has already found values of ρdmax of
the results obtained in the tests are performed.
The results of the compaction tests and Califor-
nia Bearing Ratio (CBR) are discussed. The results 2,8

obtained were compared with other results pub- '1u3 2,7

Modified
Intermediate
~
lished in the literature for different residues used in 2,6
~
the paving in order to compare the characteristics s 2,5
~
of the materials and raise their technical feasibility. '0 2,4
<; 2,3
~
0. 2,2
3.1 Physical characterization of tropical soil, iron g.
u 2,1
ore and mixtures !!3
~

Figure 2 shows the results obtained in the specific "'

(/]
1,9

mass tests of the solids for the different mixtures 1,8

proposed, pure iron residue and pure granular soil. 1,7

10 11 12 13 14 15 16
It can be seen that both the pure iron ore residue
Moisture(%)
and the three different mixtures presented fairly
high values for the specific mass of the solids. Figure 3. Compaction curves for the pure iron ore (FP)
This fact is due to the large amount of iron oxide residue for different compaction energies.
present in the waste itself.
2,8
'1u3 2,7
Modified

~
Intermediate
2,6
"'
i(l
8 2,5

I Sample! ~ 2,4
'0
ps(lO"C)
I Samplel
1
0.
2,3

2,2
ps(lO"C)
g.
u 2,1
!!3
u
I Samplel
ps(lO"C) "0.
(/] 1,9

I Arithmetic mean 1,8

ps(lO"C) 1,7
10 11 12 13 14 15 16
Pure Soil 15%Residue 25%Residue 50% Residue Residue
(g/cm') 85% Soil 75%Soil 50% Soil Soil(g/cm' ) Moisture (%)
(g/cm') (g/cm') (g/cm')
Figure 4. Compaction curves for the F50S50 blend for
Figure 2. Specific mass of solids (20ºC). different compaction energies.

23
%
2,8 -
Modified In relation to the pure iron ore (FP) residue,
2,7
~
Intermediate compacted in the intermediate energy (Table 3),
2,6
according to DNIT (2006), it can be used as sub-
~ 2,5
grade reinforcement material and pavement sub-
<:- 2,4
'0
grade (Table 4). As for the sample compacted in
1g. 2,3

2,2
the modified energy, it can be used as sub-base
"' 2,1
material, reinforcement of subgrade and subgrade
J:l of pavements. All samples of FP, SP and mixtures
"[ F50S50, F25S75 and F15S85 showed low expansion
VJ 1,9

1,8 (Tables 3 and 4) and met the requirements for appli-

1,7 cation in the floor layers (except base) as recom-
10 11 12 13 14 15 16
mended by DNIT (2006).
Moisture(%)
Table 5 shows the applicability of the different
Figure 5. Compaction curves for the F25S75 blend for mixtures evaluated and the granular soil in paving
different compaction energies. as recommended by DNIT (2006).

2,8 Table 2. CBR (Intermediate Energy).

1 ~7

2,6
- Modifid
Intermediate
Nonnal Intermediate Energy
~
.~~"'1
2,5
-r
<:-
'0 2,4 Expansion
= 2,3 Sample wot (%) ρs (g/cm3) CBR (%)
~
"'g. 2,2
!* m~ FP 10,49 2,60 19,38 0,076
2,1

t
VJ 1,9
F50S50
F25S75
9,11
9,07
2,40
2,32
23,46
30,53
0,045
0,065
1,8 F15S85 9,29 2,17 23,72 0,039
1,7 SP 9,59 2,08 16,50 0,12
W 11 U U M 15 ~

Moisture (%)

Table 3. CBR (Modify Energy).

Figure 6. Compaction curves for the F15S85 blend for
different compaction energies. Modify Energy

Expansion
Sample wot (%) ρs (g/cm3) CBR (%)

FP 10,46 2,75 37,17 0,050

F50S50 8,48 2,52 29,91 0,069
F25S75 8,70 2,33 57,68 0,036
F15S85 8,07 2,25 52,26 0,09
SP 8,86 2,11 25,02 0,12

Table 4. Requirements for pavement layers according to

DNIT (2006).

Layer Expansion (%) CBR

Subgrade ≤ 2.0 CBR ≥ 2.0

Figure 7. Compaction curves for pure tropical soil (SP) Reinforcement ≤ 2.0 (over 10 CBR ≥ 2.0 (Greater
for different compaction energies. of subgrade pounds 10 than or equal
libras) to that of the
1.65 g/cm3, 1.67 g/cm3 and 1.68 g/cm3, respectively, subgrade)
for cement contents of 3.0%, 6.0% and 9.0%. The Sub-base ≤ 1.0 (over 10 CBR ≥ 20.0
results analyzed were very satisfactory, since mate- pounds)
rials with high dry densities tend to present good Base ≤ 0.5 (over 10 CBR ≥ 60.0 para
pounds) N ≤ 5 ⋅ 106
behavior when subjected to high strength and low
CBR ≥ 80.0 para
compressibility requirements for paving materials. N > 5 ⋅ 106
Tables 2 and 3 provide the results of the CBR tests
for the two energies evaluated. N is the traffic number.

24
Table 5. Application of the mixtures and granular soil Thus, with respect to CBR and expansion
evaluated for paving according to DNIT (2006). characteristics, the different mixtures of granular
soil and iron ore residue ratios evaluated can be
Samples Intermediate Energy Modify Energy used at most in sub-base layers according to the
F50S50 Sub-base, subgrade Sub-base, subgrade norms 138/2010 – ES (DNIT, 2010a), 139/2010
reinforcement and reinforcement – ES (DNIT, 2010b) and 141/2010 - ES (DNIT,
subgrade and subgrade 2010c).
F25S75 Sub-base, subgrade Sub-base, subgrade
reinforcement and reinforcement
subgrade and subgrade ACKNOWLEDGEMENTS
F15S85 Sub-base, subgrade Sub-base, subgrade
reinforcement and reinforcement The authors express their thanks to the National
subgrade and subgrade Council for Scientific and Technological Devel-
SP Subgrade Sub-base, subgrade opment (CNPq), the Support Foundation for
reinforcement and reinforcement
Research in the Federal District (FAP DF), the
subgrade and subgrade
Institutional Programs of Servers Qualification
Grants (PIQS/IFG) and Institutional Incentive
Program Participation in Scientific and Techno-
logical Events for Servers of the IFG (PIPECT)
4 CONCLUSIONS of the Federal Institute of Education, Science and
Technology of Goiás.
With regard to the activities proposed and devel-
oped for this article, it is concluded that:
− As the iron ore residue was added to the granu- REFERENCES
lar soil increasingly in the different proportions
Associação Brasileira de Normas Técnicas—ABNT.
and compaction energies mentioned above, (1986). NBR 7.182/1986: Soil: compaction test (In
there was a density gain; Portuguese). [S.l.].
− The residue of iron ore due to having a very fine Associação Brasileira de Normas Técnicas—ABNT.
granulometry tends to fill the voids of the pure (1987). NBR 9.895/1987: Soil: California bearing ratio
soil as the mixtures are added periodically. Thus, (In Portuguese). [S.l.].
mixtures with more iron ore residues will be those Castro, S. F. (2008). Incorporation of kaolin waste into
that present higher solids specific mass results. cement soil for civil constructions (In Portuguese).
− According to the Unified Soil Classification Dissertação (Mestrado em Engenharia Civil)—
System (USCS), the iron ore residue can be clas- Universidade Federal de Campina Grande. 112 pgs.
Departamento Nacional de Infraestrutura de
sified as silty clay and granular soil as silty sand, Transportes—DNIT. (2006). Paving Manual (In
whereas in the Highway Research Board (HRB) Portuguese). 3. ed. Rio de Janeiro: [s.n.].
system, (Clay soil), non-plastic and the granular Departamento Nacional de Infraestrutura de
soil as A-6 (clay soil). Therefore, according to Transportes—DNIT. (2010a). Norma DNIT 138/2010
the current Brazilian standard, i.e., the National – ES: Paving—Reinforcement of Subgrade—Service
Department of Infrastructure and Transport Specification (In Portuguese). ed. Rio de Janeiro.
(DNIT), they are not recommended for use in 7 págs.
layers of pavement subgrade; Departamento Nacional de Infraestrutura de Trans-
− In relation to the compaction tests carried out on portes—DNIT. (2010b). Norma DNIT 139/2010 –
ES: Paving—Sub-base stabilized granulometrically (In
all samples with the three types of energy (nor- Portuguese). ed. Rio de Janeiro. 8 págs.
mal, intermediate and modified), there was vari- Departamento Nacional de Infraestrutura de
ation in the optimum moisture content of 8.07% Transportes—DNIT. (2010c). Norma DNIT 141/2010
up to 11.53% and in the apparent dry density – ES: Paving—Granulometrically stabilized base—
of 2.02 to 2.75 g/cm3, thus achieving the goal of Service specification (In Portuguese). ed. Rio de
lowering optimum moisture as the dry density Janeiro. 9 págs.
increased and the compaction energy increased; Nóbrega, R. D.; Melo, R. A. (2012). Use of aggregates
− According to the California Bearing Ratio of construction waste in urban paving in João Pessoa
(CBR) tests, both for the different mixtures of (In Portuguese). In: Encontro Nacional sobre
Aproveitamento de Resíduos na Construção, 2009,
granular soil and iron ore residue, pure granular Feira de Santana. Anais... Feira de Santana.
soil and pure iron residue, considering the three Pereira, K. L. A. (2012). Stabilization of a soil with cement
compaction energies involved, the CBR values and sludge ash for use in pavements (In Portuguese).
were between 16.50% and 57.68% and those of Dissertação (Mestrado em Engenharia Civil)—
expansion between 0.039% and 0.12%. Universidade Federal do Rio Grande do Norte. 101 pgs.

25
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Suitability assessment of soils for pavement subgrade using gyratory

compaction and bearing capacity testing

A. Athanasopoulou & G. Kollaros

Department of Civil Engineering, Democritus University of Thrace, Xanthi, Greece

ABSTRACT: The quality of subgrade soils influences the design of road structures. The maximum dry
unit weight is used to assess the quality of compacted soil materials. The standard or modified Proctor
tests do not exactly simulate the field compaction mechanism. Standard Proctor compaction curves are
presented, along with curves obtained using the gyratory compactor. The compaction curves are gener-
ally similar, regardless the rate of gyration and gyration angle. For moistures higher than the optimum,
gyratory specimens yielded densities higher than those from the Proctor compaction. Based on this obser-
vation, the gyratory compaction could be considered a feasible means of laboratory compaction. The
difference between the two compaction modes was greater at low moistures. California Bearing Ratio
specimens prepared with dynamic or gyratory compaction had lower values with an increase in moisture
contents and greater in the case of dynamic compaction. Most results confirm theoretical indications or
findings of previous researches.

1 INTRODUCTION In 2006, Browne using a procedure based on the

AASHTO compaction method for hot mix asphalt
Soil encountered in the construction of trans- (AASHTO 2015) has found that the number of
portation infrastructure is densified by different gyrations and the confinement pressure were
mechanical compaction processes. In order to rep- important parameters for the compaction by the
resent more accurately the field compaction con- SGC of soils with different characteristics and
ditions, engineers are seeking for new laboratory moisture contents (Browne 2006).
methods (Vinay et al. 2015, Mokwa 2008). These Generally, the dry unit weights of fine-grained
new testing procedures will replace the commonly soils increase with the increase of the confinement
used Standard and Modified Proctor tests for the pressure (Dantas et al. 2016). For non-cohesive,
determination of the Optimal Moisture Content granular soils an increase in the number of gyra-
(OMC) and Maximum Dry Density (MDD), since tions would bring about an increase in compaction
they do not necessarily represent field compac- dry unit weights (Wite et al. 2007).
tion pressures (static, vibratory or kneading) and An increased number of gyrations yielded
motions (Drnevich et al. 2007). high densification of pavement base or subbase
Though the intend for the development of the unbound materials (Panko et al. 2011). Similar
device called Superpave Gyratory Compactor findings were obtained for subbase materials of an
(SGC) in the decade of 1990 was to test hot asphalt airfield runway (Carry et al. 2014).
mixes, many researchers suggested that it could be In an evaluation of the use of gyratory compac-
used for the compaction of soils in the laboratory tor instead of Proctor compaction for three soils
(Ping et al. 2003). classified as sand SM, silt ML and clay CH accord-
The action of moving wheels transferred on ing to Unified Classification System, the resulting
flexible pavement structures is simulated by the curves were independent of the speed and angle of
simultaneous application of vertical loads and gyration (Perez et al. 2013). However, the gyratory
kneading action (Li et al. 2015). On the other compactor can control different variables leading
hand, specimens prepared with SGC have inter- to the standard compaction curve (Kambel 2013,
nal structure resembling that of soils in highway Kollaros & Athanasopoulou 2016).
projects. The favourable use of gyratory compactors can
When the Superpave Gyratory Compactor is be summarized in that they are more precise and
being used to compact granular soils, shear work, with higher effectiveness than impact hammers,
due to the gyration angle, is applied to the soil thus along with their capability to repeat the testing
complementing the vertical work exerted on the conditions (Harman et al. 2002, Cerni & Camilli
soil sample. 2011, Chen et al. 2015).

27
 An artificial neural network has been used for cal stress via platens to a soil mass inside a cylindri-
modelling compaction parameters from index cal mould 150 mm in diameter. The gyratory speed
properties of soil (Jayan & Sankar 2015). For its was set at 30 rounds per minute (rpm). Whilst the
verification, a set of more than 180 data has been machine kept the platens horizontal and parallel to
employed and OMC and MDD were predicted each other, the mould was gyrated along its longi-
with high accuracy. tudinal axis at a fixed angle of 1.15o (20 milliradi-
Compaction curves using the Standard Proctor ans) relative to the vertical axis.
compaction test as well as those obtained from the The CBR tests on the samples determined the
gyratory compactor in Democritus University of bearing capacity of these materials to be used as
Thrace, Xanthi, Greece, are presented in the fol- subgrades in highway construction works.
lowing sections. In Figure 1, curves representing the grain size
California Bearing Ratio (CBR) specimens were distribution of the soils tested are given. All soil
prepared both with Standard Proctor and gyra- samples tested have been characterized as soils in
tory compaction and the results of their testing the group A-2-4 according to the AASHTO soil
were compared for the two methods. In the gyra- classification system.
tory compactor procedure, the vertical pressure,
the gyration angle, and the number of gyrations
were the controlled variables. 3 RESULTS AND DISCUSSION

In Table 1, the results of the different tests per-

2 MATERIALS AND METHODS formed in the laboratory are listed for the four
soils under investigation. It must be noted that
Four sampling sites in the Municipality of Xanthi, all soils were sampled at the same period in the
Northern Greece, have been chosen in order to year. So, their natural moisture value is indicative
take soil quantities for laboratory testing. The sites of the season of the year (spring). Due to the dis-
are scattered within an area of a radius of about tance between the sampling sites moistures ranged
5 kilometres from the city centre, so as to be rep- between 2.76% and 15.22% for samples S4 and S2
resentative of different prevailing conditions. Soil respectively (the closest and most far located to
samples were assigned the names S1 to S4. Kosynthos River).
On the air dried and pulverized soil samples var- In Table 1 the values of OMC and MDD along
ious tests have been performed including the Atter- with those of CRB tests are also shown. The OMC
berg limits determination, as well as tests for the values ranged from 10.3% (S1) to 16.0% (S2), while
relationship between density and moisture (stand- MDD values were found in the range of 2040 kg/
ard Proctor). For the obtained optimum moisture m3 (S2) to 2280 (S1). The values of CBR have been
content and for three water contents under and obtained using the optimum moisture content
over this value and for the corresponding maxi- found through the Proctor procedure.
mum dry densities, cylindrical specimens have been The highest value for the Linear Shrinkage (LS)
formed using both the dynamic hammer compac- measured has been found for the soil sample S4 as
tion method and the gyratory compaction process. 7.14% along with a high Liquid Limit (LL) (17.9)
With the method of gyratory compaction the which is consistent with the nature of the soil.
assessment of the compact ability of specimens On the other hand, the lowest LS value was
was achieved by the application of a 500 kPa verti- obtained for the S1 sample (1.41%) and the low-
est LL for the S3 sample (14.0). The S2 sample
100
...
~
80 -·-
·-·-·
.&-.1..- 6
Sl
S2
SJ
S4
v
4
v"
...---' :7
Table 1. Properties of the soil specimens subjected to
laboratory tests.
OJ) V"' [,'
-~ 60 S1 S2 S3 S4
~ / / lJ
0..
/ f7 Natural Moisture 7.94 15.22 11.58 2.76
I
0..
40

20
/ /
v
Liquid Limit
Plastic Limit
19.5
14.8
27.0
24.5
14.0
12.9
26.0
17.9
v Plasticity Index 4.7 2.5 2.1 8.1
i':: t:::::: Linear Shrinkage 1.41 6.43 1.43 7.14
0.100 1.000 \ 0.000 100.000 AASHTO Classification A-2-4 A-2-4 A-2-4 A-2-4
Grain Size (rum) Optimum Moisture Content 10.3 16.0 14.3 13.5
Maximum Dry Density 2280 2040 2180 2160
Figure 1. Grain size distribution of the four soils CBR 21.4 4.54 1.39 6.17
tested.

28
presented the maximum and the S3 sample the As the moisture at which the specimen has been
minimum plastic limit (24.5 and 12.9 respectively). compacted increases, the shear is reduced. The
The curves shown in Figure 2 refer to the relation higher shear (127 KN/m2) was observed in the S1
of moisture content and the dry density values for specimen was compacted gyroscopically at 10%
the four soils taking part in the laboratory testing moisture. Finally, the density of the sample speci-
program. mens increases at the end of the rotation cycles.
The water content which corresponds to the More specifically, the higher density (2115 kg/m3)
maximum dry density is called optimum moisture has been observed in the S3 soil specimen com-
content. The part of the soil materials retained on pacted gyroscopically at 13% moisture.
the 4,75 mm sieve (No. 4) has been compacted in The highest CBR value under the gyratory com-
moulds 101.6 mm in diameter in 3 layers. paction conditions was 8.82 for the S3 specimen,
Soil samples have been compacted both dynam- compacted using 7.41% moisture content. The
ically and with the gyratory compactor. Dynamic lower CBR value (0.54) has been recorded by the
compaction is a well known technique for the S2 specimen compacted gyroscopically at 17.48%
improvement of soils since it densifies them using moisture content.
a drop weight. It is obvious that, when the dynamic compac-
In order to get comparable results with the CBR tion was used, the result obtained was sensibly
method, a limitation had been posed to the height better compared to those yielded by the gyratory
of gyratory specimens (117.6 mm); in such a way compactor. More specifically, at the low moistures,
the exactly same height and weight was deter- the difference is greater between the two compac-
mined for the specimens of the two methods. The tion modes. Also, the dynamic method leads to
height of the specimens was reduced initially with higher dry density values for all specimens tested
a higher rate after each rotation cycle. For vari- in all three different moistures used.
ous water contents, the values of the dry density Figures 3 to 6 depict the relationship developed
and CBR have been experimentally defined using between the moisture in the specimens and the
both the dynamic and the gyroscopic compaction recorded CBR values for soils S1 to S4, respec-
efforts. The S1 specimen compacted dynamically tively. The water content is also correlated to the
at 7.84% water content has yielded the maximum dry density of the specimens. In each of these
dry laboratory density (2035 kg/m3), while when graphs, two different pairs of curves are shown
it was compacted dynamically at 18.12% water referring to the way the specimen has been com-
content yielded the lowest dry laboratory density pacted (dynamically or with the use of gyratory
(1688.2 kg/m3). compactor), and to the quantity under considera-
In the case of gyroscopically compacted speci- tion (CBR or dry density).
mens, the maximum and minimum densities were The highest CBR value has been recorded in the
furnished by the S3 and S2 specimens, when they case of the specimen S3 compacted dynamically
were compacted using 13.15% and 7.11% water at 7.9% moisture content and is equal to 55.94.
contents, respectively. The dry density values On the other end of the range of values, the S1
in these cases were 2115 kg/m3 and 1501 kg/m3,
respectively.
50 2600
•0 •0 •0 Proctor
Gyratory
2400
CBR
40 Dry Density 2400

abb
~

~ 2200
M
E
eo
30 2200 c
~ 0
6 ~ ·v;
>.
·';;; 2000
S4 u c
20 2000 Q "
0"
" c
c Q
0 1800 10 1800

1600 0 1600
10 15 20 25 30 4 8 12 16 20
Moisture Content (%) Moisture Content (%)

Figure 2. Moisture content vs. dry density of soil speci- Figure 3. CBR and dry density as a function of mois-
mens using the Proctor procedure. ture content for S1 sample.

29
 20 2400 specimen yielded a CBR value equal to 0.16, when
•••
(< (< 0
Proctor
Gyratory it was compacted dynamically at water content of
16
CBR
2200
18.12%.

f .,
~

1:
On
~
12
/
}' 2000 c 4 CONCLUSIONS
~
/ 0
·v;
u t: Studies on the behaviour of compacted soils
y
1800 0 "' are of great important since those materials are
... c
0 part of the structure of many civil engineering
4 1600 projects. the quality of this behaviour can be con-
trolled by the laboratory density of the soils.
The dry unit weight of soils can be a practical
1400
tool for the analysis and further comparison of
4 8 12 16 20
gyratory compaction outcome to traditional com-
Moisture Content(%)
paction test results.
Figure 4. CBR and dry density as a function of mois- For the four soils tested in this study, the dry
ture content for S2 sample. densities of the specimens compacted with the
gyratory process yielded almost linear curves
and, for moistures higher than the optimum, have
100 - - 2400 shown values higher than those derived from the
• • • Gyratory
(< (< (<
Proctor
Proctor compaction. Based on this observation,
--- CBR the gyratory compaction could be considered a
80 --- Dry Density 2200
~ feasible means of laboratory compaction.
/0
:§ In any case, more evidence based on differ-
.---..._ Y Oil
60 2000 c ent soil types, and other variables involved in the
if
/ ' -~ procedure such as the number of gyrations, the
1\/ '* ' t:
confinement pressure the angle of rotation etc is
40

20
«\ ' -..._- -
•
1800 0"'

1600
c
0
needed. It is suggested the highest number of gyra-
tions to be used to allow the maximum of soil den-
sification to be achieved.

0
4
,._
r------ ~
"
Moisture Content(%)
12 16 20
1400
Bearing in mind the drawbacks of both the
Proctor and gyratory compaction methods, CBR
values have been proved to be a valuable guidance
for the selection of the compaction method to be
used. California Bearing Ratio specimens prepared
Figure 5. Variation of California Bearing Ratio as a with dynamic or gyratory compaction presented
function of moisture content for S3 sample. lower values with an increase in moisture contents
and were generally greater in the case of dynamic
compaction. The difference was greater at water
20 -~
2400
•••
(< (< 0
Proctor
Gyratory
contents lower than the optimum moisture.
The use of artificial neural networks is sug-
-
--- CBR
gested for the prediction of soil optimum moisture
16 --- Dry Density 2200
~
content and maximum dry density values. Because
c
1: such an analysis could be easily executed based on
\ On
12 /
2000 c simple soil index parameters, it is thought to be a
~~"'..._
//..... -~ useful tool for engineers.
--. ..._ 1800 0 ""' To develop a standardized protocol for the
..._ compaction of soils with the gyratory machine
1;// \ c
"' 0 in laboratory studies, continued research is neces-
4 1600 sary. The outcome of such efforts would provide

0
~~ ...._______.
1400
more thorough understanding of soils used in road
projects. In order for the findings of this work to
acquire a universal character more tests are needed
4 12 16 20
Moisture Content(%) especially on soils with less satisfactory engineer-
ing properties.
Figure 6. CBR and dry density vs water content for S4 Future studies may lead to a better under-
sample. standing of gyratory compaction mechanism by

30
comparing dry unit weights or energies of gyratory Kampel, C. 2013. Investigation of compaction char-
compaction to field compaction instead making acteristics of subbase material using the Superpave
this comparison with existing laboratory compac- gyratory compactor. MS in Civil Engineering thesis,
tion procedures. Rowan University.
Kollaros, G. & Athanasopoulou, A. 2016. Characteriza-
tion of pavement subgrade soil using gyratory com-
paction. 3rd International Balkans Conference on
REFERENCES Challenges of Civil Engineering, 3-BCCCE, Tirana,
Albania: 254–261
American Association of State and Highway Transporta- Li, C., White, D.J. & Vennapusa, P. 2015. Moisture-
tion Officials (AASHTO). 2015. AASHTO T 312-15 density-strength-energy relationships for gyratory
Standard method of test for preparing and determin- compacted geomaterials. ASTM Geotechnical Testing
ing the density of asphalt mixture specimens by means Journal, 38(4): 461–473.
of the Superpave gyratory compactor. Mokwa, R, Cuelho, E. & Browne, M. 2008. Laboratory
Browne, M.J. 2006. Feasibility of using a gyratory com- testing of soil using the Superpave gyratory compac-
pactor to determine compaction characteristics of tor. Transportation Research Board 87th Annual
soil. M.S. Thesis, Montana State University—Boze- Meeting Proceedings, Committee AFS10, Transpor-
man, MT. tation Earthworks, Advanced Technologies for Accel-
Cary, C.E., Kumpel, C., Bagriacik, A., Cohen, R., erated Earthwork Construction and Performance
Clark, J. & Sukumaran, B. 2014. Assessment of field Monitoring. Washington, DC.
compaction of subbase material during construction Panko, M., Stevenson, J., Hurt, C., Coffey, S., McGarvey,
and trafficking of heavy aircraft using the Superpave K., Mehta, Y.A. & Sukumaran, B. 2011. Compaction
Gyratory Compactor. FAA Worldwide Airport Tech- of granular soils using Superpave gyratory compactor
nology Transfer Conference, Galloway, New Jersey, at higher confining pressures. Transportation Research
USA. Board 90th Annual Meeting, Washington, D.C., Paper
Cerni, G. & Camilli, S. 2011. Comparative analysis Number 11–3772.
of gyratory and Proctor compaction processes of Perez, N., Garnica, P., Mendoza, I. & Reyes, M.A. 2013.
unbound granular materials. Road Materials and Behavior of fine-grained soils compacted with high
Pavement Design, 12(2): 397–421. shear stresses. Proceedings of the 18th International
Chen, J., Huang, B., Chen, F. & Shu, X. 2012. Applica- Conference on Soil Mechanics and Geotechnical Engi-
tion of discrete element method to Superpave gyra- neering, Paris: 407–410.
tory compaction. Road Materials and Pavement Ping, W.V., Leonard, M. & Yang, Z. 2003 Evaluation
Design, 13(3): 480–500. of laboratory compaction techniques for simulating
Dantas, G.H.S., Furlan, A.P., Fabbri, G.T.P. & Suárez, field soil compaction (Phase I). Research Report No.:
D.A.A. 2016. On gyratory compaction of a clayey FL/DOT/RMC/BB-890(F) Project No.: 6120-549-39
soil. EJGE, 21(17): 5725–5733. Department of Civil & Environmental Engineering
Drnevich, V., Evans, A. & Prochaska, A. 2007. A study Florida A&M University Tallahassee, FL 32310.
of effective soil compaction control of granular soils. Vinay A. & Hemanth Yadav, M.V. 2015. Study and com-
Final Report FHWA/IN/JTRP-2007/12. parison of soil compaction between laboratory and
Harman, T., Bukowski, J.R., Moutier, F., Huber, G. & field to simulate field compaction for rural roads.
McGennis, R. 2002. The history and future challenges International Research Journal of Engineering and
of gyratory compaction 1939 to 2001. Proceedings Technology (IRJET), 2(4): 2086–2092.
Annual Meeting of the Transportation Research Board, White, D.J., Vennapusa, P.K.R. & Thompson, M.J. 2007.
1789(22). Field validation of intelligent compaction monitor-
Jayan, J. & Sankar, N. 2015. Prediction of compaction ing technology for unbound materials. Partnership for
parameters of soils using artificial neural networks. Geotechnical Advancement (PGA), Center for Trans-
Asian Journal of Engineering and Technology Special portation Research and Education (CTRE) Iowa State
issue for ICETTAS’15, 3(4): 368–375. University.

31
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Experimental and statistical investigation of the mechanical properties

of limestone rocks in Lebanon

M. Mrad, G. Saad, G.R. Chehab & R. Mrad

American University of Beirut, Beirut, Lebanon

ABSTRACT: Limestone is an abundant source of rock in countries worldwide. It is widely used in the
construction industry to produce various sizes of aggregates and dimension stone for building applica-
tions. Given the extensive utilization of aggregates, knowledge of its fundamental mechanical properties
is essential for determining the properties and performance of products and projects in which it is incor-
porated. This study presents the results of a laboratory-based experimental program performed on lime-
stone rock cores collected from three different quarries in Lebanon, namely Aarsal, Kelya, and Tamnine.
A total of 80 cores were tested to determine the uniaxial compressive strength and elastic modulus (E).
Also studied is the variability in the mechanical properties of limestone in Lebanon as a function of the
source of the origin rock. The analysis extends to incorporate relevant data from literature. The studied
variability is modeled by fitting a probabilistic distribution to the experimental data obtained using the
Chi Square Goodness of Fit test. Finally, a linear regression model that relates the elastic modulus of
limestone to the uniaxial compressive strength is developed using all the data gathered.

1 INTRODUCTION Concrete and asphalt concrete in addition to

unbound base and sub base material in flexible
Aggregate is a raw material used abundantly in pavements.
the construction of buildings, transportation To incorporate limestone aggregates in large
infrastructure, among other projects of the built scale engineering projects, it is essential to study
environment. In pavement construction, aggre- their two major mechanical properties; i.e., uniaxial
gates constitute approximately 90 percent and 80 compressive strength and elastic modulus. Knowl-
percent of asphalt and Portland Cement Concrete edge of the mechanical properties of aggregates is
pavement mixes (Tepordei, 1997), respectively; essential for the design and analysis of buildings
therefore, their consumption is the largest among and pavements structures given their serious impli-
other construction materials (Sullivan, 2006; cations on the overall performance and service life
Brown et al., 2007). In particular, the production of these structures, let alone, the cost associated
of limestone aggregates accounts for 50% to 70% with their maintenance and rehabilitation.
of the total production of construction aggregates To illustrate, the role of aggregates mechanical
(Oates, 2008; Aquino et al., 2010). properties is investigated in the context of pavement
Limestone is a sedimentary rock formation engineering. Figure 1 shows that in pavements, the
extensively abundant in several countries includ-
ing Lebanon, United States of America, Japan,
Turkey among others (Bliss et al., 2008). The wide
distribution of limestone deposits throughout the
world and easy access to the material are two of the
main advantages that led to its numerous applica-
tions. Limestone has been used as a key component
in the production of cement (Hooton & Thomas,
2002), raw material in the production of lime for soil
treatment, as a filler in industrial products such as
rubber, plastic, paint and others (Oates, 2008) and
as dimension stones for both interior and exterior
buildings application (Turgut et al., 2008). How-
ever, the two main applications of limestone are
in the construction field where it is used as coarse
aggregate for the production of Portland Cement Figure 1. Factors affecting the design of pavements.

33
design of the layers is based on several input vari- each were obtained from Tamnine, and Aarsal;
ables such as the traffic loading, environment, and whereas twenty prisms were obtained from Kelya.
material properties which are stochastic in nature
(Dalla Valle & Thom, 2015). Therefore, the use of
2.2 Specimen preparation
deterministic values for these variables increases the
level of risk associated with the design. This has led Two cylindrical samples of diameter 50 mm and
highway authorities around the world to move from length 125 mm were cored and cut from each
the deterministic design approach towards a proba- prism. The dimensions of the cores meet ASTM
bilistic design (Dalla Valle & Thom, 2015). D7012 specifications, where the length to diam-
A host of researchers have attempted to model eter ratio should range between 2:1 and 2.5:1 with
the stochastic nature of aggregates mechanical a minimum core diameter of 47 mm. A total of
properties by measuring the elastic modulus of dif- 80 limestone cores were obtained for unconfined
ferent type of rocks. The reported range found in Uniaxial Compressive Strength (UCS) and Elastic
the literature falls within the range of 20 to 200 GPa modulus (E) testing.
(Chunsheng, et al., 2014). In particular, the elastic The periphery section of the drilled prisms
modulus of limestone ranges between 16.548 GPa was then crushed into coarse and fine aggregates
and 66.882 GPa (Al-Shayea, 2004; Palchik & of different sizes ranging between 75 μm (#200
Hatzor, 2002). Other studies in South eastern Mex- sieve) and 19 mm using a laboratory crusher.
ico, Turkey, China, and Canada reported averages Figures 3(a) through (c) show the preparation
of 8.34 GPa, 13.6 GPa, 17.77, and 31.5 GPa respec- process of the cores. The absorption and Saturated
tively for limestone (May-Crespo, et al., 2012; Surface Dry density (SSD) of the aggregates were
Turgut, et al., 2008; Zhang, et al., 2009; Lam, et al., then obtained (Table 1).
2007). The differences in the values of the elastic
modulus can be mainly attributed to the variations
2.3 Testing
in the lithological composition, in addition to the
physical and chemical properties of the rock beds. A closed loop servo hydraulic testing machine
Thus, the need for an accurate probabilistic char- with a capacity of 150 tons was used to test the
acterization of the material properties rises. limestone cores. The cores were tested under a
The main objective of this study is to propose monotonic uniaxial compressive load at a displace-
a probabilistic approach to determine the elas- ment rate of 1 mm/min until failure. Four vertical
tic modulus of limestone. Moreover, the study Linear Variable Displacement Transducers
presents two regression models that can be used to (LVDT) placed at 90 degrees from each other were
predict the elastic modulus of limestone from the used to measure the axial displacement of the
measured uniaxial compressive strength. cores as shown in Figure 3(d).

2 EXPERIMENTAL PROGRAM

2.1 Materials
For the case of Lebanon, a study conducted by
Darwish et al. (2011) shows that the highest number
of quarries is located in the Bekaa region which is
situated between Mount Lebanon to the west and
Anti-Lebanon Mountains to the east. The stra-
tegic location of the Bekaa between two parallel
mountain ranges made it a host to approximately
44% of the total number of quarries in Lebanon.
With this fact, this study sampled specimens from
rock beds from three quarries located in three dif-
ferent regions in the Bekaa. The three quarries are
located in Aarsal, Tamnine, and Kelya covering the
Eastern, Middle and Western Bekaa areas respec-
tively, as shown in Figure 2. This will help in identi-
fying the effect of the lithological properties, if any,
on the elastic properties of limestone aggregates.
Prismatic specimens with the following dimensions Figure 2. Map showing the locations of the three quar-
were obtained: 15 cm × 15 cm × 20 cm. Ten prisms ries in the Bekaa area, Lebanon.

34
Figure 3. (a) Prisms obtained from the quarries, (b) Drilling cores from the prisms, (c) capping of the cores for testing
and (d) testing setup for the elastic modulus and uniaxial strength measurement.

Table 1. Absorption and SSD of coarse and fine 80

crushed aggregates. r/:l
;:I 70
Aggregate Size Range Absorption (%) SSD "5

$
'"d ,-., 60
0 ell
Coarse (19 mm – 9.5 mm) 1.4 2.63 :::suO
p... 50
Fine (4.5 mm – 75 μm) 2 2.65 ...... '--'
......
r/:l
ell 40
~
3 RESULTS AND ANALYSIS 30
20
Tests for determining the uniaxial compressive
strength and elastic modulus of 80 limestone cores Arsal Kelya Temnin
were conducted and its data analysed.
Figure 4. Box and Whisker plot for the elastic modulus
of limestone cores.
3.1 Effect of source
The limestone cores were grouped according to
source: Aarsal, Tamnine and Kelya. This aims at
3.2 Probabilistic characterization of the elastic
investigating the effect of the lithological properties
modulus
of each rock bed on its elastic properties. Values
of the elastic modulus for Aarsal varied between As mentioned previously the measured elastic
31.5 GPa and 67.3 GPa, between 34.1 GPa and modulus varies significantly from one sample to
72.9 GPa for Kelya, and between 28.3 GPa and the other. Such variability necessitates considering
68.7 GPa for Tamnine. Figure 4 shows the distri- the mechanical properties probabilistic rather than
bution of the elastic modulus with the minimum, deterministic, especially for accurate and realistic
first quantile, median, third quantile and maximum prediction of mechanical properties of products,
for each source. A one-way analysis of variance on materials, layers, and other elements that they are
the means of the elastic modulus at a confidence embedded in. The example of pavement design
interval of 95% yields a p-value of 0.9. This indi- as mentioned previously can be used to illustrate
cates that there is no statistical difference between the impact of wide variability in the properties of
the medians of the different sources. This could be aggregate on the outcome. Such impact is due to
attributed to the significant variability within each the effect of aggregate properties on the modulus
source. Such large variability between specimens and strength of the asphalt concrete or Portland
of the same source is primarily attributed to the cement concrete, as well as on the modulus of the
heterogeneity of each sample, given the presence aggregate base and sub base layers.
of impurities, longitudinal cracks, voids and small To account for the inherent variability of aggre-
quantities of quartz. Figure 5 illustrates the causes gate properties, a probabilistic distribution of such
of heterogeneity that are suggested to lead to such properties, such as the elastic modulus, is necessary.
wide variability in the elastic modulus of limestone. In this study, a distribution represented by the mean

35
Figure 5. Sources of heterogeneity: (a) presence of impurities, (b and c) variable mineralogical composition of lime-
stone, (d) longitudinal cracks and (e) presence of minerals such as quartz.

versus perpendicular) and the degree of satura-

20 tion (dry versus saturated). The model presented in
Figure 7 shows a positive correlation between the
5'
c:
15 elastic modulus and compressive strength plotted in
<Ll
& 10 log-log scale. A regression analysis of the data was
~ conducted to establish a relation between the two
"" 5 mechanical properties. The power formulation was
found to best fit the data as presented in Equation 1:
0
28 34 40 45 51 56 62 67 More
σ 1.05 (1)
Elastic Modulus (GPa)
where E and σ represent the elastic modulus in GPa
Figure 6. Histogram and fitted normal distribution for
the elastic modulus of limestone sources used in this study. and compressive strength in MPa, respectively.
The fitted model can be used to estimate the
elastic modulus of limestone from the compres-
and standard deviation is selected. Figure 6 shows sive strength for a compressive strength ranging
a histogram of the experimental data fitted against between a lower boundary of 5 MPa and upper
a bounded Normal distribution with a mean of boundary of 200 MPa. The model exhibits a high
48.2 GPa, standard deviation 8.8 GPa and lower adjusted coefficient of determination (R2) of 0.89
and upper bounds of 28 GPa and 71 GPa, respec- indicating that it is able to explain 89% of the vari-
tively. A Chi—square Goodness of fit test was used ability in the data. The statistical parameters of the
to measure the suitability of the normal distribution fitted model are shown in Table 2.
to the experimental data. The test yielded a p-value The lower and upper bounds of the confidence
of 0.9 indicating that the normal distribution with interval as shown in Figure 7 present the bounds
a mean of 48.2 GPa and standard deviation of for the mean prediction of the elastic modulus by
8.8 GPa is a good representation of the data. the proposed model at a confidence level of 95%.
However, the bounds of the prediction interval
provide an estimate for which individual future
3.3 Relationship between the elastic modulus and
observations will fall. In fact, the mean predic-
compressive strength
tion of the elastic modulus is typically the average
The elastic modulus data from the experimental of several individual predictions. Therefore, the
study and literature were plotted against corre- uncertainty associated with a mean prediction is
sponding compressive strength values, as presented lower than that of the individual prediction. This
in Figure 7. The plot shows a scatter in the values is illustrated in Figure 7 where the bounds of the
of the elastic modulus which explains the factors confidence interval are narrower than those of
affecting this parameter as previously mentioned. the prediction interval. The bounds of the predic-
In addition, the data takes into consideration other tion interval show that 95% of the individual data
factors such as the direction of loading (parallel points fall within this range.

36
 - - P roposed Model
- - - Confidence Interval
100 - - Prediction Interval E= 0.346 cr1.os
+ Palchik and Hatzor, 2002
)I( Lebanon
-;;-
p., )I( Turgut et al., 2008
s X Singapore
-a" )I( LianYing et al., 2009 X
~
01)
..8
• AI-Shayea, 2004
<> United States Bureau of Reclamation, 1953
.s X Kessler and Sligh, 1927
1!10
=
• Bell, 198 1
6 Najibi et al., 201 5
X
"0
0
a
·~
"'
@

10 100
Compressive strength in log scale (MPa)

Figure 7. Proposed model for the elastic modulus of limestone versus compressive strength.

Table 2. Summary of the statistical parameters.

limestone from regions within Lebanon is consid-
Intercept Estimate of Residual ered. The relationship between the elastic modu-
standard the slope standard lus of limestone and its compressive strength
error standard error error for the case of Lebanon can be best described
using an exponential formulation as presented in
0.07 0.02 0.46 Equation 2:

--Proposed Model
E e3.3 + 0.0007σ (2)
85
---- Confidence Interval E= e 3.3+0.007cr
where E and σ represent the elastic modulus in
- - Prediction Interval
~75 GPa and compressive strength in MPa, respec-
0... tively. The fitted model fits the data most closely
0
'-' 65 for a compressive strength of limestone ranging
"'
:l from 45 MPa to 95 MPa. The model will still be
"3
"0 55 able to predict the elastic modulus for a compres-
0
:::E sive strength outside the previously mentioned
._g 45 range; however, with less reliability. The adjusted
"'Ol coefficient of determination (R2) is 0.245. This
1i:l 35 indicates that the model is able to explain approx-
imately 25% of the total variability.
25
40 50 60 70 80 90 100
Compressive Strength (MPa) 4 CONCLUSIONS AND
RECOMMENDATIONS
Figure 8. Proposed model between the elastic modulus
of limestone and compressive strength for Lebanon only. The elastic modulus of limestone aggregate is
important in the design of buildings and pave-
3.3.1 Case of Lebanon ments since aggregates constitute a major part of
Given that the experimental data obtained in this these composite models. This study presents an
study is extensive, a localized model for Lebanon approach to develop a probabilistic distribution
is presented in Figure 8. This localized model nar- for the elastic modulus of limestone along with the
rows down the variability of the predictions when associated uncertainty. The study shows that the

37
elastic modulus of limestone in Lebanon follows Bliss, J.D., Hayes, T.S. & Orris, G.J., 2008. Limestone-A
a normal distribution with a mean of 48.2 GPa Crucial and Versatile Industrial Mineral Commodity.
and standard deviation 8.8 GPa. The large value United States Geological Survey: science for a chang-
of the standard deviation proves further the need ing world.
Brown, T.J. et al., 2007. Mineral Planning Factsheet:
for a probabilistic distribution. The presented dis- Construction Aggregates.
tribution will help in decreasing the level of risk Chunsheng, Z., Kefei, L. & Fu, M., 2014. Numerical and
in design and analysis of composite materials and statistical analysis of elastic modulus of concrete as a
structures. In addition, the study proposes two three-phase heterogeneous composite. Computers and
relations between the elastic modulus of limestone Structures, Volume 139, pp. 33–42.
and its compressive strength. The first relation is Dalla Valle, P. & Thom, N., 2015. Variability in pave-
generic and described by a power formulation: ment design. The international Journal of Pavement
E = 0.346 σ 1.05 . This relation utilizes 341 data Engineering and Asphalt Technology (PEAT), 16(2),
points collected from several studies and therefore pp. 50–67.
Darwish, T. et al., 2011. Environmental impact of quar-
can be used by engineers and researchers from ries on natural resources in Lebanon. Land Degrada-
around the world to estimate the elastic modulus tion & Development, 22(3), pp. 345–358.
of limestone. The second relation is specific to the Hooton, R.D. & Thomas, M.D., 2002. The use of lime-
case of Lebanon and described by the exponential stone in Portland cements: effect on thaumasite form
formula: E e3.3+ 0.007 σ using experimental data of sulfate attack, s.l.: Portland Cement Association.
obtained from laboratory testing. Lam, T., Martin, D. & McCreath, D., 2007. Characteris-
It is important to note that the need for a rela- ing the geomechanics properties of the sedimentary
tion to estimate the elastic modulus of limestone rocks for the DGR excavations. Ottawa, Canadian
and specifically for Lebanon is timely because it geotechnical conference.
May-Crespo, J. et al., 2012. Physical, Petrographic, and
is expected that significant amount of aggregates Mineralogical Properties of Limestone Rocks from
will be quarried in Lebanon for the cement pro- the Peninsula of Yucatán. MRS Proceedings, Volume
duction during the post-war reconstruction in the 1373.
region. Oates, J.A., 2008. Lime and limestone: chemistry and tech-
nology, production and uses. John Wiley & Sons.
Palchik, V. & Hatzor, Y., 2002. Crack damage stress as
ACKNOWLEDGMENT a composite function of porosity and elastic matrix
stiffness in dolomites and limestones. Engineering
Geology, 63(3), pp. 233–245.
The research team is grateful for the support of the Standard, A.S.T.M., 2010. D7012–10 (2010) Standard
University Research Board (URB) at the Ameri- test method for compressive strength and elastic mod-
can University of Beirut (AUB). The authors uli of intact rock core specimens under varying states
acknowledge the technical support provided by of stress and temperatures. In: Annual Book of ASTM
the manager and technicians of the Structural Standards, American Society for Testing and Materi-
and Materials Lab at the American University of als. West Conshohocken: s.n., pp. 495–498.
Beirut. Sullivan, D.E., 2006. US Geological Survey Fact Sheet
2006–3127. s.l.:s.n.
Tepordei, V.V., 1997. Natural Aggregates—Foundation
of America’s Future, s.l.: U.S. Dept. of the Interior,
REFERENCES U.S. Geological Survey.
Turgut, P., Yesilnacar, M.I. & Bulut, H., 2008. Physico-
Al-Shayea, N.A., 2004. Effects of testing methods and thermal and mechanical properties of Sanliurfa lime-
conditions on the elastic properties of limestone rock. stone, Turkey. Bull Eng Geol Environ, pp. 485–490.
Engineering Geology, 74(1), pp. 139–156. Zhang, L., Mao, X. & Lu, A., 2009. Experimental study
Aquino, C. et al., 2010. The effects of limestone aggre- on the mechanical properties of rocks at high tem-
gate on concrete properties. Construction and Building perature. Science in China Series E: Technological Sci-
Materials, 24(12), pp. 2363–2368. ences, 52(3), pp. 641–646.

38
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Application of pattern classification techniques for anisotropic

characterization of pavement foundations

R.S. Ashtiani & M. Asadi

The University of Texas at El Paso, El Paso, TX, USA

ABSTRACT: Several factors such as moisture state, particle geometry, gradation parameters, fines con-
tent, and the nature of stress paths contribute to the directional dependency of material properties in the
granular layers. The motivation for this study was to evaluate the applicability of pattern classification
techniques to provide class discriminatory information of the laboratory observations. For this purpose,
several experimental permutations were subjected to variable dynamic confining pressure stress path tests
to study the synergistic influence of different factors on the anisotropic behavior of aggregate systems.
The laboratory tests and the post processed data were in turn used to evaluate the relevance of pattern
classification techniques to unravel physically meaningful information of the multi-dimensional dataset.
The results of this effort will be instrumental for the practitioners to potentially reduce the number of
features needed to be determined in the laboratory for a refined and cost-effective testing protocol.

1 INTRODUCTION in granular materials theoretically using tensor

analyses (Oda & Nakayama 1989). Adu-Osei et al.
Unbound granular layers are integral component (2001) studied the impact of aggregate characteris-
of the highway pavements and airfield runways. tics on the level of anisotropy and demonstrated its
The main role of such layers is to distribute and dependency on particle size distribution, particle
reduce the traffic induced stress to a tolerable shape, and moisture content at a certain level of
level for sub grade soils. Additionally they serve compaction.
as robust platform for the surface to maintain Several predictive models have been proposed
adequate ride quality. Traditional pavement design in the literature to estimate the anisotropic behav-
approaches characterize the Unbound Aggregate ior of the UAB (e.g. Kim et al. 2005; Masad et al.
Base (UAB) layers as an isotropic medium. How- 2006; Ashtiani et al. 2008). Ashtiani et al. (2008)
ever, it is acknowledged by many researchers that proposed a simple procedure to evaluate the level
isotropic modeling of UAB causes the mechanical of anisotropy of aggregate systems using some
response of the pavement to be evaluated unreal- easily determinable aggregate properties. Subse-
istically (Karasahin et al. 1993; Masad et al. 2006; quently, they used the level of anisotropy as an
Wang & Al-Qadi 2013). Accordingly, some efforts input to predict the performance of aggregate
have been made to develop predictive models as bases. On the other hand, a number of research-
well as testing protocols to properly capture the ers have investigated the effect of loading on
anisotropic characteristics of the UAB (Adu-Osei the anisotropic responses of granular material.
et al. 2001; Kim et al. 2005; Ashtiani & Little 2009). Tutumluer & Seyhan (1999) studied the influ-
The directional dependency of material prop- ence of multiple stress path on the anisotropic
erties of the aggregate base is twofold. It is firstly behavior of granular materials. They utilized the
due to the characteristics of the material such as University of Illinois Fast Cell (UI-FC) for simu-
particle geometry and particle size distribution, lating the dynamic stresses on aggregate material
also known as inherent anisotropy; and secondly as a result of the moving traffic loads. Adu-Osei
due to the rotation of stress fields as a result of et al. (2001) tested aggregate materials at different
moving traffic loads, which is often referred to as stress states and dynamic stress paths using Rapid
stress-induced anisotropy in the literature. Early Triaxial Testing (RaTT) cell, and calculated the
attempts to investigate the impact of particle size anisotropic parameters through system identifica-
and geometry on the anisotropic response of gran- tion method scheme. He investigated the material
ular materials were made by Parkin et al. (1968), under three different stress regimes to capture the
El-Sohby (1969) and Oda (1972). Later, a number parameters required for characterization of trans-
of contributors assessed the concept of anisotropy versely isotropic media.

39
 In the area of pavement engineering, pattern (Ey) can be represented by Equation 1. Also, the
classification techniques have been used mostly for horizontal resilient modulus (Ex) and the shear
automatic detection and classification of surface modulus (Gxy) can be determined by equations 2
distresses in place of the human labor (e.g. Saar & and 3, respectively.
Talvik 2010; Wu et al. 2014). In current research, k5 k6
these approaches were employed to study cross ⎛ I ⎞ ⎛τ ⎞
anisotropic behavior of UAB. For this purpose, Ex k4 Pa ⎜ ⎟ ⎜ oct + 1⎟ (2)
⎝ Pa ⎠ ⎝ Pa ⎠
initially a comprehensive experiment design was
k8 k9
developed to establish a multi-dimensional aggre- ⎛ I ⎞ ⎛τ ⎞
gate feature database. The database consists of fea- G xy k7 Pa ⎜ ⎟ ⎜ oct + 1⎟ (3)
tures of aggregates mixes with different lithologies ⎝ a⎠ ⎝ a
P P ⎠
and gradations which are tested at different satura-
tion levels. Nonlinear and cross-anisotropic mate- Several researchers have investigated the fac-
rial parameters were determined using Variable tors influencing the resilient behavior of aggregate
Dynamic Confining Pressure (VDCP) stress paths. systems subjected to moving traffic loads. Among
Aggregate Imaging System (AIMS) was employed these factors, moisture state, dry density, aggregate
to characterize the particle geometry parameters. geometry and stress path have been identified as
Subsequently, Hierarchical Clustering Analysis the factors that have the most considerable impact
(HCA) techniques based on different measures on the orthogonal load distribution capacity of
of distance were employed to investigate the simi- unbound aggregate systems. It is widely accepted
larities between the distributions of the param- that the resilient modulus significantly deteriorates
eters. Furthermore, Fisher’s Linear Discriminant as the saturation level increases (Dawson et al.
Analysis (LDA) and K Nearest Neighbor (KNN) 1996). Dry density (or degree of compaction) of
algorithm were used to classify the data based on aggregate materials is another factor which plays
predefined criteria. The results of this study iden- a crucial role in the resilient behavior. Generally,
tify the aggregate features with similar patterns the aggregate matrix becomes stronger and stiffer
and thus provide a deeper understanding of the as the dry density increases. It is worth mention-
underlying relations and interactions between the ing that the influence of dry density variations on
features of aggregate database. the resilient modulus is not the same for systems
with different lithologies, fine contents, and stress
states (Hicks 1970). Aggregate shape is considered
2 ANISOTROPIC CHARACTERIZATION as another factor which significantly affects the
OF UAB resilient behavior of granular materials. Roughly-
textured and angular aggregates form a stiffer
The resilient modulus (MR) of UAB is nonlin- mass through development of strong interlock-
ear, stress-dependent and anisotropic. The new ing forces. Previous studies have confirmed that
Mechanistic Empirical Design Guide (MEPDG) crushed aggregate material with high angularity
considers the following equation to capture the and rough texture shows higher resilient modulus
nonlinearity as well as hardening-softening behav- compared with the rounded and smooth counter-
ior (Uzan 1999): part (Ashtiani et al. 2008).
The moving nature of traffic loads adds another
k2
⎛ I ⎞ ⎛τ ⎞
k3 component to the stiffness anisotropy of aggregate
MR k1Pa ⎜ ⎟ ⎜ oct + 1⎟ (1) layer in the pavement structure. Figure 1 sche-
⎝ Pa ⎠ ⎝ Pa ⎠ matically illustrates the stresses felt by the mate-
rial below the surface due to moving wheel load
where I = first stress invariant; τoct is octahedral over the pavement. As is shown in this figure,
shear stress; Pa = atmospheric pressure; and k1, k2, the vertical stress (σy) reaches its maximum right
k3 are fitting parameters. In this equation, the term beneath the wheel load and reduces nonlinearly
(I/Pa)k2 is known as the hardening component, as the wheel moves away from the reference point.
which captures the stiffening effect of aggregate Similar manner is followed by the horizontal stress
matrix when subjected to repeated loading. On the (σx) but with a lower magnitude. Furthermore, the
other hand, (τoct/Pa)k3 is referred to as the soften- slope of the stress plot in the case of vertical stress
ing component, which characterizes the reduction is much higher than the horizontal stress. Conse-
of stiffness after exceeding the threshold load as quently, at any given point an extension-compres-
a result of the cumulative damage induced to the sion-extension stress regimen is experienced as
matrix. the wheel load passes over the pavement section.
In order to fully characterize the UAB as an Although the magnitude of the tensile stresses may
anisotropic medium, the vertical resilient modulus be very small, the rotation of the principal stress

40
 were tested in three moisture conditions: Optimum
Moisture Content (OMC), dry of optimum mois-
ture content (-2% of OMC), and wet of optimum
Extension Comp.r esMon Extension
moisture content (+2% of OMC). Specimens were
fabricated following (ASTM D1557) to determine
the optimum water content and maximum dry
density of each gradation variant. Table 1 provides
the labels selected to identify materials and features
of the aggregate database.
Traditional triaxial test protocols are com-
monly suitable for simulating the state of the
stresses under stationary loads. On the other hand,
dynamic stress path tests can be used to simulate
the stress states developed in aggregate layers by
the traffic loads. Such a proper test protocol is able
Figure 1. Distribution of normal and shear stresses at a
given point under a moving wheel load.
to capture the extension-compression-extension
stress regimens induced by moving wheel loads.
Particularly, VDCP type stress path tests take
plane will result in additional level of anisotropy advantage of applying dynamic vertical pressure as
induced on aggregate structure. Hornych et al. well as dynamic confinement on the specimen for
(2000) pointed out that the aforementioned stress a more realistic simulation of field conditions. In
regimen can increase the rate of plastic strains up this study, a customized triaxial setup called RaTT
to three times greater than those measured from (Rapid Triaxial Test) cell was used to perform
repeated plate load tests. VDCP stress path tests in accordance with the
Due to the role of stress states in directional stress path protocol outlined in ICAR/508 report.
dependency of material properties, anisotropic (Ashtiani & Little 2009)
characterization of granular materials using con- RaTT cell device is controlled by a multi channel
ventional tests such as plate load test or tradi- data acquisition system which is capable of apply-
tional triaxial setup is irrelevant. Therefore, some ing both vertical and dynamic confining pressures.
researchers have proposed new testing protocols The general setup for the stress path test using this
to simulate the rotation of principal stresses in the apparatus is consisted of RaTT cell mounted in the
laboratory. Universal Testing Machine (UTM), data acquisi-
Adu-Osei et al. (2001) developed a laboratory tion system, and a computer for control and stor-
testing protocol for anisotropic characterization of age of the data. This customized setup supports
unbound granular material. He applied ten static automated cell movement and multiple displace-
stress states followed by small dynamic excursions ment measurement with Linear Variable Differen-
in the stresses to obtain three stress regimes namely tial Transducers (LVDT) in both directions.
triaxial compression, triaxial shear, and triaxial The testing protocol provides a means to deter-
extension. Consequently, he determined the cross- mine cross anisotropic material properties including
anisotropic parameters of aggregate material via a elastic modulus in the horizontal and vertical direc-
back-calculation scheme called System Identifica- tions (Ex and Ey), Poisson’s ratio in the horizontal
tion (SID) method. A similar testing protocol is direction due to vertical loading (vxy), Poisson’s ratio
adopted in current study for anisotropic charac- in the horizontal direction due to horizontal load-
terization of the UAB. ing (vxx) and shear modulus (Gxy). For this purpose,
applied stresses and measured strains are used as
input parameters to an iterative error minimization
3 EXPERIMENT DESIGN
Table 1. Classification criteria and class labels of the
Various aggregate systems with distinct lithologies aggregate database.
were evaluated to study the factors that influence
the cross anisotropic behavior of unbound aggre- Classification Class
gate systems. Three gradation variants (i.e. coarse, Criterion Labels
well and fine gradations) of different granite, lime-
stone and gravel aggregates obtained from ten Lithology Limestone Granite Gravel
(L) (N) (G)
sources in Texas, Oklahoma and Minnesota, were
Gradation Coarse (1) Well (2) Fine (3)
incorporated into the experiment design. In order
Moisture Dry (D) Optimum Wet (W)
to investigate the effect of moisture state on the state (O)
mechanical response of the aggregates, the materials

41
technique (i.e. the SID) to simultaneously solve lead to a high dimensional aggregate database. In
four of the five anisotropic material properties (Ex, principle, the more information we gather about a
Ey, vxx, and vxy). The fifth material property (Gxy) phenomenon, the more accurate insight into that
is determined using elastic work potential rela- phenomenon we can achieve. However, analysis
tionships derived specifically for the shear stress of a database containing large number of data
regimen. Subsequently, the five cross anisotropic deems necessary to employ dimensionality reduc-
material properties as well as prescribed stresses in tion techniques to eliminate correlated features. In
the lab can be fitted to the material models, Equa- other words, the dimensionality reduction is syn-
tions 1 through 3, to determine the nine k param- onymous to the selection of the features with high-
eters for each aggregate system. The calculated k est significance. This is more pronounced when the
values will be used as input to characterize nonlin- development of prediction models becomes the
earity, stress dependency, and anisotropic behavior objective of the research.
of unbound aggregate systems in the lab. Dimensionality reduction is essentially a process
It is intuitively relevant that aggregate geom- of feature selection and thus can be regarded as a
etry characterized by its shape, angularity, and clustering problem. Clustering algorithms are cat-
texture greatly influence the level of anisotropy egorized into two general types namely supervised
unbound aggregate systems. This is mainly due and unsupervised approaches. The term supervised
to the contribution of the geometry to the inter- implies that the clustering is performed based on
locking mechanism in particulate systems. In this predefined training patterns. On the other hand,
study, aggregate geometry was characterized in unsupervised algorithms cluster the data only
terms of particle form, angularity, and texture according to the inherent scatter of the data. The
using the AIMS. Aggregate form refers to flat classification techniques employed in current
or elongated shape of the particles; angularity research include: Hierarchical Clustering Analysis
defines the degree of roundness or sharpness (HCA), K Nearest Neighbor (KNN), K-means
of aggregate corners; and texture is related to clustering and Fisher’s Linear Discriminant Anal-
small asperities at the surface of particles from ysis (LDA). These techniques have been success-
which the surface roughness originates (Kim et fully applied in several engineering practice such
al. 2005). Fifty-six aggregate particles from three as: geotechnical site characterization (Hegazy &
aggregate sizes of each source were evaluated Mayne 2002), detection of crack in pavements (Wu
using the AIMS device. Several two parameter et al. 2014), prediction of landslides due to rain-
and three parameter distribution functions were falls (de Souza & Ebecken 2012) and travel time
fitted to the aggregate geometry data. Based prediction on high-ways (Tak et al. 2014), to name
on the best fit analysis of the database, it was a few.
observed that the two-parameter Weibull distri- K-means is a clustering technique aiming to
bution provides a reasonable fit for both parti- generate exactly K different groups of data with
cle size distributions and geometrical properties greatest possible distinction (Duda et al. 2000).
of the aggregates data at a 95% confidence level. K-means algorithms require the number of clusters
The cumulative Weibull distribution function is (K), an initial assignment of data to clusters and
of the following form: the distance measure between data points. On the
other hand, HCA is another clustering approach
β
⎡ ⎛ d ⎞⎤ which only relies on a measure of similarity (or dis-
Q(d ) = 1 − ⎢ − ⎜ ⎟ ⎥ (4) tance) between the data points. The Euclidian dis-
⎣ ⎝α ⎠⎦ tance (also known as average distance) between all
pairs of objects in two clusters (r and s) is defined
where, d = the aggregate size; α = scale param- in Equation 5 as:
eter; and β = shape parameter. These distribution
parameters, which can be determined through fit- nr ns
1
ting the particle size distribution and aggregate d ( r, s ) =
nr ns
∑ ∑ dist( x ri , xsj ) (5)
shape properties data to the Weibull distribution i =1 j =1
function, were later imported to the aggregate fea-
ture database. where, nr and ns are the number of samples in the
clusters r and s, respectively. On the other hand,
the Mahalanobis distance is defined as:
4 PATTERN CLASSIFICATION
dist( xri , xsj ) ( xri xsj )D 1 ( xri − xsj )T (6)
Anisotropic behavior of unbound aggregate sys-
tems is influenced by several aggregate features. In this equation, the distance between data
Hence, any comprehensive experimental study will points is normalized by the covariance matrix D.

42
Consequently, this measure is less sensitive to the function applied to the multi-dimensional aggre-
outliers of the system, and typically results in bet- gate database.
ter classification rates. Classification based on the aggregates’ lithol-
The objective of LDA is, in essence, to project ogy resulted in 73% classification rate. Based on
the data onto the direction which gives the best the statistical classification analysis of the data,
classification scheme. This can be achieved if angularity parameter, dry density, and k6 param-
within class scatter (SW) is minimized and, simul- eter, i.e. exponent of the hardening term for hori-
taneously, between class scatter (SB) is maximized. zontal modulus model, were found to be the most
Mathematically, the best direction vector (w) can influential features of the database. Classification
be obtained by maximizing the following objective based on gradation resulted in 62% classification
function (Duda et al. 2000): rate. SFS algorithm selected angularity param-
eter, k3 and k9, i.e. exponents of softening terms
wT SB w for vertical and shear modulus, respectively, as the
J( ) = (7) features that best explain the system’s behavior.
wT SW w
Finally, classification based on the moisture state
resulted in 52% classification rate and the features
KNN is another widely-used classification
selected by SFS were k3, i.e. softening exponent of
method that determines the class label for unla-
the vertical modulus, and k7, i.e. multiplier in the
beled data points based on their similarity. In this
shear modulus.
method, the class of the data is labeled based on
The best classification rate of 73% was observed
the majority vote of its neighbors. Hence, in binary
when classification was performed according to
classification problems the K would be chosen as
the lithology of the aggregates in the experiment
an odd number to avoid tied votes. For most prac-
matrix. The aggregates elected in the experiment
tical purposes, K is assumed as a positive integer
design had significantly different angularity char-
typically smaller than 10.
acteristics. This was confirmed by unenforced
In current study, discriminatory information
selection of this feature in the clustering analy-
among the features of the aggregate database was
sis. Natural selection of the angularity feature by
investigated through two approaches. In the first
mathematical algorithm is in conformity with the
approach, the classification was performed by
intuitive sense of the role of particle geometry
means of HCA based on two different distance
on the anisotropic behavior of aggregate systems.
measures of the data, namely Euclidean distance
Since load transfer in particulate media is carried
and Mahalanobis distance. After choosing the
out through particle interactions and interlock-
proximity measure, K-means algorithm was used
ing effect between solid grains, aggregate systems
to find the matching samples in each cluster. The
with more angular particles are expected to be
significance of this analysis is that it identifies
less prone to develop plastic deformations when
aggregate features with similar patterns subjected
subjected to external loads. To further investigate
to stimuli such as stress path tests. This analysis can
the role of the aggregate angularity, the shape and
potentially unravel the underlying relations and the
scale parameters of the aggregate angularity were
interaction between the features of the aggregate
removed from the database and the classification
system. In the second approach, the laboratory
algorithm was executed without these parameters.
measurements and post processed aggregate data
The removal of the angularity features resulted
were classified using Fisher’s LDA. In addition,
in significant drop in the classification rates from
KNN algorithm was also applied to improve the
73% to 46%.
performance of the classifier based on pre-defined
class labels. This approach helps to analyze the sep-
arability of the features in order to develop effec- 5.2 Hierarchical clustering analysis
tive testing protocols and/or regression analyses.
In order to find physically meaningful patterns
among the features of the aggregates, hierarchical
unsupervised clustering analyses were also per-
5 RESULTS AND DISCUSSION
formed on the database. The results of hierarchi-
cal clustering of the features based on average and
5.1 Supervised clustering analysis
Mahalanobis proximity measures are illustrated in
Supervised clustering analysis was performed Figures 2, 3, respectively.
on the aggregate data based on three classifica- The general findings from these two dendograms
tion criteria namely lithology, gradation, and are very similar to each other. As demonstrated in
moisture state. Additionally, Sequential Forward Figure 2, k1 and k4, which are the multipliers of the
Selection (SFS) algorithm was applied to select elastic modulus in vertical and horizontal direc-
the features that minimize the mean square error tions, respectively, have the closest patterns in the

43
 drop of orthogonal stiffness and hence accelerates
10

I
the rutting potential in granular layers. This reduc-
tion of stiffness properties in unbound systems is

I
synonymous with increase of the softening param-
eter in Equation 1. Therefore, the patterns of the
softening parameter and water content obtained in
this study are in agreement with the expected behav-

y
ior of unbound aggregates.
S:lUBlS!O

5.3 Discriminant analysis

Fisher’s LDA was used in order to provide class
discriminatory information among the features
of the aggregate database. Initially, the eigenval-
ues and eigenvectors of Fisher’s criterion (SW−1SB)
were determined, and then the data were projected
~ ~
L ~

onto the principal components with the highest

I

'
Dry
[
~

eigenvalues. The knee in the resulted plot (Fig. 4)

n ~la rity

Density

suggests that the first two principal components

:it

.s!
Cii
2
"'
"'~

::>
8'

of Fisher’s criterion can reasonably explain the

dynamic nature of the system. In addition, KNN
Figure 2. Hierarchical clustering of aggregates’ features
using Euclidean distance as the proximity measure. classifier with different values of K (number of
neighbors) was applied to the projected data to
evaluate the classifier’s performance.
The classification rates for both raw and pro-
12

jected data after applying KNN classifier are

shown in Figure 5. According to this figure,
Fisher’s discriminant criterion accompanied by
10

KNN algorithm is well capable of providing class

discriminatory information when the class labels
are considered based on lithology. Furthermore,
8

Figure 5 suggests that this technique is not able

to provide class discriminatory information when
y
e:>UBjS!O

particle size distributions and moisture state are

regarded as true class labels.
4

~=.---=~~~~

,---.---,---.---.---.----.---.---.--~ ~
I
2

~.~~~
=---~~~.---~
0

w
Angularit y
k1
k4

k5

k3
k6
k8
k7

k9
k2

Dry Dens ~y

e n1eAua6 !3
.&
<(

ro
.$
"'~

::>
Ol
Ol

Ol

~--~~~.---=

Figure 3. Hierarchical clustering of aggregates’ features

using Mahalanobis distance as the proximity measure.
-~~~

dataset. This is in line with our prior knowledge

~. ---::

of the behavior of aggregate systems. On the other

..j ;=
==

hand, Figures 2, 3 show similar patterns for soften-

ing parameters (k3 and k9) and the water content
'15
2
z

~
!5

8_
~

(w). It is well established in the literature that ingress

of moisture in unbound aggregate systems results in Figure 4. Eigenvalues of Fishers’ criterion (SW−1SB).

44
 Figure 6 illustrates a good classification of
aggregate data through LDA accompanied by
KNN classifier. Although gravel aggregates are
fully separated using LDA+KNN classifier, some
misclassifications is evident in limestone and G
G
granite materials. Moreover, limestone and gravel G

aggregates have more condensed projections com- c

pared to granite aggregates. This could be due

"6 G
a. G
to the method of identifying aggregate types for E G N
0
construction purposes where limestone and gravel u
have more distinct definitions compared to granite.
-1
Distribution of the observations before applica-
tion of LDA projection and KNN is shown in Fig-
ure 7. This plot when compared to Figure 6, clearly -2 ~~~---1~.5~---1~~~~.5~~~~0~.5~~--~1~.5~~~~2~.5--­
demonstrates the significance of LDA+KNN
Component 2
method in providing appropriate class discrimina-
tory information between the aggregate features. Figure 7. Classification using LDA projection based on
The results of classification analyses with regard to lithology (L: Limestone, N: Granite and G: Gravel).
other criteria and the corresponding distributions
are presented in Figures 8–11.
08

0.6
100 cRaw Data 1 2
2
41 )4
«i mLDA+KNN
0::
80 ).2
]]
3
3
2
22
E 2
c:
§"'
:; 60
0
0 3 2
2
3
2
11 "'
§ 23

=
(,) 2
).2 2
·u; 40 u 1 2
VI
~.4

0"' 20
~6

0 ~. 8

Lithology Gradation Moisture -2 5 -2 -1 5 -·1 ~5 05 15

Component 2
Figure 5. Performance of the classifier based on differ-
ent criteria.
Figure 8. Classification using LDA+KNN projection
based on gradation (1: Coarse, 2: Well and 3: Fine).

2.5

1.5
G
G
G
J.5 ... oGo G
5
i§
GG

0
l G
G
G
sc. 1
c
u J.5 i5
u 1 3

-1 22 3 1 3
1)
~ 2
1' i 13
-1.5 N N
""' N N
-1
' 2 3!

-2

-2 -1.5 -1 ~.5 0.5 1.5

-2
-2 -1.5 -1 ~.5 0.5 1.5 2.5
Component 2
Component 2
Figure 6. Classification using LDA+KNN projection
based on lithology (L: Limestone, N: Granite and G: Figure 9. Classification using LDA projection based on
Gravel). gradation (1: Coarse, 2: Well and 3: Fine).

45
 2.5 this information can be used to develop more pre-
cise data-driven models such as regression models
and artificial neural networks.
1.5 The comprehensive experiment design consisted
of three aggregate types from ten different sources
;:: with three distinct gradations molded at three dif-
:J
c 0.5
0
0.
ferent moisture states were incorporated in this
g w
0 research effort. Subsequently, the aggregate sys-
v 00 tems were subjected to multiple variable dynamic
-0.5
confining pressure stress path tests to determine
·1
parameters associated with anisotropic material
·1.5 properties. Aggregate geometry was character-
-2
ized in terms of particle form, angularity, and tex-
·1 -0.5 0.5 1.5 ture using the AIMS. The distributions of these
Component 2 parameters were, in turn, quantified through the
cumulative Weibull distribution function, and
Figure 10. Classification using LDA+KNN projection the parameters of the models were imported to the
based on moisture state (D: Dry, O: Optimum and W: aggregate feature database for post processing.
Wet). Unsupervised clustering techniques were per-
formed on the features of the aggregate system to
identify natural groupings present in the aggregate
database. Dimensionality reduction techniques
were also employed to find class discriminatory
w information between the observations. The best
w classification rate was achieved when classifica-
VfN
tion was performed based on aggregates lithology.
It was also confirmed that the angularity plays a
(\)
0 0 0 0
crucial role in the anisotropic behavior of aggre-
gate systems.
"b 0 II' w w fN.,p vP
w
D W \'W 0 D
vt:ow'W"'ww
WO 0 Additionally, Fisher’s LDA accompanied by
0
0
w 0
w 00
0
~ w KNN algorithm was used to find class discrimi-
·1 o w.£ 0 0 natory information based on pre-selected features
Oo
of the aggregate database. To accomplish this, the
-2u_--~--~--~--~--~--~--~--~--~-- Eigenvectors of Fisher’s LDA were calculated, and
·2 · 1.5 ·1 -0.5 0.5 1.5 2.5
the data was projected onto the two principal com-
Component 2 ponents with highest Eigenvalues. The KNN algo-
rithm was used to calculate class rates according to
Figure 11. Classification using LDA projection based the true class labels. The highest rate was obtained
on moisture state (D: Dry, O: Optimum and W: Wet). when the classification was carried out based on
lithology using Fisher’s LDA in combination with
6 CONCLUSIONS the KNN algorithm.

The main objective of this study was to assess the

applicability of pattern classification techniques, REFERENCES
such as KNN method and Fisher’s discriminant
analysis, to explain the underlying behavior of ani- Adu-Osei, A., Little, D. N. & Lytton R. L. 2001. Cross-
sotropic granular materials. The aforementioned anisotropic characterization of unbound granular
methods were also utilized to identify the most materials. Transportation Research Record: Journal of
significant features of the aggregate database, such the Transportation Research Board 1757: 82–91.
as aggregate geometry, moisture state, and etc., Ashtiani, R. S. & Little, D. N. 2009. Methodology for
for dimensionality reduction of the system. Addi- designing aggregate mixtures for base courses. College
tionally, such pattern recognition techniques were Station: International Center for Aggregates Research
(Report No: ICAR/508).
employed to extract any physically meaningful Ashtiani, R. S., Little, D. N. & Masad, E. 2008. Material
relationship between the features of the aggregate factors that influence anisotropic behavior of aggre-
database. The findings of this research can be use- gate bases. Transportation Research Record: Journal of
ful as a means to eliminate the number of features Transportation Research Board 2059: 20–30.
needed to be determined by laboratory tests for a de Souza, F. T. & Ebecken, N. F. F. 2012. A data based
more cost-effective testing protocol. Furthermore, model to predict landslide induced by rainfall in Rio

46
 de Janeiro city. Geotechnical and Geological Engineer- Oda, M. 1972. Deformation mechanism of sand in triaxial
ing 30: 85–94. compression tests. Soils and Foundations 12(4): 45–63.
Dawson, A. R., Thom, N. H. & Paute, J. L. 1996. Oda, M. & Nakayama, H. 1989. Yield function for
Mechanical characteristics of unbound granular mate- soil with anisotropic fabric. Journal of Engineering
rials as a function of condition. Rotterdam: Balkema. Mechanics (ASCE) 115(1): 89–104.
Duda, R., Hart, P. & Stork, D. 2000. Pattern classifica- Parkin, A., Gerrard, C. & Willoughby, D. 1968. Discus-
tion. New York: Wiley. sion: deformation of sand in shear. Journal of the
El-Sohby, M. 1969. Deformation of sands under constant Soil Mechanics and Foundations Division 94(SM 1):
stress ratios. Mexico. 336–340.
Hegazy, Y. A. & Mayne, P. W. 2002. Objective site char- Saar, T. & Talvik, O. 2010. Automatic asphalt pavement
acterization using clustering of piezocone data. Jour- crack detection and classification using Neural Net-
nal of Geotechnical and Geoenvironmental Engineering works. 12th Biennial Baltic Electronics Conference
(ASCE) 128(12): 986–996. (BEC2010) Tallinn, Estonia.
Hicks, R. G. 1970. Factors influencing the resilient proper- Tak, S., Kim, S., Jang, K. & Yeo, H. 2014. Real-time
ties of granular material. PhD Dissertation, University travel time prediction using multi-level k-nearest
of California, Berkeley. neighbor algorithm and data fusion method, In R. I.
Hornych, P., Kazai, A. & Quibel, A. 2000. Modeling a Issa & I. Flood (Eds.), Computing in civil and building
full-scale experiment of two flexible pavement struc- engineering (ASCE) Orlando, USA.
tures with unbound granular bases. UK, University of Tutumluer, E. & Seyhan, U. 1999. Laboratory determi-
Nottingham, p. 359–367. nation of anisotropic aggregate resilient moduli using
Karasahin, M., Dawson, A. R. & Holden, J. T. 1993. an innovative test device. Transportation Research
Applicability of resilient constitutive models of granular Record: Journal of the Transportation Research Board
material for unbound base layers. Washington, D.C., 1687: 13–21.
TRB, National Research Council, p. 98–107. Wang, H. & Al-Qadi, I. 2013. Importance of nonlinear
Kim, S., Little, D. N. & Masad, E. 2005. Simple Methods anisotropic modeling of granular base for predict-
to estimate inherent and stress-induced anisotropy of ing maximum viscoelastic pavement responses under
aggregate base. Transportation Research Record: Jour- moving vehicular loading. Journal of Engineering
nal of the Transportation Research Board 1913: 24–31. Mechanics (ASCE) 139 (1): 29–38.
Masad, S., Little, D. N. & Masad, E. 2006. Analysis of Wu, L., Mokhtari, S., Nazef, A., Nam, B. & Yun, H.-B.
flexible pavement response and performance using 2014. Improvement of crack-detection accuracy using
isotropic and anisotropic material properties. Jour- a novel crack defragmentation technique in image-
nal of Transportation Engineering (ASCE) 132(4): based road assessment. Journal of Computing in Civil
342–349. Engineering (ASCE) 30(1): 04014118.

47
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Plastic strains of a compacted clayey soil under repeated axial loading

Zhong-Sen Li
Laboratory of Soil, Structure and Materials Mechanics, Centrale Supélec, Châtenay-Malabry, France
School of Earth Science and Engineering, Sun Yat-Sen University, China

Jean-Marie Fleureau
Laboratory of Soil, Structure and Materials Mechanics, Centrale Supélec, 92295 Châtenay-Malabry, France

ABSTRACT: An experimental study was performed in order to investigate the production and develop-
ment of the plastic strains of a subgrade soil under repeated loads. A clayey soil, compacted in a Proctor
mold with different compaction water content and energies, was subjected to repeated punching effort
ranging from 200–600 kPa for about 40000 repetitions. Cumulative plastic strain was first plotted against
the number of load repetition, and then against the water content. The results show that, under relative
dry condition, the cumulative plastic strains are negligible and are independent of the water content; then,
the cumulative plastic strains increase more and more rapidly; at last, as water content exceeds a certain
limit, the specimen cannot support the vertical stress and some soil expelled from the Proctor mold.
In order to describe the above phenomenon concerning the development of the plastic strains, soften-
ing limit (wS), expulsion limit (wE) as well as stability/transition/expulsion zones were defined and then
integrated in the compaction curve, which is widely used in engineering practice. In addition, the plastic
strains produced in the first cycle was compared and discussed.

Keywords: Plastic strains, compacted clayey soil, cyclic loading, softening limit, expulsion limit

1 INTRODUCTION interpretations will help to understand globally

the production and development of plastic strains
Recent development and construction of flexible in unsaturated soils, and to apply the laboratory
pavements have drawn much attention on the plas- experimental data to engineering practice.
tic strains of different pavement materials such as
asphalt in the surface layer, granular material in the
base course, etc. The question of the plastic strains 2 MATERIAL AND METHOD
in subgrade layers (compacted soils), however, has
not attracted so much attention because it was The material of this study was taken from a site on
thought to be insignificant compared with that in the Tours/Bordeaux High-speed Railway Line in the
surface layers and granular base. south-west of France at a depth of 1.5–7.8 meters.
Previous works suggest that plastic strains of The soil is mainly made of clayey particles but also
subgrade soils contribute to two main pavement of some sharp gravel with diameters of several cen-
defects: rutting and cracking, especially for the thin timeters, which were eliminated manually before the
pavements in rural areas or low traffic roads (Brown laboratory tests. The main geotechnical properties
et al., 1975; Elliot et al., 1998, 1999), in which exces- of the material are presented in Table 1, and it can
sive subgrade plastic strains produce high mainte- be classified as A3 according to the French classi-
nance costs and reduce riding quality (Li & Selig, fication (AFNOR, 1992) or A-7-5 in the American
1996). Association of State Highway and Transport Offi-
This paper presents some experimental results of cials (AASHTO) classification (ASTM, 2004).
cumulative plastic strains versus number of cycles
and water content using repeated axial punching
tests to simulate the traffic loads of pavements Table 1. Geotechnical properties of the material.
in earthwork projects. The concepts of softening
limit (wS) and expulsion limit (wE) are proposed Mat. wL (%) wP (%) IP (%) Gs <80 μm (%) <2 μm (%)
and the corresponding points are plotted in the
compaction curve diagram. These results and Clay 70.5 37 33.5 2.83 88 70

49
 After arriving at laboratory, the material was (Fig. 1). In order to prevent evaporation, the speci-
dried under laboratory condition (20 ± 1°C) for sev- men was covered by a thin film and sealed with two
eral weeks, and then sieved at 2 mm. The passing O-rings.
particles were dried in an oven at 105°C for about The compacted sample, together with the Proc-
48 hours, and mixed afterwards with the required tor mold, was transferred to the loading platform,
quantity of distilled water. The mixed soil was kept and fixed by two clamps. As shown in Fig. 2, the
in a sealed plastic bag for more than 48 hours, and loading system mainly consists of a hydraulic actu-
then compacted in a standard Proctor mold in 3 ator, a control center and a data acquisition center.
layers with 15, 25 or 35 blows per layer, which cor- A metallic piston of 4 cm in diameter was attached
responds to energies of 0.36, 0.6 and 0.84 MJ/m3, to the MTS machine, where the strains and stresses
and called Reduced Proctor (RP), Standard Proc- can be controlled and measured accurately during
tor (SP) and Enhanced Proctor (EP), respectively the cyclic loading/unloading test.

1.40 +Reduced Proctor (N=l S)

o Standard Proctor (N=2 5)
)<Enhanced Proctor (N=3S
1.35
~

Me
1.30
~
"'
Q.

i 1.25
6
-=
g 1.20
1.15
22 27 32 37 42 47
WatEr contmt, w (%)

Figure 1. Compaction curves of the soil.

Figure 2. Presentation of the repeated loading system: (1: Data acquisition; 2: Loading piston; 3: Compacted sample
and Proctor mold; 4: Clamp; 5: Hydraulic actuator; 6: Control center).

50
 800
~H = 0.45Hz I
'~"' 600
.;
~ 400
~
...... 200
t
>
0 5 10 15 20 25 30 35 40
Time (s)

Figure 3. Loading program of the test.

Based on the French standard (AFNOR, 2007) increases with water content until the failure of the
and engineers practice in earthworks, the load- soil. For the specimens compacted at Reduced and
ing program was chosen and set up. As shown in Standard Proctor, there is a similar trend for the
Fig. 3, the loading frequency was 0.45 Hz and the development of the cumulative plastic strain.
applied vertical stress ranged from 200 to 600 kPa,
the maximum vertical stress corresponding to the 3.2 Cumulative plastic strain versus water content
pressure of a truck tire pressure on the pavement,
and the minimum stress corresponds to that of a The observations above indicate that, for the
private car. Prior to the application of cyclic loads, specimens compacted with a certain energy, water
the contact condition between the piston and spec- content is an important factor that determines the
imen was ensured, taking care that the metallic pis- production of cumulative plastic strains. It would
ton did not penetrate into the specimen. be therefore interesting to compare directly the
When the above procedure was finished, the development of cumulative plastic strains for the
repeated load was applied. The whole test for each specimens at different compaction water contents.
specimen lasted about 24 hours for 40000 load rep- Fig. 5 presents the cumulative plastic strains after
etitions and about 20 specimens were tested. 100, 1000, 10000 and 30000 cycles for the speci-
mens at different initial compaction water contents
with the Reduced, Standard and Enhanced Proc-
3 RESULTS tor energies. Note that the plastic strain of the 10th
cycle was taken as reference because of the irregu-
3.1 Cumulative plastic strain versus number lar applied force in the first ten load repetitions.
of load repetitions In order to describe the results in Fig. 5, the
Fig. 4 presents the results of cumulative plas- cumulative plastic strain after N cycles (εp-N) is
tic strains versus the number of load repetitions expressed as a function of the water content of the
for the specimens prepared at different compac- specimen (w):
tion energies. The diagrams on the left side are
the results in normal coordinate system, and the εp–N = f (w) (1)
diagrams on the right side are the results in semi-
logarithmic coordinate system. Then, the softening limit and expulsion limit-
Taking the results of enhanced Proctor (E35) as sare defined as follows:
an example, in the normal coordinate system, it can – Softening limit (wS): the water content above
be observed that the maximum cumulative plastic which the cumulative plastic strains increase
strains are almost constant for the water contents rapidly as the soil becomes softer and softer.
smaller than 32%, followed by an increase as the Theoretically,
soil becomes more and more moist. At w = 40%, the
specimen cannot support the applied force, some f (wS)’ = 0 and f (wS + Δw)’ > 0 (2)
soil in the specimen is expelled from the Proctor
mold, and failure takes place after about 500 load – Expulsion limit (wE): the water content above
repetitions. In the semi-logarithmic coordinate sys- which the specimen cannot support the applied
tem, there seems to be a linear relationship between stress and some of the soil is expelled from the
the cumulative plastic strain and the number of Proctor mold. Mathematically,
load repetitions. The slope of the curve is almost
constant for the specimens drier than 32%, then f (wE)’ ≈ +∞ (3)

51
 25

-=-
~ 20
c.
~

·~ ----¢---- w = 32.5%

-o-w=32.8%
fl
~
15
-o-w=37.5%
~
-~:s--w = 38.4%

~ 10
~
~
~

1
a
0

--w=28.5%
-x-w=29.6%
---+-- w = 30.4%

----¢---- w = 34.7%

-o-w=34.8%
-o-w=37.5%
- f : s - w = 39.6%

--w=25.0%
--+-w = 28.0%
-x-w=32.1%
-o-w=35.7%
---<>-w = 37.2%
--+-w = 37.2%
-o-w=38.8%

0
40 000 30 000 20 000 10 000 0 10 100 1 000 10 000

Number of load repetitions, N Number of load repetitions, lg N

Figure 4. Cumulative plastic strains versus number of load repetitions for the specimens compacted with Reduced
Proctor, Standard Proctor, and Enhanced Proctor energies.

In this research, the water content for which the limited number of experimental results; more spec-
cumulative plastic strain is larger than 15% was imens at different water contents should be tested
considered to be the expulsion limit (wE). in order to determine the limits more accurately.
The softening limit of Reduced Proctor, Stand- With the concept of softening limit and expul-
ard Proctor and Enhanced Proctor are equal to sion limit, the curves in Fig. 5 can be divided into
32.8%, 34.8%, 35.7%, respectively, and the expul- three zones: i) stability zone (w < wS), where the
sion limit for these compaction energies are equal cumulative plastic strains after 100, 1000, 10000 and
to 38.4%, 39.6%, 40.4%. Note that the accuracy of 30000 cycles are almost independent of the compac-
the softening and expulsion limits determined in tion water content; ii) transition zone (wS < w < wE),
this study needs to be improved due to the relative where the production of cumulative plastic strains

52
 -
~
~
"'
11.1
15

Reduced Proctor
0N=10000

...=10
xw= 22.5%
·;
t;; + w= 26.9%
~ Stablility Zone Transition Zone Failure Zone

< >
~
~ ow= 32.8%
-a
Q,j
N=1000
;5 aw=37.5%

=e=
~ 5
ow= 38.4% /' tY
I "
u ................ :
/'
...... /
/
N=100

0
~----------------lt-.~--=~.:~: . -.- N=10

-~
~ Standard Proctor
"'
11.1
W = 28 .5%
N=1000
.e
~ 10
X

I
!::
~
+ w = 29.6% I

:as ow=34.8%
Stability Zone Transition Zone / Failure Zone

c.
Q,j
5 aw=37.5%
< ~ >
;5 / . N=lOO
~
'3
e
ow=39.6%
/ /
/./
u=
=1
0

Enhanced Proctor
N=1000
xw=28%
I
+ w = 32.1% I
I
Stability Zone Transition Zone/ Failure Zone
ow=35 .7%

aw=37.2%
< I
;> L---->
I
o I
ow=38 .8% I

¢W=40.4% I
"' /
/ N=100
- --t./
_.~:;..-··
p. · -""
=10
25 30 35 Ws 45
Water content, w (%)

Figure 5. Development of cumulative plastic strains as a function of water content.

increases rapidly with water content; iii) failure is also in safety, this will contribute to rutting and
zone (w > wE), in which the specimen cannot sup- cracks in the structure; in the failure zone, when the
port the applied force after several load repetitions soil reaches its maximum bearing capacity, it can
and some of the soil in the specimen begins to be no longer support the traffic loads; expulsion will
expelled. take place in the subgrade and then lead to subsid-
The three zones described above are of interest ence or even rupture of the pavement surface.
in engineering practice, for example: with in the sta-
bility zone, the cumulative plastic strains are insig-
3.3 Presentation of softening and expulsion limits
nificant under the applied vertical stress and the
in the compaction plan
structure is relatively safe; in the transition zone,
with the increase of the water content due to cli- With the new concepts, it will be more practical
matic or some other reasons, the cumulative plastic for the engineers if the three zones and two critical
strains increase rapidly even though the structure water contents are presented at the same time as the

53
compaction curves. Fig. 6 presents the compaction specimen should be moister in order to reach the
curves with the softening limit (wS) and the expul- new expulsion plastic strain.
sion limit (wE). As described above, the compaction
curve for each energy level can be divided into three
zones. Based on this figure, the plastic strain behav- 3.4 Synthesis of results—plastic strains during
ior of the specimen at different water contents the first cycle
under certain stress conditions can be estimated. The determination of the plastic strains during the
Further inspection of the figure shows that the first cycle requires much precision and needs to
softening limit and expulsion limit increase with be improved in the future work, as sometimes the
the compaction energy. This is logical, considering applied stress with the load system was found to
a specimen at the expulsion limit (wE), for example: be less than that expected, especially in the first ten
i) when the compaction energy increases, plastic cycles. For this reason, only the experimental results
strain produced at each load will become smaller, for the Enhanced Proctor are presented (Fig. 7).
and therefore the specimen will move into the tran- Fig. 7a presents the variation of plastic strains
sition zone; ii) under the same vertical stress, the during the first cycle as a function of water content.

Stability zone Transition zone Failure zone

1.40
1. . ·.·.· .. · .. · .. ·.1 ~~0&%'/j I.·.·.·.·. i
• Softening limit • Expulsion limit

~
~ 1.35

--
OJ)

~ 1.30
k("ll

~=1.25
~
Q 1.20

1.1 5+-~~~~~~~~~~~~~~~~~~~~~~~~~~~

18 22 27 32 37 42 47
Water content, w {0/o)

Figure 6. Presentation of softening/expulsion limits in the compaction plan.

4 120
,-_

--
~
Q

~
100

-=
.€
~
:1-1
80

60
I 2 nd to 30 ooothload 1

--
·;
:1-1
r:l.l

.~
( "ll
~
40
1st load
20
6:
0
25 30 ws 35 40 20 30 ws 40
Water content, w elo) Water content, w ( 0/o)

(a) (b)
Figure 7. Plastic strains produced during the first cycle for the specimen compacted at the Enhanced Proctor energy:
(a) plastic strain versus water content; (b) plastic strain ratio versus water content.

54
The plastic strains (εp) in the first cycle are almost load repetitions based on the plastic strains under
constant for water contents smaller than the sof- the first load, which can be easily obtained from
tening limit, and then they increase rapidly with the experiment. This method will largely reduce the
water content. This phenomenon is consistent with time and expense to predict the plastic strains, as
the cumulative plastic strains which have been pre- only the strains during the first loading are required
sented in Fig. 6, where cumulative plastic strains once the function [η(w)] is determined.
change very slightly as a function of water content
on the dry side of the softening limit, whereas they
increase more and more rapidly on the wet side. 4 CONCLUSIONS
The observation is possibly due to the structural
properties of the compacted clayey soil: at water The article presents an experimental study of the
contents smaller than the softening limit, the soil plastic strains of a compacted clayey soil by apply-
suction (s ≈ 3 MPa at wS) is much larger than the ing a relative large number of repeated axial loads
applied vertical stress (σv ≤ 600 kPa), the clayey with a metallic piston of 4 cm in diameter. The
particles of the sample form a very strong skeleton experimental results show that the cumulative plas-
structure, where the resistance of the specimen tic strains do not change significantly for the speci-
does not change a lot with water content. There- men prepared at relatively dry condition, but after
fore, there is no significant difference as regards the a certain water content limit, the plastic strains
plastic strains on the dry side of the softening limit. increase more and more rapidly until expulsion of
However, on the wet side of the softening limit, the specimen takes place after a few load repetitions.
the skeleton structure effect will be weakened, the Two critical water contents, the softening limit
specimen resistance decreases when water content (wS) and the expulsion limit (wE), were defined in
increases (suction decreases), and that is why the the plans of the cumulative plastic strain versus
plastic strains increase afterwards. the compaction water content. The critical water
Fig. 7b indicates that the plastic strains during contents are then presented in the compaction
the 1st loading are very important, representing plan, and divide the curve into three zones: stabil-
about 20 to 70 percent of the cumulative plastic ity, transition and failure zones. The classification
strains produced in the first 30000 cycles. If the of these zones are of practical use for the engineers
plastic strains in the 1st loading are compared with to estimate the potential plastic strain behavior of
those produced by the first 30000 loads, a plastic the compacted soils under a certain applied stress,
strain ratio (η) can be defined as follows: and provide necessary information for the design
of foundations, pavement, embankment, etc.
ε p −1
η= (4) The plastic strains in the first cycles, especially
ε p − 30000 in the first cycle are very important, representing
about 20% to 70% of those produced in the first
where, εp–1 represent the cumulative plastic strains 30000 cycles. With the relationship between plas-
in the 1st loading, εp–30000 represent the cumulative tic strain ratio and water content [η(w)], it is pos-
plastic strains after 30000 loads. sible to predict the cumulative strains in the first
Fig. 7b presents also the variation of the plas- 30000 cycles from the plastic strains in the first cycle.
tic strain ratio versus the water content. It can be
observed that, in general, η decreases with water con-
tent, which means that the plastic strains produced ACKNOWLEDGEMENTS
in the first loading become less important. In other
words, the influence of the first loads is reduced and The authors would like to gratefully acknowledge
the plastic strains tend to develop under the follow- the scholarship from China Scholarship Coun-
ing load repetitions as the specimen becomes moister cil (2010691009), financial supports from French
and moister. This could be due to the creep strains, National Project “Terredurable” (ANR 2011
especially for the very moist specimens in which the VILD 004 01) and China Postdoctoral Interna-
applied stress creates relatively large plastic strains tional Exchange Program (32110–41090002).
under each load. With this diagram, a relationship
can be then established between the plastic strain
ratio and the compaction water content: NOTATIONS

η = f (w) (5) The following symbols are used in this paper:

A, b = material parameters;
Combining Eq. (4) and Eq. (5), it is possible to E = compaction energy;
derive the cumulative plastic strains after 30000 f = load frequency;

55
N = number of load repetitions; AFNOR (French Association of Standardization).
s = soil suction; (2007). “Bituminous mixtures—Test methods for hot
w = water content; mix asphalt—Part 22: wheel tracking.” NF EN 12697-
wE = expulsion limit; 22, France.
ASTM (American Society for Testing and Materials).
wS = softening limit; (2004). “Standard Practice for Classification of Soils
η = strain ratio; and Soil-Aggregate Mixtures for Highway Construc-
σv = vertical stress; tion Purposes.” D2487–00, West Conshohocken, PA.
Δw = water content increment; Brown, S. F., Lashine, A. K. F. and Hyde, A. F. L. (1975).
εp = cumulative plastic strain; “Repeated load triaxial testing of a silty clay.” Géo-
εpp-N = cumulative plastic strain after N cycles; technique, 25(1), 95–114.
ξpw −σ v = Plastic strain in a single load at water con- Elliott, R., Dennis, N. and Qiu, Y. (1998). “Permanent
tent w (%), and applied vertical stress σv (kPa). deformation of subgrade soils (Phase I: A test proto-
col).” No. MBTC FR 1069.
Elliott, R., Dennis, N. and Qiu, Y. (1999). “Permanent
deformation of subgrade soils. Fayetteville. Phase II:
REFERENCES Repeated Load Testing of Four Soils.” No. MBTC
FR-1089.
AFNOR (French Association of Standardization). Li, D. and Selig, E. T. (1996). “Cumulative plastic defor-
(1992). “Earthworks—Classification of materials for mation for fine-grained subgrade soils.” J. Geotech.
use in the construction of embankments and capping Eng.,10.1061/(ASCE)122: 1006–1013.
layers of road infrastructures.” NF P11–300, France.

56
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Estimation of shakedown loads of flexible pavements

A.G. Stathas & K.V. Spiliopoulos

National Technical University of Athens, Athens, Greece

ABSTRACT: Excessive rutting, resulting from a cumulative vertical deformation due to repeated traffic
loads, is often the main cause of failure of flexible pavements. The analytical approach, however, that is
currently used to design for the flexible pavement life is based on the critical elastic strain at the top of the
subgrade and does not consider any plastic pavement behavior. An analysis for variable repeated loading
in the context of material plasticity obviously provides a more rational design criterion. Such a criterion
may be provided by shakedown analysis which guarantees the long term stability of a structure under
repeated loads. In this work, a step by step finite element numerical procedure is used to estimate the
response of the soil half-space when subjected to loadings of different intensities. The material is assumed
to follow an associated flow rule obeying a Mohr-Coulomb yield criterion. Besides examining various
convergence issues of the numerical procedure, results show that friction angle and cohesion greatly influ-
ences the shakedown load.

1 INTRODUCTION were Sharp & Booker 1984. Based on the static

shakedown theorem, they performed a linear pro-
Civil Engineering structures like bridges, buildings, gramming application using a Mohr-Coulomb
oil platforms, pavements, etc. are often subjected material.
to a loading of a cyclic nature, e.g. earthquakes, Radovsky & Murashina 1996 presented an ana-
waves, etc. A typical cyclic loading in pavements lytical shakedown analysis of a plain strain pave-
are traffic loads. In many cases these loads may ment model.
lead the structure to strain excursions well inside Using the kinematic theorem, Collins and Boul-
the plastic regime. However, should the load inten- bibane 2000, Boulbibane et al. 2005 proposed an
sity be within certain limits, a structure may adapt upper bound to shakedown of single and multi-
itself so that phenomena like low cycle fatigue or layered pavements.
ratcheting are ruled out. These limits are known as Yu 2005 proposed an analytical solution for
shakedown limits. shakedown of cohesive-frictional half-space under
In the application of pavement engineering, the moving Hertz load using Melan’s static shakedown
shakedown limit is recognized as an asymptotic theorem. Wang 2011 used elasto-plastic analyses
structural state in which the plastic deformation to obtain the shakedown load of a Mohr-Cou-
ceases to increase after a number of load cycles. lomb half space with associated or non associated
Otherwise, if the plastic deformation accumulates flow rule and also proposed an optimized analyti-
at each repetition of a load cycle, excessive rutting cal shakedown solution taking into account the
will eventually occur. Although the problem may equilibrium considerations of Melan’s theorem.
be approached by cumbersome step-by-step proce- Wang & Yu 2013 extended the analytical method
dures, usually shakedown analyses, which belong to to three-dimensional multilayered pavements.
a certain class of numerical methods called Direct Very recently Liu et al. 2016 investigated the
Methods, utilize either Melan’s statical or Koiter’s influence of a non-associated flow rule on the
kinematical shakedown theorem. shakedown limits by varying the material dilation
In 1962 Johnson used the line rolling contact angle using a step-by step procedure.
assumption to examine the shakedown condition The present study gives some insight of the
of an isotropic, elastic-perfectly plastic Tresca half step-by-step numerical investigation of the shake-
space while Bhargava et al. 1985 obtained accurate down limit for materials obeying an associated
predictions for the downward displacement of John- Mohr-Coulomb yield criterion. As a general over-
son’s Tresca half space during multiple load passes. view of the results, it is shown that friction angle
The first to consider that the concept of shake- and cohesion greatly influences the shakedown
down could be used in the design of pavements load.

57
2 PROBLEM DEFINITION 2. After each pass the load is removed so that the
residual stresses and plastic strains can be meas-
In order to investigate the load under which the ured along the path at the middle of the area of
pavement shakes down, a cylinder of infinite length interest.
is repeatedly rolled over the pavement half space. 3. After each pass the results of the plastic strains
The pressure distribution that acts on the surface and residual stresses are compared with those of
of the half space, in the contact region between the previous passes to determine whether shake-
cylinder and pavement, is considered elliptical, down takes place or not. Shakedown takes place
following a Hertz load distribution. The pressure during a load pass when both plastic strain and
function (p) is given as: residual stress diagrams of the present pass are
equal to those of the previous pass. Moreover
p0 1 − ( x a ) ( −a ≤ x ≤ a) the stress-strain diagrams at the most strained
2
p (1)
point of the path are extracted in order to
observe the behavior during the loading cycles.
where a: is half of the contact length. 4. Different values of friction angle (φ°) and dila-
p0: the maximum vertical stress. tation angle (ψ°) as well as different load magni-
The maximum vertical stress p0 depends on P tudes are used to determine the shakedown limit
and a, which are the vertical load of the wheel and of the pavement.
the half width of the contact surface respectively.
Johnson 1985states that: The numerical approach described above
demands great computational effort in order to
4RP obtain reasonably accurate results. The problem
a= (2) has been solved using a 4-core i7 CPU with16GB
π E∗
RAM.
where R: the relative radius of the half space and
the wheel.
E*: The modulus combining the stifness of 3.2 Model description
bodies in contact. A pavement model is established on ABAQUS.
Therefore: In order to create a procedure which can describe
the state of the pavement after multiple cycles of
2P PE ∗ loading, the option of restart analysis in Abaqus
p0 = = (3)
πa πR is being used. In this manner the results after each
load pass are used to describe the state of the pave-
By changing P, R different values of contact area ment on which the next load pass will take place.
and maximum stress can be found. For the analyses Each analysis consists of two steps. In the first
undertaken P, R are selected under the condition step an elliptical load distribution is applied at the
that the contact width 2a remains constant for dif- start point, then translates horizontally at constant
ferent values of maximum vertical stress p0. speed until it reaches the end. In the second step
the load is completely removed. The loading proc-
ess is controlled by a user subroutine DLOAD.
3 NUMERICAL APPROACH The simulation is processed by means of auto-
matic incrementation control while the option of
In this section a numerical step-by-step approach time extrapolation is set to ‘none’. According to
for the pavement shakedown problem is presented the Abaqus Analysis User’s Guide 2012 the above
and validated. Results including shakedown limits, option disables the characteristic of the automatic
residual stresses and plastic strains are discussed. incrementation algorithm which is responsible for
the increment size the program will choose during
3.1 Method description an analysis taking into account the previous incre-
In the present study, Finite Elements (FE) elastic- ment. This generally reduces the computational
plastic analyses are carried out to obtain the actual time needed for this analysis. The start increment
residual stresses developed in pavement structures for this analysis is 0.01 and the maximum 0.1.
under repeated moving traffic loads using the According to Abaqus Analysis User’s Guide the
Abaqus 2012 software. The following procedure is stiffness matrix of the material following an asso-
being carried out: ciated plastic flow rule is automatically selected by
the solver.
1. An elliptic load moves repeatedly from a start- A sketch of a one-layered pavement used for
ing point A to a finishing point B on the pave- this study is shown in Figure 1. The vertical barri-
ment surface. ers are restrained in the horizontal direction while

58
 I'
~

_2m.~

Figure 1. Sketch of a one-layered pavement with the 2-D moving Hertz load distribution. In the middle lies the area
of interest.

Table 1. Properties of soil materials used in analyses. Table 2. Geometrical characteristics of the analyzed
models.
E v C φ ψ
L L′ H
MPa kPa ° °
m m m
Material1 100 0.3 50 20 20
Material2 100 0.3 50 30 30 Model1 86 3 30
Model2 60 10 30

the bottom barrier is restrained at the vertical The mesh gets coarser as the vertical distance from
direction. For the influence of the vertical barri- the surface grows.
ers to be minimized, no load is applied near them. Shakedown limits obtained by the current
Eight noded, reduced integration, quadratic ele- approach are also compared with shakedown
ments (CPE8R) are selected to avoid hour glassing solutions of other researchers (Wang 2011, Liu
and interlocking problems. The soil material used et al. 2016) derived from an analogous numerical
for the analysis is homogenous, isotropic, linear procedure.
elastic in the elastic region and perfectly-plastic
using the Mohr-Coulomb criterion for plastic-
ity. The parameters necessary to define the mate- 4 ANALYSIS RESULTS
rial are E, v for the linear elastic part and c, φ°,
ψ° (cohesion c, friction angle φ°, dilatation angle For this analysis the results are collected along a
ψ°) for the Mohr Coulomb criterion. Table 1 con- path vertical to the free surface of the pavement,
tains all the properties of the soil materials needed in the middle of the model inside the region of
for the presented analyses. In the following results interest. The region of interest is the region of
compression is negative. Abaqus uses a smooth the model where the effect of the vertical barriers
approximation of the Mohr-Coulomb criterion proves to be negligible. Inside this region each ele-
for plastic flow which is close to the classical ment experiences the same load history and abides
Mohr-Coulomb, especially when the mean pres- to Johnson’s assumption (Johnson 1985) that the
sure is high. plastic strain and the residual stress fields are inde-
Following the research of Liu et al. 2016, two pendent of the travel direction. Furthermore from
models were used in order to save computational equilibrium considerations σ xz and σ zz must be
time. Table 2 contains the geometrical details for zero. The results of the analysis are presented in
both models used. High mesh density is applied Table 3. Two analyses are presented for materials
in the first two meters near the surface where the with different friction angles φ° = 20°, φ° = 30° and
stress, strain gradients are expected to be high. an associated flow rule (φ° = ψ°).

59
Table 3. Analyses results. Strain 'X

~~
p
~ ~ ~ 1": ~ § ~
p0 z

Material1 Material2 N kPa M

-- -- -- -- -- r ----- r~~ tr- - - - - --- --- - -
Model1 x 16 7.5c 0.5 ................f...... lj ~'}· ................
Model2 x 28 10.6c 0.4
~ : ............ f J ·j· ............. .
2 "' ~~~-+~--r-~~~-+~~T-+-~-+~~

Strain
in
.:.,
i - - - - - - --·· t· ·- --~
~ !No :
-----~--- ------ --- --- --
:

--; ;-- -- ~

"'"~T=~;d:::=t=:-11
~~
. ~_.....-:-·--=-~
------------------------------------ ~ --
=--------~::.--------- --
~·
~ ~ ~~~-7~--~
~ -+~~~,
~~~~~~+-~-+~~
+t +·· · · · ·· ·
~~~-r-+_,_-_·-~r-_-+-----~1~----r----Tr_-~-r----~----+---~-~----r---;-·-~
0 ------------------------------------ --------i--------- ~
.. ---------:----- r -----r ----r---------------- --
lCD : ···································· +r~ --- --- -- -~--- -. . :. - -- --- ; _-- -·- ... _.__ -- - ·- -- --- ·-- - -- --

-- -------------------------------------r -----------
T
------------------------------------ ------ ---·--------- --

--------- --------- --- -- ---- --------- --------J----- ---- --

Figure 3. Distribution of the plastic normal vertical
strain ε zzp for φ° = 20° and load p0 = 7.5c.
0~----,_----~-----+----~------r-----~

Stress

-----:----------- ~ -~~4
~ ~~~~r-~~~~~---+---7---+--~--~

: ~,;: + +
Figure 2. Distributions of the plastic shear strain ε xz
p

for φ  = 20  and load p0 7 5c.

-i
.::J.
~ :] ----- ~--- -- ------ :- ---- -----

Pavement with φ° = 20° and ψ° = 20° :ll "'

~ " ·+·1\ + +··· ·
4.1
In the first case, Model1 is being used for analysis,
and a material of φ° = 20° and ψ° = 20° is cho- ~ 0 "'L : :
sen. The shakedown load occurs at p0 = 75c after ~~ --------------------- -----r~:; --t ----- ------~ -----
sixteen load passes (N = 16). Figures 2–5 present S: in :• A p-·---+-----
'
-----~----- ·----- ~ ----
. :
the plastic strain ε xzp , ε zzp and the residual stress -----·------ ----- -----
: : :
( xx , σ yy
r
) distributions along the path in the mid- o r-~--~---+--~--~--+---+-~r-~---1

dle of the area of interest. It is observable that

after sixteen load passes, both the ε xzp , ε zzp diagrams
v• -----:------ ----- ----- --··-r··-- ---··-r··-- ---··r··--
are stabilized. It is evident that the plastic shear ---- _. . ______ ----- ----- ----- ! --- -- ------ ~----
strain stabilizes after the fourth pass while the plas-
tic normal strain stabilizes after the twelfth pass.
As it may be seen the residual stresses also stabilize
after the twelfth pass.
To further reassure the shakedown occurrence,
the evolution of the total stresses as related to the
total normal strain over the cycles is plotted next. Figure 4. Distributions of the residual horizontal
Figures 6–7 present the total stress-strain rela- stresses σ xx
r
for φ° = 20° and load p0 = 7.5c.

60
 Figure 7. Total stress-strain relations σ yy ε zz at
z 0.5 m for φ° = 20° and load p0 = 7.5c.
_j

( )
Figure 5. Distributions of the residual stresses σ yyr that as the number of cycles increases, the area of
for φ° = 20° and load p0 = 7.5c. the loops, produced by the loading, gets smaller,
indicating that small amounts of energy dissipate,
while also the relative distance Δε zz between the
[x!.E3[ loops gradually becomes zero. In the last passes
0.00
the loops have zero area and they are on top of
each other ( Δε zz = ) , a fact indicating that the
response of the pavement is elastic, a proof that
shakedown has taken place.

4.2 Pavement with φ° = 30° and ψ° = 30°

In the second case, Model2 is being used for
analysis and a material of φ° = 30 and ψ° = 30°
is chosen. The shakedown load is determined at
p0 = 10.6c after twenty-eight load passes (N = 28).
Figures 8–11 present the plastic p strain ε xzp , ε zzp
and residual stress σ xx ( r
)
, σ yy
r
distributions after
twenty-eight load passes. It is observable that both
-1.0 o.o [xi.E-3[
ε xzp , ε zzp diagrams are stabilized at the end of the
Strain analysis. It is evident that the plastic shear strain
stabilizes after the eighth pass while the plastic
normal strain stabilizes after the twentieth fourth
pass. The residual stress diagrams both also stabi-
Figure 6. Total stress-strain relations σ xx ε zz at lize after the twentieth fourth pass.
z 0.5 m for φ° = 20° and load p0 = 7.5c.
The Figures 12–13 present the total stress-strain
relations σ xx ε zz and σ yy ε zz , respectively, dur-
ing sixteen load cycles at the critical point of the
tions σ xx zz and σ yy ε zz during sixteen load model. For the above material parameters this
cycles at the critical point of the model. For the point lies at depth 0.44 m ( z = 0.4 m) from the sur-
above material parameters this point lies at depth face. Again it is evident that the half-space shakes
5 m ( z = 0.5 m ) from the surface. It is evident
0.5 down.

61
 Strain S t ress
6 6 6 m
g; ~ ~ g ~ ~
~ ~ ~ 0

0~------~-~-----~
------~-~-----~
---------;----~
--~----~M~
-~ !~
0~--~--~~--~~--~--~~~~~~~-

:~~; "~~~
, ' ' -----~---:-------+-----
:L_ _ w~~~---+-~~~~... r_ ----=
-+ ___ r:
_ _ _ ,u
1
_____ _

--~---- ---+--- ----+--- ----i----------~---- Ph~

0

-----+---- V-------1-----------~---------+-----
~ ~ -+--- ----1---- ----+--- ---+--------+--- ----¥ - ------:----- ---- ~------ -----j----- ------:----- -----:------
t6
CL

[ : --1-----

!"'
---+--- ----+--- ---+--------+---1/ i------
j. " 1\ T T +
-----+---- ----+-----~- -----+---- ----+-----
----'---------C-------'-------f - .
--~---- ----j---- -----~---- ---+----V--~---- ---+-----
0
i..n : : : : : :
0 -----+---- ----+----- --- 1---- K -----+-----
0 -- : ---- ----:---- -----:---- ----:----

"' -----+--------+----------(·------··t··-
o~~--~--+-~~-+--+---~4+--~~
--:---- ----:---- -----:---- ----:--------- -~- --- ---- ~ -- ----
-----·t----- -----1------ -----!----- ----- -~----

PASS-1 "'~====~~~~~~~~
P ASS -4 5 11
PASS-8 PASS- 1
PASS-12
PASS -16
PASS -20
PASS-24
--------
!::=!
P ASS -4
PASS-8
P ASS- 1 2
PASS-1 6
PASS-28 PASS-20
P ASS -2 4
P ASS-28

Figure 8. Distribution of plastic shear strain ε xz

p
for
φ° = 30° and load p0 = 10.6c. Figure 10. Distributions of residual horizontal stresses
σ xx
r
for φ° = 30° and load p0 = 10.6c.

S t ress

Stra in ~·

----+----- -~Jr ~~--+-----

0
0
~ ~~
: :
0

:
0

:
0

:
0

:
0

:
0

"' ------1------
~ --i--~~ et+-- --j;---1; - --- -+-- -+-- -- -- r 1
b " lf..r~ '::i: ................... b •••••••••••• -

------r. ----- -----1.1------

· ·" \ \· ·

-----1.1 --~'· -----~----- -----~------

l l
~ - --~---~-- -+-- __ L__ --j--- -- --- -- --- -- ------ -- "' '

[ ------:----- ----+----- _____ !_____ :~~ ----:----- -----:------

1: ····- · · \r· - ·
~ u,

~ --i-1-+-- -+-- -+-- -+-- -- --- -- --- -- --- -- --

~ b
"'
~ ~ ------r----------1----------- j---------- .\---------1------
~ ~r-~+-~+-~+-~+-~4-~4-~4-~~~~

~ --:-- --:--- --:--- --:--- --:--- -- --- -- --- -- ------ -- o ------~----- ----+----- -----1----- ------~-\ ----+-----
"'
.u rrr: ·········· ·········· ------:----- -----:------ -----:----- ------:----
0 ~~--~--+-~---+--+-~~~--~~

------:----- -----:------ _____ : ___________ _____ :_ --- --~------

PASS - 1 PASS- 1
PASS-4 P ASS-4
PASS-8
PASS- 1 2 PASS- 1 2
PASS- 1 6 PASS -16
PASS -20 PASS-20
P ASS-24 PASS - 2 4
PASS-28

Figure 9. Distribution of plastic normal vertical strain Figure 11. Distributions of residual horizontal stresses
ε zzp for φ° = 30° and load p0 = 10.6c. σ yy
r
for φ° = 20° and load p0 = 10.6c.

62
 1. It is possible for pavements to shakedown
depending on the loading level.
2. Friction angle φ° and cohesion c are very impor-
tant factors for the shakedown load of an asso-
ciated flow rule material. When φ° increases so
does the shakedown load.
3. The critical point moves to the surface as the
friction angle increases.
4. Plastic shear strain ε xzp stabilizes much faster
than the plastic normal strain ε zzp for the given
load level.
Research on the subject is ongoing and further
results are expected to be presented at the time of
the Conference.

0.0
REFERENCES
·3.5 ·3.0 ·2.5 ·2.0 ·1.5 ·1.0 ~-5 0.5
Strain
I- LE22-Sll·ll'ASS·16 1 Abaqus 6.12. 2012. Analysis User’s Guide.
Bhargava, V., Hahn, G. T., Rubin, C. A. 1985a. An
elastic—plastic finite element model of rolling con-
Figure 12. Total stress-strain relations σ xx ε zz at tact, part 1: analysis of singlecontacts. Journal of
z 0.4 m for φ° = 30° and load p0 = 10.6c. Applied Mechanics 52: 67–74.
Bhargava, V., Hahn, G. T., Rubin, C. A. 1985b. An elas-
tic-plastic finite elementmodel of rolling contact, part
2: analysis of repeated contacts. Journal of Applied
Mechanics 52: 75–82.
Boulbibane, M., Collins I. F, Ponter, A. R. S., Weichert D.
2005. Shakedown of unbound pavements. International
Journal of Road Materials and Pavement Design 6: 81–96.
Collins, I. F. & Boulbibane, M. 2000. Geomechanical
Analysis of Unbound Pavements Based on Shake-
down Theory. Journal of Geotechnical and Geoenviro-
mental Engineering 126: 50–59.
Johnson, K. L. 1962. A shakedown limit in rolling con-
tact. Proceedings of the Fourth US National Congress
of Applied Mechanics 971–975. Berkeley, California.
Johnson, K. L. 1985. Contact mechanics. Cambridge
University Press.
Liu, S. Wang, J., Yu, H-S., Wanatowski, D. 2016. Shake-
down Solutions for pavements with materials follow-
ing associated and non-associated plastic flow rules.
Computers and Geotechnics 78: 218–226.
Radovski, B. S. & Murashina, N. V. 1996. Shakedown
of subgrade soil under repeated loading. Transporta-
Strain tion Research Record: Journal of the Transportation
I- LE22·S33-TPASS·16 1 Research Board 1547: 82–88.
Sharp, R. W. & Booker, J. R. 1984. Shakedown of pave-
ments under moving surface loads. Journal of Trans-
Figure 13. Total stress-strain relations σ yy ε zz at portation Engineering 110: 1–14.
z 0.4 m when φ° = 30° and load p0 = 10.6c. Wang, J. 2011. Shakedown Analysis and Design of Flex-
ible Road Pavements under Moving Surface Loads.
Doctoral Thesis. University of Nottingham.
5 CONCLUDING REMARKS Wang, J. & Yu, H. S. 2013. Shakedown analysis for design
of flexible pavements under moving loads. Road
In this paper, a numerical step by step approach Materials and Pavement Design, 14:3, 703–722.
Yu, H. S. 2005. Three dimensional analytical solutions
has been developed to obtain shakedown limits of for shakedown of cohecive-frictional materials under
single layered pavements assuming an associated moving surface loads. Proceedings of the Royal Soci-
flow rule. Comparison has been made with the ety A: Mathematical, Physical and Engineering Sci-
results of other researchers. The important conclu- ence 461: 1951–1964.
sions are: Yu, H. S. 2006. Plasticity and geotechnics. Springer.

63
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Bearing capacity on stockpiled fly ash

Carl A. Lenngren
Sweco Civil, Sweden

ABSTRACT: Fly Ash is commonly used for soil stabilization. However, the demand for the product
seldom meets the supply, and stockpiles become a necessity. These in turn are usually somehow restricted
due to environmental concerns and limited availability of suitable areas. It was noticed that fly ash could
sustain heavy trucks while the material was being transported and moved around between the stockpiles.
The idea came up whether the fly ash could serve as a suitable material for industrial projects in lieu of
more traditional ones. The benefits would be reducing stockpiles and saving resources of gravel. A falling
weight deflectometer test was used to determine the mechanistic properties of fly ash. A minimum layer
thickness was further determined on a given subgrade. It was found that fly ash could be used for struc-
tures carrying rather heavy vehicles, but there is a weak zone at a depth where high shear stresses occur.
Freeze/thaw cycles may also affect this critical zone.

1 INTRODUCTION 4 BACKGROUND

Fly ash is a product that can be used for stabiliz- Fly ash has been used for many different purposes.
ing soils, and there are specifications and methods It is common, and even required, as an additive
for its use on finer soils. It is a viable alternative for Portland cement concrete. Soil stabilization,
to lime stabilization, as it does not deplete natural replacing cement and lime is another common use,
resources and reduces stock piling. which has been around for several decades. It is also
Most specified use is for traditional road being used for improving base and subbase materi-
projects and soil stabilization though. While han- als. More recently, it has been used for structurally
dling stockpiles, a Swedish company noticed that improving fills, and for waste stabilization. Most fly
the material was strong enough for rather heavy ash comes as a waste product from coal fired plants.
equipment to traverse it without any substantial Chemically, it consists of oxides of silicon, alumi-
deformation. The idea came up to use this material num, iron, and calcium. To a lesser degree, magne-
as embankment and/or subbase material or even sium, sodium, potassium, and sulfur in such ashes.
for temporary roads in conjunction with industrial In the present case, the ash comes from the burning
projects. Thus, a Falling Weight Deflectometer of waste materials. The composition of chemicals
(FWD) test was planned for a stockpile area to see vary, so only the structural properties such as elastic
what the bearing capacity was. The area was sev- and plastic response is considered. As the size of the
eral meters thick and there were tracks from haul- particles is about the same as fine silt. Singh et al
ing trucks that could be tested. (2015) pointed out that for strength the most impor-
tant parameters are physical property, chemical
reactivity and curing conditions. Gain of strength
2 OBJECTIVE is influenced by compaction, compression and con-
solidation. Santos et al. (2011) found that the opti-
The objective with the present paper is to determine mum water content for 100% fly ash is 45.5%. If
appropriate layer thicknesses for fly ash as a road kept so, the unconfined compressive strength could
material suitable for hauling trucks and similar vehi- be as highs as 9000 kPa after 14 days of curing.
cles. Further, the bearing capacity and mechanistic
properties of such layers should be determined.
5 FIELD TESTING

3 SCOPE The field-testing took place in the afternoon dur-

ing a spring day in sunny weather in a flattened
The present study deals with preliminary findings mound of fly ash near Örebro, in South Central
only to direct further studies for construction rec- Sweden. The area size was about 10 000 m2, and
ommendations and subsequent specifications. It there were some piles of fly ash, and a pile of wood
deals only with the mechanistic response. chips intended as fuel, that were not accessible for

65
the FWD (Figure 1). The pile was resting on a silty Testing on an unbound surface one cannot
moraine, and it was at least three meters thick as expect to attain as good sensor readings as on a
reported. hard surface. The measured/backcalculated sensor
To cover the area five lines in front of the wood Root Mean Square (RMS) value ranged from 0.71
chips pile and one line behind it were tested at to 32.2 percent. By sorting the data it was possible
intervals of 25 meters. As some plastic deforma- to discriminate the data set after various RMS lim-
tion on this unbound material was to be expected, its. It was found that the mean values were affected
multiple drops at three different load levels were the most by high RMS, but the other parameters
used (Table 1). like the median and percentiles between 10 and 90
The seismometer readings turned out to be percent were not influenced by the discrimination.
adequate on this surface. Thirty-six sections were Thus, in the following all basins, except the one
tested with ten drops each. One test point was omitted and the seating drop, are included in the
abandoned entirely due to a sensor reading error. study unless stated otherwise. Table 2 shows the
10, 50 (median) and 90 percentile values for drops
8–10.
5.1 FWD backcalculation and analysis
Looking at the median, all three layers seem
A backcalculation analysis was done with three to stiffen when the load, and thus the bulk stress,
layers. The fly ash extended down to several increases. The top layer is rather stiff indicating
meters, but the stiffness was expected to change that some chemical reaction has occurred. The var-
with depth. iability is high though, most likely due to broken
bonds. The intermediate layer must be considered
• The confining pressure is increasing due to the
as weak, and high shear stresses may prevent the
overburden. (Stiffness increases by depth);
material from binding.
• The compaction by traffic decreases by depth;
The “subgrade” including a large portion offly
• The surface layer is affected by precipitation,
ash has a modulus corresponding to gravel. The
freeze/thaw, and shear-stress by turning vehicles;
load influence is larger and deeper for the higher
• Frost penetration is 630 mm.
loads, so the values are not directly comparable.
With this in consideration, an approach was
tried with a surface zone, an intermediate zone, 5.1.1 Post compaction
and a subgrade. By varying the layer thicknesses, By comparing subsequent drops, it is possible
the surface layer was chosen to be 80 mm thick, to estimate the degree of compaction. The first
and the intermediate layer was 550 mm, extending repeated drop usually results in a slightly stiffer
to the depth influenced by frost. response. This is explained by the release of residual
stresses induced by a different load or temperature
change. The difference of deformation is normally
within a few percent.
As can be seen there is some compaction exerted
by the FWD as E(1) and E(2) are getting stiffer by
the number of drops. E(3) is different and there is a

Table 2. Layer E-moduli stiffness.

Layer [mm]

(1) (2) (3)

Load 0–80 80–630 >630

E-Modulus [MPa]

Figure 1. The area was tested by forming six lines by Drop 8, 10%-ile 25 133 28 83
the FWD. (Photo by the FWD operator). Drop 8, Median kN 1831 71 202
Drop 8, 90%-ile 9540 356 409
Drop 9, 10%-ile 50 166 32 83
Drop 9, Median kN 2826 95 206
Table 1. Drop sequence with multiple loads, [kN]. Drop 9, 90%-ile 7133 525 471
Drop 10, 10%-ile 70 178 33 88
Drop 1 2 3 4 5 6 7 8 9 10
Drop 10, Median kN 3160 109 202
Load 50 25 50 70 25 50 70 25 50 70 Drop 10, 90%-ile 9155 760 420

66
tendency of a softer subgrade layer. However, this The best coefficient of determination, R2 = 0.11
is consistent with the lower bulk stress, provided by was found between E(1) and E(3). Omitting, some
the top layers being stiffer. of the basins solved with higher root mean square
Looking at the median the tendency is quite values, the correlation improved marginally for lay-
clear, but there are other processes going on. ers 1 and 2. The high RMS values included basins
Figure 2 shows the top layer modulus for drops with low outer sensor readings, which the backcal-
2 and 10. The stations are sorted from lowest to culation interprets as a stiff subgrade. Hence the
highest E-modulus for Drop 2. The solid line cir- low correlation between layers 2 and 3, when all
cle in the figure show that an increase in modulus basins are included. These results indicate that the
occur for the higher loads. The dashed circle shows surface was not prepared, nor treated with com-
a decrease in modulus for the higher load levels. paction. Other studies have shown some correla-
It could be that the internal friction is building up tion between subgrade and subbase moduli, as a
in the former case. For the latter case, the friction good support facilitates the compaction.
may not be sufficient for the higher loads, and a Figure 3 shows the second layer modulus sorted
slip is taking place. on the last drop from lowest to highest. The layer
From an engineering standpoint, one would like did not develop any strong bonds like the top one
to increase the modulus for those values less than for 18 out of the 29 data points. Probably due to
200 MPa. In this case 4 of 28. It could be that these shear forces from the trucks. Two or three points
testing points were either dry, or more likely too have a very low modulus. They do coincide with
wet when subjected to the load, as the optimum rather stiff top layer properties, meaning that the
moisture content is important when compacting bending from the load could have shifted to some
the material. distance from the loading plate. The vertical strain
in the layer is high almost 0.004 in the middle, so
5.1.2 Correlation of layer moduli probably plastic deformation will take place at a
The data were tested for relationships between high rate. In effect meaning that the stress is above
the layers, but only weak correlations were found. the shakedown limit.
Figure 4 shows the subgrade modulus, includ-
Table 3. Median E-moduli [MPa] for different Load- ing a large part of the fly ash. There is some vari-
Levels (LL). ability with one low and one high value. Other
than those two outliers, the values are quite on par
Drop LL E(1) E(2) E(3) with corresponding ones for highways in the area.
2 25 1692 78 212
5 kN 1723 78 210 Table 4. Coefficient of determination between layer
moduli.
8 1831 71 202
3 50 2358 99 212 E(1) E(2) E(2) E(3) E(1) E(3)
6 kN 2766 94 208
9 2826 95 206 All basins 0.02 0.00 0.11
4 70 2503 111 211 N = 253
7 kN 2824 109 203 Basins with 0.03 0.21 0.22
10 3160 109 202 rms <5% N = 113

100000
10000
10000

1000
1000
"'
c._
:2;
100 "' 100
~

10
10
- - E(2)-10

1 3 5 7 9 11 13 15 17 19 21 23 25 27
#Station 1 3 5 7 9 11 13 15 17 19 21 23 25 27
- - - E(1)-2 - - E(1)-10 #

Figure 2. E(1) sorted lowest to highest by drop 2. Figure 3. E(2) modulus sorted after drop 10.

67
 100
10000
90
80

/
--
1000 70
60
"' ...z
~
0..
100 50
~
40
10 30
20
10
1 3 5 7 9 11 13 15 17 19 21 23 25 27
# 5 10 15 20 25 30 35 40 45 50 55 60
ms
Figure 4. Subgrade modulus sorted after tenth drop. Figure 5. Different load pulse shapes.

The material under the pile would likely have a

modulus of about 80 MPa.

6 TIME HISTORIES

If the load is plotted versus the displacements,

curves looking like hysteresis are formed. Even if
the area formed within the curves is not a direct
measure of energy losses in the layers, they have
been calibrated to rolling resistance (Lenngren,
2012). For a highway pavement the attenuated work
is about 2–8 Nm for an asphalt concrete pavement.
The subgrade modulus has a major influence, so
thicker pavements are usually better than thin ones.
The time history of a drop shows the load and
the deflections over time. There is a lag between
Figure 6. Time History software showing graphs of
the load and the displacement, due to a number time domain as well as load and displacement.
of reasons. Inertia, visco-elastic properties and
damping are common phenomena on highways. In 60
the present case, soil damping and saturated mov-
ing water are to be expected. Figure 5 shows some
so
load pulses in the time domain. The maximum
load can be changed by varying the drop height.
The length of the pulse can be changed by replac- 40
ing the stiffness of the rubber buffers in the FWD z-"'
hardware. :; 30
Figure 6 shows a screen dump from an evaluat- -' "'0
ing tool. It shows the graph for the time domain, 20
and the load-deflection curves as well. It is being
used for assessing the energy dissipation. 10
Figure 7 shows a plot of load and deflections at
the point with the lowest stiffness. This is the first
drop, so there is plastic deformation going on. 0
Further, the surface is loose, and there is a surface 0 500 1000 1500 2000 2500
wave developing, interfering with the natural elas- Displacement [mu]
tic rebound. The dissipated energy is about 45 Nm, - - DO - - 020 - - 030 - 045
which is a very high value. After repeating the drop
- - 060 - - - 090 ----· 0120
twice, the dissipated energy drops to 22 Nm for the
same load level. The maximum deflection drops
from 2.5 mm to 2.0 mm. No effect is seen on the Figure 7. Load-deflection diagram with a dissipation
stiffness of layer 2. of 44 Nm.

68
 60 8 CONCLUSIONS AND
RECOMMENDATIONS
so
The fly ash material has intrinsically a very stiff
40 response suitable for heavy industrial vehicles. At
z-" the test site, some points could without doubt serve
::;;'30 axle loads seen at harbor terminals and the like, up
"'0
..... to 1000 kN axle loads. At other points, the stiff-
20 ness of the intermediate layer was extremely low,
and the attenuation was very high, indicating that
10 the FWD load was exerting compaction to a high
degree. The test was done in May, and it could be
0 that these points had not recovered from frost pen-
0 300 600 900 etration. It is unknown if they were disturbed by
Displacement [mu] traffic or not.
- - DO - - D20 - - D30 - D45 The recommendations are to follow compac-
tion and procedures to let the layers cure properly.
- - D60 - - - D90 ----· D120
The minimum layer thickness is governed by the
frost depth penetration, which in this area is about
Figure 8. Load-deflection diagram with a dissipation 1100 mm for a 30-year period. The fly ash properly
of 4 Nm. cured is much stronger than any unbound subbase
material.
Another recommendation is to find out the
Moving on to an intermediate station with a chemical signature of the ash, which should be
stiff subgrade the response is much more elastic, different from coal fly ash. There might also be
and the attenuated energy is only 4 Nm (Figure 8). restrictions of use if some unwanted chemicals are
The stiffest response is from a section where prevalent.
layer number 2 and the subgrade are stiff. The From a mechanistic point of view, the fly ash
dissipation is only 0.2 Nm, which is on par with should be an adequate material for embankments
a Portland Cement Concrete Road. This tremen- and well as subbase and base layers.
dous range of stiffness and attenuation is indica-
tive of wanted effects being random, as no effort
was taken to abide by any specification. REFERENCES

American Coal Ash Association. Fly Ash Facts for High-

7 DISCUSSION way Engineers. FHWA Technical Report FHWA-
IF-03-019. 2003.
The idea to use fly ash in landfills is not new. There Lenngren, C.A. Pavement Contribution to Truck Rolling
Resistance, Proceedings 8th International Conference
are excellent handbooks for specifications and con-
on the Bearing Capacity of Roads, Railroads, and
struction practices published by the American Coal Airports. University of Illinois, 2012.
Ash Association (2003). As was found in the FWD Santos, F., Li, L., Li, Y., Amini, F.; Geotechnical Proper-
study the bearing capacity is quite adequate for most ties of Fly Ash and Soil Mixtures for Use in Highway
test points, but the variability is high. The weaker Embankments. World of Coal Ash (WOCA) Confer-
points may be due to lack of compaction, freeze/ ence, Denver Colorado, 2011.
thaw action and little or no control of compaction Singh, R.R; Nitin G.; Navpreet, K. Fly Ash as an
practices. After all the stockpile was not intended Embankment Material. SSRG International Journal
for any specific use; it just happened to evolve into of Civil Engineering (SSRG-IJCE) – volume 2 Issue
3 March 2015.
an area where loads where carried around. So, by
following recommendations found in the literature
the problematic stations should be alleviated.

69
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Tunnels interaction with surrounding alluvial soils

O. Naeemifar & R. Rahbari

Department of Civil Engineering, Faculty of Technical and Engineering, Malard Branch,
Islamic Azad University, Malard, Iran

ABSTRACT: Behavior of single and twin tunnels interacted with surrounding alluvial soils has been
investigated. 4 different real sites has been selected with real geotechnical and seismic down hole test
results, in order to model the real condition of surrounding soils. The results show the significant effect of
shear wave velocity and tunnel depth on the lining moment, axial forces and settlements. Also, it has been
concluded that the values of lining forces and moments of twin tunnel, has been strongly dependent on
the tunnels distance and also the type of loading (static or dynamic), where the trend of their variation is
also dependent on the soil properties considerably.

Keywords: Circular tunnels, Static, Dynamic, Single, Twin, Shear wave velocity

1 INTRODUCTION their behavior. Chehade & Shahrour (2008) con-

cluded that more distance between tunnels leads
The effect of different parameters on the static and to decrease in surface settlement, while the effect
dynamic behavior of tunnels has been one of the of tunnels will be negligible when distance between
main concerns in the previous studies. The most them is 3 times greater than their diameters.
important parameters can be type of the soil, tun- Also they concluded that for investigated sandy
nel depth and shape of tunnel section. soil, the values of axial forces and moments are not
Nicieza et al. (2008) showed that increasing in dependent on the tunnels distance resulting in simi-
depth will result more settlements around the tun- lar values for single and twin tunnels. Other results
nels, where for uncircular tunnels the relative set- of latter investigation have been shown that the
tlements are different for different points of tunnel relative position of the tunnels (vertical or inclined)
lining like crest or walls of tunnel. Lee (2009) using and also the tunnel construction process can be
the physical and numerical (finite element) meth- very effective on the lining moment or axial forces.
ods showed that excavating tunnels in coarse grain The main scope of this paper is investigating
soils will result considerable settlements in differ- the effect of the shear wave velocity, tunnel depth
ent depths, being increased with more distance and also the interaction between twin tunnels and
from the ground surface. Also, he concluded that surrounding soils. The latter has been investigated
increasing tunnel depth will lead to a more vast for circular tunnel based on the finite difference
profile for distribution of the surface settlements. method and under static and dynamic loading.
Nunes & Meguid (2009) performed some physi-
cal and numerical investigation about the soil type
and its thickness concluding that the existence of 2 NUMERICAL MODELING AND
the coarse grain layers around the tunnel will result MATERIAL PROPERTIES
in considerable reduction of the lining moment.
More thickness of coarse layers or less distance Numerical modeling has been performed based on
between coarse layer and tunnel will emphasize its the finite difference method using the Flac 2D soft-
effects. The results of Hagiwara et al. (1999) and ware. Circular tunnel are modeled supposing to be
Taylor & Grant (2000) is also emphasizing that the excavated with TBM. The modeling procedure has
existence of coarse grain layers above the tunnel been consisting of creating the model geometry,
will results in less forces values and or moments. considering material properties, applying the initial
Chu et al. (2007) have reported similar results. and boundary condition, static loading (excavat-
One of the other main problems is the interac- ing tunnels, stress relaxation and lining construc-
tion of twin tunnels that has been became of great tion), dynamic loading (applying) and sensitivity
importance in big cities. The distance between twin analysis. Sensitivity analysis has been consisting
tunnels or their position can be very effective on of investigation the effect of shear wave velocity,

71
tunnel depth and also the distance between twin Table 2a, Lining moment (Ton-m).
tunnels and their interaction.
Mohr-Coulomb model has been used for Finite difference Finite element
numerical modelling as an Elastic-Perfect Plastic
Static analysis 8.3 9.2
model. The required parameters for this model like Dynamic analysis 35 32.5
cohesion, friction and elasticity modules has been
determined based on the real geotechnical test
results for each of the sites.
Table 2.b. Lining axial forces (Ton).
4 different soil conditions have been modeled
based on the real seismic and geotechnical test Finite difference Finite element
results of different sites. These soil conditions are
selected in a way to correspond to loose to very Static analysis 70 78
dense soils. The Mohr-Coulomb parameters values Dynamic analysis 100 112
are presented in Table 1.
*Dynamic results correspond to maximum values.
The diameter of modeled tunnels is 10 meters
and the analysis has been performed considering
the tunnel center depth 15 and 25 meters below the 3 MODEL VERIFICATION
ground surface. In order to investigate the interac-
tion between twin tunnels, the ration of their cent- Verification of the models has been performed
ers distance to their diameters has been considered using the finite element method, Plaxis software.
from 1.2 to 3 variables. The static and dynamic results for axial forces,
Figure 1 shows vertical static displacement for moment and displacement were compared, result-
a tunnel model in a loose soil (Vs = 257 m/s) with ing in good accordance between them. Table 2
15 meters depth. It should be noted that a concrete presents results of model verification for an exam-
lining has been considered with 40 cm thickness as ple tunnel model (Depth: 15 m, Vs = 257 m/s).
the permanent retaining structure.

Table 1. Average model parameters for different soil 4 NUMERICAL RESULTS

condition.
In this section the numerical results of static and
Parameters Site 1 Site 2 Site 3 Site 4 dynamic analysis are presented. The effect of param-
eters like shear wave velocity, tunnel depth and also
Shear wave Vs m/s 255 465 600 800
velocity
the twin tunnel distance has been investigated. It
Friction ϕ Degree 20 26 30 36 should be noted that the lining moment variation,
angle axial forces variation and also the variations of sur-
Cohesion C Kg/cm2 0.15 0.2 0.2 0.4 face settlement and crest settlement has been investi-
Elasticity Eave Kg/cm2 265 480 650 800 gated in order to evaluate the tunnel behavior.
modulus
(average)
4.1 Static analysis results
Poisson’s υ – 0.45 0.4 0.35 0.32
ratio The static analysis results for tunnels of 15 and
25 meters depth has been presented in Figure 2. It
can be seen that the general trend of variations is
the same for two different depths. The values of
axial force, moment and settlement are the maxi-
mum values due to static loading. Dashed lines are
presenting the tunnel with 15 meters depth and
solid lines are related to tunnels with 25 meters
depth.

4.1.1 The effect of shear wave velocity

As can be seen in Figures 2a and 2b, increasing the
shear wave velocity of the soil from 250 to 800 m/s,
has resulted in reduction of the maximum lining
moment while the variation of axial force is not
significant. It seems that the axial force of lining
Figure 1. Vertical static displacements (Depth: 15 m, is more dependent on the tunnel depth more than
Vs = 257 m/s). the soil type itself.

72
 The axial force distribution in the lining is neu- velocity resulting in less effect of the overburden
tral compression, while the moment distribution soil and finally decrease of the crest settlement. In
is compression or tension in different point of lin- a similar manner, the increase in the stiffness of the
ing. Increase in shear wave velocity corresponds to soil reduces the excavation effects on the ground
more uniformity or stiffness in the soil resulting surface, resulting in less surface settlements with
in fewer tendencies for oval shape deformations increasing the shear wave velocity.
(deformation from circular to oval shape will result
in more bending moment in the lining). 4.1.2 Tunnel depth effect
Figures 2c and 2d shows the reduction of the From Figures 2a and 2b, it can be concluded that
crest and surface settlements as results of shear the values of axial forces and bending moments
wave velocity increase. The latter is due to more will be increased with increase in tunnel depth, due
stiffness in the soil by increasing the shear wave to the more stress on the tunnel lining in deeper
tunnels. The effect of depth has been more sig-
~~ ---------------------------------.~~~
nificant about the axial forces, referring to more
& uuuuuuu -.~ dependency of the axial forces on the tunnel depth.
Also the results of the figures 2c and 2d shows
the strong dependency of the crest settlement on
the tunnel depths, while the surface settlement var-
iations are more dependent on the soil type and its
shear wave velocity.
20 ------------------------------------------------------------------------
It can be noted that the graphs of different tun-
200 300 400 600 700 BOO
nels depth are tending to gather with increasing the
Vs (m/s)
shear wave velocity, referring the reduction of the
a. Max. axial force variations tunnel depth effect.

,__10.0 4.1.3 Static analysis of the twin tunnels

9 8.0
§ ----~-.....·.;;:,:··-- Figure 3 present the variations of the normalized
t: moment, axial forces and displacements versus the
-----~-<:~~:~-:::~.:~-~-'·-·-----;;~-8
c 6.0
~ (d/D) ratio for twin tunnels with 15 meters depth
~ 4.0
(d/D = distance between tunnels center to tunnel
2.0 --- ----- -- -- - -- - -- ---- - ---- -
diameter ratio).
0.0 +---~----~----~----.---~----~-----! The analysis has been performed for two soil
200 300 500
Vs(m/s)
600 700
conditions with 257 and 800 m/s the values of
b. Max. bending moment variations maximum moments, axial forces and displace-
ro ,-~------------------------~==~
ments of the twin tunnel are normalized to their
70 .... ~.!o""·······-··-··················-···········-·······--· ~ 25m corresponding values of the single tunnels for simi-
E ',
,§_ 60 ----· ----- .....,~ -------------------- ------------------- - ------ - -13- · !Sm lar condition. The surface settlement refers to set-

!=:
E
50 •••••••••• ----~,-------------··········································
',
u•'•;;<::uuuuuuuuuuuu•uuuuu
tlement of the ground surface in the middle point
between the tunnels.
~ 20 > :--..----------------------- 4.1.3.1 Distance and shear wave velocity effects
IJJ 10 From Figures 3a to 3d, it can be concluded that the
values of forces or settlements has been increased
200 300 500 600 700 BOO 900
Ys (mfs)
considerably in comparison to single tunnels.
c. Max. surface settlement variations For d/D = 1.2 (the distance between center of
tunnels is equal to 12 meters), the axial forces val-
E
14 ues has been increased about 35% as an average
! 12 value for different type of the soil, where shear
= 10
wave velocity of the soil has not shown significant
~ 8
] 6 " " " "" '" " " " " " " " ' .................................................................______
effect on the axial force variation.
~ In contrary to axial forces, the type of the soil
u 4
: : ~::~:~~:=~~~~:~====:~==~=:::::::::;;;~ has been shown a significant effect on the intensi-
fication of the moment values (Figure 3b), where
200 300
""" 500 600 700 BOO the values of intensified moments differ about 30%
Vs (m/s)
d. Max. crest settlement variations
for two different soils. This difference is became
diminished with increasing the distance between
Figure 2. Static analysis results for tunnels with 15 and tunnels, referring the reduction of the soil type
25 meters depths. effect for more far away tunnels. Similar trends are

73
 4.2 Dynamic analysis results
~
In order to investigate the dynamic behavior of the
"
~

~
1.3

1.2
--

-- single and twin circular tunnels, the models were

<
~ 1.1
--···· ............................. ,......._.............................
_ _ [i. _ _ _ _ . analyzed with Loma Prieta earthquake accelogram
] 1.0
(Figure 4). Figure 5 present some results of the lin-
-;; ing moment variations during earth quake loading.
~ 0.9
z Dynamic analysis results for tunnels with 15
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2 and 25 meters depth are presented in Figure 6. The
diD maximum moments or axial forces during earth-
a. Normalized maximum axial force variations quake are illustrated with “Max”, while the resid-
----- ~ -e ""' _______________________ _ ual moments or axial forces after the earth quake
·····;·r ;··----------,I are illustrated with “Res”.
~ 1.4 ·············'l·················-t·······································
i 1.3
- __ l _____________________________________ _
I 4.2.1 Shear wave velocity effects
~ 1.2 ''\ii'''''''''''' As can be seen in Figure 6a, the maximum and
] 1.1
~ 1.0
residual values of the dynamic axial forces doesn’t
E show significant variation with shear wave velocity
s 0.9
" 0.8 -1---~~~-~~~~~-~~~~~---.--------1
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2
diD
b. Nonnalized maximum moment variations
3.000

.-i? '
·······------ ~ -------- \···

; \ 2.000
····;,·"'-~----···--·------~
;
----~-,.. -·
; '
----·····-.' . .,-. : ·······-------------- 1.000
; '
-~---.......,~=--__..,
0.000

-1 .000
diD
c. Nommlized maximum surface settlements variations
-2.000
2.5
eg 2.4 "" '
··········-/··-----\ ···---------
------------------- ~!_ __________ )
= 2.2 ; \
--,-------------------1>{-------
j
2.0
-------' ------ -------~ -----

~
1.8
1.5
; '
········- ~ ·-·····························\······························ Figure 4. Loma Prieta accelogram used for dynamic
u '
-----7 ;·---- ------------------- ------------ ""'\·----------------- analysis.
..
11
1.4
1.2 7~- --~~-:,:: -··--------

1
z
1.0
0.8
Y-axis
Moment 1 (EI 6)
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2 (10+05 ) Moment1 (EI 7)
diD Moment 1 (EI 8)
d. Normalized maximum crest settlements variations Moment1 (EI 9)
3.000 Moment 1 (EI 10)
Figure 3. Static analysis results: normalized parameters Moment1 _(EI 11)
Moment 1 (EI 12)
variations of twin tunnels versus distance for different 2.000
soil condition (tunnel depths: 15 m).
1.000
concluded about the crest and surface settlements
(Figures 3c and 3d). For near distance (d/D = 1.2),
0.000
the type of the soil is strongly effective on the set-
tlement intensification, where loose soils with less
-1.000
shear wave velocity shows more intensified set-
tlements. With increasing d/D ratio or distance
-2.000
between tunnels, the graphs difference due to dif-
ferent soil type has been diminished.
General trend of Figures 3a to 3d shows that for 10 15 20 25 30 35 40 45
d/D ratio greater than 2.2, the forces or displace-
ment values have not any significant difference Figure 5. Some examples of the lining moment varia-
with corresponding single tunnels values. tions during earth quake loading.

74
and almost remain constant for different types of considerably dependent on the tunnel depth and
the soils. Figure 6b shows that the maximum and overburden pressure without significant depend-
residual moments of the dynamic loading decrease ency on the shear wave velocity or soil type.
with increasing the shear wave velocity, similar to
what observed for static analysis. Of course the val- 4.2.3 Dynamic analysis of twin tunnels
ues of dynamic analysis have been increased con- Figure 8 presents the variations of normalized
siderably in respect to static loading. maximum axial force and maximum moment ver-
Considering the static and dynamic analysis sus d/D ratio, for tunnel with 15 meters depth.
results, it seems that the moment variation is con- Similar to static case, the dynamic analysis is per-
siderably dependent on the soil type and shear wave formed for two different soil types, with shear wave
velocity, while the axial forces values hasn’t shown velocities equal to 257 and 800 m/s the values of
significant variations for different soil types. maximum moment and maximum axial forces are
normalized to corresponding values of single tun-
nel for similar condition.
4.2.2 Depth effect
In order to better investigate the effect of tunnel 4.2.3.1 Effect of distance and shear wave
depth on the moment and axial forces variations, velocity
the variations of maximum axial force and maxi- Considering Figures 8a and 8b, it can be con-
mum moment versus depth of tunnel is presented in cluded that the distance between twin tunnels is
Figure 7. As can be seen in Figure 7a, the axial forces less effective on the intensification of the maximum
variation with depth is considerable, but the graphs moment and axial forces in comparison to static
of different soils have not shown significant differ- case. Dynamic analysis results have shown that the
ence. In contrast, from Figure 7b it can be seen that increase of maximum moment is insignificant even
the moment variation with depth is insignificant, for very little distance between tunnels. Also the
while difference of soil type and shear wave velocity increase and intensification of the maximum axial
has resulted a great difference between the results. forces is less in comparison to the static case. As an
The above conclusion dictates again that the example, for d/D = 1.2, the maximum moment is
moment variations is mainly dependent on the intensified about 60% for static loading, while the
soil type and tunnel depth hasn’t significant effect increase of the maximum moment for similar con-
on its variations, while axial force variation is dition and due to dynamic loading is less than 10%.

--
lW , ------------------------------------------,
1W ,-----------------------------------~====~
140 ---tr--257
150 --------------------------------------------------------------
--l:::.-- 465
120 ---------------------- 13 ---------l~f -----------------------------

§ -a- 600
.. ""'""El
- - -- ~------ --- ...-----~--..-.-----~--------------6- -- - - - t:,
~ 130
140
----e--- 800

::::~:~~~~~~=~-:~~:~~~:~~~~:~~:~~~:~~;:;~;~~:::::::::_ --.!1- Max (25m)

~ 120

"'~ 110
·------------------------------------------------------- ~Res(25m)

~ 100
20 -------------------------------- ________ --------- _______ - '!!. - MaJI. (15m)
-e- Res (lSm) 90 -------------------------------------------------------------------------

80 +----------r----------r----------r--------~
200 300 400 500 700 900
10 15 20 25 30
Vs (m/s) Dcptb (m)
a. the variations of Max. and Res. axial forces a_ the variations of Max_ axial forces
40,---------------------------------~====~

35 --------------- --AA-----~-~
--~
--'"'-"'
- _-,
__:-:_:-:
__:-:
__,-:_::
__::
__::-
__:':"_"::
__::__,.,
.!>______ --o- 257
- -.!1-· 455
---------------- - ~~~~~~---~~~~~~---:~~~~~-_-.:.:.: : ::·_-_-_:;:~ ------ --8- 600

--G-· 800
!3-------------------------------fl -----------------

5 --------------------------------------------------------------------------

200 300 400 soo roo 700 900

10 15 20 25
V 5 (mls)
Dt'ptb(m)
b. the variations of Max_ and Res_Moments b. the vaTiations of Max_ moments

Figure 6. Dynamic analysis results: Variations of lining Figure 7. Dynamic analysis results: Variations of maxi-
moment and axial forces with shear wave velocity for tun- mum lining moment and axial forces with depth for soils
nel with 15 and 25 meters depth. with different shear wave velocities.

75
; 1.25
tunnel depth. Less shear wave velocity values will
""] 1.20
results in increased crest and surface settlements.
<
~ 1.15 – based on the static analysis results, the intensifi-
~ 1.10 cation of the moment, axial force and settlements
1l ---s ----------o
,!i 1.05 due to interaction of twin tunnels is consider-
~; 1.00 able especially for small distance between tun-
z
0.95 nels. Also the effect of soil type is considerable
and less shear wave velocity values will results in
O~MMMIUUUU2UUUU3U
more intensified moments and settlements. For
d/D
d/D ratio greater than 2.2, the results gradually
a. variati ons of norm alized maximum axial force
tend to the single tunnel results and the interac-
tion between twin tunnels became insignificant.
~ 1.10 – under dynamic loading, the interaction effects
g of twin tunnels on the intensification of axial
~ 1.08

"
~ 1.06
forces is less in comparison to static case, while
"' the moment variations is also insignificant.
~ 1.04
Also it was concluded that the effects of soil
~ 1.02
type and shear wave velocity on the interaction
i
1.00 of twin tunnels is insignificant under dynamic
loading.
O~MMM I UUUU2UUUU3U

diD
b. variations of normalized maximum moment
ACKNOWLEDGEMENT

Figure 8. Dynamic analysis results: Variations of The writers would like to thank the respected
the normalized maximum axial force and moment reviewers for their worth full comments.
versus d/D ratio for different soils (depth of twin
tunnels = 15 meters).
REFERENCES
The graphs of different soils in Figures 8a and Bin-Lin Chu (2007). Mechanical behavior of a twin-tun-
8b are almost similar and their difference is insig- nel in multi-layered formations. Tunnelling and Under-
nificant. It means that in contrary to what observed ground Space Technology, Vol. 22, pp 351–362.
for static analysis, the effect of soil type and shear Chehade, F. & I. Shahrour. (2008). Numerical Analysis
wave velocity on the interaction of twin tunnels is of the interaction between twin-tunnels: Influence of
insignificant under dynamic loading. the relative position and construction procedure. Tun-
nelling and Underground Space Technology, Vol. 23, pp
210–214.
5 CONCLUSIONS Grant, R.J. & R.N Taylor (2000). Stability of tunnels
in clay with overlying layers of coarse grained soil.
GeoEng 2000, An International Conference on Geo-
The excavation effect of the single and twin tun- technical and Geological Engineering, Melbourne,
nels under dynamic and static loading and also the Australia.
effect of soil type, tunnel depth and the distance Hagiwara, T., R.J. Grant, M. Calvello, R.N Taylor
between twin tunnels have been investigated. The (1999). The effect of overlying strata on the distribu-
most important results are as follows: tion of ground movements induced by tunnelling in
clay. Soils and Foundations, 39 (3), pp 63–73.
– based on the dynamic and static analysis results, Nicieza, C., A. Vigil, A. Diaz & C. Palacio. (2008).
the lining axial forces variations are mainly Influence of the depth and shape of a tunnel in the
dependent on the tunnel depth and overbur- application of the convergence confinement method,
den pressure, while the moment variations are Tunnelling and Underground Space Technology, Vol 23,
mainly dependent on the soil type and shear pp 25–37, (2008).
wave velocity without significant dependency on Nunes, M.A, & M.A. Meguid, (2009). A study on the
the tunnel depth. effects of overlying soil strata on the stresses devel-
oping in a tunnel lining. Tunnelling and Underground
– increasing the tunnel depth and over burden Space Technology, Vol 24, pp 716–722.
pressure will result in considerable crest settle- Yong-Joo Lee (2009). Investigation of subsurface defor-
ment increase. Surface settlements are mainly mations associated with model tunnels in a granular
dependent on the soil type and shear wave mass, Tunnelling and Underground Space Technology,
velocity without significant dependency on the Vol. 24, pp 654–664.

76
Unbound granular materials
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Structural characteristics of unbound aggregate materials

in conventional flexible pavements

Long Wang & Xiao Guang Xie

School of Transportation Science and Engineering, Harbin Institute of Technology,
Harbin Heilongjiang, P.R. China

ABSTRACT: The objective of this study was to investigate the dynamic structural characteristics of
unbound aggregates with quantitative expressions. Specimens were fabricated with top-down surface
vibrations, saturated and unsaturated specimens were administrated to load dynamically with 3000 rep-
etitions at three different stress states. Three coefficients, including dynamic structural stability, structural
variability, and dynamic load structural coefficient were defined that are useful to investigate the dynamic
characteristics of water susceptibility in fine aggregates and skeleton stability of coarse aggregates. It was
also found that dynamic load structural coefficient of unbound aggregates was not sensitive when the con-
fining pressure was more than 150 kPa, It may suggest that the rather lower confining pressure, less than
150 kPa, should be adopted. The outcome of this study showed that it is feasible to evaluate structural
characteristics of unbound aggregates with dynamic load structural coefficient.

1 INTRODUCTION Soil structure is a issue that has attracted many

attentions of the scholars in recent years(Shen Zhu-
The gradation types of pavement base aggregate jiang1996). In 1925, the sensitivity of clayey soil st,
materials are traditionally classified based on the Karl Terzaghi proposed, is used to show the degree
extent to which the air void is filled by the fine of unconfined compressive strength reduction of
aggregates, i.e., skeleton-dense, skeleton-void, undisturbed soil after being disturbed, of which
and suspension-dense types. Such a classification, the essence is, to certain extent, a reflection of the
which is very subjective and qualitative, is not suf- structural characteristics of the clay. For clayey
ficient to characterize structural features of differ- soils, researchers (Xie Ding-yi & Qi Ji-lin 1999, Luo
ent gradation types quantitatively and consistently. Ya-sheng, Xie Ding-yi & Sheng-jun 2004, Hu Zai-
Currently, there does not exist quantitative index qiang, Shen Zhu-jiang & Xie Ding-yi 2000, Hu Zai-
for describing the structural behavior of different qiang, Shen Zhu-jiang & Xie Ding-yi 2004), defined
gradation types under dynamic loads, thus hinder- the soil structural parameters on the basis of com-
ing the accurate and objective evaluation of the pression tests, proposed soil structural param-
aggregate materials structural behavior of pave- eters experiment method and calculated structural
ment base. parameters of the clay under static load through the
The structural characteristics of soils are defined three line method (the ratio of undisturbed satu-
as the concentration of aggregate particles as well rated and disturbed soil compressive strain product
as the size, shape, alignment, and connectivity of between undisturbed soil compressive strain square
air voids among aggregate particles (He Li-hong, under the same stress). For cohesionless soils, lit-
Wang Ren & Zhang Hui-mei 2003). Therefore, the erature (Chen Cun-li, Hu Zai-qiang & Xie Ding-yi
structural characteristics of soils consist of not 2004) used the ratio of deviatoric stress difference
only geometrical features of skeleton aggregate and that generate the same axial strain ε1 in static triaxial
air void but also particles arrangement, in other tests for saturated undisturbed and disturbed soils
words, the structural characteristics of soils include to represent the degree of structural characteristics.
both coarse particles arrangement and fine parti- Pavement materials are always subject to the
cles cohesion properties (Xie Ding-yi, Qi Ji-lin & dynamic wheel loads, therefore the definition
Zhu Yuan-lin 1999). In terms of unbound pave- of structural parameters of unbound aggregate
ment aggregate materials, the arrangement feature materials under dynamic loading should take
of coarse particles is especially pronounced, which into account the change of its using condition
exhibits strong anisotropy depending on the con- and the development of pavement distresses,
tent and property of fine aggregates. and combination with laboratory experiments

79
method. The differences of structural character- 2.2 Test method
istics of unbound aggregate materials in pave-
The structural characteristics of unbound aggre-
ment are essentially attributed to varying particle
gate materials under dynamic load are measured
arrangement, air void composition, etc., which
by a medium-scale dynamic triaxial shear appara-
in turn leads to the differences and changes of
tus (Kong Xian-jing, Zhang Tao, etc., 2005). The
mechanical properties. From this perspective,
specimens were fabricated 20 cm in diameter and
the structural parameters should reflect not only
50 cm in height. In order to achieve the efficiency
the differences mechanical properties but also the
of a vibratory roller compactor, the specimens were
transition of structural characteristics. Accord-
molded and compacted by a vibratory compactor
ing to the two major distress types of asphalt
(Wang Long, Xie Xiao-guang, 2005).
pavements with unbound pavement base layer,
The vibration compaction frequency is 32 Hz
i.e., rutting (PIARC 1995) and fatigue (Garg
with centrifugal force of 10 kN and nominal ampli-
Navneet & Thompson Marshall R 1997), the
tude of 1.2 mm. The specimen was divided into
structural characteristics of unbound aggregates
three layers during compaction, with the compacted
under dynamic load should consist of two parts,
thickness of each layer is 20 cm, 15 cm and 15 cm,
i.e., one representing stiffness change at differ-
respectively. Specimens were mixed at the optimum
ent status and termed as stiffness structure, and
moisture content, the targeted dry density level is
the other representing the change of permanent
98% of the maximum dry density (ASSHTO T180,
strain at different conditions and termed as the
Modified Proctor compaction), three different
deformation structure. The former reflects the
confining pressure levels were applied, i.e., 50 kPa,
stability of aggregate particles arrangement, while
100 kPa and 150 kPa, respectively. To mimic the
the latter represent the deformability of unbound
dynamic vehicular load on pavements, certain
aggregates and the stability of cohesion proper-
dynamic stress with a half-sine wave was applied up
ties of fine aggregates. Those two parts should be
to about 3000 cycles, the load frequency of the half-
integrated in order to realistically characterize the
sine wave is 1 Hz with a loading time of 0.1 seconds
structural behavior of unbound aggregates under
and a rest period of 0.9 seconds.
dynamic load.
The best method to study the soil structural
properties is to destroy the structural integrity thor-
oughly, i.e., fully releasing the structural potential,
2 TEST MATERIALS AND TEST METHOD
which can be achieved essentially through distur-
bance, loading, and saturation. Since unbound
2.1 Test materials
aggregate base materials are in a closed-loop sys-
The unbound aggregate materials are limestone in tem of the pavement, excess water infiltrated into
test, with the gradation shown in Figure 1. Table 1 the pavement structure through surface cracks
summarizes materials basic physical properties. often plays a role of releasing the connectivity and
arrangement characteristics of unbound aggregate
structure under the effect of traffic loads, which fur-
ther causes pavement damage. Therefore, water is
used as a disturbing factor to categorize laboratory
tests into saturated and unsaturated conditions.

I .j::.
~
Vl
;.....
00
particles size /mm
0
w
. 3 STRUCTURAL STABILITY

According to the definition of the soil structure

stability, it characterizes the extent which the soil
structure is destructed. It mainly reflects the cohe-
Figure 1. Particles distribution curve of unbound sion properties of aggregate particles. Therefore,
aggregate materials. subjecting aggregate specimens to water saturation

Table1. Basic physical properties of unbound aggregate specimen.

Fabricating Optimum moisture Maximum dry No uniform Curvature Average particle

method content, % density g/cm3 coefficient Cu coefficient Cc size D50, mm

Vibrating 4.7 2.30 37.5 1.58 16

Fabrication

80
can weaken/dissolve the chemical substance in fine beginning of the loading (N < 1200), the struc-
aggregates and eliminate the cohesion among aggre- tural stability parameter me< 1, indicating strong
gate particles, under the application of dynamic structural stability, water destructed the cohe-
loading, the micro-structural change can be mani- sion among particle interfaces, as expected. For
fested into macro-mechanical change. As for N>1200, the structural stability parameter starts to
unbound aggregates in pavement, there is no exist- increase, i.e., the resilient modulus after saturation
ence of undisturbed and disturbed specimens, hence, is greater than that prior to saturation, indicating
the structural stability index (me) under dynamic the phenomenon of water-hardening.
load can be defined as the ratio of resilient modu- The reason behind this is due to the structural
lus between saturated and unsaturated specimens coupling effect. The initial structural stability of
subjected to the same number of loading cycles with specimens compacted by the vibratory compac-
identical stress magnitude, As shown in Eq. 1. tor is destructed during loading or saturation,
coupled with the generation of secondary struc-
me Ew / E y (1) tural stability, therefore, the structural stability
parameter fluctuates with initial decrease repre-
where: Ey = Resilience modulus of unsaturated senting the destruction of initial structure and
aggregate specimen; Ew = Resilience modulus of subsequent increase representing the generation
saturated aggregate specimen. of secondary structure.
When the structural stability coefficient me < 1,
it means that the stiffness of unbound aggregate
materials decreases after saturation. The less me, is, 1.4
the more sensitive to water of the skeleton stability -;:::
il) 1.2
is, and the more stronger of the structural stability ·u
is. When me = 1, the stiffness of unbound materi- sil)
1.0
0
als is not affected by water, indicating no bonding u
0.8
effect (or cohesion) or in other words, no structural ..0
:.=:l
stability. When me > 1, the materials are moisture- :E 8 0.6
hardening (moisture-hardening means after satu- ~ --<>--- 50kPa
'@ 0.4
---{]- 1OOkPa
rated, materials stiffness increased, in other words, 3u 0.2 --f::r- 150kPa
the negative structural stability), the larger me is,
and the greater of the negative structural stability ~ 0.0
is. 0 500 1000 1500 2000 2500 3000
Table 2 lists all the test result and Figure 2 loading cycles N/time
depicts the relationship between the structural
stability coefficient and the number of load cycle Figure 2. Relation between structural stability coeffi-
times N. It can be seen from Figure 2 that at the cient and load cycles N.

Table 2. Resilient modulus between saturated and unsaturated specimens.

σ3 = 150 kPa σ3 = 100 kPa σ3 = 50 kPa

Loading Ey Ew Ey Ew Ey Ew
cycle times (MPa) (MPa) me (MPa) (MPa) me (MPa) (MPa) me

200 556.5 442.4 0.79 292.7 302.2 1.03 230.0 210.8 0.92
400 516.7 400.3 0.77 330.0 308.0 0.93 249.2 221.3 0.89
600 464.0 363.7 0.78 315.6 290.0 0.92 243.8 234.8 0.96
800 445.3 430.3 0.97 314.7 310.3 0.99 249.1 223.5 0.90
1000 453.3 404.2 0.89 310.5 300.4 0.97 259.6 200.0 0.77
1200 437.9 424.4 0.97 332.6 330.2 0.99 188.1 205.9 1.09
1400 422.8 427.8 1.01 297.2 350.0 1.18 93.9 100.7 1.07
1600 419.9 402.3 0.96 263.7 286.7 1.09
1800 381.8 344.9 0.90 163.7 205.8 1.26
2000 284.2 351.3 1.24 131.0 123.3 0.94
2200 295.7 305.7 1.03
2400 208.1 209.2 1.01
2600 191.3 189.0 0.99

81
4 STRUCTURAL VARIABILITY the plastic deformation is larger at the early stage
of loading. when loading cycle increases, the struc-
The structure variability evaluates the magnitude ture variability coefficient decreases significantly,
of soil deformation induced by the destruction and the decreasing rate changes from a large value
of the original structure. It is an indicator of the to a small value and then gradually tends to be
extent to which aggregate particles are disturbed stable. This means that plastic deformation of
and re-arranged under external disturbing factors saturated specimens changes from being unsta-
such as saturation and dynamic loading. There- ble to being stable under dynamic loading. At the
fore, the structural variability coefficient (mε) is beginning of the loading cycles, the deformation
defined as the ratio of plastic strain between satu- stability curves align successively from top to bot-
rated and unsaturated specimens subjected to the tom in Figure 3 in accordance with the increasing
same number of loading cycles with identical stress confining pressure and extend to a wide range.
magnitude, As shown in Eq. 2. With the advance of loading cycles, the degree of
destruction increases, and the deformation stabil-
mε = ε pw ε ppy (2) ity curves under different confining pressure levels
are distributed in a relatively narrow range. When
N(loading cycles), ≥1200 times, the initial struc-
where: εpw = Plastic strain of saturated aggregate ture of unbound aggregates is close to complete
specimen; εpy = Plastic strain of unsaturated aggre- destruction, and secondary structure starts to gen-
gate specimen. erate, as indicated by the overlapping of structural
When the structure variability coefficient mε < 1, variability curves. This means that unbound aggre-
it means that the plastic deformation of unbound gates have the same residual deformation stability
aggregate materials decreases after water satura-
tion, and that the structure variability is small. 30
When mε = 1, the plastic deformation of unbound 5
·n 25 ---+-- 50kPa
materials is not affected by water, and there is no iE 1OOkPa
particle rearrangement. When mε > 1, it means that (!)
0
(.) 20
---e--
----ts- 150kPa
the plastic deformation of unbound aggregate ..0
materials increases after saturation, and that the ~ w15
·§ s
structural variability is large. The larger mε is, the ro
;>
10
greater the structural variability is, and the more ""§
sensitive to water the materials are. B
(.)

Table 3 summarizes the plastic deformation data ~ 0

obtained under three different confining pressure 2500 3000
-5
levels. Figure 3 illustrates the relationship between loading cycles N/time
the structure variability and the number of loading
cycles. As it can be seen from Figure 3, regardless of Figure 3. Relation between the structure variability
the relative magnitude of the deformation stability, coefficient and load cycles.

Table 3. Deformation between saturated and unsaturated specimens.

σ3 = 150 kPa σ3 = 100 kPa σ3 = 50 kPa

Loading
cycles εpy εpw mε εpy εpw mε εpy εpy mε

0 0.012 0.009 0.74 0.002 0.023 12.45 0.002 0.049 29.56

200 0.061 0.055 0.90 0.027 0.095 3.48 0.034 0.192 5.67
400 0.122 0.143 1.18 0.077 0.187 2.43 0.086 0.315 3.67
600 0.201 0.219 1.09 0.157 0.299 1.90 0.159 0.466 2.93
800 0.306 0.365 1.19 0.286 0.445 1.55 0.257 0.696 2.71
1000 0.425 0.510 1.20 0.446 0.596 1.34 0.623 1.081 1.73
1200 0.554 0.656 1.18 0.650 0.766 1.18 2.229 2.176 0.98
1400 0.749 0.854 1.14 1.060 1.051 0.99 5.086 4.504 0.89
1600 1.055 1.174 1.11 1.945 1.775 0.91
1800 1.606 1.759 1.10 4.086 4.365 1.07
2000 2.619 2.631 1.00 5.824 8.590 1.47
2200 4.544 4.122 0.91
2400 8.472 6.829 0.81
2600 11.795 11.185 0.95

82
under different confining pressure, and that the σd = deviator stress of dynamic triaxial test, or (σ1–
ultimate level of deformation stability is independ- σ3), where: σ1 = axial stress, σ3 = confining pressure.
ent with the initial deformation stability level. As can be seen from Equation 3, the structural
As the specimen gradation, physical state and fab- coefficient (mN) under dynamic load is the ratio of
ricating method are the same under three different the product of plastic strain εpw and elastic strain
confining pressure levels, therefore, it can be seen εew of saturated specimen to the product of plastic
from Fig. 3 that the confining pressure has a great strain εpy and elastic strain εey of unsaturated speci-
influence on the structural variability of unbound men at the same number of loading cycles, if plastic
aggregates materials. When the confining pressure strain εpw and elastic strain εew of saturated specimen
is low, the structural variability coefficient mainly of saturated specimen increase, dynamic load struc-
reflects the deformation stability properties (perma- tural coefficient becomes large, the structural of
nent deformation behavior) of unbound aggregates unbound aggregate materials is strong, conversely,
materials. on the other hand, when the confining when integrated dynamic load structural coefficient
pressure increases, the additional structure generated becomes small, the structural is weak.
by the confining pressure is enhanced, and the sen- If the εpw becomes larger and εew becomes small,
sitivity of structural variability coefficient decreases. or the εpw becomes small and εew becomes larger,
Under such circumstance, the structural variability but the product of these two parameters remains
coefficient reflects both initial structural of speci- the same, according formula 3, the structural coef-
mens and the additional structural caused by con- ficient (mN) is still the same, hence, the structural
fining pressure. When the confining pressure further behavior is the same too. The changes of both
increases (σ3 = 150 kPa), the specimen initial struc- plastic and elastic deformation are involved in the
tural will be suppressed, at this moment, the struc- structural coefficient under dynamic load.
tural parameters only characterize the additional The scope mN can be defined like this, when
structure. Therefore, lower confining pressure level MN > 1, specimen has a positive structural property.
should be used for measuring structural parameters MN = 1, specimen has no structural property.
of unbound aggregate materials. MN < 1, specimen has negative structural property.
Figure 4 illustrates the relationship between the
structural coefficient and the loading cycles N under
5 DYNAMIC LOAD STRUCTURAL different confining pressure levels, which reflects
COEFFICIENT OF UNBOUND stability and variability of the unbound aggregate
AGGREGATE MATERIALS materials. When the load cycle N < 1200, the scat-
tering of the curves shows that the structural coef-
The aforementioned analysis of the structural ficient is sensitive to confining pressure, when load
stability and the structural variability are the two time N > 1200, the clustering of curves shows the
aspects of the structural characteristic of unbound uniformity of the damage state of unbound aggre-
aggregate materials, to develop a coefficient for gate materials. The fluctuation of curves reflects the
structural characteristic, these two aspects should evolution of the initial and secondary structures.
be integrated so that the contribution of both par- Broadly speaking, the method presented in this
ticle cohesion/connectivity and particle arrange- paper can be adopted to evaluate the structural prop-
ment properties to structural behavior can be taken erty of unbound aggregate materials with different
into account. Therefore, the structural coefficient dynamic load testing and varying disturbing condi-
of unbound aggregates under dynamic load is tions, the structural coefficient of unbound aggre-
defined as the ratio of structural variability to the gate materials can not only reflects the structural
structural stability, as shown in Eq. 3.

ε ppw σd 7.0
mε ε py ε ppw E y ε ppw ε ey ε pwε ew 6.0
-'>- SOkPa
mN = = = = = (3) --fr- 1OOkPa
me Ew ε ppy Ew ε ppy σ d ε pyε ey
Ey ε ew "
0
Tl
tt:
5.0
4.0
---fr- 150kPa

"0 § 3.0
"
e
where: mN = structural coefficient of unbound .: 2.0
aggregate materials; εpw = plastic strain of saturated (.)
1.0
aggregate specimen; εpy = plastic strain of unsatu- ~
0.0
rated aggregate specimen; Ey = resilience modulus of 0 500 1000 1500 2000 2500 3000
unsaturated aggregate specimen; Ey = resilience mod- loading cycles N/time
ulus of saturated aggregate specimen; εew = resilience
strain of saturated aggregate specimen; εey = resil- Figure 4. Relation between structural coefficient and
ience strain of unsaturated aggregate specimen; and load cycles.

83
Table 4. Structural coefficient in different confining ACKNOWLEDGMENTS
pressures.
Thanks to Ministry of Transportation Ji Lin prov-
Loading ince to provide financial support, thanks to the
cycles σ3 = 150 kPa σ3 = 100 kPa σ3 = 50 kPa
guidance of Professor Erol Tutumluer of UIUC,
200 1.13 3.37 6.18 thanks to doctoral candidate Yuan-Jie Xiao of
400 1.52 2.60 4.13 UIUC selfless help.
600 1.39 2.07 3.04
800 1.23 1.58 3.02
1000 1.35 1.38 2.25 REFERENCES
1200 1.22 1.19 0.89
1400 1.13 0.84 0.83 Chen Cun-li, Huzai-qiang & Xie Ding-yi. 2004. Research
1600 1.16 0.84 on the relationship between deformation-strength
1800 1.21 0.85 property and structure of red mud[J]. Rock and Soil
Mechanics, 25(12):1862–1866.
2000 0.81 1.06
Garg Navneet, Thompson Marshall R. 1997. Triaxial
2200 0.88 characterization of Minnesota road research project
2400 0.80 granular materials[J]. Transportation Research Record
2600 0.96 1577: 27–36.
He Li-hong, Wang Ren & Zhang Hui-mei. 2003. The
status quo of research methods of soil structural
characteristics[J]. Soil Engineering and Foundation,
17(3): 84–86.
Hu Zai-qiang, Shen Zhu-jiang & Xie Ding-yi. 2000.
stability and structural variability but also the effect Research on structural behavior of unsaturated
of test conditions on those properties, those coeffi- loess[J]. Chinese Journal of Rock Mechanics and Engi-
cients have good stability, sensitivity, operability and neering, 19(6):775–779.
applicability. Hu Zai-qiang, Shen Zhu-jiang & Xie Ding-yi. 2004.
Deformation properties of structural loess[J]. Chinese
Journal of Rock Mechanics and Engineering, 23(24):
6 CONCLUSION 4142–4146.
Kong Xian-jing, Zhang Tao, Zou De-gao, et al. 2005.
Development of medium scale dynamic triaxial appa-
According to analysis results of the modulus and ratus and application of small strain measurement
permanent deformation of saturated and unsatu- technique [J]. Journal of Dalian University of Technol-
rated unbound aggregate specimens subjected to the ogy, 45(1):79–84.
same number of loading cycles during dynamic tri- Luo Ya-sheng, Xie Ding-yi, & Shao Sheng-jun. 2004.
axial test, the following conclusions can be reached: Variation characteristics of soil structure of unsatu-
rated loess[J] Journal of Northwest Sci-Tech Univ of
1. Unbound aggregate materials have the struc- Agri and For (Natural Science Edition), 2004(8):
tural stability and variability under the dynamic 115–119.
load, the methods for measuring and calculating PIARC. 1995. Technical Committee on Flexible Roads.
the structural coefficients are proposed, which Bituminous Materials with a High Resistance to Flow
achieves the objective of quantitatively assessing Rutting[M].
the structural behavior of unbound aggregates. Shen Zhu-jiang. 1996. Mathematic model of soil struc-
2. Skeleton structure performance can be reflected ture-the key question of mechanics in 21 century[J].
by the structural variability parameter, while the Chinese Journal of Geotechnical Engineering,
18(1):95–97.
performance of filling materials can be char- Terzaghi K., Peck R.B. & Mesri G. 1996. Soil mechanics
acterized by the structural stability parameter, in engineering practice[M]. A Wiley-Inter science Pub-
those two parameters are two aspects of aggre- lication. Third edition. John Wiley and Sons, 127–133.
gates structural characteristic, when confining Wang Long, Xie Xiao-guang. 2005. Research on the
pressure σ3 150 kPa, the unbound aggregate vibrating compaction method of graded aggregate
materials structural behavior is suppressed. material[J]. Journal of Highway and Transportation
therefore, the structural parameters should be Research and Development, 22(7): 26–30.
measured under a small confining pressure level. Xie Ding-yi, Qi Ji-lin & Zhu Yuan-lin. 1999. Soil struc-
3. It is rational to use the dynamic load structural ture parameter and its relations to deformation
and strength[J]. Journal of Hydraulic Engineering,
coefficient to evaluate structure characteristic 1999(10):1–6.
of unbound aggregate materials, it integrally Xie Ding-yi, Qi Ji-lin. 1999. Soil structure characteris-
reflects the development of stiffness and defor- tics and new approach in research on its quantitative
mation and is related to fatigue and rutting parameter[J]. Chinese Journal of Geotechnical Engi-
damage of asphalt pavements. neering, 21(6): 651–656.

84
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Investigation of resilient modulus values for base aggregates

B. Cliatt
Laboratory of Pavement Engineering, National Technical University of Athens (NTUA), Athens, Greece

A. Loizos
Department of Transportation Planning and Engineering, National Technical University of Athens (NTUA),
Athens, Greece

ABSTRACT: Characterization of base aggregates layers that underlay the bound layers within a pave-
ment system is vital to correctly design road pavement structures. One method to determine these char-
acteristics is resilient modulus testing which is becoming a more common method that provides a more
complete overview of the aggregate behavior under a variety of loading and confining conditions that
may be present in a road pavement structure. The present paper present the results of the investigation of
a variety of aggregates, with variations in gradations and moisture contents for a typical Greek aggregate
utilized in base layers of flexible pavements.

1 BACKGROUND AND OBJECTIVES The resilient modulus can be defined as,

Characterization of base aggregates layers that Mr = σ1 - σ3 / εr = σd/εr (1)

underlay the bound layers within a pavement
system is vital to correctly design road pavement where Mr is the resilient modulus (kPa), σ1 is the
structures. One method to determine these char- major stress (kPa), σ3 is the minor stress (kPa),
acteristics is resilient modulus testing which is σd = deviatoric stress (kPa), and εr = recoverable
becoming a more common method that provides axial strain.
a more complete overview of the aggregate behav- The resilient modulus is utilized to characterize
ior under a variety of loading and confining con- the mechanical response of unbound materials in
ditions that may be present in a road pavement the pavement structure. It is essentially a measure
structure. For the unbound granular layers of a of stiffness suitable for usage under variations in
pavement a main structural function of the mate- conditions such as stress level changes, density
rials is to distribute stresses from the overlying changes and responses under various moisture
bound pavement layers onto the underlying sub- conditions.
grade layers. Current pavement design processes Figure 1 below provide an typical illustration
for many agencies include the utilization of static of the generated stresses/strains during a loading
data, for example the California Bearing Ratio cycle for resilient modulus testing in the laboratory.
(CBR) which provides limited data concerning the Figure 2 below illustrates the applied stress, the
behavior of the material under investigation. In deviator stress σd and the difference between the
order to provide more information concerning the
behavior of the materials laboratory characteriza-
tion of the unbound granular base layers need to
be more accurately simulated through laboratory
testing procedures. The unbound granular material
behavior for a pavement design and analysis can ,Permanen}
• slraJn •
be determined from repeated load triaxial testing
on prepared specimens. This testing provides infor-
mation for material characterization through the
Resilient Modulus (Mr) of the material is becom-
ing more utilized and valuable for pavement design
and analysis. Thompson (1998) provides a good Strain
overview of the approaches and material charac-
terization models that attempt to describe the resil- Figure 1. Strains in granular material during load
ient behavior of granular materials. application.

85
 produced a material with significantly improved
resistance to rutting and a slight increase in load
spreading capability. More recent research from
Hossain (2015) in contrast stated that aggregate
gradation did not show any definitive trend in resil-
ient modulus value based on the gradations tested.
In regards to moisture content Lekarp et al (2000)
discussed the moisture content relationship con-
cluding that it is generally agreed that that resilient
response of dry and most partially saturated granu-
lar materials is similar. Bilodeau et al (2012) exam-
ined the issue of water sensitivity of base aggregates
in relationship to the degree of saturation conclud-
ing a significant reduction of modulus as the mate-
Time(s) rial approaches saturation. Other research into the
resilient modulus and gradation of base aggregates
include amongst others Austin (2009), Bilodeau
Figure 2. Stress/Strains in granular material during
(2011), Khoury (2007), Kim et al (2012),
load application.
The Mr values for a new pavement design can
be determined from repeated load triaxial testing
recoverable resilient strain and permanent plastic on prepared specimens. The determined values and
strain on the material with time (s) for a given load- their usage in the analysis and design have signifi-
ing cycle of the material under investigation. cant influence on projected pavement performance.
The quality of the material used as base road Proper characterization of unbound materials is
construction can have a significant impact on the important. Currently for road new projects and
performance and life of the pavement it is intended motorways in Greece they are generally designed
to support Gandara et al (2005). From this it can be with static values (ex. CBR).
seen that the proper selection of materials is vitally The intent of the current study was to investi-
important for future road performance. Lekarp et al gate the materials for their eventual usage in the
(2000) outlined the factors that with vary degrees of design processes. The modulus of the unbound
importance affect the resilient behavior of unbound layer is vital to know during design and can signifi-
granular material. These factors include amongst cantly affect the overall final pavement design of
others, the effect of stress, the effect of density, effect a motorway. The AASHTO T307 test for Deter-
of grading, fines content, and moisture content. mining the Resilient Modulus of Soils and Aggre-
According to Puppala (2008) only a limited gate Materials was utilized for the investigation
number of studies were found for unbound base of the Resilient Modulus. Limited information is
aggregate material modulus values. In regards to currently available concerning Resilient Modulus
gradation some of these studies are outlined as values of various gradations of aggregate utilized
follows. Hopkins et al (2007) performed on several in Greece for unbound pavements layers. The pri-
different types of aggregates to establish and effec- mary objectives of the investigation were to deter-
tive means of predicting resilient modulus of gran- mine modulus values of typical unbound granular
ular materials. Conclusions included that the Mr aggregates utilized in Greece at various moisture
of well-graded aggregates increase with dry den- conditions. In addition to the moisture content the
sity with more noticeable improvements for well gradation of the material was altered to be at the
grade aggregates. They also recommend to study upper and lower limits for grains sizes less 2 mm
the effect of different gradations of base materi- and greater 0.063 mm to determine the effects of
als to determine its influence. Gandara et al (2005) gradation The current investigation examines a
investigated the effect of gradation of unbound limited number of aggregate gradations in Greece
granular materials. The research found there was to determine the Resilient modulus values that can
a limit to the mount of fines (passing 200 sieve) of be expected both at optimum moisture content
approximately 10% above which material behavior (OMC), and values above and below the OMC.
was adversely affected. Uthus (2005) investigated
the grain size distribution with the fuller curve
2 MATERIALS TESTED
for n = 0.5 and n = 0.35 concluding as had Ekblad
(2007) that materials with high grading coefficients
2.1 Basic information
are less affected by a change in water content.
Brown et al (1996) conclude the selection of an The basic material tested was to be utilized for base
optimum grading to maximize compacted density layer construction of a to be built motorway. From

86
this material the other two material gradations 2.5 Optimum moisture content
were derived in the laboratory by sieving the mate-
The determination of the optimum moisture
rials and removing the required materials to create
content for compaction of the mentioned mate-
gradations A and C. The gradations A and C were
rials was carried out by utilizing an appropriate
designed to be near or at the upper and lower lim-
testing standard. Samples of the materials were
its for grain sizes less 2.00 mm.
compacted using modified proctor procedures
with compaction mold dimensions of 120 mm for
2.2 Aggregate sampling height and 150 mm for the diameter. The hammer
weight was equivalent to 4.5 kg and was dropped
Samples of the aggregate were obtained from from a height to fall distance of 457 mm. The
a section of a motorway under construction. materials were compacted in five separate lay-
The material was sample according to EN932- ers with 56 distributed blows per layer as per the
1/1997. The obtained aggregates meet the cur- related standard.
rent standard for usage in base and sub-base The materials were compacted at a variety of
layers for motorway construction according to moisture contents for each of the three materials in
the relevant Greek standard (Hellenic Technical order to determine the corresponding relationships
Specifications 1501/2009—Type I gradation). between moisture content and dry density of the
Approximately 800 kg of bulk material samples materials. Figure 4 below shows the gradation of
were sourced from the under construction base the materials investigated.
layer of the motorway. The materials were test- The optimum moistures contents for the three
ing for gradation (EN 933-1/2012), maximum materials are 6% for material A, 6.2% for material
dry density and moisture density relationship B and 7.2% for material C.
(EN13282-2/2010). The maximum dry density for material A is
2162 kg/m3, for material B 2248 kg/m3 and for
2.3 Classification material C 2280 kg/m3.
According to Unified Soil Classification System
(USCS) all materials are well-graded gravel (GW)
and according to AASHTO classification they are
classified as A-1-a materials. 4.5 Kg Hammer
457 rnm Drop height
5 Material layers
2.4 Gradation 56 Blows per layer
A sieve analysis was performed in accordance with
EN 933-1/2012 on the three materials and the gra-
dation curves are shown in Table 1. All three mate-
rials meet basic Greek requirements for gradation
including S4mm ≥ S2mm +6, S2mm ≥ S1mm +6, & S1 ≥
1.8* +S0.063mm.

Table 1. Material gradation. Figure 3. Modified proctor compaction.

Greek standard—
A B C Type 1

Sieve
(mm) % Passing Upper % Lower %

40 100 100 100 100 100

31.5 100 97.4 97.9 99 85
16 76.8 80.4 84.1 85 55
8.0 56.9 63.6 65.0 65 35
4.00 34.6 44.7 49.2 50 25
2.00 18.5 31.1 40.1 40.5 19
1.00 12.0 21.0 29.8 31 12 M OIS T UR E CO NTENT %
X B + C
0.500 8.1 14.3 20.3 24 8
0.063 0.9 1.4 2.0 11 0
Figure 4. Moisture content versus dry density.

87
3 RESEARCH METHODOLOGY testing is begun and to minimize any contact issues
between the base plate, test specimen and stone
3.1 Research approach platens. The remainder of the testing cycle consist
of 15 load sequences as per the standard and the
Samples from the is-situ obtained material and
loading sequence is shown in Table 2.
the laboratory sieved materials were prepared in
The load (Fig. 5) for each sequence is a
the laboratory at the various moisture contents in
haversine load pulse with a 0.1 second duration
order to preliminarily test the various materials and
and a 0.9 second rest period. In total 100 cycles
the effect moisture content and gradation exert on
are performed per sequence with the final 5 cycles
the resilient modulus of the materials under inves-
utilized to determine the resilient modulus. Only
tigation. The materials were prepared and tested
the final 5 values were utilized to ensure the mate-
immediately after preparation to ensure moisture
rial adjusted to each sequence of confining and
conditions were regulated.
axial stresses.
Drainage valves were open to atmospheric pres-
3.2 Sample preparation sure to simulate drained conditions during the tests.
The material to be prepared for testing was dried, After the tests were completed collected data
the appropriate water was added to achieve the was saved and available for analysis, Data included
desired moisture content and was thoroughly
kneaded to ensure complete mixture of the water Table 2. Test sequence for material.
and aggregate. The samples mixture was placed in
a double sealed plastic bag, to ensure no moisture Max
escaping and was allowed to cure for 16-24 hours. Confining axial Cyclic Constant No.
The material was removed and immediately avail- pressure stress stress stress of
able for sample preparation. (kPa) (kPa) (kPa) (kPa) loads
The sample material for compaction was uti-
lized to prepare samples for resilient modulus (Mr) 0 103.4 103.4 93.1 10.3 1000
testing with a vibratory compaction device imple- 1 20.7 20.7 18.6 2.1 100
2 20.7 41.4 37.3 4.1 100
mented for the compaction of the materials. The
3 20.7 62.1 55.9 6.2 100
samples were compacted in a split mold, lined with
4 34.5 34.5 31 3.5 100
a rubber membrane in 6 near equivalent lifts of
5 34.5 68.9 62 6.9 100
approximately 50 mm. The prepared sample was
6 34.5 103.4 93.1 10.3 100
encased with porous stone platens both on the bot-
7 68.9 68.9 62 6.9 100
tom and top of the sample. After the compaction 8 68.9 137.9 124.1 13.8 100
of each lift to the calculated height, the upper sur- 9 68.9 206.8 186.1 20.7 100
face of the lift was scored with a spatula to ensure 10 103.4 68.9 62 6.9 100
cohesion between each successive lift. The final 11 103.4 103.4 93.1 10.3 100
dimensions of the prepared samples were 151 mm 12 103.4 206.8 186.1 20.7 100
in diameter by 300–305 mm in height. All samples 13 137.9 103.4 93.1 10.3 100
were compacted to above 95% of the calculated 14 137.9 137.9 124.1 13.8 100
proctor density required for the sample. 15 137.9 275.8 248.2 27.6 100

3.3 Sample testing

The prepared samples available to be tested in the
Mr testing apparatus were removed from the split
mold ensuring the rubber membrane was intact Haversine Load Pulse
and then placed immediately within the Triaxial
chamber. Two external Linear Variable Transduc-
ers (LVDT’s) placed on the sample were utilized to
measure the axial deformation during testing with
the resilient modulus being based upon the average 0 _1 sec Load Duration

of the two LVDT’s.

The testing sequence for unbound granular base
materials consisted of an initial loading sequence
of the specimen with a confining pressure of 0.9 sec Rest Period
103.4 kPa and a vertical stress of 103.4 kPa and Time (s)
1000 loads were applied. The initial sequence is
to ensure to proper seating of the material before Figure 5. Load cycle.

88
at minimum: load data, axial stress, cyclic stress,
contact stress, recoverable deformation of the
LVDT’s, specimen height and diameter, strain and
calculate modulus.

4 TESTING RESULTS AND ANALYSIS

In total 36 Mr samples were prepared, 12 for each

aggregate gradations and for each aggregate grada-
tion 4 samples sample were prepared at each of the
50.0 +-- --t---+---+----+---+-----1
3 moisture conditions (OMC+2, OMC, OMC-2).
100 200 300 400 500 600 700
The result for the resilient modulus testing of BuU• Slress (KJta)
the three gradation of aggregates (A, B, C) and • OMC+2 • OMC
at the three moisture conditions (OMC-2, OMC, & OMC-2 - - - Pow~· (OM C+ 2)
- -Power O!vlC - - Power OMC-2
OMC+2) are presented in the following section.
Overall for all tested samples the trend was to
Figure 7. Gradation B—modulus versus bulk stress.
increase the modulus with increase bulk stress. This
is the expected behavior of the unbound materi-
als and is consistent with previous studies. This
indicates the aggregates are stress dependent and
stress hardening. The follow sections present the
results of the laboratory determined modulus in
relationship to the bulk stress, deviator stress and
the moisture conditions for the tested aggregates.

4.1 Modulus versus bulk stress

Figures 6–9 present the results of the tested sam-
ples for the laboratory determined resilient modu-
lus of the three gradations (A,B,C) and the three
moisture conditions tested (OMC+2, OMC, OMC-
2). The results presented in the figures illustrate Bulk Slress (Kpa)
the influence of the bulk stress on the determined • OMC+2 • OMC
OMC-2 - - - Pow~· (OMC+2)
resilient modulus. The results overall indicated &
- -Power WvfC - - Power o:MC-2
increases with the bulk stress. The three gradations
each had variations in regards to the moisture con- Figure 8. Gradation C—modulus versus bulk stress.
ditions producing the largest modulus values.

100 200 300 400 500 600 700

50.0 +'IIIL----+---+-- -+---+-------1
Bulk St ress (Kpa)
100 200 300 400 500 600 700 + OMC-A • OMC-B
Bu1k Slress (K1~<1) & OMC-C ----- Power (OMC-A)
• OMC+2 • OMC
& OMC-2 - - - Pow~ (OMC+2) - - Power (OMC-6) - - - Power (OMC-C)
- -Powu· OfvlC - -Power OMC-2

Figure 9. Gradations A,B,C at optimum moisture

Figure 6. Gradation A—modulus versus bulk stress. conditions.

89
 For gradation A at the lower bulk stresses the Mr test sequence, which represent the maximum
modulus at optimum moisture conditions was axial loading at each of the 5 confining pressures
lower than both OMC+2 and OMC-2. While at of the protocol are shown. The legend below is for
higher stresses the value was slightly larger at opti- Figures 10–12.
mum moisture. For gradation A at the lowest confining pressure
For gradation B at the lower bulk stresses the of 20.7 KPa the modulus is higher both above and
modulus at the optimum moisture conditions was below the OMC. OMC +2 and OMC-2 were 24%
higher than both OMC+2 and OMC-2. While and 35% greater respectively than the values deter-
at higher stresses the value was slightly larger mined at OMC. While for cycles 9 and 12 mini-
at optimum moisture. For gradation B OMC+2 mal difference was observed (less than 5 percent)
exhibited the lowest modulus for all bulk stresses. between the two cycles for the determined modulus
For gradation C at the lower bulk stresses the values. The influence of the moisture condition is
modulus at the optimum moisture conditions was limited for cycles 9 and 12. These similar modulus
higher than both OMC+2 and OMC-2. While at values indicate the axial stress is more influential
higher stresses the value was slightly larger at opti- than the confining pressures.
mum moisture. For gradation B OMC-2 exhibited For gradation B at the lowest confining pressure
the lowest modulus for all bulk stresses. of 20.7 KPa the modulus is lower both above and
Overall the three gradations exhibited signifi- below the OMC. While for cycles 9 and 12 minimal
cant variations in their determined resilient modu- difference was observed between the two cycles
lus in regards to the bulk stress.
From Figure 9 above it can be seen that grada- - -3 - -6 - -9 - -1 2 - - 15
tion B produced the highest resilient modulus val-
ues at lower bulk stresses. Gradation A produces 500.0
the lowest modulus values for all bulk stresses. ,..., f--
~
As the bulk stresses reached their maximum for ~

the test protocol the modulus values approached ~

similar values for all three of the gradations. For '
=
the lowest bulk stress of 87 KPa (cycle 1) for Gra- '6
dation A the resilient modulus was more than ~
50 percent below gradation B. Gradation A was
2..
a much coarse gradation containing less mate- ~
=
~

----------
rial passing the sieves 2.00 mm and below. This ~

absence of finer materials may have been the cause

of this lower values at lower stresses. Gradation C 5:J
was a more fine gradation with higher percentages OMC+2% OMC OMU%
passing the 2.00 mm sieves and below. For the low- MoishueConient
est bulk stress of 87 KPa (cycle 1) for Gradation C
the resilient modulus was approximate 18 percent Figure 10. Gradation A—moisture content versus dry
below gradation B. These large variations indicate density.
an effect of the gradation at the lower confining
pressures. 500.0
On the opposite end of the spectrum at the I--
highest bulk stress of 689 KPa the values of Gra- ,...,
~ I--
~
dation A versus B were 10% lower. Additionally for
Gradation C versus B at 689 KPa bulk stress the ~ I--
value were only 5% lower. The variations in grada-
tion have a more limited affect with increasing bulk
':
]
~ 1--
f-
=-
stress, possibly due to aggregate interlock material 2.. ~
with increased axial stresses.

4.2 Moisture content versus confining stress

Figures 10–13 present the results of the tested
samples for the laboratory determined resilient
:
~
'&:

5:0 I
~
OMC+2% OMC
----
MoishucContent
OMU%
modulus of the three gradations (A,B,C) and the
three moisture conditions tested (OMC+2, OMC,
OMC-2). In order to illustrate the influence of the Figure 11. Gradation B—moisture content versus dry
moisture conditions, 5 cycles (3,6,9,12,15) of the density.

90
 500
I-
""'•
r-
6"' r---
~

]=
~ I-
~

]=
- -----
----------
·~
~
•
so 100 150 200 250 300
OMC+2% OMC OMC-2% Deviator Stress (<HI)
MoistureContent + OMC-A • OMC-B
"- OMC.C --- - - Poly.(OMC·A)
- - Poly. (OMC-B) - - - Poly. (OMC-q

Figure 12. Gradation C—moisture content versus dry

density. Figure 14. Gradations A,B,C modulus versus deviator
stress σd.
_soo.o

f
~ modulus of the three gradations (A,B,C) for the
~ •- - ---=-- =--=--:.=--=--= -~- -=--=--::.:.
~ OMC. The results presented in the figure illustrate
=
-o the influence of the deviator stress on the deter-
:E mined resilient modulus. The results overall indi-
c
cated increases with the deviator stress.
J!~ 50.0
Gradation B exhibited the largest resilient
140tsture Co ntent
----- A - - B -- - C
modulus values over the entire range of deviator
stresses in the test protocol. Gradation C exhib-
ited the second largest modulus values, while
Figure 13. Modulus at maximum bulk stress. gradation C exhibited the lowest overall modulus
values for the entire range of the test protocol.
for the determined modulus values. The influence The variation in the modulus for the lowest devia-
of the moisture conditions is limited for cycles 9 tor stress of 21 KPa of gradation C to B is equal
and 12 below at OMC-2 is significantly lower than to 19 percent while the variation in the modulus
the other two moisture conditions. OMC-2 values for the lowest deviator stress of 21 KPa of grada-
were 13 and 15% lower than those at the OMC. tion A to B is equal to 54 percent. The variation
While for OMC-2 the values for cycle 9 and 12 in the modulus for the largest deviator stress of
were respectively 3 and 4% lower. These similar 276 KPa of gradation C to B is equal to 3 percent
modulus values for cycles 9 and 12 which have while the variation in the modulus for the lowest
confining pressures of 68.9 KPa and 103.7 kPa deviator stress of 276 KPa of gradation A to B is
indicate overall for the specimen the axial stress equal to 12 percent. There are significant differ-
applied is more influential than the applied confin- ence at lower deviator stresses for gradation A.
ing pressure. The lack of material in the lower sieve sizes nega-
For gradation C at the lowest confining pressure tively affect the modulus of the material at low
of 20.7 KPa the modulus is lower both above and deviator stress. On the other hand at higher devia-
below the OMC. While for cycles 9 and 12 minimal tor stresses this effect in greatly reduced, perhaps
difference was observed between the two cycles for due to the aggregate interlock of the larger graded
the determined modulus values. The influence of materials in the mixture.
the moisture conditions is limited for cycles 9 and
12 below at OMC+2 is significantly lower than the
other two moisture conditions. These similar mod- 5 DISCUSSION AND CONCLUSION
ulus values indicate the axial stress is more influen-
tial than the confining pressure. Characterization of base aggregates layers that
underlay the bound layers within a pavement
system is vital to correctly design road pavement
4.3 Modulus versus deviator stress
structures. The present paper presented the results
Figure 14 below present the results of the tested of the investigation of a variety of aggregates, with
samples for the laboratory determined resilient variations in gradations and moisture contents

91
for a typical Greek aggregate utilized in base lay- Overall gradation B near the midrange of the
ers of flexible pavements. In total 36 Mr samples Greek standard for base aggregate in pavement
were prepared, 12 for each aggregate gradations construction produced the highest laboratory
and for each aggregate gradation 4 samples sample determined modulus values.
were prepared at each of the 3 moisture conditions
(OMC+2, OMC, OMC-2). The AASHTO T307
test for Determining the Resilient Modulus of REFERENCES
Soils and Aggregate Materials was utilized for the
AASHTO T307. Standard Method of Test for Deter-
investigation of the Resilient Modulus. Limited mining the Resilient Modulus of Soils and Aggregate
information is currently available concerning Materials, American Association of State Highway and
Resilient Modulus values of various gradations Transportation Officials, Washington, DC, USA.
of aggregate utilized in Greece for unbound pave- Austin, A. 2009. Fundamental characterization of
ments layers and goal of the paper was to provide unbound base course material under cyclic loading,
more information on the subject. PhD Dissertation, Louisiana Tech University, USA.
The results from the aggregates of the labora- Bilodeau, J.P. & Dore, G. 2011. Influence of moisture
tory determined modulus values were compared content on the resilient modulus of base granular
against the bulk stress, deviator stress and moisture materials, Conference of Transportation Association of
Canada.
content of the specimens. In regards to the bulk Bilodeau, J.P. & Dore, G. 2012. Water Sensitivity of resil-
stress versus the modulus values each of the aggre- ient modulus of compacted unbound granular mate-
gate gradations exhibited different behavior. For rials used as base pavement, International Journal of
aggregate B and C the optimum moisture condi- Pavement Engineering 13:5 pp. 459–471.
tions produced the largest corresponding modulus Brown, S. & Chan, F.W.K.. 1996. Reduced rutting in
values in comparison to the other moisture condi- unbound granular pavement layers through improved
tions tested. This is not however true for gradation grading design, Proceedings of the Institution of Civil
A in which the largest modulus values depended Engineers: Transport pp. 40–49.
both on the bulk stress and the moisture condi- Ekblad, J. 2007. Influence of water on coarse granular
road material properties, PhD Dissertation, Royal
tions combined. This variation may be caused by Institute of Technology, Stockholm, Sweden.
the lower amount of material in the smaller sieves Gandara, J., Kancherla, A., Alvarado, G., Nazarian, S. &
sizes. In regards to the bulk stress of the three Scullion, T. 2005. Impact of aggregate gradation
tested gradations at optimum moisture content on base material performance, Texas Department of
gradation B produced the highest modulus values Transportation Research Report 4359–2.
for the lowest to highest bulk stresses. However as Hopkins, T., Beckham, T. & Sun, C. 2007. Resilient
the bulk stress increased all gradation approached modulus of compacted crushed stone aggregate base,
similar modulus values perhaps indicating aggre- Kentucky Transportation Center, Research report
gate interlock playing a larger role as the material KTC-05-27.
Hossain, S. & Lane, S. 2015. Resilient modulus values for
was more compacted. base aggregate from Virginia, Transportation Research
The three material were also tested in regards Board 94th Annual Meeting.
to the moisture conditions OMC, OMC+2 and Khoury, N., Zaman, M. & Laguros, J. 2007. Optimizing
OMC-2. Gradation A at lower stresses indicated a Aggregate Base Gradation for Maximum Perform-
lower modulus value at OMC in comparison to the ance, Proceedings of the International Conference:
other moisture conditions. This behavior should Advanced Characterization of Pavement and Soil
be further investigated. For all gradations at cycles Engineering Materials.
12 and 15 of the protocol the material had similar Kim, S., Gopalalakrishnan, K. & Ceylan, H. 2012.
modulus values indicating perhaps that the axial Unbound material characterization with Nottingham
asphalt tester, Proceedings of the Institution of Civil
stress was playing a larger role in the determina- Engineers—Construction Materials, Volume 165 Issue
tion of the modulus values. At the highest stress 6 December 2012, pp. 355–365.
conditions in cycle 15 at OMC+2 all gradations in Lekarp, F., Isacsson, U. & Dawson, A. 2000. State of
regards to the determined modulus were similar, the Art I: Resilient Response of Unbound Aggregate,
with the difference increasing as all the gradation American Society of Civil Engineer, Volume 126 No1
approached OMC-2. pp. 66–75.
In regards to the materials tested versus the Puppala, A. 2008. Estimating stiffness of subgrade and
deviator stress gradation B, produced the largest unbound materials for pavement design, NCHRP
modulus values for the entire range of deviator Synthesis 382, Transportation Research Board,
Washington, D.C.
stresses. Gradation C with the largest percentage Uthus, L., Hoff, I. & Horvli 2005. A study on the influ-
of materials in the 2.00 mm–0.063 mm range pro- ence of water and fines on the deformation properties
duced the next largest values and close to grada- of unbound aggregate, Proceedings of the 7th Inter-
tion B. While gradation A produce significantly national Conference on the bearing Capacity of Roads,
lower modulus values. Railways and Airfields.

92
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Unsaturated resilient strain behaviours of a granular material

P. Jing, C. Chazallon & H. Nowamooz

Laboratoire des sciences de l’ingénieur, de l’informatique et de l’imagerie (ICube) UMR7357,
CNRS, INSA de Strasbourg, Strasbourg Cedex, France

ABSTRACT: Road pavement structures are generally composed of unsaturated granular materials. The
influence of the fine content is significant as well as unsaturated state on the resilient strain behaviour of
granular materials for pavements. In this paper, based on the soil water retention curves and repeated load
triaxial tests for a granular material with three fine contents, the simple exponential function relationship
between resilient strain behaviours and s/s* values (suction) could be observed, which can reduce the
number of tests required to determine the unsaturated resilient strain behaviours of this kind of granular
material.

1 INTRODUCTION granular material is then studied by a series of

Repeated Load Triaxial Tests (RLTT).
Granular materials are usually used in low traffic
pavements as base layer or sub-base layer. During
the service life, low-traffic pavements are subjected 2 STUDIED MATERIAL
to variable hydraulic and mechanical impacts,
which have a significant influence on mechanical The studied granular material is the Missillac fine
behaviours (both resilient and permanent strain sand. It is an alluvial sand coming from the quarry
behaviours) of granular materials. In fact, the of Missillac in France. The particle size varies
granular materials in low traffic pavements are between 0 and 4 mm. This sand is used as subgrade
commonly in unsaturated states. soil in low traffic pavements for full-scale pavement
Several researches have shown that besides the tests at IFSTTAR (Institut Français des Sciences
effect of water content on the mechanical behav- et Technologies des Transports, de l’Aménagement
iours of the unsaturated granular materials, it is et des Réseaux) in Nantes, France. This kind fine
necessary to take into account the effect of suction sand is sensitive to moisture variation, and its in
(Yang et al., 2008, Cary & Zapata, 2011, Nowa- situ elastic modulus typically varies between 50
mooz et al., 2011, Nowamooz et al., 2013, Salour and 100 MPa.
et al. 2014, Han et al., 2015 and Jing et al., 2016,
2017 in press) as well as hydraulic hysteresis (Miller
2.1 Particle-size analysis
et al., 2008, Yang et al., 2012 and Ho et al., 2014).
Different experimental researches showed that the In this work, the Missillac sand are studied in three
variation of fine content (particles passing the sieve different fine contents, named respectively M4.0,
75 μm or sieve No. 200 based on American classifi- M7.5 and M15.3:
cation) has also an important effect as well as water
– M4.0 samples contain 4.0% of fine content.
content on mechanical behaviours of granular
– M7.5 samples contain 7.5% of fine content.
materials as reported by Babić et al., 2000, Duong
– M15.3 samples contain 15.3% of fine content.
et al., 2013 and Jing et al., 2016, 2017 in press.
However, the relationship between resilient Figure1 shows particle size distribution curves
strain behaviours and unsaturated state with dif- for all of three Missillac sands (XP P94-041, 1995).
ferent fine contents has been rarely studied. Table 1 presents all of the characteristic param-
In this work, the objective is to study the effect eters of these curves, such as Cc and Cu. The
of unsaturated state on the resilient strain behav- Coefficients of curvature (Cc) (estimated between
iours for a granular material constituted of three 1 and 3) show a well-graded composition of three
fine contents in low-traffic pavements. studied Missillac sand. The methylene blue values
The Soil Water Retention Curves (SWRC) are (VBS) of M7.5 and M15.3 (NF P94-068, 1993) are
obtained by suction tests with filter paper method. also introduced in Table 1, which shows an obvi-
The resilient strain behaviours of the unsaturated ous increase of VBS value from M7.5 to M15.3.

93
It can be stated that the main component of the are prepared at a water content ranging from 7%
fine content for Missillac sand is the clay which is to 12.3% and for the drying path, the samples ini-
really sensitive to moisture variation. tially saturated are dried in the ambient tempera-
The materials are also classified based on the ture (20°) to reach the desired water content from
VBS values and particle size distribution (NF P11– 7% to 12.3%. All of samples are compacted at an
300, 1992 or USCS ASTM D2487-06) as reported initial dry density ranging of 2 ± 0.06 Mg/m3. The
in Table 1. measured soil water retention curves for M4.0 and
M15.3 are illustrated in Figure 2.
In unsaturated soil mechanics, various empiri-
2.2 Soil Water Retention Curve (SWRC)
cal equations have been suggested to describe the
When the soil pores are filled by water and air, the SWRC. Among these equations, the relationships
porous material is unsaturated. Unsaturated soils (proposed by van Genuchten, 1980 and Fredlund &
can exert an attraction on water, either by capil- Xing, 1994) have been widely used in geotechnical
lary action in the pores, between soil particles, or engineering. The van Genuchten model which is
through physicochemical effects. The pressure dif- simple and has meaningful parameters is used in
ference is referred to as matric suction: this study. The van Genuchten equation is written
as follows:
s ua uw (1)
(ws wr )
w = wr + (2)
where, ua is pore air pressure and uw is pore water [ + s ]
n m

pressure.
A Soil Water Retention Curve (SWRC) is usu-
where w is the actual soil water content at the suc-
ally used to illustrate the evolution of saturation
tion s; ws and wr are the saturated water content
or water content, as a function of matric suction in
and the residual water content; α is a parameter
unsaturated soil mechanics.
related to the air entry suction; m and n are the
In this study, filter paper method is used to
model parameters with the relationship: m = 1−1/n.
obtain SWRC. For the wetting path, the samples

Figure 2. Matric suction obtained by filter paper

Figure 1. Particle size distribution curves of Missillac method as well as model prediction (M4.0, M7.5 and
sand (M4.0, M7.5 and M15.3). M15.3).

Table 1. Characteristics of the studied materials.

Fraction (%) Particle size Classification

Dry
density 0/80 0.08/0.4 0.4/2 2/4 Blue
Material (Mg/m3) μm mm mm mm d60 d30 d10 Cu Cc value NF USCS

M4.0 2.00 4 10 76 5 0.95 0.55 0.30 3.17 1.06 – B2 SP

M7.5 2.00 7.5 6.5 76 5 1.40 0.60 0.25 5.60 1.03 0.56 B2 SP-SC
M15.3 2.00 15.3 14.7 55 10 0.85 0.40 – 8.50 1.88 0.85 B5 SC

94
Table 2. Parameters of van Genuchten model.

M4.0 M15.3 M7.5 (Prediction)

Parameters
of VG model Wetting Drying Wetting Drying Wetting Drying

α 0.803 0.016 0.049 0.007 0.180 0.012

n 1.929 1.947 2.261 3.025 2.100 2.100
m 0.482 0.486 0.558 0.669 0.524 0.524
ws (%) 14.3 11.2 11.8 11.0 12.3 11.0
wr (%) 6.5 0.1 6.0 0.1 6.0 0.1
s* (kPa) 1.8 ± 0.18 12 ± 1.2 4.2 ± 0.42

13 where a, b and c are constant. In other words, the

•
12
11 .... ..•
M4.0
M7.5
M15.3
soil water retention curves for three different mate-
rials coincide together by using s/s* value instead
10
9
t--. Fittino curve of suction value.
~
~
c 8 ""'.
J!l 7
c
0
u 6
""·• w= a -e <-sls"/bl+c 3 REPEATED LOAD TRIAXIAL TESTS
..
2
!i: 4
5 a
b
6.478
1.967
(RLTT)

3
c 7.032 RLTT is widely used to investigate the mechanical
2 behaviours of granular materials. It can simulate
the variation of pavement loading conditions to
0 describe the resilient strain behaviours or perma-
0.1 10 100
nent strain behaviours.
s/s"

Figure 3. Evolution of water content, as a function of 3.1 Principle of RLTT

s/s*.
For triaxial tests, the mean normal stress p and the
deviatoric stress q are usually used to describe the
The fitting curves of M4.0 and M15.3 are stress state of samples, which are defined as:
also plotted in Figure 2, which shows the van
Genuchten model fits well with the measured val- σ 1 + 2σ 3
ues for both studied materials. The parameters of p= (4)
3
van Genuchten model are summarized in Table 2.
In Figure 2, the s* value is defined as the suc- q = σ1 − σ 3 (5)
tion value corresponding to the intersection point
of wetting and drying paths, which is significantly where σ1 is the vertical stress (kPa); σ3 is the confin-
related to fine content. Besides, since the s* value is ing pressure (kPa).
very sensible to any variation of the model param- The volumetric strain εv and the deviatoric
eters, this value is presented in Table 2 in a range of strain εq are used to describe strain behaviours of
between 90% ∙ s* and 110% ∙ s* for the lower bound samples, which are defined as:
and the upper bound of s* respectively. Table 2
summarizes the s* values for M4.0 and M15.3. ε v = ε 1 + 2ε 3 (6)
Based on the parameters of van Genuchten
model of M4.0 and M15.3, the model parameters 2 (ε ε )
εq = (7)
and the s* values are predicted for M7.5 presented 3
in Table 2. Figure 2 also shows the prediction
SWRC for M7.5 both wetting and drying paths. where ε1 is the axial strain; ε3 is the radial strain.
Figure 3 shows the evolution of water content, As shown in Figure 4, in RLTT, the axial cyclic
as a function of s/s*. It can be observed that a deviatoric stress q and the cyclic cell pressure σ3 are
simple exponential function could represent the applied to samples in phase in each cycle. The axial
relationship between s/s* value and water content, and radial strain behaviour separates two parts:
which is defined as: permanent strain and reversible strain.
For a given relatively low stress state, without
w a ⋅ e ( − s s ∗ b ) +c (3) failure, the plastic strain will not increase with

95
 the last cycle of 100 loading/unloading cycles are
used to determine the resilient strain behaviours.

Cyclic I Cyclic 2 Cyclic 3 Cyclic n 4 TEST RESULTS AND ANALYSIS

Tim<

]Cltamd)
---
Deformation
'

--
- - --_

-- --
P -1
deformation
,_,..:-::_--+ r<~·crsiblc
__
-~
deformation
4.1 Resilient strain behaviours
Figure 6 and Figure 7 present respectively the evolu-
tion of resilient volumetric strain ε vr and the resil-
ient deviatoric strain ε qr in the last cycle for three
different Missillac sands (M4.0, M7.5 and M15.3)
Cyclic I Cyclic 2 Cyclic 3 Cyclic n
at two different water contents (8% and 11%). Based
on these figures, it can be stated that ε vr is positive
in the stress paths of Δq/Δp = 0; 0.5; 1 and 2 (Con-
Figure 4. Principle of repeated load triaxial test.
traction) while ε vr is negative in the stress paths of
Δq/Δp = 3 (Dilation). The resilient deviatoric strain
are positive in the stress paths of Δq/Δp = 2 and 3
100
and negative in the other stress paths.
Besides, the effect of water content on resilient
80 strain behaviours is obvious: Higher the water
content, higher the ε vr in each stress path for each
iV 60 material. Higher the water content, higher the ε qr
~
:. in each stress path for each material. For M15.3
c:r 40 material, there are large increases of ε vr and ε qr
with an increase of water content from 8% to 11%,
20 especially for the stress paths of Δq/Δp = 0; 0.5
pq/fl =0 and 1. The effect of fine content on resilient strain
0 ' behaviours is not significant when the fine content
0 10 20 30 40 so 60 70 80 90 100 increases from 4% to 7.5%. At the same time, a
p (kPa) large open loop can be observed for M15.3 at the
stress paths of Δq/Δp = 0; 0.5; 1 and 3.
Figure 5. Resilient strain behaviour: stress paths
applied.
4.2 Effect of suction
In section 2.2, we defined a new parameter s*
increase of number of cycles after enough loading which is the suction value of the intersection point
cycles. Then the reversible strain could be treat as of wetting and drying paths in SWRC.
resilient strain. Figure 8 compares the maximum resilient vol-
umetric strain ε vrmax in the last cycle of resilient
strain test as a function of water content (Fig-
3.2 Resilient strain tests
ure 8a) and s/s* value (Figure 8b) for three mate-
In this study, the repeated triaxial tests are per- rials (M4.0, M7.5 and M15.3) having a water
formed with three Missillac sands (M4.0, sand content range of 7% to 11.3% in stress paths of
M7.5 and M15.3) for the water contents rang- Δq/Δp = 0 and 3.
ing from 7% to 11% with the same dry density As shown in Figure 8a, the ε vrmax increases
of 2 ± 0.06 g/cm3. The samples are prepared at with the increase of water content (even the results
a diameter of 150 mm (160 mm for M7.5) and a of M7.5 are somehow scatter) like it has been
height of 285 ± 5 mm (320 ± 5 mm for M7.5). described in Figure 6. Besides, the effect of fine
The samples are first subjected to a condition- content is important as well as water content: the
ing phase that consisted of 104 loading/unloading ε vrmax is different with different fine content and
cycles to stabilize the plastic strains. At the end of same water content.
the conditioning phase, the increase in axial plastic In Figure 8b, all of ε vrmax values are plotted in
strain was lower than 10−7 per cycle confirming the s/s* plane. The lower bound of s/110% ⋅ s* and the
stabilized plastic deformation. upper bound of s/90% ⋅ s* are also illustrated in this
After the conditioning phase, 5 different stress figure to take into account the sensitivity to varia-
paths (Δq/Δp = 0; 0.5; 1; 2; 3) are applied to each tion of the parameters of van Genuchten model.
sample as shown in Figure 5. For each stress path, From this figure, it can be stated that a simple

96
 200 ,------""'•""
-'"'M"'4'"'.o""-w""="'7'"'
.8""%'"'-q:;:/p:-:=~s=; 200,------r=-..~~~~~~~~~
M 4 _0-7.8% •
M4.0-w=7.8%-q/p=2-test M 4. 0-11.0% r-:- ~: ~:::~ ~ ~~~;~:~:~::~ ~:~:
b"'- 150 •
•
M4.0-w=7.8%-q/p=1 -test
M4.0-w=7.8%-q/p=0.5-tes
b::::. 150 --T-
• - M4.0-w= 11.0 %-qlp= 1-test data
M4.0-w=11 .0% -q/p=0.5-test dat
tJ' * M4.0-w=7.8% /p=O-test data !..! • M4.0-w= 11.0%:,gm:=o-test data
c
-~ " 100
·e;
100
'"
~
0
'"
~
0

E E
50 50
g=> g=> - ";E.r-
~-
*--*
~~~~ tli
-
c c .........
~ • ~
-~
cr:
0
"81
cr: -
-50 -50
0 20 40 60 80 100 20 40 60 80 100
Mean normal stress p (kPa) Mean normal stress p (kPa)

200 , - - - - - -"r-
""'••, .-;Mc;:7;c.s-"'w
" "-"8" ta9
' .o""%'"'-q/:;:p:-:-"'3--::te:-:s;-:td"'a"' 2oo -r;;~:--;--;-;:;~===w.~w)%-q~4.St"Clat.'ii
M7 .5-8.0% - • M7.5-w=8.0%-q/p=2-test data M7 .5-11.0% :- M7:5-w=11 :0%-q/p=2-test data

"
~
tJ'
150
•
...-
M7 .5-w=8.0%-q/p=1-test data
M7 .5-w=S.O%-q/p=O.S-test data
M7 .5-w=B.O%-<itP=o-test data
"
~ 150 -..-
•
M7 .5-w=1 1.0%-q/p=1-test data
M 7 .S~w-1 1 .0%-qlp;Q_S~test da
M7 .5-w-11.0%· -0-test data
"
~ 0
100
:s
1'1 50
~
g
~ " i
.,
·;;;
cr:
0

-50 -50 +-~--.--~--.-~-.-~---.--~--1

0 20 40 60 80 100 0 20 40 60 80 100
M ean normal stress p (kPa) Mean normal stress p (kPa )

200 • 15.3-w= 11 .1°o-Qip=3-lestdata

• M15.3-w=8. 1%-q/p-3-test data • M15.3-w=11 .1%..q/p=2-testd.ala
• M1 5.3-w=8. 1%-q/p=2-test data - .a. M15.3-w=11 .1%-q/p=1 - testdata

"
0
150
• - M15.3-w=8.1%-q/p= 1-test data
-. M15.3-w::8.1 %-qlp=0.5-test data
f
200 .,
•
M1 5.3-w : 11 . 1% -q/p=0 .5 -test dat
M1 5.3-w=11 .1%- I =0-testdata

""
-~
100
,.._ M15.3-w=8.1%-Q.rP; Q.test data
:;,: 150
.5
'".s ~ 100
1'=>1 50 ·"~
g §
c g 50
~ 10
-~ 0 .!!I
cr: ~ 0

-50
M15.3-8.1% "' Ml5.3-ll.l%
-50+-----~------~----~------~------1
0 20 40 60 80 100 0 20 40 60 80 100
M ean normal stress p (kPa) Mean normal stress p (kPa)

Figure 6. Evolution of resilient volumetric strain ε vr .

exponential function could represent the relation- for three different materials (M4.0, M7.5 and
ship between all of the ε vrmax values and s/s* values, M15.3) having a water content range of 7% to
which is defined as: 11.3% in stress paths of Δq/Δp = 0 and 3.
In Figure 9a, the ε qr max increases with the
ε vrmax = A e ( s / s*/ B )
+C (8) increase of water content and the fine content also
plays an important role.
where A, B and C are constant as shown in In Figure 9b, another simple exponential func-
Figure 8b. As a result, the ε vr max could be deter- tion can be observed to represent the relationship
mined by s/s* only. These results are useful to between all of the ε qr max values and s/s* values,
understand problem of dual variation of the which is defined as:
maximum resilient volumetric strain ε vrmax with the
water content and the fine content. ε qrmax = D e ( s / s*/ E )
+F (9)
As ε vr max , the maximum resilient deviatoric
strain ε qr max are also plotted in water content plane where D, E and F are constant as shown in
(Figure 9a) and s/s* plane (Figure 9b) respectively Figure 9b.

97
 20 ~----~----------------------, 20~----------------------~----,

~ 10 k 10

~"' 0 ~., 0

~ -10
.g
~ -10
.g
B B
-~ -20 ~ -20
~ ~
c -30 • - M4.0-w= 7.8%-qlp=3--test data c -30 M4.0-w=1 1.0%-q/p=3-test data
~ ~
·;; • M4.0-w=7 .8%-q/p=2-test data ·;; • M4.0-w=1 1.0%-q/p=2-test data
&-40 a- M4 .0-w=7.8%-qtp=1-test data ~ -40 a- M4.0-w=1 1.0%-q/p=1-test data
.- M4.0-w=7.8%-q/p=0.5-test data M 4 0 II 0 01 • M4.0-w=1 1.0%-q/p=0.5-testdat
_50 +-M 7 _.8_"_
_ 4_._0_-T Yo_b,;..._
==fM~4~.0-fOw~=;=7=l.B!.f%!!i-~P=:,O-i"t'"
es~t~da~ta~l · ---r-·--1_0' ~*~M~4~.0~-~w=
-~ +---- =1~1~.0~%~-g~i~=~OF·ffi~s~t~
da~ta~
0 20 40 60 80 100 0 ~ 40 00 M 100
Mean normal stress p (kP a ) Mean normal stress p (kPa)

20~----~----------------------, 20~----------------------~----,

• M7.5-w- 8.0%--qlp- 3-test data M7.5-w=11.0%-:q/Ji=3-test data

• M7.5-w=8.0%-q/p=2-test data • - M7.5-w=11.0%-q/p=2-test data
f-.-. M7.5-w=8.0%-q/p=1-test data - M7.5-w=11 .0%-q/p=1-test data
8 00 • M7.S.w=B.O%-q/p=0.5-test dat
M7 .5-ll.Oo/o _;___~~:;~~~:~~~~~g;~~~~~a~;j
·SO +-M
_:_;7_:_.5:_-...;:..:·_::..:.V.!;:_o...!;H,.
~~M::,;7..5:;:·w:;;=FB
;; ..: :,O%~.-<;:;!J'~I/ o:;:'=O;,::·t::;_es~t~da;::ta::d.j
-~
0 20 40 60 80 100 0 20 40 00 80 100
Mean normal stress p (kPa) Mean normal stress p (kPa)

20 ~--~-------------, • - M15.3-w=1 1. 1%-qlp=3-test data

20 ~------! • M15.3-w=11 . 1%-q/p=2-test data
• M15.3-w=11 .1%-q/p=1-test data
~ 10
f 10
• M15.3-w=11. 1%-q1p=0.5-test data
- M15.3-w=11. 1%- =0-test data
7.¢" 0 ;:;., 0
c

~ -10
c
!2 -10

·"_g '"
u
~ -20
-~ -20
~ -30
"~ -30 • - M15.3-w=8.1%-qlp=3-test data
~ • M 15.3-w=8 .1%-qlp=2-test data ~ -40
&-40 10/ •
& - M15.3 -w=8.1%-q/p=1-test data !'i
"'-~
M 15 3 o M15.3-w=8.1%-q/p=0.5-test data
M15.3-ll.l%
-so +-- -· _-To_._1_'0.L.:•:,....:M""-!;
1 5"'.3"·w:::=;o•o.·1'-!%,:·"'1"'P=:;O;=!·t,es,_t~da,_,ta"-''J
0 20 40 60 80 100
Mean normal stress p (kP a ) Mea n normal stress p {kPa)

Figure 7. Evolution of resilient deviatoric strain ε qr.

,- ~----------------~---------------,,-
~ 200 • M4.0-6ql6p=O-test data .-- --+--=::."""""
.. <uur+-
. - - - - - - - ,r.. • M4.0-Aq/lip=O-test data
• M7.5-6q/6p=O-test data - • M7.5-oqlop=O-test data
..r.J'
"' M15.3-6q/6p=O-test data "' ~ "' M15.3-oqlop=O-test datal
o M4.0-aq/Ap=3-test data ..~.. 150 o M4.0- qlop=3-test data
c: 150 \ o M7.5-Aq1Ap=3-test data
·~ o M7.5-oq/6p=3-test da ta -~
6 M15.3-Aql.&p=3-test data
\
;; "' M 15.3-oq/6p=3-test data '"
u u [Thetowerbound: st(110%1.s·t Fitting curve
~ ~ 100
§
100
.. . E
\

~ ,....
1
The upper booncl: s/(90%-s")

<-- =A-e(-sJs•tB) + C
"
~
c
~
·;;;
~
50
.• .. ~

E (J - cra-cA - ~ - o - v a e-ll.- - - 5. - - - -
E n. 4., - - -
..
·;;c
::; -50
a
6 10 11 12 0.1 10 100
Water content (%) sJs•

Figure 8. Relationship between maximum resilient volumetric strain ε vr max and water content w and s/s* value.

98
Figure 9. Relationship between maximum resilient deviatoric strain ε qr max and water content w and s/s* value.

5 CONCLUSIONS Cary, C.E., & Zapata, C.E. 2011. Resilient Modulus for
Unsaturated Unbound Materials. Road Materials and
In this paper, we address the problem of variation Pavement Design, 12(3): 615–638.
of unsaturated state with different fine contents on Duong, T.V., Tang, A.M., Cui, Y.J., Trinh, V.N., Dupla,
J.C., Calon, N., et al. 2013. Effects of fines and water
resilient strain behaviours of a granular material in
contents on the mechanical behavior of interlayer soil
low-traffic pavements. in ancient railway sub-structure. Soils and Founda-
The RLTT experimental results in s/s* plane tions, 53(6): 868–878.
show that there are two simple exponential func- Fredlund, D.G., & Xing, A. 1994. Equations for the soil-
tions could represent respectively the relation- water characteristic curve. Canadian Geotechnical
ship between ε vr max and ε qr max and s/s*. Hence, Journal, 31(6): 1026–1026.
the maximum resilient strain behaviours of this Han, Z., Mihambanou, B., & Vanapalli, S.K. 2015.
unsaturated granular material in low-traffic A New Approach for Estimating the Influence of
pavements could be predicted only by SWRC of Soil Suction on the Resilient Modulus of Pavement
Subgrade Soils. Airfield and Highway Pavements,
each material. These findings are helpful for an
861–872.
easier interpretation of the results and reduc- Ho, X.N., Nowamooz, H., Chazallon, C., & Migault, B.
ing the number of tests required to predict the 2014. Effective stress concept for the effect of hydrau-
unsaturated resilient strain behaviours of missil- lic hysteresis on the resilient behavior of low traffic
lac sand. pavements. International Journal of Pavement Engi-
Besides, the SWRCs for three different materials neering, 16(9): 842–856.
locate together as a unique curve in s/s* plane. This Jing, P., Nowamooz, H., and Chazallon, C. 2016. Influ-
is another new point, and it will be continued in ence of Fine Content and Water Content on the Per-
the future studies. manent Mechanical Behavior of a Granular Material
Used in Low Traffic Pavements. Proc. Geo-China
2016., Jinan, 25–27 July 2016: 215–223.
Jing, P., Nowamooz, H., and Chazallon, C. 2017 in press.
ACKNOWLEDGMENTS Permanent Deformation Behavior of a Granular
Material Used in Low Traffic Pavements. Road Mate-
This work is supported by the China Scholarship rials and Pavement Design. Accepted: 02/11/2016;
Council. Acknowledgments to the previous work Number of pages: 26; Editor: Catubay, Praise Ann.
of Pierre Hornych (IFSTTAR) and Xuan Nam Ho Miller, G.A., Khoury, C.N., Muraleetharan, K.K., Liu,
C., & Kibbey, T.C.G. 2008. Effects of solid deforma-
(INSA-Strasbourg).
tions on hysteretic soil water characteristic curves:
Experiments and simulations. Water Resources
Research, 44, W00C06.
REFERENCES NF P11-300. 1992. Exécution des terrassements-Classi-
fication des matériaux utilisables dans la construction
ASTM D2487-06. 2006. Standard Practice for Classifica- des remblais et des couches de forme d’infrastructures
tion of Soils for Engineering Purposes (Unified Soil routières.
Classification System). NF P94-068. 1993. Sols: reconnaissance et essais-Mesure
Babić, B., Prager, A., & Rukavina, T. 2000. Effect of fine de la quantité et de l’activité de la fraction argileuse-
particles on some characteristics of granular base Détermination de la valeur de bleu de méthylène d’un
courses. Materials and Structures, 33(7): 419–424. sol par l’essai à la tache.

99
Nowamooz, H., Chazallon, C., Arsenie, M.I., Hornych, Soils1. Soil Science Society of America Journal, 44(5):
P., & Masrouri, F. 2011. Unsaturated resilient behav- 892.
ior of a natural compacted sand. Computers and Geo- XP P94-041. 1995. Sols: reconnaissance et essais-Iden-
technics, 38(4): 491–503. tification granulométrique-Méthode de tamisage par
Nowamooz, H., Ho, X.N., Chazallon, C., & Hornych, voie humide.
P. 2013. The effective stress concept in the cyclic Yang, C., Sheng, D., & Carter, J.P. 2012. Effect of hydrau-
mechanical behavior of a natural compacted sand. lic hysteresis on seepage analysis for unsaturated soils.
Engineering Geology, 152(1): 67–76. Computers and Geotechnics, 41: 36–56.
Salour, F., Erlingsson, S., & Zapata, C.E. 2014. Model- Yang, S.R., Lin, H.D., Kung, J.H.S., & Huang, W.H.
ling resilient modulus seasonal variation of silty sand 2008. Suction-Controlled Laboratory Test on Resil-
subgrade soils with matric suction control1. Canadian ient Modulus of Unsaturated Compacted Subgrade
Geotechnical Journal, 51(12): 1413–1422. Soils. Journal of Geotechnical and Geoenvironmental
van Genuchten, M.T. 1980. A Closed-form Equation for Engineering, 134(9): 1375–1384.
Predicting the Hydraulic Conductivity of Unsaturated

100
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Laboratory investigation on unbound materials used in a highway with

premature damage

D.M. Barbieri, I. Hoff & H. Mork

Norwegian University of Science and Technology, Trondheim, Norway

ABSTRACT: A highway road section was built from 2004 to 2006 in the southern part of Norway.
Soon after the opening significant damage in form of alligator cracking and rutting occurred. The aim of
the research described in this paper is to characterize the behaviour of the pavement unbound layers that
may have possibly caused the issue.
At a first stage, a collaboration among different partners carried out field surveys and samplings in five
locations with and without visible surface damage.
At a second stage, laboratory analyses investigated the properties of the base layer and especially
focused on repeated triaxial load testing. The specimens were analysed with varying moisture content and
degree of compaction. These laboratory tests gave significantly lower stiffness and resistance to perma-
nent deformation for the damaged sections; this could partly explain why distresses are more evident just
on some parts of the road.

1 INTRODUCTION Table 1. Section designs as derived from Handbook

018. Construction of the wearing course was postponed.
The research described in this paper focuses on a
highway section situated in the southern part of Layer Thickness (mm) Material
Norway. The new road link was built in 2004; the wearing course 35 hot-mixed asphalt
length is about 17.5 km, including a bridge and binder course 45 hot-mixed asphalt
several tunnels. upper base 50 paved gravel
Damage in form of both alligator cracking and lower base 150 crushed rock
rutting occurred soon after the opening to traffic subbase 550 blasted rock
in 2006: a series of interconnecting cracks devel-
oped along the wheel paths.
This type of damage is uncommon in new
2 METHODOLOGY
asphalt pavements in Norway: excessive rutting
and unevenness are the typical damage sometimes
2.1 Material sampling
seen early in a pavement life.
A collaboration comprising NTNU (Norwegian Figure 1 portrays the condition of the wearing course
University of Science and Technology), NPRA (Nor- related to sites I, II, III, V.
wegian Public Roads Administration) and the con- The ruts along the wheel paths were measured
tractor organized a field investigation in September using a three-meter straightedge. The condition of
2014 in five areas. Three locations had clear distresses: the road pavement was inspected according to a
site I (km 0.844), site II (km 1.195), site III (km 1.547). preset excavation pattern (Figure 2).
Alligator cracking was distinctly visible and the frac- The excavation surface was rectangular with
ture pattern was approximately 10 × 10 cm. 4 m × 5 m in size. The asphalt layer was sawn with
Two locations had no visible distresses: site IV a diamond blade and removed using an excava-
(km 8.709), site V (km 9.720). Table 1 describes the tor. A transversal excavation ca. 0.5 m deep took
road pavement structure designed according to the place in the central area to examine both base
pavement design manual in use at the time (Statens and subbase layers. There were also some lesser
Vegvesen 1998). digs extending longitudinally from the main
The aim of the research presented is to analyse excavation to compare the base course materials
the mechanical behaviour of the unbound layers in along the inner and outer as well as between the
the pavement and highlight any possible relation- wheel-paths.
ships between their properties and the premature The lower base course material was so hard
damage encountered. RTLT (Repeated Triaxial to displace that the use of a pneumatic drill was
Load Test) was used to characterize the materials. necessary. This action could have generated some
101
 Figure 2. Framework of excavation and its fulfilment.

extra fines. However, it is believed that this did not

significantly change the grain size distribution; it
is reasonable to assume any crushing happened in
the same proportion for all the sites.

2.2 Sieving analysis

Laboratory tests enabled a more careful examina-
tion of the materials extracted from each location,
the unbound lower base course (crushed rock)
were studied.
The sieving analysis identifies the grain sizes
and highlights possible significant discrepancies
among the five locations and the grading envelope
specified by the code Handbook 018 (also defined
in EN 933–1, CEN 2012).

2.3 RTLT (Repeated Triaxial Load Test)

RTLT is the main testing procedure used in this
research. It gives a comprehensive insight into
the properties of the material by assessing the
stiffness (resilient modulus, which replaces the
modulus of elasticity to indicate the nonlinearity
of the behaviour) and the resistance to permanent
deformation.
Barksdale (1971) found that the RTLT is one of
the best methods available for laboratory simula-
tion of traffic loading on unbound materials. It
reproduces the stress conditions in flexible pave-
ments more adequately than other available meth-
ods like the CBR test.
A confining pressure and a vertical load are the
actions exerted on the specimen. For unbound
granular aggregates, it is often convenient to
separate deformations into two types: one elastic
(recoverable or resilient) and one permanent (not
recoverable or residual). Rutting is the most com-
mon damage caused by permanent deformations.
Seed et al. (1962) and Hicks & Monismith
(1971) discussed the main factors influencing the
stiffness and the deformation for unbound materi-
als; Lekarp et al. (2000a, b) presented a thorough
state-of-the-art of the mechanical behaviour. They
found that the resilient modulus and the perma-
Figure 1. Road surface condition of the surveyed nent deformation were mainly influenced by the
sites. same factors (Uthus 2009): mineralogy, stress level,

102
moisture content, dry density, grading, particle Table 2. Loading sequences for the multi-stage low
size and fines content. stress procedure (data in kPa).

2.3.1 Sample preparation Seq. 1 Seq. 2 Seq. 3 Seq. 4 Seq. 5

Stress level, moisture content and dry density are σt σd σt σd σt σd σt σd σt σd
the input parameters used in this research to ana-
lyse the sensitivity of the collected materials. The 20 20 45 60 70 80 100 100 150 100
amount of water in the samples varies between 20 40 45 90 70 120 100 150 150 200
3%, 5% and 7%. A Kango 950X vibratory hammer 20 60 45 120 70 160 100 200 150 300
(total weight 35 kg, frequency 25 ÷ 60 Hz, ampli- 20 80 45 150 70 200 100 250 150 400
tude 5 mm) compacts the specimen layers for 20 s 20 100 45 180 70 240 100 300 150 500
or 40 s leading to different densities. Both the min- 20 120 45 210 70 280 100 350 150 600
eralogy and the grading are given.
In order to prepare the sample, the desired
amount of water is added to the testing material,
ing deviator stress. Table 2 contains the five load-
which rests for 24 h to let the moisture distribute
ing sequences and the respective loading steps
uniformly. Five equal layers are compacted with
(combination of confining and deviatoric stresses).
the vibratory hammer according to the desired
The hydraulic jack applies the deviatoric stress
time. The bulk density and dry density are assessed
according to a sinusoidal pattern. It ranges from a
as specified by EN 13286-4 (CEN 2003).
minimum value of 5 kPa to assure contact between
All the samples have a diameter of 150 mm
the end plate and the jack itself to the maximum
and the height of the mould is 240 mm. This
values specified in Table 2. Each load step consists
parameter differs from the indication given by EN
of 10,000 load pulses at 10 Hz frequency. A load-
13286-7 (CEN 2004) where the height is recom-
ing sequence is interrupted if the axial permanent
mended to be twice the diameter of the sample.
deformation reaches 0.5%; the operator then con-
Dongmo-Engeland (2005) studied the influence of
tinues with the next loading sequence.
the height to diameter ratio with respect both to
resilient modulus and the resistance against perma-
2.3.3 Interpretation of results
nent deformations. It was concluded that samples
The resilient modulus MR expresses the stiffness of
with a ratio ranging from 1:1 to 1.5:1 showed little
the material. The resilient modulus associated with
difference.
a change in the dynamic deviatoric stress σddyn and
Water is the confining medium, the triaxial cham-
a constant confining pressure σt is:
ber is made of plexiglass. There is one metal plate at
each end of the specimen, two rubber membranes
surround it and prevent the entrance of water. Δσ ddyn
MR =
d
(1)
ε ela
2.3.2 Loading procedure
Two and three LVDTs (Linear Variable Differ-
ential Transducer) measure the axial and radial where εael is the axial resilient strain.
deformations, respectively. There are models describing the relationship
The code EN 13286-7 (CEN 2004) specifies dif- between the resilient modulus and the applied
ferent ways to apply the load. The multi-stage load- stress. The k-θ model is a well-known curve-fitting
ing procedure gathers data regarding the resistance equation based on the sum of principal stresses or
against permanent deformation and the resilient bulk stress. The k-θ model is a non-linear, stress-
properties from one sample. The alternative of test- dependent power function model described by
ing one sample for each stress combination would Hicks & Monismith (1971). The model is given as
be very time consuming (Gidel et al. 2001). follows in its dimensionless form:
The confining medium applies a uniform con-
stant pressure in all directions (σt, triaxial or ⎛ θ ⎞ k2
confining stress). The hydraulic jack exerts an M R = k 1σ a⎜ ⎟ (2)
additional vertical dynamic stress (σd, deviatoric ⎝ σ a⎠
stress), which stepwise increases at different levels
of confining stress. where θ is the bulk stress sum of the principal
The triaxial test apparatus performs the multi- stresses, σa is a reference pressure (100 kPa) and
stage loading procedure in five different sequences k1 and k2 are model parameters assessed from the
with different confining stresses (σt = 20, 45, 70, regression of the test results.
100, 150 kPa). In addition, six steps form a given Figure 3 is a typical RTLT output. The resilient
sequence and each of them has a different increas- modulus is evaluated as expressed by equation 1.

103
 ~
900
to define failure. The strain rate ε‘ here refers to the
800
~
development of permanent deformation per cycle.
~ 700
:::> 600 cr, = 100 kPa Table 3 defines the two boundary lines between
:::>
't:l
e suu
.__..1"'__ cr, = 70 kPa
the three ranges (Figure 5): each load step is cat-
E
.,c:
400
cr, = 45 kPa egorized considering the average strain rate for the
300
., 200 cr, = 20 kPa last 5000 to 10,000 cycles (Hoff at al. 2003).
The equations for the elastic limit line and fail-
.j 100
ure line are, respectively:
10000 20000 30000 40000 50000 600()()

2 sin ρ ⋅ (σ + a )
N, number of load cycles

σd = (4)
Figure 3. Example of resilient modulus data obtained 1 − sin ρ
from a repeated triaxial load test.
2 sin ϕ ⋅ (σ + a )
σd = (5)
1 − sin ϕ
1
0 at = 20 kl'a

O at = 45 kPa
A regression analysis is used to find the two best
" at = 70 kPa fit boundary lines. As a simplification, the appar-
• at = 150kPa ent attraction is interpreted to be 20 kPa for all the
· k-0 model
samples to make comparison easier between them.
Unbound granular materials develop some
0, bulk stress (kPa)
permanent deformations at each load step.
Figure 6 shows a typical output obtained from a
Figure 4. Example of a k-θ regression curve from the RTLT. The permanent deformation is assessed at
resilient modulus data obtained from a repeated triaxial every load step of each load sequence. The strain
load test. rate is normally highest for the first few pulses of
loading and then decreases gradually. There are five
loading sequences corresponding to five confining
Figure 4 portrays a distribution of the resilient (triaxial) stresses. It is quite common to have extra
moduli obtained when using the bulk stress as the permanent deformation for the first load sequence
abscissa. The continuous line is the general trend
obtained from the k-θ model (Equation 2) adopted
as the regression curve.
There are several models used to interpret the per-
manent deformation behaviour of an unbound gran- !failure zone I
ular material tested in a cyclic load apparatus; the lelasto-plastic zonel
. · t \\1\e
shakedown approach (Werkmeister et al. 2001) and e\ast\.c\'-~
the Coulomb approach (Hoff et al. 2003) are two of
them. The Coulomb criterion relates the mobilized Ielastic zone I
shear strength to the development of permanent
deformations and the maximum shear strength to
failure. The mobilized angle of friction ρ and the
failure angle ϕ, respectively, express the degree of Figure 5. Degrees of mobilized shear strength ρ and
mobilized shear strength and the maximum shear maximum shear strength ϕ: different ranges of material
strength. The mobilized angle of friction is: behaviour.

σ1 σ3
sin ρ = (3) crt= 20kPa
σ 1 + σ 3 + 2a cr,= 45 kPa

cr, = 70 kPa
where σ1 is the maximum principal stress, σ3 is the
minimum principal stress and a is the apparent i
; 1 cr, = 150 kPa
attraction of the material. The angle of friction and
of failure identify three different ranges of material
5c. 0
0 10000 20000 30000 40000 50000 60000
behaviour: elastic, elasto-plastic and failure. The N, number of load cycles
Coulomb criterion is shown in Figure 5 in a σ-τ plot.
The strain rate ε‘ is a measure of the speed of the Figure 6. Example of permanent deformation data
permanent deformation and is used as the parameter obtained from a repeated triaxial load test.

104
Table 3. Permanent strain rate values defining the Table 4. Field surveys data collection.
ranges of material behaviour.
Wheel path Lower
Permanent strain rate Range depth (mm) Asphalt sub-
layer base
ε‘ < 2.5 ⋅ 10−8 elastic zone Site Condition inner outer (mm) (mm)
2.5 ⋅ 10−8 < ε‘ < 1.0 ⋅ 10–7 elasto-plastic zone
ε‘ > 1.0 ⋅ 10−7 plastic (failure) zone I damage 24 17 80 135
II damage 24 20 84 158
III damage 13 13 96 123
700
IV no damage 17 15 101 95
~ 600 V no damage 15 16 94 148
.. 500
400
'J ····• ··· site I ···-+···site II ····•··· site III
...0 300
~ 200 "" - e- site !V -JK- site V - HO!Slow limit
·~ - HOI 8 high limit
100 ,.,, ~~~:::: 8 [] [] []
"0 100
.......
J -40
·····' 40 80 120 160
90

~~~.\
80
a 1, triaxial stress (kPa) ~
70

\~:;~;, ._:;: ; ;..,;':•

60
g_
Figure 7. Example of permanent deformation data
treatment according to the Coulomb approach. The
.
~
50
40
30
shapes of the points (square, triangle, circle) stand for the
deformation range. The two boundary lines corresponds
l 20
10
to the degree of mobilized shear strength ρ and degree of
100 10 1 0.1 0.01
maximum shear strength ϕ. sieve size (mm)

Figure 8. Sieving test of the materials.

·· · . 1
indicating that the following sequences are influ-
enced by the load history of the previous steps. ···• ··· site I -e-site IV
The limits given in Table 3 are the criteria defin-
e-
2.4
ing the boundaries between the different ranges
2.3 I~
___J_ ..········
=--o
··········• -
according to the Coulomb approach. The steps
in the elastic, elasto-plastic and failure range are
respectively marked by square, triangle and circle
-
~
"0
~
2.2
- : . ;....... ~--~·---
I
€)----

symbols. The first load sequence is interpreted as 2.1

per part of the compaction procedure of the mate- 2 3 4 5 6 7 8
W (o/o)
rial, thus only the other four load sequences contrib-
ute to the assessment of the boundary lines (Uthus Figure 9. Dry densities for samples coming from site I,
et al. 2007). Figure 7 portrays a typical output. IV; w = 3%, 5%, 7%; tc = 20 s.

allowed limit. Even if these grain size distributions

3 RESULTS AND DISCUSSION
were somewhat outside of requirements for some
of the samples, it was difficult to explain the dif-
3.1 Material sampling
ference in performance among the five sites from
Table 4 lists the results obtained from the field these tests.
investigations.
3.3 RTLT, resilient modulus
3.2 Sieving analysis
The first RTLTs inspect the materials coming from
The sieving test describes the grain dimensions of site I and from site IV to establish some initial
the lower base relating to all the five inspected sites. comparisons between a damaged site and a not
The sieving curves according to the standard EN damaged site. The compression time (tc) is 20 s and
933-1 (CEN 2012) are displayed in Figure 8. the water contents are 3%, 5% and 7%. Figure 9
The grain sizes for the base layer are between portrays the dry density corresponding to each
the oversize and the undersize specified by Hand- water content after compaction.
book 018 (Statens Vegvesen 1998) for site I, IV and Both site I and IV show a little variation in den-
V. The grain size distribution for sites II and III is sity and the optimum water content determined
very close to the oversize limit, and the quantity from these samples is around 5%. The field densi-
of material smaller than 2 mm is higher than the ties were not measured.

105
 - w=7% - - w=S% -----w=3% --•-- site I ···+···site II ····•··· site HI - e- site IV -*- site V

-
6""
=
=
"tl
900
800
700
600
500
-
6""
:g
~
700

600

500

9 400
9 400
c 300
5 300
~ 200
~
e 100 e
200
Ci 100
~
0
200 400 600 800 1000 i
9, bulk stress (kPa) 200 400 600 800 1000
9, bulk stress (kPa)

Figure 10. k-θ regression curves for site I; w = 3%, 5%,

7%; tc = 20 s. Figure 12. k-θ regression curves for site I, II, III, IV, V;
w = 7%; tc = 20 s.
Table 5. Model parameters k1 and k2 for site I; tc = 20 s.
Table 7. Model parameters k1 and k2 for site I, II, III,
Site w (%) k1 k2 IV, V; w = 7%.

I 7 1.4 0.7 Site tc (s) k1 k2

I 5 2.0 0.7
I 3 2.7 0.4 I 20 1.4 0.7
II 20 1.1 0.5
III 20 0.6 0.6
-w=7% - -w=S % -----w=3%

- 900 IV 20 1.5 0.7

800 V 20 1.5 0.6
6"" 700
600
=
:§ 500

9 400 ··•·· site I ···+···site II ····•··· site HI -e-site IV -)1(- site V

s 300
l1200
~ 200
e 100
~ 1000
Ci
~ :g 800
200 400 600 800 1000
0, bulk stress (kPa)
9 600
5
~ 400
Figure 11. k-θ regression curves for site IV; w = 3%, 5%, e 200
7%; tc = 20 s. i
200 400 600 800 1000

Table 6. Model parameters k1 and k2 for site IV; tc = 20 s. 9, bulk stress (kPa)

Site w (%) k1 k2 Figure 13. k-θ regression curves for site I, II, III, IV, V;
w = 7%; tc = 40 s.
IV 7 1.5 0.7
IV 5 1.2 0.5 Table 8. Model parameters k1 and k2 for site I, II, III,
IV 3 3.1 0.5 IV, V; w = 7%.

Site tc (s) k1 k2
The k-θ regression curves for site I are displayed
in Figure 10. Table 5 specifies the k1 and k2 param- I 40 0.9 0.8
eters corresponding to the curves. II 40 0.8 0.4
The k-θ regression curves for site IV are dis- III 40 1.1 0.5
played in Figure 11. Table 6 specifies the k1 and k2 IV 40 1.2 0.8
parameters corresponding to the curves. V 40 2.5 0.6
The resilient moduli data from site I and IV
show a sensitivity to water content. Further analy-
ses are considering just w = 7% because this is of 20 s. Table 7 specifies the k1 and k2 parameters
sufficient to highlight discrepancies among the corresponding to the curves.
five sites; moreover, running RTLTs with further Figure 13 shows the different k-θ regression curves
water content values would be quite time consum- with compaction time of 40 s. Table 8 specifies the k1
ing. The specimens from sites II and III reach the and k2 parameters corresponding to the curves.
maximum permanent deformation of 0.5% just The stiffness values in Figure 12 and 13 underline a
after some tens of cycles. Figure 12 shows the dif- significant variation in the resilient properties of the
ferent k-θ regression curves with compaction time materials. The specimens from sites II and III have

106
the lowest curves throughout the bulk stress range: Moreover, the specimen from site IV has higher
their resilient modulus values are almost one third of range boundary limits than speci mens from site
those belonging to the other sites when tc = 20 s. The I. As for the resilient modulus, further analyses are
not damaged sites IV and V always have the highest performed just considering w = 7%. The specimens
stiffness values, the damaged site I presents high val- from sites II and III reach the maximum perma-
ues when the compaction time is 20 s. nent deformation of 0.5% just after some tens
of cycles. The analyses regarding these two loca-
tions are not given because they reach the failure
3.4 RTLT, permanent deformation
very quickly (they would lie along the X axis).
As specified in section 3.3, the specimens com- Figure 16 shows the different range boundary lines
ing from sites I and IV are tested initially with with compaction time of 20 s. Table 11 specifies
compression time (tc) of 20 s and water contents the corresponding angle of mobilization ρ and
equal to 3%, 5% and 7%. The elastic limit line and angle of failure ϕ.
the failure line for site I are displayed in Figure 14. Figure 17 shows the different range boundary
Table 9 specifies the angle of mobilization ρ and lines with compaction time of 40 s. Table 12 speci-
the angle of failure ϕ corresponding to site I. fies the corresponding angle of mobilization ρ and
The elastic limit line and the failure line for site angle of failure ϕ.
IV are displayed in Figure 15. Table 10 specifies the Figures 16 and 17 show a significant variation
angle of mobilization ρ and the angle of failure ϕ in the permanent deformation. The performances
corresponding to site IV. of sites IV and V are similar. The angle of mobi-
The permanent deformation data from site I lization from site I is approximately 50% of the
and IV also show a sensitivity to water content. corresponding value of sites IV and V, the ratio
becomes approximately 15% when considering the
- w=7%,p ---*- w=7%, !jl angle of failure. The specimens from sites II and
- - w=5%, p ~- w=5%,!jl
----- w=3%, p --X-- w=3%, III reaches the 0.5% limit very quickly and they are
~
!p

~
500
600
400
I
I ---<"~
not reported in the graphs.

~.
~ 300 Table 10. Angle of mobilization ρ and angle of failure
~ 200 ϕ for site IV; tc = 20 s.
•t
"CC 100
J 0 ~-------------------------------------- Site w (%) ρ (°) ϕ (°)
0 40 80 120 160
a 0 triaxial stress (kPa)
IV 7 53 61
IV 5 48 58
Figure 14. Range boundary lines for site I; w = 7%, 5%,
3%; tc = 20 s. IV 3 68 71

Table 9. Angle of mobilization ρ and angle of failure ϕ .... ... site I, p .... ... site I, <P - e - site IV, p
for site I; tc = 20 s. - e - site IV, <P - • - site V, p - • - site V, <P

Site w (%) ρ (°) ϕ (°) ~ 600

500

~... 400 -- ...

~.:.~.~.~:.~.;;1;~::::~::::~~~~~;
I 7 32 50 ... 300
I 5 59 65 ~ 200
-~
I 3 66 68 "CC 100

J
40 80 120 160
a 0 triaxial stress (kPa)
- w=7%,p ---*- w=7%, !jl
- - w=5%, p ~- w=5%,!jl
----- w=3%, p Figure 16. Range boundary lines for site I, IV, V;
~~
--x-- w=3%, <P

w = 7%; tc = 20 s.
500
600 I
400
~ 300 Table 11. Angle of mobilization ρ and angle of failure
~ 200 ϕ for site I, IV, V; w = 7%.
•"CCt 100
J 0 ~-------------------------------------- Site tc (s) ρ (°) ϕ (°)
0 40 80 120 160
a 1, triaxial stress (kPa)
I 20 32 50
IV 20 53 61
Figure 15. Range boundary lines for site IV; w = 7%, V 20 64 67
5%, 3%; tc = 20 s.

107
 ···• ··· site I, p ···• ··· site I, cp - e - site IV, p
REFERENCES
- e - site IV, cp - " - site V, p - " - site V, cp

~ 600 Barksdale, R. 1971. Compressive stress pulse times

~,,,,,~:::::::~~~~:~:~~=~~~~
500 in flexible pavements for use in dynamic testing.
~ 400 Highway Research Record 345: 32–44. Washington
..."' 300 DC: Highway Research Board.
~ 200 CEN 2003. EN 13286-4 Unbound and hydraulically
·~ bound mixtures. Part 4: test methods for laboratory
"CC 100
J
• ····················· reference density and water content. Vibrating ham-
40 80 120 160 mer. Brussels: CEN.
a 1, triaxial stress (kPa) CEN 2004. EN 13286-7 Unbound and hydraulically
bound mixtures. Part 7: cyclic load triaxial test for
Figure 17. Range boundary lines for site I, IV, V; unbound mixtures. Brussels: CEN.
w = 7%; tc = 40 s. CEN 2012. EN 933-1 Tests for geometrical properties
of aggregates. Part 1: determination of particle size
Table 12. Angle of mobilization ρ and angle of failure distribution. Sieving method. Brussels: CEN.
ϕ for site I, IV, V; w = 7%. Dongmo-Engeland, B. 2005. GARAP, Influence of
sample’s height on the development of permanent
Site tc (s) ρ (°) ϕ (°) deformation. Rapport STF50 A05075. Trondheim:
SINTEF.
I 40 29 48 Gidel, G. & Hornych, P. & Chauvin, J.J. & Breysse, D. &
IV 40 62 66 Denis, A. 2001. A new approach for investigating the
V 40 67 70 permanent deformation behaviour of unbound gran-
ular material using the repeated load triaxial appara-
tus. Bulletin des Laboratoires des Ponts et Chaussées
233: 5–21. Champs-sur-Marne: IFSTTAR.
4 CONCLUSION Hicks, R.G. & Monismith, C.L. 1971. Factors influ-
encing the resilient response of granular materials.
This research has dealt with the road distresses Highway Research Record 345: 15–31. Washington
which occurred in a highway in southern part of DC: Highway Research Board.
Norway. Five locations with and without visible Hoff, I. & Bakløkk, L. & Aurstad, J. 2003. Influence
of laboratory compaction method on unbound
damage were chosen to survey the conditions. granular materials. Proceedings for the 6th Inter-
Laboratory tests evaluated the properties of the national Symposium on Pavements Unbound: CD-
unbound base layer. Specimens from two sites con- ROM. Nottingham: University of Nottingham.
tained somewhat more material smaller than 2 mm Lekarp, F. & Isacsson, U. & Dawson, A. 2000a.
than allowed in the design guidelines. State of the Art. I: Resilient response of unbound
The curves resulting from the k-θ regression aggregates. Journal of Transportation Engineering
model highlighted that the base layers of not 126(1): 66–75. Reston: ASCE.
Lekarp, F. & Isacsson, U. & Dawson, A. 2000b.
damaged locations were much stiffer than the ones
State of the Art. II: Permanent strain response
from damaged areas. of unbound aggregates. Journal of Transportation
The permanent deformation results also under- Engineering 126(1): 76–83. Reston: ASCE.
lined a significant discrepancy in damaged and not Seed, H. B. & Chan, C. K. & Lee, C.E. 1962. Resil-
damaged sites: both the angle of mobilization and ience characteristics of subgrade soils and their
of failure from site I were lower than the respective relations to fatigue in asphalt pavements. Pro-
ones from sites IV and V. Sites II and III reached ceedings of International Conference on Structural
the maximum permanent deformation of 0.5% Design of Asphalt Pavements 1: 611–636. Ann
Arbor: University of Michigan.
just after some tens of cycles in the triaxial test Statens Vegvesen 1998. Håndbok 018 Vegbygging.
apparatus. Oslo: Vegdirektoratet.
The overall performance of the damaged sites Uthus, L. & Tutumluer, E. & Horvli, I. & Hoff I.
proved to be significantly different from the ones 2007. Influence of grain shape and surface texture
which did not undergo distresses. on the deformation properties of unbound aggre-
However, the differences in stiffness and resist- gates in pavements. International Journal of Pave-
ance against permanent deformation in the base ments 6(1). United Kingdom: Abingdon.
Uthus, L. 2009. Effect of grading and moisture on
materials are not likely to represent the only the deformation properties of unbound granular
explanation for the distresses observed. The aggregates. 8th International Conference on the
research could be extended by inspecting more Bearing Capacity of Roads, Railways and Airfields
areas of the road and testing the fatigue prop- 8: 167–177. Rotterdam: Balkema.
erties of the asphalt mix. The estimation of the Werkmeister, S. & Dawson, A. & Wellner, F. 2001.
stress and strain levels for the total structure Permanent deformation behaviour of granular
could be a basis for deciding on the best rehabili- materials and the shakedown concept. Transpor-
tation Research Record 1757: 75–81. Washington
tation strategy.
DC: Highway Research Board.

108
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

An investigation into relevance between fractal dimension and dynamic

behavior of pure granular materials

S. Altun, A. Sezer, A.B. Göktepe, T. Günay & P. Ahmedzade

Department of Civil Engineering, Ege University, Bornova, İzmir, Turkey

ABSTRACT: This study presents an investigation into the effect of particle shape on dynamic behav-
ior of granular sub grade materials. The properties of flexible pavement base materials were analyzed
using appropriate geotechnical tests and analysis. Additionally, finite element analyses were carried out to
find out response of these materials to simulated standard traffic loads. Lastly, the variation of different
fractal parameters on dynamic response of particular media was investigated. As a consequence, it can
be stressed that fractal dimension is a good identifier for evaluation of dynamic response of base course
layers of similar grain size distributions.

1 INTRODUCTION expresses a self-similarity of the shape of an object

that is repeated at different scales. Although the
Road pavements are subjected to greater stresses due shape of soils may be expressed with parameters
to heavy traffic and higher tire pressures. Problems including sphericity, roundness, shape and rough-
related to fracture; permanent deformation and ness; which seem to be independent from each other;
surface wear are often being reported. On the other the fractal dimension is a single identifier of all these
hand, soils exhibit extremely different engineering parameters (Gori and Mari 2001). Millan et al.
behaviours due to their heterogeneous, anisotropic, (2003) underlined that grain size distributions may
three-phased, particulate, and initial conditions exhibit different behaviours with respect to soil tex-
dependent characteristics. It is therefore a quite ture characterized by fractal geometry. In this con-
complex task to simulate any soil behaviour with a text, the authors made the definition of two scaling
certain precision, which is assumed to be a very hard domains with a piecewise fractal model, analyzing
task for many problems in soil science. Nevertheless, the relationships among scaling exponent and soil
there are several engineering properties of earth texture. Wang et al. (2006) applied fractal approach
materials changing due to the particle size as well for the consideration of land use patterns with
as controlling the behaviour that is crucial for the particle size distribution of a soil. The researchers
problem under consideration. Geometric attributes revealed that the fractal dimensions of particle size
(i.e. particle roundness, angularity, and roughness) distribution can be used for the monitoring of soil
as well as the size distribution play significant role degradation induced by land use patterns. Ersahin
in correct characterization of the porous media con- et al. (2006) estimated the cation exchange capacity
sisting of granular materials. Apart from the tradi- and specific surface area of the soils using fractal
tional geometric measures considering straight lines geometry. Dathe et al. (2003) and Bird et al. (2006)
and circles, fractal geometry has emerged as a fun- made investigations for evaluation of soil pores
damentally different tool which is capable of better interacting with soil skeleton by fractal dimensions.
characterizing the particulate media. In this context, Several studies in soil science concerning fractal
geometrical multifractals and entropy-based analy- geometry focused on the evaluation of grain size
ses are the basic types of fractal applications utilized distribution and its use in assessment of physical
for the modelling of soil variability (Millan et al. properties (Hyslip and Vallejo 1997, Posadas et al.
2003, Pechepsky et al. 2006, Wang et al. 2006). 2001, Su et al. 2004, Filguera, et al. 2006).
Fractal geometry is based on a measurement In this study, it is investigated whether relation-
scale which quantifying natural scales by focus- ships underlie among the deformations of a sub
ing on relatively homogeneous components of the grade layer exposed to dynamic loading and aver-
object. Fractal dimension was firstly instructed by age roughness of grains. Furthermore, the depend-
Mandelbrot, and is capable of modelling the shape ence of dynamic properties of sub grade materials
of irregular objects, which are widely encountered in on fractal information is questioned. In this scope,
nature (Mandelbrot, 1975; 1982). The idea behind relationships among parameters obtained from
the fractal dimension concept is that the parameter triaxial tests on pure sands and fractal dimension

109
identifiers are established. Results revealed that, 2.2 Line-divider method
it is possible and plausible to obtain a knowledge
Line divider method is able to calculate the fractal
of soil dynamic properties using fractal analysis,
dimension of objects individually. As can be seen in
which is a laboratory study without expense.
Figure 1, Mandelbrot proposed that if lines of equal
step lengths ( ) are used to form the inscribed poly-
gon of a particle, which may be expressed as:
2 METHODS FOR DETERMINATION
OF FRACTAL DIMENSION
P ( ) n1− DR (5)
It should be noted that, while the fractal dimension
has many definitions, the most popular and easily where P( ) is the total length of polygon length
applied methods can be enumerated as the area- obtained using a step length or “yardstick” ; and
perimeter, line divider, and box-counting methods DR is the roughness fractal dimension of the soil in
(Kaye, 1989; Xie, 1993; Hyslip and Vallejo, 1997). question. The fastest application of the method can
be performed on digitized images which the yardstick
size is decreased hierarchically. Plotting the length of
2.1 The area-perimeter method polygon against yardstick size and drawing the best-
The area-perimeter method is known as the easiest fit line with a slope m, the fractal dimension equals:
application of fractal geometry. After taking the
digital images of a reasonable number attaining a DR 1 m (6)
statistical meaningfulness, the area and perimeter
of the particles are measured by image analysis. As a consequence, it is worth mentioning here
In this manner, Mandelbrot’s proposal of “linear that the while line divider method calculates the
ratio of extensions” is utilized: fractal dimension for every particle, the area-
perimeter method results a single fractal dimen-
P1 DR sion for the whole of the measured particles. In
c= (1) addition, for images of acceptable resolution, line
A0.5
divider method enable to determine the roughness
where, c is a constant value for the fractal shape better with the use of relatively small yardstick
possessing similarity, P is the perimeter of the par- sizes. It is required that the step lengths must not
ticle, and A is the area of the particle and DR is the be greater than 0.3 times the maximum particle
mean roughness fractal dimension of the particle. diameter (Hyslip and Vallejo, 1997).
Taking the logarithm of the two sides of the Eq.1,
the following equation is obtained: 2.3 Box counting method
The fractal dimension calculated by box count-
1
DR
log P (
log c A ) (2) ing method is the fragmentation fractal dimension,
which is related to the particle size distribution, frag-
mentation process, and hydraulic properties. The
Then, making necessary arrangements, method is based on the idea that number of particles
having a dimension greater than a specific measure
DR is exponentially related with the fractal dimension:
log (log c log A) (3)
2
t• ¥." %o"

I\ I
100
and, Eq.4 is finally obtained: I
' Con
···· ·Medium
- •fine
80
DR
log
2
( log A) (4) ,I \ II' I
""~ 60
It is understood from Eq.4 that the slope of
the best-fit line belonging to plot of particle areas
li:
=
~ 40
\\ \

against the particle perimeter equals to 2/frac-

tal dimension. However, this simple model is not
~

20
I I

II
\\ \
capable of determining fractal dimension for each
·'!~!!_U~I
particle individually. In essence, in area-perimeter t"INt:

10 0. 1 0 .0 1
method, the fractal dimension is determined for Grain size (mm)
the whole group of particles under investigation,
which is an advantage of the line-divider method. Figure 1. Grain size distributions of the materials.

110
N (X x ) kx − DF (7)

where, x is a specific linear dimension and X is the

dimension of the particles greater than x, N is the
number of particles, and k is a constant and DF
is the fragmentation fractal dimension (Tyler and
Wheatcraft, 1989; Yang, et al. 1993; Wang, et al.,
2001; Huang and Zhan, 2002). Simplifying Eq. 7,
Figure 2. Sample images of a) S3 b) S3 analyzed.
it can be acknowledged that:

DF m (8) 4 IMAGE ANALYSIS

It should be noted that a similar relationship can In the light of the fractal dimensions depicted in the
be set up using the area of the objects instead of a preceding paragraphs, a study concerning the esti-
linear dimension, resulting a fragmentation fractal mation of dynamic analysis of pavement base uti-
dimension of (−2m) (Hyslip and Vallejo, 1997). lizing the fractal dimension of various materials is
Utilizing the brief theories depicted above, it is conceived. High resolution images of minimum 100
investigated whether these fractal parameters are in sand grains belonging to each sand type were taken
relationship with the dynamical properties of sub via a stereo microscope. Later, information concern-
grade sands. In order to accomplish fractal analyses ing the shape of the grains is obtained by image
on a considerable number of sand particles, auto- analysis (Matlab, 2006; Qwin, 2003). Sample image
mated image processing approach should be preferred analysis is given in Figure 2a. After taking the images
because of its fast and accurate data mining ability. of a known scale, the images are instructed to Qwin
software. By thresholding the pixels of interest, the
undetected pixels inside the boundaries of the grains
3 MATERIALS AND METHODS are filled. Making successive erosion/dilation opera-
tions, a binary image comprising thresholded pixels
At the first part of the experimental part of this of interest is obtained (Fig. 2b). The images analyses
study, the sands are sieved to predetermined particle employed on these materials resulted a knowledge
size distributions. The sieving process on sands from belonging to many topographical properties of soil
five different origins led to three soil classes: Coarse, grains like area, perimeter, etc. Therefore, with the aid
medium and fine sand soils of SP class according to of the computed values of shape features; it can be
Unified Soil Classification System, which equals a revealed that these soils are ready for fractal analysis.
total of 15 soils. Afterwards, image analyses are per-
formed on sands of different origin in order to find
the different fractal dimension of the soils. Later, 5 LABORATORY TESTS
laboratory triaxial compression tests are performed
on 15 sands which are used as base material in order Triaxial compression testis employed on wet speci-
to obtain the strength properties of sands. mens in order to obtain the stress-strain and shear
Sands from five different origins are obtained strength parameters of the base materials. Triaxial
from Western Anatolia. Three crushed rock and compression tests are performed on 15 types of to
two natural soils are sieved into three soil classes, find the modulus of elasticity and the internal fric-
namely, coarse (S1-S5), medium (S6-S10) and fine tion angle of these materials. The tests have been
(S11-S15) sands. The first three sands of each group carried out under isotropic and undrained condi-
are crushed rocks (limestone, basalt and andesit, tions. The modulus of elasticity obtained from
and the remainder two soils are natural soils (river triaxial compression tests are depicted in Table 1.
and shore sands). Sands obtained from different ori- It should be expounded that as the fractal dimen-
gins ensure diverse fragmentation surfaces, which is sion increases, the elasticity index of soil increases.
the main reason of the alteration in grain particle The changes in elasticity index of coarse, medium
shape. Besides, sands are brought into same grain and fine sands are 20.4, 31.8 and 29.2% respec-
size distributions for minimizing the effect of grain tively. This is not an unexpected behaviour since
size distribution on the mechanical and dynamic the strength parameters should develop as the
behaviour of these soils. The grain size distributions fractal dimension increase. When sands of differ-
of coarse, medium and fine limestone sands are ent fineness are evaluated separately, the isotropic
given in Figure 1 as an example. While the uniform- undrained test results depicted in Table 1 demon-
ity coefficient of coarse, medium and fine sands strated decreases up to 3–5° in the internal friction
range between 1.55 and 1.67; the coefficient of cur- angle of these sands, which is the sole Mohr-
vature range between 0.90 and 1.03; respectively. Coulomb shear strength parameter of these soils.

111
Table 1. Base Layer material parameters obtained from analyses. Due to the symmetry of the geometry
laboratory tests. and load distributed around the horizontal x axis,
a semi-infinite half space model was adopted. The
Unit Modulus of Poisson Friction thickness of the base layer was changed in the
Sand Weight Elasticity E Ratio Angle
Type (kN/m3) (MPa) ν φ(˚)
model for evaluation of the effect of thickness on
results obtained from dynamic analysis (Figure 3)
S1 17.1 81.6 0.30 41 In PLAXIS, the geometry is modelled with an
S2 16.8 75.7 0.30 39 axisymmetric model using 15-node elements for
S3 16.6 71.1 0.30 38 modelling pavement layers. A three-layered system
S4 16.7 73.7 0.30 37 representing pavement structure and consisting of
S5 16.7 72.3 0.30 38 asphaltic, base and subgrade layers was chosen.
S6 16.5 76.2 0.30 39 Absorbent boundary conditions for dynamic load-
S7 16.3 72.9 0.30 38 ing were selected in order to avoid wave reflection
S8 16.0 66.3 0.30 35 effects. Absorbent boundaries were applied at bot-
S9 15.8 61.5 0.30 35 tom and right-hand vertical boundaries. The pulse
S10 15.4 62.3 0.30 36 load distributed on a radius of 0.15 m was used to
S11 15.0 67.7 0.30 37 represent the transient haversine traffic load. The
S12 14.5 64.3 0.30 35 load starts at t = 0.052 sec and ends at t = 0.076 sec
S13 18.0 56.7 0.30 37 (Al-Khoury et al, 2001).
S14 16.7 58.8 0.30 35 In this scope, ESAL (one side dual wheels,
S15 16.0 57.6 0.30 34 330 mm apart, and supporting a load of 50 kN with
the tire pressure of 700 kPa) was applied in the anal-
yses. It is considered that the load was transferred
6 FINITE ELEMENT ANALYSIS to the pavement surface through a contact pressure
of a single tire. In the 2D finite element analysis, the
In the literature, two or three-dimensional finite pulse load was distributed on a plate with a radius of
element studies were carried out to investigate 0.15 m. The properties of constituting layers, which
the mechanical behavior of pavement systems were collected by experimental studies, are given in
due to the applied traffic loading (Zaghloul and Fig. 3. The behavior of the transient applied load
White 1993, Huang 1993, Tutumluer 1997, Saad is shown in Fig. 4. Finally, absorbent boundaries
et al. 2005, 2006). In these studies, several mate- and null material damping were taken into account
rial models, namely elastoplastic strain hardening, in the calculations. In Fig. 5, time-domain deflec-
elastic isotropic, and different traffic loading con- tion curves are illustrated for different points in the
ditions were used to estimate the pavement system calculation area. In addition, vertical displacement
performance more accurately, considering a wide variations due to dynamic load applications are
range of constitutive laws for pavement materials. illustrated using surface plots from Fig. 6.
In this study, dynamic finite element analyses by
use of the pavement, granular base and subgrade
with different material models (strain harden-
ing, elastic isotropic) were carried out to investi-
gate the effects of base thickness and the material
properties on performance of the layered pave-
ment structure. In selection of the material param-
eters, data and structure in studies of Brinkgreve
(2005) and Surarak et al. (2012) were determined
as guides The dynamic 2D finite element analyses
were performed by PLAXIS software, which is a
widely-used finite element code for soils and rocks.
Finally, the fractal properties of base materials
were compared and correlated with performance
of pavement structure determined by finite ele-
ment analyses and geotechnical studies.

6.1 Model geometry and loading system

In order to simulate the traffic load and soil pro-
file, ESAL (Equivalent Single Axle Loading) and Figure 3. Pavement structure used in Finite Element
2D model geometry were used in the finite element Modelling.

112
Load (kN) Table 2. Mechanical properties of materials.
60 r ······· ············
Properties

45 !-················· Layer Thickness (cm) γ E (MPa) υ

Asphalt 5 1.0690 28.000 0.30

30 !-················· Base 20–40 14.5–18.0 56.7–81.6 0.30
Subgrade 2000 1.0294 50.6 0.40
15 !-·················

subgrade material was obtained from an earlier

study of Desai and Siriwardane (1984).
0.05 It is known that using an elastoplastic model for
Dynamic time(s)
representing the asphalt concrete is not sufficient.
Figure 4. Transient input loading curve (After Al-
In addition, though the viscoelastic model best
Khoury et al., 2001). simulates the actual performance of the asphalt
concrete, it needs many laboratory tests to iden-
tify its characteristics. Nevertheless, elastoplastic
+ model is considered to be viable for simplification
purposes. Linear elastic behavior was selected for
~

~ ""\
\
AC layer. This is an acceptable assumption con-
sidering short period of the traffic load wave. The
model data belonging to the layered structure is
.
I
\ summarized in Table 2.
In the FEM analyses, elastoplastic soil model
\-
-- - t-
(namely hardening soil model) was utilized to
characterize base the material behavior. In addi-
tion to its abilities in simulation of non-tensioning
1\ response of granular materials, elastoplastic soil
.. .. ~ .. .. .. .. ~

' ..----.
.. models mimic significant characteristics of the
granular materials including elastic response at
lower stress levels and plastic response at higher
Figure 5. Time domain deflection curves.
stress levels (Baladi and Chen 1985). The subgrade
layer was modeled using hardening soil model
(Brinkgreve, 2005).

7 RESULTS AND DISCUSSION

After fractal and mechanical properties of 15

sand types were determined, different material
models for base layer with varying thickness were
applied, where linear elastic and hardening soil
models were utilized for asphaltic and subgrade
layers.

7.1 Influence of base course material attributes

and thickness
Figure 6. Vertical displacements due to dynamic
loading. Fifteen types of sands and two different mate-
rial models were selected to employ finite element
analyses on layered pavement structure subjected to
6.2 Material behavior and modelling
dynamic loading conditions. Model results revealed
The material data of AC were obtained from a pre- that base material quality, which can be quantified
vious study (Zaghloul and White 1993). The mate- by internal friction angle, had a considerable influ-
rial properties of granular base layers required for ence on the strain properties of pavements. The
the simulations were determined by laboratory influences of the granular base layer thickness and
testing (Table 3). The inputs for a typical sand soil type on vertical surface deflection at the bottom

113
of AC layer was determined at t = 0.064 sec (the time Soillype
of application of peak load). As can be derived from
the results, the vertical surface deflection decreased
'.:: I
as the base thickness increased. The quality of the
subgrade material has relatively less influence on sur-
face deflections. This conclusion agrees with those
· I.OE-01

·UE-(11
I
obtained in previous studies (Huang 1993; Dawson
and Plaistow 1996).

7.2 Dependence of FEM outcomes on fractal

dimension identifiers Soil lype
S8 S9 SlOSllSl2SllSl4 S lS
As indicated before, fractal dimension is a measure
of the irregularity of a particle, which may play
crucial roles in interlocking and friction mecha-
nisms. Therefore, increasing fractal dimension may
play a significant role in the strength characteristics
Ji
',',~: II
·l.OE-(11

-I.SE-01

·2.0E-(11
t
of the particulate media. In this context, it would
be better to investigate variations of several fractal
dimension identifiers with surface deflections.
Figure 7 shows the effect of model selection and
base course layer thickness on surface deflection Soillype

of the layered pavement structure. It is obvious

that as the base thickness increases, deflections
are lower, regardless of the model selection. As
'.: : I
l.OE-(11

expected, deflections obtained by using strain 'i" : U E-(11

hardening model was lower than those obtained '

J · 2.0E-(11

by elastic isotropic model. Moreover, coarser soils

resulted in smaller deflections. It is underlined
in the preceding paragraphs that the roughness
parameter, namely the fractal dimension, should
affect the stress-strain behavior of granular media. Soil lypt
Sl Sl Sl S4 SS Sl> S7 S8 S9 SIOSll SJ2Sl3Sl4SJS
Meanwhile, it is initially questioned which fractal
dimension is the best identifier for assessment of
dynamic soil behavior. The Dtot parameter, which
is the best identifier of the strength parameter
variations, was also found to be the best identifier
of the deflection changes in particulate media. A
closer look to Figure 7 and the Dtot values listed in
Table 1 revealed that, regardless of the base thick-
ness and selected model for base layer, the Δzmax is
inversely related with the Dtot parameter. In detail,
comparing these values with the remaining fractal Figure 7. The variation of surface deflection (Δz,max)
dimension definitions, it can be figured out that considering 15 soils in material models: a) Elastic iso-
the strain properties of the sands are not barely tropic, base thickness 200 mm. b) Elastic isotropic, base
affected by the textural or structural fractal dimen- thickness 400 mm. c) Strain hardening, base thickness
sion of the soil: In other words, neither micro nor 200 mm. d) Strain hardening, base thickness 400 mm.
macro shape characteristics solely control the
strain properties of the soils. (R2) values ranging between 0.72 and 0.94. Moreo-
The Dtot parameter, comprising the shape ver, it is evident that Dtot of coarse grained soils is
attributes in both scales is the most effective effective on interlocking behavior. This is observed
parameter on strain properties. Therefore, refer- from increasing internal friction angles with
ring to Figure 7 and the fractal dimension values increasing fractal dimension, which caused better
in Table 1, Δzmax values are in agreement with the interlocking of discrete particles in medium, and
total fractal dimension values, which were approved then accompanied by decreasing strain levels. The
by the inversely proportional behaviors observed changes in fractal dimension of soils can be respon-
between vertical deflections and the total fractal sible for abrupt changes in strength parameters of
dimension values with coefficient of determination these soils, which were supported by remarkable

114
 Bird, N., Díaz, M.C., Saa, A. and Tarquis, A.M. Frac-
tal and multifractal analysis of pore images of soil.
J.Hydr., 322(1-4) (2006) 211-219.
Brinkgreve, R. Selection of soil models and parameters
for geotechnical engineering application. Soil Const.
Mod. (2005) 69-98.
Carbone, A., Chiaia, B.M., Frigo, B. and Türk, C. Snow
metamorphism: A fractal approach,, Physical Review
E 82(2010), 036103.
Chen, W.F. and Baladi, G.Y., Soil plasticity: Theory and
implementation. (Elsevier, Amsterdam, 1985), pp.
23-62.
Dathe, A., Eins, S., Niemeyer, J. and Gerold, G. The
Figure 8. Total fractal dimension-internal friction surface fractal dimension of the soil–pore interface
angle-surface deflection variation in coarse sands. Mate- as measured by image analysis. Geoderma, 103 (1-2)
rial model layer thickness 200 mm. (2003) 203-229.
Desai, C.S. and Siriwardane, H. J. Constitutive laws for
engineering materials: with emphasis on geologic
changes in deflections. The effect of layer thickness materials (Prentice Hall, Inc., New Jersey, 1984).
on variation of total fractal dimension-internal fric- Ersahin, S., Gunal, H., Kutlu, T. Yetgin, B. and Coban, S.
tion angle−surface deflection values is given in Fig- Estimating surface area and cation exchange capacity
in soils using fractal dimension of particle-size distri-
ure 8. Similar behaviors were observed for medium bution. Geoderma, 134(3-4) (2006) 588-597.
and fine sands. Changes in deflection values show Filguera, R.R., Fournier, L.L., Cerisola, L.I., Gelati, P.
that the Δzmax values are strongly influenced by vari- and García, M.G. Particle size distribution in soils:
ations in Dtot values, up to 45 and 47%, when elastic A critical study of the model validation. Geoderma,
isotropic base material model is preferred. 134(3-4) (2006) 327-334.
Fractalyse. Computer code developed by Théma Institute
(Gillés Vuidel) (2002) www.fractalyse.com.
8 CONCLUSIONS Gori U. and Mari, M. The correlation between the frac-
tal dimension and internal friction angle of different
granular materials. Soils Found., 41(6) (2001) 17-23.
In this study, the relationships among fractal
Huang, Y. H. Pavement analysis and design (Prentice–
identifiers of a base course of a layered pavement Hall, New Jersey, 1993), p.275.
structure and its performance of under transient Huang, G.H. and Zhan, W.H. Fractal property of soil
loading conditions are investigated. The total particle size distribution and its application. Acta Ped.
fractal dimension (Dtot) parameter, was found Sin., 39 (2002) 490–497.
to be highly correlated with the shear strength Hyslip, J.P. and Vallejo, L.E. Fractal analysis and the
and deflections. Micro structural analyses on dif- roughness distribution of granular materials. Engi-
ferent sands proved that total fractal dimension neering Geology, 48(3-4) (1997) 231-244.
(Dtot) parameter is good in estimation of strain Kaye. B.H. A random walk through fractal dimensions
behavior of sands subjected to dynamic loading. (VCH, New York, 1989), p.427.
Kindratenko, V.Development and application of image
The internal friction angles are ascertained to be
analysis techniques for identification and classifica-
directly proportional with Dtot, Deflection values tion of microscopic particles, (1997) MSc Thesis, Uni-
are inversely proportional to Dtot. Moreover, it was versity of Antwerp, Belgium.
found that the rate of change in deflection is inde- Mandelbrot. B.B. Les objects fractals: forme, hazard et
pendent from that of base course thickness. Nev- dimension. (Flammarion, Paris, 1975), p.190.
ertheless, rate of changes in fractal dimension and Mandelbrot, B.B. The fractal geometry of nature (W.H.
deflections are not correlated with each other. In Freeman, New York, 1982), p.109.
medium sands, deflections are found to be sensitive MATLAB, Image Processing Toolbox, (2007)
to changes in fractal dimension. In forthcoming Mathworks.
studies, the results of this study can be revisited or Millán, H., González-Posada, M., Aguilar, M.,
Domínguez, J. and Céspedes, L. On the fractal scal-
broadened to establish relationships between frac- ing of soil data. Particle-size distributions. Geoderma,
tal dimension and deflection values. 117(1-2) (2003) 117-128.
Pechepsky, Y., Perfect, E. and Martin, M.A. Fractal
geometry applied to soil and related hierarchical sys-
REFERENCES tems. Geoderma134(3-4) (2006) 237-239.
Posadas, A.N.D., Giménez, D., Bittelli, M., Vaz, C.M.P.
Al-Khoury, R., Scarpas, A., Kasbergen, C. and Blaau- and Flury, M. Multifractal characterization of soil
wendraad, J. Spectral element technique for efficient particle-size distributions. Soil Sci. Soc. Am. J.65(5)
parameter identification of layered media. I. Forward (2001) 1361-1367.
calculation. Int. J. Sol. Struct., 38(9) (2001) 1605-1623. Qwin Reference Guide, 2003, Leica Microsystems.

115
Saad, B., Mitri, H. and Poorooshasb, H., 3D FE Analysis Wang, X., Li, M-H., Liu, S. and Liu. G. Fractal charac-
of flexible pavement with geosynthetic reinforcement. teristics of soils under different land-use patterns in
ASCE J. Trans. Eng., 131(6) (2005) 460–469. the arid and semiarid regions of the Tibetan Plateau,
Su, Y.Y., Zhao, H.L., Zhao, W.Z. and Zhang, T.H. Fractal China, Geoderma, 134(1-2) (2006) 56-61.
features of soil particle size distribution and the impli- Xie, H. 1993. Fractals in Rock Mechanics, (Balkema,
cation for indicating desertification. Geoderma122(1) Rotterdam, 1993), p. 41.
(2004) 43-49. Zaghloul, S.M. and White, T.D. Use of a three-dimen-
Surarak, C., Likitlersuang, S., Wanatowski, D., Balasub- sional, dynamic finite element program for analysis of
ramaniam, A., Oh, E., Guan H., Stiffness and strength flexible pavement. Transp. Res. Rec., 1388 (1993) 60–69.
parameters for hardening soil model of soft and stiff Zhang, Q., Yin, G., Fan, X., Wei, Z., Wang, W. and Nie,
Bangkok clays, Soils Found., 52(4) (2012) 682-697. W. Loading capacity and deformation characteristics
Tutumluer, E. and Thompson, M. R. Anisotropic mod- of tailings based on a fractal geometrical analysis of
eling of granular bases in flexible pavements. Trans. the particle microstructure. Minerals, 5 (2015) 86-103.
Res. Rec., 1577 (1997) 18–26.
Turk, C., Carbone, A., Chiaia, B.M. Fractal heterogene-
ous media, Physical Review E 81 (2010) 026706.

116
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Experimental characterization of unbound granular materials

subjected to high harmonic loads

G. Canon Falla, S. Leischner, F. Wellner & T. Spanier

Institute of Urban and Pavement Engineering, Technische Universität Dresden, Dresden, Germany

ABSTRACT: Unbound Granular Materials (UGM), used in base and subbase layers, play an impor-
tant role in the structural performance of low volume roads with thin asphalt surface layers. They show
complex stress dependent elasto-plastic behavior under cyclic loading. A proper understanding of the
elastic and permanent deformation behavior of UGM is crucial to assess their suitability for the particular
service. Full understanding of the material behavior is only possible by means of sophisticated test proce-
dures such as the repeated load triaxial test. This testing method is quite complicated and the human and
technological resources needed usually exceeds the budget of low volume road projects. Therefore, there
is a need for simple and economical test methods to characterize the mechanical behavior of UGM under
the action of high traffic loads. The paper shows the results of a test campaign using two different testing
methods: a simple repeated load volumetric compression test and a more advanced repeated load triaxial
test. The results of both tests showed a strong dependency of the bearing capacity and the permanent
deformation behavior on the amount of water and stress level. The tests results are also used to develop a
classification method for UGM.

1 INTRODUCTION of pavements with less than 50 mm asphalt thickness

proved to be satisfactory provided that the design is
In high volume roads the thicknesses of the bitumi- performed considering the non-linear behavior of the
nous layers are large enough to ensure that the traf- UGMs in the base course. It was concluded that the
fic loads are carried by the asphalt layers. On the granular materials in the vicinity of the load must be
other hand, in rural areas, where the traffic volume modeled using non-linear stress dependent models.
is less of 500 ADT, the thickness of the asphalt Despite the above mentioned findings, the design
layer is minimum (25 mm to 60 mm thickness) of thin flexible pavements worldwide is almost
and the structural function falls on the Unbound always done empirically and the cross sections and
Granular Materials (UGM) of the basecourse. construction materials of these roads are based
The upper bituminous surface acts in this case as a on country specific experience in high-standard
waterproofing layer with relatively small contribu- asphalt pavements. Such empirically based design
tion to bending stiffness (Arnold, Werkmeister, & methods have the disadvantage in that the materi-
Alabaster (2008), Werkmeister, Steven, & Alabaster als and layer thicknesses are selected in accordance
(2006), Rahman & Erlingsson (2014)). with inflexible predetermined design inputs and
The European road network consists to a large extrapolation is required when the input param-
extent of thin flexible pavements. This type of eters lay beyond the original interference space.
structure has been widely adopted because it is cost Nowadays, this is of great concern because of an
effective, requires less energy resources and it is fast increased growth rate in vehicles, use of new materi-
to build compared with traditional thick asphalt als, and changing environmental conditions.
pavements. Thin flexible pavements are widely It is therefore, that the UGM used in the base-
used in low volume road networks of scarcely pop- course of thin asphalt pavements, especially the
ulated areas. These roads are vital to provide the materials at the the vicinity of the load, shall be
local permanent population and associated rural selected based on structural criteria rather than
industry an all-year round transportation solution. on consensus properties. This requirements brings
Furthermore, a well maintained road network is together the demand of new testing techniques to
essential in attracting tourist to these areas. characterize the strength and stiffness of UGM as
Recently, a research project (Wellner, well as experimental procedures to identify their
Werkmeister, & Falla (2014)) aiming to understand long term plastic behavior.
the response of thin asphalt pavements under design Due to the lack of resources available in low
conditions of Germany showed that the durability traffic volume projects, the technology used in the

117
characterization of the properties of UGM is very supposed linear elastic, with Poisson’s ratio of 0,4
limited. In most cases the load bearing capacity is and E-modulus of 113 MPa. The basecourse behav-
obtained indirectly through empirical based tests ior was modeled using the modified universal model
such as the California Bearing Test value. The Cali- for UGM (Equation 1). The model parameters
fornia bearing test is a relatively simple laboratory were determined from the results of RLT tests on
test from which a unique value is obtained which a sandy gravel with maximum grain size of 32 mm
represents the rigidity of the material. In the opin- and moisture content of 4,1%.
ion of the authors, the CBR is a useful test that can
be used as an initial filter in the selection process ⎛ θ ⎞ ⎛τ ⎞
k2 k3

of granular materials, however, when the structural Mr k1 ⎜ ⎟ ⎜ oct + 1⎟ (1)

function of the granular layer is vital, it is neces- ⎝ pa ⎠ ⎝ pa ⎠
sary to perform a more comprehensive laboratory
test campaign including test procedures under where Mr is the resilient modulus of the UGM, θ
cyclic loading as the ones described in this paper. is the bulk stress, τoct is the octahedral shear stress,
pa is a constant stress of 100 kPa and k1, k2, k3 are
material dependent variables.
2 AIM OF THE RESEARCH A stone mastic asphalt with a polymer modied
bitumen was selected for the asphalt layer. The
Based on the need for laboratory tests that provide modulus of elasticity was determined using the
reliable results to characterize the elasto-plastic indirect tensile test (E = 3500 MPa for a surface
behavior of UGM, this paper presents the out- temperature of 32.5C and 5 Hz).
come of a research project carried out at The Tech- The resulting stress regimes in the unbound gran-
nische Universität Dresden, Germany. The project ular layers are illustrated in Figure 2. The results
(Wellner, Werkmeister, & Falla (2014)) was aimed show that the stress distribution in the basecourse
at developing a methodology to characterize the of the thin surfaced pavements (structures 2 and 3)
response of UGM under high cyclic loading using extend to higher levels in the deviatoric-hydrostatic
performance related laboratory tests. Part of the
project’s goal was to rank UGM of varying qual-
ity, in order to be able to use them adequately in
the base course of thin asphalt pavements.
To provide performance data, the resilient and
permanent strain behavior of different UGM has
been investigated using a volumetric compression
test and a Repeated Load Triaxial (RLT) test.

3 IN PAVEMENT STRESSES
(a) SWr:lure No. l (b) Sir\r:I~~~~:No.'l (c) Stroct~ No.J

Before beginning with the laboratory experiments

it is necessary to determine the stress loci that needs Figure 1. Pavement structures investigated (A is Asphalt,
to be covered. For that reason the stress regimes BC is base course, SB is subbase).
in the unbound granular layer of two thin asphalt
·600.0
pavements (10 mm and 40 mm of asphalt) and one
thick asphalt pavement (100 mm of asphalt) were -500.0
calculated numerically using the in-house finite ele-
ment code EasyFEM. The thickness of the pave- -400.0

ment structures are depicted in Figure 1. ! -300.0

The stresses within the pavement were calculated
for a uniformly distributed wheel contact pressure -200.0

of 750 kPa over a circular area of 150 mm radius.

This contact pressure corresponds to a wheel load
of 53 kN.
The subgrade was modeled assuming an iso- 0.0 -50.0 ·100.0 -150.0 -200.0
p (kPa)
-250.0 -300.0 -350.0 -400.0

tropic linear elastic material behavior with Poisson’s

c s tructure No. l
ratio of 0,4 and E-modulus of 45 MPa. The sup- <~. Structure No. 2
oStructureNo.3
port provided by the subgrade was chosen follow-
ing the guidelines of the German code of pavement Figure 2. Stress regimes of the analyzed pavement
design, RSTO (2001). The frost protection layer was models.

118
plane (q and p, respectively) compared to tradi- 100

tional pavements. This plot enables us to establish 90 10

testing stress paths to cover the actual stress spec- 80 20

trum of thin asphalt pavements. 70 30

~
~ 60 40
!
! so so "C
Cll
c:
4 MATERIALS i 40 60
~a::
:f 30 70
Three different unbound base materials and one 20 18 80
frost protection material were investigated in the 10 90
2,5
laboratory. The materials were selected based on 100
0,063 O,S 1 2 4 8 16 32
experience on their quality in terms of stiffness and
Grain size [mm]
susceptibility to permanent deformation.
100
The base course aggregates, labeled mixtures A, --o-- • Gradin2 band
90 ll ! 10
B and C, have been widely used in Germany in the -<- S•l•ct•d RradlnR lin• ~,~ ·''
..0
j
88

80 20
base layers of different pavement structures. The .l ···' 73
70 30
base materials were taken from three different Saxon Mixture D
~ ,ll
~
60 61 40
quarries located close to Dresden. The forth mate- i ,./' i
rial, labeled mixture D, is a variation of mixture B. !'
so 51[
J>
so !
"C
,,l
~Cl.
Cll
40 I 60 c:
The modification consists in adjusting the particle
~a::
size distribution to bring it to the acceptable grading
zone for frost protection purposes. Table 1 shows
30
20 /' .•
21,7
31.9 70
80

some properties of the selected aggregates. 10

3,3
l' 5~; ·;,;·;;:: 90
100
The materials were tested within the grading 0,063 o,s 1 2 4 8 16 32
limits of the German specification of UGM for Grain size [mm]

road constructions, TL SoB-StB (2004). The mean

curve of the reference band was used as the design Figure 3. Grading lines of tested materials.
grading for the base course materials. For the
frost protection material a grading line close the
5 TEST CAMPAIGN
lower limit of the reference band was chosen. The
design grain size distribution curves and reference
5.1 Repeated load triaxial tests
bands are illustrated in Figure 3. The maximum
dry density and optimum moisture content were In the initial phase of the project, the resilient and
established by means of standard Proctor compac- permanent deformation behavior of the materi-
tion tests. The tests were conducted following the als were investigated based the on results of RLT
German standard DIN 18127 (1993). tests. All tests were performed with the RLT test
apparatus of the Institute of Urban and Pavement
Engineering at TU Dresden (Figure 4, right). The
Table 1. Materials tested.
test samples are cylindrical specimens of 150 mm
diameter and 300 mm high (Figure 4, left). The large
diameter of the specimen allows to incorporate the
whole spectrum of aggregate sizes into the test. The
materials were tested at target dry density of 97% of
Property Mixture A Mixture B Mixture C Mixture 0
the maximum dry density and moisture content var-
ying between 70% and 90% of the Optimum Water
Natural Crushed Crushed
Type sandy- rock Gran-
Crushed rock
rock Gran- Content (OWC). These target values were based on
Andesite
gravel odiorite odiorite expected in-situ values. Every specimen was sub-
Grain size
[mm] 0/32 0/32 0/32 0/16 jected to a test program that is divided into two
stages named post-compaction test and resilient test.
Maximum
dry density 2,20 2,17 2,10 2,20
[g/cm 3] 5.1.1 Post-compaction test
Optimum The samples were initially subject to 20.000 load
moisture
content
6,00 5,80 12,1 6,7 cycles at 10 Hz with a reference stress level of
[m-%] constant confining pressure of 150 kPa and cyclic
Fines con-
stress deviator of 300 kPa. Figure 5 shows the
tent
2,50 2,50 2,50 3,3
stress-strain behavior observed in the test of Mix-
{<0,063mm)
[%)
ture B at 70% OWC. The results show a large per-
manent deformation at the beginning of the test

119
 -600.0

-500.0

-400.0

-300.0
<T

-200.0

-100.0

0.0
0.0 -100.0 -200.0 -300.0 -400.0 -500.0
p (kPa)
+ EN Low stress level [EN 13286-7, 2004]
• EN High stress levels [EN 13286-7, 2004]
... Stress levels after Numrich [Numrich, 2004]
-<>-Chosen stress sequence

Figure 4. Triaxial test machine and specimen.

Figure 6. Stress levels used in the resilient test.
400,0
First 10 load cvcles Last 10 load cycles
350,0 pressure and sinus shaped cyclic axial stresses at
300,0
1 Hz were applied. The cell pressure was set at lev-
els of 100, 130, 150, 180, 210 and 250 kPa and the
σ
- 250,0
axial stress ratio σ 13 was varied from 1 to 3.
~
t; In a RLT test with constant confining pressure
200,0
·.::
0 the resilient modulus is determined by dividing the
150,0
~ cyclic deviator stress by the resilient axial strain.
100,0 The resilient modulus was calculated with Equa-
50,0
tion 2 from the last 10 cycles of each loading step.
Due to difficulties at experimental recording of
0,0
0,00 1,00 2,00 3,00 4,00 5,00
radial deformation in the triaxial testing appara-
tus, the Poissons ratio was not calculated. Instead,
Deviatoric strain [thousandth]
a constant Poissons ratio of 0,4 was assumed.
Figure 5. Typical stress—strain diagram during the
post-compaction test. σ1 − σ 3
Mr = (2)
Δ ∈1
(first 10 load cycles), due to post compaction phe-
nomena. At the end of the test (last 10 load cycles) where Mr is the resilient modulus of the UGM, σ1
the behavior of the material is nearly elastic. is the vertical stress, σ3 is the cell pressure, Δ ∈1 is a
increment of the vertical strain.
5.1.2 Resilient test Figure 7 shows the results of the resilient test on
After the post-compaction test, a resilient test was Mixture B at 70% OWC. Qualitatively, the results
performed in order to characterize the material of the other mixtures are similar. It is observed that
resilient modulus for varying stress combinations. the effect of the deviator stress in the material stiff-
The testing stress sequence was based on the fol- ness is less pronounced compared with the effect
lowing criteria: of the confining pressure. High cell pressures leads
to strong compression between aggregates that
• The stress levels are representative of those
results in an increase of the deformation resist-
found in the granular layers of thin asphalt
ance. The degree of dependence on the deviator
pavements.
or shear stress is not so evident. This is also seen
• The loading steps were chosen to avoid excessive
in the results of similar studies conducted by other
permanent deformations
authors, however, there are some contradictions
The tests were conducted at constant confining on this regard. Hicks & Monismith (1971) cited
pressure following the guidelines of the European by Lekarp, Isacsson, & A. Dawson (2000) sug-
Standard for cyclic load testing for unbound gran- gested that the resilient modulus does not depend
ular materials, EN 13286-7 (2004). A modification on the deviator stress provided that excessive plas-
to the standard protocol was adopted in order tic deformation is not occurring. Other authors
to extend to higher levels of hydrostatic stresses. like Uzan (1985) cited by Rondon & Reyes (2008)
Figure 6 shows the adopted testing stress paths observed that the resilient modulus decreases
in comparison with the stress levels provided by slightly with increasing deviator stress under
the European Standard. A series of 42 load steps constant confining pressure. Other researchers
consisting of different levels of constant confining like Kamal, Dawson, Farouki, Hughes, & Sha’at

120
 700 content. This loss of stiffness can be introduced
ro
600
.,.:". -~-·. into Equation 1 by replacing the constant k1 by a

~-
~ 500
moisture dependent function. For small changes in
::J
:;
moisture the parameter k1 decreases linearly with
400
"0
0 increasing moisture. Thus, a moisture dependent
E
c 300
Cell pressure [kPa)
model can be formulated as follows:
~
·v; 200
QJ ......:r-100 -D-130
a::
--150 ---180 ⎛ ⎛ WC ⎞ ⎞ ⎛ θ ⎞ 2 ⎛ τ oct ⎞
k k3

kw1 + kw 2 ⎜ + 1⎟
100
Mr ⎟ (3)
Copt ⎟⎠ ⎠ ⎜⎝ pa ⎟⎠ ⎜⎝ pa
---210 -+-250
⎝ ⎝ WC ⎠
100 200 300 400 500 600
Deviator stress lkPa 1
where WC is the water content, WCopt is the opti-
Figure 7. Results resilient test on Mixture B, at 70% mum water content and kw1, kw2, k2, k3 are material
OWC. dependent variables.
Table 2 shows the model parameters for all the
materials. The parameters were determined by a
(1993) claim that the resilient modulus increases non-linear least square technique that minimizes
with increasing the deviator stress. the error between experimental and numerical data.
The results of the resilient tests are used to find
the parameters of a non-linear material model. For 5.1.3 Material ranking
thin surfaced asphalt pavements, a linear elastic The results of both tests, post-compaction and
assumption is not valid and the UGM in the vicinity resilient, were used to perform a macro-mechan-
of the load must be modeled using non-linear stress ical classification. The classification uses the fol-
dependent models. The importance of this non-lin- lowing two characteristic values representative of
ear behavior on critical stress and strains values is a material stiffness and material long term behavior,
function of the induced stress gradients within the respectively:
pavement and the inherited material non-linearity.
Over the years, many non linear models that • a characteristic elastic modulus (Ec), determined
describe the stress dependent elastic behavior of with Equation 3 using a hydrostatic stress of
UGM used in pavements have been developed. 250 kPa and a deviator stress of 500 kPa.
One of the most widely used model that has been • a characteristic permanent axial strain rate,
proved to be efficient and simple is the Modified determined with the slope of the plastic strain
Universal Model developed by Uzan (Equation 1). curve at the end of the Post-compaction test.
The model includes the effect of mean and devia- Based on this classification philosophy, three
toric stress in the magnitude of the elastic modulus. material qualities are defined (Table 3) in which a
After the stress level, the moisture content is good material is characterized by a high stiffness
the most significant factor affecting the elastic and a low susceptibility to permanent deformation.
response of UGM. Repeated load triaxial tests Figure 8 presents the results of the mechanical
performed by Numrich (2004) with test samples classification of all the materials. The influence of
prepared at constant density but different mois- the water content in the mechanical properties of the
ture contents showed an increase of the resilient materials is clearly seen in the chart. It is observed a
strain with an increase in moisture content. Some general trend in which the resilient modulus decreases
authors attribute this effect to pore water pres- and the permanent strain rate increases when the
sure development. Other authors indicate that the water content increases. This effect is more pro-
excess of water between the aggregates acts as a
lubricant reducing the friction between particles
and this increases the deformation. Table 2. Material model parameters: Moisture depend-
In order to account for the effect of moisture ent modified universal model.
content on the bearing capacity, we propose a slight
refinement to the Modified Universal Model: close ⎛ ⎛ WC ⎞ ⎞ ⎛ θ ⎞ ⎛ τ
k2
⎞
k3

examination of Equation 1 shows that the parame- Er k + k w2 ⎜⎝ WC ⎟⎠ ⎟⎠ ⎜⎝ p ⎟⎠ ⎜⎝ p

oct
+ 1⎟
ter k2 characterizes the stiffening effect due to bulk ⎝
w1

opt a a
⎠
stress and the parameter k3 represents the soften-
ing effect due to shear stress. The parameter k1 rep- Material kw1 kw2 k2 k3
resents the magnitude of the E-modulus. For small
changes of moisture content it was observed that Mixture A 209,03 −23,459 0,6757 −0,0672
the stress stiffening and softening behavior remains Mixture B 201,65 −110,34 0,5923 −0,1946
almost constant. However, the magnitude of the Mixture C 641,71 −701,38 0,4641 0,0056
material stiffness decreases with increase of water Mixture D 159,09 −12,647 0,5976 −0,1219

121
Table 3. Material qualities defined based on the results free radial expansion of the specimen. Analysis of
of post-compaction and resilient tests. the applied vertical stress and the resulting verti-
cal strains gives a measure of the material bulk’s
Characteristic modulus.
Material elasticity Characteristic plastic
quality modulus Ec strain rate (ε p1)rate
The volumetric tests were performed using a
cylindrical mold of 150 mm diameter and 175 mm
Good Ec < 500 [MPa] (ε p1)rate < 30 [‰/106 LW] high installed into a servo hydraulic universal test-
Moderate 500 [MPa] < 30 [‰/106 LW] < (ε p1)rate ing machine of 100 kN. The vertical load was
Ec < 250 [MPa] < 60 [‰/106 LW] measured by a transducer in the loading piston.
Poor 250 [MPa] < Ec 60 [‰/106 LW] < (εp1)rate Three equally spaced LVDTs attached to the
loading plate were used to measure the vertical
deformation. Figure 9 shows the cyclic volumetric
Ch aract eristi c p lasti c st ra in rat e
compression test arrangement at TU Dresden.
['Yoo/ 10 6 LW]
10 20 30 40 50 60 70 80 90 100 110 120 The materials were tested at a target dry density
ro 600 of 97% of the maximum dry density and moisture
i GOOD I MODERATE POOR

::> 500 70% 80% 90% contents of 70% and 80% of the optimum.
::>
"0 70%~~~-- ----- <> Analogous to what has been done with the RLT
0
E 400 D 8o -- - - - - - -6. 90"Ai test, each material undergoes a test program that is
> 70% - -
·u divided into two parts. Initially, the materials were
·~ 300
a:; 80% subjected to a post-compaction test to characterize
·.;:::;
·;:: 200 --o the resistance to volumetric permanent deforma-
- l:l - Mixture A: Sandy Gravel (Zwickau)
tion. Afterwards, a bulk modulus test was done
Q)

l
u
100 - <> - Mixt ure B: Granodiorite (Oberottendorf)
-0 - Mixture C : Andesite (Wurgwit zl aiming to get an indication of the elastic response
at high hydrostatic stress levels.
Ch aracteristic plast ic stra in rat e [%o/106LW]
0 10 20 30 40 50 60 70 80 90 100 110 1 20 5.2.1 Post-compaction test
600
GOOD I MODERATE POOR
Poorly compacted mixes have less resistance to rutting
"'
i ~
500 ()..80%
due to secondary consolidation under traffic. This
mode of rutting is stress history dependent. In prac-
E 4ao 70?. '<> 90"Ai tice this type of rutting is seen as a vertical depression
~
·u
near the wheel path (Figure 10, Roadex (2000)).
~ 300 In order to characterize the propensity of the
a:;
materials to permanent deformation due to a sec-
~ 200
ondary consolidation, a post-compaction test was
~
~ 100 performed. The test consists of applying a vertical
- <> - Mixture D: Granodiorite (FSS)
u load of 550 kPa during 50.000 load cycles at 5 Hz
and measuring the accumulated permanent defor-
Figure 8. Results of mechanical classification. mation after load removal.
Figure 11 shows the development of permanent
strain during the post-compaction tests of mixtures
nounced at high water contents (near the optimum C and D. As expected, the permanent deformation
moisture content) where saturation is approached increases with a higher rate at the beginning of the
and there is pore water pressure development. How- test and it tends to level-off at the end.
ever, the sensitivity of the resilient and permanent The accumulated plastic strain at the end of the
deformation behavior of the material to the amount test is used as a characteristic value to identify the
of water is not the same for all materials. materials that are sensitive to volumetric plastic

5.2 Cyclic volumetric compression test

It was observed in section 3 that large hydrostatic
stresses are present in the granular base course of
thin asphalt pavements. Therefore, obtaining an
estimation of the bulk modulus is very important
for such applications. For this reason, in the second
stage of the project, cyclic volumetric compression
test were carried out.
Volumetric compression tests are performed by
compressing the material inside a metallic cylinder. Figure 9. Cyclic volumetric compression test arrange-
The fixed boundary of the metalic mold prevents ment.

122
 12 .-------------------------------------

D 70%0 WC
c: 10 +--------------r--r--------------------
'jij

~ ~ 8 +-------------~
'tiIV"C
£
! ~ 6 +-------------~
~]
~ ~4
a
;},

Mixture A Mixture C Mixture D

Figure 10. Example of road that exhibits post-compac- Figure 12. Results of the volumetric post-compaction test.
tion rutting.

Δσ υ ool
K= (4)
Δ ∈υ ool
10.00 1 1 1-
<?8.00
where K is the resilient Bulk modulus of the UGM,
~ Δσ υ ol is the amplitude of the volumetric stress and
Δ ∈υ ol is the amplitude of the volumetric strain.
'!:fs.oo• • Figure 13 shows the strain development during
Ci5 4 .00
the test of Mixtures C and D at 70% OWC. It is
J 200
observed that both materials follow a non-linear
stress-strain relationship that can be represented
0.00
0 10,000 20,000 30,000 40 ,000 50,000
by a power function given by Equation 5. Further-
Number of Load Cycles, N more, it is clear that the magnitude of the Bulk
-.-Mixture e, 80% OWe -o- Mixture e, 70% OWe modulus as well as the stress hardening behavior of
....... Mixture D at 80% OWe -Q- Mixture D at 70"A; OWe
Mixture D are higher than the ones of Mixture C.
Figure 11. Results of volumetric post-compaction test.
Mixtures C and D. σ υ l = a Δ ∈υa2ol (5)

where a1, a2 are material parameters.

deformation. Figure 12 shows the results of mix-
tures A, C and D. The data of the post-compac- 5.2.3 Material selection critera
tion test of mixture B is not shown due to errors The results of both volumetric compression tests
in the installation of the displacement transducers. are used to develop a selection criteria that uses
It is evident that the mixture C shows much higher two characteristic values:
plastic deformation than the other two mixtures.
• a characteristic Bulk modulus (Kc), determined
The authors suggest a limiting value of 0,5% in the
with Equation 4 and Equation 5 using a volumet-
accumulated plastic strain.
ric stress of 250 kPa and a Poisson’s ratio of 0,4.
• a characteristic volumetric permanent axial
5.2.2 Bulk modulus test
strain, determined with the magnitude of the
After the post-compaction a multistage bulk mod-
accumulated permanent strain at the end of the
ulus test was performed. In this test 100 load cycles
post-compaction test.
at 5 Hz are imposed to the specimen at five increas-
ing vertical stress amplitudes: 65 kPa, 165 kPa, The authors propose a selection methodology in
265 kPa, 365 kPa and 465 kPa. which an UGM can be used in the base course of
The main characteristic of this test is that there thin asphalt pavements provided that it has a char-
is no change in the radial strain direction, therefore, acteristic bulk modulus higher than 150 MPa and
the volumetric stress can be obtained as a function a characteristic volumetric permanent axial strain
of vertical stress amplitude and Poisson’s ratio of less than 0,5%.
(assumed as 0,4 for all materials). Due to the radial The results of the volumetric ranking are gath-
displacement constrain, the volumetric strain cor- ered into a selection chart shown in Figure 14. It is
responds to the measured strain amplitude in the concluded that the plastic deformation of Mixture
vertical direction. Thus, the Bulk modulus can be C is excessive and therefore it cannot be used in the
determined as follows: basecourse layer of thin asphalt pavements.

123
 400
350
T r .··r;J
l1
ACKNOWLEDGMENTS
'ii
~ 300 y• 135.4t..l..JWI
r;:J/ ..t>···· The authors gratefully acknowledge the financial
~ 250
/ . y•125.171t1.H4iJ
support of the Deutsche Forschungsgemeinschaft,
~ 200
-~ 150
. ..: :<~i:-~··· +
DFG (Project FOR 2089).
E ,,,$"!1
~ 100 ~· +
~ li'f_,,/ I REFERENCES
50 r + +

Arnold, G., S. Werkmeister, & D. Alabaster (2008). Per-

0.5 1.5 2.5
formance tests for road aggregates and alternative
Volumetric strain [thousandth]
materials. Technical report, Land Transport New Zea-
t:. Mhcture C, 70% OWC D MiKture 0, 70% OWC
land Research Report 335.
DIN 18127 (1993). Proctorversuch (in German). Deut-
Figure 13. Volumetric strain development during bulk
sches Institut für Normung e.V., Berlin.
modulus test. Mixtures C and D at 70% OWC.
EN 13286-7 (2004). Unbound and hydraulically bound
mixtures—Part 7: cyclic load triaxial test for unbound
mixtures. European committee for standarization,
Characteristic plastic strain[%] Brussels.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.1 Hicks, R. & C. Monismith (1971). Factors influencing
250 , - - - - - - - - - - - - -- .- - - - - - - - - - - - - - - - - , the resilient properties of granular materials. Techni-
'ii
.,_ GOOD POO R
cal Report 335, Hwy. Res. Rec.
::; 70% 80%
-: zoo I:J····:e 80% Kamal, M., A. Dawson, O. Farouki, D. Hughes, &
~
:; 70% 80% D. A. Sha’at (1993). Field and laboratory evaluation of
., o·· the mechanical behavior of unbound granular materi-
~ 150 1----''----------'
-" als in pavements. Transportation research record 1406,
:;
~ 100
Transportation research board.
:; ·"-6.··· Mixture A Lekarp, F., U. Isacsson, & A. Dawson (2000, January-
~ so ·0 ··· Mixture(
February). State of the art I: resilient response of
unbound aggregates. Journal of Transportation
~ ·<>··· Mixture D
.<:
u Engineering.
Numrich, N. (2004). Modellierung des nichtlinear-elas-
tischen Verformungsverhaltens von Tragschichten ohne
Figure 14. Selection chart. Bindemittel (in German). Ph. D. thesis, University of
Technology Dresden.
Rahman, S. & S. Erlingsson (2014). Predicting permanent
deformation behaviour of unbound granular materi-
6 CONCLUSIONS als. International Journal of Pavement Engineering.
Roadex (2000). Road condition management of low traf-
The paper shows the principle of operation of fic volume roads in the northern periphery. Technical
two test methods, RLT test and volumetric com- report, Northern Periphery Roadex.
pression test, used to characterize the elastic and Rondon, H. & F. Reyes (2008). Comportamiento de mate-
plastic response of UGM to harmonic loads. riales granulares en pavimentos flexibles (in spanish).
Different materials commonly used in the base- Universidad Catolica de Colombia, Pontificia Univer-
course of German roads were tested at different sidad Javeriana.
water contents. The results of the tests showed a RSTO (2001). Richtlinie für die Standardisierung des
Oberbaues von Verkehrsflächen (Empirical German
strong dependency of bearing capacity and per-
pavement design guideline– in German).
manent deformation on the amount of water and TL SoB-StB (2004). Technische Lieferbedingungen für
the stress level. Based on the tests results it was Baustoffgemische und Böden zur Herstellung von
possible to develop a classification method for Schichten ohne Bindemittel im Straßenbau (in German).
UGM intended to be used in the basecourse or Uzan, J. (1985). Characterization of granular materials.
thin asphalt pavements. Transportation research record 1022, Transportation
The proposed material selection criteria seems research board.
to be able to rank and differentiate the quality of Wellner, F., S. Werkmeister, & G.C. Falla (2014). Straen-
materials. Based on the findings from laboratory befestigungen aus duennen asphaltschichten auf
qualitativ hochwertigen tragschichten ohne bind-
results, the Granodiorite (Mixtures B and D) can
emittel fuer bundesstrassen. Research project no.
be categorized as good quality material while the 09.0175/2011/erb, German Federal Highway Research
Andesite (Mixture C) is categorized as poor qual- Institute.
ity material. In order to validate the defined qual- Werkmeister, S., B. Steven, & D. Alabaster (2006). A mech-
ity limits, more tests should be carried out with anistic-empirical approach using accelerated pavement
different grading, moisture contents and material test results to determine remaining life of low volume
types. roads. Road and Transport Research Journal 1.

124
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Numerical simulation and laboratory testing of unbound base course

materials considering the effect of temperature

J. Patzak & F. Wellner

Chair of Pavement Engineering, Technische Universität Dresden, Dresden, Germany

ABSTRACT: The climate conditions in the Federal Republic of Germany make it necessary to ensure
the frost-resistance of pavements. The aim is to prevent pavement damage by frost impact in the defined
service life. The main pavement damage potential by frost penetration results from the thawing proc-
ess (loss of load-bearing capacity due to insufficient drainage) as well as the freezing process (volume
expansion of included water). The thickness of the construction is the main factor to guarantee the frost-
resistance of pavements. The Chair of Pavement Engineering at the Technical University of Dresden
performed a numerical simulation of the temperatures in pavements based on a climate database. This
thermal simulation shows the possible depth and frequency of frost penetration. Due to extreme climatic
conditions these freeze and thaw cycles are very common in most of the granular base layers of flex-
ible pavements in central European countries. The deformation behaviour of the Unbound Granular
Materials (UGM) used in structural pavement layers is very complex due to their inherit nonlinear stress
dependency and the effect of bulk material factors, such as water content, density, grain size distribution,
etc. Furthermore, thermal effects play an important role in the mechanical deformation of UGM if the
following two conditions are met: (1) water is present and (2) the temperature fluctuates in the saturation
range (i.e. changes from positive to negative °C, or vice versa). In this context, this paper presents selected
results in addition of an investigation about the effect of the temperature on the mechanical deformation
of UGM under cyclic triaxial conditions using the results of Repeated Load Triaxial (RLT) tests with
Variable Confining Pressure (VCP).

1 INTRODUCTION tion of load capacity is a result of insufficient or

interrupt drainage abilities of soils. Thus melting
The temperatures in pavements result from dif- water can not sweep away because deeper layers
ferent processes of energy transportation (trans- are still frozen. The definition of the frost proof
portation by rays, conduction and convection). pavement thickness is either regulated by guide-
The environmental conditions (global radiation, lines for the standardisation of road construc-
air temperature, humidity, wind speed, rainfall tions RStO [FGSV 2012] or by guidelines for
rate and cloud cover) are changing continuously the asphalt design process RDO Asphalt [FGSV
and cause non steady state temperature condi- 2009]. In addition to the frost proof dimension-
tions. These problems can be solved based on ing of pavements the knowledge of temperature
thermal-physical legalities, especially with the help changes in Unbound Granular Materials (UGM)
of energy and thermal balance equations using is of major importance. It is well known that the
an iterative calculation algorithm (Kayser, S.; deformation behaviour of Unbound Granular
Wellner, F. 2008). Next to the so called environ- Materials under cyclic loading consists of an elas-
mental conditions the thermal-physical properties tic and a plastic part and the magnitude of the
of pavement materials like thermal conductiv- stress/strain response mainly depends on the type
ity, specific heat capacity, water content, density, and magnitude of the applied loading (Lekarp,
reflection properties and heat transfer properties F. et al. (1) and (2) 2000). Furthermore, there are
influence temperature conditions considerably. several material related factors that influence the
Frost in pavement sub layers and the associated mechanical response of UGM noticeable result-
damage process are well known. In addition to ing in a high complexity. Some of these factors are
lifting effects (formation of ice lenses) as a result the material density, moisture content, grain size
of the increase in volume of water included dur- distribution and particle shape (Barksdale, R. D./
ing the freezing process, the temporary reduc- Itani, S. Y. 1989; Van Niekerk et al. 1998).

125
 It is important that the thermal effects play a where
significant role in the mechanical deformation of qU Area-related energy level [W/m2]
UGM if the following two conditions are met: (1.) qG global radiation [W/m2]
water is present and (2.) the temperature fluctuates qAG atmospheric radiation [W/m2]
in the saturation range (i.e. changes from posi- qFO radiation of the pavement surface [W/m2]
tive to negative °C, or vice versa). Due to extreme qk, ref, l, ref short-wave and long-wave reflectivity [W/m2]
climatic conditions these freeze and thaw cycles qV latent heat flow [W/m2]
are very common in most of the granular base qL sensible heat flow [W/m2]
layers of flexible pavements in central European qB ground heat flow [W/m2]
countries.
The properties of frozen and unfrozen soils In order to calculate the individual terms of
under different loading conditions differ signifi- the thermal equilibrium extensive calculated or
cantly. The driving factor for this difference is measured climatic data is required. Measured by
the presence of both, ice and liquid water. Freez- Germany’s National Meteorological Service (DWD),
ing generally results in an increase in strength and global radiation, air humidity, air temperature, wind
stiffness of the soil (Mancuso, C. et al. 2012; Hass, speed, degree of cloudiness and rainfall for example
H. 2012). Simonsen reported a large temperature can be taken into account in the calculation proce-
dependency in all soils investigated but different dure. The detailed explanation and calculation of the
stress dependency at the same time (Simonsen, E. individual terms is omitted within this paper. Details
et al. 2002). can be found in the literature [FGSV 2008].
To the knowledge of the authors there are no
previous investigations about the mechanical
2.2 Detection of condition-based material
deformation behaviour of natural unbound granu-
parameters
lar materials considering the effect of temperature
(frost-thaw cycles) under small strains. UGM basically are 3-phases mixtures including
Therefore, this paper presents both, numerical grains, air and water. The very complex mecha-
simulations of temperatures in pavements, espe- nisms of their interference make the detection
cially the simulation of frost penetration as well as and characterisation very difficult. Thus, there
results of an experimental investigation to deter- are predefined states for modeling, e.g. density,
mine the effect of temperature on the mechani- particle size distribution, saturation, phase ratio
cal deformation of UGM under cyclic triaxial or temperature. It is well known that the thermo-
conditions. physical importance of the single phases differs
considerably. Pore volume air has very poor heat
conductivity. It is nearly four times smaller than
2 BASICS OF NUMERICAL SIMULATION the heat conductivity of water and grains. Prin-
OF TEMPERATURES cipally the heat conductivity of soil or unbound
granular materials increases with growing water
The temperatures in pavements result from differ- content. This is caused by the heat conductivity
ent processes of energy transportation (transpor- of water which is 20 times greater compared to
tation by rays, conduction and convection). The air. The behaviour of water content and specific
environmental condition (global radiation, air heat capacity are proportional as well. Increasing
temperature, humidity, wind speed, rainfall rate water content results in a higher ability to save heat
and cloud cover) are changing continuously and energy. There is a distinction between the complete
causes non steady state temperature conditions. unfrozen and complete frozen state of unbound
These problems can be solved based on thermal- granular materials. The thermo-physical param-
physical legalities, especially by the help of energy eters differ considerably. In completely solid state
and thermal balance equations using an itera- of water the heat conductivity is approximately
tive calculation algorithm [Kayser, S.; Wellner, F. 2.20 W/(m⋅K) and therefore nearly four times
(2008)]. higher than in fluid state. In contrast to the heat
conductivity the specific heat capacity of ice is
approximately 2.12 kJ/(kg⋅K) and therefore only
2.1 Calculation methods-Basics about 50% of the specific heat capacity of water
(fluid state). Thus, the ability to transport heat
The mathematical formulation of the energetic energy increases with reduced heat storage and
problem (the heat balance equation) can be repre- the frost penetration will be encouraged. Water
sented as follows. naturally exists in three aggregates (solid, fluid
and gaseous). Due to heat extraction or heat input
qU qG − qk ,reff + qAG − ql ,ref − qFO + qV + qL + qB (1) an aggregate change takes place at constant tem-

126
perature. The energy level increases from the solid 30 .-------------------------------------~

over the fluid to the gaseous condition. The dif- 25

ferent kinds of bindings in grains are the cause
20
of a variable melting point of water, which means
the change of the aggregate phase takes place at ~ 15 '\
different temperatures. The freezing temperature ~ '\
of water depends on the free enthalpy, which is .a 10
~ - - measured temperature .,

l v-~-"/;"f
reduced by decreasing binding and ion concentra-
----- calculated temperature (melting
tion (Unold, F. 2006). Water is freezing first where energy consitered)
the influence of binding forces is low (high pore ---------calculated temperature (melting
energy excluded)
volume). Just with further falling temperatures, -5
also the bound water changes its aggregate phase c: ..ci ·ro ci. > N
~
Q)
lL
1::!
'ro ~
<( :::2:
Q)
en
0
z
Q)
0
:::2:
and becomes solid. The detailed material specific
functional description can be realized only within
Figure 2. Exemplary comparison of measured and
by a massive experimental effort. The description
numerical simulated temperatures in soils in depth of
of the correlation between temperature and the 100 cm; station Dresden, year 2003.
unfrozen water part can be realized to approxima-
tion. The used functional classification is based
on references (Smoltczyk, U. 2001; Farouki, O. T. drainage ability of soils. Thus, melting water can
1986). Knowing that the temperature depending not sweep away because deeper layers are still fro-
water part depends on temperature it is possible zen. On the one hand Figure 3 shows exemplary the
to convert the heat energy for calculation of heat maximum frost penetration (zF max) exemplary. On
capacity (Brehm, D.R. (1989). In the result of this the other hand, the thickness of subgrade still frozen
process the heat capacity can be calculated depend- (at the time of subgrade thawing) is shown (dkrit). In
ing on temperature and soil type. The results of order to avoid damages as a result of frost penetra-
the coupling effect are shown in Figure 1 und 2 tion in pavement structures a thickness of 60% of
exemplary. It appears that under consideration of the maximum frost penetration (zF max) is enough
freezing and thawing process the calculated tem- according to the existing experience [FGSV 1994].
peratures represent the real temperatures consider- Based on the weather data recording of the
ably better. Therefore, the temperature conditions Germany’s National Meteorological Service (DWD)
in pavements in general and the frost penetration in 10 weather stations over the years 1991 to 2004, the
particular can be calculated as accurate as possible. temperatures in pavements are simulated continu-
ously for different German regions and one fixed
structure. The following conventions of temperature
2.3 Frost in pavement structures—damages
simulation and parameter definitions are used:
Frost in pavement sub layers and the associated Total thickness of the numerical model = 250 cm
damage process are well known. In addition to lift-
ing effects (formation of ice lenses) as a result of Simulated structure
the increase in volume of water included during
Total thickness asphalt layer = 30 cm
the freezing process, the temporary reduction of
Thickness subbase = 20
load capacity is a result of insufficient or interrupt
Thickness subbase = 200
30 .-------------------------------------~ Reflective properties of asphalt:
25 − Reflectivity for short wavelength radiation
αk = 0.15
− Reflectivity for long wavelength radiation
αl = 0.05
− Thermal conductivity λ = 1.05 W/m2
− Specific heat capacity cp = 878 W⋅s/(kg⋅K)
− Density ρ = 2240 kg/m3
---------calculated temperature (melting Properties of sub base (UGM):
energy excluded)
-5 − Thermal conductivity (for T > 0°C) λ = 2.14 W/m2
c: cil ci. > N
~
..ci
− Thermal conductivity (for T ≤ 0°C) λ = 1.70 W/m2
"ffi
~ ~
1::!
~
Q)
lL 'ro <( :::2: ~ ~
Q)
en z
0 Q)
0
:::2:
− Specific heat capacity (for cp > 0°C) cp = 837 W⋅s/
Figure 1. Exemplary comparison of measured and (kg⋅K)
numerical simulated temperatures in soils in depth of − Specific heat capacity (for cp ≤ 0°C) cp = 767 W⋅s/
50 cm; station Dresden, year 2003. (kg⋅K)

127
 (!) (!) (!) (!) (!) (!) 1'- 1'- 1'- 1'- 1'- 1'- 1'- 1'- 1'- 1'- 1'- 60%
en en en en en en en en en en en en en en en en en 'if<
~ ~ ~ ~ ~ en en en en en en en en en en en en <0
ro- • frost in pavement
N N N N N N 50%
0 0 0 0 0 0 0 0 0 0 0 '<t"
(") ui ,....:
N N N N
oi
N (;; N
0 ~ 0<Ci 0<ri c:i N ~ <Ci <ri Nc:i N
N
0 40%
~
T> 0 C e.:.
'E 20 i:j'3o%
cCl)
~
:I
LL
N
40 ~20% 'if<
<0
r:: II..
0 ~
~ 60 10%
Qj
r::
Q)
c. 80
0-50 50 55 60 65 70 75 80 85 90 95 100
Reference depth [em]

120 ~------------------------------~----~ Figure 4. Frequency of exceedances of frost in pave-

ment structures.
Figure 3. Qualitative freezing-thawing process pres-
entation of the maximum frost depth (zF max) and the
critical layer thickness (dkrit) according the reference
depth. 25%
'l()
if< • frost in pavement > 50 em
ci
N
'if< • eritieallayer
20% N
cD 'if<
<0

− Density ρ = 2050 kg/m3 ~15%

;:!:-

− Gravimetric water contend wg = 3.5 M.-% ~

'if<
'<t"
~ ~
− Approximation coefficient (temperature depend- c
~10%
'if<
<0
a:)
'if<
<0
ro-
0
a)

ent unfrozen water part) Z = 5 (Brehm, D.R. ~

II..
1989) 5%

Properties of sub ground:

0%
− Thermal conductivity (for T > 0) λ = 1.60 W/m2 50 55 60 65 70 75 80 85 90 95 100
− Thermal conductivity (for T ≤ 0) λ = 1.82 W/m2 Reference depth [em]

− Specific heat capacity (for cp > 0°C)

Figure 5. Frequency of exceedances of frost deeper
cp = 1173 W⋅s/(kg⋅K)
than > 50 cm (normalized to 100%) and the part of
− Specific heat capacity (for cp ≤ 0°C) cp = 900 exceedances of critical layers.
W⋅s/(kg⋅K)
− Density ρ = 1780 kg/m3
− Gravimetric water contend wg = 15 M.-%
− Approximation coefficient (temperature depend- 120 120

ent unfrozen water part) Z = 1.5 (Brehm, D.R. 100 100

1989)
80 80
Figure 4 shows the calculated frequency of
'E
exceedance for a pavement system according to a 'E 60 60~
~ !C. middle critical layer thickness d"" ·E
fixed structure and the defined material specific ..r-
"0~

parameters. 40 -- L L - -- Trend d~, 40

About half (48.6%) of all frost penetrations

do not reach depths over 50 cm and therefore
20 1 l 1 Lfll
1 1 1- 1
1
20

no sub ground. Figure 5 shows the distribution 0 0

of frost penetrations with depths greater than 40 50 60 70 80 90 100 110
Reference depth [em]
50 cm (normalized to 100%) and the part of
exceedances of critical layers and thus, reduced
Figure 6. Average frost penetration and average of crit-
bearing capacity. As excpected, the exceedances ical layer thickness according to the reference depth.
of frost are reduced with increasing reference
depth. Depths greater than 50 cm (important for
the dimensioning process) are reached by 50% of
all frost events. Critical layers appear compara- However, the influence of such temperature con-
tively rare but with thicknesses of up to 30 cm ditions on the mechanical material behaviour of
(Figure 6). UGM is of great importance. This very complex
Due to the frost-proof dimensioning of pave- material behaviour is the subject of the used labo-
ments the damages described should be excluded. ratory tests and part of the studies.

128
3 LABORATORY TESTS order to avoid these problems the particle size distri-
bution of the initial material has been modified and
The RLT tests were performed with the RLT test screened (max. grain size was limited to 22.4 mm).
apparatus of the Institute of Urban and Pavement The material was placed in layers in a cylindrical
Engineering at TU Dresden (Figure 7). The RLT steel mold and compacted to the target density. The
apparatus consists of a test frame, a hydraulic unit, height of the rotationally symmetrical test speci-
a test cylinder and a climatic chamber. In the axial mens was 300 mm and the diameter was 150 mm.
direction, tension and compressive dynamic loads
up to 5 MPa can be applied. The cell pressure can
3.2 Determining the time for the tests
be both static and dynamic and it ranges between
0.05 MPa and 1.5 MPa. Dynamic loads can be up To ensure accurate testing it was necessary to deter-
to 10 Hz. The climatic chamber supports tempera- mine the duration of the freezing and thawing
tures between −20°C and +70°C. For measuring process within the test specimen. Hence, pre-tem-
axial deformations two inductive displacement perature tests were conducted on unloaded speci-
measuring systems (external) and a magnetic mens (temperature measurements in the center of
measuring system (internal) are available. In the the specimen—specimen core) in order to develop
radial direction the measuring system consists of a reference temperature profile. The specimens used
nine inductive displacement transducers. for the pre-temperature tests had the same proper-
ties (DoC and MC) as the specimens used for the
tests under repeated loading conditions. A separate
3.1 Material tested
temperature sensor was placed in the middle of
Within this research a Sandy Gravel 0/45 (KTS these specimens (Dummies) to control the speci-
0/45) typically used in frost protection layers in Ger- men core temperature. Figure 9 presents the results
many was tested. Figure 8 presents the grain size
distribution of the Sandy Gravel and the grading 100
~ I //
limits according to the German guideline TL SoB-
80
p Pr =
w opt=
2.215 g/cm
6.1 M.-%
3

I ~----- ----" ~/ ro 20
StB. The Sandy Gravel was taken from the Zwickau I ~
/ I
quarry near Dresden. This roundly grained material ~ / / ~
was chosen because of the sensitive test and meas- ~
60
/ v ,; 40
~
bO / 1/ / ~7 "d
v
urement equipment of the RLT apparatus. ·~
40
/ / v
60
·§
One problem in the preparation of the test ro
p... 20 / __..-V /
____ v 80
v
~
specimens experienced was that the large grains / l---::::
~
often were very close to the axial measuring system 0.5
100

(magnets integrated during specimen production)

of the RLT apparatus. The larger grains may lead
to changes in position of the magnets during test-
ing and thus, to inaccurate test results. In addition, Figure 8. Grain size distribution of the Sandy Gravel
damage of the measurement system may occur. In 0/45.

30
l - T-Chamber -T-Cell - T-Core Specimen
r
~
20
- I

v---~
1\/ phase transition
~ I
I
I

~r
I

\ A pre-tempering
B cooling I
I

~~ ~ I
C warming
I I

-20
I I

A I B I c I
¢::::::::::>I I I
-30
0 10 15 20 25 30 35
Time [hl

Figure 9. Time dependent temperature in chamber, cell

and specimen core; right: measured (grey points) and
approximated values (red line). As an example of a speci-
Figure 7. RLT test apparatus of the Institute of Urban men with MC = 4 M.-%. DoC = 97% and a temperature
and Pavement Engineering at TU Dresden. change of 40 K (Patzak 2015).

129
of a typical temperature run in terms of measured be described in dependence of the average stress p
temperature in the chamber, cell, and specimen and the deviatoric stress q.
over a period of about 29 h for a complete freezing
and thawing cycle. The period A covers the speci- ε11_ el = f (q; p ). (2)
men pre-tempering, period B the cooling or freez-
ing cycle and period C the warming or thawing. The deviatoric stress q is calculated using the
The duration of the freezing and thawing period second invariant of the deviatoric stress tensor:
in the triaxial cell and in the specimen are shown.
The discontinuous shape of the temperature curve q 3 J2 =
at nearly −1°C can be explained by the phase transi- (3)
tion period. During this phase transition period the [ ⋅ − + − + − ].
aggregate state of the existing water changes from
liquid to solid (frozen) and back to liquid; the water and the following equation:
as a liquid and the ice crystals coexist.
The temperature of the whole system remains q = σ1 − σ 3 (4)
nearly equal to the freezing/melting point due to
slow removal of heat when being in contact with The average stress p can be calculated using the
air. It is observable that the freezing process is first invariant of the stress tensor:
greatly slowed and the temperature will not drop
for about 30 minutes once the freezing starts but 1 1
continues dropping once the freezing process has σm I1 (σ 1 + σ 2 + σ 3 ). (5)
been completed. 3 3

Under the assumption of isotropy (independ-

3.3 Test procedure ency of direction) the material behaviour for
In order to determine the deformation behaviour any stress state can be expressed by two relevant
of UGM under conditions that simulate the physi- material parameters (Young’s modulus and Pois-
cal conditions and stress states of these materials son’s ratio) using Hook’s law. However, in case of
in the layers beneath the surface layer of a pave- non-linear material behaviour, as it is known for
ment which is subjected to moving loads, Variable unbound granular materials, the Hook’s law can
Confining Pressure (VCP) tests were conducted on only be applied for infinitesimal stress and strain
specimens with a size of 150 × 300 mm. To deci- changes. Following, the required calculations for
sively test the elastic material response it was neces- analysing the laboratory test results must be car-
sary to define stress limits for every stress path. The ried out incrementally. For rotational symmetry and
following procedure for stress-path limitation has isotropy follows:
been chosen. The turning point of the stress-strain-
relationship at different radial stress amplitudes σ xx σ yy = σ xx / yy / σ = σ σ / 33 . (6)
was calculated for each studied material condition ε xx ε yy = ε xx / yy / ε = ε ε . (7)
/ 33
(DoC, MC) (Patzak 2015). This point defines the
transition from the so called material hardening 1
to the material softening. Based on this tempera- ε11
E
(σ 11 − 2 μ σ 22 33 ). (8)
ture distribution within the specimen depending
on time the measuring points for the tests under
repeated loading were defined. In total, 10 meas-
ε 22 /33
1
E
(
σ 22 /33 − μ (σ 22 /33 + σ 11 ) . ) (9)
uring points were chosen during the freezing and
thawing period to measure the change of stiffness The Young’s modulus can now be calculated
during this period. The measurement was stopped by:
after the freezing process has been completed at a
specimen core temperature of −10°C. In the posi- E=
tive temperature range the measurements were (Δ − Δ ) (Δ + ⋅ Δ ) . (10)
Δε11 ⋅ ( Δ + Δ )
completed after the thawing process has been fin-
− 2 ⋅ Δε 22 / 33 ⋅ Δσ 222 / 33
ished at a temperature in the center of the specimen
of + 3°C.
and the Poisson’s ratio by:
3.4 Test analysis μ=
Δε 22 / 33 ⋅ Δσ 11 − Δε11 ⋅ Δσ 22 / 33
For the characterisation of the mechanical defor- . (11)
mation behaviour the elastic axial strain ε11_el can 2 ⋅ Δε 22 / 33 Δσ 22 / 33 − Δε11 ( Δσ 11 + Δσ 111 + Δ 22 33 )

130
 Only the vertical deformation of the specimens significantly reduced due to the greater resistance
during loading and discharge are accurately meas- to deformation. Depending on the stress level the
ured and evaluated. Consequently, the stress point strains are reduced by a factor of 30. The sample
only is in the focus of consideration. From with 97% degree of compaction and 2% water con-
tent represents an exception in this case. The axial
ε11_ f (σ 11;σ 22 /33 ) (12) elastic strains achieve significantly larger values
(about 5-fold) than the ones at comparable condi-
and by differentiation with respect to the desired tions. The reason for this can be found in the rel-
independent variable (with the same direction atively low water content of the specimen. In the
as the dependent variable) the calculation of the quasi-completely frozen state the Young’s modulus
Young’s modulus as a tangent modulus follows: increases with an increasing water content of the
material. The UGM tested in the positive tempera-
d f (σ 11;σ 22 /33 ) 1 ture range containing higher amounts of water
11
= (13) in the samples is known to lead to larger elastic
dσ 11 E strains (under drained conditions); this relationship
reverses in the quasi-completely frozen state.
It is important that the calculated Young’s Depending on the material condition the
modulus, considered as differential quotient, is Young’s modulus reaches, calculated values
exclusively valid for the stress point. Based on the between about 7.500 MPa and 35.500 MPa
Young’s modulus calculated at the stress point, the (Figure 10 and 11). In this case the values are com-
strain change in axial and radial direction and fol- parable to the stiffness of asphalt mixtures in the
lowing also the Poisson’s ratio can be calculated for same temperature range.
an incremental infinitesimally small change of the
principal stresses. 40.000
no significant difference

3.5 Results
35.000
-8-- ---- -cp----- - -i=j-
30.000 I-33.9D.I': 3-3; P5u 33.925-
The following chapter contains selected results ~ 25.000
focusing on the calculated Young’s modulus. ~
-20.000
~ L> KTS-97-04
parameter overview (main factors)
15.000
OKTS-100-04
test procedure: 10.000
VCP (multi-stage): DKTS-103-04
applied stress paths: (total range) 5.000
96 98 102 104
σ22/33 [N/mm2] 0.075 to 0.45 (depending on
Temperature)
σ11 [N/mm2] 0.075 to 0.75 (depending on
Temperature)
degrees of compaction DoC (DPr): Figure 10. Young’s modulus at different degrees of
97% / 100% / 103% compaction (DPr = 97% to 103%) and constant water
moisture contents MC (w): content (w = 4%); quasi-completely frozen (Patzak 2015).
2.0 M.-% / 4.0 M.-% / 6.0 M.-%
temperature range cooling/heating phase: 40.000
ΔT from −10.0°C to 10.0°C 35.000
}4 ll
The specimen identification is presented as 33.906 } ~ ;' 35. 96
follow: 30.000 ;'
;'
Material-DoC-MC; for example KTS-100-04 ~ 25.000 /
/.

~20.000
/

/
~ O KTS-97-02
3.6 Tests with thermal influences (quasi- 15.000
_, / trend _
/
completely frozen) /
I
L> KTS-97-04
10.000
Based on the test results of the examined stress
5.000
A D7.606 D KTS-97-05

levels for the quasi-completely frozen state of the 3 4

material (sample core temperature = cell tempera- w[%]
ture) linear elastic material behaviour was assumed.
Nonlinearities could not be determined explicitly. Figure 11. Young’s modulus at different water content
In comparison to the laboratory tests with- (w = 2% to 5%) and constant degrees of compaction (DPr
out thermal stresses the axial elastic strains are = 97%); quasi-completely frozen (Patzak 2015).

131
3.7 Tests with thermal influences (phase 40.000 6 KTS 97-04 o KTS 97-02 D KTS 97-05

transition) D L,
I
30.000
Apart from the basic reduction of elastic strains ~
~
I
during the quasi-completely frozen state of the I
I

water contents as part of UGM, analyses have been ~20 . 000 I

run on how the nonlinearity of the stress-strain w I

I

relationship is influenced while the temperatures trend I

10.000 ,
change from positive to negative and consequently ,'ill 0 trend
- --E)
the aggregate state of the water changes; regardless rtr/; 6-G- -- -- ----
0
of the material condition. After evaluating all 10 0 -1 -3
test points of the freezing phase (during the phase-
transfer from fluid to ice—shown in Figure 9) it
could be found that the decreasing amount of fluid Figure 13. Young’s modulus as a function of sample
water does not only lead to a reduction of the elastic core temperature and different water content (w = 2%
axial strains, but also to an elastic material response to 5%) and constant degrees of compaction (DPr = 97%),
which can be described as “solid-state-like” whereby (Patzak 2015).
the latter results in an increasing linear elastic
behaviour. Opposite behaviour can be found during
where the main part of the phase-transfer takes
the thawing process. When the temperature rises the
place the water content affects the value of the
freezing line retreats farther and farther to the inside
Young’s modulus considerably. A difference in
of the specimen and the elastic strains increase.
the water content in the mixture of 2 to 3 M.-%
Although the strains already increase, linear elas-
causes a difference in the specific example of the
tic material behaviour can still be assumed during
calculated Young’s modulus between factor 3 and
this process. Firstly, with the start of measuring
4. Regarding samples with 4 M.-% and 5 M.-%
core temperatures < 0°C, the expected nonlineari-
water content as well as a core temperature allows
ties (as in the unfrozen state) could be found again.
to assume linear elasticity.
Figure 12 and 13 show the temperature-depending
increase of the Young’s modulus with respect to the
different degrees of compaction and water content.
4 CONCLUSIONS
Due to a very rapid decrease of the elastic strains as
a result of the freezing progresses the Young’s mod-
The results of the numerical simulations show that
ulus also rises relatively quickly and significantly.
even for highly stressed pavements with a compar-
At the end of the transition phase of the water the
atively great thickness of bound layers sufficient
Young’s modulus reaches nearly identical values as
frost safety is available. A reduction of the thick-
the quasi-complete freezing.
ness of UGM is possible without a significantly
Significant differences regarding the influence
larger damage risk of the unbound layer, especially
of the degree of compaction can not be deter-
with regard to postulated climate change.
mined within the framework of the conducted
In addition to the well known temperature
studies. It is very clear that even within the period
independence within a positive temperature range
unbound granular materials shows a tempera-
40 .000
ture independence when the water existing in the
6 KTS97-04 o KTS 100-04 D KTS 103-04
unbound mixture is quasi-completely crystallized
30 .000 ~R and this state remains unchanged. Consequently,
the mass fractions of water in fluid and crystalline
~
!--'

~ I
I form can be considered constant. Temperatures
~0 .000 I
between the two aforementioned situations (inde-
w trend ,1
pendent of whether a cooling or heating process
,,
10.000
,. ,. exists) influence the deformation behaviour of the
unbound granular material. The period of phase
__ ,-/ ~ transition of water is of particular importance due
0
11--- to the fact that the ongoing change of the state
0 -1,5 -2 of aggregation (fluid-crystalline) has a significant
influence on the material structure and perform-
Figure 12. Young’s modulus as a function of sam- ance. The degree of compression is of great impor-
ple core temperature and different degrees of compac- tance for the deformation behaviour of UGM
tion (DPr = 97% to 103%) and constant water content given that no thermal influence is present. How-
(w = 4%), (Patzak 2015). ever, within the research presented here the degree

132
of compression could not be defined as a relevant Lekarp, F. et al. (2) State of the Art. II: Permanent Strain
influencing factor on the material behaviour; both, Response of Unbound Aggregates. Journal of trans-
during the phase of transition as well as the frost portation engineering, 2000.
penetration. Variations in the material properties Mancuso, C. et al. Unsaturated Soils: Research and
Applications, 2012, 2. Aufl. Springer Verlag.
and the small specimen deformations due to ther- Patzak, J. Mechanical deformation behaviour of a base
mal influence are possible reasons for this. It could course material with particular effect of tempera-
be deduced that the water content seems to be the ture—Experimental analysis using the example of a
most influencing factor not only in case of no ther- gravel base course. Dissertation, TU Dresden, 2015.
mal influence, but also under thermal influence. Smoltczyk, U. (2001). Grundbau-Taschenbuch 2.
The stress dependence of rock grain mixture Geotechnische Verfahren, 6. Auflage, Ernst W. + Sohn
which is known for a thermally uninfluenced state Verlag, Berlin.
in this case could be determined for the thermally Unold, F. (2006). Der Gefriersog bei der Bodenfros-
influenced state only at the beginning of the tung und das Kompressionsverhalten des wieder
aufgetauten Bodens. Dissertation, Universität der
cooling phase or at the end of the heating phase. Bundeswehr München, Fakultät für Bauingenieur-
Even at a relatively low cooling time the unbound und Vermessungswesen.
granular materials showed a linear elastic material Van Niekerk et al. Estimation of mechanical behaviour
behaviour under the defined test procedure and of unbound road building materials from physical
test conditions including the defined stress levels. material properties. Proceedings, 5th International
Conference on the bearing Capacity of roads and air-
fields, 1998, R. S. NORDAL and G. RAFSDAL, eds.,
REFERENCES VOL. 3, pp. 1221–1233.
FGSV 1994. Entstehung und Verhütung von
Barksdale, R. D./Itani, S. Y. Influence of aggregate shape Frostschäden an Straßen. Arbeitsausschuss Frost.
on base behaviour. Transportation Research Record, Schriftenreihe der Forschungsgesellschaft für
1989, No. 1227, pp. 173–182. Straßen- und Verkehrswesen 105. Bundesministe-
Brehm, D.R. (1989). Entwicklung, Validierung und rium für Verkehr, Bau und Stadtentwicklung, Kirch-
Anwendung eines dreidimensionalen, strömungsge- baum Verlag, Bonn.
koppelten finite Differenzen Wärmetransportmod- FGSV 2008. Grundlagen zur Erfassung der Temper-
ells. Justus-Liebig-Universität Giessen, Fachbereich aturbedingungen für eine analytische Bemessung von
Geowissenschaften und Geographie. Asphaltbefestigungen. Forschung Straßenbau und
Farouki, O. T. (1986). Thermal Properties of Soils. Series Straßenverkehrstechnik, Heft 996, 2008.
of Soils. Series of Soil Mechanics, Vol. 11. FGSV 2009. Richtlinien für die rechnerische Dimen-
Hass, H. North American Tunneling. Proceeding. Soci- sionierung des Oberbaus von Verkehrsflächen mit
ety for Mining, Metallurgy, and Exploration. Inc, Asphaltdecke RDO Asphalt). Forschungsgesells-
2012. chaft für Straßen- und Verkehrswesen, FGSV Ver-
Kayser, S.; Wellner, F. (2008). Grundlagen zur Erfas- lag, Köln.
sung der Temperaturbedingungen füreine analytische FGSV (2012). Richtlinien für die Standardisierung des
Bemessung von Asphaltbefestigungen. Forschung Oberbaues von Verkehrsflächen (RSTO). Forschungs-
Straßenbau und Straßenverkehrstechnik, Heft 996, gesellschaft für Straßen- und Verkehrswesen, FGSV
Bundesministerium für Verkehr, Bau und Stadtent- Verlag, Köln.
wicklung, Bonn. TL SoB-StB Technische Lieferbedingungen für Baust-
Lekarp, F. et al. (1) State of the Art. I: Resilient Response offgemische und Böden zur Herstellung von Schich-
Of Unbound Aggregates. Journal of Transportation ten ohne Bindemittel im Straßenbau. Ausgabe 2004/
Engineering, 2000. Fassung 2007.

133
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Some recent research on the hydraulic conductivity of road materials

P.J. Vardanega, S. Feng & C.J. Shepheard

University of Bristol, Bristol, UK

ABSTRACT: Hydraulic conductivity is an important material parameter for pavement engineers. The
ability to make a priori estimates of the coefficient of permeability of road pavement materials is very
useful when studying issues of pavement durability and predicting pavement performance. This paper
presents two laboratory databases assembled from data sourced from the literature. Using a database of
permeability tests on fine-grained soils, the ability of the Kozeny-Carman void ratio function to interpret
the data is examined alongside a simple power-law relationship. A second database of permeability meas-
urements on asphalt concrete specimens is used to re-examine the ‘representative pore size’ concept as a
method of predicting the coefficient of permeability. The influence of Nominal Maximum Aggregate Size
(NMAS) on the coefficient of permeability is also studied.

1 INTRODUCTION 2014). Strictly speaking, the percentage of con-

nected (accessible) voids should be evaluated dur-
1.1 Literature review ing testing (cf. Bear 1972). This is not always done
in practice but can be done using the ‘hand-pump-
The hydraulic conductivity of road building materi-
ing’ method (see Smith & Gotolski 1969, p.24 - the
als is a useful engineering parameter and should be
method was also used by Kumar & Goetz 1977).
considered when Asphalt Concrete (AC) mixtures
are designed and specified (e.g. McLaughlin and
Goetz 1955, Liebenberg et al. 2004) to avoid mois- 1.3 Study aims
ture damage (e.g. Chen et al. 2004) which is detri-
This paper presents the analysis of two databases.
mental to pavement performance (e.g. Abdullah
Database I (soil) contains over 100 measurements
et al. 1998, Mohammed et al. 2003). A recent review
of the hydraulic conductivity of fine-grained soils.
on the permeability of asphalt concrete (focussing
Database II (AC) contains over 1300 measure-
on empirical prediction models) has been published
ments of the hydraulic conductivity of asphalt
(Vardanega 2014). The hydraulic conductivity of
concrete. Statistical analysis is performed with both
fine-grained materials has been reviewed in many
databases using power-law functions. The results
publications (e.g. Olsen 1960, Olson & Daniel
are compared to a previously published empirical
1981). Fine-grained materials may constitute the
model for the prediction of asphalt concrete per-
natural materials that are used as road subgrades.
meability (Vardanega & Waters 2011, 2015).
Comparisons of laboratory and field perme-
The Nominal Maximum Aggregate Size (NMAS)
ability measurements on AC have been made (e.g.
has been suggested as a key influence on the hydrau-
Gogula et al. 2003). Methods for measuring perme-
lic conductivity of asphalt concrete (e.g. Cooley
ability in field conditions have been developed by
et al. 2002, Yan et al. 2016). Therefore, the effect of
various researchers using water as a permeant (e.g.
NMAS will also be studied using Database II (AC).
Zube 1962, Gerke 1982, Cooley 1999, Fwa et al.
2001) and air as a permeant (e.g. Kari & Santucci
1963). Construction issues such as lift thickness and
2 DATABASE I (SOIL)
time for compaction have also been investigated
(e.g. Hainin et al. 2013). Neural network approaches
2.1 Permeability models
for interpreting databases of AC permeability have
been published (Tarefder et al. 2005). Permeability testing usually employs falling head or
constant head approaches (e.g. Bear 1972). Dołźyk
and Chmielewska (2014) recently reviewed various
1.2 Influence of porosity
empirical and semi-empirical approaches to describe
A key predictor of asphalt concrete permeability is permeability data, emphasizing amongst other fac-
percentage air voids in the mixture (e.g. Zube 1962, tors the importance of the temperature of the per-
Mullen 1967, Vivar & Haddock 2007, Vardanega meant—which is not always reported in the literature.

135
 Carrier (2003) recommended that formulations Table 1. Summary of Database I (soil),.
based on the work of Hazen (Hazen 1892, 1911)
for modelling permeability should not be used in No. Source Symbol Soil type
geotechnical practice (the formulation is shown as
Al-Tabbaa & Wood (1987) Kaolin
equation 1 using the notation of Carrier 2003): Chung et al. (2002) ◊ Pusan Clay
Lehka et al. (2003) IIT Clay
k = (CH)(D10)2 (1) Lehka et al. (2003) Δ Calcium
Bentonite
where k is the coefficient of permeability (expressed Leroueil et al. (1990)* × Backebol Clay
in units of length per unit time); CH is an empiri- Leroueil et al. (1990)* Matagami Clay
cal coefficient and D10 is the sieve aperture through Leroueil et al. (1990)* ♦ St. Esprit Clay
which exactly ten percent of the material would Leroueil et al. (1990)* Louisville Clay
pass. Carrier (2003) recommended that the Pane et al. (1983) • Kaolin
Kozeny-Carman formulation (Kozeny 1927, Car- Walker & Raymond (1968) ×| Leda Clay
man 1938, 1956) should be preferred for use in
practice (the formulation is shown as equation 2 * Vertical permeability data included in the database.
using the notation of Carrier 2003):

k = (γ / μ)(1 / CK-C)(1 / S02)[e3 / (1+e)] (2) -6.0

ln(k) = 1.3441n([esoi/J3/(1+[esoi/lll -13.889
R'=0.564
~ -8.0 RD= 66.0%

where γ is the unit weight of the permeant (usually E SE= 0.940

n= 119

water), μ is the dynamic viscosity of the permeating .~ -10.0 p < 0.0001

'0
fluid, CK-C is an empirical coefficient, S0 the specific ~ -12.0
VI
surface area of the soil per unit volume of particles, ~
Q.-14.0
and e is the void ratio, hereafter denoted as (esoil). ><
Q)

Equation 2 relies on the principles of viscous ~-16.0

~
fluid flow from the Navier-Stokes equation to treat ::E-18.0
the assembly of soil particles as creating a series
of capillary tubes through which a fluid travels -20.0 +-----r--..------r--..------r--.....-----.
-3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0
(Chapuis & Aubertin 2003). Through this method,
it can be shown that the coefficient of permeability
(k) strongly varies with the void ratio (esoil). Chapuis Figure 1. The relationship between k and the Kozeny-
& Aubertin (2003) showed that for plastic soils Carman void ratio function.
that are “hydrated, saturated, and consolidated
before permeability testing”, the Kozeny-Carman is 66.0%. The Relative Deviation (RD) is defined in
formula may be used. Masad et al. (2004, 2006) Waters & Vardanega (2009) as:
presented semi-empirical approaches based in part
on the Kozeny-Carman formulation to predict the RD = 100 (1 – R2)0.5 (3).
permeability of asphalt concrete using databases. Figure 2 shows the data re-plotted with ln(k)
predicted using ln(esoil). The resulting trend-lines
2.2 Statistical analysis (re-arranged into power functions) are shown in
the figure. The exponent on esoil varies from 1.53 to
Table 1 summarises the sources of the data in Data- 5.32 for each of the ten test series analysed. The RD
base I (soil). Permeability test data from six pub- values for the analysis of the individual test series
lications (ten soils) was collated and statistically varies from 4.5% (R2 = 0.998) to 55.5% (R2 = 0.692).
analysed. The Matagami Clay and the Louisville The exponent on esoil for the whole dataset is 3.29.
Clay were reported as being tested using the con- For the regression of the entire dataset, R2 = 0.573,
stant head approach whereas for the other clays RD = 65.4% and n = 119. If the possible outlier test
falling head testing was reported. The relationship series (No. 3) is excluded from the main regression
between the coefficient of permeability (k) and the the R2 increases to 0.624 (RD = 61.3%, n = 113).
Kozeny-Carman void ratio function [esoil3/(1+esoil)]
is plotted in Figure 1. The coefficient of determina-
tion (R2) of the regression function is 0.564, based 3 DATABASE II (AC)
on 119 data points (i.e. n = 119). The scatter could
probably be further reduced if the other parameters 3.1 Original database
in equation (2) were known (cf. Chapuis & Auber- Waters (1998) used the D50 i.e. the sieve aperture
tin 2003). The standard error (SE) of the regression through which 50% of the granular material in the
is 0.940. The p-value of the regression is < 0.0001. mix would pass in conjunction with the percent-
The RD value for the regression shown in Figure 1 age air voids, to define the Normalised Voids (NV)

136
 -8.0 No. Trendline Equation R2 RD(%) n
Q) k =5.81 x1 0·7[esoi1p.o2 0.972 16.7 33
-9.0 Q) ........ k =7.99x10·B[esoi1]4.78 0.692 55.5 13
® k =8.13x1Q·7[esoil]1 -53 0.969 17.6 6 ® •
@ -·· k= 1.37x10·7[esoii]2-9B

....··· _.- ---

0.993 8.4 6

V)
-10.0 ®
®
---
-··
k =5.21 x1 O·B[esoi1p.ss
k =2.80x10·7[esoil]4.03
0.998
0.996
4.5
6.3
4
4
"" __

eE -11.0 (j) -.- k =1.08x1Q·7[esoilp.ss 0.985 12.3 13

® --
........ k= 1.01x10·7[esoi1]3-B2 0.993 8.4 5
® k =5.43x10·7[esoil]4-45 0.887 33.6 17
c::: @ ---- k =2.06x10·7[esoilJ5.32 0.880 34.6 18
"'C
GJ
-12.0 All
All (excl. ®l
--
---
k= 3.61x10-7[esoi1]3-29
k =2.70x1Q-7[esoilp.as
0.573
0.624
65.4
61 .3
119
113
f/)
f/)
GJ -13.0
....
a. ®

=.
><
GJ -14.0
______ :).
~ -15.0
c 0------
-16.0

-17.0

-18.0
-0.8 -0.3 0.2 0.7 1.2
ln(e50u)

Figure 2. The relationship between ln(k) and ln(esoil) for the soils in Database I (soil).

as (NV) = [AV(%)][D50(mm)]/4.75. Vardanega & pore size (based on numerical simulations). Waters
Waters (2011, 2015) assembled a laboratory data- (1998) compared the Normalised Voids param-
base and after performing regression analysis pro- eter (NV) with asphalt concrete texture depth and
posed equation (4): showed that a positive correlation existed (i.e. as
NV increased so did the texture depth).
k = 0.46(Rp)3.70 [R2 = 0.74; n = 467] (4)
3.2 Expanded database
where k has units of mm/s and Rp has units of
mm and where Rp is the Representative Pore Size In this paper, the database of Vardanega & Waters
(RPS), where, Rp = (2/3)[AV(%)/100][D75(mm)]. (2011, 2015) has been expanded with data from an
Vardanega & Waters (2011, 2015) selected the additional ten publications. This has increased the
effective particle size statistically and determined number of permeability measurements available for
that the D75 (i.e. the sieve aperture through which regression analysis to over 1300. The first ten entries
75% of the granular material in the mix would pass) in Table 2 are the newly added data sources in this
was the best fit to the entire dataset (although any paper.
effective particle size upwards of D50 i.e. the ‘coarse In those publications where percentage Air Voids
fraction’, was shown to yield a good fit to the data). (AV) were measured using a ‘vacuum method’ in
Both the normalized voids and the representative addition to other methods the ‘vacuum’ data was
pore size concept are based on channel theory work entered into the database. When the NMAS was
detailed in Taylor’s classic textbook (e.g. Taylor not stated in the original reference, it was assigned
1965, p.109–110). A more extensive discussion of based on the supplied gradation curve.
the derivation of these parameters is given in Waters Table 2 shows the R2 of the regression of ln(k)
(1998) and Vardanega & Waters (2011, 2015). with ln(AV) and ln(Rp) with the Deff being assigned
Goode & Lufsey (1965) observed that “Air per- as D10, D25, D50, D75 and D90. For seven sub-datasets
meability is a function of aggregate gradation as the Deff is not required statistically or weakens the
well as air voids. The effect of gradation was found regression (this may be because the range of grada-
to be much more pronounced at high air voids tions in the sub-dataset is not significant) but for
than at low air voids”. Zhang et al. (2014) used the nine of the sub-datasets improvement is shown.
representative pore size to study pavement perme- Analysis of the entire data-set (n = 1318) shows
ability and skid resistance; reporting that perme- that D50 is actually the preferred choice for the
ability increases with increasing representative effective particle size. The D50 was used by Waters

137
Table 2. Summary of Database II (AC).

R2 of regression ln(k) vs. ln(Rp, Deff=Dx)***

Stated AV Stated k
NMAS measurement measurement ln(Rp, ln(Rp, ln(Rp, ln(Rp, ln(Rp, Best
Reference (mm) (*) n** method method ln (AV) D10) D25) D50) D75) D90) fit

Choubane 9.5(1), 151 FM 1-T Falling head 0.508 – 0.432 0.536 0.499 0.494 D50
et al. (1997, 12.5(85), 166(4) &
1998) 19(65) FM1-T
209(5)
Al-Omari 12.5(8), 13 ASSHTO Karol-Warner 0.916 – 0.382 0.588 0.457 0.455 AV
et al. 19(2), T-166 & permeameter
(2002) 25(3) Corelok
Bhattacharjee 9.5(25), 88 Vacuum Falling head 0.632 0.751 0.728 0.678 0.706 0.693 D10
& Mallick 12.5(45), seal &
(2002) 19(18) Saturated
Surface Dry
Brown et al. 9.5(131), 404 Vacuum ASTM PS 0.532 0.341 0.375 0.421 0.365 0.342 AV
(2004 12.5(209), seal & 129-01;
a,b,c,d) 19(64) ASSHTO NCAT Field
T-166 Permeameter
Haddock & 9.5(5) 5 FDOT method 0.384
Prather
(2004)
Chen et al 19(42) 42 AASHTO Falling head 0.786 0.385 0.963 0.940 0.917 0.893 D25
(2004) T269
Norambuena- 11.2(18), 72 geometric BS 1377–6 0.845 0.812 0.831 0.846 0.846 0.832 D75
Contreras 16(54)
et al. (2013)
& Noram-
buena-
Contreras
(2014)
Schmitt et al. 12.5(19) 19 Falling head 0.072 0.001 0.018 0.021 0.003 0.027 AV
(2007) test with
NCAT device
Pease et al 12.5(6) 6 ASTM D5084 0.379 0.736 0.518 0.424 0.650 0.487 D10
(2010)
Yan et al. 9.5(12), 51 Falling head 0.519 0.646 0.534 0.591 0.647 0.642 D75
(2016) 12.5(13), (FDOT)
19(7),
25(6),
37.5 (13)

Previous database (Vardanega & Waters 2011, 2015)

Hewitt (1991) 10(27) 27 Falling head 0.800 0.865 0.879 0.923 0.911 0.921 D50
Maupin (2000, 9.5(38), 189 VDOT TM120 0.606 0.362 0.325 0.463 0.557 0.514 AV
2009) 12.5(151)
Cooley Jr. 9.5(30), 130 Field 0.430 0.410 0.564 0.683 0.735 0.762 D90
et al. (2001) 12.5(45), permeameter
19(20),
25(35)
Kanitpong 12.5(19) 19 ASTM D5084 0.913 0.724 0.679 0.678 0.791 0.877 AV
et al. (2001)
Mallick et al. 9.5(20), 49 Falling head 0.413 0.620 0.606 0.789 0.807 0.875 D90
(2003) 12.5(10),
19(9),
25(10)
Vardanega 14(53) 53 Falling head 0.380 0.702 0.687 0.722 0.682 0.630 D50
et al. (2008)
ALL n = 1318 0.516 – 0.444 0.554 0.548 0.517 D50

* Number of samples at this NMAS;

** Number of measurements where both a k value and AV value are given (tests where k or AV reported as zero were excluded from
the regression analyses);
*** Values in bold indicate the highest R2 for the data subset.

138
(1998) when developing the ‘normalised voids’ Figures 5 and 6 show the database subdivided into
approach. Statistically, the D75 remains an accept- mixtures that have NMAS ≤ 12.5 mm (Figure 5)
able choice for Deff (the R2 is similar to that of D50 and those with NMAS > 12.5 mm (Figure 6).
for the whole dataset). In order to compare the
analysis of the updated database with the regres- 3.3.1 NMAS ≤ 12.5mm
sion relationships reported in Vardanega & Waters Figure 5 shows (as with Figure 4) that the fitted
(2011, 2015) the D75 will be used to compute the trend-line is similar to that proposed in Vardanega
RPS in the analysis that follows in this paper. & Waters (2011, 2015). Therefore, for mixtures
Figure 3 shows that ln(AV) is a key predictor with NMAS ≤ 12.5 mm we can write:
of ln(k) (as is void ratio for fine-grained soils, see
Figure 2). Figure 4 shows that the inclusion of the ln(k) = 3.668 ln(Rp) – 1.313 (6)
effective particle size via the Rp parameter does [R2 = 0.603; n = 917; SE = 1.351; RD = 63.0%]
slightly increase the R2 (and visually reduce some
of the scatter) but the effect is not as marked as where k has units of mm/s and Rp has units of mm.
that reported in Vardanega & Waters (2011, 2015). Equation (6) can be re-arranged to give:
The updated trend-line is given as equation (5):
k = 0.269(Rp)3.67. (6a)
ln(k) = 3.578 ln(Rp) – 1.645 (5)
[R2 = 0.548; n = 1318; SE = 1.581; RD = 67.2%] For this sub-dataset (NMAS ≤ 12.5 mm) the R2
value when ln(k) is regressed against ln(AV) is 0.555.
where k has units of mm/s and Rp has units of mm.
Equation (5) can be re-arranged to give:
3.3.2 NMAS > 12.5 mm
k = 0.193(Rp)3.58. (5a) Figure 6 shows (as with Figure 4) that the fitted
trend-line is similar to that proposed in Vardanega
3.3 Influence of NMAS & Waters (2011, 2015). Therefore, for mixtures
with NMAS > 12.5 mm we can write:
The NMAS may affect the co-efficient of perme-
ability as recently reported by Yan et al. (2016). ln(k) = 4.123 ln(Rp) – 1.872 (7)
NMAS is not explicitly accounted for in equation 5.
[R2 = 0.540; n = 401; SE = 1.875; RD = 67.8%]
4.0
ln(k) = 4.6291n(AV) + 7
R' =0.516
RD=69.6%
SE=1 .635 VardanegaandWaters(201
~ 0.0 n = 1318
p<0.0001
E -2.0
c:
~ -4.0
..
"tl

~ -6.0 ln(k)=3.6681n(Rp)-1.313
[ R' =0.603
~ -8.0 ~ RD=63.0%
SE=1.351
n=917
~ -10.0
~ -10.0 p<0.0001

-14.0 -14.0

-16.0 +-----.----.------.---.------,.-----r------. -16.0-4-l-.0--~-----r---~-1~.0---.-----.---~

-3 .0 -2.5
ln(Rp)[mm)
ln(AV) (numerical)

Figure 3. ln(k) versus ln(AV) for the expanded database Figure 5. ln(k) versus ln(Rp) for the mixtures in the
(air voids expressed numerically not as a percentage). database with NMAS ≤ 12.5 mm.
6.0

VardanegaandWaters(201
4.0 VardanegaandWaters(2011)

~ 2.0

~ 0.0
·= -2.0
"tl
=3.5781n(Rp)-1 .645
~
"tl
R' =0.548 -4.0
Ill
RD=67.2%

~ SE=1 .581
n= 1318
p<0.0001
!::::
~-10.0
ln(k) = 4.1231n(Rp) -1 .872
R'=0.540
RD=67.8%
SE=1.875
~-10.0 n=401
-12.0 p<0.0001

-14.0

-16.0 +-----.----.------.---.------,.-----r------. -16.0 +----.------.------.-----..-----..--------.

-4.0 -1.0
-4.0 -1 .0 0.0
ln(Rp)[mm] ln(Rp)[mm)

Figure 4. ln(k) versus ln(Rp) for the expanded database Figure 6. ln(k) versus ln(Rp) for the mixtures in the
(D75 was used as the effective particle size). database with NMAS > 12.5 mm.

139
 l.OE+02
measurements on road building materials. The
-;;;- l.OE+01 analysis of Database I (soil) revealed that the expo-
E"
E l.OE+OO nent on void ratio varies and on average is about
::::
g l.OE-01 3.3. Database II (AC) was used to show that the
~ original empirical equation presented in Varda-
~l.OE-02
nega & Waters (2011, 2015) is similar to that gen-
0 l.OE-03
erated using the expanded database presented in
·u
]
l.OE-04
this paper. The revised exponent on the Rp param-
u l.OE-05 eter was computed to be around 3.6. However, the
l.OE-06 +--"""""~......_"""+-_......_.._. .............t--__...__..................... influence of the grading parameter on reducing the
O.Dl 0.1
Representative Pore Size (Rp, mm)
10
scatter about the regression line is not as marked as
reported in Vardanega & Waters (2011, 2015). The
Figure 7. Comparison of regression equations derived effect of NMAS was statistically detectable but rela-
from the database analysis. tively minor compared to the influence of air voids.

where k has units of mm/s and Rp has units of mm.

ACKNOWLEDGMENTS
Equation (7) can be re-arranged to give:
CJS undertook some of this work with the support
k = 0.154(Rp)4.12. (7a)
of a University Research Committee Interdiscipli-
For this sub-dataset (NMAS > 12.5 mm) the R2 nary Research Internship in 2015 at the University
value when ln(k) is regressed against ln(AV) is 0.470. of Bristol. SF thanks Dr Erdin Ibraim for co-
Figure 7 shows equations (4, 5a, 6a and 7a) supervising her MSc project. PJV thanks Mr Max
compared. Bearing in mind the degree of scat- James for his helpful comments.
ter in the data shown in Figures 5 and 6 there
is not a significant difference between the two
sub-sets in the database (NMAS > 12.5 mm and REFERENCES
NMAS ≤ 12.5 mm), however for smaller datasets
Abdullah, W.S., Obaidat, M.T. and Abu-Sa’da, N.M.
with consistent testing methods the effect may be 1998. Influence of Aggregate Type and Gradation
more visible (cf. Yan et al. 2016). on Voids of Asphalt Concrete Pavements. Journal of
Materials in Civil Engineering, 10(2): 76–85.
Al-Omari, A., Tashman, L., Masad, E., Cooley, A. and
4 DISCUSSION Harman, T. 2002. Proposed Methodology for Predicting
HMA Permeability. Journal of Association of Asphalt
The results of the analysis of Database II (AC) Paving Technologists 71: 30–58 (includes discussion).
reveal that while the representative pore size (Rp) Al-Tabbaa, A. and Wood, D.M. 1987. Some measure-
remains a useful parameter to predict the coefficient ments of the permeability of kaolin. Géotechnique
of permeability, there are limitations—including: 37(4): 499–503.
Bear, J. 1972. Dynamics of fluids in porous media, Ameri-
i. use of power laws means that coefficient of can Elsevier Publishing, New York, NY.
permeability values reported as ‘zero’ cannot be Bhattacharjee, S. and Mallick, R.B. 2002. An Alterna-
included (even if some air voids are reported) in tive Approach for the Determination of Bulk Spe-
the database; cific Gravity and Permeability of Hot Mix Asphalt
ii. some of the scatter on Figure 4 is probably (HMA). International Journal of Pavement Engineer-
due to the mixing and matching of various air ing, 3(3): 143–152.
void measurement methods being present in Brown, E.R., Hainin, M.R., Cooley, A. and Hurley, G.
2004a. Relationships of HMA in-place air voids, lift
the database (connected voids are not always thickness, and permeability Volume One. NCHRP web
explicitly measured in practice). document 68 <http://onlinepubs.trb.org/onlinepubs/
Therefore, further work studying the influence nchrp/nchrp_w68v1.pdf> (19 Nov. 2016).
of the ‘shape of the grading curve’ on asphalt con- Brown, E.R., Hainin, M.R., Cooley, A. and Hurley, G.
2004b. Relationships of HMA in-place air voids, lift
crete permeability is proposed. Some preliminary thickness, and permeability Volume Two. NCHRP web
work using ‘grading entropy’ (e.g. Lőrincz 2005) document 68. http://onlinepubs.trb.org/onlinepubs/
has recently been undertaken (James 2015) as part nchrp/nchrp_w68v2.pdf (19 Nov. 2016).
of this on-going study. Brown, E.R., Hainin, M.R., Cooley, A. and Hurley,
G. 2004c. Relationships of HMA in-place air voids,
lift thickness, and permeability Volume Three.
5 SUMMARY NCHRP web document 68. http://onlinepubs.trb.org/
onlinepubs/nchrp/nchrp_w68v3.pdf (19 Nov. 2016).
This paper has presented the results of the analy- Brown, E.R., Hainin, M.R., Cooley, A. and Hur-
sis of two databases of hydraulic conductivity ley, G. 2004d. Relationships of HMA in-place air

140
 voids, lift thickness, and permeability Volume Four. Haddock, J.E. and Prather, M. 2004. Investigation of
NCHRP web document 68. <http://onlinepubs.trb.org/ permeability on Indiana SR-38. Journal of Perform-
onlinepubs/nchrp/nchrp_w68v4.pdf> (19 Nov. 2016). ance of Construction Facilities, 18(3): 136–141.
Carman, P.C. 1938. The determination of the specific Hainin, M.R., Yusoff, N.I. Md., Satar, M.K.I.M.S. and
surface of powders. I. Journal of the Society of Chemi- Brown, E.R. 2013. The effect of lift thickness on
cal Industry (Transactions), 57, 225–234. permeability and the time available for compaction
Carman, P.C. 1956. Flow of gases through porous media, of hot mix asphalt pavement under tropical climate
Butterworths Scientific Publications, London. condition. Construction and Building Materials, 48:
Carrier, W.D. 2003. Goodbye, Hazen; Hello, Kozeny- 315–324.
Carman. Journal of Geotechnical and Geoenvironmen- Hazen, A. 1892. Some physical properties of sands and
tal Engineering, 129(11): 1054–1056. gravels, with special reference to their use in filtration.
Chapuis, R.P. and Aubertin, M. 2003. On the use of the 24th Annual Report Pub. Doc. 34. Massachusetts State
Kozeny-Carman equation to predict the hydraulic Board of Health, Boston, pp. 539–556.
conductivity of soils. Canadian Geotechnical Journal Hazen, A. 1911. Discussion of ‘Dams on sand founda-
40(3): 616–628. tions’ by A.C. Koenig. Transactions American Society
Chen, J.-S., Lin, K.-Y. and Young, S.-Y. 2004. Effects of of Civil Engineers 73: 199–203.
Crack Width and Permeability on Moisture-Induced Hewitt, C. 1991. A study of asphalt permeability. BE the-
Damage of Pavements. Journal of Materials in Civil sis, University of Central Queensland, Rockhampton,
Engineering, 16(3): 276–282. Qld.
Choubane, B., Page, G.C. and Musselman, J.A. 1997. James, M.A. 2015. The Grading Entropy and Permeability of
Investigation of Water Permeability of Coarse Graded Road Surfaces. Undergraduate research project report
Superpave Pavements. Research Report FL/DOT/ (unpublished), University of Bristol, Bristol, UK.
SMO/97-416, Department of Transportation, FL. Kanitpong, K., Benson, C.H. and Bahia, H.U. 2001.
Choubane, B., Page, G.C. and Musselman, J.A. 1998. Hydraulic conductivity (permeability) of laboratory-
Investigation of Water Permeability of Coarse Graded compacted asphalt mixtures. Transportation Research
Superpave Pavements. Journal of Association of Asphalt Record 1767, Paper No. 01-2997, Transportation
Paving Technologists 67: 254–276 (includes discussion). Research Board, Washington, DC, 25–32.
Chung, S.G., Giao, P.H., Kim, G.J. and Leroueil, S. 2002. Kari, W.J. and Santucci, L.E. 1963. Control of Asphalt
Geotechnical properties of Pusan clays. Canadian Concrete Construction by the Air Permeability Test.
Geotechnical Journal, 39(5): 1050–1060. Journal of the Association of Asphalt Paving Technolo-
Cooley, L.A., Jr., Brown, E.R. and Maghsoodloo, S. gists, 32: 148–170 (includes discussion).
2001. Development of Critical Field Permeability and Kozeny, J. 1927. Üeber kapillare Leitung des Wassers in
Pavement Density Values for Coarse Graded Super- Boden. Wien. Akad. Wiss. 136(2a): 271–306. (in German).
pave Pavements. Report no. 01–03, National Centre for Kumar, A. and Goetz, W.H. 1977. Asphalt Hardening
Asphalt Technology, Auburn University, Auburn, Ala. as Affected by Film Thickness, Voids and Perme-
Cooley, L.A., Jr., Prowell, B.D. and Brown, E.R. 2002. ability in Asphaltic Mixtures. Journal of Association
Issues Pertaining to the Permeability Characteris- of Asphalt Paving Technologists 46: 571–605 (includes
tics of Coarse-Graded Superpave Mixes. Journal of discussion).
Association of Asphalt Paving Technologists 71: 1–29 Lekha, K.R., Krishnaswamy, N.R. and Basak, P. 2003.
(includes discussion). Consolidation of Clays for Variable Permeability and
Cooley, L.A., Jr. 1999. Permeability of Superpave Mix- Compressibility, Journal of Geotechnical and Geoenvi-
tures: Evaluation of field permeameters. Report no. ronmental Engineering, 129(11): 1001–1009.
99-01, National Centre for Asphalt Technology, Leroueil, S., Bouclin, G., Tavenas, F., Bergeron, L. and
Auburn University, Auburn, Ala. LaRochelle, P. 1990. Permeability anisotropy of natu-
Dołźyk, K. and Chmielewska, I. 2014. Predicting the ral clays as a function of strain. Canadian Geotechni-
co-efficient of permeability of non-plastic soils. Soil cal Journal, 27(5): 568–579.
Mechanics and Foundation Engineering 51(5): 213–218. Liebenberg, J.J.E., Rossmann, D. and Fletcher, C. 2004.
Fwa, T.F., Tan, S.A., Chuai, C.T. and Guwe, Y.K. 2001. Asphalt Mix Design and Construction: A Selection
Expedient Permeability Measurement for Porous of Possible Pitfalls. In Proceedings 8th Conference on
Pavement Surface. International Journal of Pavement Asphalt Pavements for Southern Africa (CAPSA ’04)
Engineering, 2(4): 259–270. 12–16 September 2004, Sun City, South Africa, Paper
Gerke, R.J. 1982. In situ testing infiltration of water No. 035.
through road pavements. In Proceedings 11th Austral- Lőrincz, J., Imre, E., Gálos, M., Trang, Q.P., Rajkai, K.,
ian Road Research Board Conference. Part 2. Australian Fityus, S. and Telekes, G. 2005. Grading Entropy Var-
Road Research Board, Melbourne, VIC, pp. 178–192. iation Due to Soil Crushing. International Journal of
Gogula, A., Hossain, M., Romanoschi, S. and Fager, Geomechanics, 5(4): 311–319.
G.A. 2003. Correlation between Laboratory and Field Mallick, R.E., Cooley, L.A., Jr., Teto, M.R., Bradbury,
Permeability Values for the Superpave Pavements. In R.L. and Peabody, D. 2003. An Evaluation of Factors
Proceedings of the 2003 Mid-Continent Transporta- Affecting Permeability of Superpave Designed Pave-
tion Research Symposium, Ames, Iowa, August 2003. ments. Report no. 03–02, National Centre for Asphalt
<http://www.ctre.iastate.edu/pubs/midcon2003/gogu- Technology, Auburn University, Auburn, Ala.
lasuperpave.pdf> (19 Nov. 2016). Masad, E., Al-Omari, A. and Lytton, R. 2006. Simple
Goode, J.F. and Lufsey, L.A. 1965. Voids permeability, Method for Predicting Laboratory and Field Perme-
film thickness vs. asphalt hardening. Journal of Asso- ability of Hot-Mix Asphalt. Transportation Research
ciation of Asphalt Paving Technologists 34: 430–463 Record 1970, Transportation Research Board, Wash-
(includes discussion). ington, DC, pp. 55–63.

141
Masad, E., Birgisson, B. and Al-Omari, A. 2004. Analyti- Tarefder, R.A., White, L. and Zaman, M. 2005. Neural
cal Derivation of Permeability and Numerical Simula- Network Model for Asphalt Concrete Permeability.
tion of Fluid Flow in Hot Mix Asphalt. Journal of Journal of Materials in Civil Engineering, 17(1): 19–27.
Materials in Civil Engineering, 16(5): 487–496. Taylor, D.W. 1965. Fundamentals of Soil Mechanics, John
Maupin, G.W. Jr. 2000. Asphalt permeability testing in Wiley & Sons Inc, New York, NY.
Virginia. Transportation Research Record 1723, Paper Vardanega, P.J., Natatmadja, A., Waters, T. and Ramanu-
No. 00–1206, Transportation Research Board, Wash- jam, J. 2008. A study of asphalt permeability: Empiri-
ington, DC, pp. 83–91. cal permeability models. In Proceedings of the 23rd
Maupin, G.W. Jr. 2009. Personal Communication. ARRB Conference – Research Partnering with Practi-
McLaughlin, J.F. and Goetz, W.H. 1955. Permeability, tioners (CD-ROM), ARRB Group, Melbourne, VIC.
Void Content, and Durability of Bituminous Concrete. Vardanega, P.J. 2014. State of the Art: Permeability of
Highway Research Board Proceedings, 34: 274–286. Asphalt Concrete. Journal of Materials in Civil Engi-
Mohammad, L.N., Herath, A. and Huang, H. 2003. neering 26(1): 54–64.
Evaluation of Permeability of Superpave ® Asphalt Vardanega, P.J. and Waters, T.J. 2011. Analysis of
Mixtures. Transportation Research Record 1723, Paper Asphalt Concrete Permeability Data Using Repre-
No. 03-4464, Transportation Research Board, Wash- sentative Pore Size. Journal of Materials in Civil Engi-
ington, DC, pp. 50–58. neering 23(2): 169–176.
Mullen, W.G. 1967. Beam flexure and permeability testing Vardanega, P.J. and Waters, T.J. 2015. Erratum for “Anal-
of bituminous pavement samples. Journal of the Asso- ysis of Asphalt Concrete Permeability Data Using
ciation of Asphalt Paving Technologists 36: 615–631. Representative Pore Size” by P.J. Vardanega and T.J.
Norambuena-Contreras, J., Izquierdo, E.A., Castro- Waters. Journal of Materials in Civil Engineering,
Fresno, D., Partl, M.N. and Garcia, Á. 2013. A New 27(8): 08215002.
Model on the Hydraulic Conductivity of Asphalt Vivar, E. and Haddock, J.E. 2007. Hot-mix asphalt per-
Mixtures. International Journal of Pavement Research meability and porosity. Journal of the Association of
and Technology, 6(5): 488–495. Asphalt Paving Technologists 76: 953–979.
Norambuena-Contreras, J. 2014. Personal Communication. Walker, L.K. and Raymond, G.P. 1968. The prediction
Olsen, H.W. 1960. Hydraulic Flow Through Saturated of consolidation rates in a cemented clay. Canadian
Clays. Clays and Clay Minerals 9(1): 131–161. Geotechnical Journal, 5(4): 192–216.
Olson, R.E. and Daniel, D.E. 1981. Measurement of Waters, T.J. 1998. A study of water infiltration through
the Hydraulic Conductivity of Fine-Grained Soils. asphalt road surface materials. In Proceedings of the
In Zimmie, T.F. and Riggs, C.O. (Eds.) Permeability International Symposium on Subdrainage in Roadway
and Groundwater Contaminant Transport ASTM STP Pavements and Subgrades. Asociación Técnica de Car-
746, American Society for Testing and Materials, West reteras, Madrid, Spain, 311–317.
Conshohocken, PA, pp. 18–64. Waters, T.J. and Vardanega, P.J. 2009. Re-examination of
Pane, V., Croce, P., Znidarcic, D., Ko, H.-Y., Olsen, H.W. the coefficient of determination (r2) using road mate-
and Schiffman, R.L. 1983. Effects of consolidation rials engineering case studies. Road and Transport
on permeability measurements for soft clay. Géotech- Research 18(3): 3–11.
nique, 33(1): 67–72. Yan, Y., Zaniewski, J.P., Hernando, D. 2016. Develop-
Pease, R.E., Stormont, J.C., Hines, J. and O’Dowd, D. ment of a predictive model to estimate permeability of
2010. Hydraulic Properties of Asphalt Concrete. Geo- dense-graded asphalt mixture based on volumetrics.
technical Testing Journal, 33(6): 445–452. Construction and Building Materials 126: 426–433.
Schmitt, R.L., Owusu-Ababio, S., Crovetti, J.A. Zhang, L., Ong, G.P. and Fwa, T.F. 2014. Influences of
and Cooley, A.L., 2007. Development of In- Pore Size on the Permeability and Skid Resistance
Place Permeability Criteria for HMA Pavement of Porous Pavement. 3rd International Conference on
in Wisconsin (No. WHRP 06–15). <http://minds. Transportation Infrastructure, ICTI 22–25 April 2014,
wisconsin.edu/bitstream/handle/1793/53421/06– Pisa, Italy. pp. 621–632.
02%2520Final%2520Report.pdf> (22 Feb. 2017). Zube, E. 1962. Compaction of asphalt concrete pave-
Smith, R.W. and Gotolski, W.H. 1969. A study of physi- ment as related to the water permeability test. In:
cal factors affecting the durability of asphaltic pave- Proceedings 41st Annual Meeting Highway Research
ments. PB 189 832, Pennsylvania State University, Board, Washington DC, pp. 12–37.
University Park, PA.

142
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Laboratory investigations of thermal properties of crushed

rock materials

K. Rieksts, I. Hoff & E. Kuznetsova

Norwegian University of Science and Technology, Norway

J. Côté
Laval University, Québec, Canada

ABSTRACT: During the last decades in Norway, natural (fluvial, glaciofluvial) gravel resources have
been substituted with crushed rock material for road and railway construction. However, the thermal
properties of these materials in some cases are significantly different compared to natural materials.
Therefore, evaluation of thermal conductivity of these diverse crushed rock materials is important.
For engineering purposes, empirical models are widely used to assess the thermal conductivity. Most of
the currently developed models are based on soil properties. However, it is observed that thermal conduc-
tivity of crushed aggregates can vary significantly due to their particles shape and size, porosity, mineral-
ogy, quartz content and density. Therefore, the use of these models might lead to false results.
In a frost design method, the required parameters of crushed rock aggregates are thermal conductiv-
ity, density and water content. The heat transfer during the freezing of natural soils is assumed propor-
tional to thermal conductivity of the material. In the current calculations used by Norwegian Public Road
Administration, thermal conductivity of granular layers is considered mostly being below 2 W/mK based
on the water content. However, it has been shown that the thermal conductivity of crushed rock materials
can vary from 1 to 4 W/mK depending on other material properties in addition towater content.
The main objectives of our investigations are: i) to measure thermal conductivity of crushed aggregates
and assess the effect of various mineralogies ii) to use the obtained data to validate an existing empirical
model.

1 INTRODUCTION frost depth penetration is one of the key param-

eters to design a sufficiently thick structure layer
During the last decades road construction prac- to eliminate freezing of frost susceptible soils.
tice in Norway has changed significantly. This is Thermal conductivity is the most important soil
mainly due to increasing environmental concern to characteristic that defines the heat transfer in road
exploit natural mineral resources, such as gravel, structures (Farouki, 1981; Johansen, 1975).
and subsequent replacement with crushed rock For engineering and design purposes, empirical
material. These materials, unlike natural soils, can models are often used. Two ofthe most widely used
have different mineralogy and fabric, which refers models proposed by Johansen (1975) and Kersten
to the size and arrangement of particles as well as (1949) were originally develop for fine and medium
pore space distribution. size soils. As such, these models have been adopted
Although in vast majority of the land bed- to crushed rock materials neglecting the contrast-
rock is typically found in shallower or deeper lay- ing properties of these two classes of materials. In
ers, a great part of the populated area is covered fact, heat and mass transfer for road construction
by fine-grainedsoils. This makes design process, materials have been systematically assessed only
both mechanically and thermally, complicated to during the last decade. Still larger quantities of
ensuresufficient structural integrity. Some silty experimental data are required to develop more
and clayey materials are frost susceptible causing precise soil type based models with higher degree
frost heave issues during the wintertime and conse- of accuracy (Côté & Konrad, 2005a, 2005b).
quently significant loss of bearing capacity during These models used by engineers completely or
thawing season. partly neglect the influence of mineralogy on the
Frost action analysis in cold regions like overall thermal conductivity of crushed material.
Norway is crucial to sustain good driving condi- However, it has been found that exact mineralogy
tions for both, roads and railways. Calculation of has to be known to obtain reliable modeling results.

143
Moreover, todays strategical target is aimed at the 100
use of local materials and it should be noted here
that mineralogical content can vary significantly
just within a relatively small region. Besides, ther-
90
;t. 80
~ 70 1-
If
mal conductivity values for materials with different -~ 60
mineralogical origin can vary between 1–5 W/mK.
..
~50
Therefore, it is important to consider these factors
when any thermal calculation is performed (Côté
& Konrad, 2005a, 2005b; Ramstad et al., 2015).
5 40
=
~ 30 1-
vv
~ 20
The guidelines and specification for frost pro-
tection of road are defined in Norwegian Road
Administration issued handbook N200. Last revi-
10
0
0.01
-
0.1
t;
10 100
sion of N200 handbook (Handbook N200, 2014) Sieve size, mm
took place at 2014 when new specifications of frost
protection layer were introduced. Among them, Figure 1. Sample gradation curve.
allowable content of fines were increased upto 15%
for materials with particles smaller than 22.4 mm. Table 1. Sample mineralogy.
The fundamental intension behind frost protec-
tion layer is the prevention offrost penetration Quarry
down into frost susceptible subsoils. Objective of
increased fines content was based upon the argu- Mineral Lauvåsen Meraftåsen Hello Vassfjell
ment that this will retain more water and subse-
quentlyincrease latent heat offusion. However, in Quartz 22% 5% 8% <1%
a research done by Kuznetsova et al. (2015) it was Plagioclase 23% 33% 20% 21%
shown that increased fines content does not neces- K-feldspar 2% 3% 2%
sarily result in increased water content. Pyroxene 1% 2% 1% 1%
Amphibole 10% 49% 35%
Furthermore, specifications of N200 handbook
Clinozoisite 30%
tend to neglect the importance of other factors
Epidote 18% 16%
affecting thermal conductivity appart from the
Mica 15% 3%
water content. However, in research by Côté and
Chlorite 17% 19% 6% 10%
Konrad (2005b) they demonstrated that thermal Calcite 20% 6% <1%
conductivity of crushed rock materials can vary Sphalerite <1% <1%
significantly up to 4 W/mK based on mineralogy, Rhodochrosite <1%
dry density and water content.
The study presented in this paper is a part of
a larger project looking into specifics about heat
50 blows per layer. This ensured high material den-
and mass transfer as well as frost heaving issues
sity and low porosity varying between 0.16–0.17.
in road construction materials. The objectives of
Test materials were collected from four quar-
the current research described in this paper is to
ries chosen to obtain diverse mereologies. Table 1
i) measure thermal conductivity of crushed aggre-
presents mineralogy data for all four materials.
gates and assess the effect of various mineralogies
ii) to use the obtained data to validate an existing
empirical model.
3 EXPERIMENTAL PROCEDURE

3.1 Test setup

2 MATERIALS
Thermal conductivity measurements are done
The current Norwegian regulations states that using a steady state test method. Test sample
fines content in frost protection material should (Fig. 2) is placed in a temperature controlled cabi-
be between 2–15% from material below 22.4 mm net. The temperature in the cabinet was set as the
(Handbook N200, 2014). For this study, a medium mean temperature of the boundary conditions
value of 7%was chosen. The maximum particle imposed to the sample.
size was set to 16 mm and the gradation curve was Test sample comprises of test material com-
designed using Fuller’s equation to obtain a dense pacted into cylindrically shaped PVC mold with
graded material (Fig. 1). inner diameter of 100 mm and height of 75 mm.
Target water content during sample preparation Test sample is placed between two 25.4 mm thick
was set to 4% that led to experimental values vary- borosilicate glass discs (Borofloat®) with thermo-
ing between 3.70–4.30%. Samples were compacted couples embedded in the center of the flat surfaces.
with manual proctor hammer in three layers with To obtain a steady temperature gradient through

144
 +7/-3
,,
...
',BOROSILICATE GLASS DISC
>thennocouples

\\
\
\
\
\
\\ temperature
SAMPLE \ gradient
\
\
\
\
\
\
\
\
BOROSILICATE GLASS DISC', ) them1ocouples

+3 /-7
Figure 2. Experimental test setup.
Figure 3. Cross section of test setup.

this three layer system, it is placed between two concept of normalized thermal conductivity with
independent heat exchanging plates setting the respect to dry and saturated states:
boundary conditions at the top and the bottom.
Figure 3 illustrates cross section of test setup. k kdrry
100 mm thick Styrofoam insulation is placed kr = . (2)
around both, the discs and test sample to minimize ks kdry
any radialheat loss and impose onlyone dimen-
sional heat transfer. The model was later improved by Côté and
Tests are carried out in unfrozen and frozen Konrad (2005a) and Côté and Konrad (2005b)
state applying +7/+3°C and −3/−7°C at the top and as such it has been adopted in this rearch. In
and the bottom, respectively. Temperatures from their study they proposed more accurate empiri-
all four thermocouples are recorded every three cal equation for defining dry conductivity (kdry) as
minutes with Campbell Scientific CR100 data a modified form of geometric mean with fitting
logger. parameters α = 0.59 and β = 0.73, respectively:
Heat flux is calculated as the average heat flux
α
ks( − n) kan .
through both glass discs. Borosilicate glass discs β
kdrry (3)
are materials with known thermal conductivity
that varies little with temperature. For temperature
range used in this study, it varies only from 1.03 W/ They also proposed a new soil type dependent
mK at −7°C to 1.05 W/mK at 7°C. Temperature relationship between normalized thermal conduc-
gradient in glass discs is calculated using thermo- tivity (kr) and degree of saturation (Sr):
couple data obtained from both faces of the disc.
κ Sr
Thermal conductivity of sample material is calcu- kr = , (4)
lated by dividing the average heat flux by tempera- 1+ ( − ) Sr
ture gradient in the sample:
where κ is soil type dependent factor to account
quf + qlf 1 for different soil types in the unfrozen and frozen
k= ⋅ . (1) states (κ = 4.7 for unfrozen base course and sub-
2 ∇T
base material and κ = 1.8 for frozen state).
Temperature gradient is equal to the temperature
difference at both sample-disk interface divided by 3.3 Thermal conductivity of solid particles
the sample height.
Thermal conductivity for solid particles is a crucial
element in order to have a high accuracy of estima-
3.2 Model
tion of thermal conductivity for porous material.
The model used in this paper was originally It can be measured or estimated in several ways. In
proposed by Johansen (1975). It is based on the this study, it was obtained two ways.

145
 The first way implies measuring material min- Table 2. Hokssund sand mineralogy.
eral content using XRD analysis. Geometric mean
method was used to calculate thermal conductivity Thermal
for solid rock particles (ks) (eq. 5). This data is con- conductivity
Mineral Percentage W/mK
sidered as an external parameter and independent
of test setup. Quartz 35 7.69
Na-feldspar 25 2.25
ks ∏k xi
mi , (5) K-feldspar 20 2.25
Mica 10 2.03
where kmi and xi are the thermal conductivity and Amphibole 5 3.46
volume fraction of mineral i, respectively. It is Others 5 2.00
noted that the sum of xi is equal to 1 and that typi-
cal km values can be found in Horai (1971).
The second way implies measuring thermal con-
ductivity of two phase material. This means that 2
thermal conductivity (ksat) of a water saturated
sample is measured and thermal conductivity for
solid rock particles is backcalculated employing
eq. (6) (Côté & Konrad, 2007). This method gives 1.5
good results since it is generally accepted that this
provides good estimation if thermal conductivity
contrast between solid particles and saturating
fluid is less than one order of magnitude.

⎛ 1 ⎞
⎜ ⎟ --calculated
⎛ k ⎞ ⎝ 1− n ⎠
ks = ⎜ sant ⎟ (6) 0.5 0 measured
⎝k ⎠ w

0
4 RESULTS AND DISCUSSION 0% 5% 10% 15%
w[%]

4.1 Test setup validation

Figure 4. Hokssund sand thermal conductivity.
Test setup validation was performed running a
series of tests with sand sample being gradually
dried. The general procedure was that the sam- the sample after drying or poor contact between
ple was tested and afterwards it was taken out the sample and borosilicate disc. The difference
and slowly dried in an oven in temperature range between calculated and measured values is very
between 30–40°C. After drying, it was placed back small for most data point and less than 15% for
in test equipment to run the next test at lower those two data points at medium moisture content.
moisture content. This sequence was continued It is thus assumed that the test setup operates with
until the last test with a completely dry material. sufficient degree of confidence.
It is generally accepted that thermal conductiv-
ity models perform fairly well with sand materi-
als and the Côté and Konrad model (2005) will 4.2 Thermal conductivity measurements
be used here as a means of validation for the test Figure 5 illustrates a typical test result. A steady
setup with a sample of Hokssund sand (mineral- state temperature distribution is obtained in both,
ogy shown in Table 2). The sample was compacted positive and negative temperature range. During
at a dry density of 1.83 g/cm3 and a moisture con- transition between unfrozen and frozen states, the
tent of 13.77%. effect of phase change of water can be observed as
Test results are displayed in Figure 4. Thermal a lag in temperature due to the latent heat of solidi-
conductivity of Hokssund sand is shown as a func- fication. A typical test requires less than 20 hours
tion of water content (w%) by weight. It can be to obtain steady state conditions in both tempera-
seen that measured values fits the model calcu- ture ranges.
lated values fairly well with some discrepancies at Figure 6 summarizes ks obtained from geo-
moderate water contents. This could be attributed metric mean and backcalculation methods,
to insufficient water redistribution throughout respectively. It is apparent from this figure that

146
 25 3.0
,,
oo, ,
,,
20
steady-

~
15
,\ ....state
,..,i l!hase "' "'
"' "'
f 10
B ~ I I change 2.0
f 5 • •"auy-
, , "' "'
e"'=- o I I J state
,, "' "'
~ ~ I+ +I
, "' "'
-5
I
: , , "' "' "' OLauvasen
·10
20 , "' , "' "' DMeraftasen

, , "' "'
0 10 15
6 Hello
Tlme,b
, "' X Yassfjell

Figure 5. Typical test sequence. .,

A. ,f;., "'

0.0
0.0 LO 2.0 3.0
5.0

4.0 0 3.89 Figure 7. Calculated thermal conductivity values com-

pared with measured for unfrozen conditions.
x- 2.89
I{ 306 0
~ 3.0 3.03
2.96 f{ 2.66
e X 2.48 2.61
~ 2.0
0 geometric mean
3.0
~-

LO x backcalculated

0.0
2.0
~

~
E
, " "'
1 " , "'
Figure 6. Thermal conductivity for solid rock particles ~ " "
ks for test materials from different locations. } LO
" " "' "'
/ """" "'
significant discrepancies might exist between "'"' "' Ogeometric mean
X l>ackcalculated
both ks estimation methods. However, the high- 0.0
est difference between thermal conductivities can 0 .0 LO 2.0 3.0
be observed for Lauvåsen material, varying from kmmund [W/mKJ
2.89 W/mK backcalculated from saturated sam-
ple to 3.89 W/mK from geometric mean method. Figure 8. Calculated thermal conductivity versus meas-
It is expected that this difference might introduce ured values using different ks.
a notable difference between thermal conductivity
values obtained from testing and from the model
for compacted moist materials. from some discrepancies in mineral content meas-
Figure 7 shows predicted thermal conductivity urements. This will be further investigated.
values compared to measuredvalues of the four Figure 8 displays calculated thermal conductiv-
tested crushed rock samples under unfrozen condi- ity values compared with measured values using
tions. Line of equality is added together with model different ks values. Comparing the two results, it
accuracy boundaries of ±20%. From the graph it can be seen that varying ks values can have a strong
can be seen that most of the data points fall within influence on the final result of the calculation.
the set boundaries. Yet, values calculated with the When ks value of 3.89 W/mK for Lauvåsen material
model for Lauvåsen material strongly overpredicts is substituted with value of 2.89 W/mK obtained
the actual values obtain from thermal conductiv- with backcalculation method, prediction accuracy
ity tests. Since the rest of the data points fit the greatly improves when compared to measured
model predictions fairly well, it is assumed that the values. Similar effect can be observed with Hello
overestimation of Lauvåsen values may originate values to a lesser extent where ks substitution results

147
from slight overprediction into minor underpredic- complicated to distinguish between particular
tion of thermal conductivity values. minerals coming from the same class or subclass.
This indicates that the precision of model per- A decision by the technician evaluating measure-
formance is highly sensitive to ks variations. If ks ments has to be made and this might introduce a
values are not accurate, estimation of overall ther- human error leading to unprecise measurements.
mal conductivity of crushed rock material will not When XRD analysis are performed, it is often
be accurate either. recommended to use subclasses rather than partic-
However, general idea of the model is to give the ular mineral. However, if geometric mean (eq. 5)
user a way of estimating thermal conductivity val- method is employed to define ks, exact minerals
ues without running an actual test. It is therefore has to be known. Minerals coming from the same
important that agreat care be taken in the estab- subclass does not necessarily have the same ther-
lishment of ks, either with typical published values mal properties (Horai, 1971).
or through mineral content analyses using eq. 5. Johansen (1975) proposed in his work a sim-
Figure 9 displays predicted thermal conductiv- plified method to assess the ks. He proposed to
ity values compared with measured ones under consider only quartz content and to approximate
unfrozen and frozen conditions. Predicted values thermal conductivity of the rest of the minerals as
of thermal conductivity are calculated using ks val- 2 or 3 W/mK if quartz content is higher or less
ues obtain from backcalculation method. that 20%, respectively. However, Côté and Kon-
The result, as shown in Figure 9, indicated a rad (2005b) demonstrated that geometric mean
good fit between model data and measured values. method should be used for determining thermal
A clear shift towards higher thermal conductiv- conductivity of solid particles, especially if quartz
ity values can be seen for frozen conditions com- content is lower than 20%. Other minerals, e.g.
pared to unfrozen conditions (Figure 8). This is chlorite (also found in materials used in these tests)
attributed to water, with a thermal conductivity of can highly contribute to the overall thermal con-
0.6 W/mK, changing into ice that has a thermal ductivity of solid rock.
conductivity of 2.2 W/mK. Materials examined in this study had the quartz
content below 22% yielding ksvalues from 2.48–
3.03 W/mK. Materials used in these tests can be
4.3 Thermal conductivity of solid particles
considered with relatively low thermal conductiv-
Early models of thermal conductivity of mineral ity. Using Stefan’s equation as a simple estimation
materials tended to underestimate the importance of frost depth penetration it states that the frost
of thermal conductivity of solid particle. This is depth varies as a square root of frozen thermal
understandable as obtaining precise measurements conductivity. This yields that even with a small var-
of mineralogy requires sophisticated measuring iation observed in these tests, possible frost depth
devices and procedures. In addition, it is sometimes penetration could increase by roughly 9% if mate-
rial with highest thermal conductivity would be
used. However, if quartz content would be as high
3.0 as 50%, approximate ks values would be around
" " 4 W/mK. This could increase potential frost depth
" " "' penetration even more.

""" "' "'

"'
2.0 " "' "' 5 CONCLUSION
~
s "" "' "'
~ " "' "' This paper presents results from a small scale
1 " " " "' "' thermal conductivity tests performed under stae-
e
~~ 1.0 " " "' "' "' ady state conditions. Crushed rock materials
were tested under unfrozen or frozen conditions.
";
" "' "' "' Ofrozen
X unfrozen Four different mineralogies were used for sample
/. " "'
;
" preparation.
~ "' Thermal conductivity of solid particles were
0.0
"' estimated in two ways yielding some significant
0.0 1.0 2.0 3.0 variance. This consequently indicated a strong
kmeasu<ed [W/mK] influence on the predictions of thermal conduc-
tivity with the adopted model. Model should use
Figure 9. Calculated thermal conductivity values external parameter, such as XRD results, to allow
compared with measured ones in unfrozen and frozen estimating ks values, hovewer great care should be
conditions. takenevaluating their reliability.

148
 In general, model adopted in this study for cal- Quebec marine clays. Canadian Geotechnical Journal,
culation of thermal conductivity of crushed rock 44(9), 1117–1127. doi:10.1139/t07–049
material show a good estimation. Measured values Farouki, O. T. (1981). Thermal properties of soils. US
are fairly well within margin of error less than the Army Corps of Engineering, Cold Regions Research
and Engineering Laboratory, Hanover, NH, CRREL
usual 20% found in the literature. Monograph 81−1.
In order to fully adopt the model for Norwe- Handbook N200. (2014). Vegbygging (Road construc-
gian conditions itshould be examined with higher tion). Statens vegvesens håndbokserie. Norway.
variety of mineralogy, different water content and Horai, K. i. (1971). Thermal conductivity of rock-form-
material particle gradation. ing minerals. Journal of Geophysical Research, 76.5,
1278–1308.
Johansen, O. (1975). Thermal conductivity of soils.
ACKNOWLEDGMENT Ph.D. Thesis. University of Trondheim, Trondheim,
Norway. US Army Corps of Engineering, Cold Regions
Research and Engineering Laboratory, Hanover. N.H.
The authors wish to acknowledge the technical CRREL Draft English Translation 637, 291.
contribution of laboratory technicians NTNU/ Kersten, M. S. (1949). Laboratory research for the deter-
SINTEF: Bent Lervik, Jan Erik Molde and Lis- mination of the thermal properties of soils. Research
beth Johansen. This research was supported by the Laboratory Investigation, Engineering Experiment
Norwegian Research Council (NRC) under grant Station, University of Minnesota, Minneapolis, Minn.
246826/O70. Technical report.
Kuznetsova, E., Henri, G., Hoff, I., Solås, L. A., Vignali,
V., Lantieri, C.,... Miranda, I. (2015). Investigation
REFERENCES of crushed rock material used in the frost protection
layer (Norway). GEO Quebeck 2015 conference.
Ramstad, R. K., Kirsti, M., Liebel, H. T., Frengstad,
Côté, J., & Konrad, J.-M. (2005a). A generalized thermal
B. S., & Willemoes-Wissing, B. (2015). Thermal con-
conductivity model for soils and construction materi-
ductivity map of Oslo region based on thermal dif-
als. Canadian Geotechnical Journal, 42(2), 443–458.
fusivity measurements of rock core samples. Bulletin
Côté, J., & Konrad, J.-M. (2005b). Thermal conductivity
of Engineering Geology and the Environment, 74(4),
of base-course materials. Canadian Geotechnical Jour-
1275–1286.
nal, 42(1), 61–78.
Côté, J., & Konrad, J.-M. (2007). Indirect methods
to assess the solid particle thermal conductivity of

149
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Stiffness and permanent deformation characteristics of open-graded

unbound granular materials

M.S. Rahman, S. Erlingsson & F. Hellman

Pavement Technology, VTI, Linköping, Sweden

ABSTRACT: This paper presents the investigation of the deformation characteristics of two Open-
Graded (OG) Unbound Granular Materials (UGMs) that are generally used in permeable pavement
structures. Based on multistage Repeated-Load Triaxial (RLT) tests, the stiffness and resistance against
Permanent Deformation (PD) of these materials were compared to their well graded (standard) counter-
parts. The OG UGMs showed similar stiffness characteristics as standard UGMs. On the other hand, the
resistance against PD was slightly worse than the standard UGMs in a relatively dry state. However, the
OG UGMs showed less sensitivity to moisture variation, yielding similar performance to that of standard
materials at higher moisture content. The parameters of a stiffness model and a PD model were evaluated
for these materials which may be used for prediction of rutting, estimating bearing capacity and for layer
thickness design, using any suitable pavement design software.

1 INTRODUCTION types are built where the OG base and sub base lay-
ers allow for the movement of water, act as a water
Pavements are generally composed of several lay- reservoir for the trees and vegetation, and allow
ers. The functional performance of a pavement their roots to grow and penetrate (Pratt, 1990; Lar-
structure greatly relies on the selection of materials son, 1990). Permeable pavements are usually built
and thicknesses of the different layers. Structural with interlocking paving blocks or porous asphalt
design of pavements involves selection of materi- with OG UGM layers underneath and are mostly
als and thicknesses of the different layers based utilized in parking lots or low trafficked areas. For
on load carrying requirements, topography and the base and sub base layers several permeable and
climatic conditions, bearing capacity of the exist- structural or skeletal soils have been proposed or
ing sub grade, availability of the materials and eco- utilized to meet the drainage requirements while
nomic considerations. The base and subbase layers maintaining the load bearing capacity of the struc-
of pavements are most commonly constructed of ture. The OG UGMs generally refer to the parti-
Unbound Granular Materials (UGMs) of well cle size gradation where the fine particles up to a
graded rock aggregates with dense configuration to certain size have been removed to create voids to
provide the desired stiffness and resistance against allow for the passage of water. Since the fine parti-
permanent deformation. However, these layers of cles are absent from the material, its density is less
well graded aggregates are relatively impermeable than that of the well graded material. As a result,
to water and therefore excess water may get trapped it is presumed that these kinds of materials may
inside the structure, which can accelerate damage perform worse than the well graded counterparts
and premature failure of the pavement (Hajek, due to lack of skeletal support from the fine par-
et al., 1992). Another issue is that, due to increased ticles. The existing design guidelines for standard
urbanization and construction of pavements, the materials therefore need to be revised when using
proportion of impervious land surface is increas- OG materials. Since practical experience regarding
ing, which poses a challenge for storm water the field performance of OG materials is relatively
management as well as hampering the growth of limited, it is necessary to carry out studies on the
urban trees and vegetation (Ahmed et al., 2016). mechanical properties of these materials. Today’s
One possible approach to solve the issue with demand for increased traffic load and load bearing
water trapping is to build pavements with Open- capacity necessitates the development of a robust
Graded (OG) drainage layers so that the pavement analytic design approach. For this it is necessary to
cannot retain any water (Hajek, et al., 1992). To formulate reliable predictive models to character-
address the storm water management and vegeta- ize the deformation behavior of OG UGMs that
tion growth issues permeable pavements of various can be used for the design of pavements.

151
 This paper presents a comparative study of range. Particles smaller than that size simply fill the
the stiffness properties and Permanent Deforma- voids with no significant contribution to the load
tion (PD) behavior of some OG UGMs against bearing capacity. In that regard, OG UGMs may
well graded UGMs of similar origin (referred to possess similar mechanical strength as standard
here as standard or reference material), based on ones if fines are removed up to a certain size in
MultiStage (MS) Repeated-Load Triaxial (RLT) compliance with the DASR concept.
tests. The aim was to implement the findings and The deformation characteristics of OG UGMs
enhanced understanding of the material behavior are similar to those of well graded UGMs. The
to improve or modify the existing design guidelines total deformation due to compressive cyclic stresses
in future. Furthermore, two existing simple models in a UGM consists of two parts: (a) elastic or recov-
were implemented to characterize the stiffness and erable or Resilient Deformation (RD) and (b) irre-
PD behavior of these materials and thus a set of versible or plastic or Permanent Deformation (PD).
material parameters was generated. The objective For a single load pulse, this can be expressed as:
was to come up with a simple analytical design tool
which may be used for the layer thickness design and ε tot = ε r + ε p (1)
performance prediction of the pavement structure.
where, εtot is the total axial strain, εr is the axial
resilient strain and εp is the axial permanent strain.
2 PROPERTIES OF OG UGMS Usually, the resilient strain is much larger than the
permanent strain (Erlingsson and Magnusdottir
OG UMGs are UGMs where the fine portions up 2002, El Abd et al., 2005). Both kinds of defor-
to a certain size are absent from the grading. The mation in UGMs are non-linear stress depend-
objective is to create voids in the aggregate assem- ent (Kolisoja 1997, Lekarp 1999, Englund 2011).
bly so that water can easily penetrate and flow Deformations occurring in a UGM assembly have
through. Selection of the Particle Size Distribution been attributed to the elastic deformation, bend-
(PSD) of the materials is governed by the desired ing, breaking, crushing, sliding and rolling of the
use of the materials and infiltration capacity individual particles (Lekarp et al., 2000).
requirements. Figure 1 shows a typical distribution
of hydraulic conductivities for a UGM with differ-
ent PSDs as determined by the Swedish Cement 2.1 RD characteristics
and Concrete Institute (CBI). In the figure, for The resistance of a UGM against the RD is
example, 2/32 refers to the minimum particle size/ defined using the resilient modulus (or the resilient
maximum particle size in millimeters. stiffness), MR, which is an estimate of the stiffness
For UGMs, it is generally accepted that the PSD modulus of the specimen for rapidly applied loads,
with a Fuller curve having n = 0.45 yields the dens- expressed as (Huang 2004):
est aggregate assembly. Since for the OG UGMs the
fine portion of the particles is absent, the density σd
of the aggregate assembly is comparatively lower. MR = (2)
It is therefore likely that the mechanical resistance εr
of OG materials will be inferior. However, accord-
ing to the Dominant Aggregate Size range (DASR) where σd (or q) is the cyclic deviator stress. The
concept, the load bearing skeleton of a UGM resilient strain, εr, is taken after several load rep-
assembly is comprised of particles of a certain size etitions when the deformations are stabilized. The
MR is dependent on the state of stress, measured
as the sum of the principal stresses, called the bulk
stress, θ = σ1 + σ2 + σ3 or the mean normal stress
or the hydrostatic stress, p = θ/3. Other factors
• 0/32 that affect the MR are density, grading, moisture
• 2/32
• 2/32-rep content, stress history, aggregate type and shape
• 2/4 (Lekarp et al., 2000).
2/4-rep The variation of the MR with θ can be expressed
• 2/90
o 2/90-rep with the well-known k-θ model (Seed et al., 1962,
• 0/90 Hicks and Monismith 1971, Uzan 1985) in its
0 0/90-rep dimensionless form:
Materials
k2
⎛θ⎞
Figure 1. Hydraulic conductivities of UGMs with dif- MR k1 pa ⎜ ⎟ (3)
⎝ pa ⎠
ferent PSDs.

152
where k1 and k2 are material parameters and pa εˆ p ( N ) aN bS f S f (4)
is a reference pressure usually taken equal to the
atmospheric pressure as 100 kPa.
where εˆ p ( N ) is the accumulated permanent strain,
The MR is an important parameter for pave-
N is the total number of load cycles, a and b are regres-
ment design. A higher MR is desirable for better
sion parameters associated with the material and the
resistance against fatigue cracking of the asphalt
term Sf takes into account the effect of stress state on
concrete layer. Furthermore, a higher MR provides
permanent deformation accumulation given as:
better downward spreading of the load and thus
reduces stresses in the lower layers.
⎛ q⎞
⎜⎝ p ⎟⎠
2.2 PD characteristics Sf = a
α
(5)
⎛ p⎞
The PD in UGMs accumulates with the number ⎜⎝ p ⎟⎠
of load applications and contributes to the rut- a
ting in pavements. The PD is dependent on stress
levels, stress history, number of load cycles, degree where α is a parameter obtained from regression
of compaction, particle size distribution, moisture analysis. The advantages of this model are that a) it
content (w) and aggregate type (Lekarp, 1999). contains only three material parameters, b) unlike
Based on the shakedown theory, Dawson and Well- some other models, it does not require the shear
ner (1999) and Werkmeister et al., (2001) have iden- strength parameters, and c) it directly utilizes the
tified that the accumulation of the PD in UGMs applied stress level as a predicting variable. These
falls within the three shakedown ranges, depend- simplify its application and it is convenient to use
ing on stress levels (Figure 2). Range A occurs for for tests with a series of moisture contents. In this
relatively low stress levels when permanent strain study, this model was used for MS RLT tests by
accumulates up to a finite number of load appli- applying the time hardening formulation proposed
cations, after which the response becomes entirely by Erlingsson and Rahman (2013).
resilient with no further permanent strain. For
stress levels higher than this, Range B occurs where 3 LABORATORY TESTING
the accumulation of permanent strain continues at
a constant rate (per cycle). When stress levels are This study was based on RLT tests in the labora-
even higher, Range C behavior is observed where tory. The advantages of the RLT tests are that it is
the permanent strain accumulates at an increasing a widely used and convenient approach that reveals
rate that may eventually lead to failure. a great deal of information regarding the material
Several models are available in the literature to behavior. The RLT tests were conducted in accord-
characterize the PD behavior of UGMs. The fol- ance with the European standard EN-13286-7
lowing model proposed by Rahman and Erlings- (CEN, 2004a). The MS loading approach was pre-
son (2015) was used for this study: ferred because (a) it allows for a more comprehen-
sive study of the PD behavior for a large number
of stress conditions, (b) it includes the effect of
stress history, and (c) all of these can be attained
using a single specimen with reduced effort.
Range C The size of the cylindrical specimens was
150 mm in diameter and 300 mm in height. The
tests were carried out applying a set of different
Increased stress level stress paths according to the standard referred
to as ‘Low Stress Level’ (LSL) as presented in
Table 1. Each of the stress paths was applied for
10,000 cycles with a frequency of 10 Hz (Havers-
ine pulse) with no rest period. The total number of
load cycles applied during the tests were 300,000
Range A (30 stress paths). The tests were performed under a
free drainage condition.

Number of load cycles, N 4 MATERIALS TESTED

Figure 2. Different types of PD behaviour, depending The UGMs used for this study were assorted
on stress level. from three different origins. Material 1 was a

153
Table 1. Stress levels used for the MS RLT tests (LSL from the European standard).

Sequence 1 Sequence 2 Sequence 3 Sequence 4 Sequence 5

Confining Deviator Confining Deviator Confining Deviator Confining Deviator Confining Deviator
stress, σ3 stress, σd stress, σ3 stress, σd stress, σ3 stress, σd stress, σ3 stress, σd stress, σ3 stress, σd
kPa kPa kPa kPa kPa kPa kPa kPa kPa kPa

constant min max constant min max constant min max constant min max constant min max

20 0 20 45 0 60 70 0 80 100 0 100 150 0 100

20 0 40 45 0 90 70 0 120 100 0 150 150 0 200
20 0 60 45 0 120 70 0 160 100 0 200 150 0 300
20 0 80 45 0 150 70 0 200 100 0 250 150 0 400
20 0 100 45 0 180 70 0 240 100 0 300 150 0 500
20 0 120 45 0 210 70 0 280 100 0 350 150 0 600

Table 2. Material properties.

Test condition

Maximum Dry
Specific dry density Degree of density Void
Material wopt [%] Gravity [Mg/m3] w [%] saturation [%] [Mg/m3] ratio

Material 1 5.5 2.64 2.11 1 9.1 2.00 0.29

(crushed rock)
3.5 31.9 2.00 0.29
Material 1 OG 3.5 2.64 1.96 1 5.9 1.84 0.45
(crushed rock)
3.5 21.3 1.84 0.45
Material 2 11.9 2.22 1.74 11.9 95.8 1.74 0.30
(recycled)
Material 2 OG 4.1 2.22 1.52 4.1 19.8 1.52 0.47
(recycled)
Material 3 6.9 2.68 2.35 2 26.7 1.44 0.23
(crushed rock)
4 53.5 2.23 0.23

crushed rock aggregate commonly employed in 100

~~ Mater i al l (c rushedrock)

Swedish road construction. Two different PSDs 90

- Material I OG (crushed rocl)
of this material were studied. The first one was a :E 80
~~ Matcrial2(rccyclod)

standard PSD referred to here as ‘Material 1’ and ·~ 70 -=-eo e Matcri:JI 2 00 (recycled)

~ 60 - M ater iai J(crushedrock)
the second one a OG variant referred to here as .0
>-.
50
‘Material 1 OG’ where the fine particles below ~
40
2 mm were essentially absent from the sample. The 00

tests on these materials were conducted at identi- ·~ 30

0... 20
cal moisture content regardless of their optimum 10
moisture contents (wopt). The second UGM was a 0
recycled material referred to here as ‘Material 2’ 0.0625 0.125 0.25 0.5 16 32
Sieve size [mm]
and ‘Material 2 OG’ where, for the OG variant,
fines below 1 mm were removed from the sample.
Figure 3. Particle size distributions.
These materials were tested and compared at their
individual wopt. The third material, ‘Material 3’,
was also a crushed rock aggregate used in Swedish and was included in this study to serve as a ref-
road construction. This was tested at two different erence. The properties of these materials are pre-
moisture contents identical to those of Material 1 sented in Table 2. The PSDs are shown in Figure 3.

154
The modified Proctor method in accordance with Equation 4, calibrated using the test data. Analyses
the European standard EN 13286-2 (CEN, 2004b) of these plots show that at w = 1%, Material 1 dem-
was used to determine the maximum dry density onstrated a lower PD compared to Material 1 OG.
and wopt of the materials. The specimens were pre- On the other hand, at w = 3.5%, Material 1 showed
pared using a vibrocompactor. Some of these tests an identical PD to that of Material 1 OG. In fact,
were replicated to take into account any experi- similar to the stiffness behavior, the PD perform-
mental dispersion. ance of the OG variant was essentially unaffected
by moisture. At their individual wopt, Material 2
OG showed better resistance against PD compared
5 RESULTS AND DISCUSSION
100 0
• Materiall (w ~ l %) (a)
From the RLT tests, the MR and the accumulated 900 • Materiall (w ~ 3.5%)
o Materiall OG (w ~ 1%)
permanent strain for each specimen were calcu- . o Material l OG (w ~ 3.5%)

~
800
• Material 2 (w ~ 11.9%)
lated. Figure 4(a) presents the measured MR val-
ues as a function of θ for all the tests. Figure 4(b) i
~
700

600
• Material 2 OG (w ~ 4. 1%)
• Material 3(w~2%)
• Material 3 (w ~ 4%)
.o

."
" II
... .
"'
shows the modelled results using the k-θ model
calibrated from the RLT test data. From this fig-
:g
~
500

400
... ;~ ~ ~'t ~
I
t

ure, it is observed that the OG variant of Mate-

c::
~ 300
0 ~~ ~~ : t! J.
'='li ~ : ;l<. &.
rial 1 demonstrates slightly higher stiffness values. ~ 200

Furthermore, when w was increased from 1% to 100

~~~·
-~

3.5%, the MR values of the material with stand- o~~~~~~~~~~~~~~~~~~~~~~

0 200 400 600 800 1000 1200

ard PSD decreased slightly, whereas the MR val-
Sum of principal stresses, 8 [kPa]
ues of the OG variant were essentially unaffected.
For Material 2, the MR values of the two variants 1000 r-----~------~-----,------,------,,-----,
Materiall (w ~ t %)

were identical; however, it should be noted that the 900 - --+- Material! (w ~ 3.5%) -t------+----+-- (b)
'<;' _ - o - Material! OG (w ~ 1%)
specimen with standard PSD had a much higher 800

w (11.9% compared to 4.2% for the OG variant). ~ 700 - =-=-=~:::~::: ~~G~(~l~9~~%)

~ --~ -- Materia l 2 OG (w ~ 4 .1%)

Although Material 3 at w = 1% showed the high- ; 600 - - - Material 3 (w ~ 2%)

:; _ -- Material 3(w ~ 4%)
est MR values, it was reduced more significantly 500

compared to Material 1 when w was increased to ] 4 00 i--------+----~~~~~~~--~----~~----~

3.5%. The probable reason was that Material 3 had ] 300

more fine fractions which contributed to the lubri- ~ 200

cation effect of water, thus decreasing the MR of 100

the material. 0
0 200 400 600 800 1000 1200
The accumulated permanent strain as a function Sum of principal stresses, 8 [kPa]
of N is plotted in Figure 5 for the individual tests.
This also includes the fitted model presented in Figure 4. Resilient moduli: (a) measured, (b) modelled.

• Material ! (w = 1%) • Material ! (w = 3.5%) o Material ! OG(w = 1%) <> MateriallOG(w=3.5%) .. Material 2(w=l l.9%)
t:. Material20G(w = 4.1%) • Material3 (w = 2%) " Material3 (w = 4%) - Model
< "" 0.07 .------------------"""T'"-------------------,;------------------...-------------------.-------------------,
"' Sequence 1 Sequence 2 Sequence 3 Sequence 4 Sequence 5

0.05

0.04

0.03

0.02

0.01

60000 120000 180000 240000 300000

Nmnber ofload cycles, N

Figure 5. Accumulated permanent strain.

155
Table 3. Model parameters. the laboratory. This study was limited to 2/32 mm
(minimum/maximum particle size) UGMs where
PD model the fine fraction up to 2 mm was removed from the
samples (Figure 3). Results showed that the stiff-
Dry k-θ model (α = 0.75)
w density ness and PD characteristics of these materials were
Material [%] [Mg/m3] k1 k2 a b still comparable to the well graded reference materi-
als. Because of the lack of fines, the OG materials
Material 1 1 2.00 1520 0.62 0.00027 0.15 appeared to be less sensitive to moisture variation
(crushed 3.5 2.00 897 0.87 0.00108 0.11 for which they may exhibit more stable behavior
rock) with seasonal variation of moisture and better pro-
Material 1 1.84 1529 0.67 0.00076 0.12 tection against frost heaving. This result suggests
1 OG 3.5 1.84 1777 0.57 0.00226 0.08 the suitability of the 2/32 mm UGMs for use in
(crushed drainage layers and in permeable pavements. The
rock)
models used for this study were quite well suited to
Material 2 12 1.74 1107 0.69 0.00123 0.13
(recycled)
describe the MR and PD behavior of these materi-
Material 2 4.1 1.52 870 0.82 0.00088 0.13 als. Thus the model parameters obtained here may
OG be used for prediction of rutting, estimating bear-
(recycled) ing capacity and for layer thickness design using
Material 3 2 1.44 1947 0.57 0.00001 0.24 any suitable pavement design software. Since the
(crushed 4 2.23 1100 0.75 0.00178 0.12 study was limited to RLT tests in the laboratory,
rock) the results obtained here should be further vali-
dated with field studies and large scale tests such
as using a Heavy Vehicle Simulator (HVS). The
RLT test database should be further extended with
to Material 2, while it should be noted that the wopt
tests on more materials and replicate testing. The
of Material 2 OG was also much lower. Material
study should also be extended to other OG UGMs
3, again in terms of PD, showed the similar high
with different PSDs, for instance, using 4/32 mm
sensitivity to moisture as in the case of MR val-
materials.
ues. It showed the highest resistance against PD at
w = 1% while its resistance against PD was the low-
est when w was 4%. Again, in this case the prob- ACKNOWLEDGEMENTS
able explanation for this behavior was the relatively
higher amount of fines present in Material 3 which This work was sponsored by the Swedish Trans-
enhanced the lubrication effect of moisture. port Administration (Trafikverket). Part of the
In general, it appears that the OG UGMs did tested materials was supplied by NCC (Sweden).
not perform any worse than the well graded mate-
rials. It also seems that the OG materials were
less sensitive to moisture and as a result they per- REFERENCES
formed relatively better than the reference materi-
als at higher moisture content. For the well graded Ahmed, A., Hellman, F., and Erlingsson, S., 2016. Full
materials, the fines might have contributed to the Scale Accelerated Pavement Tests to Evaluate the
lubrication effect of water which increased the PD. Performance of Permeable and Skeletal Soil Block
The k-θ model and the PD model used in this Pavement Systems, The Roles of Accelerated Pave-
study worked quite well for both kinds of mate- ment Testing in Pavement Sustainability, Springer,
rials, as can be realized from Figures 4 and 5. pp. 131–144, ISBN: 978-3-319-42796-6 (Print) 978-3-
319-42797-3 (Online).
The calibrated parameters of these models for CEN-European Committee for Standardization, 2004a.
the materials used in this study are presented in Cyclic Load Triaxial Test for Unbound Mixtures.
Table 3. The material parameter data obtained Brussels: European Standard. EN 13286-7.
here may be used as a guideline for future studies. CEN-European Committee for Standardization, 2004b.
Furthermore, by establishing the relation of these Test Methods for the Determination of the Labora-
parameters to moisture, it can be used for predict- tory Reference Density and Water Content: Proc-
ing the stiffness and PD behavior of the materials tor Compaction. Brussels: European Standard. EN
with the seasonal variation of moisture. 13286-2.
Dawson, A.R. and Wellner, F., 1999. Plastic Behaviour of
Granular Materials. Nottingham: The University of
Nottingham. Reference PRG99014, April, 1999, Final
6 CONCLUSIONS Report ARC Project 933.
El Abd A., Hornych P., Breysse D., and Denis D., 2005.
The stiffness and PD characteristics of some OG Prediction of permanent deformations of unbound
UGMs were studied by means of MS RLT tests in pavement layers. In: Proceedings from the 7th Inter-

156
 national Conference on the Bearing Capacity of Roads Loading, Doctoral thesis, TRITA-IP FR 99–57,
and Airfields (BCRA), June, Trondheim, Norway. KTH Royal Institute of Technology, Stockholm,
Englund, J., 2011. Analyses of Resilient Behavior of Sweden.
Unbound Materials for the Purpose of Predicting Lekarp, F., Isacsson, U., and Dawson, A., 2000. State of
Permanent Deformation Behavior. Doctoral thesis. the art. i: Resilient response of unbound aggregates.
Chalmers University of Technology, Gothenburg, Journal of Transportation Engineering, 126 (1), 66–75.
Sweden. DOI:10.1061/(ASCE)0733-947(2000)126:1(66).
Erlingsson, S. and Magnusdottir, B., 2002. Dynamic Pratt, C.J., 1990. Permeable pavements for stormwa-
triaxial testing of unbound granular base course ter quality enhancement. In: Proceedings of Urban
materials. In: Proceedings of the 6th International Con- Stormwater Quality Enhancement—Source Control,
ference on Bearing Capacity of Roads and Airfields, Retrofitting and Combined Sewer Technology, edited
BCRA ’02, Lisbon, Portugal. Lisse, Netherlands: by: Torno, H.C. pp. 131–155. ASCE, New York.
Balkema, 989–1000. Rahman, M.S. and Erlingsson, S., 2015. A model for pre-
Erlingsson, S. and Rahman, M.S., 2013. Evaluation of dicting permanent deformation of unbound granular
permanent deformation characteristics of unbound materials. Road Materials and Pavement Design, 16:3,
granular materials by means of multistage repeated- 653–673, DOI: 10.1080/14680629.2015.1026382.
load triaxial tests. Transportation Research Record: Rahman, M.S., 2015. Characterising the Deformation
Journal of the Transportation Research Board, Trans- Behaviour of Unbound Granular Materials in Pave-
portation Research Board of the National Acad- ment Structures. Doctoral thesis, TRITA-TSC-PHD
emies, Washington, D.C., No. 2369, 11–19. DOI: 15-004, KTH Royal Institute of Technology, Stock-
10.3141/2369-02. holm, Sweden.
Hajek, J.J., Kazmierowski, T.J., Sturm, H., Bathurst, R.J., Seed, H.B., Chan, C.K., and Lee, C.E., 1962. Resilient
and Raymond G.P., 1992. Field Performance of Open- characteristics of subgrade soils and their relations
Graded Drainage Layers. Transport Research Record to fatigue in asphalt pavements. In: Proceedings of
1354, Transportation Research Board, Washington the International Conference on Structural Design
DC, USA, Accession Number: 00625419, pp. 55–64. of Asphalt Pavements, Ann Arbor, USA, Vol. 1,
Hicks, R.G. and Monismith, C.L., 1971. Factors influ- 611–636.
encing the resilient response of granular materials. Uzan, J., 1985. Characterisation of Granular Mate-
Highway Research Record, 345, pp. 15–31. rial. Transportation Research Record. Journal of the
Huang, Y.H., 2004. Pavement Analysis and Design. 2nd Transportation Research Board, 1022, 52–59.
ed. Prentice Hall, Inc., Englewood Cliffs, NJ. Werkmeister, S., 2003. Permanent Deformation Behav-
Kolisoja, P., 1997. Resilient Deformation Characteristics ior of Unbound Granular Materials. Doctoral thesis,
of Granular Materials. Doctoral thesis. Tampere Uni- University of Technology, Dresden, Germany.
versity of Technology, Tampere, Finland. Werkmeister, S., Dawson, A.R., and Wellner, F., 2001.
Larson, R., 1990. Swedish experience with design and Permanent Deformation Behavior of Granular Mate-
construction of pervious asphalt constructions. In: rials and the Shakedown Concept. Transport Research
Proceedings of the Standing Conference on Stormwater Record 1757, Transportation Research Board, Wash-
Source Control, edited by: Pratt, C.J. Coventry, UK: ington DC, USA, Paper No. 01-0152, pp. 75–81.
Coventry University. Vol I.
Lekarp, F., 1999. Resilient and Permanent Deformation
Behavior of Unbound Aggregates under Repeated

157
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Characterisation of permanent deformation behaviour of unbound

granular materials using repeated load triaxial testing

N. Zhalehjoo, A. Tolooiyan & R. Mackay

Geotechnical and Hydrogeological Engineering Research Group (GHERG), Federation University Australia,
Victoria, Australia

D. Bodin
Australian Road Research Board (ARRB), Victoria, Australia

ABSTRACT: Unbound Granular Material (UGM) used in the base/subbase layers of a flexible
pavement structure constitutes the vast majority of the material found in roads around the world. The
permanent deformation of a compacted UGM layer due to cyclic deviatoric loading has a significant effect
on the performance of the pavement structure. The accurate prediction of the magnitude of accumulated
permanent strain at varying load cycles and stress levels plays an important role in improving the design
and maintenance of flexible pavements. In this study, samples of two road base UGMs are tested to
evaluate the characteristics of permanent deformation using the laboratory Repeated Load Triaxial (RLT)
test. Three permanent deformation models are used to predict the magnitude of strain accumulation of
the studied UGMs. The permanent strain results predicted by the models are compared against those
measured by laboratory RLT tests to evaluate the prediction ability of each model.

1 INTRODUCTION loadings of different magnitude to a single speci-

men (Gidel et al., 2001). As UGMs experience cyclic
UGMs used in construction of the base/subbase lay- stress of variable magnitude under real conditions,
ers in flexible pavements show complex elastoplastic multi-stage RLT testing seems to offer an appropri-
behaviour when subjected to cyclic stresses caused ate approximation to reality.
by moving traffic (Lekarp et al., 2000). Rutting in Several factors affect the accumulation of per-
an unbound granular layer due to the accumulation manent deformation of UGMs. In this paper, the
of permanent deformation is one of the primary influence of moisture content and amount of fines
damage modes for low traffic flexible pavements content are considered. The RLT test specimens
(Gidel et al., 2001). The accumulation of perma- are prepared with a range of moisture contents.
nent deformation may lead to pavement deteriora- Several researchers have studied the influence
tion and eventual failure of the structure during its of moisture content on deformation behaviour
service life. Improving the design and enhancing of UGMs and reported that permanent strain
the long term performance of a pavement struc- accumulation is principally affected by moisture
ture require an accurate and reliable prediction of content (Lekarp et al., 2000, Gidel et al., 2001,
cumulative permanent strain. For this aim, there is Rahman & Erlingsson, 2012).
a need for an appropriate understanding and char- After performing the material characterisation
acterisation of the deformation behaviour of the and the RLT tests, the measured accumulation of
UGMs. RLT tests are generally performed to evalu- permanent strain is compared against predictions
ate the deformation behaviour of pavement materi- by rutting models developed by other researchers.
als under repeated loading. In this study, a series of Three permanent deformation models are applied
RLT tests are carried out to measure the accumula- to investigate the ability of each model to predict
tion of permanent deformation of two UGMs. The the permanent strain. For this aim, the permanent
Austroads test method AG-PT-T053 (Austroads, strain values predicted by the models are compared
2007) for measuring permanent deformation of against the results of the laboratory experiments to
UGMs is used, which contains three different stress identify the model that best predicts the magnitude
stages. The multi-stage RLT test applies cyclic of strain accumulation.

159
2 LABORATORY TEST PROCEDURE The specimens were prepared with a range of
moisture contents and dry densities, representing
2.1 Materials the conditions of laboratory wheel-tracking test-
ing performed on the same materials by Austral-
The laboratory tests are performed on two unbound
ian Road Research Board (ARRB) in Melbourne,
granular base materials sourced from quarries in
Australia (Aguiar & Bodin, in prep.).
Victoria, Australia. The first material is crushed
hornfels, complying with the class 2 VicRoads
specification for road bases with standard plastic- 2.3 RLT test equipment
ity, entitled as Hornfels Standard Plasticity (HSP).
For this study, a triaxial apparatus consisting of a
The second material named as Hornfels Increased
servo-pneumatic actuator with the +/- 10 kN cyclic
Plasticity (HIP) is crushed hornfels with increased
load capacity, and the air/water bladder to generate
plasticity having fines content passing the No.200
a controlled confining pressure was used to carry
(75 μm) sieve of around 13%. This amount of fines
out the RLT tests. The triaxial system is instru-
content exceeds the allowable fines content for class
mented with a set of four axial gauges; the on-
1 and 2 road base products according to the Vic
specimen axial Hall-Effect transducer, an internal
Roads specification. The plasticity index of HSP
Linear Variable Differential Transformer (LVDT),
material is equal to 2. However, the HIP material
an external LVDT, and the actuator LVDT. The
has the plasticity index of 8 due to its higher fine
aim was to compare the specimen’s axial defor-
fraction. Also, the optimum moisture content of
mation measured by each of the four gauges. The
HSP and HIP materials are equal to 5.6% and 6.6%
internal LVDT measures the deformation of the
respectively (Aguiar & Bodin, in prep.).
full length of specimen inside the triaxial cell and
The particle size distribution curves of these
the external LVDT is fixed to the loading piston
materials are presented in Figure 1.
outside the triaxial cell. The axial Hall-Effect trans-
ducer is mounted to the specimen and measures
2.2 Sample preparation the deformation on the middle third of the sample.
In addition, radial deformation is measured during
The materials were compacted in six layers in a
the test by radial Hall-Effect transducer mounted
steel mould at specific moisture contents and den-
at mid-height of the specimen. Analysis of the
sities using a vibratory compaction method to pro-
radial deformation measurements is not addressed
duce cylindrical specimens of 100 mm in diameter
in this paper. Figure 2 shows the RLT test speci-
and 200 mm in height. The compaction was per-
men and triaxial cell placed in the loading frame.
formed with an electric vibratory hammer accord-
In this paper, the axial deformation results are
ing to the AASHTO standard T307-99 (AASHTO,
reported based on the internal LVDT measurements.
2007). After compaction, each specimen was
extruded using a hydraulic extruder from the ini-
tial steel mould into a plastic split mould with a
rubber membrane placed inside it. When extruded,
the specimen was ready to be mounted on the base
of the triaxial cell with porous stones placed at the
top and bottom ends of the sample.

100

~
80 -
-<>- HSP Mat erial l
1>0
c:
-•- HIP Material
I
....."'
·v; 60

-
.s: 40
1>0
"Qi
5: 20
l'l' =""~
_v/
0
0.01 0.1 1 10 100
Sieve Size (mm) Figure 2. Triaxial cell placed in the loading frame, and
triaxial test specimen instrumented with transducers for
Figure 1. Particle size distribution of HSP and HIP measuring the axial strain (internal LVDT, Hall-Effect
materials (after Aguiar & Bodin, in prep.). transducer, and external LDVT).

160
Table 1. Stress stages applied for permanent deforma- The magnitude of accumulated permanent
tion test. deformation of UGMs is affected by several fac-
tors such as applied stress level, number of load
Stress stage Confining stress Deviatoric stress applications, stress history, moisture content, den-
no. kPa kPa
sity, particle size distribution, and aggregate type
1 50 350 (Lekarp et al., 2000). In this paper, the effect of
2 50 450 moisture content and amount of fines content on
3 50 550 the permanent deformation of two UGMs are
evaluated.

The comparison of the on-specimen Hall-effect

3.1 The shakedown concept
transducer, internal LVDT and external LVDTs
measurements are presented elsewhere (Zhalehjoo The shakedown concept describes two types of
et al., 2016). material response to loading. In the first type, the
material behaviour will be entirely resilient after a
finite number of load applications but in the sec-
2.4 RLT test procedure
ond, permanent deformation does not stabilise and
Resilient modulus testing was conducted on the may lead to failure with excessive accumulated per-
specimens according to the AASHTO T307-99 manent deformation. Based on these two types of
testing procedure (AASHTO, 2007). This proce- behaviour, the term “shakedown limit” describes
dure consists of a preconditioning sequence with the critical stress level that separates the stable and
up to 1000 load cycles, and 15 loading sequences unstable response of pavement materials to load-
with 100 load cycles each. Subsequently, per- ing (Werkmeister, 2003). The use of shakedown
manent deformation testing was performed on theory in pavement design was first introduced by
the same specimen following the Austroads test Sharp & Booker (1984).
method AG-PT-T053 (Austroads, 2007). The pro- Recently, Werkmeister et al. (2001), Werkmeister
cedure for permanent deformation testing consists (2003), and Arnold (2004) conducted studies to
of three consecutive stress stages with a constant investigate the application of the shakedown con-
confining pressure as shown in Table 1. Each stress cept in pavement design by carrying out extensive
stage is applied for 10,000 loading cycles making RLT tests.
the full permanent deformation test 30,000 loading Dawson & Wellner (1999) and Werkmeister
cycles in total. A haversine-shaped load pulse with et al. (2001) proposed three different categories
a frequency of 1 Hz with 0.9 second rest period in for the permanent strain accumulation with the
each loading cycle was applied to the specimens. number of load cycles by analysing the permanent
Each material was tested at three different mois- strain rate development of a series of RLT test
ture contents with constant dry density to study the results. The assigned three ranges (A, B, and C),
effect of moisture on the performance of UGMs. which describe the stress levels at which the type of
Considering the experimental variability in RLT permanent strain response changes, are as follows:
tests on UGMs, the tests were repeated two to four
times for each moisture content and density condi- i. Range A, plastic shakedown range: for the
tion to ensure the results were not arbitrary. first few thousand load cycles, the deforma-
tion response to loading is plastic, but after
completion of the post-compaction, the per-
3 PERMANENT DEFORMATION manent strain rate progressively diminishes
BEHAVIOUR resulting in no further plastic strain and the
material response becomes purely resilient.
With each load application, a small amount of This behaviour is therefore permitted in the
permanent deformation occurs in unbound granu- pavement providing sufficiently small amount
lar layers. As the permanent deformation in base/ of accumulated permanent strain.
subbase layer causes irreversible deformation at the ii. Range B, intermediate response (plastic creep):
pavement surface, the accumulation of permanent for the first few thousand load cycles, the per-
deformation with the number of load cycles may manent strain rate is high, but after these the
lead to excessive rutting and pavement deteriora- rate decreases to a low level. Then the rate
tion during its service life. reaches a nearly constant level in a number
To avoid the eventual failure of a pavement, it of load cycles depending on the material
needs to be designed in such a way that no, or only and the stress level. Kolisoja (1998) reported
small, permanent deformations accumulate in the that although the material behaviour for the
pavement layers (Werkmeister, 2003). first 100,000 cycles seems to stabilise, it can

161
 progress to failure if the same load level con- parameters (ε0, ρ, β ) that are obtained by curve
tinues cycling for a long enough time. Werk- fitting of permanent strain against the number of
meister (2003) found similar results at 850,000 load applications (N) from RLT tests. The mate-
load cycles when the sample approaches an rial parameters are different for each specimen
incremental collapse. Range B behaviour may depending on the material type, its physical proper-
therefore be acceptable for pavements for a ties such as moisture content and density, and test
finite number of load cycles. conditions such as temperature, and stress level.
iii. Range C, incremental collapse: as a conse- The second permanent deformation model con-
quence of a much higher cyclic stress level sidered in this paper is presented by Gidel et al.
compared to range B, there is a progressive (2001) as shown in Equation 5.
increment in permanent strain with each load
cycle and the response is always plastic. If ⎡ ⎛ N ⎞ − B ⎤ ⎛ lmax ⎞ n
ε p ( N ) ε 0 ⎢1 − ⎥
this behaviour occurs in an unbound granular ⎣ ⎝ 100 ⎠ ⎦ ⎜⎝ pa ⎟⎠
layer, it would cause failure of the pavement.
⎛ 1 ⎞ (5)
As a result, in a well-designed pavement, range ×
⎜ s qmax ⎟
C is not allowed and UGMs should not be ⎜m+ − ⎟
permitted to experience this behaviour. ⎝ pmax pmax ⎠
Werkmeister (2003) proposed the following
criteria (Equations 1–3) to determine shakedown This model, which is developed based on multi-
range limits based on RLT test results. stage RLT tests on UGMs, links the accumulated
permanent strain (εp) to both the number of load
Range A : (ε p ( ) − εp( )) .045 × 10 −3 applications (N) and stress levels explicitly. ε0, B,
(1) and n are material parameters. Parameters pmax and
Range B : 0.045 × 10 3 < ( p (5000 ) qmax are the maximum applied mean and deviator
(2) stress, respectively and lmax= √(pmax)^2+(qmax)^2.
− εp( )) .4 × 10 −3
Parameters m and s are obtained from static tri-
RangeC : (ε p ( ) − εp( )) .4 × 10 −3 (3) axial tests and are the slope and the intercept of
the Mohr-Coulomb failure line in p-q space. Pa is a
where, εp (5000) and εp (3000) are accumulated ver- reference stress and equals 100 kPa.
tical permanent strain at the 3000th and 5000th The third permanent deformation model evalu-
load cycles, respectively. ated in this study is developed by Korkiala-Tanttu
(2008) as given in Equation 6.
3.2 Permanent deformation models for UGMs R
ε p (N ) C
CN b (6)
Several permanent deformation models have been A R
developed for UGMs to represent the relationship
between permanent strain and applied stresses, or where C and b are material parameters. The param-
permanent strain and number of load applications, eter R is the deviatoric stress ratio and is obtained
or the combination of both. Although the effect of by R = q/qf, where q is the deviatoric stress of the
moisture content, density, and material type are not load application and qf is deviatoric stress in failure
directly considered in the current models, there is a and is defined by qf = m.p + s. The parameters m,
constant factor(s) existing in these models which can p, and s are expressed in the previous model. A is
be determined through a regression analysis to take a material independent parameter and is the maxi-
the effect of these material properties into account. mum value of the deviatoric stress ratio R, which
In this paper, three recent models for evaluation theoretically is 1 but Korkiala-Tanttu suggested to
of permanent deformation behaviour of UGMs consider A equals to 1.05 for the pavement response
are employed to estimate the permanent deforma- calculation (stress state). This model combines the
tion and compare them with the measured defor- cyclic loading function with the permanent shear
mations in the laboratory. strain component to calculate the axial permanent
The first permanent deformation model used in strain. The value of the parameter C depends on the
this study is developed by Tseng & Lytton (1989) material, its degree of compaction, and moisture
as given in Equation 4. content. The value of parameter b mostly depends
on the stress state and failure ratio. Parameter b is
β (4) also dependent on the degree of compaction and
ε p (N ) ε 0 e− ( ρ / N )
moisture content. This value represents the damp-
ing shape of the permanent deformation curve.
This model predicts the accumulated permanent To obtain the slope (m), and the intercept (s)
strain (εp) by taking into account three material of the Mohr-Coulomb failure line in p-q space

162
to be utilised in the permanent deformation small, it gradually accumulates with increasing
models, a static failure triaxial test is carried out number of load cycles. Also, the cumulative per-
immediately after completion of the permanent manent strain is strongly affected by the stress
deformation test procedure. For performing this level in such a way that by increasing the devia-
test, an axial strain at a rate of 1 percent per toric stress from 350 kPa to 450 kPa and then to
minute is applied under a strain-controlled load- 550 kPa (in a constant confining pressure), there
ing procedure. Three identical test specimens are is a rapid increase in the permanent strain rate
required to be subjected to a static shear test with especially in the specimens with higher moisture
three different confining pressures to obtain the contents.
static parameters. In this paper, the materials In addition, moisture content is one of the
static parameters measured by Australian Road most significant factors affecting the long-term
Research Board (ARRB) (Aguiar & Bodin, in performance of UGMs. Figure 3 shows that in
prep.) are used. both materials, permanent strain increases with
increasing moisture content (ω). The significance
of moisture content on the accumulation of per-
3.3 Factors affecting the permanent strain
manent strain is reported by several researchers.
response
Thom & Brown (1987) and Korkiala-Tanttu et al.
Development of axial permanent strain with the (2003) reported that even relatively small changes
number of load cycles for HSP and HIP materials in moisture content may influence the permanent
having a range of moisture contents is illustrated deformation dramatically. This is due to the lubri-
in Figure 3. The permanent strain was measured cation effect of water and loss of internal friction
for the first 10 loading cycles and every 20 cycles of the materials (Lekarp, 1999).
after that making 510 readings in total for each The results demonstrate that the HIP mate-
stress stage. The results show that although the rial with higher fines content and plasticity has a
permanent strain in each loading repetition is higher resistance to permanent deformation under
repetitive loading compared to HSP material. At
the end of stress stage 3, HSP material with the
2.6
2.4 moisture content of 3.8% has experienced an accu-
2.2
Stress Stage 1 Stress Stage 2 Stress Stage 3 ,
mulative permanent strain of 1.1%. However, the
;< 2.0
71.8
cumulative permanent strain at the end of stress
~ 1.6 stage 3 for HIP material with the same moisture
content and dry density is equal to 0.26%.
~ ~:~ Generally, it is believed that the materials stiff-
~ 1.0 - w= 3.5%
D. 0.8 -- w=3.8% ness and the resistance against permanent defor-
0.6
0.4
- · w=4.3% mation decrease with increase in the amount
0.2 of fines content. However, Yideti et al. (2013)
0.0 introduced a new parameter entitled Disrup-
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ § ~ tion Potential (DP) for evaluating the permanent
Numbe r of load Cycle s deformation behaviour of unbound granular
(a) materials. The DP is defined as the ratio of the
volume of potentially disruptive fine material
2.6
2.4 ,. over the free (available) volume within the pri-
2.2 Stress Stage 1 Stress Stage 2 Stress Stage 3 / mary structure material. They reported that
'ii 2.0 #.,-"' UGMs having a DP ranging from 0.5 to 0.9
~ 1.8
·~ 1.6 I exhibit the best performance in terms of resist-
Vi 1.4
I ance to permanent deformation. This finding
~ 1.2
indicates that a granular material that contains
:; 1.0 - w = 3.8%
/"
Eo.s -- w= 4.3% some or a limited amount of fines content is
~ 0.6
0.4
- · w=5.3%
....... ......... _..... ../ / desirable to effectively resist the deformation
0.2 -·- and distribute loads but an extensive amount
0.0 of fines content reduces the resistance against
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ permanent deformation. This is due to the fact
Number of Load Cycles that in the case of existence of high percentage
(b) of fines content, the interconnection between
the large particles decreases resulting in reduc-
Figure 3. Measured axial permanent strain with tion of load-carrying capacity and stiffness of
number of load applications; (a) for HSP material, (b) materials. The results indicate that fines content
for HIP material. passing the No.200 sieve of 13% in HIP material

163
provided stability of coarse particles producing 0.14

the higher resistance to permanent deformation 0.12

under repeated loading. g 0.10

·~ 0.08
.:;
~ 0.06
3.4 Shakedown range calculation -Measured Data (w = 3.5%)
~ 0.04 - ·Tseng & Lytton Model
Tables 2 and 3 present the shakedown ranges for ~ 0.02 ••• Gidel et al. Model
- Korkiala-Tanttu Model
HSP and HIP materials for different moisture con- 0.00
tents in the three stress stages. The ranges are cal- § ~ ~ ~ ~ § ~ ~
culated using the criteria presented in section 3.1 Number of Load Cycles

by analysing the RLT test results. (a)

It is clear that by increasing the deviatoric stress
0.14
and moisture content, the ranges shift from A to
0.12
B or from B to C. It is important to consider the
shakedown ranges for each specific material as one
of the key objectives of pavement design is that
the unbound granular layers must be able to resist
g 0.10
-~ 0.08
.:;
~ 0.06
r.::
... . . . . · · · · · · · -Measured Data (w = 3.8%)
against permanent deformation beyond a certain ~ 0.04 - ·Tseng & Lytton Model
• •• Gidel et al. Model
level. (Werkmeister, 2003). ~ 0.02
- Korkiala-Tanttu Model
0.00

3.5 Permanent strain prediction by the models

~ ~ § ~ ~
Number of Load Cycles
a~
(b)
Figures 4 and 5 compare the magnitudes of per-
manent strain predicted by the three rutting mod-
els against those measured from the laboratory

Table 2. Shakedown ranges for HSP material.

- Measured Data (w = 4.3%)
Moisture Stress Confining Deviatoric - ·Tseng & Lytton Model
••• Gidel et al. Model
content stage stress stress Shakedown - Korkiala-Tanttu Model

% no. kPa kPa range

~ ~ ~ ~ ~ ~
8
§
Number of load Cycles
1 50 350 A (c)
3.5 2 50 450 B
3 50 550 B Figure 4. Measured and predicted axial permanent
1 50 350 A strain with number of load applications for HSP mate-
3.8 2 50 450 B rial; (a) ω = 3.5%, (b) ω = 3.8%, (c) ω = 4.3%.
3 50 550 C
1 50 350 B
4.3 2 50 450 C RLT tests. The material parameters for each model
3 50 550 C are obtained using Minitab statistical software
(Minitab, 2010) based on least square fitted esti-
mates of the non-linear model parameters. The
Table 3. Shakedown ranges for HIP material. parameters for the Tseng and Lytton model, Gidel
et al Model, and Korkiala-Tanttu model are pre-
Moisture Stress Confining Deviatoric sented in Tables 4 and 5. In addition the Root Mean
content stage stress stress Shakedown
% no. kPa kPa range
Square Error (RMSE) is included in the tables to
evaluate the prediction accuracy and to show how
1 50 350 A well a model fits the laboratory measured values.
3.8 2 50 450 A RMSE presents the differences between the pre-
3 50 550 B dicted values by each model and the laboratory
1 50 350 A measured values. Based on this definition, a lower
4.3 2 50 450 B RMSE indicates a better prediction.
3 50 550 C Permanent strains are only predicted for the
1 50 350 B
first stress stages. Further studies are required to
2 50 450 C
include the effect of different stress levels on the
5.3
3 50 550 C accumulation of permanent deformation. Time
hardening approach may be utilised in further

164
 0.14 Table 4. Material parameters for each model for HSP
0.12 material.
~ 0.10 ········ MC M1* M2** M3***
·~ 0.08
:;;
~ 0.06
IU · ·· · · · · % ε0 ρ β RMSE ε0 B n RMSE C B RMSE
- Measured Data (w = 3.8%)
e 0.04

3.20E-7 3.77E-7 7.47E-7

- ·Tseng & Lytton Model

0.1507
45.508
0.1872

0.0019

64.302 59.374 41.356

4.8721 4.7494 4.4440

0.0231

0.0124
0.1084

0.0032
~ ••• Gidel et al. Model

3.5
0.02 - Korkiala-Tanttu Model
0.00
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0

0.02452
~
0 0 0 0 0 0 0

0.2001
188.20
0.1376

0.0017

0.0130
0.1109

0.0026
N ,..., '<T U") 0.0 ,....
Number of load Cycles

3.8
(a)
0.14

0.12

0.2850
3181.8
0.2055

0.0041

0.0187

0.0052
0.2269

0.0019
4.3
~ 010

·~ 0.08
:;;
~ 0.06
o ··· · · - Measured Data (w =4.3%)
*M1: Tseng & Lytton model, **M2: Gidel et al. model,
***M3: Korkiala-Tanttu model.
e 0.04
- ·Tseng & lytton Model
All the parameters are obtained based on the permanent
~ 0.02
••• Gidel et al. Model
- Korkiala-Tanttu Model strain in percent.
0.00

§ ~ ~ ~ ~ § ~ ~ ~ ~
Number of load Cycles Table 5. Material parameters for each model for HIP
(b) material.
0.14

0.12 MC M1* M2** M3***

~ 0.10
0
'i! 0.08
% ε0 ρ β RMSE ε0 B n RMSE C B RMSE
:;;
3.68E-7 7.91E-7 6.45E-7
~ 0.06
0.1435
61.096
0.1829

0.0012

60.570 37.400 41.538

4.8116 4.4742 4.4868

0.0209

0.0110
0.1125

0.0026
- M easured Data {w =5.3%)
3.8

e 0.04 --Tseng & lytton Model

~ 0.02
•• • Gidel et al. Model
- Korkiala-Tanttu Model
0.2046
900.76
0.1210

0.0015

0.0204

0.0105
0.1194

0.0022
0.00
4.3

0 0000000
§ ~~~~§~~
Number of load Cycles
(c)
0.1891
138.93
0.1977

0.0024

0.0246

0.0115
0.1347

0.0020
5.3

Figure 5. Measured and predicted axial permanent

strain with number of load applications for HIP mate-
rial; (a) ω = 3.8%, (b) ω = 4.3%, (c) ω = 5.3%. *M1: Tseng & Lytton model, **M2: Gidel et al. model,
***M3: Korkiala-Tanttu model.
All the parameters are obtained based on the permanent
works to consider the influence of the stress his- strain in percent.
tory on the materials deformation behaviour.
(Lytton et al., 1993, Erlingsson & Rahman, 2013).
According to the comparison between the meas- 4 CONCLUSIONS
ured and predicted cumulative permanent strain,
based on the visual examination of the curves This study evaluated the suitability of three per-
and the RMSE values, it is found that the predic- manent deformation models developed by Tseng
tions by Korkiala-Tanttu and Tseng and Lytton and Lytton, Gidel et al. and Korkiala-Tanttu for
models show a good agreement with the experi- predicting the accumulated permanent strain due
mental results. However, comparing the RMSE to the cyclic loading on two UGMs. For this aim,
values indicates that Tseng and Lytton model per- RLT testing was performed on HSP and HIP
forms better and tends to predict the permanent materials sourced from quarries in Victoria, Aus-
deformation behaviour more precisely by hav- tralia to investigate the permanent deformation
ing generally lower RMSE values compared to responses. The materials were tested under a range
Korkiala-Tanttu model. Gidel et al. model, on the of moisture content with constant dry density fol-
other hand, shows low prediction accuracy and the lowing the Austroad test method AG-PT-T053.
predicted permanent strain do not match with the The laboratory measured permanent strain values
measured ones. are compared with the predicted values obtained

165
by the three models. It is found that both Tseng Korkiala-Tanttu, L. 2008. Calculation method for per-
and Lytton and Korkiala-Tanttu models are able manent deformation of unbound pavement materials.
to predict the accumulated permanent deformation VTT Publications 702, Espoo, Finland.
accurately for the tested materials. However, the Korkiala-Tanttu, L., Laaksonen, R. & Törnqvist, J. 2003.
Effect of spring and overload on the rutting of a low-
Tseng and Lytton model prediction results show volume road. HVS-Nordic—research. Finnish Road
better fit to the measured values by having, gener- Administration, Finnra Reports 22/2003.
ally, a lower RMSE. The Gidel et al. model does Lekarp, F. 1999. Resilient and permanent deformation
not perform very well for the tested materials. behavior of unbound aggregates under repeated load-
The effect of moisture content on the accumula- ing. Doctoral thesis, Royal Institute of Technology
tion of permanent strain for different stress levels (KTH), Stockholm, Sweden.
under repeated loading is investigated. It is shown Lekarp, F., Isacsson, U. & Dawson, A. 2000. State of the
that the moisture content has a dominating influ- Art. II: Permanent strain response of unbound aggre-
ence on the deformation behaviour of unbound gates. Journal of Transportation Engineering, ASCE
126, 76–83.
pavement layers since by increasing the moisture Lytton, R.L., Uzan, J., Fernando, E.G., Roque, R.,
content, the permanent strain increases markedly. Hiltumen, D. & Stoffels, S.M. 1993. Development
The results of measured permanent strain also and validation of performance prediction models and
indicate that HIP material with increased plas- specifications for asphalt binders and paving mixes.
ticity having fines content of around 13% shows The Strategic Highway Research Program Project,
increased resistance to permanent deformation Rep. No. SHRP-A-357.
under repetitive loading, which may improve the Minitab 2010. Minitab 16 Statistical Software (2010).
long term performance of the materials compared [Computer software]. State College, PA: Minitab, Inc.
to HSP material with the standard plasticity. www.minitab.com.
Rahman, M.S. & Erlingsson, S. Moisture sensitivity of
The study presented in this paper is part of unbound granular materials. Proceedings of the 4th
an ongoing project to consider a wider range of European Pavement and Asset Management Confer-
UGMs with different moisture contents. ence (EPAM4), September 2012 2012 Malmö, Sweden.
Sharp, R.W. & Booker, J.R. 1984. Shakedown of Pave-
ments under Moving Surface Loads. ASCE Journal of
REFERENCES Transportation Engineering, 110, 1–14.
Thom, N.H. & Brown, S.F. 1987. Effect of moisture on
AASHTO 2007. Standard method of test for determin- the structural performance of a crushed-limestone
ing the resilient modulus of soils and aggregate mate- road base. Transport Research Record 1121, Trans-
rials. Washington, D.C: American Association of portation Research Board, Washington DC, USA.,
State Highway and Transportation Officials. 50–56.
Aguiar, L. & Bodin, D. in prep. Characterisation and Tseng, K.H. & Lytton, R.L. 1989. Prediction of per-
performance evaluation of granular bases: laboratory manent deformation in flexible pavement materials.
data report. in preparation: Austroads Project No: Implication of aggregates in design, construction,
TT1819. and performance of flexible pavements. ASTM STP
Arnold, G.K. 2004. Rutting of granular pavements. Doc- 1016, H.G. Schrauders, and C.R. Marek, eds. Ameri-
toral thesis, University of Nottingham, England. can Society for Testing and Materials, Philadelphia,
Austroads 2007. Determination of permanent deforma- 154–172.
tion and resilient modulus characteristics of unbound Werkmeister, S. 2003. Permanent deformation behaviour
granular materials under drained conditions. AG:PT/ of unbound granular materials in pavement construc-
T053. Austroads, Sydney. tions. Dresden University of Technology.
Dawson, A. & Wellner, F. 1999. Plastic behaviour of Werkmeister, S., Dawson, A. & Wellner, F. 2001. Per-
granular materials. Final Report ARC Project 933. manent deformation behaviour of unbound granular
Department of Civil Engineering, University of Not- materials and the shakedown-theory. Transportation
tingham, Reference PRG990. Research Record No. 1757, Transportation Research
Erlingsson, S. & Rahman, M.S. 2013. Evaluation of per- Board, Washington, D.C., 75–81.
manent deformation characteristics of unbound gran- Yideti, T.F., Birgisson, B., Jelagin, D. & Guarin, A. 2013.
ular materials by means of multistage repeated-load Packing-based theory framework to evaluate perma-
triaxial tests.. Transportation Research Record: Journal nent deformation of unbound granular materials.
of the Transportation Research Board, 2369, 11–19. International Journal of Pavement Engineering, 14,
Gidel, G., Breysse, D., Hornych, P., Chauvin, J.-J. & Denis, 309–320.
A. 2001. A new approach for investigating the perma- Zhalehjoo, N., Tolooiyan, A., Mackay, R. & Bodin, D.
nent deformation behaviour of unbound granular 2016. Determination of resilient modulus for unbound
material using the repeated load triaxial apparatus. Bul- granular materials using on-specimen Hall-effect
letin des laboratoire des Ponts et Chaussées, 233, 5–21. transducers. International Conference on Geotechnical
Kolisoja, P. 1998. Large scale dynamic triaxial tests III. Engineering and Soil Mechanics (5thICGESM)- Full
Tampere University of Technology. paper accepted.

166
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Implementation of shakedown and packing theories for unbound

granular materials

Tatek F. Yideti
Norwegian Public Roads Administration, Region East, Lillehammer, Norway

Andrew Dawson
Department of Civil Engineering, The University of Nottingham, University Park, Nottingham, UK

Bjorn Birgisson
Center for Infrastructure Renewal, Texas A&M University, Texas, USA

ABSTRACT: The shakedown theory is a tool to characterize the state of the structural system of an
unbound granular material that is subjected to repeated loading when its response has become resilient
in nature and no further accumulation of plastic strain occurs. This paper demonstrates the application
of both shakedown and packing theories to characterize permanent deformation behaviour of unbound
aggregate materials. Ranges of shakedown phenomenon (Ranges A, B and C) for some selected unbound
materials have first been investigated based on their permanent strain responses. The Disruption Potential
(DP) of the Primary Structure (PS) of the aggregate assemblage has then been determined for each grain
size distribution based on a packing theory model. The DP was found to characterize satisfactorily the
three types of shakedown ranges. Further, a finite element analysis was performed to determine the shake-
down limit load as a mean stress and a dimensionless shakedown load. The analysis was simulated using
the Mohr-coulomb yield criteria for a three-dimensional cylindrical sample. From this study, the DP value
was found to compare favorably with the predicted mean stress and dimensionless shakedown load.

1 INTRODUCTION Brown (1996a, 2004b) conducted Repeated Load

Triaxial Testing (RLTT) in order to demonstrate the
In pavement structures, the development of per- existence of a threshold deviator stress level, below
manent deformation over time is of fundamental which the magnitude of accumulated permanent
importance to determine its serviceability over the strain is quite small. Similarly, the concept of shake-
intended design period. The permanent deforma- down was applied by Sharp & Booker (1984) that
tion response of the pavement structure is strongly a material shakes down after a certain number of
associated with the properties of Unbound Granu- load repetitions into an entirely resilient response
lar Material (UGM) layers. UGMs are widely used and there is no further accumulation of permanent
in base and sub base layers of a pavement structure strain. The observed permanent strain has, further,
in many countries. These layers exhibit two types been categorized into different shakedown ranges
of deformation when subject to repeated loading: using RLTT, namely Ranges A, B and C (Dawson &
resilient and permanent deformation. Accelerated Wellner 1999, Werkmeister et al. 2001a).
pavement tests show that the unbound layers con- During the last four decades a number of differ-
tribute 30% to 70% of the total surface rutting of ent types of modeling approach have been devel-
a pavement structure (AASHTO 1962, Larsen & oped to describe the accumulation of permanent
Ullidtz 1997, Korkiala et al. 2003). The ability to deformation in granular material using analytical
determine the magnitude of deformation in each theory (Werkmeister et al. 2002b, Peres et al. 2006),
layer of a road structure is, therefore, of interest plastic theory based models (Chazallon et al.
and each deformation needs to be related to dif- 2006b) and shakedown theory (Collins et al. 1993).
ferent properties of the each material. In a road The permanent deformation response of granular
structure, in order to avoid premature failure, the materials is highly dependent on the stress level.
accumulation of permanent deformations should The shakedown concept was proposed on the
eventually cease, resulting in a stable (resilient) basis that accumulation of permanent deformation
response during the service life of the pavement of unbound materials gradually stabilizes when the
(Lekarp & Dawson 1998). applied stress remains low for a large number of

167
loading cycles. At the other extreme, at high stress shakedown limit for evaluating the permanent strain
levels, the materials exhibit a gradual failure with condition in two-dimensional (Shiau & Yu 2000,
rapid increase of accumulated permanent strain Nguyen et al. 2008) and in three-dimensional (Yu &
(Barksdale 1972, Chan 1990, Lekarp & Dawson Wang 2012) pavement analyses. A few researchers
1997). Therefore, this implies that there is a criti- (Collins & Cliff 1987) further studied the kine-
cal load condition that distinguishes the stable matic upper-bound shakedown theorem for a two
condition from an unstable condition. This stress dimensional pavement model and Boulbibane et al.
level was named by a few researchers as ‘endur- (2005a, b) investigated this for three-dimensional
ance limit’ (Lerew & Leonards 1962). Sangrey et al. cases under repeated moving surface loadings.
(1969) defined it as ‘critical level of repeated stress’, Failure of layered pavement structures by subsurface
and Heath et al. (1972) and Loach (1987) named it slip in the direction of travel (Collins et al. 1993a, b)
‘threshold level’. However, Trollope et al. (1962) and and preliminary results, perpendicular to the line of
Werkmeister et al. (2001a) used the most common travel (Collins & Boulbibane 1997 1998b), were also
term, ‘shakedown limit’, found in the literature. presented by researchers.
Methods to model and predict this kind of In unbound granular materials, the response to
response, so far, do not incorporate the physical repeated load applications can be categorized into
properties of the materials into their models—they three major types according to the level of loading.
simply describe observed behavior. Thus, there First, when the applied load is lower than the elastic
is a clear need to develop theoretical models that limit of the structure, the response is purely elastic.
include the physical properties such as gradation However, when the load is larger than the first yield
and packing arrangement. limit so that some part of the structure is deforming
In that context, the purpose of the present paper plastically, but it is less than a critical load limit, and
is to incorporate the packing theory parameters into then no further accumulation of permanent strains
the determination of the shakedown range bound- occur. In this case the material is said to be in the
ary parameters (Range A, Range B and Range C) state of shakedown. Finally, if the applied load is
and perform a finite element analysis to evaluate greater than the shakedown limit load, the material
the critical shakedown load parameter. The pack- reaches the state of incremental collapse.
ing theory-based model was previously developed Dawson and Wellner (1999) conducted a
by the present authors to identify the load carrying study categorizing the different permanent strain
aggregate particles in UGMs and eventually used response behaviour in terms of shakedown limit
to evaluate the material’s permanent deformation ranges. As shown in Figure 1, the expected rutting
behaviour and resilient modulus (Yideti 2013a, b). responses of the pavement are categorized into
In the framework developed, Disruption Potential 3 Ranges Range A (plastic shakedown), Range B
(DP), porosity, coordination number and void ratio (plastic creep) or Range C (incremental collapse).
of the skeleton are the key packing parameters. The response properties of the three ranges are
In order to investigate this, the responses of now described in Figure 1.
five Finnish crushed rock aggregates and two Range A is the plastic shakedown range and
Norwegian crushed gravel and gneiss rock have the response shows a high permanent strain rate
been abstracted from the test results reported by (permanent strain per load cycle) for a finite
Kolisoja (1995, 1998). Additionally, three different
materials corresponding to three different mor-
phologies are used for the simulation using finite
element analysis. RangeC
The shakedown ranges and load parameters
have been described as a function of Disruption
Potential (DP). All experimental data were favour- ..··//// Increasing Stress Ratio
ably compared with the shakedown and packing .•...
(or reducing material strength)
theories’ parameters. "; RangeB
l
I

2 SHAKEDOWN OF UNBOUND
GRANULAR MATERIALS I
I

I
I
;
The application of the shakedown concept to pave-
ment design was first introduced by Sharp and N umber ofload repetitions
Booker (1984). They introduced a method of con-
ics using the static shakedown theorem. After this Figure 1. Shakedown range behaviors for permanent
work, several researchers proposed a lower-bound strain versus cumulative loading (Dawson & Wellner 1999).

168
number of load applications during the initial com- than those in the PS, which resides in-between the
paction period. After the initial compaction period PS particles to provide stability. A summary of the
the permanent strain rate per load cycle decreases process to determine the PS is given in Figure 2.
until the response becomes entirely resilient and
no further permanent strain occurs. This range
3.2 Disruption potential
occurs mostly at low stress levels. In the design of
pavement thickness, the applied load on top of the The Disruption Potential (DP) is defined as the
base and sub base layers is recommended to be in ratio of the volume of potentially disruptive fine
the stress levels in which Range A will result. This material over the free volume within the primary
range limit can be used as a tool to determine the structure (Yideti et al. 2013a). As can be seen in
stability of the layers. Figure 3, DP values provide a measure of the
Range B is the plastic creep response at which stability of unbound granular materials.
the response behaviour is initially like Range A
during the compaction period. After this time, the
permanent strain rate becomes decreasing, con-
stant or slightly increasing. This range will prob- Select consecutive sieve tizes D 1 and D 2
ably be acceptable for low volume road pavements
where the total number of vehicles carried is small.
In this range, a large resilient deformation can be
seen and particles exhibit abrasion as the number Calculate weighted average
p:n1icle diameter
Calcula te w eighted average
void dirunetet' (part:ide
of load cycles increases. The particle abrasion is diameter w ithin PS)

thought to decrease the angle of internal friction

by polishing the particle contact points thus low-
ering the coefficient of friction between particles
and accelerating the accumulation of plastic defor-
mation without increasing the applied stress which Ch eck interaction to det ermine PS
t. t •D1D 1 d 1. 1•D1D 2
might, in time, change the response to Range C
~D;+2.36*D; s VP'I s lJ236*D;+~3
(incremental collapse). In general, the strains
developed never become entirely resilient.
Range C is incremental collapse where an initial
compaction period may be observed and after this
time, the permanent strain rate remains constant or
increases with further load cycles. It cannot be per-
mitted to occur in an unbound granular layer of a
pavement due to the uncontrolled plastic deforma-
tion that prevents the pavement layer from support-
ing traffic. It may be associated with high abrasion
and particle crushing (Werkmeister et al. 2004). Select next D 1 (= previou!i D 1 ) , select next D 1 =next smaller !iieve size.
Rcpcot procedure with new D 1 nnd D 1 until D 1 = 2mm or 2.36mm

3 PACKING THEORY-BASED
FRAMEWORK Figure 2. Primary structure identification (Yideti et al.
2013b).
3.1 Primary and secondary structures
A packing theory-based theoretical and analytical Contacts lost between PS
Contact points
model of unbound granular materials was previ- between PS particles
particles due to excessive
volume of SS particles
ously developed by the present authors to identify
the load carrying aggregate particles and, eventu-
ally, was used to evaluate the permanent defor-
mation behaviour of unbound granular materials
(Yideti et al. 2013a). The framework identifies
the two basic components of unbound granu-
lar materials, namely the Primary Structure (PS)
and the Secondary Structure (SS). The Primary less SS Optimum SS Excess SS
(Unstable) (Stable) (Disrupted)
Structure (PS) is a network of interacting grains,
0.5 $DP$1.0 DP > I
of a range of coarse sizes that forms the load DP <0.5

carrying network of the UGM. The Secondary

Structure (SS) is formed of grains, of sizes smaller Figure 3. The disruption potential of PS due to SS.

169
 In general, a low DP represents a situation with the shakedown critical load resulting from finite
a low amount of secondary structure in the voids element analysis in terms of disruption potential.
of the PS and this leads to unstable conditions. Unbound materials in the first group are
Good performance is expected materials with an Crushed Rock (designated CR) and gravel (CG)
optimum amount of secondary structure that can aggregates as presented in a report from Kolisoja
provide stability within the PS. A high DP repre- at Tampere University of Technology [39]. The
sents a disruptive condition that eventually leads five crushed rock aggregates were Finnish test
the material into incremental collapse under load materials with four of the aggregates being the
repetition, because the PS particles are no longer in same rock type, vulcanite, quarried from Sievi
intimate contact (see right hand part of Figure 3). (designated SI), whereas the fifth was gneiss quar-
In this study, the value of DP will be used to ried from Hallila (HA). In this group there are also
identify the ranges of shakedown limits. The math- two Norwegian test materials which were crushed
ematical expression for DP is given in Equation (1). gravel from Hovinmoen (HO) and crushed gneiss
from Åndalen (AN). Each of the seven types of
SS
VDM aggregates has two similar grain size distributions
DP = (1) (designated 1 or 2) with different properties. The
V frPSee
basic properties of each of the aggregates are sum-
marized in Table 1 and the intended grain size
where, is the volume of disruptive material and is distributions for both Finnish and Norwegian test
the free volume (voids) within the primary struc- materials are shown in Figures 4 and 5, respectively.
ture. The disruptive material volume ( ), includes SS Combining the designators gives the notation (see
particles larger than 0.225 times the smallest grain Table 1) with the additional T or N in the suffix
size of the PS. The 0.225 factor corresponds to the representing Tampere or Norway, respectively.
ratio of the average void size to the spherical parti- All the permanent deformation tests were con-
cle size for the densest possible packing of uniform ducted using State Highway Research Project
spheres (Cubic and Hexagonal close packing). (SHRP) protocol P46 (AASHTO T294-92) with
large-scale test specimen diameter of 300 mm and
height of 600 mm. The tests were conducted at
4 MATERIAL CHARACTERIAZATION the Tampere University of Technology, Finland
in which each specimen was exposed to about one
There are two groups of unbound granular million loads under a constant confining pressure
materials that are used for the purpose of this study. of 50 kPa. The axial load pulse was varied from
The first group is used in the verification of the 200 kPa to 400 kPa.
shakedown ranges based on their disruption poten- Permanent deformation accumulation during
tial values. The second is used for the evaluation of the first few thousands or tens of thousands of load

Table 1. Basic properties of the test specimens (Kolisoja 1999).

Basic properties of the test specimens

Specific gravity Dry density Water content Unit volume Unit volume
Notation Gs ρd(g/cm3) w [%] of aggrega-tes of voids

CRSI_T5/1 2,77 2.223 2.0 0.803 0.197

CRSI_T5/2 2.215 2.4 0.800 0.200
CRSI_T6/1 2.131 1.4 0.769 0.231
CRSI_T6/2 2.127 1.4 0.768 0.232
CRSI_T7/1 2.260 2.4 0.816 0.184
CRSI_T7/2 2.246 3.0 0.811 0.189
CRSI_T8/1 1.751 0.3 0.632 0.368
CRSI_T8/2 1.745 0.4 0.630 0.370
CRHA_T5/1 2,72 2.205 2.7 0.811 0.189
CRHA_T5/2 2.186 3.1 0.804 0.196
CGHO_N1/1 2,66 2.142 4.5 0.805 0.195
CGHO_N1/2 2.153 4.5 0.809 0.191
CRAN_N2/1 2,74 2.137 4.0 0.780 0.220
CRAN_N2/2 2.130 4.0 0.777 0.223

170
 120 I I I '"' Table 2. Permanent deformation results at different
-+- CRSI_T611 & 2 load repetitions.
100 -+- CRSI_TS/ 1 & 2
Permanent strain % Permanent strain %
-r- CRSI_TI/ 1 & 2
80 Materials @200000 @END
___. CRSI_T8/ 1 & 2
-~"" 60 ~
f/, The seven materials with large permanent strains
_.,_ CRHA_TS/ 1 &2
~v >2.5
~
~
CRSI_T5/2 0.508
40 CRSI_T6/2 0.671 2.414
~1 1 Vlf CRSI_T7/1
CRSI_T7/2
0.537
0.858
>2.5
>2.5
20

0
0.01
....-~.

0.1
I
rt J
10 100
CRHA_T5/2 0.326
CGHO_N1/2 0.466
>2.5
>2.5
CRAN_N2/1 0.277 >2.5
Sieve Size [nun)
The five stable materials
Figure 4. Five finnish crushed rock gradations. CRSI_T8/1 0.396 0.5
CRSI_T8/2 0.604 0.696
CRHA_T5/1 0.076 0.189
120 ' I
CGHO_N1/1 0.181 0.265
"
100
-+- CRAN_N2/ 1 & 2 II CRAN_N2/2 0.099 0.126

80 -+- CGHO_N l/1 & 2

~ materials, the strain readings were abstracted for
f 11 all materials at both 200000 and 106 load repeti-
- ""
~
~ 60 tions. The results are presented in Table 2.
J
~
~
40
5 RESULT AND DISCUSSIONS
20
v-/
,f-J:: v 5.1 Disruption potential and shakedown concept
0 Before describing the permanent deformation
0.01 0.1 10 100
results, the packing parameters for the same mate-
Sieve Size [nun)
rials are summarized in Table 3. The table lists the
Figure 5. Two norwegian crushed rock gradations.
DP values of all the materials including their PS
ranges, available free volume of voids within the
PS and the volume of disruptive SS aggregates.
cycles was very rapid, but the rate was decreasing Figure 6 shows the larger permanent axial
with the number of cycles. After that, the total per- strain results (relatively large when compared
manent axial strain increased for a while at a more with those of the other materials) as a function
or less constant rate. Seven materials (i.e., CRSI_ of the number of load repetitions for the seven
T5/2, CRSI_T6/2, CRSI_ T7/1, CRSI_T7/2, crushed rock materials. It can also be observed
CRHA_T5/2, CGHO_N1/2 and CRAN_N2/1) from the figure that, except for one of the crushed
showed large permanent strains and the test were rocks (CRSI_T6/2), all exhibited large permanent
stopped before the intended number of load rep- strains before the intended number of load rep-
etitions, N = 106 (Table 2), had been applied. etitions (N = 106 cycles) had been applied. From
Two of the crushed rocks from Sievi (CRSI_T5/1 the shakedown concept these kinds of permanent
and CRSI_T6/1) exhibited failure behavior at the strain behavior can be categorized either Range B
intended number of load repetitions and these or B/C. At the beginning of the permanent strain
are excluded from the evaluation described in this tests, Range B materials behave like Range A.
paper. Five of the remaining aggregate materials After a while the permanent strain rate (perma-
(CRSI_T8/1, CRSI_T8/2, CRHA_T5/2, CGHO_ nent strain per load cycle) decreases or is constant
N1/1 and CRAN_N2/2) showed a stabilizing type or only slightly increasing. Range B/C represents
of behavior during the application of the intended a marginal response between Range B and Range
number of load repetitions. C (incremental collapse). Among the materi-
Initial densification of the specimens was mostly als, CRSI_T5/2, CRSI_T6/2, CRHA_T5/2 and
complete after the first 50000–100000 load repeti- CRAN_N2/1 behaved according to Range B by the
tions. Therefore, in order to determine the perma- end of the tests. Similarly, CRSI_T7/1, CRSI_T7/2
nent deformation resistance behavior of the test and CGHO_N1/2 exhibited a Range B/C response.

171
 Table 3. Packing theory parameters. 2.50
2.25
Materials PS range VPSfree(%) VssDm(%) DP 2.00
~

CRSI_T5/2 32.0–2.0 16 8.8 0.55 -_g- 1.75

1.50
CRSI_T6/2 32.0–2.0 20.79 4.9 0.236 "'
CRSI_T7/1 32.0–2.0 14.8 5.09 0.344 ] 1.25
+ CRSI_TS/2
CRSI_T7/2 32.0–2.0 15.21 5.06 0.333 ~ LOO
• CRSI_T6/2
CRHA_T5/2 32.0–2.0 15.64 8.981 0.574 ~ 0.75 &
X
CRSI_T7/ I
CRSI_T7/2
CGHO_N1/2 16.0–2.0 13.33 10.946 0.821 ~ 050
0< CRH<\_TS/2
CRAN_N2/1 8.0–2.0 8.14 6.2392 0.767 0.25
CRSI_T8/1 32.0–16.0 36.16 1.11 0.031 0.00
CRSI_T8/2 32.0–16.0 36.35 1.1 0.03 200000 400000 600000 800000 1000000 1200000
CRHA_T5/1 32.0–2.0 15.08 9.059 0.601 Nt.unber of Load Repetitions
CGHO_N1/1 16.0–2.0 13.61 10.892 0.8
CRAN_N2/2 8.0–2.0 8.24 6.2192 0.753 Figure 7. Permanent axial strain versus disruption
potential for Range B and B/C.

I I

0.9
I I • CRSI_T S/2
250
.••.•• "=200000
+ I I
:1( CR SI_T612
2.2 5
-;;<.
0.8
0.7
I I •+ CRSI_T7/1
CRSI_T 7/2
~2.00
+
•
CRSI_TS/1
CRSI_TS/2
~
·a
X ' ·,
I I
-~
X CRR<\_T S/2 b 1.75 & CRH<\_TS/1
b
rn 0.6 til

] 0.5 • ·,_I

~ ---- .. •
I

I
•... CGHO_N! /2
CRAN_N 2/1
] 1.50
1.25
0< CGHO_N 1/ 1
• CRAN_N212
0.4 ~
~
~ I 1- - - limit :i LOO
a X
•-
~~~~~~~~~~~~~~~~~~~==~
0.3
~ 0.2
I
:-I I ~ 0.75

"" 0.1
I I
E0.50
I I 0.25
0 0.00
0 0.5 1.5 200000 400000 600000 800000 1000000 1200000

Disruption Potential, DP Number of Load Repetitions

Figure 6. Range B and B/C crushed rock and gravel Figure 8. Crushed rock and gravel materials exhibiting
materials (Kolisoja 1999). Range A response (Kolisoja 1998).

Figure 7 plots the results of materials with a i.e. the response is of Range A type. In Figure 9 it
Range B or Range B/C response and shows that can be observed that the DP values of these mate-
those materials with a DP value between 0.5 and 1 rials were found to be between 0.5 and 1 except
exhibited lower permanent strain (measured at for the two uniformly graded crushed rocks from
200000 load repetitions) than materials having a Sievi (CRSI_T8/1 and CRSI_T8/2). These have an
DP value less than 0.5. Thus the results show that exceptional gradation with a large percentage
an adequate volume of SS particles between PS volume of PS aggregates of a very narrow range
particles enables an UGM to exhibit good resist- (32–16 mm) of grain sizes (Table 3). Thus, due to
ance to permanent deformation as postulated the small volume of SS aggregates within the uni-
previously [36]. Thus, a coarse grained granular formly graded material, the DP values do not satis-
material should have an optimum amount of SS factorily characterize the deformation behavior of
particles that reside in between PS particles in the materials. Apparently, coarse-grained crushed
order to resist deformation. The stable and unsta- aggregates of uniform size are able to interlock
ble responses of UGM during the initial compac- each other to resist deformation without getting
tion stage are greatly affected by the interplay support from fine-grained aggregates. This sug-
between PS and SS particles. In general, it can be gests some limit to the PS-SS schema described in
observed in Figure 7 that the DP value has a sig- this paper, but the limits of its applicability have
nificant effect on the shakedown response of the not yet been determined. Despite this limitation,
unbound granular materials. practical road aggregates have very different gra-
As shown in Figure 8, the other five crushed rock dations, which, clearly, do not come close to this
and gravel materials are stable until N = 106 cycles, limitation.

172
 + CR!S_TS/ 1
0.9
X CRSI_TS/2
0.8
e CRIL'\_TS/1
~ 0.7 A CGHO_NI /1
·~ 0.6 • CRAN_l'1212
rn
-a; 0.5 e.,-----~------_._ - - lim it

~ 0.4
c
c OJ
~ 0.2
--" - Figure 10. Modeled sample of material.
"" 0.1
0
•
~ 1.5 f t n φ c) ≤ 0 (2)
Disruption Potential, DP
where, σt and σn are the tangential stress and nor-
Figure 9. Permanent axial strain versus disruption mal components, respectively. Since the aim of this
potential for materials exhibiting Range A response. finite element analysis was to evaluate the shake-
down critical load for several confining pressures,
6 CALCULATION OF SHAKEDOWN an axial displacement, Δl = 5 mm, was imposed on
LOAD LIMIT the top of the specimen and confining stresses,
σ3 = 20, 45, 70, 100 and 150 kPa, were applied uni-
Lower and upper bound limit loads calculations formly around the sample (Figure 10).
have been suggested by several researchers using In such cases, it is simple to read from the analy-
Melan’s static and Koiter’s kinematic theorems sis the mean stress (ps) at which the plastic deforma-
respectively. However, in this study, a simple finite tion starts. Finally, the following equation was used
element analysis for critical shakedown load (Ps) to calculate a dimensionless load parameter (μ).
has been proposed in order to evaluate the experi-
mental results from triaxial tests and eventually Ps p
μ= = s (3)
correlate these with the disruption potential value π a 2c c
of primary structure particles.
where Ps = λP and ps = λp are the values of mean
load and mean stress, respectively, at which shake-
6.1 Finite element modelling
down occurs, λ is the shakedown load factor, P is
In the finite element model, a three-dimensional the applied load and p the applied stress, respec-
cylindrical sample with a diameter of 150 mm and a tively, πa2 is the cross-sectional area of the loaded
height of 300 mm was used; the size being exactly the area. The results give the lower limit of shakedown
same as the experimental sample in the triaxial test. load at which plastic strain starts; with the mate-
1800 eight-node brick elements were used for the rial behaving purely elastically below this mean
finite element simulations and, for computational stress condition. Before explaining the results of
reasons, only a quarter segment (full height) of this model, an overview is given of sources and
the specimen was meshed with the appropriate types of materials used.
boundary conditions (Figure 10). The material was The group of unbound granular materials which
modelled as elastic perfectly-plastic, yield being have been used for the simulation by this finite ele-
described by the Mohr-Coulomb criterion (Eq. 2), ment analysis are three different materials corre-
characterized by a cohesion (c) and angle of internal sponding to three different morphologies. These
friction (ϕ). In the reported calculations here the c unbound granular materials have been collected
and ϕ were calculated from a series of static failure from a reference paper [39]. One of the materi-
tests which had been conducted by Erlingsson and als was a crushed granite rock from Skärlunda
Rahman (Erlingsson & Rahman 2013). in Sweden (Skärlunda, SE). The second material
The load at which the material starts yielding, (Hallinden, SE) was also homogenous crushed
was adopted as the shakedown load; the approach granite with no foliation, which was obtained from
widely used in upper bound limit analysis Hallinden, Sweden. The third material (SG1, DK)
(Collins & Boulbibane 2000). The stress compo- was natural gravel mixed with crushed gravel; it
nents should satisfy the Mohr-Coulomb condition was obtained at a gravel pit in Denmark.
expressed as in Eq. 2, in which the yield function The grain size distributions of all the materi-
f( ), evaluates as less than or equal to zero. The als are shown in Figure 11. The material para-
Mohr-Coulomb yield criterion is given by: meters are also summarized in Table 4. The RLTT

173
 100 Table 5. Finite element analysis results.
90 - - - Skat-lunda, SE 1
80
/) DP σ3 [kPa] Ps [kPa] μ
- -Hallinden, SE
~ 70 Skärlunda, SE 0.369 20 59 1.64
~ 60
- - SG1 , DK ./~ 45 60 1.67
~
;>,
..0
50
I / I 70 62.3 1.73
I
:=......
40 / 100 64.4 1.79
·i 30
I 1--" 150
20
68
52.1
189
1.8
"" 20 : /
.,.. ~~
k:i"
45 54.6 1.88
10 I Hallinden, SE 0.488 70 57 1.97
I I
0 I
100 60 2.07
0 .063 0.63 6.3 150 65 2.24
Sieve size [mm] SGI, DK 1.227 20 110 3.14
45 118 3.37
Figure 11. Grain size distribution for the three 70 127 3.63
materials. 100 138 3.94
150 160 4.57
Table 4. Material parameters used in the finite elements
simulations.
Sequence 1
0.03
Friction Young - SGl, DK [DP =1.227) at High stress level
Material Cohesion, angle, modulus Poisson’s
.:; - Skarlunda. SE [DP = 0.369) at low stress level
0.025
parameters c (MPa) φ (°) E (MPa) ratio (v) f 0.02
- Hallinden, SE [DP = 0.488] at l ow stress level

Skärlunda, SE 0.036 45 459 0.35 1

Hallinden, SE 0.029 48 303 0.35
[ 0.015
~

SGI, DK 0.035 65 537 0.35 0.01

J 0.005

tests were carried out based on the European 0

standard EN-13286–7 (CEN-European 2004) 0 10000 20000 30000 40000 50000 60000

by using low stress levels for Skärlunda. SE and Number of load cycles, N
(• )
Hallinden, SE materials, and high stress levels for
Sequence5
SG1, DK material, according to that standard,
the standard tests were divided into five sequences 0.06

with different stress levels. .:;

0.05
f
6.2 Results 1i
;;
004

The results show the influence of the disruption [ 0.03

~

potential on the shakedown zone of unbound

granular materials at different stress levels. l"
~
0.02
- SGI , DK [DP = 1.227} at High stress level
Table 5 shows the mean stress level at which plastic 0.01 - Hallinden, SE [DP = 0.488] at Low stress level
deformation starts and the dimensionless shake- - S~unda, SE [DP =0.369] at low stress level
down load. The DP values for the three materials 0
230000 250000 270000 290000 310000
are also listed in the table. In this analysis, it should
be borne in mind that the mean stress at which the Number of load cycles, N

material starts yielding is defined as a critical stress (b)

in which the material starts to behave plastically.

Finally, the dimensionless shakedown load was Figure 12. Permanent axial strain versus number of
obtained by using the cohesion of the materials as load cycles (Sequence number is as defined by the stand-
ard (CEN-European 2004)).
stated in Equation 3 above.
Before explaining the influence of DP value
on the dimensionless shakedown load of UGM, At low stress level, the Skärlunda SE material,
Figure 12a and b describe the permanent axial having a DP value of 0.369, showed a higher plas-
strain test results for the three materials at the begin- tic strain throughout the RLTT compared with
ning and end of the test sequences respectively [62]. the Hallinden, SE material which had a DP value

174
 potential—in the shakedown theory was investigated.
4.5 From this, the following main conclusions were
"-
1. &
drawn:
.2
3.5 &
• An influence of DP value on the three shakedown
.,~
&
0
& response ranges (Range A, B and C) was observed
~ 2.5
for all unbound materials. From the results, it was
• found that the shakedown concept satisfactorily
•• •• •• • supported the concept of an optimum disruption
0

'
.Q
~ 1.5 + Skiirlunda, SE (DP=0.369)
potential limit (from 0.5 to 1) for both shakedown
E
Ci Range A and Range B responses.
0.5
• Hallinden, SE (DP = 0.488)
• The critical shakedown load limit at which the
0
& SGI , DK (DP = 1.227) unbound materials become resilient in nature
0 50 100 150 200 and no further accumulation of plastic strain
Confining stress, 0 3 occurs after a number of load application,
depends on the amount of SS particles within
Figure 13. Comparison of values of dimensionless PS particles (i.e., the effect of the DP value).
shakedown load for the three materials. • The Finite element analysis validated satisfacto-
rily the dimensionless shakedown loads of the
three materials with their experimental results.
• Aggregates with a very uniform (sing-sized)
of 0.488—the expected result for such a low DP.
grading maintain a stability against permanent
However, at high stress level, the disrupted SG1,
deformation by another mechanism. Thus there
DK material having a DP value of 1.227 exhibited
is some limit to the Disruption Potential model
(as expected) a high plastic deformation as shown
for assessing likelihood of unstable response to
in Figures 12a and 12b.
repeated loading of UGMs, but the limiting cri-
Figure 13 presents the dimensionless shakedown
terions (or criteria) are not yet known.
load (µ) with the corresponding confining stress
for the three unbound granular materials. As
mentioned above, those materials with a low ACKNOWLEDGMENTS
amount of SS (DP < 0.5) have low values of shake-
down load, due to the unstable condition of the The authors would like to extend their sincere
PS network of particles. However, in the disrupted gratitude to the Swedish Road Administration
material (SG1, DK) the dimensionless shakedown (Trafikverket) for providing the financial support
load is higher, which usually means an excessive for the project. The authors are also indebted to
amount of SS resists permanent deformation for Prof. Pauli Kolisoja at Tampere University of
a short period of time and eventually leads to Technology, Finland for providing the experimen-
dilation effect. Nevertheless, from Figure 12b it is tal data for this paper.
clear that this higher non-dimensional load causes
greater accumulation of permanent deformation
as would be expected for a material with a high DP. REFERENCES

AASHO. Road Test Report 7: Summary Report; 1962.

7 CONCLUSIONS AASHTO T 294-92. Interim method of test for resilient
modulus of unbound granular base/subbase materials
and subgrade soils—SHRP protocol P46. American
This study has been implemented by incorporating Association of State Highway and Transportation
a packing theory-based framework into shakedown Officials.
theory. The packing theory framework provides a Barksdale RD. Laboratory Evaluation of Rutting in
gradation—based approach to evaluate fundamen- Basecourse Materials. Proceedings of 3rd Interna-
tal mechanical properties of unbound granular tional Conference on the structural Design of Asphalt
materials. Similarly, shakedown theory provides a Pavements, London, UK; 1972:161–174.
generalized performance-based approach using a Boulbibane M, Collins IF, Ponter ARS, Weichert D,
number of load applications to evaluate permanent 2005. Shakedown of unbound pavements. Road
deformation. It is very important to determine the Mater Pave 2005; 6:81–96.
Boulbibane M, Ponter ARS. Extension of the linear
physical basis of shakedown limits for interpreting matching method to geotechnical problems. Comput.
the permanent deformation behavior of unbound Methods intApplMechEng 2005; 194:4633–4650.
granular materials. Brown SF. 36th Rankine Lecture: Soil Mechanics in
The interpretation and implementation of one Pavement Engineering. Géotechnique 1996a; 46(3):
of the key packing theory parameters—disruption 383–426.

175
Brown SF. Design considerations for pavement and rail Kolisoja P. Large scale dynamic triaxial tests, III. Tampere
track foundations. Proc. Inst. Seminar on Geotech- University of Technology, Helsinki, Finland; 1998.
nicin Pavement and Railway Design and Construc- Korkiala-Tanttu L, Laaksonen R, Törnqvist J. Effect of
tion, Athens, Greece 2004b; 61–72. the Spring and Overload to the Rutting of a Low-
CEN-European Committee for Standardization. Volume Road. HVS-Nordic-research. Finnish Road
Cyclic Load Triaxial Test for Unbound Mixtures. In Administration. Finnra Reports 22/2003; Helsinki
European Standard, No. EN 13286–7, Brussels, 2004. 2003.
Chan, F.W.K. Permanent Deformation Resistance of Larsen H, Ullidtz P. ‘Pavement Subgrade Performance
Granular layers in Pavements. PhD Thesis, University Study in the Danish Road Testing Machine’, Interna-
of Nottingham, Nottingham, UK; 1990. tional Conference on Asphalt Pavements, 8th, 1997,
Chazallon C, Hornych P, Mouhoubi S. An elasto-plastic Seattle, Washington, University of Washington, Seat-
model for the long term behaviour modelling of tle, Washington, USA, 1997; 1:843–857.
unbound granular materials in flexible pavements. Int Lekarp F, Dawson A. Modelling permanent deformation
J Geomech 2006; 6(4):279–89. behaviour of unbound granular materials. Constr.
Collins IF, Boulbibane M. (1998b). Shakedown analy- Build Mater 1998; 12(1):9–18.
sis and design of unbound pavements. Proc Roading Lekarp F, Dawson AR. Analysis of permanent defor-
Geotechnics 1998b, The Insti Profess Eng, Auckland, mation behavior of unbound granular materials.
New Zealand, 24:59–64. Proc. Symposium on Thinly Surfaced Pavements,
Collins IF, Boulbibane M. Geomechanical analysis of Univ. New Brunswick, Fredericton, Canada; 1997:
unbound pavements based on shakedown theory. 91–99.
J Geotech Geoenviron Eng 2000; 126, 50–59. Loach SC. Repeated Loading of fine grained soils
Collins IF, Boulbibane M. Pavements as structures sub- for pavement designs. PhD Thesis, University of
jected to repeated loadings. The mechanics of struc- Nottingham, Nottingham, UK; 1987.
tures and materials, Grzebieter, Al-Mahaidi, and Melan E. TheoriestatischunbestimmterSystemeaus ideal-
Wilson, eds., Balkema, Rotterdam, The Netherlands; plastischemBaustoff. Sitzber. Akad. Wiss. Wien, IIa
1997:511–516. 1936; 145:195–218.
Collins IF, Cliffe PF. Shakedown in frictional materials Nguyen AD, Hachemi A, Weichert D. Application of the
under moving surface loads. Int. J. Numer Anal Meth- interior point methods to shakedown analysis of pave-
ods Geomech 1987; 11:409–420. ments. Int J Numer Methods Eng 2008;. 75:414–439.
Collins IF, Wang AP, Saunders LR. Shakedown in layered Sangrey DA, Henkel DJ, Esrig MI. The effective stress
pavements under moving surface loads. Int. J. Numer response of a saturated clay soil to repeated loading.
Anal Methods in Geomech 1993a; 17:165–174. Canadian Geotechnical J 1969; 6:269–277.
Collins IF, Wang AP, Saunders LR. Shakedown theory Sharp RW, Booker J. Shakedown of Pavements under
and the design of unbound pavements. Road and Moving Surface Loads. J Trans Eng 1984; 110(1):
Transport Res 1993b; 2:29–38. 1–14.
Collins IF, Boulbibane M. The application of shake- Shiau SH, Yu HS, 2000. Load and displacement predic-
down theory to pavement design. Metals and materi- tion for shakedown analysis of layered pavements.
als 1998a; 4(4):832–837. Trans Res. Record 2000; 1730:117–124.
Collins IF, Wang AP, Saunders, LR. Shakedown Theory Werkmeister S, Dawson A, Wellner F. Pavement Design
and the Design of Unbound Pavements. Australian Model for Unbound Granular Materials. J Trans Eng
Road Research Board, Road and Transport Research 2004; 130(5):665–674.
Vol. 2, No. 4, 1993, 28–37. Werkmeister S, Dawson A, Wellner F. Permanent Defor-
Dawson A, Wellner F. Plastic Behaviour of Granular mation Behaviour of Unbound Granular Materials
Materials. Final Report ARC Project 933. University and the Shakedown-Theory. Trans Res Record 2001a;
of Nottingham, Nottingham. UK, Reference PRG990; 1757:75–81.
1999. Werkmeister S, Numrich R, Wellner F. The Develop-
Erlingsson S, Rahman S. Evaluation of Permanent defor- ment of a Permanent Deformation Design Model for
mation Characteristics of Unbound Granular Mate- Unbound Granular Materials with the Shakedown-
rials from Multi-stage Repeated Load Triaxial Test. Concept. Proc of the 6thint symposium on the
TRB (2013) Annual Meeting. Bearing Capacity of Roads and Airfields (BCRA),
Heath DL, Shenton MJ, Sparrow RW. Design of Con- Lisbon, Portugal; 2002b, 24–26.
ventional rail track foundations. Proceedings of the Yideti TF, Birgisson B, Jelagin D, Guarin A. Packing-
Institution of Civil Engineers 1972; 51(2):251–267. based theory framework to evaluate Permanent
Koiter WT. General theorems for elastic-palstic solids. in: Deformation of Unbound Granular Materials. Int.
progress in solid Mechanics (Eds. I.N. Sneddon and R. J Pave Eng 2013a; 14(3):309–320.
Hill), Amsterdam, The Netherlands; 1960:165–221. Yideti TF, Birgisson B, Jelagin D, Guarin A. Packing-
Kolisoja P. Deformation modulus of unbound pavement based theory framework for Evaluating the Resilient
material. Proceedings of the 11th European Confer- Modulus of Unbound Granular Materials. Int. J Pave
ence on Soil Mechanics and Foundation Engineering. Eng; 2013b, Published online: 29 Nov (2013b), DOI:
Copenhagen 1995; 3:119–124. 10.1080/10298436.2013.857772.

176
Asphalt mixes
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Experimental study on the effects of angular rate on the torsional shear

property of asphalt mixture

Yuhua Li, Yalong Li, Guangyuan Wang, Shuan Zhang, Yejia Liu & Jingyun Chen
School of Transportation and Logistics, Dalian University of Technology, Dalian, China

ABSTRACT: As the obvious viscoelasticity property of asphalt mixture, the loading rate is surely one
of the most important factors affecting the mechanical performance of asphalt mixture. Therefore tor-
sional shear test of asphalt mixture is definitely a reasonable proper method to determine the shear resist-
ance. In this paper, based on self-assembled torsional shearing test device, AC13 dense-graded asphalt
mixture was prepared to form 10 sets of square specimens to undergo the shearing failure test through
which the maximum shear stress τmax, the maximum shear strain γmax, the shear modulus G and the cor-
responding load time Δt were obtained under different torsional loading angular rates between 0.1 ∼ 1.0
rad/s (interval 0.1 rad/s) at 20°C. It turns out that with the increase of ω, τmax gradually increases and
reaches the maximum at the point of ω = 0.6–0.7 rad/s after which τmax declines slowly, and γmax rises little
by little all the way, and G goes a little bit up firstly and then steadily decreases to the end, and Δt is obvi-
ously rapidly falling down and finally to a stable state.

Keywords: Asphalt mixture, Torsional shear test, Angular rate, Failure form, Maximum shear stress,
Maximum shear strain, Load time

1 INTRODUCTION pression and torsion condition (Marc E. Novak,

2007). As the obvious viscoelasticity of asphalt
The theoretical calculation shows that the maximum mixture (Zhao Yanqing, Huang Da-xi, 2008), the
shear stress analyzed within the asphalt pavement loading rate and the test temperature are the main
can reach 0.4 MPa and ever more under the stand- factors to be considered. This paper, based on self-
ard vehicle load or the most unfavorable situation assembled torsional shearing test device, focuses the
(Li Yuhua, 2013). Since the shear resistance property research on the affection of loading angular rate
of asphalt mixture is closely related to rutting, wrap- on torsional shear failure performance of asphalt
ping and other permanent deformation, the shear mixture.
strength of asphalt mixture can never be ignored
in asphalt pavement design. There have been many
methods developed to determine the shear strength
2 THE EXPERIMENT DEVICE AND
of asphalt mixtures, including tri-axial compression
PRINCIPLE
test (Zhu Haoran, Yang Jun, Chen Zhiwei, 2009),
Super pave shear test (AASHTO, 2000), direct
2.1 Device composition
shear test, rotary compaction shear test (Guler M.,
Bahia H.U., Bosscher P.J., etc. 2000), uniaxial pen- Figure 1 shows an self-assembled torsional shear-
etration test (Bi Yufeng, Sun Lijun, 2005), coaxial ing test device of asphalt mixtures, divided into
shear test (Feng Junling, Guo Zhongyin, Yang Qun torsion loading subsystem, data acquisition sub-
et al., 2008), and torsional shear test (Li Yuhua, 2013, system, specimen thermal insulation subsystem
Ravindran, Parag et al., 2007, Pham, Linh Viet, and automatic control subsystem, consisting of
2003). etc. As to the torsional shear method, which platform board, transducer, micro motor, coupling,
applies an axial torsional load to the specimen, test- torsion shaft, clamping, specimen and a connecting
ing equipment with a special torsional loading part shaft, torque sensor, angle sensor, temperature sen-
is necessarily needed. At present, complex shear- sor, RS232/RS485 communication, thermal insula-
ing modulus G* can be obtained under pure tor- tion pipe, circulating pump, high-low temperature
sional condition with a Dynamic Shear Rheometer experimental box, and the computer program. The
(DSR) (ASTM, 2014), and lots of shear resistance output shaft of the micro motor is connected with
properties can be obtained by a Hollow Cylindrical the torsion shaft, the clamping and one end of the
Torsional shear tester (HCA) under tri-axial com- test specimen successively through the coupling,

179
 <D Connected to t he
111icro 110tor shaft
®Torsion load inc shaft

r
®CI1111Pin1 dev ice
@Square columar speci~~en
(Asphalt mi rtures)
®Couplin1 axis
@ Torque sensor
T : load torque value
Lo : Spec inen free length
s : Spec inen side length

Figure 1. The experiment device for torsional shear of Figure 2. Shear stress and shear strain calculation of
asphalt mixtures. free torsion.

while the other end of the test specimen is fixed on mation in the process of twisting which is treated as
the torque sensor through the connecting shaft, and free torsion approximately. As to the analysis, the
the angle sensor on torsion shaft. torsional shear stress states of each cross section
are the same, but the shear strain is different. The
2.2 The method of fixing specimen shear stress and strain of the point A in Figure 2
are the maximum value of the specimen. Thus, the
A square groove (deep 10 mm) is located in the τ(t) and γ(t) curves of the point A over the cross sec-
clamping. Firstly, put glass cement on the end of tion of the specimen can be obtained consequently.
the specimen and the square groove. Secondly,
both ends of the specimen are directly embedded T
τ= (1)
into the groove, and then laid at room temperature 0.208s 3
20°C for 12 hours. To avoid the specimen eccen- where τ(MPa) is the shear stress of the edge mid-
tric, the length of square groove slightly larger (less point of the cross section of specimen. T ( N mm )
than 1 mm) than specimen and located in the center is the torque of specimen. s(mm) is the length of
position. The consolidation of square groove and the square specimen.
the bonding effect of glass cement to ensure the
connection requirements of the specimen. as
γ = (2)
2 L0
2.3 Experiment principle
where γ is the shear strain of midpoint A on the
The output frequency of the transducer, accurately edge over cross section near the end of torsion
controlled under the computer program, adjusts loading shaft. a(rad) is torsional displacement at
the output shaft rotating angular rate ω of the point A. L0(mm) is the free length of the specimen
micro motor, and then the torsion load is applied between the clamping.
on the bar specimen through force component like
coupling, torsion shaft, clamping. At the same
time, data, like the sensor torque (T), the angular 3 TORSIONAL SHEAR FAILURE
displacement (a), the temperature (F), are collected EXPERIMENT OF ASPHALT MIXTURE
and stored by computer program through RS233/
RS485 communication, also included correspond- 3.1 Material, forming process and dimension
ing time. Heat preservation is controlled by the parameters
high-low temperature experimental box, the circu- AC13 dense-graded asphalt mixture is selected,
lation pump and the insulating pipe together. made from the asphalt AH90 and limestone. The
According to the torque-time T(t) and angular asphalt mixture plate (300 mm × 300 mm × 50 mm)
displacement-time a(t) data, as well as the section are prepared according to the wheel rolling method,
size of the specimen s and the free length of the and then are cut into specimens by the rock cutting
specimen L0, the shear stress and strain of the mid- machine (Fig. 3). The theoretical size of the right
point on the edge, whose cross section comes near square prism specimen is 50 mm (length) × 20 mm
the end of loading shaft, are calculated by formu- (width) × 20 mm (high). While before the experi-
las (1) and (2) respectively. ment, the actual cross-section dimensions (s × s)
It must be pointed out that, due to the constraint and the actual free length between clamping (L0)
of the specimen clamping, there is warping defor- of each specimen are measured carefully in turn.

180
3.2 Experimental temperature and loading rate 3.3 Data and curve of original result
The torsional loading experiment is carried out The torque-time T(t) data curves obtained from the
after 1.5 h of specimen heat preservation, before original experiment of the specimens in 10 angular
which the specimen, together with the double layer rates under torsional loading conditions are shown
quartz glass insulation tube, has been put into the in Figure 4, and the angular displacements—time
constant temperature test box under 20 ± 0.01°C, a(t) are neglected because it’s simple linear prop-
which is also the test temperature. In order to erty. Figure 4 shows that the maximum torque error
obtain the affection of the torsional loading rate value and K value at each loading angle, as shown in
on the torsional shear failure performance of the Table 1. (K is the ratio of the difference between test
asphalt mixture, 10 kinds of torsional loading value and mean value to the standard deviation.)
angle rate are adopted ranging from 0.1∼1.0 rad/s.
A total of 60 specimens are divided into 10 groups,
each group have 6 specimens.
4 AFFECTION OF LOADING RATE ON
THE PERFORMANCE OF ASPHALT
MIXTURE

4.1 Affection of loading rate on the maximum

shear stress
According to the data in Figure 4, the shear stress
τ(t) curve can be obtained according to the for-
mula (1), including the maximum shear stress τmax,
which is shown in Figure 5.
As is shown clearly in Figure 5 that, with the
increase of the angular rate ω, τmax gradually
increases and reaches the maximum at the point of
ω = 0.6–0.7 rad/s after which τmax declines slowly.
Figure 3. Right square primal asphalt mixture As to this experiment, the average τmax is about 2.8
specimen. ∼ 4.9 MPa.

load Time. t { s) load Time , t (s)

load Time , t(s)

Load Time, t (s)

Figure 4. Torque-time curves of torsion shear loading process (Note: without curve from group 5th).

181
Table 1. The maximum torque error value and K value at each loading angle.

Angular rate (rad/s) 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Maximum error (%) 5.49 9.27 13.40 9.85 11.87 10.74 7.47 12.46 2.45 9.83
Kmax 2.08 1.97 1.69 1.63 1.89 2.05 1.39 2.07 1.62 1.68

45
6.0
40

-;;- 35
a..
:iiii
....-- v- ............:
r---.... 5' 30
.."'
~ 4.0 v ~
------- ~
E
e 3.5 ~ - - - - r- - N~
~~ .6427 1)2 + 8 58111 ~ + 2. 549
~ 3.0 / ... z -v
........., ....-.
ft

15
"-..

2.5
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 10 -
angular rate, w(rad/s) 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
angular rate, w(rad/s)
Figure 5. Affection of loading rate on the maximum
shear stress. Figure 7. Affection of loading angle rate on the shear
modulus.

0.40

;... 0.35 I
~t
f
l
§ 0.20
·~
0.30

0.25

. V:
.............:
L.--"E-"'
~p-

v=o.ft~(A
'
I
l 0 .4082
-(/) 4
+r
<1 3
\
\
~
b. t = f:.7123 w ·O.B37

~
0.15 - --- - R2 = 0 .9i 05
2
R 0.77 ~2
:iiii
~
0.10 "C

"' 1 ~ r---.J
I ,3
.......
0.05 - -
u u u u u u ~ u u u 0
angular rate . w(rad/s) 0. 1 0. 2 0. 3 0. 4 0. 5 0. 6 0. 7 0. 8 0. 9 1. 0
angular rate , w(rad/s)
Figure 6. Affection of loading rate on the correspond-
ing shear strain. Figure 8. Affection of loading angle rate on loading
time.

4.2 Affection of loading rate on the corresponding

between the tangent shear modulus G of each
shear strain at maximum shear stress
specimen and ω, which is shown in Figure 7.
Based on the angular displacement-time a(t) data, As is shown clearly in Figure 7 that, with the
the corresponding shear strain γ(t) curve can be increase of the angular rate ω, the shear modu-
obtained consequently according to formula (2), lus G goes a little bit up firstly and then steadily
which is easy to get the corresponding shear strain decreases to the end with a serious discreteness. As
γ t ,τ max , denoted by γmax, when τmax occurs, which is to this experiment, the average shear modulus G is
shown in Figure 6. about 34.8 ∼ 16.9 MPa.
As is shown clearly in Figure 6 that, with the
increase of the angular rate ω, γmax increases slowly
4.4 Affection of loading rate on the correspond
in a power function form. As to this experiment,
loading time to achieve maximum shear stress
the average γmax is about 0.14 ∼ 0.34.
Based on the curve τ ( ) , it is easy to obtain the
relationship between the loading time Δt, corre-
4.3 Affection of loading rate on the shear
sponding to the maximum shear stress τ max , and the
modulus
loading angle rate ω, which is shown in Figure 8.
According to the previous analyzing, it is easy to get As is shown clearly in Figure 8 that, with the
the γ τ curve, and subsequently the relationship increasing of the angular loading rate ω, the

182
loading time Δt rapidly falls down with an obvi- stability from circular section and rectangular
ous power function form until ω = 0.9 / s, section should be compared carefully, together
after which it becomes not that obviously. As to with the maximum aggregate size of asphalt
this experiment, the average loading time Δt cor- mixture, to determine the minimum section size
responding to the maximum shear stress is about and specimen length.
0.77 ~ 5.33 s.
REFERENCES
5 CONCLUSIONS AND PROBLEMS AASHTO (2000). Standard experiment for determining
the permanent shear strain and stiffness of asphalt
5.1 Conclusions mixtures using the Superpave Shear Experimenter
1. With the increase of the angular rate ω, τ max (SST)[S]. Washington D.C: American Association of
State Highway and Transportation Officials Designa-
gradually increases and reaches the maximum tion TP7-00.
at the point of ω = 0.6–0.7 rad/s after which ASTM (2014). Standard Experiment Method for Deter-
τ max declines slowly. As to this experiment, the mining the Complex Shear Modulus (G*) Of Bitumi-
average τ max is about 2.8 ∼ 4.9 MPa. nous Mixtures Using Dynamic Shear Rheometer[S].
2. With the increase of the angular rate ω, γ max , (D7552−09).
the corresponding shear strain, increases slowly Bi Yufeng, Sun Lijun (2005). Research on Experiment
in a power function form. As to this experiment, Method of Asphalt Mixture’s Shearing Properties[J].
the average γ max is about 0.14 ∼ 0.34. Journal of Tongji University (Natural Science),
3. With the increase of the angular rate ω, the Vol.33(8), pp.1036–1040.
Feng Junling, Guo Zhongyin, Yang Qun, Xu Shao-guo
shear modulus G goes a little bit up firstly and (2008). Study on Co-axial Shear Experiment Method
then steadily decreases to the end with a serious of Asphalt Mixes[J]. Journal of Tongji University
discreteness. As to this experiment, the average (Natural Science), Vol.36(10), pp.1395–1398.
shear modulus is about 34.8 ∼ 16.9 MPa. Guler M., Bahia H.U., Bosscher P.J., etc. (2000). Device
4. With the increase of the angular loading rate ω, for measuring shear resistance of hot-mix asphalt
the loading time Δt rapidly falls down with an in gyratory compactor[J]. Transportation Research
obvious power function form and then steadily Record, 1723:116–124.
decreases to the end. As to this experiment, the Li Qiang, Lee Hyun Jong, Hwang Eui Yoon (2010).
average loading time Δt corresponding to the Characterization of permanent deformation of
asphalt mixtures based on shear properties[J]. Trans-
maximum shear stress is about 5.33 ~ 0.77 s. portation Research Record, 2181:1–10.
Li Yuhua (2013). Study on Experimental Method and
Rotational Shear Experimenting Apparatus of
5.2 Problems Asphalt and Asphalt Mixtures[D]. Dalian University
1. Due to the limited experimental conditions, the of Technology, China.
shear modulus obtained in this paper has not Marc E. Novak (2007). Creation of a Laboratory Experi-
been compared with other tests (such as SST). menting Device to Evaluate Instability Rutting in
Asphalt Pavements[D]. University of Florida.
2. According to the test results, the cause and Pham, Linh Viet (2003). Dynamic torsional shear experi-
effect relationship between the maximum shear ment for hot mix asphalt[D]. University of Florida.
stress τ max or maximum shear strain γ max Ravindran, Parag; Murali Krishnan J.; Rajagopal, K.R.
and the specimen failure can still not be fully (2007). Characterization of the non-linear response
obtained. of asphalt mixtures using a torsional rheometer[J].
3. 20°C, not 60°C or even higher, is selected in this Mechanics Research Communications, Vol.34(5),
paper, which is insufficient to reflect the high pp.432–443.
temperature stability of asphalt mixture, and it Zhao Yanqing, Huang Da-xi (2008). Viscoelastic Behav-
should be considered in the future study. ior of Asphalt Mixtures with Damage Stage[J].
China Journal of Highway and Transport, Vol.21(1),
4. Specimen size and boundary effect should also pp.25–28.
be taken into consideration. As to the boundary Zhu Haoran, Yang Jun, Chen Zhiwei (2009). Tri-axial
effect, there are specimen clamping, bonding, shear experiment on anti-shear properties of asphalt
and natural anchoring to be considered when mixture[J]. Journal of Traffic and Transportation
it comes to the effect on the test result. Results Engineering, Vol.9(3), pp.19–23.

183
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Comparison of flexural strength and crack propagation test

characteristics of SMA mixtures including the ageing impact

Pavla Vacková, Jan Valentin & Petr Mondschein

Department of Road Structures, Faculty of Civil Engineering, CTU in Prague, Prague, Czech Republic

ABSTRACT: Low temperatures decrease the ability of an asphalt mixture to relax after repeated traf-
fic loading and therefore the probability of cracking is increased and becomes even more critical with
ageing. In order to prevent cracking appropriate mix design with suitable aggregates and high quality
bituminous binders are the prerequisites. Additionally, it is always important to study and understand
the performance behaviour, which in this case is related mainly to temperatures below 0°C. For such
assessment several test methods are presently available. This research compares two different destructive
methods of asphalt testing in low temperature range with focus on flexural strength (three-point bending
beam test) and crack propagation test. Three-point bending beam test according to the Czech technical
specifications TP151 was performed on un-notched beams. In addition, half of the test specimens was
tested after artificial laboratory long-term ageing according to pr EN 12697–52 (5 days at 85°C). The
load-deformation diagram was used to calculate the fracture energy. At the same time SCB test accord-
ing to EN 12697-44 was performed on notched semi-cylindrical test specimens. For some of the tested
mixtures the same laboratory long-term ageing protocol was used and aged test specimens were tested
again for the same crack propagation test. In both cases ageing index was defined and gained values for
particular SMA mixtures compared.

1 INTRODUCTION binders 45/80-65, nevertheless in the actual Czech

national appendix to the standard EN 13108-5 it is
Stone Mastic Asphalt (SMA) mixtures are a type allowed to use also traditional paving grade binders
of hot asphalt mixture used in surface layers of (usually with gradation of 50/70), even if this appli-
heavy-duty pavements, like motorways. SMA mix- cation is for many years a matter of discussions.
tures are characterized by discontinuous granu- With respect to common bitumen quality this topic
larity and higher bitumen content compared to becomes even more actual and needs necessary cor-
asphalt concrete. The mixture has coarse aggre- rection in the national appendix/standard. On the
gate skeleton, which provides good stone-on-stone other hand, the standard does not focus in detail on
contact between the particles. The cohesion of the the use of various additives to bituminous binders,
skeleton is secured by the mastic, which is formed including the application of low-viscous options of
by a mix of increased content of a Polymer Modi- bituminous binders or the bitumen modification by
fied Bituminous binder (PMB), mineral filler and crumb rubber, pulverized rubber powder and/or alter-
cellulose fibres. native additives like Poly Phosphoric Acid (PPA).
Because of a specific design, these mixtures In the standard EN 13108-5 it is not allowed to
indicate increased fatigue resistance, high resist- use any Recycled Asphalt material (RAP), even
ance to permanent deformations and reflection though many of researchers already proved that
cracking as well as good skid resistance. The macro use of RAP might have and usually has a positive
texture of the wearing course brings not only good effect on asphalt mixture properties. Aged bitumen
comfort of ride for drivers but it can also lower the has lower penetration grade and therefore higher
noise pollution. stiffness characterizations. Already aged bitu-
SMA mixtures are designed and tested accord- men is also less predisposed to additional ageing.
ing to standard EN 13108-5. This standard requires Problem which is usually punctuated, might be the
to fulfil given volumetric characteristics, resistance behaviour in the range of low temperatures. For
to rutting and limiting value of bitumen drainage. this reason, presented research was aimed on two
Other tests and results are not required. methods used in the Czech Republic for assess-
In the Czech Republic SMA mixtures are usu- ment and characterization of asphalt mixtures in
ally produced with a polymer modified bituminous low temperature range.

185
2 TESTING METHODS the fracture energy, which can be defined as energy
expended for beam to break.
In the presented research two destructive testing The testing beams were divided into two sets.
method were chosen for assessment of the asphalt First set (“referential”) was left at laboratory con-
mixture characteristics related to the ability or ditions and the second set (“aged”) was experimen-
disability to cracking in the range of low temper- tally aged by conditions specified in pr EN 12697-52
atures—tensile bending (flexural) strength and (simulation of long-term asphalt mix ageing). This
crack propagation test characterized by fracture simulation of ageing was realized by conditioning
toughness. the beams in laboratory oven with air circulation
at temperature of 85°C for 5 days. After the ageing
period the beams were left in oven to cool down in
2.1 Tensile bending strength (Three-point
the same position to eliminate potential deforma-
bending)
tion of the test specimens during any manipulation
Tensile bending strength (simple destructive 3-point with hot beams. The aged beams were then tem-
beam test) was tested in observance to the method pered to 0°C and tested by the same method as the
given by the Technical specifications of the Minis- un aged test specimens. It is nevertheless necessary
try of Transport of the Czech Republic—TP 151. It to point out, that even the “un aged” specimens
is necessary to point out, that this is a specific test were expose to a certain level of ageing. As already
used mainly in the Czech Republic and required described the beams were always cut from the slab
presently only for High-stiffness Modulus Asphalt used for wheel tracking test. This test takes at least
mixtures (HMAC). The test was used in this 6 hours at a temperature of 50°C, so the slabs are
research for its long history and sufficient results already mildly aged after it.
with HMAC and its practicability and repeatability The resulting parameter is flexural strength and
which can be provided by a regular road laboratory. flexural modulus of elasticity, both in [MPa]. These
Simultaneously there is a long practice with this parameters were thanks to force-deformation dia-
kind of test in the Czech Republic. gram extended by the fracture energy. Fracture
Tensile bending strength is tested on beams, energy is calculated as an area under the curve of a
which can be cut from slabs used for wheel track- force-deformation diagram. Using an appropriate
ing test. Wheel tracking test (resistance to rutting) data recorder it is possible to get 50 data points of
is one of the test required for Initial Type Testing force and deformation per second.
(ITT) Protocol for SMA mixtures in the Czech All of these parameters were determined for
Republic. The slab can be used only in the case, referential (virgin/unaged) and aged test speci-
when it is not too deformed from wheel tracking mens with calculation of ageing indexes—simple
test. The standard limits for wheel tracking test are rate of the values showing the thermo-oxidation
quite low, so in almost all cases the slab can be re- degradation potential of the bitumen in an asphalt
used for tensile bending strength test. If the track mixture.
would be too deep or the slab would be anyhow
deformed from the test, it would be necessary to
2.2 Resistance to crack propagation
prepare new slab, especially for bending test. Thus
(in almost all cases) the tensile bending strength test The resistance to crack propagation (semi-cylin-
provides another use of already produced slabs. drical bending test) was determined in compliance
The testing temperature in TP 151 is specified with the requirements of EN 12697-44. The essence
in a range from -20°C to 0°C. For sake of this of the test is a three-point bending using semi-cylin-
research, temperature of 0°C was chosen for com- drical specimens with a defined crack in the middle
parability with the subsequent test. TP 151 speci- of bottom specimen face. The test procedure was
fies two testing velocities of loading (loading performed at a test temperature of 0°C applying a
rates): 50 mm/min and 1.25 mm/min. The slower standardized loading rate of 5 mm/min.
speed was chosen to allow comparability of the The specimens were manufactured and tested
test results with the later Semi-Cylindrical Bending during long period of time, so some of the variants
(SCB) test. were enlarged by additional testing procedures.
For the lower testing velocity it is possible to The specimens were manufactured according to
observe certain rate of relaxation. The lower veloc- EN 12697–30 (Specimen preparation by impact
ity has also defined a minimum limit strength value compactor) using 2 × 50 blows of the Marshall
given in TP 151, nevertheless this threshold value is hammer (50 blows from each side of the specimen).
valid only for HMAC. For this loading velocity it is Most of the variants were tested not only at the
also possible to accurately record the test data and temperature of 0°C, but also at −10°C to get addi-
draw the force-deformation (stress-strain) diagram. tional conception about the behaviour of tested
The diagram was used additionally to determine asphalt mixtures. Very last series of Marshall

186
specimens were exposed to the same artificial ageing were compared. The fibers were labelled as S-CEL
as at the previous test method. Currently at CTU in (commercially produced and regularly used fibre
Prague the scope of another research is to compare for SMA mixtures) and S-CEL R (innovative fibre
ageing of whole specimens, which are cut after the which should increase the strength characteristics).
ageing (like in this case), and in parallel ageing of The first applied chemical additive was Inno-
already cut un-notched semi-cylindrical specimens. dur (further labelled as “Inn”). It is an amide wax
The resulting required quality parameter is based additive containing polyamines and fatty
defined as the force needed for the specimen to acids. This additive is primarily used to improve
crack and the volume of the test specimen. It is the bitumen viscosity and it should have a posi-
called fracture toughness and usually shown in tive influence on the deformation behaviour of the
[N*mm−3/2]. For the specimens exposed to artificial asphalt mixture. The additive was dosed directly
ageing the ageing index was calculated following to the mixing device with aggregate and other
the same principle like for the previous test. component.
The next additive was an inorganic Poly Phos-
phoric Acid (PPA). It was used for improving per-
formance of the bitumen, which was afterwards
3 RESULTS OF EXPERIMENTAL TESTS
used for mixture manufacture. PPA has an effect
on increased binder consistency, which usually
3.1 Asphalt mixture variants
leads to increased stiffness and resistance to per-
For this research stone mastic asphalt mixture manent deformation. Simultaneously this modifier
with maximum particle size of 11 mm (SMA 11S) improves creep characteristics as well (i.e. the flow
was designed. The mixture was manufactured number as tested e.g. in the U.S.).
with crushed spilit aggregate from quarry Litice, The alfa-polyolefin based additive (further
limestone filler from Velke Hydcice and different labelled as “VP”) is like Innodur a substance based
types of bitumen and selected additives. Most of on a kind of synthetic wax. The additive should
the mixtures were manufactured with 5.8 M% of improve adhesion of bitumen to aggregate and
bitumen according to valid ITT protocol, except of increase the strength characteristics. The additive
mixtures with 30% of RAP and pulverized rubber. was firstly dosed to the bitumen, which was sub-
According to the standard EN 13108-5 it is in sequently used for mixture manufacture. In one of
the Czech Republic not possible to use any RAP the variants VP was combined with regularly used
in SMA mixtures. For the research purpose 10% SBS modifier.
and 30% of RAP was used. In mixture with 10% Last used modification was an activated pulver-
RAP 5.8 M% bitumen was kept as in referential ized rubber of a particle size 0.0–0.8 mm added to
mixture—the surplus of bitumen because of bitu- the bitumen by 10 M% of the binder. In the asphalt
men in RAP was not considered. In mixture with mix variant where the CRMB binder was used, no
30% RAP the amount of bitumen was reduced to cellulose fibres were applied, since it was believed
4.6 M% – it was expected that some of the bitumen that the pulverized rubber might be able to substi-
in RAP will be re-activated/re-mobilized during tute the properties of fibres.
the heating and mixing process. In mixture with The tested variants with used additives and fib-
pulverized Rubber Modified Bitumen (CRMB) ers are summarized in Table 1.
the amount of bitumen was increased by 0.5 M% For better understanding and overview the
to 6.3 M% (for referential mix). results were divided into 5 testing groups accord-
The used RAP material was crushed and ing to the used materials:
screened to a grading of 0–11 mm. By repeated
number of extraction it was determined that the 1. mixtures with non-modified bitumen;
content of bitumen in RAP was 4.8 M%. The gran- 2. mixtures with refinery-modified bitumen;
ularity of the mixtures with RAP was designed as 3. mixtures with 10% RAP;
close to the reference mixture as possible. 4. mixtures with 30% RAP;
Besides RAP many of the additives and modi- 5. mixtures with modifying additives and mixtures
fiers were used to determine their influence on low manufactured with bitumen modified addition-
temperature characteristics. ally in the lab.
The very first additive were cellulose fibres. The
cellulose fibers are regularly used in SMA mix-
3.2 Volumetric characteristics
tures to prevent bitumen drainage. With respect
to the character of this asphalt mixture (discon- As it was described higher, the Marshall test
tinuous granularity) the binder is predisposed to specimens for SCB test were manufactured in
sag and the mixture can simply become inhomo- compliance with EN 12697-30. The volumetric
geneous. In this case two different cellulose fibers characteristics were compared with the required

187
Table 1. Assessed asphalt mix variants.

Virgin Additives
bitumen S-CEL S-CEL R RAP [% of bitumen]

ref I referential 5.8 M% 0.3 M% – – –

ref II referential 5.8 M% – 0.3 M% – –
PmB I PmB 25/55-55 5.8 M% 0.3 M% – –
PmB II PmB 25/55-55 5.8 M% – 0.3 M% – –
PmB III Polybit. 45 5.8 M% – 0.3 M% – –
10RA_PmB Polybit. 45 10% RA 5.8 M% – 0.3 M% 10 M% –
10RA_2INN 2% INN + 10% RA 5.8 M% – 0.3 M% 10 M% 2% Innodur
10RA_4INN 4% INN + 10% RA 5.8 M% – 0.3 M% 10 M% 4% Innodur
30RA_PmB Polybit. 45 30% RA 4.6 M% – 0.3 M% 30 M% –
30RA_2INN 2% INN + 30% RA 4.6 M% – 0.3 M% 30 M% 2% Innodur
30RA_4INN 4% INN + 30% RA 4.6 M% – 0.3 M% 30 M% 4% Innodur
2INN 2% INN 5.8 M% 0.3 M% – – 2% Innodur
4INN 4% INN 5.8 M% 0.3 M% – – 4% Innodur
PPA I PPA 5.8 M% – 0.3 M% – 1% PPA
PPA II 70/100 + PPA 5.8 M% – 0.3 M% – 1% PPA
5VP 5% VP 5.8 M% 0.3 M% – – 5% Vestoplast
6VP 6% VP 5.8 M% 0.3 M% – – 6% Vestoplast
VP+SBS 3% SBS a 1% VP 5.8 M% 0.3 M% – – 1% Vestoplast 3% SBS
10CR 70/100 + 10% CR 6.3 M% – – – 10% pulverized rubber

limits given in the standard EN 13108–5. Allowed not the most important parameter. SMA mixtures
limits of air void content according to this stand- must more likely indicate good resistance to rut-
ard are for initial type testing ITT very narrow ting and fatigue.
(3.0–4.5%) and more than third of the experimen- This test is usually not performed for traditional
tally tested variants did not reach this limit. Lim- AC or SMA mixtures and therefore there isn’t any
its for control testing are wider (2.0–6.0%) and all specified or recommended value of this param-
variants (except one) were within these limits. eter. For sake of this research, 6 MPa according
All mixtures with 10% RAP have an air voids to TP 151 for HMAC mixtures was chosen as ref-
content lower than it was expected. It can be erential value, even though the flexural strength is
caused either by improper grading curve design or not expected as high as for high modulus asphalt
more likely by higher bitumen content in the mix. mixtures. It is necessary to point out that 6 MPa is
In the mixtures with 10% RAP, as already men- value used only for comparing the particular SMA
tioned, there was an identical content of 5.8 M% variants, it is not a required minimum limit.
bitumen like in the referential mixture. RAP was For better orientation the results were divided
added to mixing device preheated so there was a re- into two figures, both with delineating of referen-
activation of the aged bitumen in RAP. It can lead tial mixtures.
to a slightly increased bitumen content and there- Strength characteristics increased due to artifi-
fore lower voids content. The variant 10RA_2INN cial ageing as it was expected. The specimens were
has voids content of only 1.3%-vol., which might exposed to flow of hot air what resulted in oxida-
have negative impact on the test results. tion/hardening of the bitumen. This fact leads to
an increase of maximum force which is necessary
for the rupture of the test specimen. Generally it is
3.3 Tensile bending strength (Three-point
possible to express that the higher ageing index is,
bending)
the higher would be the predisposition of a bitu-
As it was mentioned in part 2.1, this test is pro- men to age and therefore also to collapse due to
ceeded by a method given in TP 151. The techni- the impact of changing weather conditions and
cal specifications determine the minimum value repeated traffic loading. On the other hand the
of flexural strength, but only for HMAC mix- closer is this parameter to 100%, the more might be
tures. The HMAC mixtures are used in base or the bitumen susceptible to ageing and degradation.
binder layers where high strength characteristics For some variant there was even decrease in age-
are required. For SMA mixtures (used in wear- ing index. It is possible that specimens were dam-
ing course) the high strength characteristics are aged during the ageing or there were changes in

188
their structural cohesion. Mixture 30RA_2INN was .... 9,5 9,0
apparently damaged, even thought it was not vis- Q.
~ 8,5 7,9
ible at first sight. Decrease along 10% can be cause g. 7, 7.
7.
7,
~ 7,5 7,1 7,0
by used statistic calculation method. This issue is o;
still being verified with usage of higher number of !"' 6,5
test specimens used in one tested series. In the case ~ 5.5
of test results presented in this paper minimum 3 ~
~ 4,5
beams were used for determination of one result.
Beams with application of 10% RAP evince the
most stable results of flexural strength. The results
1,02 1.26 1,09 1.25 1,09 1,05 0,92 0,94 1.20 0,69 1.23
of aged specimens are less influenced by hardening
of the bitumen, which leads to an assumption that 9.5
the bitumen is more resistant to ageing and the age- ....
~ 8.5
ing process will be slower leading to a slightly lower .t::
g. 7,5 7,1 7.1
degradation of the asphalt mixture. It is necessary ~

to point out, that already aged bitumen reactivated ~ 6,5

~
from RAP will aged in a new mix more slowly than ~ 5,5
is expected for virgin bitumen. This assumption can ~ 4,5
be made since it is expected that the bitumen from ~ ,;;.'
RAP is already partly aged so the process of fur-
1,02 1.26 1,30 0,94 1,06 0,92 1,34 1.26 1.24 1,35
ther ageing is less vigorous. For this reason, asphalt
mixtures containing RAP might be more durable, • unaoed u aaed

but this assumption needs another deeper analysis Figure 2. a, b Tensile bending strength (MPa) of SMA
especially from the point of view of assessing addi- 11S.
tional asphalt mixtures after their ageing and even
after various periods of ageing.
The highest strength values are reached by A summary of all experimental results related
the SMA mixtures with application of refinery- to tensile bending strength, flexural modulus and
modified bitumen and RAP. On the other hand, calculated fracture energy is given by the Table 2.
mixtures modified subsequently in the laboratory Very important but often neglected parameter is
reached values even lower than referential mix with fracture energy—energy needed for a specimen to
a standard paving grade bitumen 50/70. Applica- fail. This value comes from force-deformation dia-
tion of 4% of Innodur seems to be too high and gram and for its determination it is necessary to get
therefore superfluous. Mixtures with application data about exact time-dependent development of
of only 2% Innodur reach better results. It might force and deformation during the test. From this
be worth for further consideration about applica- parameter it can be seen that even though the spec-
tion of such additive to further test mixtures with imen after artificial ageing has usually higher flexu-
dosing between 2% and 4% Innodur. Nevertheless, ral strength (maximum force), the fracture energy
in this group of mixtures and their results there was necessary for specimen to be broken is lower. And
a substantial increase of the strength characteris- so the aged specimen will mechanically fail sooner
tics after ageing. This leads to a clear assumption than a test specimen with lower flexural strength
that these asphalt mixtures will be more sensitive but higher total fracture energy. In general and
to ageing processes and they can potentially have a even in the technical specifications TP 151 frac-
shorter lifetime. ture energy is almost always disregarded and the
asphalt mixtures are compared only from the point
4,5 of maximum strength values.
4,5% ---.r--3-,9------~4.-0:---- 4 ' 2 - 4·1 4 0 On the Fig. 3 there is an example of force-defor-
3,7 '
4,0%
3,5 3,7 3,7 mation diagram. The mixture with application of
~ 35% 3,4
~ . 2.8 2,9
3.1 3,0
5% of VP shows that in the case of aged specimen
~ 3,0o/o
2,5 :l,b it is necessary to apply higher loading and the col-
1; 2,5% lapse of such mixture will come at a lower defor-
~ 2,0%
1.5%
-1-12,3-
1,3

• I mation by usage of lower energy.

3.4 Resistance to crack propagation

The test method was done in compliance with
the requirements given by EN 12697-44. The test
Figure 1. Voids content (%) of assessed SMA 11S. method is in Europe relatively new, so there are

189
Table 2. Tensile bending beam test results of SMA 11S mixtures.

Flexural strength Flexural modulus Fracture energy

[MPa] [MPa] [kJ]
Ageing Ageing Ageing
ref. aged index ref. aged index ref. aged index

ref I 7.1 7.2 1.02 1440 1538 1.07 1.3 1.3 1.01
ref II 5.2 6.5 1.26 868 1289 1.48 1.5 1.1 0.74
PmB I 7.2 7.9 1.09 898 1141 1.27 2.3 2.2 0.98
PmB II 7.3 9.1 1.25 1094 1520 1.39 2.1 1.8 0.85
PmB III 6.0 6.5 1.09 912 1223 1.34 1.9 1.6 0.86
10RA_PmB 7.6 8.0 1.05 1280 1333 1.04 1.7 1.8 1.06
10RA_2INN 7.9 7.2 0.92 983 1176 1.20 1.9 1.3 0.67
10RA_4INN 7.0 6.5 0.94 1245 1367 1.10 1.2 1.1 0.92
30RA_PmB 7.3 8.8 1.20 1207 1556 1.29 1.7 1.8 1.07
30RA_2INN 9.0 6.2 0.69 1386 1475 1.06 1.1 1.5 1.35
30RA_4INN 5.6 6.9 1.23 1554 1952 1.26 0.8 0.7 0.90
2INN 5.2 6.8 1.30 890 1498 1.68 1.1 1.1 1.01
4INN 5.4 5.1 0.94 1038 1178 1.14 1.1 0.8 0.74
PPA I 6.1 6.4 1.06 678 1107 1.63 2.1 1.5 0.71
PPA II 7.1 6.6 0.92 1050 1106 1.05 1.7 1.6 0.92
5VP 5.5 7.4 1.34 971 1466 1.51 1.5 1.4 0.92
6VP 5.6 7.0 1.26 1190 1523 1.28 1.0 1.1 1.10
VP+SBS 5.4 6.6 1.24 969 1472 1.52 1.2 1.2 1.01
10CR 5.4 7.3 1.35 684 892 1.30 1.8 1.7 0.95

For regular asphalt mixtures there is an assump-

tion that the fracture toughness will increase with
1,5
lowering of the test temperature—the lower the
~ testing temperature is, the higher the toughness
j ---+- 5VP- ref. might be. All presented mixtures are supporting this
0,5
- - - SVP - aged assumption with exception of the SMA variant Pm
0 B I. Another “exception” are mixtures containing
0 0,5 1,5 RAP. For these asphalt mixtures this assumption is
Deformation [mm)
usually exactly opposite—the lower the testing tem-
perature is, the lower is the fracture toughness.
Figure 3. Force-deformation diagram of SMA 11S
with 5% of Vestoplast. Fig. 4 shows results of asphalt mixtures which
were tested at 0°C and −10°C. Highlighted is the
value of 43 N*mm-3/2, which is not a limit value but
not any given or recommended limits for fracture only a value which was reached for the referential
toughness either for AC or for SMA mixtures so far. mix used for comparison of all other variants.
This applies not only to the standard but especially Again mixtures modified additionally in the
in Central Europe also to any other specifications or laboratory evince worse results than referential
regulation—as same as for flexural strength, which mixture with 50/70. Surprisingly even mixtures
is however more specific and in the past was used with refinery-modified bitumen did not reach the
only in the Czech Republic or partly in Germany. expected values. Very good results are again shown
The specimens were manufactured and tested with application of 10% of RAP. This combination
during longer period of time under more complex and this level of RAP seems to be convenient not
research, so not all variants were tested at the same only from the point of economic and environment
parameters. For the first group of test specimens view but also from the view of tested characteristics.
only temperature of 0°C was applied. In case of sec- Marshall test specimens with application of
ond group of tested asphalt mixtures temperature VP additive were artificially aged and then cut to
of −10°C was added and tested. If looking on the semi-cylindrical specimens. Assumption for this
last group of tested specimens, there the evaluation test method was completely opposite to flexural
was extended by test specimens subjected to artificial strength. The assumption was that the fracture
ageing according to pr EN 12697-52 (see part 2.1). toughness will be influenced negatively by the

190
 so 49
48 47 SMA mixture with application of 6% VP. On the
if.E 4S 44
contrary application of only 1% lower content of
43
~
···41· ·~r·········4f····
42 VP caused noticeable decrease of the results.
The results of maximum fracture stress which

r _I I _I__
39
fo can be calculated for the maximum force at speci-
i 35 men rupture have the same trend as fracture tough-
~
30
ness, therefore they are not further presented in this
~' ~.'' paper.

1.09 1.08 0.85 1.11 X X 0.96 0.90 X 0.85 0.97

so 4 COMPARISON OF TESTING METHODS

fE 4S
. .... . . . "- --~· · ·~·· .. i""ll
46 46

~ On the beginning of this research there was an

assumption of certain correlation between val-
ues determined from tensile bending strength and
resistance to crack propagation test and its charac-
teristics. This correlation was not found and could
~~~ <~;.~~ f.~"' f.~"'' <! .,~ ~·4>"' '"'"~ not be proven. Both of the test methods are reli-
Rat/g: 1.09 1.08 1.12 1.02 X X 0.8S 1.02 0.93 0.87 ant on precise preparation of test specimens and
adequate conditioning at the low test temperatures.
Specimens should be always stored in a suitable air-
Figure 4. a, b Fracture toughness of SMA 11S conditioned box with regulation of temperature for
(N*mm−3/2). sufficient amount of time according to specimen’s
volume. In this research all specimens were store
ref. aged always for at least 10 hours at the given temperature.
50 The preparation of a test specimen (cutting) can
cause a damage, which does not have to be apparent
48 46 at the first sight, but the result will be influenced.
The semi-cylindrical specimens are notched and it
45 is important that this notch is located directly in
the middle of the base of the semi-cylindrical test
43 41
specimen and has the given proportions, including
40 its depth. The notch also have to be horizontal with
edges of the test specimen.
38 Further, from Fig. 6 and Fig. 7 it is evident, that
low temperature characteristics do not have any sta-
35 tistically apparent correlation between each other. It
is not even possible to say that if an asphalt mixture
33 demonstrates good results by one method, that it
would reach good results by the other method as
30 well. For this reason Fig. 8 was drawn. This figure
f~JC:>
# # .§X- ..,.#" 9,5 A ref I
Brefll
9,0 C PmB I
DPmBII
AI• I'll 0,85 1.02 0;93 0,87 - 8,5
,. •o:
i 8,0 D
G
E PmBIII
F lORA_PmB
G 10RA_21NN
−3/2 .£ 7,5 F H lORA 41NN
Figure 5. Fracture toughness of SMA 11S [N*mm ].
""ai 7,0 0
A
c 130RA_PmB
t:
~ 6,5
N J 30RA_21NN
K 30RA 41NN
ageing process—thus the results of aged specimens L21NN
will be decreased. Semi-cylindrical specimens are ~ 6,0 E K M41NN
u::: 5,5 P ~ RM N PPAI
notched and at low temperatures fragile fracture Q
LB 0 PPAII
appears after the test specimen is loaded. Aged bitu- 5,0 P SVP
men has to a certain extend higher strength charac- 4,5 Q6VP
30 35 40 45 SO RVP+SBS
teristics but it is more fragile and more liable during Fracture t oughness O"C {N/mm3/2) S lOCR
low temperatures.
This assumption was verified by all tested mix- Figure 6. Comparison of flexural strength and fracture
tures. The most stable results were reached for a toughness at 0°C for un aged test specimens.

191
 shows comparative ratio of referential mixture determined by EN 12697–26, method C (IT-CY),
ref I to tested mixture—by how many percent the at test temperature of +5°C. This test method was
characteristic differs from the referential value. The not part of the performed research, but the results
chart curves, given in this figure, should have similar could help to determine the correctness of the
trend, nevertheless result of some mixtures are scat- trend the results might show.
tered on opposite side of the figure.
In the Fig. 9 comparative ratio from stiffness
modulus was added. The stiffness modulus was 5 CONCLUSION

9,5 A ref I It is not possible resolutely mark out which test

B ref II
9,0 D C PmBI
method is the correct and most appropriate
- 8,5
OPmBII one. Both of the compared test methods were
i 8,0 c
E PmB III
F lORA_PmB
G 10RA_21NN
precisely done according to the requirements
given in standards, but still their results are not
.£ 7,5 H lORA 41NN
"'c 7,0 p SA G 130RA_PmB
comparable.
Q
"
!;: Kt_
R
J30RA_21NN Primary the crack propagation test is very
6,5 N 0 E B K30RA 41NN
~
J L21NN
dependent on the precise preparation of test speci-
~ 6,0
M41NN mens. In order to objectively determine which
u:: 5,5 N PPAI
OPPAII
test method provides meaningful results in point
5,0 M of quality assessment, it would be necessary to
PSVP
4,5 Q6VP
RVP+SBS
extend the research by influence of test specimen
30 35 40 45 50
Fracture toughness ooc (N/mm3/2) S lOCR preparation.

Figure 7. Comparison of flexural strength and fracture

toughness at 0°C for aged test specimens. ACKNOWLEDGEMENT

This paper was prepared as part of the activities

of the research project supported by the Technol-
£:20% ogy Agency of the Czech Republic, project no.
~ 10% TA04031328.
,;:
-o '"'
]
~ ·10* REFERENCES
;;_ -20%

EN 12697-44, Bituminous mixtures—Test methods for

hot mix asphalt—Part 44: Crack propagation by semi-
circular bending test, 2011.
EN 13108–5, Bituminous mixtures—Material
specifications—Part 5: Stone Mastic Asphalt, 2008.
prEN 12697-52, Bituminous mixtures—Test methods—
Figure 8. Proportional difference for selected charac- Part 52: Conditioning to address oxidative ageing,
teristics of tested and referential (ref I) mixture. standard in development.
TP 151, Technical specifications for high-stiffness modu-
30%
lus asphalt mixtures (HMCA), 2010, (available only in
Czech).
20% Vacková et al., Combined effect of new type of cel-
~
lulose fibers and reclaimed asphalt on perform-
~ 10%
•
,;:
ance characteristics of stone mastic asphalt mix,
-o Insights and Innovations in Structural Engineer-
""
I
ing, Mechanics and Computation. 2016, ISBN
·10%
978-1-138-02927-9.
Valentin et al., Asfaltové směsi typu sma s použitím
"'"" různých druhů celulózového vlákna a R-materiálu,
CESTI competence centre—Partial Research Report,
CTU Prague, 2015, (available only in Czech).
Valentin et al., Stone mastic asphalt mixtures with use of
organic additive Vestoplast, Research report, 2016.

Figure 9. Proportional difference for selected charac-

teristics of tested and referential (ref I) mixture.

192
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Computational study of the influence of form and angularity of coarse

aggregates in the linear viscoelastic properties of asphalt mixtures

D. Castillo & S. Caro

Universidad de los Andes, Bogotá, Colombia

ABSTRACT: Aggregates account for most of the mixture’s weight and volume and, consequently, they
play a key role in defining the overall mechanical response of these materials. In asphalt mixtures, the
aggregate phase is composed of thousands of different discrete aggregates, each having a unique set of
morphological properties that strongly contribute to the heterogeneity of the material. The objective of
this work is to use a computational approach to study the isolated impact of the form and angularity
of coarse aggregates on the linear viscoelastic material properties of an asphalt mixture, i.e. their axial
Dynamic modulus (|E*|) and Phase angle (δ). To accomplish this goal, a recently developed Random
Microstructure Generator was adapted and used to computationally produce two-dimensional micro-
structures of asphalt mixtures with coarse aggregates having controlled average values of Angularity and
Form indexes (AI and FI, respectively). These microstructures were then implemented in Finite Elements
(FE) and subjected to controlled—stress cyclic loading to determine |E*| and δ. In summary, the results
show that variations in the morphology of the aggregates affect the overall contribution of the linear vis-
coelastic properties of the mixture, and they also impact the final variability of these parameters.

1 INTRODUCTION have also been conducted by other researchers with

promising results in both two and three dimensions
The application of computational mechanics to (Papagiannakis et al. 2002; You et al. 2009; Coleri
model the microstructure of Hot Mix Asphalt et al. 2012; Ying et al. 2014; Shakiba et al. 2015).
(HMA) in two and three dimensions (2D and 3D) When analysing these existing works on the
has been extensively explored in recent years. The topic, it is observed that there are several pos-
increasing interest in these numerical tools vali- sibilities regarding both the specific phases to
dates the idea that the characteristics of the micro- be considered in the numerical model and the
structure (i.e. the geometrical disposition of its constitutive formulation used to represent their
components: aggregates, fine aggregate matrix and response. Microstructural models in asphalt mix-
air voids, and their corresponding constitutive for- tures consider the material to be composed of
mulations) could be used to predict the mechanical three or two phases. In the first case, the model
response of HMA specimens when subjected to dif- includes aggregates, air voids and mastic or fine
ferent types of loading, without performing labo- aggregate matrix (i.e. a mixture of binder, air voids
ratory tests with whole specimens. From this point and fine aggregates, usually smaller than 1.18 mm),
of view, it is expected that computational software as independent entities (You et al. 2009; Castillo et
will work in the future as a ‘virtual laboratory’. al. 2015; Ying et al. 2014). In the second, which is
A good example of the capabilities of computa- the most common approach, the mixture is simpli-
tional models is the work conducted by Dai and You fied as the combination of two individual phases:
(2008). This study follows and summarizes what is coarse aggregates and mastic or a fine aggregate
now a common approach to modelling HMA, spe- matrix (Papagiannakis et al. 2002; Caro et al. 2016;
cifically by combining image acquisition techniques Kim et al. 2005; Aragão et al. 2011). Usually, the
and Finite Element (FE) models. Here, the authors decision on the number of phases and the level of
used an image of an asphalt mixture, separated detail in their geometrical representation depends
the coarse aggregate phase, and then implemented both on the available computational power and
this phase into FE along with the mastic portion on experimental characterization data, which is
of the mixture. The two-dimensional computa- required to count with reliable constitutive formu-
tional specimen was subjected to cyclic loading to lations for each constituent. Since the focus in the
calculate the complex modulus of the mixture and geometry of the microstructure is herein given to
to perform creep compliance tests. Similar efforts the shape properties of the coarse aggregates, in

193
this work the mixture is modelled as a simplified and to determine the isolated influence AI and FI
two-phase material. on the resulting properties of the mixtures.
The objective of this study is to use compu-
tational mechanics to evaluate the impact of the
morphological properties of the coarse aggregate 3 RANDOM GENERATION OF HMA
phase on the dynamic linear viscoelastic properties MICROSTRUCTURES
of an asphalt mixture. This is an important task for
the pavement engineering community, considering 3.1 Process of generation of the aggregates
that even though aggregates typically compose
more than 85% of the total weight of a mixture, The aggregate portion hasa Nominal Maximum
the impact of their shape properties has rarely Aggregate Sizeof 12.5 mm, and the gradation cor-
been studied using computational techniques. responds to that reported by Masad et al. (2006).
To isolate the effect of the morphology of the
aggregates, the gradation and the proportion of
the total coarse aggregate area with respect to the
2 AGGREGATE SHAPE
fine aggregate matrix area were kept constant dur-
CHARACTERIZATION
ing the random generation of the microstructures
of the HMA, while the shape properties of the
The morphological properties of aggregates are
coarse aggregate particles were modified.
usually characterized using three shape param-
The random microstructures used in this study
eters: angularity, form and texture (Al-Rousan
were created through a two-step process using an
et al. 2005). In this study, the following Angularity
improved version of a numerical tool developed
and Form indexes (AI and FI, respectively), were
by Castillo et al. (2015): (1) the generation of a
selected to characterize these two shape proper-
random set of two-dimensional closed shapes that
ties of aggregate particles (Masad et al. 2001; Al-
represent coarse aggregate particles, whose shape
Rousan et al. 2005):
parameters and size are fully controlled, and (2)
the random location of the aggregates inside a geo-
1 N −1
AI = ∑ Gi − Gi +1
N − 1 i =1
(1) metric domain, which in this case corresponds to a
square of 20 by 20 cm.
The result of this process is a 2D geometrical
θ − Rθ
θ =360 − θ
Rθ +
FI = ∑
θ =0 Rθ
(2) representation of the microstructure of the HMA,
composed by a set of coarse aggregates embedded
within the fine aggregate matrix of the mixture.
Briefly stated, Equation 1 calculates the angu- Typically speaking, each computational HMA
larity of an aggregate (AI) as an accumulation microstructure has between 275 and 575 coarse
of the changes in the ‘direction’ of the edges of aggregate particles (depending on their specific
the particle. The edges are divided into N seg- shape properties), as discussed in the next section.
ments, with Gi being the gradient of i-th segment
in radians. Meanwhile, Equation 2 computes the
3.2 Characterization and control of AI and FI
form (FI) as a summation of the relative changes
in the distances between the centre of the particle To control the resulting values of AI and FI of an
and points located on its perimeter; the points are aggregate sample (i.e. the aggregate fraction of one
located at even angles (Δθ) from the centre, and random microstructure), three geometric proper-
Rθ is the distance from the centre to a point on ties of every individual particle were considered:
an edge. (1) its minimum and maximum probable number
The methodology used to reach the objectives of vertices, (2) its total elongation, and (3) its mini-
of this study consists of the following stages: (1) mum and maximum probable ‘radius’, i.e. the dis-
the generation of random microstructures of an tance between the centre and each of the vertices
asphalt mixture with aggregates having different of the aggregate. The average AI or FI values of all
morphological properties, using a computational aggregates in a specimen were used to characterize
microstructure generator, (2) the implementation its overall AI or FI.
and modelling of these microstructures in FE; in Five different values of AI and FI were targeted
the models, the fine aggregate matrix is considered a for the microstructures. These AI and FI values were
linear viscoelastic material and the specimen is sub- selected to range between what are considered ‘typi-
jected to a cyclical controlled-stress loading, and (3) cal’ values for coarse aggregates in actual mixtures
the evaluation of the mechanical results obtained (AI between 0.07 and 0.13 and FI between 2 and 3)
from the computational simulations to estimate the to some other ‘extreme’ values (i.e. very high and very
dynamic modulus and phase angle of the specimens, low FI and AI).This implies that certain random

194
microstructures produced with this methodology of AI of 0.07and variable values of FI (from
will result in sets of coarse aggregates with angu- 1.24 to 3.5), and
larity and form properties that are not expected to – Five sets of HMA microstructures (30 per set)
occur in reality. However, these cases were included with coarse aggregates having a constant value
in the analysis to evaluate the potential impact of a of FI of 3.0 and variable values of AI (0.068 and
wide range of possible aggregate shape properties on 0.15).
the viscoelastic properties of asphalt materials.
To capture the random nature of the aggregate
Through careful calibration, different sets of
shapes, a total of 30 microstructures were gener-
these parameters were selected to create 10 cases
ated and evaluated in FE for each case of study
of study:
(i.e. a case is composed of 30 microstructures with
– Five sets of HMA microstructures (30 per set) aggregates having common average values of AI
with coarse aggregates having a constant value and FI). Since 10 different cases with combinations
of AI and FI were considered, this means that a
total of 300 HMA microstructures were generated
(a) AI = 0.07, fixed
and later used in the FE models.
Notice that these combinations enable the study
of the isolated effect of each of the two morpho-
logical properties of the aggregates on the mechan-
ical properties of the mixture; a task that is very
difficult or even impossible to perform in the labo-
ratory. Figure 1 illustrates some of the microstruc-
tures that were randomly generated as part of this
study. Besides, a summary of the average AI and
FI values of all the computational samples in the
10 cases is presented in Figure 2.

(a) 4

3.5
Fixed AI, • 111_,.,., • • • Case 5
variable FI
u: 3
x·
,...._,. Case4
:s"' 2.5
,....,.... Case3
E
& 2 •• ..._.,._ Case 2

1.5
~ Case 1

0.062 0.064 0.066 0.068 0.07 0.072 0.074 0.076

Angularity Index, AI

(b) 3.3
Fixed Fl,
3.2
va riable AI
i:i::: 3.1
x·

.s"'
'0
3
.. .;:.·' .;;:
J!: .;)~
...
...·M'
~
E
2.9
J Case 2
Case 3
Case 5
;;;: Case 1 Case 4
..c
-~
2.8
~
u: 2.7

~ !,~~~~~~~ 0.06 0.08 0.1

Angularity Index, AI
0.12 0.14 0.16

"'~
u"' Figure 2. Individual AI and FI of all randomly gener-
ated cases, (a) fixed AI, (b) fixed FI. A point represents
Figure 1. Microstructures with (a) fixed AI and increas- the average of the AI and FI values of all aggregates in a
ing FI, (b) fixed FI and increasing AI. computational specimen.

195
4 FINITE ELEMENT MODELING AND (a) 0.95 0.96 0.97 0.98 0.99
BOUNDARY CONDITIONS 0 # . .............. . - -. . .-'- -!

Fixed AI, // ....

-~ -0.2
The previously created computational HMA :<> variab le FI /
~
microstructures were modelled in FE using the
commercial software Abaqus®. The linear viscoe- "<:: -0.4
"'
~
~

lastic mechanical behaviour of the fine aggregate i!:' -0.6 a norm l,./ enorm
~
phase of the material was characterized by means /

of a Prony series calculated from data reported

".~ -0.8
-- - - Case 1 (low Fl)
- - Case2
E
------------/--
r·
for these materials (Perico 2014). The Poisson’s 0 - - Case3
ratio was 0.35, and, at the reference temperature z -1 - - Case4
- - Case 5 (high Fl)
(25°C), the FAM material presented an instanta- -1.2
neous modulus of 5142.6 MPa. Coarse aggregates
were assumed to be linear elastic materials with a (b) 0.95 0.96 0.97 0.98 0.99
Young’s modulus of 25 GPa and a Poisson’s ratio 0 ____________________,
of 0.25 (Papagiannakis et al. 2002). Both phases Fixed Fl,
-~ -0.2
were modelled using type CPE3 triangular ele- :<> va riable AI /
~
ments and a mesh size of 1.5 mm.
A 400 kPa stress-controlled condition was "~ -0.4 /

~
~

applied by means of haversine wave (10 Hz) ~ -0.6 a norm l/" Enorm
through a rigid plate on top of the specimens, dur- /

ing 1 second (10 cycles). Vertical displacement was

".~ -0.8 /
/
- - Case 1 (low AI)

restricted from the lower boundary of the speci- ~ _....................

Case 2
- - Case3
0
mens, as well as horizontal displacement for the z -1
- - Case4
leftmost node on this boundary. The rigid plate - - Case 5 (high AI)
-1.2
was also restricted from rotation and horizontal
displacement. Figure 3. Normalized strain signals (average per case)
during the last half-cycle of the simulations. (a) Fixed AI,
(b) fixed FI.
5 RESULTS AND DISCUSSION
(a) 10000

The primary result of the simulations was the dis- Fixed AI,
9500
placement at the top of the specimens caused by Case 1
variable FI
the applied stress. This displacement, which rep- 9000

resents the overall response of the microstructure,

was converted to strain and used together with
'"6
0..
8500 Cas~ 2
·-----------'··-4 Ca; e 3 Cas; 4 Case 5
1:u
the input stress to calculate the dynamic modulus 8000 'f.'• ... , _____ ,& ~~~
(E*) and the phase angle (δ) of the asphalt mix- .. z,-------------1·
ture specimens. For all samples, E* was calculated
7500 ~· ~.
as the ratio of the stress to the strain amplitudes. 7000
To calculate δ, a haversine cosine function was 1.5 2.5 3.5
superimposed to the output force signal, and it was Form Index (FI), average
adjusted using SOLVER until the error between
the fitting curve and the extracted data was mini- (b) 10000
mized. The lag between the input stress signal and
the output strain signal, or δ, was then calculated. 9500
Fixed Fl, ..
The behaviour between cases can be appreciated in 9000
variable AI . ·-;.··4.i
."J:;------
Figure 3, which presents the average normalized ~
'"6 •• , .,,... '1

... ·;·
0..

strain response of half-cycle for the 30 computa- 8500 '·L------~'

.. Case4
Case 5
tional microstructures of each case of study. 1:u
8000 j----------
Case 3
7500 Case 2
5.1 Dynamic modulus (E*) Case 1
7000 ~
Figure 4 presents a summary of the E* values cal- 0.05 0.07 0.09 0.11 0.13 0.15
culated for both AI and FI cases, together with Angularity Index (AI), average
a curve connecting the average value among the
30 simulations conducted in each case. It can be Figure 4. Calculated value of dynamic modulus (E*)
observed right away that the dynamic modulus for all cases. (a) Fixed AI, (b) fixed FI.

196
decreased for high values of FI, at fixed AI, and for (a) 22
low values of AI, at fixed FI. In fact, E* presented Fixed AI,
a 13.5% reduction between the cases with the low-
est and the highest FI, i.e. Cases 1 and 5 (fixed AI).
20

18
variable FI
j--------------~~...,
This reduction was of 4% between Cases 2 and 4, ~-----l------ ~ Cas; 5
c
which can be considered more ‘typical’ representa- <0
16
.................
-- ,""
. ...
Case 3
Case4

tions of actual mixtures. On the other hand, the 14 Case 2

average E* increased 13.8% between the cases
with lowest and highest AI at fixed FI, and 7.0% 12

between the ‘typical’ cases 2 and 4. 10

·----
Case 1

At constant AI, an increase of 2.8 times in FI 1.5 2.5 3.5 4

(between Cases 1 and 5) caused E* to decrease by Form Index (FI), average
13.5%, while at constant FI, an increase of 2.0 times
in AI (again between Cases 1 and 5) generated an (b) 22
increase in E* of 13.8%. Calculating the normal- Fixed FI,
ized effect of incrementing the shape parameters, it 20
variable AI
is found that increasing AI increases E* around 1.4 18
times more than what it decreases due to the same Case 1
percentage change in FI. In summary, variations in c
<0
16 !------------- ....
angularity have a greater influence over E*. Never- Case 2 -------- ~
theless, it is worth mentioning that for those micro-
14
Case 3
-~--
•
~~
... ·•
structures with AI and FI values representing the 12 Cas: •4 \ --- ~~;
shape of aggregates more likely to be found on HMA Case 5
10
mixtures (cases 2 to 4 for both fixed AI and fixed AI 0.05 O.D7 0.09 0.11 0.13 0.15
conditions), the results suggest a mild influence of the
Angularity Index (AI), average
morphological properties of the coarse aggregates on
the linear viscoelastic properties of the mixture. It is
Figure 5. Calculated value of phase angle (δ) for all
noteworthy, however, that this phase can have a more
cases. (a) Fixed AI, (b) fixed FI.
relevant influence in the development of mechanical
degradation processes within the mixture.

resulting viscoelastic properties. Such an analysis

5.2 Phase angle (δ)
is possible due to the random nature of the proc-
Figure 5 presents the results of the phase angle ess used for the generation of the microstructures
obtained for all the cases under consideration. In (i.e. every HMA microstructure is unique), which
the cases with fixed AI, the phase angle increased causes that the resulting values of E* and δ are dif-
with an increase of FI up to 47.0% between the ferent among numerical replicates, even within the
extreme cases (Cases 1 and 5) and 11.3% between same case of study, as observed in Figures 4 and 5.
the typical cases (Cases 2 and 4). This means an The dispersion in the behaviour of E* and
increment in the ‘viscous’ behaviour of the mate- δvaries between cases, as observed in the previous
rial due to increasing FI. On the other hand, a figures. It should be acknowledged, however, that
reduction of 28.2% between the extreme cases and part of the observed variability in the resulting vis-
of 15.7% between the typical cases ensued with the coelastic properties comes from the ‘input’ varia-
increase in AI, suggesting that more angular aggre- bility of the microstructures. This is the variability
gates contributed to a more ‘elastic’ behaviour of among the values of FI of the 100 specimens (fixed
the mixture. AI cases) and the values of AI of the 100 specimens
Currently, there are no clear hypotheses to (fixed AI cases). To account for this, the Coeffi-
explain the impact of the shape parameters on the cients of Variation (CV) of E* and δ (Figure 6a)
phase angle of the material. However, the results were normalized with their corresponding ‘input’
do suggest that variations in the morphology of CV of the AI or FI of the original microstructures
the aggregates affect the overall contribution of the (Figures 6b and 6c). Therefore, if the normal-
viscous and elastic components of the mixture. ized CV increases or decreases among cases, this
change can be understood to be caused only by the
different values of AI or FI of the cases. In other
5.3 Variability inthe resulting values of E* and δ
words, this parameter could be used to quantify
Given that several comparable numerical speci- the impact of the morphological properties of the
mens or replicates compose each case of analysis, aggregates on inducing uncertainty in the determi-
it is possible to analyse the variability related to the nation of the viscoelastic properties of a mixture.

197
(a) • CV of Form Index (100 realizations, f ixed AI) 6 CONCLUSIONS
• CV of Angularity Index (100 realizations, fixed Fl)
1.8
1.6 In this study, several two-dimensional computa-
1.4 tional random HMA samples with coarse aggre-
- 1.2 gates having controlled morphological properties
~ 1 of form and angularity were subjected to cyclic
>
u 0.8 loading and their linear viscoelastic properties
0.6 were determined. This analysis was performed to
0.4 study the influence of morphological properties
0.2 (AI and FI) of the aggregates in the behaviour
0
and uncertainty of the dynamic modulus (|E*|)
4
and phase angle (δ) of the asphalt material.
Case
The aggregate phase within the microstructures
of the samples was created using a random genera-
(b) 7.5 7.5 tor of HMA microstructures, in which the angu-
CV E*
_. CVB Fixed AI, larity and form of the particles were controlled
6.5 6.5
----8---- CV E* (norm) variable FI through two indexes (AI and FI, respectively). A
5.5 ----6---· CV B (norm.) 5.5
\2::> total of 10 different cases with variable values of
~ 4.5 4.5 FI at a fixed value of AI, and variable values of AI
•------8-......... A 0
at a fixed value of FI were studied. The microstruc-
(j 3.5 /" A -., 3.5 3
6------------~/ ~----- G------------i!J tures were modelled in FE and the isolated impact
2.5 2.5
A ,. of the shape parameters of the coarse aggregates
1.5 8-------------'--- 1.5 on the linear viscoelastic properties of the mixtures
was studied.
1.5 2.5 3.5 4 Under the constraints of this study, i.e. angu-
Form Index (Fl), average larity increases while FI is kept constant and
vice versa, it was found that increasing the form
(c) 7.5
r (FI) of the aggregates produced a reduction in

~~ ~.
7.5
6.5
----G---· CV E* (norm.)
r 6.5
E*, and that increasing the angularity (AI) gen-
erated a rise in E*. The modulus was found to
be more sensitive to changes in angularity than
5.5 Bj
:·· ·.\ ·.............., ..
----6---- CV B (norm.) l 5.5 \2
for equivalent changes in FI. Meanwhile, the
~ 4.5 4.5 ::>

,A
0 behaviour of the phase angle δ was the opposite:
(j 3.5
', \ A A 3.5 ~ it increased when FI increased, and decreased
2.5 \ \ }!{/ _.!"! [ 2.5 when AI increased.
1.5
Fixed FI, \,8_~~~--"..-:::~:::'.E(,/ 1.5
In general, the impact of the aggregate shape
variable AI parameters was found to be mild when the HMA
0.5 0.5 microstructures had coarse aggregates with
0.05 0.07 0.09 0.11 0.13 0.15
angularity and form indexes typical of actual
Angularity Index (AI), average
mixtures (i.e. changes between 4 and 7% for E*
and between 11 and almost 16% for δ). However,
Figure 6. Coefficients of variation of (a) input values
recognizing and quantifying the actual influ-
of AI and FI, (b) E* and δ for fixed AI cases, (c) E* and
δ for fixed FI cases. ence of these morphological properties on the
macro-properties of the mixture is important to
better understand the contribution of the aggre-
By studying Figure 6b, a slightly increasing gate phase to the mechanical response of asphalt
trend can be found between the Form Index and mixtures.
the resulting variability of E*. These results sug- Finally, the results suggest that AI and FI do
gest that increasing FI (at constant AI) may be impact the resulting variability of E* of δ. FI
inducing an increase in E* variability, while the appears to affect the variability of E* in a more
effect of FI in the variability of δ is less certain. direct way than δ (i.e. increases in FI produce an
On the other hand, it can be observed that E* and increase in the variability of E*), and a value of
δ are more sensitive to changes in AI (Figure 6c) AI of 0.10 was found to minimize the variability
than the FI parameter (Figure 6b). However, it is in the determination of both E* and δ. In other
difficult to establish a trend in the variability of E* words, the morphological properties of the aggre-
and δ due to the increase of AI, other than the fact gates induce certain level of uncertainty when
that high values of CV of E* and δ appear away quantifying the mechanical properties of asphalt
from an ‘optimum’ AI value close to 0.10. mixtures.

198
ACKNOWLEDGEMENTS of asphalt mixtures. Materials and Structures, 41(6),
pp.1025–1037.
This publication was partially made possible by the Kim, Y., Allen, D. & Little, D., 2005. Damage-Induced
call for proposals “Research Program 2012” from Modeling of Asphalt Mixtures through Computa-
tional Micromechanics and Cohesive Zone Frac-
the Office of the Vice President for Research at ture. Journal of Materials in Civil Engineering, 17(5),
Universidad de los Andes (Bogotá, Colombia). Its pp.477–484.
contents are solely the responsibility of the authors Masad, E. et al., 2001. Correlation of Fine Aggregate
and do not necessarily represent the official views Imaging Shape Indices with Asphalt Mixture Per-
of the University. formance. Transportation Research Record: Journal of
the Transportation Research Board, 1757, pp.148–156.
Masad, E., Castelblanco, A. & Birgisson, B., 2006.
REFERENCES Effects of Air Void Size Distribution, Pore Pressure,
and Bond Energy on Moisture Damage. Journal of
Al-Rousan, T. et al., 2005. New Methodology for Shape Testing and Evaluation, 34(1), pp.1–9.
Classification of Aggregates. Transportation Research Papagiannakis, A., Abbas, A. & Masad, E., 2002. Micro-
Record: Journal of the Transportation Research Board, mechanical Analysis of Viscoelastic Properties of
1913, pp.11–23. Asphalt Concretes. Transportation Research Record:
Aragão, F.T.S. et al., 2011. Micromechanical Model for Journal of the Transportation Research Board, 1789,
Heterogeneous Asphalt Concrete Mixtures Subjected pp.113–120.
to Fracture Failure. Journal of Materials in Civil Engi- Perico, C., 2014. Caracterización de Mezclas Asfálticas
neering, 23(1), pp.30–38. Finas Modificadas con Aceite Refinado, Blanqueado
Caro, S. et al., 2016. Influence of different sources of y Desodorizado (RBD). Universidad de los Andes,
microstructural heterogeneity on the degradation of Bogotá, Colombia.
asphalt mixtures. International Journal of Pavement Shakiba, M. et al., 2015. Three-dimensional microstruc-
Engineering. tural modelling of coupled moisture-mechanical re-
Castillo, D. et al., 2015. Studying the effect of micro- sponse of asphalt concrete. International Journal of
structural properties on the mechanical degradation Pavement Engineering, 16(5), pp.445–466.
of asphalt mixtures. Construction and Building Mate- Ying, H. et al., 2014. Heterogeneous Finite-Element
rials, 93. Modeling of the Dynamic Complex Modulus Test
Coleri, E. et al., 2012. Development of a micromechani- of Asphalt Mixture Using X-ray Computed Tom-
cal finite element model from computed tomogra- ography. Journal of Materials in Civil Engineering,
phy images for shear modulus simulation of asphalt 10.1061/(ASCE)MT.1943-5533.0000949.
mixtures. Construction and Building Materials, You, Z., Adhikari, S. & Kutay, M.E., 2009. Dynamic
30(783–793). modulus simulation of the asphalt concrete using the
Dai, Q. & You, Z., 2008. Micromechanical finite ele- X-ray computed tomography images. Materials and
ment framework for predicting viscoelastic properties Structures, 42(5), pp.617–630.

199
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Linear viscoelastic behaviour of bituminous mixtures with

multi-Recycled Asphalt Pavement

A. Pedraza, H. Di Benedetto & C. Sauzéat

University of Lyon/ENTPE, Laboratory of Civil Engineering and Building (LGCB) and Laboratory of
Tribology and System Dynamics (LTDS) (UMR CNRS) 3 Rue Maurice Audin, Vaulx-en-Velin, France

S. Pouget
Eiffage Infrastructures, Research and Innovation Department 8 rue du Dauphiné, Corbas Cedex, France

ABSTRACT: Multi-recycling of Reclaimed Asphalt Pavement (RAP) in bituminous mixtures is a rela-

tively new procedure of growing interest for the research community. The aim of this paper is to show the
influence of several RAP recycling cycles on thermo-mechanical properties of bituminous mixtures.
Three hot bituminous mixtures were studied. The first mixture, considered as reference, was fabricated
without RAP, using a 35/50 pen grade binder. The second one contains 40% of RAP and was used, after
aging procedure, as RAP for the third mixture. Therefore, this last mixture, containing 40% of the “aged”
second one, contains only 16% of the original RAP, subjected to two recycling cycles. A 50/70 pen grade
bitumen was used as fresh binder for the two mixtures containing RAP.
Complex modulus tests were performed to determine the Linear Viscoelastic (LVE) properties of all
mixtures. 2S2P1D (2 Springs, 2 Parabolic Elements and 1 Dashpot) model was used to fit all experimental
data. The evolution of LVE properties and 2S2P1D model parameters were analysed as a function of the
different composition of mixtures. Similar results were obtained for the two mixtures containing RAP,
regardless of the number of recycling cycles.

1 INTRODUCTION asphalt pavement have not been done, except for

binder index characteristics. (Chen et al., 2009).
Nowadays, Reclaimed Asphalt Pavement (RAP) is High RAP contents (at least 40%) are considered
widely used in different road related applications. in the project. Bituminous mixtures are produced
The materials present in old asphalt pavements have in laboratory condition, with classical hot mixing,
residual value even when they complete their service warm mixing with additives and warm mixing with
lives. RAP has proven its economic and environ- foam bitumen. The thermo-viscoelastic behaviour
mental interest. In addition to the aforementioned, of bituminous mixtures was first characterized. Fur-
in terms of thermomechanical properties, bitumi- ther studies will focus on thermal cracking and crack
nous mixtures with added RAP fulfil the require- propagation. This paper presents the first results.
ments for pavement as shown by many researches Tension-compression 3-dimensional complex mod-
(Kaur et al., 2013; Mogawer et al., 2012; Valdés ulus tests were carried out on cylindrical specimens
et al., 2011; Visintine et al., 2013; Widyatmoko, at the Laboratory of Civil Engineering and Building
2008). A logical question is now rising concerning (LGCB) of the “Ecole Nationale des Travaux Publics
the number of times pavements might be recycled. de l’Etat” (ENTPE). Complex modulus was meas-
In this context, a French project called ured on a wide range of temperature and frequency.
IMPROVMURE aims at studying the behaviour LVE behaviour was then modelled with 2S2P1D
of bituminous mixtures containing RAP, with sev- rheological model, developed at ENTPE.
eral recycling cycles (multi-recycling). IMPROV-
MURE stands for “Innovation en Matériaux et
PROcédés pour la Valorisation du MUlti-Recyclage 2 COMPLEX MODULUS TEST
des Enrobés”. It is founded by the French Research
National Agency (ANR). This project is pio- Sinusoidal tension-compression loadings were
neer in the study of the multi-recycling. Previous applied to cylindrical specimens (140-mm height
researches regarding the thermo-mechanical prop- and 75-mm diameter) using a hydraulic press.
erties of bituminous mixtures with multi-recycled Tests were performed in strain controlled mode.

201
Bituminous mixtures behaviour is related to the where ε0 and σ0 are the strain and stress amplitudes
applied strain level and the number of loading respectively. There exists a time lag between the
cycles. In order to remains in the viscoelastic lin- stress application and the material’s deformation.
ear domain, the applied axial strain amplitude, ε0, It is defined by the phase angle (ϕE).[4] Figure 2
was maintained equal to 50 μm/m and the number shows an example of both sinusoidal signals meas-
of applied cycles very low. As shown by several ured during test.
studies (Gayte et al., 2016, 2014), this ensures that From amplitude and phase angle of measured
behaviour is linear viscoelastic. signal, the complex modulus E* can be calculated
The considered axial strain value is the average of as in Equation 3 (Di Benedetto and Corté, 2005):
three strain measurements, from three extensome-
ters. They were located at 120° around the specimen σ0 j
as shown in Figure 1. Their length is 72.5 mm and E* e E = E * e jϕ E (3)
ε0
their measurement range ±1 mm. Two non-contact
transducers were used to measure the radial strain
Performed tests consist in determining the
of the sample, which is not presented in this paper.
complex modulus for the widest possible range of
A load cell was used to measure the force, in order
temperatures and loading frequencies (Figure 3).
to calculate the axial stress (σ). The testing tempera-
ture was imposed with a thermal chamber regulated. o Strain fEl • Stress fal
~
A thermal gauge (PT100 probe) measured the tem- 60 150
perature at the surface of specimens.
40 100
The measurements performed during the test
enables to determine the complex modulus E* in the 20 so
vertical direction, which corresponds to the direc- :[
0
~
E
tion I (Figure 1). Directions I, II and III correspond ~
c
~
v;
to specific material direction determined during the .@ -20
v;
-50

fabrication (slab compacting and samples coring). -40 aO ; 126 kPa -100
As indicated in Figure 6. Direction I is the direction ED; 50 -
~m

m
of wheel rolling, direction II is the vertical compac- -60
0.0 5.0 10.0 15.0 20.0
-150

tion loading direction and direction III is the direc- nme(s)

tion transverse to the rolling wheel.
Considering the LVE behaviour of tested mix- Figure 2. Example of stress and strains measurements.
tures, when a sinusoidal strain ε is applied in the (f = 0.1 Hz and T = 15°C).
axial direction I, as expressed in Equation 1, a
60
resulting sinusoidal stress signal can be observed
(Equation 2). E 40
(t ) ε 0 sin (ωt ) (1) ~ 20
~
~
σ (t ) σ 0 sin (ωt + ϕ E )
0
(2) Q.
E
~ -20
-25
I (Traffic) -40
0 10 20 30 40 so 60
Test for 35"C
II (Vertical)- \ .A
Ill r Test for -25 to 25"C """\
100 ---~.A __ _ _
Axial r SO cycleS\
PTlOO surface
extensometer (3) 10 40 cyde(OHz
30 cycles 3Hz
1H•
Target for non 16 cvcles
12 cycles 0.3Hz
contact sensor
I 0 .1 8 cvcles O.l Hz
0 .0 3Hz
System to maintain
non-contact sensor Lower cap 0.01

0.001

0.0001
Radial non contact 4.3 4.6 4.9 S.2 s.s
Time (h)
sensor (2)
Figure 3. Procedure of the complex modulus test: tem-
Figure 1. Instrumented sample for complex modulus peratures chronology, loading frequencies and applied
test. cycles number.

202
The samples were tested at nine different tempera- The pulsation is defined as ω 2π f with f
tures going from −25°C to 52°C. Eight loading the loading frequency. E0 is the glassy modulus
frequencies were applied (0.003 Hz to 10 Hz). The (ω → ∞) and E00 is the static modulus (ω → ).
number of loading cycles varied from eight cycles at E0 and E00 are parameters that depend on sample
0.03 Hz up to 50 cycles at 10 Hz. The full procedure compaction and aggregate skeleton (Olard and Di
is detailed in Figure 3. In order to verify the integ- Benedetto, 2003). The complex number j is defined
rity of the sample throughout the test, loadings at by j2 = −1. δ is a calibration constant. k and h are
15°C were carried out three times at the beginning, constants that vary between 0 and 1. β is a con-
middle and end of the test (15(1), 15(2) and 15(3) stant that depends on the viscosity of the dashpot:
respectively). η ( ) βτ and τ is characteristic time value.
When TTSP is verified, τ depends on the tem-
perature, as shown on Equation 5:
3 LVE MODELLING
τ (T ) T (T ) τ 0 (5)
Experimental results were modelled with the
2S2P1D rheological model, developed in the where aT is the shift factor at temperature T and
LGCB of the University of Lyon/ENTPE. τ = τ0 at reference temperature Tref.
The model is a generalization of the Huet- The shift factors, introduced to take into
Sayegh model (Huet, 1963) and consists of 2 account TTSP, may be approximated by the
Springs, 2 Parabolic creep elements and 1 Dash- William-Landel-Ferry (WLF) equation (Equation
pot (Mangiafico, 2014; Olard and Di Benedetto, 6) (Ferry, 1980; Williams et al., 1955). Two WLF
2003; Tapsoba et al., 2014). The 2S2P1D model constants (C1 and C2 in Equation 6), determined
has been widely used to model the LVE behav- at the reference temperature, are then needed if the
iour of bituminous materials, including binders, temperature effect is to be considered. In that case,
mastics and mixes on a wide range of frequen- 9 constants are required to model the complex
cies and temperatures (Pouget et al., 2012, 2009; modulus at any temperature.
Tiouajni et al., 2011). A 3D version of the model
has been developed to completely describe the 3D
LVE behaviour of an isotropic material. In this log (aT ) =
−C (
C1 T − Treef ) (6)
paper only the modelling of complex modulus C2 T − Treef
is considered. The model describes E* with only
7 parameters (Equation 4) for one temperature. In this study, the reference temperature Tref was
If the Time Temperature Superposition Princi- fixed at 15°C.
ple (TTSP) is valid, E* may be modelled for any
temperature with 2 supplementary parameters, as
explained further.
Some parameters are easily identified in a Cole- 4 MATERIALS
Cole diagram, as shown on Figure 4:
The experimental campaign was carried out on
E0 E 00 three different hot bituminous mixtures. Each
E * (ω ) = E 00 + (4) material is a so-called BBSG-03 0/10 bituminous
1 + δ ( jω ) ( jωτ ) + ( jωβτ )
k h −1

mixture, as specified in the European standards

for classification of bituminous mixtures (NF EN
13108-1 - 2007). The same grading curve was used
for the 3 configurations as shown on the Figure 5.
Total binder content was 5.4% for all mixtures.
The three hot bituminous mixtures were produced
by adding RAP material in various proportions (0%,
40% of the total mass). Two different unaged bitu-
men were added according to the RAP content.
The first mixture called LH0-0, considered as
the reference bituminous mixture is made without
RAP and with a 35/50 pen grade bitumen.
The second mixture called LH40-1 contains 40%
Figure 4. Analogical representation of the 2S2P1D RAP and a 50/70 pen grade bitumen. The RAP
Model (left) and representation of the model parameters used had a binder content 4.16% and a 10 ⋅10 −1 mm
on a Cole-Cole diagram (right) 1 Dimension (Olard and pen grade aged bitumen. The RAP comes from a
Di Benedetto, 2003). mix plant samples which was homogenised.

203
 - - LH0-0 - + - LH40-1 - - LH40-2 Table 1. Estimated penetration of binder blend from
100 log-log rule, assuming a perfect blend.
~ 80 LH0-0 LH40-1 LH40-2
"
OD
!!! 60
c
a (%) 100 69 59
"~ 40 Pena 41 60 60
""-
OD
b (%) – 31 41
c 20
..
·~
Q.
0
Penb – 10 14
Penmix 41.0 34 33
0.01 0.1 10 100
Grain dimensions (mm)

Figure 5. Aggregate grading curves.

The third mixture called LH40-2 was made with

40% RAP and the same 50/70 pen grade fresh bitu-
men as for the second mixture. This mixture ena-
bles to study a second recycling cycle. Indeed, the
used RAP comes from the second mixture, which
was aged in the lab, according to the following
procedure.
After mixing the LH40-1 bituminous mixture,
one part was not compacted and was submitted
to aging process according to a method developed
during RILEM works (International Union of
Laboratories and experts in construction materi- 400mm
als, systems and structures) (Partl et al., 2012). This
method involves to warm up the material during Figure 6. Slab from French Rolling compactor and
4 h to 135°C and then nine days to 85°C. Result- specimen material directions.
ing material is considered to be equivalent to RAP,
which may be used for a second recycling cycle. Table 2. Characteristics of tested specimens.
This RAP had a binder content 5.4% and a pen RAP Recycling Voids Type of
grade bitumen of 14 ⋅10 −1 mm. Sample [%] cycle (%) binder added
As the total binder content, the RAP and the
fresh added bitumen are fixed, the binder blend LH0-0 1–6 0 – 7.2 35/50
is also completely fixed. The penetration of each 1–7 0 – 6.8 35/50
binder blend was then estimated from the log-log LH40-1 1–3 40 1 5.9 50/70
rule, considering a perfect blend (Mangiafico, 1–8 40 1 6.0 50/70
2014): LH40-2 1–7 40 2 6.0 50/70
1–8 40 2 6.7 50/70
log ( blend ) log ( pena ) b log ( penb ) (7)
Figure 6. Cylindrical specimens were cored in
where a and b are mass fractions of fresh added direction I, parallel to rolling direction. Dimen-
bitumen and RAP binder respectively. Table 1 sions are 75 mm in diameter and 140 mm in height.
presents used parameters value and results of the Two specimens of each mixture were tested.
calculations for the three mixtures. Characteristics of the tested specimens are shown
Resulting blend penetrations are close to the in Table 2.
35/50 bitumen of the reference mixture. Conse-
quently it seems relevant to compare the results of
the three mixtures. 5 EXPERIMENTAL RESULTS
The bituminous mixtures were produced in the AND ANALYSIS
Research Center of the French company EIFF-
AGE Infrastructures. 600 × 400 × 150 mm slab Results of complex modulus tests are first plot-
were compacted using a French wheel compactor ted in the Cole-Cole diagram (imaginary part in
(NF EN 12697-33+A1 - 2007). Material direction function of real part of E*) and Black’s diagram
I, II and III were defined from the slab compac- (norm of E* in function of phase angle). Cole-
tion, as explained previously and indicated in Cole diagram shows rather good test repeatability

204
for the three mixtures, at low temperature. In the l.E+OS
~ 15. 1(1)"(

Black’s diagram some discrepancy appears at very ~ -25.2"(

high temperature, mainly for the phase angle. At ~ -15.1'(

low and intermediate temperature, the mixtures l.E+04 ____._ -5.2' (

LH40-1 and LH40-2 show very similar results. ~ s .o·c

In these axes, data form continuous curves. The

time temperature superposition principle can be
applied. Therefore, materials are thermo-rheolog-
!. l.E+03
- - 15.0(2)' (

~ 2 5. 2'(

ically simple. (Chehab et al., 2002; Nguyen et al., - - 35.1' (

2013). l.E+02
- - 45.0' (

Hence, master curves of the norm and phase ~ 52.0'(

angle of complex modulus can be built. Figure 8 ____._ 15.3 (3)'(

shows LH40-2-1-7 isotherms curves of |E*| and l.E+Ol Master

the master curve plotted at a reference temperature l.E-07 l.E-03 l.E+Ol l.E+OS l.E+09
Curve

Tref = 15°C. Frequency (Hz)

Master curves and shift factors aT were deter-

mined for all the tested samples. Figure 9 presents Figure 8. Illustration of the |E*| master curve construc-
tion procedure for LH40-2-1-7 sample.
aT shift factors as a function of the temperature. aT
values are very close. Data were then approached
with the WLF equation (Equation 6). C1 and C2 l .E+09 aT EXP LH0-0-1-6
WLF parameters were obtained from the best
- - aT WLF LH0-0-1-6
fitting for each material. Considering the small 1.E+07
difference between aT values, a set of C1 and C2 aT EXP LH0-0-1-7

parameters is also proposed to approach all aT l .E+OS

- - aT WLFLH0-0-1-7

aT EXP LH40-1-1-3

3.E+03
.
•i>(jf~~ . .. •
-';;;
1.E+03 - - aT WLF LH40-1-1-3

aT EXP LH40- 1- 1-8

.~··
·~"'' l :\-~. . cz:g>~ ·e l .E+Ol - - aT W LF LH40-1-1-8

aT EXP LH40-2-1-7
-;o
Q._

;; l:Ji. l! ·. ~: .· ~, l.E-01
- - aT WLF LH40-2-1-7
j .~ .g
2.E+03
~ t: ~ o
aT EXP LH40-2-1-8
~
1ij
c
·c;,
J f
• LH0-0-1-6 (7.2%)
.."66'!i.
r- ~
0
l .E-03
- - aTWLF LH40-2-1-8

- - Average model WLF

~ I . LH0-0-1-7 (6.8%) .. . ....

l .E-05
-40 -20 20 40 60
~f. 0~0
l.E+03
o LH40-1-1-3 (5.9%) Temperature ("C)
• LH40-1-1-8 (6.0%) 4, o
o LH40-2-1-7 (6.0%) Figure 9. Factually coinciding shift factors aT for E* the
O.E+OO • LH40-2-1-8 ~ tested materials based on WLF law fitting Equation 6.
O.E+OO l. E+04 2.E+04 3.E+04
Real (E*) (MPa)
Table 3. WLF constants of tested specimens, Tref = 15°C.

Sample C1 C2
l.E+04
LH0-0 1–6 27.83 182.44
1–7 20.03 134.54
l.E+03
-;o LH40-1 1–3 24.16 162.79
Q._
;; 1–8 25.8 166.52
k l.E+02 LH40-2 1–7 25.37 172.15
1–8 22.11 145.56
Average Model WLF 25.13 167.5
l.E+01

l.E+OO
0 10 20 30 40 so 60 shift factors independently of the specimen or the
Phase angle <p (' ) mixture. C1 and C2 parameters are given in the
Table 3. WLF curves are also plotted in Figure 9.
Figure 7. Complex modulus test results in Cole-Cole Figure 10 presents modulus and phase angle
plan (up) and Black space (down). master curves plotted at Tref = 15°C using the

205
 1.E+OS Table 4. E0 and E00 values of tested specimens.
- · %_

_a,~ ~ Sample E0 (MPa) E00 (MPa)

1.E+04
jp
LH0-0 1–6 32300 10
li' l.E+03 J I T,.t=lS' C 1–7 31700 22

t!~· LH40-1 1–3 35400 40

a.
! 1–8 34950 19
~ l.E+02 ,g1
LH40-2 1–7 35000 19
~~·
. 1–8 34700 20
0

Jt.~{· A
1.E+01

36,000 E, = · 2443.8x + 49730

1.E+OO so E00 = -11.514x + 95 .739
l.E·09 l.E·OS 1.E· 01 l.E+03 l.E+07 35,000 R2 = 0.4

\
aT . frequency (Hz)
40 0

• LHQ-0·1·6 (7.2%) • LH0·0·1·7 (6.8%) o LH40-1-1-3 (5.9%) 34,000 30

Q
w 8

~"'
• LH40-1· 1·8 (6.0%) o LH40·2· 1·7 (6.0%) • LH40-2· 1·8 (6.7%) w

60 .. 33,000 20

: ~~t A I T,.t=lS' C I
32,000
.. 10
50 " ·~ ~
·· · ~~.. · 31,000 0
~~~
·-- ~p ~
5.0 6.0 7.0 8.0 5.0 6.0 7.0 8.0
40
:· t
Voids(%) Voids (%)

c-e- 30 ..,. "'

• C Jt.O .. Figure 11. E0 function of voids and (left) and E00

l ~i~
0
••• o function of voids (right).
0
20
\ l ... Table 5. 2S2P1D unique simulation model calibration
10
parameters for the tested materials.
0
""~ ~
k h δ β
l.E-09 l.E-05 l.E·01 l.E+03 l.E+07
aT. frequency (Hz)
0.179 0.575 2.250 150
Figure 10. Complex modulus (up) and phase angle
(down) master curves of all the tested samples. viscous behaviour of the mixtures, which is mainly
influenced by the binder.
*
In, E nor data are plotted in Cole-Cole diagram.
constants C1, C2 of the WLF law (Table 3). The
It is possible to describe a unique curve for all
modelling of these data enables to describe the
tested specimen. This observation indicates that
complete LVE behaviour of tested materials.
bituminous mixtures have very similar LVE behav-
In order to fit data with 2S2P1D model, E0
iour. k, h, δ, β 2S2P1D parameters were fitted
and E00 values were first estimated. As indicated
using this diagram. Unique values were determined
in Table 4, these parameters are different for each
for all tested specimen so that a unique curve was
tested specimen. Previous studies have shown that
obtained. The parameters used for 2S2P1D are
these parameters are related to the aggregates skel-
shown in Table 5. It should be underlined that, it
eton and the air voids content. In Figure 11 E0 and
is not possible to observe the influence of tempera-
E00 are plotted in function of the voids content. A
ture in this diagram. Then τ parameter could not
rather clear tendency is visible.
be determined with this representation, and master
One E0 and E00 are obtained, it is possible to
curves have to be used.
normalised E* as follows:
Table 6 shows the values of characteristic times
τ determined at Tref. = 15°C, thanks to the mas-
E * E00
*
E nor = (8) ter curves. The experimental master curves and
E0 E00 the calculated with 2S2P1D model are plotted for
all the specimens at Tref. = 15°C on Figure 13. This
Such normalized E* enables to compare the evo- Figure shows that experimental data are satisfac-
lution of the complex modulus between the two torily modelled on the whole range of considered
asymptotic behaviour (at very low and very high temperatures and frequencies, thanks to 2S2P1D
frequencies). This evolution is dependent on the model.

206
Table 6. Characteristic times τ determined at Tref. = 15°C. 6 CONCLUSIONS
Sample τ(s)
IMPROVMURE project allows address the multi-
LH0-0 1-6 0.042 recycling, a topic which is not sufficiently studied,
1–7 0.085 but the environmental and economic advantages
LH40-1 1–3 0.079 justify more research on the thermomechanical
1–8 0.045 behaviour on the topic.
LH40-2 1–7 0.098 The TTSP was verified and applied for the tested
1–8 0.073
bituminous mixtures. The shift factors used were
very closes for all materials and a unique WLF
0.12 LH0-0-1-6 (7.2%)
LH0-0-1-7 (6.8%) model could be fitted.
-0 0.10
LH40-1-1-3 (5 .9%) Normalized curves shown a similar LVE behav-
LH40-1-1-8 (6.0%)

s LH40-2-1-7 (6.0%) iour of mixtures, this means that RAP percentage

s
:::;
0.08
LH40-2-1-8 (6.7%)
and number recycling cycles (study limited to two
0 cycles), have no significant impact on the perform-
q: 0.06
ance of mixtures.
- A 2S2P1D model was satisfactorily fitted to
.~
0.04
., experimental results on the whole range of consid-
~ 0.02 ered temperatures and frequencies for the complex
modulus of bituminous mixtures.
0.00
0.00 0 .20 0.40 0.60 0 .80 1.00
Thus, the RAP multi-recycled can be considered
Real (E* - EOO)/(EO-EOO) (-) as a suitable material for road related applications.
Figure 12. Complex modulus test results in plan Cole-
Cole normalized.
ACKNOWLEDGMENTS
l.E+OS

Authors would like to acknowledge Eiffage Infra-

1.E+04 structures, Research & Innovation Department for
providing the materials and the French research
~l. E+03 agency (ANR) for its support.
!
;_ l.E+02

REFERENCES

Chehab, G.R., Kim, Y.R., Schapery, R.A., Witczak,

l.E+OO _(__~~~~~----~-~;::::::;:;~::;:;;:::::~ M.W., Bonaquist, R., 2002. Time-temperature super-
l.E-09 l.E-05 l.E-01 l.E+03 l .E+07 position principle for asphalt concrete with growing
aT . frequency (Hz) damage in tension state, in: Journal of the Associa-
LH0-0-1-6 (7.2%) LH0-0-1-7 (6.8%)
tion of Asphalt Paving Technologists. Presented at the
LH40-1-1-3 (5.9%) LH40-1-1-8 (6.0%)
LH40-2-1-7 (6.0%) LH4Q-2-1-8 (6.7%)
Asphalt Paving Technology 2002.
- - 252P10 LH0-0-1-6 ---- 252P1D LH0-0-1-7 Chen, J.-S., Wang, C.-H., Huang, C.-C., 2009. Engineer-
- - 252P10 LH40-1-1-3 ---- 252P1D LH40-1-1-8 ing Properties of Bituminous Mixtures Blended with
- - 252P10 LH40-2-1-7 ---- 252P10 LH40-2-1-8
Second Reclaimed Asphalt Pavements (R 2 AP). Road
Mater. Pavement Des. 10, 129–149. doi:10.1080/14680
629.2009.9690240
Di Benedetto, H., Corté, 2005. Matériaux routiers bitu-
mineux Tome2.
Ferry, J.D., 1980. Viscoelastic Properties of Polymers.
John Wiley & Sons.
Gayte, P., Benedetto, H.D., Sauzéat, C., Nguyen, Q.T.,
2016. Influence of transient effects for analysis of
complex modulus tests on bituminous mixtures. Road
Mater. Pavement Des. 17, 271–289. doi:10.1080/14680
629.2015.1067246.
l.E-09 l.E-05 l.E-01 l.E+03 1.E+07
Gayte, P., Di Benedetto, H., Sauzeat, C., 2014. Three dimen-
aT . frequency (Hz) sional behaviour of bituminous mixtures in the linear
viscoelastic and viscoplastic domains: The DBN model,
Figure 13. Complex modulus (up) and phase angle in: Asphalt Pavements. CRC Press, pp. 1079–1090.
(down) experimental and modelled master curves of all Huet, C., 1963. Etude par une méthode d’impédance du com-
the tested samples. portement viscoélastique des matériaux hydrocarbons.

207
Kaur, K., Swamy, A.K., Das, A., 2013. Constituent Pro- Pouget, S., Sauzéat, C., Benedetto, H.D., Olard, F., 2009.
portioning in Recycled Asphalt Mix with Multiple Numerical simulation of the five-point bending test
RAP Sources. Procedia—Soc. Behav. Sci. 104, 21–28. designed to study bituminous wearing courses on
doi:10.1016/j.sbspro.2013.11.094. orthotropic steel bridge. Mater. Struct. 43, 319–330.
Mangiafico, S., 2014. Linear viscoelastic properties doi:10.1617/s11527-009-9491-1.
and fatigue of bituminous mixtures produced with Tapsoba, N., Sauzéat, C., Di Benedetto, H., Baaj, H.,
Reclaimed Asphalt Pavement and corresponding Ech, M., 2014. Behaviour of asphalt mixtures con-
binder blends. École Nationale des Travaux Publics de taining reclaimed asphalt pavement and asphalt shin-
l’État. gle. Road Mater. Pavement Des. 15, 330–347. doi:10.1
Mogawer, W., Bennert, T., Daniel, J.S., Bonaquist, R., 080/14680629.2013.871091.
Austerman, A., Booshehrian, A., 2012. Performance Tiouajni, S., Di Benedetto, H., Sauzéat, C., Pouget, S.,
characteristics of plant produced high RAP mixtures. 2011. Approximation of Linear Viscoelastic Model in
Road Mater. Pavement Des. 13, 183–208. doi:10.1080/ the 3 Dimensional Case with Mechanical Analogues
14680629.2012.657070. of Finite Size: Application to Bituminous Materials.
Nguyen, Q.T., Benedetto, H.D., Sauzéat, C., Tapsoba, Road Mater. Pavement Des. 12, 897–930. doi:10.1080/
N., 2013. Time Temperature Superposition Principle 14680629.2011.9713899.
Validation for Bituminous Mixes in the Linear and Valdés, G., Pérez-Jiménez, F., Miró, R., Martínez, A.,
Nonlinear Domains. J. Mater. Civ. Eng. 25, 1181– Botella, R., 2011. Experimental study of recycled
1188. doi:10.1061/(ASCE)MT.1943-5533.0000658. asphalt mixtures with high percentages of reclaimed
Olard, F., Di Benedetto, H., 2003. General “2S2P1D” asphalt pavement (RAP). Constr. Build. Mater. 25,
Model and Relation Between the Linear Viscoelastic 1289–1297. doi:10.1016/j.conbuildmat.2010.09.016.
Behaviours of Bituminous Binders and Mixes. Road Visintine, B., Khosla, N.P., Tayebali, A., 2013. Effects
Mater. Pavement Des. 4, 185–224. doi:10.1080/14680 of higher percentage of recycled asphalt pavement on
629.2003.9689946. pavement performance. Road Mater. Pavement Des.
Partl, M.N., Bahia, H.U., Canestrari, F., Roche, C. de 14, 432–437. doi:10.1080/14680629.2013.779310.
la, Benedetto, H.D., Piber, H., Sybilski, D., 2012. Widyatmoko, I., 2008. Mechanistic-empirical mix-
Advances in Interlaboratory Testing and Evaluation ture design for hot mix asphalt pavement recycling.
of Bituminous Materials: State-of-the-Art Report of Constr. Build. Mater. 22, 77–87. doi:10.1016/j.
the RILEM Technical Committee 206-ATB. Springer conbuildmat.2006.05.041.
Science & Business Media. Williams, M.L., Landel, R.F., Ferry, J.D., 1955. The Tem-
Pouget, S., Sauzéat, C., Benedetto, H.D., Olard, F., 2012. perature Dependence of Relaxation Mechanism in
Modeling of viscous bituminous wearing course Amorphous Polymers and Other Glass Forming Liq-
materials on orthotropic steel deck. Mater. Struct. 45, uid. J. Am. Chem. Soc. 77. doi:10.1021/ja01619a008.
1115–1125. doi:10.1617/s11527-011-9820-z.

208
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Apps for modal analysis to characterize the complex modulus of

asphalt concrete

A. Gudmarsson
Peab Asfalt AB, Stockholm, Sweden

N. Ryden
Engineering Geology, Faculty of Engineering, Lund University, Lund, Sweden

ABSTRACT: The time and temperature dependent dynamic modulus of asphalt concrete is a key
parameter to modern pavement design and to pavement quality. However, accurate measurements of the
complex modulus through conventional cyclic testing are expensive regarding both time and costs. Novel
test methods based on modal analysis facilitate economic, accurate and faster determination of linear
viscoelastic material properties of asphalt concrete. The modal analysis is performed by using an impact
hammer and an accelerometer to measure frequency response functions of asphalt concrete specimens
with free boundary conditions. The complex modulus and complex Poisson’s ratio are determined by
matching finite element computed frequency response functions to the measurements. This paper presents
one software application to perform the modal testing and one application to compute and optimize fre-
quency response functions to determine the complex modulus and complex Poisson’s ratio. The software
applications greatly simplifies the methodology to determine material properties through modal analysis
of asphalt concrete specimens.

1 INTRODUCTION (1994) applied impact resonance testing and a con-

crete standard test method (ASTM C215-08), valid
The dynamic modulus (|E*|), defined as the abso- for a length to diameter ratio of at least two, to
lute value of the complex modulus (E*), is a key determine the elastic modulus of the fundamental
parameter to modern pavement design and to pave- resonance frequency of asphalt concrete specimens.
ment quality. To characterize the linear viscoelastic A good repeatability and reproducibility of the test
behavior of asphalt concrete, measurements of the method was concluded (Whitmoyer & Kim 1994).
complex moduli are performed over a limited range The impact resonance testing methodology was
of temperatures and frequencies. The time-tempera- further developed to account for the viscoelastic
ture superposition principle are then normally used behavior of asphalt concrete enabling the characteri-
to extrapolate the complex moduli to frequencies zation of one complex modulus per test temperature
higher and lower than those actually applied in the (Kweon & Kim 2006, Lacroix et al. 2009). However,
testing (Levenberg 2011, Nguyen et al. 2013). By this, being limited to the modulus of the fundamental
a master curve is determined, which characterizes resonance frequency, shift factors were needed to
the complex moduli over a wide range of tempera- be known beforehand to characterize the complex
tures and frequencies. The complex moduli measure- modulus master curve. Ryden (2011) applied three-
ments are typically performed through cyclic loading dimensional numerical calculations to estimate the
(Levenberg 2011, Daniel et al. 2004). However, due complex modulus for arbitrary specimen dimensions
to the significant investment in cost and time to per- and for several resonance frequencies at each testing
form the cyclic testing accurately (Daniel et al. 2004), temperature. In common for the above-mentioned
there is a need for new test methods to facilitate the impact resonance testing of asphalt concrete is
determination of the complex modulus. that only the response of the specimens (resonance
The development of test methods based on reso- frequencies) were measured. However, for viscoelas-
nance testing to derive elastic constants of different tic materials it is often necessary to determine the
types of materials have introduced new opportuni- modulus at more closely spaced frequency intervals,
ties to testing and characterization of asphalt con- not only at the resonances, to accurately determine
crete (Demarest 1971, Visscher et al. 1991, Migliori master curves (Buchanan 1987, Gudmarsson et al.
et al. 1993, Buchanan 1987). Whitmoyer & Kim 2012a, Gudmarsson et al. 2012b). This is possible

209
by performing modal analysis in which the applied analysis software applications presented in this
force is measured in addition to the response of the paper includes two parts; measurements of FRFs
specimen (Gudmarsson et al. 2012b). Modal analy- and optimization of computed FRFs to match the
sis applied to asphalt concrete have proven to enable measured FRFs.
a wider frequency characterization of the complex
modulus by using the measured acceleration and
2.1 Measuring FRFs
impact force to calculate Frequency Response Func-
tions (FRFs). The FRFs are determined by using an impact ham-
In the modal analysis applied to asphalt con- mer to apply and measure the load and an acceler-
crete, the complex modulus and complex Pois- ometer to measure the vibrations of the specimen
son’s ratio are initially assumed to compute FRFs over time. The following equipment, shown in
through the Finite Element Method (FEM). Then, Figure 1, are used to measure the FRFs: impact
an iterative fitting of the FEM computed FRFs to hammer (PCB model 086E80), accelerometer (PCB
measured FRFs is performed in order to charac- model 352B10), signal conditioner (PCB model
terize the stiffness properties over a range of fre- 480B21), data acquisition device (NI USB-6251 M
quencies (Gudmarsson et al. 2012b). The complex Series) and a computer. The specimen is placed on
modulus and complex Poisson’s ratio master curves soft foam to provide free boundary conditions. To
are determined by performing the modal testing excite and measure the flexural modes of vibration
and analysis at several measurement temperatures. for a disc-shaped specimen shown in Figure 1, the
The modal analysis can be applied to speci- impact is applied close to the edge of the specimen
mens with arbitrary geometries and measurements in the direction of the normal to the top surface.
on beams (Gudmarsson et al. 2012b), cylinders To measure the longitudinal modes of vibration, the
(Gudmarsson et al. 2014, Gudmarsson et al. 2015) impact is applied in the center of the specimen also
and disc-shaped specimens (Bjurström et al. 2016, in the direction of the normal to the top surface.
Gudmarsson 2014) have been reported. The com- Figure 2 presents an example of the measured
plex modulus and complex Poisson’s ratio deter- force from five impacts in time domain (a) and
mined through modal analysis have also shown a transformed to frequency domain (b) using the fast
good agreement with tension-compression cyclic Fourier transform. The vibrational response of
loading performed at ENTPE, University of Lyon the specimen due to the five impacts are shown in
(Gudmarsson et al. 2014, Gudmarsson et al. 2015). Figure 3 by the measured acceleration plotted in
In addition, modal analysis and in-situ measurements time (a) and frequency domain (b).
performed through Multichannel Analysis of Surface According to the Nyquist-Shannon sampling
Waves (MASW) have recently been demonstrated as theorem the sampling rate should be at least twice
very promising tools for accurate stiffness based qual- of the highest measurement frequency (e.g. 20
ity control of pavements (Bjurström et al. 2016). to 30 kHz). However, due to the short duration
Although the modal testing is economic and accu- time of the load pulse, the sampling rate was set
rate, the signal processing and optimization of FEM to 500 kHz. This gives approximately 25 samples
computed FRFs that is required to characterize over the duration of the load pulse (∼0.05 ms). The
material properties might be challenging to imple- recording of the force and acceleration is triggered
ment. Therefore, this paper present two software
applications that have been developed to facilitate
the practical implementation of the modal analy-
sis applied to asphalt concrete specimens. The first
application is used to perform the modal testing. The
second application is used to characterize the com-
plex modulus and complex Poisson’s ratio through
optimization of FEM computed FRFs. The aim of
this paper is to demonstrate the potential of software
applications to simplify the use of modal analysis
applied to asphalt concrete. Results that are obtained
using the software tools show a good accuracy for
both modified and unmodified asphalt mixtures.

2 METHODOLOGY

The methodology to determine the complex mod- Figure 1. Equipment used to perform the modal testing
ulus and complex Poisson’s ratio using the modal (Gudmarsson 2014).

210
 100

I r I II I r I
(a) .. (b)

~ ,.
.:·,.
... lime or space
doma in
~ L.......J
~ 1~ r. Dati

"
!!
50 ~·; ~ ~
dt - 1/f,
0
u. i
;·
i ~
I Yp+1 = L. oJPx j+1
n- l

0.05 0. 1 0.15 0.2 10 15 20

Timc(ms] Frequency (kHz)
I j =O

I
Figure 2. Measured force from the impact hammer in
time (a) and frequency (b) domain.
Frequency
domain
r I r I 1 y r r '? r r r •
p+l L.......J
Frequency
content
abs(y)
df = (. / 11
2000 40
(a) (b)
~ Figure 4. Illustration of discrete Fourier transform
.=. 1000
.§ ll.l from time to frequency domain (MathWorks, User’s
~
"
Guide R2016a).
1l - 1000
<-2000
0 10 15 20
Time[ms] Frequency [kll7.] The application to perform the modal testing (see
Figure 5) requires information of the date, sample
Figure 3. Measured acceleration in time (a) and fre- identification, mode of vibration and the measure-
quency (b) domain. ment temperature. This information is used to save
time domain data in a file that can be read by the
application at any time. In this way, several meas-
by the hammer impact. Note in Figure 2a and 3a urements can be repeated after each other and the
that there is a pre-trigger time defined (0.1 ms) in frequency domain transformation and calculation
order to capture the complete load pulse, which is of the FRFs can be performed afterwards. Also
important for accurate determination of FRFs. required information are; the record length, sample
The record length is depending on the damp- rate and the channels of the data acquisition device,
ing characteristics of the asphalt concrete, which used for the input (hammer) and output (accelero-
changes with the temperature. Typical record meter). The record length should be adjusted for
lengths can vary between 2 (at ∼ 40°C) to 30 (at ∼ each measurement temperature in order to ensure
−20°C) milliseconds. that the complete time of vibration of the specimen
Figure 4 shows the principle of discrete Fou- is recorded. The sample rate, which determines the
rier transform to identify the frequency content in resolution of the measurements, may be limited to
the time domain data (x). The sampling rate (fs) the specifications of the data acquisition device.
and record length gives the resolution (dt) and Once the information of the measurements are
the length (n) of data in time domain. The ampli- given, the testing may start (“Start test”) and five
tude in frequency domain (y) is given by summing impacts should then be applied to the specimen.
the data in time domain at each frequency, where The test status displays “DONE” when the testing is
= e −2π i / n , i is the imaginary unit, and p and j are completed. The FRFs are calculated and displayed
index from 0 to n − 1. along with the measured force and acceleration in
Equation 1 is used to calculate the averaged frequency domain by the “Calculate FRFs” button.
FRF from the measured force and acceleration in The calculation of the FRFs requires information
frequency domain by averaging the five measure- of the sensitivity of the hammer and accelerometer
ments (n = 5) at each frequency. used in the testing. Figure 5 shows the application
with the input fields and graphs of measurements in
⎛1 n ⎞ time and frequency domain. The calculated averaged
⎜⎝ n ∑Yk ( f ) ⋅ X k ( f )⎟⎠
*
FRF (black dashed line) is automatically saved and
H ( f )= k =1
, (1) can be further used to evaluate the complex modulus
⎛1 n ⎞
⎜⎝ n ∑ k X ( f ) ⋅ X *
k ( f )⎟⎠ and complex Poisson’s ratio. There is also a possibil-
k =1 ity to pick key points based on the resonance peaks
of the FRF to reduce the computational time in the
where H( f ) is the frequency response function, following stiffness evaluation, which is performed in
Y( f ) is the measured acceleration, X( f ) is the the application to optimize computed FRFs.
measured applied force and X*( f ) is the complex A text file that can be directly uploaded to the
conjugate of the applied force. FEM application is automatically saved from

211
 Info Input (DAQ device ID = Dev1) Test status FRFs Pick FRF points
PolntsPerPeak: G
Date : ~ Record length= ~ (@40 oc =...2 ms:@ -20 oc =... 30 DONE Ham.sensltlvty= ~
ExcludeNrOfPeaks: ~ (fromright)
SampleiD : ~ Sample rate=l sooooo I Ace. sensitivity = ~
AmplitudeOfFIHer. ~
Mode Type : ~ Ham. channel: ~ Max. frequency = ~ SlopeOffiner. G J

Temperature = ~ Ace. channel: I:=J Start test New test Calculate FRFs AmplltudeOfPolnts: ~ Pick

Plots Plot Frequency Response Functions

Impact hammer Accelerometer FRFs

0.8

1.5

-0.2

-0.5 L__-~-~-~-~-~-~ -0.4 L__-~-~-~-~-~-~

0 0.5 1.5 2.5 0 0.5 1.5 2.5

Time (s) Time (s) Frequency (Hz) x 10 4

Figure 5. Screenshot of the application to perform modal testing.

the “Pick” button. The file(.txt) consists of one

column of frequencies and one column of the
averaged FRF amplitudes. Measurement temperature

Specimen data
2.2 Computing and optimizing FRFs Height=

The FEM based application to determine the Density =

complex modulus and complex Poisson’s ratio by

Mode type
optimizing computed FRFs to measured FRFs is @ Flexuralmodesofvibration (fl)
presented in Figures 6 to 8. The application used 0 l ongitudinalmodesofvibration (f2)

for disc-shaped specimens (see Figures 6 to 8) con-

Import measured FRF*
sists of six tabs where the first tab (Input) requires C:\Users\andgud\ Documen· l Uploadfile j Yj _x
the following data: measurement temperature (°C),
height (mm), diameter (mm), density (kg/m3), mode
type, measured FRF and the size of the mesh (see
Figure 6). The default maximum mesh size (mm) Figure 6. Screenshot of the application to opti-
can be changed to an appropriate size that is related mize FEM computed FRFs to match measured FRFs
to the specimen dimensions. The choice of mode (Input).
type between the flexural and longitudinal mode
of vibration governs the position of the point load
applied in the model so that it corresponds to the Input Check startvalues OptimizeFRF PlotsofE*andv* TableofE*andv* Objectivetable

actual position of the applied load in the measure- Start values stiffness parameters
ments. A 1-Newton point load is positioned to the FO 1·10 8

edge opposite to the accelerometer position when alpha:

choosing the flexural modes of vibration. The longi-
tudinal mode of vibration positions the load in the
center of the modelled disc. The load is applied in Havriliak-Negami model
E* = E1 + (Eo- E1)/(1 + (iwrar Y't ~~ '56
the negative z-direction (see Figure 6). v*=v1 + (v0 -v1)/(l+ (iwrp(lr'f"/ 18

The second tab (Check start values) can be used

to estimate appropriate start values to the optimi- WLF shift factor equation
Iogar = -cl· (T - Tref )/(c2 + T - Tref )
zation of the complex modulus and complex Pois- 11

son’s ratio by comparing a computed FRF to the c2 120

measured FRF. Figure 7 presents the measured Tref 15 Refereocetemperature("C)

FRF and a FRF computed by using an assumed

stiffness given by the displayed parameter values. Figure 7. Screenshot of the application to optimize
The complex modulus and complex Poisson’s FEM computed FRFs to match measured FRFs (Check
ratio are expressed by the Havriliak-Negami (HN) start values).

212
 100 MPa and 0.5, respectively (Gudmarsson et al.
GJ
Input Checkstartvalues OptimizeFRF PlotsofE*andv* TableofE*andv* Objectivetable

Start values Optimized 2014). This assumption is made since the FRFs
parameters _st~rt.
optimize parameter opt1m1zat1on are not sensitive to low frequency parameter values
lfl El E, = 3.52HO" l!l l!l ,(j). :±J CD ua •., m ..
within a realistic range for asphalt concrete. The
~ alpha 0
~ beta fJ = 0.1785 WLF equation is in the application only used to
[il vl
facilitate the optimization of the τ-parameter in the
ril tauP HN model since this parameter is strongly depend-
Plotfrequency: ~
ent on the temperature. Using the WLF equation, a
l!ll!l ·<i>. :±J '-' • cr"' eo IIII IB change in temperature corresponds to a change in
[] J!'l l'"l m iii
the start value of τ. The evaluation of start values
(see Figure 7) is recommended to facilitate the opti-
mization process and to save computational time.
In the third tab (Optimize FRF), there is a pos-
sibility to select which parameters to optimize and
Figure 8. Screenshot of the application to optimize
FEM computed FRFs to match measured FRFs (Opti-
to start the optimization of the FRFs. The optimi-
mize FRF). zation is performed by using the Bound Optimi-
zation BY Quadratic Approximation (BOBYQA)
solver in COMSOL Multiphysics® (Optimization
model (Hartmann et al. 1994) and the Williams- Module User’s Guide 2016). Figure 8 presents the
Landel-Ferry (WLF) shift factor equation start values and the final optimized values of the
(Williams et al. 1955) to account for the viscoe- parameters in the HN model. The computed FRF
lastic behavior of asphalt concrete. Other models is a result from the complex modulus and complex
can also be used for this purpose. For example, the Poisson’s ratio given by the optimized HN param-
2P2S1D model has also proven to be applicable to eters (see Equation 3 and 4). The computed FRF
characterize the stiffness properties of asphalt con- fits well to the measured FRF in this example.
crete through modal analysis (Gudmarsson et al. There are three additional tabs in the FEM appli-
2014, Olard & Di Benedetto 2003). Equation 2 cation that presents results from the optimization.
presents the WLF model and Equation 3 and 4 Two tabs present plots and tables of the complex
presents the HN models for the complex modulus modulus and complex Poisson’s ratio and one tab
and complex Poisson’s ratio, respectively. shows the objective value for each iteration. The fol-
lowing objective function is used to minimize the dif-
−cc1 ( ) ference between the computed and measured FRF,
(T ) =
ref
e
log , (2)
c2 T − Trrefe
T
N ⎛ H HTi ⎞
Objective = ∑ ⎜
Mi
⎟, (5)
where T is the test temperature, Tref is the reference i =1 ⎜
⎝ H Mi ⎟⎠
temperature and c1 and c2 are material constants.
where HM is the measured FRF, HT is the theoreti-

E * ( ,T ) = E1 +
(E E ) , (3)
cal FRF, N is the number of data points and i is the
index of the data point.
β
[ + (i T (T ) )α ]

v* ( ,T ) = v1 +
(v v)
, (4)
2.3 Materials
β The modal analysis software applications have
⎡⎣1 + (i (T ) p )α ⎤⎦
T been used to determine the complex modulus and
complex Poisson’s ratio of two disc-shaped asphalt
where E0 and v0 are the low frequency values of the concrete specimens with specifications according
modulus and Poisson’s ratio, E1 and v1 are the high to Table 1. The two specimens have the same type
frequency values of the modulus and Poisson’s of aggregates and gradation curve (Stone Mastic
ratio, α governs the width of the loss factor peak, Asphalt) with the nominal aggregate size of 16 mm.
β governs the asymmetry of the loss factor peak, However, the two specimens have different binders;
and τ and τp is the relaxation time of the complex one consist of unmodified conventional penetration
modulus and complex Poisson’s ratio, respectively. grade binder (70/100) and one of modified binder
Note that α and β are assumed the same for the with 7.5% SBS polymers (25/55–80). The specimens
complex modulus and complex Poisson’s ratio. have been cored from a newly built highway that
Six parameters are estimated in the software was not opened to traffic at the time of coring. The
application (E1, v1, α, β, τ and τp), while the low cores were sawn on the top and bottom to obtain
frequency parameters E0 and v0 are assumed to parallel sides. Note that exact measurements of the

213
Table 1. Specifications of disc-shaped asphalt concrete differences in dynamic Poisson’s ratio between the
specimens. two mixes. However, the modified mix give a lower
Specimen 70/100 25/55–80
phase angle and a lower viscous part of the com-
plex Poisson’s ratio compared to the unmodified
Binder type Neat penetration Polymer modified binder. Table 2 presents the complex modulus and
grade 7.5% SBS complex Poisson’s ratio master curve parameters.
Binder content 6.4 6.6
(%) 40 50
(b)
Air voids (%) 3.5 3.6

/:
:w E 4o
Height (mm)
Diameter (mm)
31
148,6
30,7
148,7
;E
·S2, 20
t JO
~ 20
Density (kg/m3) 2346 2336 ~
10
0: 10

IO' JO' IO' 1012 10° 104 108 10

Reduced frequency fHzl uccd frequency fHz1
height and diameter of the specimens are important O. lS
r--------'----'--'--'----'r-------
R-cdr o

for accurate determination of the stiffness. (c) • ~;IVOO~{~:~ ~~%

- Master curve
~ 0.1
~ • 25/55-80 (6.6 % )
E 7.5% SBS
3 RESULTS -
Air voids = 3.6 %
Master curve

The modal analysis were performed at seven tem- 0.2 OA 0.6 0.8
peratures between −20 to 40°C to estimate the mas- E' l max(E')

ter curves. The complex modulus master curves

of the two specimens are presented in Figure 9. Figure 9. Complex modulus of disc-shaped asphalt
concrete specimens with conventional penetration grade
Figure 9a presents the dynamic modulus master binder (70/100) and SBS-polymer modified binder
curves, Figure 9b the phase angle master curves, (25/55–80).
and Figure 9c shows the modulus data at each tem-
perature for the measured frequency ranges as well

v
as the estimated master curves in the Cole-Cole 0.4
space. The loss (E’’) and storage (E’) modulus have _ 0.3
-; -2

been normalized to the maximum storage modulus }

in the Cole-Cole space to compare the loss modu- "
- 0.2
"
~-4
lus part of the complex modulus of the two mixes. 0. 1
(") -6 L_~-~--~----'(h_:_j
i
Figure 8a shows that the polymer modifica- IOo 104 lOs 1012 too 104 I 08 10
tion reduces the high frequency/low temperature r--.;-
R_ cd_u_
cc_d_fr_c.,
q_uc_
n~cy--'r~
H-zl'-------R
-c--,duccd frequency fHz1
dynamic modulus and increases the low frequency/ • 70/ 100 (6.4 %)
Air voids = 3.5 %
high temperature dynamic modulus. Figure 8b -0.01 - Must~·r curvc

shows a reduction of the phase angle for the poly- :~ -0.02 • 25/55-80 (6.6 %)
7.5% SBS
mer modified mix. Figure 8c shows that the vis- Air voids = 3.6 %
-0.03
cous part of the complex modulus is smaller for - Muster curve

the modified mix in comparison to the conven-

035 0_4 0_45 0 _5
tional penetration grade binder. These results of ,;
the complex modulus are indictors of an improved
performance in the polymer modified asphalt Figure 10. Complex Poisson’s ratio of disc-shaped
mixture. In addition, Figure 8c shows that the asphalt concrete specimens with conventional penetra-
estimated master curves in general have a good fit tion grade binder (70/100) and SBS-polymer modified
to the modulus determined at each measurement binder (25/55–80).
temperature. A challenge of modal analysis is to
measure the response of asphalt concrete at high Table 2. Complex modulus and Complex Poisson’s
ratio master curve parameters of the HN and WLF
temperatures when the damping is high. This may model (Tref = 15°C).
be a possible reason for the slight deviation shown
for the specimen with the 70/100 binder at the Specimen E0 (MPa) E1 (MPa) α β τ (s)
highest temperature (40°C) in the Cole-Cole space.
Figure 10 presents the complex Poisson’s ratio, 70/100 100 34354 0.552 0.246 0.024
where Figure 10a show the dynamic Poisson’s ratio 25/55–80 100 32011 0.407 0.258 0.035
master curves, 10b show the phase angle master v0 v1 τ p(s) c1 c2
curves, and 10c the complex Poisson’s ratio at each 70/100 0.5 0.2 0.024 11.4 100.3
measurement temperature and the estimated mas- 25/55–80 0.5 0.2 0.525 16.3 131.3
ter curves in the Cole-Cole space. There are small

214
4 CONCLUSIONS Gudmarsson, A., Ryden, N., Birgisson, B. 2012b. Char-
acterizing the low strain complex modulus of asphalt
The results of the modal analysis using the soft- concrete specimens through optimization of fre-
ware applications displays a good accuracy and quency response functions, J. Acoust. Soc. Am., Vol.
132, Issue 4, pp. 2304–2312.
capability of characterizing the complex modulus Gudmarsson, A., Ryden, N., Di Benedetto, H., Sauzéat,
of asphalt concrete with modified and unmodified C., Tapsoba, N., Birgisson, B. 2014. Comparing linear
binders. The comparison of the two mixes indicate viscoelastic properties of asphalt concrete measured
improved performance of the polymer modified by laboratory seismic and tension-compression tests,
mix due to a lower viscous component shown by Journal of Nondestructive Evaluation, Vol. 33, Issue 4,
the phase angle and the Cole-Cole space. pp. 571–582.
The software application to perform modal test- Gudmarsson, A., Ryden, N., Di Benedetto, H., Sauzéat,
ing provides a tool that does not require any addi- C. 2015. Complex modulus and complex Poisson’s ratio
tional signal processing or data transformation by the from cyclic and dynamic modal testing of asphalt con-
crete, Construction and Building Materials, Vol. 88, pp.
operator in order to accurately measure and calculate 20–31.
FRFs of asphalt concrete specimens. Furthermore, Gudmarsson, A. Resonance Testing of Asphalt Con-
the combination of computing and optimizing FRFs crete, PhD Thesis, ISBN 978-91-87353-50-5, KTH
in a FEM application reduced the computational Royal Institute of Technology, Stockholm, 2014.
time and simplified the practical aspect of character- Hartmann, B., Lee, G.F., Lee, J.D. 1994. Loss fac-
izing the complex modulus through modal analysis. tor height and width limits for polymer relaxations,
The two presented software applications facili- J. Acoust. Soc. Am., Vol. 95, No. 1, pp. 226–233.
tate the application of modal analysis to determine Kweon, G., and Kim, Y.R. 2006. Determination of the
the complex modulus and complex Poisson’s ratio complex modulus of asphalt concrete using the impact
resonance test, J. Transp. Res. Board 1970, pp. 151–160.
of asphalt concrete. Lacroix, A., Kim, Y.R., and Far, M.S.S. 2009. Construct-
ing the dynamic modulus mastercurve using impact
resonance testing, Assoc. Asph. Paving Technol. 78, pp.
ACKNOWLEDGMENTS 67–102.
Levenberg, E. 2011. Smoothing Asphalt Concrete Com-
The Development Fund of the Swedish Construc- plex Modulus Test Data, Journal of Materials in Civil
tion Industry (SBUF), the Swedish Transport Engineering, Vol. 23, No. 5, pp. 606–611.
Administration (Trafikverket) and the Nordic Road MathWorks (2016). User’s Guide (R2016a). Mathemat-
Association (NVF) are acknowledged for the finan- ics, Fourier Analysis and Filtering, Discrete Fourier
cial support. Hervé Di Benedetto and Cédric Sauzéat Transform.
are gratefully acknowledged for their support in the Migliori, A., Sarrao, J.L., Visscher, W.M., Bell, T.M., Lei,
M., Fisk, Z., Leisure, R.G. 1993. Resonant ultrasound
validation of the modal testing methodology. spectroscopic techniques for measurement of the elas-
tic moduli of solids, Physica B, 183, pp. 1–24.
Nguyen, Q.T., Di Benedetto, H. Sauzéat, C., Tapsoba,
REFERENCES N. 2013. Time Temperature Superposition Principle
Validation for Bituminous Mixes in the Linear and
ASTM C215-08, Standard Test Method for Fundamental Nonlinear Domains, Journal of Materials in Civil
Transverse, Longitudinal, and Torsional Frequencies Engineering, Vol. 25, No. 9, pp. 1181–1188.
of Concrete Specimens (American Society for Testing Olard, F., Di Benedetto, H. 2003. General “2S2P1D”
and Materials, West Conshohocken, PA), 2008. Model and Relation Between the Linear Viscoelastic
Bjurström, H., Gudmarsson, A., Ryden, N., Stark- Behaviours of Bituminous Binders and Mixes, Road
hammar, J. 2016. Field and laboratory stress-wave Mater. Pavement, Vol. 4, No. 2, pp. 185–224.
measurements of asphalt concrete, Construction and Optimization Module User’s Guide, COMSOL Multi-
Building Materials, Vol. 126, pp. 508–516. physics® v.5.2. COMSOL AB, Stockholm, Sweden,
Buchanan, J.L. 1987. Numerical solution for the dynamic 2016.
moduli of a viscoelastic bar, J. Acoust. Soc. Am., Vol. Ryden, N. 2011. Resonant frequency testing of cylindri-
81, pp. 1775–1786. cal asphalt samples, Eur. J. Environ. Civ. Eng., Vol. 15,
Daniel, J., Chehab, G., Kim, Y.R. 2004. Issues Affecting pp. 587–600.
Measurement of the Complex Modulus of Asphalt Visscher, W.M., Migliori, A., Bell, T.M., Reinert, R.A.
Concrete, J. Mater. Civ. Eng. 16, Special Issue: 1991. On the normal modes of free vibration of inho-
Strengthening of Concrete Structures with Advanced mogeneous and anisotropic elastic objects, J. Acoust.
Composite Materials-Prospects and Problems, Soc. Am. 90(4), pp. 2154–2162.
pp.469–476. Whitmoyer, S.L., and Kim, Y.R. 1994. Determining
Demarest, H.H. 1971. Cube-resonance method to deter- asphalt concrete properties via the impact resonant
mine the elastic constants of solids, J. Acoust. Soc. method”, J. Test. Eval. 22(2), pp. 139–148.
Am. Vol. 49, pp. 768–775. Williams, M.L., Landel, R.F., Ferry, J.D. 1955. The tem-
Gudmarsson, A., Ryden, N., Birgisson, B. 2012a. Appli- perature dependence of relaxation mechanisms in
cation of resonant acoustic spectroscopy to asphalt amorphous polymers and other glass-forming liquids,
concrete beams for determination of the dynamic J. Am. Chem. Soc., Vol. 77.
modulus, Mater. Struct., Vol. 45, pp. 1903–1913.

215
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

An implementation of the Illinois flexibility index testing protocol for

balanced asphalt mix designs

M.K. Barry, I.L. Al-Qadi, H. Ozer & F. Safi

University of Illinois at Urbana-Champaign, Urbana, IL, USA

ABSTRACT: The Illinois Flexibility Index Test, or I-FIT, was developed specifically to analyze the
flexibility and cracking potential of asphalt mixtures. The introduction of recycled materials, such as
Reclaimed Asphalt Pavement (RAP) and Recycled Asphalt Shingles (RAS), has increased mixture brit-
tleness to unacceptable levels, leading to increased cracking. The I-FIT method is ideal for identifying
potentially brittle mixes and analyzing acceptable flexibility in recycled mixes.
The results from mixes with varying levels of RAP and RAS content is presented in this paper. Mate-
rials with low recycled content often have Flexibility Index (FI) values above 10, while mixes with high
recycled content and proven high cracking potential often haveFI values below 4. The I-FIT method
combined with the Hamburg Wheel Tracking (HWT) test, which quantifies rutting potential, can ensure
a balanced mix design, which will resist both rutting and cracking failure.

1 INTRODUCTION 2 MATERIALS AND METHODS

Asphalt material surface failure typically occurs in 2.1 Surface AC mixes

one of two modes. At high temperatures, asphalt
Eleven surface AC mixes were included as part
materials may permanently deform under loads,
of a project titled “Construction and Perform-
which is known as rutting. Widely accepted as a test
ance Monitoring of Various Asphalt Mixes in
of high temperature rutting failure is the Hamburg
Illinois”(Lippert et al., 2016). All projects were
Wheel Tracking test (HWT) (AASHTO, 2015a).
located within 40 miles of each other, in Joliet
At low temperatures, materials might crack under
and the western and south suburbs of Chicago.
thermal contraction; while at service temperature,
This minimizes any environmental effects on the
materials might experience fatigue cracking due to
performance.
heavy repeated loading.
The 11 surface AC mixes had varying types
There are various test methods to evaluate
and levels of ABR, with basic details presented
cracking resistance of Asphalt Concrete (AC)
in Table 1. This permits the study of the effect of
mixes. The flexibility index or Illinois Flexibil-
various types and levels of recycled materials.
ity Index Test (I-FIT) protocol was developed as
a simple yet reliable laboratory test method to
2.2 Illinois Flexibility Index Test (I-FIT)
analyze AC mixture flexibility to quantify the
mixture’s cracking potential (Al-Qadi et al., 2015 All plant AC mixes were evaluated via the I-FIT
and Ozer et al., 2016a). The Flexibility Index (FI) method in accordance with AASHTO TP124 and
is ideal for characterizing AC mixes with high Illinois IL-45 specifications (AASHTO, 2015b).
Asphalt Binder Replacement (ABR) because high Semicircular Bending (SCB) geometry which
levels of Reclaimed Asphalt Pavement (RAP) and utilizes semicircular specimens of 150-mm diam-
Recycled Asphalt Shingles (RAS) decrease flex- eter and 50-mm thickness. I-FIT method uti-
ibility and increase cracking potential when aged lizes Superpave gyratory specimens fabricated
binder is used with no softer binder or rejuvenator. in the SCB geometry. A notch is placed at the
Plant AC mix was tested for 11 surface AC center face (at ∼10% of specimen diameter) and
mixes in order to analyze the flexibility and rutting loaded in three-point bending fashion in a strain-
resistance of high-ABR and Total Recycle Asphalt controlled test at 50-mmper minute at an inter-
(TRA) mixes. The results are then analyzed from a mediate temperature of 25ºC. Load response
balanced design perspective. curves are analyzed to find their fracture energy

217
Table 1. Summary of surface AC mixes.

Virgin binder Crushed Steel

Mix ID N-Design grade RAP (%) RAS (%) ABR (%) concrete (%) slag (%)

157M N70 58–28 9.9 5.0 29 None None

156M N70 64–22 4.9 2.5 15 None None
140M N70 58–28 20 3.1 30 None None
159M (Y02) N70 58–28 34 None 29 None None
185M (N08) N70 52–34 39 5.0 48 29 27
185M (N07) N70 52–28 39 5.0 48 29 27
177M N70 58–34 20 3.1 30 None None
159M (Y04) N70 58–34 34 None 29 None None
137M N50 52–28 51 4.6 60 30 15
338 K N50 52–28 53 5.0 56 27 15
138Z N50 58–28 27 None 29 None 73

and their shape by calculating the slope at the

post-peak inflection point. Four specimens were
tested for I-FIT and compacted at target air voids
of 7 ± 0.5%.
Fracture energy is defined as the area under-
neath the load displacement curve, normalized to
the ligament area, which is the area over which the
crack propagates. High fracture energy is benefi- Figure 1. Pointed VS Rounded Fracture Curves.
cial to AC cracking resistance.
Fracture energy is insufficient to quantify I-FIT
fracture curves because curves of different shapes Resultant rut depths are measured with Linear
may display equivalent fracture energy. More Variable Differential Transformers (LVDTs). Rut
rounded curves are more ductile (beneficial), while depth as a function of wheel tracking cycles is
more pointed curves are more brittle (detrimental, analyzed to determine the rut resistance of the AC
shown in Figure 1). The slope at the post-peak mix.
inflection point is used to quantify the shape of
I-FIT fracture curves. Low slopes represent more
ductile fracture and are beneficial to cracking 3 I-FIT RESULTS
resistance(Ozer et al., 2016b).
These two parameters are combined in the FI, Several hundred pounds of each plant AC mix were
presented in Equation 1. High FI values indicate sampled and transported to the Illinois Center for
increased cracking resistance. For reference, the Transportation (ICT) in Rantoul, Illinois. Table 2
Illinois Department of Transportation (IDOT) presents I-FIT results, including fracture energy,
considers a mix with an FI of 8.0 or higher a slope, and FI for each surface AC mix.
“good” or “acceptable” mix. Virgin binder softening is shown to be an effec-
tive method of increasing FI. PG 64–22 was shown
Fracture Energy (G f ) to be an improper binder grade to use for ABR
FI = *A mixes in the Illinois climate, according to Mix 156
Absolute Value oof Slope (abs( m )) results.
(1) Mixes with no RAS performed better than
A = 0.. ( fo
f r more convenient index values) those with RAS present. AC mixes with moderate
levels of ABR solely from RAP showed good FI
values. For AC mixtures with moderate RAS levels
2.3 Hamburg wheel tracking test
(e.g., 3.1%), FI can be brought to acceptable levels.
The Hamburg Wheel Tracking (HWT) device However, in certain cases, such as those with high
evaluates rutting potential of AC specimens. A RAS content (e.g., 5%), binder softening may be
47-mm-wide steel wheel, with a load of 158 lb is inadequate.
repeatedly tracked across the AC specimens sub- All TRA mixes (those with near or above 50%
merged in 50°C water, until either 20,000 cycles or ABR) had low FIs. Despite attempts to mitigate
a 20-mm rut depth. stiffness though binder softening, these AC mixes

218
Table 2. I-FIT Results.

Mix Design ID Fracture Energy (joules/m2) Slope (kN/mm) FI

156M 2095 −4.5 4.8

157M 1912 −5.5 3.5
140M 1791 −2.7 6.6
159M (Y02) 2138 −2.2 10.4
185M (N08) 1271 −2.9 4.7
185M (N07) 1396 −2.0 7.2
159M (Y04) 1739 −1.6 11.9
177M 1872 −1.8 10.6
137M 1686 −4.5 3.8
338 K 1146 −13.5 0.9
138Z 2476 −4.9 4.1

experienced low FIs. Finally, high steel slag con- to withstand cracking. TheI-FIT method and
tent was shown to significantly increase the frac- HWT test, when combined, are ideal for evaluat-
ture energy, which occurs due to strong bonding ing the balanced mix design properties of asphalt
between steel slag and binder content. Also, the mixtures.
use of crushed concrete could reduce the flexibility Investigation of I-FIT results suggests that
index, however, further investigation is needed to mixes with higher recycled content display lower
confirm this statement. FIs, and thus lower cracking resistance. HWT
results, on the other hand, display lower rut depths
for higher recycled content mixes, indicating
4 HAMBURG WHEEL TRACKING higher rut resistance. A trade-off exists between
RESULTS the parameters, with increased cracking resistance
resulting in decreased rutting resistance and vice
All 11 plant mixes reach 20,000 passes without versa.
reaching a 20-mm rut depth. Therefore, the rut To effectively present this trade-off, the Rutting
depth (mm) at 20,000 passes is a proper metric Resistance Index (RRI) is introduced (Wen et al.,
by which to judge rutting resistance. For refer- 2016) and presented in Equation 2.
ence, IDOT considers failure to be a 12.5-mm rut
depth, with the minimum allowable passes to fail- I N×( .4 − RD )
ure dependent on the climate. The hottest climates
require the highest minimum allowable passes to N N Number oof Cycles to Completio
C n (2)
failure, 20,000 passes (Osman, 2014). Three speci- RD = Rut Depth ( mm )
mens having diameter of 150-mm and thickness
of 60-mm were compacted at target air voids of
7 ± 1%.Table 3 presents HWT results for each sur-
face AC mix. Table 3. Hamburg wheel tracking results.
All surface AC mixes pass even the most strin-
gent IDOT requirements. Mixes with softer virgin Mix Design Rut Depth @ 20,000
binder grades display higher rut depths. Higher Passes (mm)
levels of ABR lead to lower rut depths, with TRA
mixes performing to an even higher standard than 156M 2.64
typical ABR mixes. 157M 3.14
140M 3.86
159M (Y02) 4.60
185M (N08) 4.77
5 BALANCED MIX DESIGN ANALYSIS
185M (N07) 6.48
159M (Y04) 8.60
The concept of balanced mix design suggests that
177M 9.21
mixes must be designed against both rutting at 137M 4.83
high temperatures and cracking at low tempera- 338 K 2.09
tures. A mix design must be of proper stiffness 138Z 4.51
to withstand rutting but be of proper flexibility

219
 Because all of the AC mixes achieved 20,000 ing both requirements, and will likely resist both
passes, the RRI is directly proportional to the rut cracking and rutting. Mixtures in the Stiff & Brit-
depth at 20,000 passes. However, it does convert rut tle quadrant risk cracking failure but should not
depth to a parameter that increases with increased experience rutting issues. Mixtures in the Soft &
rut resistance, which provides a more intuitive Flexible quadrant risk rutting failure but should
comparison with the FI. Figure 2 presents the FI not experience cracking issues. Mixtures in the
compared with the RRI for each mix, clearly dis- Soft & Unstable quadrant are undesirable, failing
playing the trade-off between the two parameters. both requirements, and may fail in both cracking
The trade-off displayed in Figure 2 validates the and rutting. Figure 3 presents the 11 surface AC
concept of balanced mix design. Designing overly mixes imposed on an interaction plot.
stiff mixtures with high rutting resistance may lead High recycled content led to stiffer designs,
to low cracking resistance, and designing overly which resulted in excellent rutting performance for
flexible mixtures with high cracking resistance may each of the mixes. All AC mixes tested are to the
lead to low rutting resistance. With this in mind, left side (beneficial) of the rutting cut-off in the
it is possible to achieve a balance between the two table.
modes of failure. Unfortunately, most mixes are too stiff, push-
Interaction plots between rut depth and FI ing them into the Stiff & Brittle quadrant. In this
visualize the balanced mix design properties of quadrant, mixes should not experience rutting
AC mixes. Mixtures are separated into quadrants, issues, but they are likely prone to cracking failure.
with an FI cut-off of 8.0 and a rut depth cutoff Mixes with high ABR content are typically likely
of 12.5-mm at 20,000 passes. Mixtures in the to be in this quadrant.
Stiff & Flexible quadrant are desirable, exceed- However, three mixes fall into the desirable
Stiff & Flexible quadrant, where mixes are likely
to resist rutting and cracking. Mixes 159(Y02) and
159(Y04) both fall in this quadrant. These mixes
have 29% ABR, which, unlike most of the mixes, is

•..... ········-~..
...
.. solely from RAP, with no RAS present. Addition-
ally, Mix 177 falls in the desirable quadrant. This
mix has 30% ABR, with a moderate level of RAS
(3.1%). It also has a softer virgin binder than other
•······..
e e ··· ..... R' = 0.6508
·· ..
• mixes with moderate RAS levels (i.e., Mixes 156
and 140).
... ...
• ... ...
......
The balanced mix design analysis suggests that
high ABR mixes can fall in the desirable quad-
··... rant if designed properly. Generally, mixes with
• • higher RAS levels (i.e., >3%) fall in the Stiff &
Brittle quadrant and are likely prone to cracking.
0 4 6 8 10 12
However, Mix 177 shows that mixes with moder-
Fl
ate levels of RAScan fall in the desirable Stiff
& Flexible quadrant when properly soft virgin
binder is employed. Mixes with no RAS, and ABR
Figure 2. FI vs. RRI.
solely from RAP, are shown to fall in the desirable
quadrant.

14 -
6 SUMMARY AND CONCLUSION
12
STIFF & FLEXIBLE
e
.-
[CELLRAN GE]

[CELLRANG EJe [CELLRAN(

SOFT&
l FLEXIBLE
!;':_ 10
X
Eleven high ABR AC mixes were tested for their
Q)
"0
8 rutting and cracking resistance. The I-FIT method
E
.~
.-!CELLR$JGE] was employed for cracking resistance, and the
:0 [CELLRANGE] , j [CELLRANGE] HWT was employed for rutting resistance.
·;;: e [CELLRANGE] SOFT&
(CELLR~GE~ELLRANGE]STIFF & I-FIT testing yielded several interesting con-
Q) 4
u:: UNSTABLE
clusions. Most mixes fell below the acceptable FI
0
e-- [CELLRANGE] BRrJl"LE threshold due to high levels of stiff ABR. Binder
0 5 10 15
softening was found to be an effective method of
Rut Depth (mm) @ 20,000 Passes increasing the FI. Mixes with no RAS present
were found to have higher FIs than when RAS
Figure 3. FI and rut depth interaction plot. was present, even if ABR levels were equivalent.

220
If levels of RAS are moderate, FIs can be brought American Association of State Highway and Transporta-
to acceptable levels via binder softening. However, tion Officials (2015b).AASHTO TP124. Determining
high levels of RAS lead to unacceptably low FIs. the Fracture Energy Parameters of Asphalt Mixtures
Using the Semicircular Bend Geometry (SCB) at
All AC mixes performed well in the HWT. Recy-
Intermediate Temperatures.
cled materials lead to stiffer designs, with rutting Lippert, D.L., H. Ozer, I.L. Al-Qadi, J.F. Meister,
failure not being a concern in any of the 11 AC G. Renshaw, ASM T.U. Khan, T.R. Murphy, J.S.
mixes. Trepanier, and J.W. Vespa (2016). Construction
A loose but relatively clear trade-off between FI and Performance Monitoring of Various Asphalt
and RRI is presented. This further validates the Mixes in Illinois: 2015 Report. Publication FHWA-
concept of balanced mix design and the trade-off ICT-16–009. Illinois Center for Transportation.
that exists between rutting and cracking resistance. Osman, O.M. (2014). Special Provision for Hot-Mix
Interaction plots between FI and rut depth Asphalt—Mixture Design Verification and Pro-
duction. Memorandum, Illinois Department of
show that it is possible for recycled mixes to display
Transportation.
desirable performance parameters. Limiting RAS Ozer, H., Al-Qadi, I.L., Lambros, J., El-Khatib, A.,
use and employing proper virgin binder softening Singhvi, P., & Doll, B. (2016a). Development of the
are promising methods of increasing the FI—and fracture-based flexibility index for asphalt concrete
subsequently placing recycled mix designs in the cracking potential using modified semi-circle bending
desirable Stiff & Flexible quadrant. test parameters. Construction and Building Materials,
115, 390–401.
Ozer, H., and P. Singhvi, T. Khan, J. Rivera, I.L. Al-Qadi.
REFERENCES (2016b). Fracture Characterization of Asphalt Mix-
tures with RAP and RAS using the Illinois Semi-
Circular Bending Test Method and Flexibility Index.
Al-Qadi, I.L., H. Ozer, J. Lambros, A. El-Khatib,
Transportation Research Record: Journal of the
P. Singhvi, T. Khan, J. Rivera-Perez, and B. Doll
Transportation Research Board, 2575, 130–137.
(2015). Testing Protocols to Ensure Performance of
Wen, H., S. Wu, L.N. Mohammad, W. Zhang, S. Shen,
High Asphalt Binder Replacement Mixes Using RAP
and A.F. Faheem (2016). Long-Term Field Rutting
and RAS. Publication FHWA-ICT-15–017, Illinois
and Moisture Susceptibility Performance of Warm
Center for Transportation.
Mix Asphalt Pavement. Transportation Research
American Association of State Highway and Transpor-
Record: Journal of the Transportation Research Board,
tation Officials (2015a). AASHTO T324. Hamburg
No. 2575, Transportation Research Board of the
Wheel-Track Testing of Compacted Hot Mix Asphalt
National Academies, Washington, D.C.
(HMA).

221
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Laboratory evaluation of stiffness and fatigue susceptibility of asphalt

paving materials incorporating environmental factors

Ahmed Aljuboryl
Nottingham Transportation Engineering Centre, University of Nottingham, Nottingham, UK
AL-Mustansiriya University, Baghdad, Iraq

G.D. Airey & J.R.A. Grenfell

Nottingham Transportation Engineering Centre, University of Nottingham, Nottingham, UK

ABSTRACT: Fatigue life assessment of bituminous mixtures has a great significance in the design of asphalt
pavements which may be more susceptible to fatigue cracking if the asphalt binder is prone to environmental
issues such as ageing and moisture damage. Therefore, it is important to incorporate these environmental
conditions into the assessment to ensure an accurate in service pavement fatigue life prediction. Currently, not
enough data exists regarding the moisture effects on the fatigue susceptibility of asphalt concrete mixtures. In
addition, ageing has not been considered in any detail in the current fatigue life prediction relationship. Conse-
quently, this study aims to investigate the susceptibility of asphalt concrete mixtures to fatigue damage incor-
porating ageing and moisture conditions. The effects of ageing and moisture using British Board of Agrément
(BBA) protocols on the stiffness and fatigue of asphalt mixtures were carried out using ITSM (Indirect Tensile
Stiffness Modulus) and ITFT (Indirect Tensile Fatigue Test) in strain controlled mode respectively. After mois-
ture conditioning, the ITSM test results show that the stiffness of control asphalt mixes increased significantly
at 20°C and slightly at 10°C. Similarly, there was a slight increase in the stiffness of wet-aged asphalt mixtures
at both temperatures. On the other hand, ITFT results show a decrease of the control and aged asphalt mix-
tures fatigue life at 10°C and 20°C due to moisture effects compared to that of dry condition. Finally, there was
an observed increase in the fatigue life when those mixtures become stiffer due to LTOA.

1 INTRODUCTION studies have investigated the effects of different fac-

tors; different material constituents, testing condi-
The fatigue cracking phenomenon is one of the tions and testing methods to quantify the fatigue life
most significant detrimental distresses in asphalt of asphalt concrete mixtures (Baburamani, 1999, Di
pavement industry, which takes place in pave- Benedetto et al., 2003, Miller and Bellinger, 2003,
ments due to repeated stresses or strains resulting Maggiore, 2014). In fact, this literature may not ade-
from the repetitive applications of traffic loads. quately identify some of the factors and conditions
This phenomenon will be accelerated when the that affect the real in-service performance. In other
repeated traffic loading is combined with different words, environmental factors such as moisture, age-
environmental conditions, such as the presence of ing and their combined effect have not been consid-
water, oxidation and temperature fluctuation, and ered enough in the current fatigue life studies.
with improper paving materials or inefficient con- Based on these findings, this study is an ambi-
struction quality. Therefore, new perspectives are tious programme to explore more about the stiffness
seriously required to associate and understand the and fatigue of asphalt paving materials incorporat-
effects of these environmental issues on the fatigue ing environmental factors using the Indirect Tensile
prediction of flexible pavement structures. Tests; ITSM and ITFT in strain controlled mode.
Regarding moisture effect, it is possible that the In particular, it aims to understand the fatigue sus-
absorption of moisture by asphalt mixtures may ceptibility of asphalt mixtures in the presence of
result in a stiffness degradation and in turn a poten- individual and combined impacts of ageing and
tial reduction of fatigue life, but this has not been moisture conditions.
conclusively demonstrated. In addition, ageing has
not been considered enough in the current fatigue
1.1 Water effect on asphalt concrete mixture
life prediction relationship or in the suggested shift
fatigue resistance
factors. Probably, if the asphalt binder ages and
stiffens excessively, the asphalt mixture may be more The structural integrity of asphalt concrete can be
susceptible to fatigue cracking. Currently, extensive deteriorated by water damage through either the

223
cohesion loss of bitumen or the failure of adhesive occurs during asphalt mixture production and
bond between the bitumen and aggregate (Kennedy identify as long-term ageing when it occurs during
1985, Airey and Choi 2002, Huang et al., 2010). the in-situ service period of asphalt roads. (Glover
Epps and Monismith (1972) indicated that no data et al., 2005, Walubita et al., 2005, Wu, 2009).
are available on the effects of moisture on the fatigue Vallerga et al. (1967) tested mixtures containing
resistance of asphalt pavement. They supposed that aged and un-aged asphalt materials in controlled
the action of water on asphalt mixtures may lead stress mode and found those mixtures with high
to increase the rate of stiffness modulus reduction stiffness values and harder bitumen possessed
and in turn a potential decrease in fatigue resistance. longer fatigue lives, although the slopes of the
This assumption had not been conclusively demon- strain versus fatigue life relationships were nearly
strated at that time. More recently, Virgili et al. (2008) identical for all tested mixtures. A similar relation-
indicated that conventional techniques commonly ship between asphalt penetration and fatigue life
used for asphalt testing do not take into considera- have presented by Jimenez and Gallaway (1962).
tion the combined action of environmental factors; Harvey et al. (1995) evaluated the effect of dif-
temperature and water, and repeated loading which ferent binder (4.0, 4.5, 5.0, and 5.5%) and air void
may affect adversely mix design performance. They (4 to 6% and 7 to 9%) contents on the fatigue
found that the fatigue damage process of asphalt life of a typical California asphalt concrete mix-
pavements can considerably be accelerated by envi- ture using a three-point flexural test at 150 and
ronmental factors; water and temperature. However, 300 micro-strain, a frequency of 10 Hz and at a
water effects being more detrimental than the tem- temperature of 19 ± 1°C. They tentatively found
perature. Barra et al. (2012) found that the action that the applied long-term ageing condition of
of water can lead to a conclusive reduction of the (0, 3, and 6 days) has a slight effect on the labora-
fatigue life of asphalt mixes and lead them to rupture tory fatigue life, although asphalt concrete mix
before the end of the predicted designed period. Fur- stiffness increased. In addition, they indicated that
thermore, they observed that the moisture damage ageing was affected by air void content to a great
is more intense by including the effect of tempera- extent. The lower air voids level of 1% to 3% have
ture, increasing the potential of asphalt binder film the higher resistance to ageing and the susceptibility
removal from the aggregate particles. In addition, of asphalt mix to ageing increased at 7% to 10% air
Jaskula and Judycki (2014) found that the decrease void content.
of the fatigue life of open graded asphalt base course In another study, Harvey and Tsai (1997) found
material by the water and frost detrimental actions that the effect of ageing on the predicted fatigue
could be critical to the service life of the asphalt pave- life was binder type dependent. For Valley asphalt
ment. They indicated that the observed decrease of mixes (PG 58-10 or PG 64-10), there was no signif-
fatigue life of the asphalt base course should be taken icant effect of LTOA on beam fatigue life as shown
into account to precisely determine the long–term in the first study. For Coastal asphalt mixes (PG
performance of pavement structures. 58–28), 6 days of LTOA significantly decreased
beam fatigue life. Walubita et al. (2005) stated
that ageing of asphalt concrete mixtures should
1.2 Ageing effect on asphalt concrete mixture
be incorporated in fatigue analysis and design
fatigue resistance
because they found that ageing reduced the ability
Ageing of bituminous binder is a phenomenon of asphalt mixtures to heal and in turn reduced the
which can significantly affect the paving materials fatigue resistance.
durability in long-term performance. This generally Walubita (2006) investigated the effect of
takes place because the asphalt binder experiences binder oxidative ageing incorporating some vari-
a gradual hardening (ultimately perpetual harden- ables on the fatigue of Hot Mix Asphalt (HMA)
ing) due to two factors. The first factor is the oxi- under strain-controlled laboratory testing. It was
dation process. Oxidation is a chemical reaction concluded that HMA resistance to fracture and
that happens when the oxygen in the air attacks its ability to heal has been reduced due to binder
the asphalt material resulting in more polar com- oxidative ageing. It was also found that the fatigue
pounds of strong associations with each other. resistance of HMA is both environmental condi-
Due to this process, the binder stiffness modulus tions and pavement structure dependent.
increases accompanied with a higher viscosity (bet- Finally, Jung et al. (2007) investigated the binder
ter resistance to flow). Volatilization is the second oxidation effects on the fatigue response of hot-
factor which affects the hardening of bitumen. It is mix asphalt measured under laboratory strain-
the evaporation of the lighter hydrocarbons, volatile controlled conditions. It was concluded that the
(unstable) or oily constituents, from the bituminous laboratory measured strain control fatigue resist-
material to the air or from aggregates or from both. ance decreases significantly due the binder oxida-
This loss may identify as short-term ageing which tion hardening.

224
 100

90
80
70

60 IL------,_J
50

30 t----,-+--+---~

20
10

0.1 10 100
Sleveslze,mm

Figure 1. Gradation curve for 14 mm DBM.

Figure 2. Indirect Stiffness Modulus Test (ITSM).

2 EXPERIMENTAL PROGRAM The captured signals from the transducers were

used to calculate the stiffness modulus as a function
2.1 Material of vertical load, horizontal deformation, specimen
In this research project, the experimental pro- dimensions and an assumed Poisson’s ratio for the
gram was carried out on limestone aggregate of tested material. The relationship used to calculate
a continuously graded Dense Bitumen Macadam stiffness modulus Sm (in MPa) is as follows:
(DBM) asphalt mixture with a maximum aggre-
gate size of 14 mm, in accordance with the speci- F
fication for 14 mm size binder course mixtures in Sm = ( + .27 )
BS 4987-1:2005 standard as shown in Figure 1. An z h
unmodified 40/60 penetration grades bitumen was
used according to the same standard. where F is the peak value of the applied vertical
load (N); z is the peak horizontal diametric defor-
mation resulting from the applied load (mm); h is
3 TESTING PROGRAMME the mean thickness of the test specimen (mm); and
υ is Poisson’s ratio for the bituminous mixture at
The Indirect Tensile Stiffness Modulus test (ITSM)
the test temperature.
was performed to determine the stiffness modulus
During the ITSM test, ten conditioning pulses
of DBM asphalt mixtures. In addition, the Indirect
are followed by five test pulses, which are used to
Tensile Fatigue Test (ITFT) in strain control mode
determinean average stiffness modulus. The test is
was carried out in order to find the fatigue resist-
then repeated after rotating the specimen through
ance of the cylindrical core specimens at different
90° and the average stiffness from the two tests is
conditions and variables.
recorded as the indirect stiffness modulus of the
asphalt mixture specimen.
3.1 Indirect Tensile Stiffness Modulus (ITSM)
The ITSM test was used to determine the stiffness
modulus of the cylindrical specimens conform- 3.2 Indirect Tensile Fatigue Test (ITFT) in strain
ing to BS EN 12697-26: 2004. In this test, a com- control mode
pressive load pulse is applied by means of a load
In general, no permanent deformation is involved
actuator along the vertical diameter of a cylindri-
in pure fatigue tests such as four point and two
cal specimen. The resultant (indirect tensile) peak
point bending beam tests which are characterised
transient deformation is measured along the hori-
by a sinusoidal loading waveform about the zero
zontal diameter using Linear Variable Displace-
stress point. Consequently, the developed ITFT
ment Transducers (LVDTs), whereas the vertical
in strain control mode based on BS EN 12697-24:
load is measured using a load cell as shown in
2004 was designed on recoverable deformation to
Figure 2.
not include the permanent deformation that iden-
The following test parameters were applied in
tified as a main drawback in the traditional ITFT
this test:
(Maggiore, 2014).
a. Test temperature: 10°C and 20°C In this project, it was decided to carry out the
b. Poisson’s ratio: 0.35 developed ITFT in strain control mode to over-
c. Loading rise-time: 124 ms come the permanent deformation issue of the
d. Peak transient horizontal deformation: 5 μm. conventional ITFT. In the new test, the testing

225
equipment, the Nottingham Asphalt Tester (NAT) lier if 9 mm vertical deformation occurs. The whole
shown in Figure 3 is exactly the same as the one fracture along the specimen’s vertical diameter will
used in the ITFT in stress control method while the not occur at failure as in the case of a stress control-
specimen preparation involves gluing two defor- led mode. It is typically characterised by an appar-
mation strips by using a position rig on the oppo- ent crack located along the vertical plane of the
site sides of the horizontal diameter plane. These specimen and with a localised deformation at the
two deformation strips have a concave surface to edge of the two loading strips as shown in Figure 7.
hold two LVDTs located horizontally to capture
the variation of the horizontal diameter by meas- • Rl: Remver.ble df'focm.Iim for !<:rrm IIIH- cootrd lfix- wet mmit:ion
uring the variation of the distance between the two • R2: ReoJverabledEflll'n.I:im for l<lrrmll!H- ,artra mix- !lycmdtlln

glued deformation strips as shown in Figure 4. • RJ: Remver.bledH!Trmtim for l<lrrmii!H-LTGA.nU- wit cauit[m

This fatigue test is carried out by applying a

constant deformation, the constant parameter
.. • R4: Reawcrableddorm.I:1M forl<~rrvnOOM-LT~nix- drycoodtlion

~
of the test, with the recoverable deformation that
occurs during the unloading time of one pulse. .. I I"
Typically, the single pulse of 0.5 sec consists of
a loading time of 0.1 sec followed by 0.4 sec rest ~! I R2

:~
time. Permanent deformation is the amount of
deformation that is unrecoverable after each sin-
gle pulse and it is calculated as the difference
"'
150
··! :
between the total deformation and the recover-
able deformation as shown in Figures 5 and 6.
,., lOO
"'
..,
Time (ms)
600 700 ..
Finally, the failure criteria in this test should be
specified by the user in terms of the stiffness reduc- Figure 5. Recoverable deformation for 14 mm DBM at
tion percentage but sometimes the test may end ear- 20°C and 200 με.

13P2: Penn""ent ~~or~(mfo• 14rrm DBH- control nu- oty rondtion

!IP3: Pe~manentdefor""'=onfor 14rrm DBH- LTOAmx- ..et oo~dbon

1:11'4: Pe~manent deformahmfor 14mm DBH- LTOA mx- «y condi:ton

~~====================~

10

15

10

Figure 6. Permanenthorizontal deformation for 14 mm

DBM at 20°C and 200 με.
Figure 3. Indirect Tensile Fatigue Test (ITFT).

Figure 4. The setup of two deformation strips to hold Figure 7. Mechanism of failure in ITFT strain control-
two horizontal LVDTs. led test.

226
4 AGEING CONDITION cool after LTOA. After that they were collected
once more to measure the stiffness modulus and
4.1 Short-Term Oven Ageing (STOA) fatigue characteristics.
Short-Term Oven Ageing was applied to the loose
mixtures at 135 ± 3°C for 4 h ± 5 minutes before
5 MOISTURE CONDITIONING
compaction according to AASHTO PP2, 1994 as
shown in Figure 8. After that the loose mixtures
Moisture conditioning was considered in this study
were compacted using a roller compacter. In order
as one of the main variables that affects the stiff-
to evaluate the effect of STOA on the progress of
ness of asphalt mixtures and in turn fatigue resist-
stiffness modulus from the virgin condition to LTOA
ance. The effect of moisture was evaluated for the
circumstances, ITSM of short-term aged specimens
control asphalt mixtures (un-aged) and after apply-
was measured before moving the compacted speci-
ing LTOA to asphalt mixtures. This conditioning
mens to the Long-Term Oven Ageing (LTOA) stage.
was carried out according to British Board of
Agrément (BBA)—Appendix A.2: 2008 as follows:
4.2 Long-Term Oven Ageing (LTOA)
1. Determine the unconditioned stiffness charac-
This conditioning was carried out in accordance teristics of the specimen in accordance with BS
with AASHTO PP2, 1994 and British Board of 12697-26:2004 for ITSM,
Agrément (BBA) - Appendix A.12: 2008. The 2. Place the specimens in the vacuum desiccator,
ageing simulates the hardening of the bitumen cover it with distilled water at 20(±1)°C, seal
in the mixture subsequent to construction. The the apparatus, and apply a partial vacuum of
practice should result in ageing representative 510(±25)mm Hg or 680(±33) mbar for 30(±1)
of 5 to 10 years in service for dense-graded mix- minute. Remove the specimen from the vacuum
tures. The prepared specimens were collected and desiccator,
supported in the wire mesh baskets, then they 3. Place it in a hot water bath at 60(±1)°C for
were placed on a rack in the forced-draft oven at 6(±1) hours. The samples may be supported by
a temperature of 85(±2)°C for 120(±0.25) hours a wire mesh baskets if necessary to minimise
shown in Figure 9. The specimens were left to distortion.
4. Remove the specimen from the hot water bath
and immediately place it in a cold water bath at
5(±1)°C for 16(±1) hours.
5. Remove the specimen from the cold water bath
and immediately place it in a water bath at
20(±0.5)°C (or at the specified testing tempera-
ture) for at least 2 hours.
6. Determine the conditioned stiffness/fatigue
characteristics of the specimen in accordance
with BS EN 12697-26:2004 for ITSM and BS
EN 12697-24:2004 for ITFT.

Figure 8. Short Term Oven Ageing (STOA) of loose

mixture at 135°C. 6 TEST RESULTS

6.1 Effect of ageing on the ITSM of asphalt

mixtures
In this work, STOA and LTOA was applied on the
cored specimens and the indirect tensile stiffness
was measured for both conditions and the ITSM
ratio for each specimen is expressed in terms of
ageing index as follows:

Ageing Index = ITSMLTOA/ITSMSTOA

In addition the ITSM was measured for un-aged

specimens (control mix) for comparison purposes
with STOA and LTOA conditions. Figure 10 shows
Figure 9. LTOA at 85°C for 120 hours of the com- ITSM values of control, STOA and LTOA mixes at
pacted specimens. different air void contents and testing temperatures.

227
 lll 7% air voids - dry condition - zooc
1Il 7% air voids - wet condition - 20"C
Ell lO% air Yoids - dry condition - zooc
l!l lO% air Yold:s - wet condition - zooc

AlrYold scontant, %

Figure 10. ITSM of control, STOA, LTOA mixtures. Figure 12. Effect of moisture conditioning at different
air void contents on the ITSM of 14 mm DBM.

~Ageing Index a t 20°C

IJ 10%airvolds - 20" C
!::! 10% airvoids

~ ~========"='='%==•=;c=vo=;'='==================~

Air voids content, o/o

Figure 11. Ageing index at 10°C and 20°C.

It is observed as the air voids increase the ITSM Figure 13. Effect of moisture conditioning on the
ITSM of 14 mm DBM at different testing temperatures.
decreases. From Figure 11, one interesting finding
isthe ageing index at 20°C increases as the air voids
increase. That means air voids increase allows to Moreover, Figure 13 shows the effect of moisture
more access of oxidation and in turn more ageing. conditioning on the ITSM of 14 mm DBM at two
However, at 10°C the ageing index slightly decreases different testing temperatures 10°C and 20°C. Simi-
as the air voids increase. That might be attributed to larly, there is a slight ITSM increase at 10°C after
three factors working at the same time. Two of them moisture conditioning. It is difficult to explain this
causing a significant ITSM increase; the 10°C tem- result, but it might be related to binder ageing taking
perature and the ageing conditions while air void place during the procedure of moisture conditioning
contentsincrease causes a reduction in ITSM values. and increasing the conditioned ITSM.

6.2 Effect of water on the ITSM of asphalt mixtures 6.3 Combined effect of moisture and LTOA
conditioning on the ITSM of asphalt mixtures
In this work, one cycle of moisture condition-
ing was applied to the cored specimens and the Another important finding was the combined effect
achieved saturation level was around 30% for most of moisture and ageing conditioning. It wasevaluated
cores. The indirect tensile stiffness was measured at two different testing temperatures, 10°C and 20°C,
for both dry and wet conditions and the ITSM as shown in Figure 14. Similarly, there is a slight
ratio for each specimen is calculated as follows: increase in ITSM of LTOA mixtures after applied
moisture conditioning at both testing temperatures.
ITSMratio = ITSMC/ITSMU In addition, Figure 15 and 16 show a com-
parison between the ITSM of aged and un-aged
Figures 12 shows the effect of moisture condition- mixtures at 20°C and 10°C. There is a significant
ing on the ITSM of 14 mm DBM at two levels of air increase in the ITSM due to the application of
voids; 7% and 10%. There is an observed reduction LTOA conditioning at both testing temperatures.
in the dry ITSM due to air void content increases but It was observed there is a relative increase in the
surprisingly, ITSM was found to increase slightly at ITSM of both mixtures after applying mois-
7% air voids and significantly at 10% air voids after ture conditioning. However, this increase is less
the moisture conditioning protocol was applied. pronounced for aged mixtures because aged mix-

228
 • LTOA mix- dry condition at 10°C
+ LTOA mix- wet condition at 10°C
t Control mix- dry condition at 10°C
• Control mix - wet condit ion at 1ooc
1000

. ·~
J
~
~ 100

~ l y= 2453.6:Cl 129 1 y = 2621x4m l l y = 2969.5~-165 , ,y = 4034.Sx-03118 1

eu R2 =0.9995 R2 =0.9863 R"=0.9846 R1 =0.9914
f
Figure 14. Effect of moisture conditioning and ageing
at different testing temperatures. 10
1000 10000 100000 1000000

Cycles Number, Nf50

Figure 17. Fatigue Life of 14 mm DBM at different

conditions and at 10°C.

:t~g: ~:~ = ~:t ~~~~wi~~ ~~ ~~:~

• Control mix- dry condition at 20°C
1000 i====='"=c=o="'='•='=m=lx=-=w="='=o=
nd="='•=n=•'=2'=o='•c"======!

Figure 15. Effect of moisture conditioning and LTOA

at 20°C.
lv =2684.4x0-l'l4 11v = 4768.7x-<'' "5 l lv = 6585.9J(D-m l lv - 7827..3x0.411~ 1
I Rl=0.9983 II Rl=0.9791 II Rl =0.992 II Rl=09757 I

10 +-----------~----------------------~
1000 10000 100000 1000000

Cycles Number, Nf50

Figure 18. Fatigue Life of 14 mm DBM at different

conditions and at 20°C.

fatigue life at 10°C is greater than at 20°C regard-

less of the condition of the tested mixtures. LTOA
increases the fatigue life regardless of the testing
Figure 16. Effect of moisture conditioning and LTOA temperature and applied conditions. The reason
at 10°C. of this can be attributed to the stiffness increase at
10°C and after applying the LTOA. That means the
resultant fatigue life from this test increases as the
tures already acquired some ageing during ageing material is stiffer. On the other hand, it is observed
conditioning itself. Once more, this may give fur- from figures below that fatigue life decreases
ther evidence to the manifestation of binder ageing due to moisture effects irrespective of whether the
during the procedure of moisture conditioning. tested mixtures was aged or not and whether the
testing temperature was 10°C or 20°C.
6.4 Indirect Tensile Fatigue Test (ITFT) in strain
control mode
7 CONCLUSIONS
In this study, ITFT was applied to samples under
different conditions; dry, wet and LTOA and at In this study the effect of moisture and ageing con-
different strain levels (120, 150, 200, 300, 400) με. ditions on the stiffness and fatigue of asphalt mix-
Fatigue life of asphalt mixtures was determined in tures were investigated.
terms of 50% reduction in stiffness (Nf50). From
Figures 17 and 18, there is an obvious convergence − The present study raises the possibility that age-
between fatigue lives at high micro-strain levels ing might take place during the procedure of
(300 and 400) με in particular at 20°C. In addi- moisture conditioning causing a relative increase
tion, it is noticed from applying the ITFT, that the in ITSM values.

229
− These findings may be somewhat limited by the British Standards Institution (2005), Coated Macadam
number of wet conditioning cycles which were (asphalt concrete) for roads and other paved areas—
applied. Therefore it is an important issue for Part 1: Specification for constituent materials and for
mixtures, BS 4987—1:2005, London.
future research to apply more moisture condi-
Epps, J. A. and Monismith, C. L. 1972. Fatigue of asphalt
tioning cycles and measuring the ITSM after concrete mixtures-summary of existing information.
each cycle. Fatigue of compacted bituminous aggregate mixtures,
− Ageing index in terms of air void contents at 20°C ASTM STP, 508, 19–45.
is different from that at 10°C. It is important to Glover, C. J., Davison, R. R., Domke, C. H., Ruan, Y.,
bear in mind the possible interaction of different Juristyarini, P., Knorr, D. B. And Jung, S. H. 2005.
factors at the same time; ITSM increases due to Development of a new method for assessing asphalt
decreased temperature and ageing but decreases binder durability with Field Validation.
due to the increase of air void content. Harvey, J. T., Deacon, J. A., Tsai, B.-W. and Monismith,
C. L. 1995. Fatigue performance of asphalt concrete
− One of the issues that emerges from this work is
mixes and its relationship to asphalt concrete pave-
the contrast in reduction and increase of fatigue ment performance in California.
life by water damage and ageing respectively. Harvey, J. and Tsai, B. W. 1997. Long-term oven-aging
This finding has important implications for effects on fatigue and initial stiffness of asphalt con-
developing a better understanding of fatigue crete. Transportation Research Record: Journal of the
damage in the presence of environmental fac- Transportation Research Board, 1590, 89–98.
tors. Testing a recovered binder after the applied Huang, B., et al. (2010). “Laboratory evaluation of
conditions may assist in a better understanding. moisture susceptibility of hot-mix asphalt containing
− Based on the findings of ITFT, moisture effects cementitious fillers.” Journal of materials in civil engi-
neering 22(7): 667–673.
should be considered in fatigue life prediction rela-
Jaskula, P. and Judycki, J. 2014. Durability of asphalt
tionships. On the other hand, there is still unan- concrete subjected to deteriorating effects of water and
swered questions about the limit of how much frost. Journal of Performance of Constructed Facilities.
ageing can affect the fatigue behavior of asphalt Jimenez, R. A. and Gallaway, B. M. Behavior of asphaltic
materials. This suggests more ageing cycles and concrete diaphragms to repetitive loadings. Interna-
different fatigue methods should be studied. tional Conference on the Structural Design of Asphalt
Pavements, 1962.
Jung, S. H., Walubita, L. F., Martin, A. E. and Glover, C. J.
Mixture versus neat-film binder oxidation and hard-
ACKNOWLEDGEMENTS ening and the Impact of Binder Oxidation on Esti-
mated mixture fatigue life. Transportation Research
My sincere thanks are due to AL-Mustansiriya Board 86th Annual Meeting, 2007.
University in Baghdad, Iraqi Ministry of Higher Kennedy, T. W. (1985). Prevention of water damage in
Education and Scientific Research, Iraqi Cultural asphalt mixtures. Evaluation and Prevention of Water
Attachéin London for their support and financial Damage to Asphalt Pavement Materials, ASTM
International.
funding of this research.
Maggiore, C. 2014. A comparison of different test and
analysis methods for asphalt fatigue. PhD, the Univer-
sity of Nottingham.
REFERENCES Miller, J. S. and Bellinger, W. Y. 2003. Distress identifica-
tion manual for the long-term pavement performance
AASHTO 1994. Standard practice for short and long program.
term aging of hot mix asphalt. PP2. Washington, DC, Vallerga, B., Finn, F. and Hicks, R. Effect of asphalt
AASHTO Provisional Standards. aging on the fatigue properties of asphalt concrete.
Airey, G. D. and Y.-K. Choi (2002). “State of the art Intl ConfStruct Design Asphalt Pvmts, 1967.
report on moisture sensitivity test methods for bitumi- Virgili, M.N. Partl, A. Grilli and F.A. Santagata. Dam-
nous pavement materials.” Road Materials and Pave- age model for environmental conditioned fatigue test
ment Design 3(4): 355–372. with CAST. Fatigue and Fracture engineering materi-
Baburamani, P. 1999. Asphalt fatigue life prediction als and structures31, 967–979, 2008.
models: a literature review. Walubita, L. F., Martin, A. E., Jung, S. H., Glover, C. J.,
Barra, B., Momm, L., Guerrero, Y. and Bernucci, L. Park, E. S., Chowdhury, A. and Lytton, R. L. 2005.
2012. Fatigue behavior of dense asphalt mixes in dry Comparison of Fatigue Analysis Approaches for Two
and environmental-conditioning states. Construction Hot Mix Asphalt Concrete (HMAC) Mixtures.
and Building Materials, 29, 128–134. Walubita, L. F. 2006. Comparison of fatigue analysis
British Standards Institution (2004), Bituminous mix- approaches for predicting fatigue lives of Hot-Mix
tures—Test methods for hot mix asphalt—Part 24: Asphalt Concrete (HMAC) mixtures. Texas A and M
Resistance to fatigue, BS EN 12697—24: 2004, London. University.
British Standards Institution (2004), Bituminous mix- Wu, J. 2009. The influence of mineral aggregates and
tures—Test methods for hot mix asphalt—Part 26: binder volumetric on bitumen ageing. Dissertation
Stiffness, BS EN 12697–26: 2004, London. University of Nottingham, UK.

230
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Laboratory test and numerical simulation of microwave heating

properties of asphalt mixture

H. Wang, P. Apostolidis, X. Liu & T. Scarpas

Section of Pavement Engineering, Faculty of Civil Engineering and Geosciences, Delft University of Technology,
Delft, The Netherlands

J. Yang
School of Transportation, Southeast University, Nanjing, Jiangsu Province, China

L. Xu
School of Mathematical Science, Beijing Normal University, Beijing, China

ABSTRACT: Microwave heating has been applied for asphalt pavement recycling and maintenance. To
investigate the microwave heating properties of asphalt mixture, laboratory test and numerical simulation
were done and compared. Dense asphalt mixture samples with and without graphite powder were heated
using microwave oven for different time. Numerical simulation of microwave heating of asphalt mixture
was developed with FEM software COMSOL Multiphysics. Both laboratory measured and simulated
surface temperatures were recorded and analyzed. Results show that graphite powders can significantly
increase the microwave heating efficiency of asphalt mixture. Numerical simulation results have a good
correlation with laboratory test results. It is feasible to use COMSOL Multiphysics to simulate the micro-
wave heating process of asphalt mixture.

1 INTRODUCTION in asphalt mixtures for better use of microwave to

repair pavement defects, to recover and improve
The mechanism of microwave heating is that pavement performance. They showed that incor-
dielectric material can generate heat in microwave porating 2.0 percent magnetite by weight in the
field due to its dielectric loss, including polarized asphalt reduced the time required for the asphalt to
relaxation loss and conductive loss (Meredith, reach its softening temperature of 250°F from 240
1998). Asphalt mixture usually consists of about seconds to 45 seconds when heated in a standard
5% of asphalt, about 95% of coarse aggregate, fine 2.45 GHz microwave oven. Long (2001) also noted
aggregate and other mineral powders (Kim et al., that pothole repairs by the use of applied micro-
2003). As a dielectric material, asphalt mixture also wave energy could be done more rapidly and effec-
can be heated by microwave. Microwave heating tively with a patching compound with enhanced
has the characteristics of fast heating, good uni- microwave absorption. Hopstock & Zanko (2005)
formity, high efficiency, safety and environmental found that bitumen has a very low loss tangent of
friendly (Metaxas & Meredith, 1983). It has been about 0.001. The natural magnetite in taconite is
widely used in asphalt pavement hot recycling and an outstanding microwave absorber. Gallego et al.,
maintenance technology. (2013) used a kitchen microwave oven to heat the
The earliest on-site microwave heating regen- asphalt mixture mixed with steel fibers, and studied
eration of asphalt pavement was completed by the heating process and the temperature changing
Bosisio et al. (1974), which patched potholes rule. The results showed that the microwave heat-
through a microwave heating system with a power ing is a feasible method of heating and healing the
of only 1.2 kW. Al-Ohaly & Terrel (1988) performed asphalt mixture; the addition steel fiber can greatly
a simple elastic modulus and tensile test on the enhance the heating rate. Norambuena-Contreras
asphalt mixture which was heated by a household & Garcia (2016) evaluated the effect of microwave
microwave oven. The results show that microwave and induction heating on the self-healing of asphalt
heating can improve the adhesion of asphalt binder mixture test samples containing steel wool fibers. It
to mineral aggregate, and decrease the moisture sus- was found that microwave technology is more effec-
ceptibility. Osborne & Hutcheson (1989) proposed tive than induction heating to heal cracks in asphalt
the incorporation of microwave-absorbing materials roads.

231
 Through various time and material consuming
laboratory tests, it can be found microwave heat-
ing is a promising technology for asphalt pavement
recycling and maintenance. However, fewer studies
applied numerical modelling to investigate the micro-
wave heating process and mechanism of asphaltic
materials (Apostolidis et al., 2016). This study aims
to investigate the microwave heating properties of
asphalt mixture through both laboratory test and
numerical simulation. Furthermore, it aims to give a
first insight into the feasibility of numerical simula-
tion for microwave heating of asphalt mixture.

2 EXPERIMENTAL INVESTIGATION Figure 1. Appearance of graphite.

2.1 Materials
In this study, basalt aggregates and limestone fillers
were used to produce asphalt mixtures. The basic 80

asphalt binder used in this study was SHELL-70,

which is equivalent to PG 64–22. The properties of
asphalt binder are listed in Table 1. Natural flaky
graphite was used to enhance the electro-thermal
properties of asphalt mixture. The basic physical
parameters and appearance are shown in Table 2
and Figure 1 respectively. Graphite powder, together ;;;;
with the limestone, work as fillers in the mixture. Seive size (m m)

Figure 2. Mix gradation of AC-13.

2.2 Mixture design
Dense graded asphalt concrete (AC-13) with 13.2-
was designed in accordance with standard Mar-
mm nominal maximum aggregate size was used in
shall Design method (ASTM D1559). The optimal
this research. Gradation AC-13 shown in Figure 2
asphalt content for the control mixture was 4.8%.
In order to make the addition of graphite as the
Table 1. Basic properties of asphalt binder. only factor effecting the microwave heating prop-
erties of asphalt mixture. No separate mix designs
Properties Value
were performed for the mixtures containing graph-
Penetration (25°C, 100 g, 5 s, 0.1 mm) 71 ite powders. Two types of asphalt mixture sam-
Ductility (5 cm/min, 5°C, cm) 32.2 ples, control mix (AC) and mix containing 1.3%
Softening point (R&B, °C) 47.5 graphite powder by weight of asphalt mixture
Flash point (°C) 272 (AC-Graphite), were prepared. Cylindrical shape
Rotational viscosity (60°C, Pa.s) 203 specimens with 100 mm diameter and 65 mm
Wax content (%) 1.6 height were fabricated for microwave heating test.
Density (15°C, g/cm3) 1.032
2.3 Thermal conductivity measurement
Table 2. Basic parameters of graphite powder. Thermal conductivity is one of the important fac-
tors that affect the microwave heating process
Parameters Value which refers to heat transfer. The thermal conduc-
Chemical component C tivity measurement was done through steady-state
Density (g/cm3) 32.2 method using heat flow meter according to ASTM
Particle size (μm) 75 C518. A specimen (15 cm × 15 cm × 4 cm) shown in
Carbon content (%) 96.1 Figure 3 is placed between a hot and a cold plate,
Specific surface area(m2/g) 7.12 and the heat flow created by the well-defined tem-
Electrical resistance (Ω ⋅ m) 10−6 perature difference is measured with a heat flux
Thermal conductivity (W/(m⋅K)) 151 sensor. In this case, the cold plate was at a constant
temperature of 5°C and the hot plate at a constant

232
 and then the averaged was calculated to obtain a
“near-uniform temperature”. Then, the oven was
turned on again. This process was repeated twelve
times for a total heating period of 120 s. Each type
of asphalt concrete has two replicates.

3 NUMERICAL SIMULATION

3.1 Model definition

The microwave oven is a metallic box connected
to a 2.45 GHz microwave source via a rectangular
Figure 3. Asphalt mixture plate sample.
waveguide. Near the bottom of the oven there is
a cylindrical glass plate with a cylindrical asphalt
mixture sample placed on top of it. The microwave
operates at 1000 W, but because we use symmetry
to reduce the model size by one half, we only input
500 W in simulation. The symmetry cut is applied
vertically through the oven, waveguide, asphalt
mixture sample, and plate. Figures 5, 6 below show
the reduced geometry and 3D mesh respectively.

Figure 4. Surface temperature measurement of speci-

men in the microwave oven.

temperature of 45°C. The thermal conductivity can

be obtained directly from the data acquisition sys-
tem according to Fourier’s Law for heat conduction.

2.4 Microwave heating and temperature

measurement Figure 5. Geometry of microwave oven, asphalt mix-
Microwaves, with their ability to rapidly heat die- ture sample, and waveguide feed.
lectric or magnetic materials, are commonly used
as a source of heat and an alternative to conven-
tional conductive heating methods. To heat the AC
samples in this study, a microwave oven was used
with an input of 1200 W and a 220 V, 50 Hz power
supply. The oven can produce microwaves of up to
1000 W, with a frequency of 2.45 GHz, which cor-
responds to an approximate wavelength of 120 mm.
The cylindrical Marshall sample was placed in
the microwave oven and heated for a period of 10 s.
Between each two consecutive periods, the oven
was opened quickly and the surface temperatures
were taken using an infrared temperature-meas-
ured gun (Figure 4). The temperature loss during
the measuring process was not considered. Due to
the temperature difference between coarse aggre-
gates and asphalt mastic, six temperatures were Figure 6. Mesh of microwave oven, asphalt mixture
taken randomly from the surface of the test sample, sample, and waveguide feed.

233
3.2 Simulation parameters Therefore, for the numerical simulation part, only
control asphalt mixture was done.
The model uses copper for the walls of the oven
The values of relative permittivity, relative perme-
and waveguide. Although resistive metals losses
ability and heat capacity at constant pressure were
are expected to be small, the impedance bound-
chosen according to the material manual. The remain
ary condition on these walls ensures that they get
property parameters were tested in the lab (Wang
accounted for. The symmetry cut has mirror sym-
et al., 2016). In this study, the permittivity of asphalt
metry for the electric field and is represented by
mixture sample was set to εr = 8.6 – 0.52 j where the
the boundary condition as shown in Equation 1
imaginary part accounts for dielectric losses.
(Metaxas & Meredith, 1983).
n H=0 (1)
4 RESULTS AND DISCUSSION
where n is the outward unit normal vector to the
port boundary; H is the magnetic field vector. 4.1 Microwave heating speeds of asphalt concrete
The rectangular port is excited by a Transverse test samples
Electric (TE) wave, which is a wave that has no elec- Figure 7 presents the surface temperature changes
tric field component in the direction of propagation. of different asphalt mixture specimens in the
The port condition requires a propagation constant microwave oven. It can be observed that the sur-
β, which at the frequency ν is given by the expression: face temperatures of both types of asphalt mixture
2π 2 2 specimens increased significantly after microwave
β= v − vc (2) heating for 240 s. The final surface temperatures
c reached to 101°C and 122°C for control mix and
With the stipulated excitation at the rectangular mix containing graphite respectively. In addition,
port, the following equation is solved for the elec- AC samples containing graphite powder had higher
tric field vector E inside the waveguide and oven: heating speeds than the control ones. This phenom-
enon can be interpreted that graphite powders in
⎛ jσ ⎞ the mixtures can be used as microwave absorbing
∇ × ( μr−1∇ × ) − k02 ⎜ ε r − E=0 (3)
ε 0 ⎟⎠
enhancers for they can dramatically increase the
⎝ ωε
dielectric constant of asphalt mixture according to
where μr denotes the relative permeability, j the generalized average effect of composite materials.
imaginary unit, σ the conductivity, ω the angular As mentioned before, asphalt mixture, as a com-
frequency, εr the relative permittivity, and ε0 the posite dielectric, will alternate electromagnetic
permittivity of free space. energy into heat under the action of microwave
The model uses material parameters for air: due to dielectric loss, which eventually leads to
σ = 0 and μr = εr = 1. The glass plate has σ = 0, the increase of temperature. The mechanism of
μr = 1 and εr = 2.55. The thermal properties of the microwave heating includes polarized relaxation
glass plate were not defined as it was not involved loss and conductive loss. Specifically, the dielec-
in the thermal part of the model. tric loss enhancing mechanisms of graphite can
For asphalt mixture sample without graphite, the be summarized as follows (Liu et al., 2010): (1)
following material parameters were implemented Conductive powder, considered as dipole in the
in the numerical model as shown in Table 3. It is process of damped vibration, can attenuate the
quite difficult to obtain all the necessary material electromagnetic wave. (2) Graphite powder used in
parameters for asphalt mixture containing specific this study has high specific surface area and more
percentage graphite, as well as the implementa-
tion of graphite as a separated part in the model.
140

Table 3. Material property parameters. 110

p
........ 100
Property Value Unit ~
~ 80

Relative permittivity 8.6–0.52 j 1 ! 60

40
Relative permeability 1 1
Electrical conductivity 4.2e-9 S/m 20

Thermal conductivity 1.13 W/(m⋅K) 20 40 60 80 100 120 HO

Density 2536 kg/m3 Time exposed to microwave (s)
Heat capacity* 1485 J/(kg⋅K)
Figure 7. Surface temperature evolution of test
* at constant pressure. samples.

234
dangling bonds due to small particle size, so when metric treatment of the model. It is obvious that
electromagnetic wave enters into the absorber, they the internal temperatures are higher the surface
can be attenuated by multi-scattering and reflect- temperatures, which can be explained by the fact
ing. (3) The leakage conductance between graphite that heat dissipation is greater on the surface of a
particles can attenuate electromagnetic wave. specimen than in its interior. This simulation result
From another viewpoint, the addition of graph- coincides with the laboratory result which was
ite slightly increases the thermal conductivity of found by Gallego et al. (2013).
asphalt mixture (from 1.13 W/(m⋅K) to 1.21 W/ When heating the asphalt mixture to certain
(m⋅K)), which also makes contributions to the temperatures, the inside water contents start boil-
higher microwave heating speeds of asphalt mix- ing and transporting heat as steam to outer layers.
ture containing graphite. Bitumen may start flowing due to softening, result-
ing the change of air voids and skeleton structure.
4.2 Numerical simulation results These above also affect the electromagnetic proper-
ties of the asphalt mixture. The simple microwave
4.2.1 Microwave heat source distribution absorption and heat conduction model used here
Figure 8 below shows the distributed microwave does not capture these nonlinear effects. However,
heat source as a slice plot through the centre of the model can serve as a starting point for a more
the asphalt mixture sample. It indicates that the advanced analysis.
dissipated microwave power distribution shows
a complicated oscillating pattern, which has sev- 4.2.3 Comparison between laboratory and
eral strong peaks inside the sample. This pattern simulation results
is different from spherical potato sample in food In order to make a direct comparison between
engineering which has only one strong peak in the the laboratory tested surface temperatures and
centre, acting as a resonant cavity for the micro- numerical simulated surface temperature, the col-
wave field. The reasons are probably due to the oured temperature contour plots of top surface
higher homogeneity of potato and the sample of asphalt mixture were summarized in Figure 10.
shape dependent microwave heating property. With the increase of microwave heating time, the
Through a volume integral of the microwave
heating, the amount of the energy absorbed in the
asphalt mixture sample can be found out as 426 W. Time= l20 s Surface: Temperature (degC)

The power absorbed in the sample amounts to

about 85% of the input microwave power (500 W).
Most of the remaining power is reflected back
through the port.

4.2.2 Temperature distribution of test samples

Figure 9 shows the temperature distribution of
asphalt mixture after 120 s simulative microwave
heating. It should be emphasized that the rectan-
gular cross section seen as surface area is actually
the internal part of asphalt mixture due to the sym-

Figure 9. Surface temperature distribution of asphalt

freq(1)=2.45E9 Slice: Resistive losses (W/m3 )
mixture.

1.8
1.6 Top surface temperature (degC)
1.4
1.2
1
0.8
0.6
0.4
0.2
¥ Time=80 s Time=IOO s Tm~e=1 20 s
0 z- x

Figure 8. Dissipated microwave power distribution Figure 10. Top surface temperatures at different micro-
(W/m3). wave heating time.

235
 120 For further research, it is recommended that
100 more accurate electromagnetic properties of spe-
cific asphalt mixture samples should be measured
E so and implemented in the numerical model. In addi-
;
60 tion, the size effect of test samples and specific
1 material parameters, such as moisture content,
~ 40
air voids, asphalt content, aggregate properties,
lO
etc. also should be considered. Developing more
0 advanced numerical models which consider the
0 20 40 60 so 100 120 1<0
Time exposed to microwave (s)
nonlinear effects will be a challenge.

Figure 11. Top surface temperature comparison

between laboratory test and numerical simulation. ACKNOWLEDGEMENTS
The corresponding author would like to acknowl-
edge the scholarship from China Scholarship
top surface temperature increases gradually. More Council.
precisely, the top surface temperatures of labora-
tory test and numerical simulation were compared
in Figure 11. Here the simulative top temperature REFERENCES
values were averaged through area integral. It is
obvious that numerical simulation results have a Al-Ohaly, A. A. & Terrel, R. L. (1988) Effect of micro-
wave heating on adhesion and moisture damage of
good correlation with laboratory test results. How-
asphalt mixtures.
ever, the numerical simulated temperatures are Apostolidis, P., Liu, X., Scarpas, A., Kasbergen, C. &
somewhat lower than the laboratory test results. Van De Ven, M. F. C. (2016) Advanced evaluation of
This could be due to the nonlinear effects during asphalt mortar for induction healing purposes. Con-
the heating process or the uncertainty of material struction and Building Materials, 126, 9–25.
parameter inputs used in the model, which needs Bosisio, R., Spooner, J. & Granger, J. (1974) Asphalt road
to be further investigated. Nevertheless, it is feasi- maintenance with a mobile microwave power unit.
ble to use numerical method to simulate the micro- Journal of Microwave Power, 9, 381–386.
wave heating process of asphalt mixture. Gallego, J., Del Val, M. A., Contreras, V. & Paez, A.
(2013) Heating asphalt mixtures with microwaves to
promote self-healing. Construction and Building Mate-
rials, 42, 1–4.
5 CONCLUSIONS AND Hopstock, M. D. & Zanko, M. L. (2005) Minnesota
RECOMMENDATIONS Taconite as a Microwave-Absorbing Road Aggregate
Material for Deicing and Pothole Patching Appli-
This study investigated the microwave heating cations. University of Minnesota Duluth, Natural
properties of asphalt mixture through both labora- Resources Research Institute.
tory tests and numerical simulation. The effect of Kim, Y. R., Little, D. N. & Lytton, R. L. (2003) Fatigue and
graphite on the thermal property and microwave healing characterization of asphalt mixtures. Journal of
Materials in Civil Engineering, 15, 75–83.
heating speed of asphalt mixture were explored
Liu, L., Duan, Y., Ma, L., Liu, S. & Yu, Z. (2010) Micro-
through laboratory tests. The main conclusions are wave absorption properties of a wave-absorbing coat-
as follows: ing employing carbonyl-iron powder and carbon
1. The addition of graphite can slightly increase black. Applied Surface Science, 257, 842–846.
Long, H. W. (2001) Asphaltic compositions and uses
the thermal conductivity of asphalt mixture
therefor. Google Patents.
from 1.13 W/(m⋅K) to 1.21 W/(m⋅K). Meredith, R. J. (1998) Engineers’ handbook of industrial
2. Asphalt mixture sample containing graphite microwave heating, IET.
has a higher microwave heating speed than Metaxas, A. A. & Meredith, R. J. (1983) Industrial micro-
common asphalt mixture sample. This is mainly wave heating, IET.
because graphite powders can act as microwave Norambuena-Contreras, J. & Garcia, A. (2016) Self-heal-
absorbing enhancers in the asphalt mixture and ing of asphalt mixture by microwave and induction
increase the thermal conductivity. heating. Materials & Design, 106, 404–414.
3. There is a good correlation between laboratory Osborne, T. L. & Hutcheson, W. R. (1989) Asphalt com-
pounds and method for asphalt reconditioning using
measured temperatures and numerical simu-
microwave radiation. Google Patents.
lated temperatures of asphalt mixture sample. Wang, H., Yang, J., Liao, H. & Chen, X. (2016) Elec-
4. It is feasible to use COMSOL Multiphysics trical and mechanical properties of asphalt concrete
to simulate the microwave heating process of containing conductive fibers and fillers. Construction
asphalt mixture. and Building Materials, 122, 184–190.

236
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Influence of lateral confining pressure on flow number tests

E. Santagata, O. Baglieri & P.P. Riviera

Politecnico di Torino, Turin, Italy

M. Lanotte
Michigan State University, East Lansing, MI, USA

M. Alam
Abasyn University, Peshawar, Pakistan

ABSTRACT: The study described in this paper aimed to investigate the rutting resistance of bituminous
mixtures by means of flow number tests, with a particular emphasis placed on the analysis of the effects
produced by variations of lateral confinement. Two wearing course mixtures, one standard dense-graded
and one rubberized gap-graded, were employed in the experimental investigation in order to highlight the
influence of confining pressure on materials with significantly different bulk structure and rheological
properties of the binder phase. Flow number tests were carried out at a single test temperature and in
three confinement conditions on cylindrical samples compacted at two air void contents. Obtained results
suggest that confining pressure can significantly affect flow number values to an extent that depends on
mixture type and composition, indicating the importance of selecting proper testing conditions in order
to achieve a reliable performance-based ranking of materials.

1 INTRODUCTION tests (Kaloush & Witczak 2002, Hinislioglu & Agar

2004, Tayfur et al. 2007, Fontes et al. 2010, Perra-
Accumulation of permanent deformation caused ton et al. 2011, BaghaeeMoghaddam et al. 2014a,
by heavy and slow-moving traffic represents one b, Katman et al. 2015, Morea & Zerbino 2015).
of the main distress types affecting asphalt pave- Wheel tracking testsare attractive since they are
ments. The presence of ruts or distortions on road capable of simulatingtraffic-related moving loads;
surfaces may jeopardize storm water run off, thus however, a non-uniform stress distribution is gen-
increasing the possible occurrence of hydroplan- erated in the slabs, thus making these tests not suit-
ing events; moreover, it may lead to the premature able for the development of constitutive models
failure of pavement structures due to undesirable (Perraton et al. 2011, Sohm et al. 2012, Hofko 2015,
wheel impact and dynamic load amplification Yu et al. 2015). For this reason, triaxial cyclic com-
effects (Sousa et al. 1991). pression tests currently represent the most popular
Formation of ruts results from densification option for the evaluation of the anti-rutting prop-
and shear strain (Zaniewski & Nallamothu 2003, erties of bituminous mixtures (Kaloush & Witczak
Sousa & Weissman 1994) which are influenced by 2002, AASHTO 2008, Fontes et al. 2010, Katman
several factors such as aggregate characteristics et al. 2015, Abdelfattah et al. 2016).
(type, texture and gradation), binder rheology, The importance of considering the behavior
binder dosage, mixture volumetrics (Muniandy & of bituminous mixtures in the triaxial stress con-
Huat 2006, Radziszewski 2007, Fontes et al. 2010, figuration has been underlined by several studies
Baghaee Moghaddam et al. 2011, Xu & Huang since it is more representative of actual conditions
2012, Baghaee Moghaddam et al. 2014a, b). occurring in the field (Brown & Foo 1994, De
It is widely recognized that evaluation of the rut- Visscher et al. 2006, Taherkhani et al. 2007, Clec’h
ting potential of bituminous mixtures in the labo- et al. 2009, Hofko & Blab 2010, Roy et al. 2015b,
ratory represents a challenging task due to their Yu et al. 2015,Gayathri et al. 2016, Zhu et al. 2016).
complex rheological behavior and to lab-to-field However, the selection of appropriate triaxial stress
scale effects (Blanc et al. 2015). Over the years, sev- paths still represents a crucial issue which needs
eral methods have been proposed by researchers, to be addressed thoroughly (Blanc et al. 2015).
although those which are more widely accepted are A contribution to this topic has been recently given
wheel tracking tests and triaxial cyclic compression by proposing the adoption of a cyclic confining

237
pressure which was shown to have a significant consists in applying repeated compressive stresses
impact on materials response (Hofko& Blab 2014, (equal to 600 kPa) to cylindrical specimens. A sin-
Hofko 2015). gle test temperature was adopted (equal to 58°C,
The experimental investigation described in this which corresponds to the high performance-grade
paper focused on the evaluation of rutting prop- temperature commonly used in northern Italy), in
erties of bituminous mixtures by means of flow combination with four different confining pres-
number tests carried out in unconfined and con- sures (0, 14, 28 and 42 kPa).
fined conditions. The flow number test is a dynamic Samples for the flow number test were prepared
creep and recovery test in which a slender cylindri- in accordance with the AASHTO PP60 specifica-
cal specimen is subjected to cyclic haversinepulse tion. Over-height specimens (100 diameter, 150 mm
loads followed by rest periods. Two wearing course height) were first produced with the gyratory shear
bituminous mixtures, one standard dense-graded compactor by employing the mixture mass needed
and one rubberized gap-graded, were employed in to reach the target value of air voids. Cores with
the investigation. Obtained results were analyzed 100 mm diameter were then extracted from the
and discussed with the specific goal of highlight- center of gyratory samples using a diamond cor-
ing the influence of lateral confining pressure on ing stand and thereafter subjected to trimming by
the behavior of materials with significantly differ- means of a masonry saw in order to obtain smooth
ent bulk structure and rheological properties of the and parallel end surfaces. In order to evaluate the
binder phase. influence of volumetric characteristics on mate-
rials response, specimens were compacted at two
different air void contents 5 ± 0.5% and 7 ± 0.5%.
2 MATERIALS AND TESTING Measurements were conducted in triplicate runs

Bituminous mixtures employed in the experi-

mental investigation were provided by a local Table 1. Actual binder contents (Bactual) determined
contractor operating in north-western Italy. The from ignition tests, optimum binder dosages (Bopt) and
Dense-Graded mixture (DG) was produced with TMD.
siliceous aggregates and neat 50/70 penetration
Mixture code Bactual (%) Bopt (%) TMD (kg/m3)
grade bitumen. Target aggregate gradation was
defined according to Italian technical specifica-

,
DG 4.8 5.0 ± 0.5 2477
tions for standard wearing courses (CIRS, 2001), GG 7.8 8.0 ± 0.5 2540
with an optimum binder dosage of 5.0% by weight
of dry aggregates. The rubberized Gap-Graded
mixture (GG) contained basaltic aggregates com- 100
bined with a commercially available asphalt rubber ---o- Actual DG comp~~ition
prepared with a nominal percentage of crumb rub- 80 - - -Target DG composition
--l:r- Actual GG composition '1 '
ber equal to 18% by weight of total binder. Target
..
~

I f
aggregate gradation was defined according to tech- e:. 60 ········· Target GG composition

nical specifications commonly adopted in Italy for =

·~ 40 I rk
rubberized gap-graded mixtures (ARI, 2013), with
an optimum binder dosage of 8%.
~ "
20
AJ4"
.·
"
Both mixtures sampled from the hot mix plant
were subjected to a preliminary laboratory char- 0
-- ·'1f 'Vf l
t;l

acterization in order to check the coherency of 0.01 0.1 10 100

their composition with job mix formulae. Binder
contents were determined by means of ignition Diameter (mm)
tests (EN 12697-39) which were followed by sieve
analyses of recovered aggregates (EN 933-1). Mix- Figure 1. Comparison between actual and target aggre-
tures were also tested by means of the pycnometer gate size distributions.
method (EN 12697-5) to determine their theoreti-
Table 2. Volumetrics properties of tested mixtures.
cal maximum densities (TMD), to be used for the
evaluation of volumetric properties of compacted Mixture code v (%) VMA (%) VFB (%)
specimens. Obtained results are reported in Table 1
and Figure 1, which prove a very good agreement DG 5.3 15.7 66.4
between actual and target composition. 7.2 17.4 58.6
Flow number tests were performed with the GG 4.8 23.2 80.0
Asphalt Mixture Performance Tester (AMPT) 7.4 25.3 71.0
according to AASHTO TP 79. The testing protocol

238
on the slender specimens for both compaction In general terms, the evolution of accumulated
conditions. axial strains as a function of number of loading
Volumetric properties of specimens after com- cycles is expected to show the presence of three
paction (air void content v, voids in the mineral consecutive stages: a primary stage, in which the
aggregate VMA, voids filled with bitumen VFB), rate of strain decreases as the number of loading
determined according to EN 12697–8, are reported cycles increases; a secondary stage, characterized
in Table 2. A very good agreement between actual by an almost constant rate of strain; and a third
and target air void contents was observed. stage (of tertiary flow), in which strain rate rises
dramatically, leading to failure. Flow Number
(FN) represents the number of loading cycles cor-
3 RESULTS AND DISCUSSION responding to the transition point between the sec-
ondary and tertiary stage and is considered as a
Examples of experimental data obtained from reliable indicator of the anti-rutting potential of a
flow number tests carried out on mixtures at differ- mixture (Bonaquist, 2008; Witczak et al., 2002).
ent levels of confining pressure (ranging between 0 It was observed that the typical behavior
and 42 kPa) are reported in Figure 2. described above, referred to as Type I creep pattern
by some authors (Roy et al., 2015b), was exhibited
120,000
GO (no con finement) by specimens tested in unconfined conditions and
100,000
by those subjected to lower lateral confining pres-
c
80,000 I sures (14 and 28 kPa). In the case of specimens
] /
.
"'·;;
<
60,000

40,000
,...., v tested at 42 kPa, the tertiary stage was not reached
(Type II creep pattern), making the determination
of the FN value impossible. Although cell pres-
20,000

0 r
0 50 100 150 200
sure values up to 200 kPa have been used by other
authors (Sohm et al. 2012; Hofko & Blab, 2014;
Number of Cycles Roy et al., 2015a; Blanc et al., 2015), in this experi-
80,000 mental investigation the level of lateral confine-
ment was limited to 28 kPa since the specific focus
60,000 of the study was on FN assessment.
c
-~
FN values obtained for mixtures DG and GG
40,000
<ii plotted as a function of confining pressure are
]
.:;; 20,000
reported in Figures3 and 4, respectively. A linear
trend in the semi-log scale is observed, indicat-
ing the increase of resistance to flow of materials
500 1000 1500 with the increase of imposed lateral confinement.
Number of Cycles
Such an evidence, which is consistent with expec-
80,000 tations, can be attributed to the “stiffening effect”
GG (28 kPa)
/
v produced by the decrease of deviatoric stress that
60,000 reduces the rate of accumulation of axial strain in
~
c
.E / the specimen and extends the secondary flow stage.
..
<ii

~
40,000

20,000 /
/ Moreover, due to the fact that confining pressures
are relatively low, the behavior of materials can be
F
0 100,000
0 500 1000 1500 2000 2500
Number of Cycles
10,000
... ................~·:

-
60,000
GG (42 kPa) "'
,.Q
·····
....;:;_·:::::.. ::.. ··: . .:::X ---
E 1,000
z=
,..- ..--
c
40,000
~ 100
~ 0

..
"'
~
20,000 / rz
10
0 Void content (nominal)= 5%
-
to. Void content (nominal)= 7%
0
0 5000 10000 15000 20000 25000 0 5 10 15 20 25 30
Number of Cycles Confining pressure (kPa)

Figure 2. Examples of strain evolution at various con- Figure 3. FN as a function of confining pressure (mix-
fining pressures. ture DG).

239
 100,000 Table 3. FN values at 58°C for mixtures DG and GG at
various confining pressures and air voids.
10,000 ...... .....<>
... DG GG
"'
,Q
1,000
..,....··
E
:z= Confining pressure (kPa) 5% 7% 5% 7%
~
~
fi:
100 --- ---t>

<>Void content (nominal)= 5%

0 112 77 140 43
10 - 14 788 398 1100 358
6 Void content (nominal)= 7% 28 8575 1852 17106 3552

0 5 10 15 20 25 30
Confining pressure (kPa)
compaction) creep behavior is strongly affected
Figure 4. FN as a function of confining pressure (mix- by shear flow and therefore improved stiffness and
ture GG). elasticity of asphalt rubber are emphasized, result-
ing in a reduced aptitude of the corresponding
mixture to accumulate permanent deformation.
assumed to be dilatant, i.e. characterized by a vol-
At the higher air void content (lower degree of
ume increase (Sohm et al., 2012).
compaction)densification mechanisms are predom-
For both mixtures, specimens compacted at the
inant and materials response is mainly governed by
lower target air void content (5%) exhibited higher
internal structure and behavior of the mixtures was
FN values than those compacted at the higher tar-
found to be more susceptible to testing conditions.
get air void content (7%) in all testing conditions.
In particular, the anti-rutting characteristics of the
Build-up of permanent deformation in mixtures
gap-graded aggregate skeleton were highlighted
under repeated creep loading is the result of two
when lateral confinement reached a sufficiently
mechanisms related to shear flow and densification.
level for the promotion of stone-to-stone contact.
The first one is in most part governedbythe visco-
On the contrary, such a skeleton was more prone
elastic characteristics of the bituminous phase,
to densification than the continuous one of the
while the second one is related to the bulk structure
DG mixture when tested in the unconfined state
of the material. Given that tests were performed
or at the lower confining pressure, thus offering an
at the same temperature and loading frequency
inferior resistance to permanent deformation.
(which therefore do not imply any variation of the
rheological response of the binder phase), for each
mixture type considered in the investigation it can
4 CONCLUSIONS
be assumed that the contribution of shear flow
to total deformation was the same regardless of
In the study presented in this paper, rutting proper-
degree of compaction. As a consequence, observed
ties of a standard dense-graded and a rubberized
differences in creep response can be linked to the
gap-graded mixture were investigated and com-
occurrence of densification phenomena, which
pared. The experimental program included flow
were proven to be more pronounced in specimen
number tests carried out at a single test tempera-
characterized by a more porous and weaker inter-
ture and four confining pressures on specimens
nal structure.
compacted at two different air void contents.
Further considerations can be drawn from the
From the analysis of obtained results the follow-
comparative evaluation of the anti-rutting poten-
ing conclusions can be drawn:
tial of mixtures DG and GG.
By referring to the mean FN values summa- – Depending upon the composition of the con-
rized in Table 3, it can be noticed that the relative sidered bituminous mixtures, there may be a
ranking of mixtures varies depending upon the limitation to the maximum confinement pres-
considered combination of confining pressure and sure which can be applied when focusing on the
volumetrics. determination of the FN value;
In the case of 5% air void content, the rubberized – Lateral confining pressure has a stiffening effect
gap-graded mixture showed a superior anti-rutting as a consequence of the reduction of deviatoric
potential with respect to the traditional dense- stress;
graded material at all levels of lateral confinement, – Lateral confinement can affect the anti-rutting
while in the case of higher air voids (7%), this was ranking of bituminous mixtures based on the
confirmed only at the highest confining pressure results of flow number tests;
(28 kPa). – At high air void contents (e.g. 7%) bituminous
The possible reason for these findings is that mixtures with a non-continuous aggregate size
at the lower air void content (higher degree of distribution can exhibit a better anti-rutting

240
 performance than that of dense-graded mixtures Proceedings of the 8th International Conference on
only under the effect of higher confinement the Bearing Capacity of Roads, Railways and Air-
pressures. fields, Urbana—Champaign, IL, USA, 29 June-2 July
2009.
In summary, it was confirmed that proper De Visscher, J., Maeck, J. & Vanelstraete, A. 2006. The
testing conditions need to be selected in order to permanent deformation law of asphalt mixtures:
obtain a reliable evaluation of performance-based Investigation of the effect of mix composition and
mixture properties. material properties. Proceedings of 10th international
conference on asphalt pavements, Quebec City, Canada,
12–17 August 2006.
EN 933-1. Tests for geometrical properties of aggregates—
REFERENCES Part 1: Determination of particle size distribution—
Sieving method.
AASHTO PP 60. Standard practice for preparation EN 12697-5. Bituminous mixtures—Test methods for
of cylindrical performance test specimens using the hot mix asphalt—Part 5: Determination of the maxi-
SUPERPAVE Gyratory Compactor (SGC). mum density.
AASHTO TP 79. Standard method of test for determin- EN 12697-8. Bituminous mixtures—Test methods for
ing the dynamic modulus and flow number for Hot hot mix asphalt—Part 8: Determination of void char-
Mix Asphalt (HMA) using the Asphalt Mixture Per- acteristics of bituminous specimen.
formance Tester (AMPT). EN 12697-39. Bituminous mixtures—Test methods
Abdelfattah, H.F.H., Al-Shamsi, K.& Al-Jabri, K. for hot mix asphalt—Part 39: Binder content by
2016. Evaluation of rutting potential for asphalt ignition.
concrete mixes containing copper slag. Interna- Fontes, L.P.T.L., Triches, G., Pais, J.C. & Pereira, P.A.A.
tional Journal of Pavement Engineering, DOI: 2010. Evaluating permanent deformation in asphalt
10.1080/10298436.2016.1199875. rubber mixtures. Construction and Building Materials
American Association of State Highways and Trans- 24(7): 1193–1200.
portation Officials (AASHTO). 2008. Mechanistic- Gayathri, V.G., Rajasekar, Y.P., Lakshmi Roja., K. &
Empirical Pavement Design Guide—A Manual of Krishnan, J.M. 2016. Influence of confinement pres-
Practice (2nd Edition). sure on the development of three stage curve for
ARI. 2013. Gap-graded asphalt rubber mixture (wet bituminous mixtures. Transportation in Developing
method), Asphalt Rubber Italia (in Italian). Economies 2(10).
Baghaee Moghaddam, T., Karim, M.R. & Abdelaziz, M. Hinislioglu, S. & Agar E. 2004. Use of waste high density
2011. A review on fatigue and rutting performance polyethylene as bitumen modifier in asphalt concrete
of asphalt mixes. Scientific Research and Essays 6(4): mix. Materials Letters 58(3–4): 267–271.
670–682. Hofko, B. & Blab, R. 2010. Assessment of permanent
Baghaee Moghaddam, T., Soltani, M. & Karim, M.R. deformation behavior of asphalt concrete by improved
2014a. Evaluation of permanent deformation charac- triaxial cyclic compression testing. Proceedings of 11th
teristics of unmodified and Polyethylene Terephtha- international conference on asphalt pavements, Nagoya,
late modified asphalt mixtures using dynamic creep Japan, 1–6 August 2010.
test. Materials and Design 54: 317–324. Hofko, B. & Blab, R. 2014. Enhancing triaxial cyclic
Baghaee Moghaddam, T., Soltani, M. & Karim, M.R. compression testing of hot mix asphalt by introducing
2014b. Experimental characterization of rutting per- cyclic confining pressure. Road Materials and Pave-
formance of Polyethylene Terephthalate modified ment Design 15(1): 16–34.
asphalt mixtures under static and dynamic loads. Con- Hofko, B. 2015. Addressing the permanent deformation
struction and Building Materials 65: 487–494. behavior of hot mix asphalt by triaxial cyclic compres-
Blanc, J., Gabet, T., Hornych, P., Piau, J.-M. & Di Bene- sion testing with cyclic confining pressure. Journal of
detto, H. 2015. Cyclic triaxial tests on bituminous Traffic and Transportation Engineering 2(1): 17–29.
mixtures. Road Materials and Pavement Design 16(1): Kaloush, K.E. & Witczak, M.W. 2002. Tertiary flow char-
46–69. acteristics of asphalt mixtures. Journal of the Associa-
Bonaquist, R. 2008. Ruggedness testing of the dynamic tion of Asphalt Paving Technologists 71: 248–280.
modulus and flow number tests with the simple perform- Katman, H.Y., Ibrahim, M.R., Karim, M.R., Salim,
ance tester. NCHRP Report 629. National Coop- Mashaan N. & Koting S. 2015. Evaluation of per-
erative Highway Research Program, Transportation manent deformation of unmodified and rubber-re-
Research Board, National Research Council. Wash- inforced SMA asphalt mixtures using dynamic creep
ington, D.C., USA. test. Advances in Materials Science and Engineering
Brown, E.R. & Foo, K.Y. 1994. Comparison of uncon- 2015.
fined- and confined-creep tests for hot mix asphalt. Morea, F. & Zerbino, R. 2015. Wheel Tracking Test
Journal of Materials in Civil Engineering 6(2): (WTT) conducted under different standards. Study
307–326. and correlation of test parameters and limits. Materi-
CIRS. 2001. Performance-related technical specifica- als and Structures/Materiauxet Constructions 48(12):
tion for the construction and maintenance of road 4019–4028.
pavements, Inter-University Road Research Center, Muniandy, R. & Huat, B.B.K. 2006.Laboratory diame-
Ancona, Italy (in Italian). tral fatigue performance of Stone Matrix Asphalt with
Clec’h, P., Sauzeat, C. & Di Benedetto, H. 2009. cellulose oil palm fiber. American Journal of Applied
Multidirectional behavior of bituminous mixture. Sciences 3(9): 2005–2010.

241
Perraton, D., Di Benedetto, H., Sauzéat, C., De La Tayfur, S., Ozen, H. & Aksoy, A. 2007. Investigation of
Roche, C., Bankowski, W., Partl, M. & Grenfell, J. rutting performance of asphalt mixtures containing
2011. Rutting of bituminous mixtures: Wheel track- polymer modifiers. Construction and Building Materi-
ing tests campaign analysis. Materials and Structures/ als 21(2): 328–337.
Materiauxet Constructions 44(5): 969–986. Witczak, M.W., Kaloush, K., Pellinen, T., El-Basyouny,
Radziszewski, P. 2007. Modified asphalt mixtures resist- M. & Von Quintus, H.L. 2002. Simple performance
ance to permanent deformations. Journal of Civil tests for SUPERPAVE mix design, NCHRP Report
Engineering and Management 13(4): 307–315. 465, National Cooperative Highway Research Pro-
Roy, N., Veeraragavan, A. & Krishnan J.M. 2015a. Inter- gram, Transportation Research Board, National
pretation of flow number test data for asphalt mix- Research Council. Washington, D.C., USA.
tures. Proceedings of the Institution of Civil Engineers: Xu, T. & Huang, X. 2012. Investigation into causes of
Transport 168(3): 191–199. in-place rutting in asphalt pavement. Construction and
Roy, N., Veeraragavan, A. & Krishnan, J.M. 2015b. Influ- Building Materials 28(1): 525–530.
ence of confinement pressure and air voids on the Yu, B., Zhu, H., Gu, X., Ni, F. & Guo, R. 2015. Modified
repente creep and recovery of asphalt concrete mix- repeated load tri-axial test for the high-temperature
tures. International Journal of Pavement Engineering performance evaluation of HMA. Road Materials and
17(2): 133–147. Pavement Design 16(4): 784–798.
Sohm, J., Gabet, T., Hornych, P., Piau, J.-M. & Di Bene- Zaniewski, J.P. & Nallamothu, S.H. 2003. Evaluation of
detto, H. 2012. Creep tests on bituminous mixtures binder grade on rutting performance. Asphalt Technol-
and modelling. Road Materials and Pavement Design ogy Program. Department of Civil and Environmen-
13(4): 832–849. tal Engineering. Morgantown, WV, USA.
Sousa, J.B., Craus, J. & Monismith, C.L. 1991. Summary Zhu, T., Ma, T., Huang, X. & Wang, S. 2016. Evaluat-
report on permanent deformation in asphalt concrete. ing the rutting resistance of asphalt mixtures using a
Strategic Highway Research Program, report SHRP- simplified triaxial repeated load test. Construction and
A/IR-91–104. Washington, DC, USA. Building Materials 116: 72–78.
Sousa, J.B & Weissman, S.L. 1994. Modeling permanent
deformation of asphalt aggregate mixes. Journal of
the Association of Asphalt Paving Technologists 63:
224–257.
Taherkhani, H., Grenfell, J., Collop, A., Airey, G. &
Scarpas, A. 2007.Characterization of repeated creep
recovery behaviour of asphaltic mixtures. Proceeding
of the International conference on Advancedcharac-
terisation of pavement and soil engineering materials,
Athens, Greece, 20–22 June 2007.

242
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Degradation of asphalt mixtures due to impact method of compaction

P. Singh & A.K. Swamy

Department of Civil Engineering, Indian Institute of Technology, Delhi, India

ABSTRACT: For best mechanical performance, loose mixture of asphalt concrete needs to be com-
pacted. The achievable density of mixture depends on several variables including aggregate properties
(source, type, texture, shape and gradation), binder properties (grade, content) and process (compaction
temperature, compaction method, compaction energy). Among all the impact compaction methods,
Marshall Mix Design is more widely used. During the impact method of compaction, aggregate particles
may disintegrate due to the impact load, and abrasion. This disintegration may affect the aggregate
gradation and volumetric properties of mixtures. In this work the effect of impact compaction in aggregate
degradation and subsequent volumetrics using is presented. The four different mixtures using two binders
(neat binder-VG 30, modified binder-PMB 70) and two aggregate gradations were designed using Marshall
mixture design. These mixtures were compacted using 4 inch and 6 inch diameter Marshall moluds using
standard and modified Marshall compactor for different levels of compaction. The aggregates from the
compacted specimens were recovered and changes in aggregate gradation were examined. The statistical
analysis indicates that the compaction level, aggregate type and size of mould affect the degradation of
aggregates and mixture volumetrics significantly.

1 INTRODUCTION (Airey & Collop 2016, Hunter, Airey, & Collop

2004, Khan, Al-Abdul Wahab, Asi, & Ramadhan
To produce/design better performing Hot Mix 1998, Masad, Muhunthan, Shashidhar, & Harman
Asphalt (HMA) mixture, aggregate and asphalt 1999, Tashman, Masad, D’Angelo, Bukowski, &
binder needs to be combined optimally. The opti- Harman 2002, Yue, Bekking, & Morin 1995) have
mal conditions include proportion of individual been conducted to examine the effect of different
constituents, and volumetrics of resultant HMA compaction methods and their suitability to simu-
mixture. Some examples for mixture volumetrics late field compaction. The compaction is affected
include air voids, Voids in Mineral Aggregate by numbers of parameter such as compaction
(VMA), and Voids Filled with Bitumen (VFB). temperature, method of compaction, compaction
These volumetric properties control resistance energy, binder grade, binder content, aggregate
against rutting, fatigue as well as permeability source, aggregate type, aggregate texture, aggregate
(Brown 1990, Brown & Gibb 1999, Khan, Al-Abdul shape and aggregate gradation, etc. (Copple 1998,
Wahab, Asi, & Ramadhan 1998). The compacted Dubois, Roche, & Burban 2010, Linden, Mahoney,
HMA density indicates mechanical performance & Jackson 1989).
of mixture (Linden & Van Der Heide 1987). Since During the compaction, aggregate particles may
major portion of asphalt concrete mixture is occu- disintegrate due to the impact of the hammer, con-
pied by aggregate, properties of aggregate affect fining pressure, inter-particle friction, and applied
performance of mixture significantly. In order boundary conditions. Among all the compaction
to achieve optimum mechanical performance of methods used, Marshall compaction approach is
asphalt mixture, the specimen should have well dis- more susceptible for aggregate degradation (Airey,
tributed and representative aggregate structure as Hunter, & Collop 2008). Due to this degradation
exist in field samples (Goetz 1989, Hunter, Airey, of the aggregates the actual mixture properties
& Collop 2004). Such a representative aggregate could be different from the designed/calculated
structure can be guaranteed through proper selec- properties such as air voids, voids in mineral aggre-
tion of gradation and compaction procedure. gates, etc.
Due to differences in compaction mecha- The present research studied the effect of com-
nism, interaction between different aggregate paction method used in Marshall mixture design
particles, and boundary conditions, aggregate approach. The specific objectives of this research
skeleton found in laboratory specimen can be dif- are (i) to investigate the aggregate degradation
ferent from those observed in field. Several studies during Marshall compaction, (ii) to examine the

243
 factors affecting the aggregate degradation, (iii) 2.1.1 Binder
to study the effect on VMA due to aggregate Two different types of binder VG30 (unmodi-
degradation. fied) and PMB 70 (polymer modified) are selected
for the study. The suitability of these binders are
evaluated as per relevant Indian standards (IS 73
2 MATERIALS EMPLOYED FOR 2006; IRC SP53 2010) and was found satisfactory.
THE STUDY The test results along with requirements for VG 30
and PMB 70 are presented in Table 2 and Table 3,
2.1 Aggregate respectively.
Two gradations recommended in Indian specifi-
cation (IRC: 111 2009) with Nominal Maximum Table 2. Test results on unmodified bitumen (VG 30).
Particle Size (NMPS) of 13.2 mm & 19 mm were
selected. These two aggregate gradations were VG-30
obtained from the mid points of the upper and Relevant
lower limits specified in the IRC: 111–2009. The Test standard Requirement Result
curve of aggregate gradation corresponding to
19 mm and 13.2 mm were designated as A and B, Viscosity ASTM D Min. 350 487
@135°C, cSt 4402-2006
respectively. The selected gradations are plotted as
Penetration IS 1203-1978 50–70 58
Figure 1. All the aggregates used in present study @25°C, 0.1 mm
were evaluated for its suitability as per relevant Softening Point IS 1205-1978 Min. 47°C 48°C
Indian standards and were found to be satisfactory. Ductility, IS 1208-1978 Min. 75 Full
The results of these tests are presented in Table 1. @ 27°C, cm scale

Table 3. Test result on modified bitumen (PMB 70).

0>
-~ 80 ~-+--------------~--------------~---~--~
(/)
PMB 70
ro Relevant
0...
Q) Test standard Result Requirement
Er 60 ~-+---------------+---------?=---79------~
c:
Penetration IS 1203: 1978 55 50–90
~
Q) test, dmm
0... 40 ~-+--------------~~----~L-----~------~
Softening point IS 1205: 1978 61.1 Min 55
~
~ test, °C
3 Ductility test, IS 1208: 1978 81.2 60
§ 20 ~-+----~----~L-~--------------~------~
cm
()
Flash point IS 1209: 1978 317 Min 220
test, °C
0.1 1 10 Viscosity test at ASTM D 3.675 2–6
150°C, poise 4402-06
Seive Size (mm)
Elastic Recovery ASTM D 78 Min 75
Figure 1. Aggregate gradations. at 15°C, % 6084-06
Separation IS 1205: 1978 1.4 Max 3
difference
Table 1. Test results on aggregates. in softening
point, °C
Indian
Property Results Specification standard Thin Film Oven Test (TFOT) on Residue (IS: 9382–1992)

Cleanliness 1.61% Max. 5% passing IS 2386 Loss in IS 9382: 1992 0.56 Max 1
0.075 mm sieve Part 1 weight, %
Particle shape 20.22% Max. 30% IS 2386 Increase in IS 1205: 1978 4.3 Max 6
(Combined) Part 1 softening
Strength 27% Max. 30% IS 2386 point, °C
Part 4 Reduction in IS 1203: 1978 30.77 Max 35
21% Max. 24% IS 2386 penetration of
Part 4 residue, %
Water <2% Max. 2% IS 2386 Elastic recovery ASTM D 68 Min 50
absorption Part 3 at 25°C, % 6084-06

244
3 TESTING METHODOLOGY specimens were prepared at designed optimum
binder content. These specimens were prepared at
Using the selected aggregate and binders, four four different levels of compaction using standard
different mixtures were designed in this study. (4 inch Dia. 4.5 kg.weight) and modified (6 inch Dia.
The preheated aggregates were mixed with the 10.2 kg. weight) Marshall compactor (AASHTO T
four different content of binder and conditioned 245 2013, Asphalt Institute 1997). The number of
(AASHTO R30 1990) in convection oven for 2 blows corresponding to four levels of compaction
hour prior to compaction. Three replicates were using standard compactor were 25, 50, 75, and 100
fabricated at each binder content. These condi- one either face of specimen. The number of blows
tioned mixtures were compacted with 75 blows corresponding to four levels of compaction using
on each faces using standard Marshall compactor modified compactor was 36, 74, 112, and 150 on
(AASHTO T 245 2013, Asphalt Institute 1997, BS either face of specimen. Two aggregate gradation
EN 12697-30 2004). The compacted specimens and two binder types mentioned previously were
were investigated for Marshall parameters and used. Five replicates were prepared at each testing
volumetric properties (BS EN 12697-34 2004). condition. The mixture designations corresponding
Marshall mixture design approach was to deter- to each test condition is presented Table 6. The mix-
mine the optimum binder content of these mix- ture designation is combination of three alphabets
tures (Asphalt Institute 1997). and one numeric number. The first alphabet of the
To study the effect of compaction in aggregate designation is taken from the binder grade i.e. the
degradation, standard and modified Marshall V and P were used for VG 30 and PMB 70 binder,
respectively. The second alphabet (A or B) indicates
Table 4. Mix designations. the gradation while the third alphabet represent the
level of compaction. The numeric suffix (in fourth
4 inch dia. Standard Marshall samples
place) indicates the specimen size i.e. 1 is used for
Gradation A Gradation B 4 inch specimen and 2 is used for 6 inch specimen.
The compacted specimens were examined for the
Nos. of blows VG 30 PMB 70 VG 30 PMB 70 volumetric properties. The aggregates from the
compacted specimens were recovered using ignition
25 VAa1 PAa1 VBa1 PBa1 oven (?). Subsequently sieve analysis was conducted
50 VAa1 PAa1 VBa1 PBa1 to determine their modified gradation. The method-
75 VAa1 PAa1 VBa1 PBa1 ology adopted in this study is presented in Figure 2.
100 VAa1 PAa1 VBa1 PBa1

6 inch dia. Modified Marshall samples

4 INDICATORS OF AGGREGATE
36 VAa2 PAa2 VBa2 PBa2 DEGRADATION
74 VAb2 PAb2 VBb2 PBb2
112 VAc2 PAc2 VBc2 PBc2 The mixture volumetrics and aggregate grada-
150 VAd2 PAd2 VBd2 PBd2 tion (before and after impact compaction) thus

Binder VG30 & r--

PMB70

_____. 6 inch dia

1--
Samples

_____. Compaction at volumetrics Recovery of

four level
~
measurement r------. aggr~gate

Aggregate
f-+
4 inch dia
Samples
1--
•
Sieve analysis
-

•
Gradation A & D

Comparison of
di-fferent
gradations

Figure 2. Methodology of the study.

245
collected was used to address issues including (i) the change in VMA, PCV was calculated using the
are there any aggregates of a particular size that Equation 4.
is degrading more when compared to other sieve
size?, (ii) how degradation of individual sieve sizes VMA e re − VMA
M Befo MAAfter
affect overall changes in aggregate gradation?, and PCV = 100 × (4)
VMA
MABeefore
(iii) How aggregate degradation affects the mixture
volumetrics. To address these aspects three param-
eters i.e Percentage Change in Individual Particle where, VMAAfter is the void in mineral aggregates
Size (PCIPS), Gradation Change Index (GCI) and calculated using the modified aggregate grada-
Percentage Chance in VMA (PCV) are introduced tion after compaction, and VMABefore is the void
in this research. These parameters are discussed in in mineral aggregate calculated using the original
detail in following paragraphs. aggregate gradation. VMAAfter and VMABefore were
calculated using Equation 5.

4.1 Change in Passing (from Individual Sieves in PG

P
Percentage) M = 100 −
VMA s mb
(5)
Gsb
Due to the impact method of compaction, bigger
aggregates breaks to form smaller sized aggregates. where, PS is percentage of aggregate by weight of
The PCIPS was calculated using weight of aggre- total mix, Gmb is bulk density of mix, Gsb is bulk
gate retained at a particular sieve size before and specific gravity of aggregate. Gsb is dependent on
after impact compaction. The expression used aggregate gradation. The VMA before is calcu-
to calculate PCIPS is presented in Equation 1. lated by putting the specific gravity of the original
The impact compaction may cause breaking of (undegraded) aggregate.
the aggregates which will increase or decrease the
aggregate fraction passing from a particular sieve.
This parameter shows how the aggregate from a
5 RESULTS AND DISCUSSION
particular sieve got degraded with respect to the
original gradation.
5.1 Mixture design
Wti , Before − Wti , After As mentioned previously, Marshall mix design
PCIPS = 100 × (1) method was used to arrive at optimum proportion
Wti , Before
of individual constituents. The binder content cor-
responding to 4% of air void is adopted as Opti-
where Wti,Before = weight of aggregate at particle mum Binder Content (OBC) (Asphalt Institute
size i, before compaction, and Wti,After = weight of 1997). The OBC of mixtures containing VG30
aggregate at particle size i, before compaction. binder for aggregate gradation A an d B are 5.3 and
6.2, respectively. The OBC of mixtures containing
4.2 Change in overall gradation PMB70 binder for aggregate gradation A and B are
5.2 and 6.0, respectively. All mixtures designed ful-
To capture overall change in aggregate gradation, filled the requirements, specifiedin IRC:111–2009.
GCI was calculated using Equation 2. GCI is accu-
mulation of changes in aggregates of different
sieve sizes. 5.2 Change in Passing (from Individual Sieves
in Percentage)

C =
GCI
∑δ i ,C
CPP
(2)
As mentioned previously, to evaluate breakage
100 of bigger particles PCIPS was used. Variation
of PCIPS with sieve sizes for mixtures prepared
δ i ,CPP
CPP = Wti , Before
f − Wti , After (3) using PMB70 binder and aggregate gradation B
with standard and modified Marshall compactor
where, δi,CPP = change in cumulative percentage are presented in Figures 3 and 4, respectively. At a
passing at sieve size i. particular sieve size, the value of PCIPS increased
with increase in number of blows. This was true for
all mixtures indicating more degradation occurred
4.3 Changes in volumetrics
under impact loading. In general, numerical value
The VMA is a parameter that depends upon the of PCIPS increased with decrease in particle size
aggregate gradation and binder content. There- (except 0.3 mm sieve size). This indicates that parti-
fore any change in the aggregate gradation will cles of certain sizes are more resistant to impact load.
result in changes in VMA. In order to quantify This was true for all mixtures evaluated in this study.

246
 O.D75
+ O.D75
+
0.15
+ 0.15 _'}<
0.3
! 0.3
+
0.6
+ E 0.6
I
1.18 I .sw 1.18
I
.~
VJ
2.36 I ~
2.36
.ffi
4.75 I VJ 4 .75 I

B 136 Blows I

b 174 Blows
9.5 9.5

13.2 132

E' 19 19
0
.s
Q)
10 20 30 40 50 60 10 20 30 40 50 60

N
Ui O.Q75 ~I' O.Q75
' '
Q)
>
·a;
U)
0.15
+ 0.15

0.3
+ 0.3
+
0.6
+ E
.sw
0.6
~
1.18
+ 1.18 _t
.~
2.36 VJ 2.36
" '
-~
4.75
I VJ 4.75

B B
95 9.5

13.2 [ii2Blows 13.2

1150 Blows I
19
IT
10 40
19
!J 10 40
20 30 50 60 20 30 50 60

Change in Passing (from Individual Sieves in Percertage)

Figure 3. Variation of PCIPS with number of compaction blows and sieve size for mixture prepared using PMB 70
binder, Gradation B and standard Marshall compactor.

0.075
+ 0.075
+
+ ~·
0.1 5 0.15

0.3 ! 0.3
+
0.6 I E 0.6
I
1.18 .sw 1.1 8
I
.~
VJ
2.36 i w 2.36 I
·~
4.75 t VJ 4 .75 I

E'
.s
9.5

13.2

19
B 10 20 30
[36" Blows I

40 50 60
9.5

13.2

19
n~ 10 20 30 40
174 Blows

50 60
Q)
.!::!
U)
Q)
>
0.075
+ 0.075
·I·
·a; 0.15 0.15 I•
U)
0.3
+ 0.3
J
0.6
+ E
.s
0.6

1.18
+ -~
1.18
J
VJ
2.36
+ ~
.ffi
2.36
t
4.75
t VJ 4.75
I
9.5

13.2

19
B
IT
1112 Blows
9.5

13.2

19
B
tJ
[150 Blows [

10 20 30 40 50 60 10 20 30 40 50 60

Change in Passing (from Individual Sieves in Percertage)

Figure 4. Variation of PCIPS with number of compaction blows and sieve size for mixture prepared using PMB 70
binder, Gradation A and modified Marshall compactor.

247
5.3 Gradation Change Index (GCI)
VBd2I;;!!~~Do:~
VBc2
VBb2
VBa2
To investigate the overall effect on final grada- ~ -·······························································································

tion, GCI was computed for all mixtures evalu- VAc2

VAd2
VAb2
J::::::::::::~===<=
ated in this research. A comparison of GCI for
~:~ ~-
all mixtures evaluated in this study is presented in PBc2

Figure 5. For all mixtures, more number of blows PBa2

PBb1
resulted in higher numerical value of GCI. This (Jl

§
PAd2
PAc2
indicates that impact load on aggregate resulted 'g PAb2

in more degradation. For a particular mixture -~~:~~

~ VBc1
and number of blows, higher GCI was observed x VBb1
:::2: VBa1
with standard Marshall compactor. This again VAd1

indicates standard Marshall compaction leads to VAc1

VAb1
higher aggregate degradation. For a particular VAa1
PBd1
number of blows and compaction method, grada- PBc1
PBb1
tion A resulted in more degradation than grada- PBa1
PAd1
tion B. This indicated relatively coarser mixture PAc1
PAb1
results in more degradation than relatively finer PAa1
mixture. This is due to fact that more number of 10 12
points of contact are available with finer mixture Percent Change in VMA

than coarser mixture. Higher degradation was

observed with mixtures prepared with PMB 70 Figure 6. Percentage Change in VMA (PCV) for dif-
when compared to mixtures prepared with VG 30 ferent mixes.
binder. This can be attributed to lower binder con-
tent available for lubrication and and subsequent
restriction on aggregate reorientation.
as change in VMA. In this research, change
in the VMA is assessed using parameter PCV
5.4 Percent Change in VMA (PCV) Figure 6 shows the observed PCV or different
mixtures.
VMA is widely accepted volumetric property
For all mixtures, more number of blows
for HMA which ensures minimum binder con-
resulted in higher numerical value of PCV. This
tent in asphalt mixes. Further, VMA depends on
indicates impact load on aggregate resulted in
the aggregate gradation. This it is expected that
more aggregate degradation. For a particular
any change in aggregate gradation is reflected
mixture and number of blows, higher PCV was
observed with modified Marshall compactor. This
indicates modified Marshall compaction leads
VBd2
VBc2 ~
······································;;=
···· .............................................. to higher aggregate degradation as compare to
VBb2
VBa2
standard Marshall compactor. For a particular
VM2
VAc2
number of blows and compaction method, grada-
VAb2 tion A resulted in more degradationthan grada-
VAa2
tion B. This indicated relatively coarser mixture
PBd2~====~
PBc2]
results in more degradation than relatively finer
PBa2
PBb1
oo PAd2 mixture. This is due to fact that more number of
§ PAc2
points of contact are available with finer mixture
~ PAb2
~PAa2
·v.; VBd1
than coarser mixture.
~ VBc1
x VBb1
To check the significance of the difference in
~ VBa1 the aggregate gradation Least Significant Differ-
VAd1
VAc1 ence (LSD) (Abdi 2009, Williams & Abdi 2010)
VAb1
VAa1 test was performed on the values of different
PBd1
PBc1 mixes. Statistical analysis indicated that for stand-
PBb1
PBa1
ard Marshall compaction, obtained with 25 blows
PAd1
PAc1
and 75 blows were comparable with obtained
PAb1
PAa1
with 50 blows and 100 blows, respectively. Fur-
0.0 0.1 0.2 0 .3 0.4 0.5 0 .6 0.7
ther, obtained with 50 blows was significantly dif-
Gradation Change Index (CGI)
ferent from with 100 blows. This was true for all
mixtures compacted with standard compaction
Figure 5. Gradation Change Index (GCI) for different method except mixture prepared with VG30 and
Mixtures. gradation B.

248
6 CONCLUSIONS mixture design, and construction. NAPA Research and
Education Foundation, Lanham, MD.
In this study various HMA mixtures were prepared Brown, E.R. (1990). Density of asphalt concrete: How much
and evaluated for aggregate degradation dur- is needed? Technical report, National Center for Asphalt
Technology.
ing Marshall compaction. Two binder types, two Brown, S.F. & J.M. Gibb (1999). Effects of compaction
aggregate gradation and four levels of compaction on mechanical properties of asphalt mixtures. In Pro-
were considered in this study. It was found that ceedings of the 7th Conference on Asphalt Pavements for
aggregate gradation used before impact compac- Southern Africa. Gaborone Botswana: CAPSA.
tion was different from gradation found after com- BS EN 12697–30 (2004). Bituminous mixtures Test meth-
paction.The following conclusions can be drawn ods for hot mix asphalt Part 30 Specimen preparation by
based on the results obtained in this study. impact compactor. British Standards.
BS EN 12697–34 (2004). Bituminous mixtures-Test meth-
• The level of compaction significantly affected ods for hot mix asphalt Part 34 Bituminous mixtures.
the degradation of aggregate. Test methods for hot mix asphalt. Marshall test. British
• The aggregate gradation affects the degradation Standards.
more than other factors. In general coarser gra- Copple, F. (1998). Compaction the only way to quality
dation degraded more when compared for finer asphalt pavements, Materials and Technology Engineer-
gradation. ing and Science. The Materials and Technology Division
of the Michigan Department of Transportation (18).
• The size of the specimen affected the degrada- Dubois, V., C.D.L. Roche, & O. Burban (2010, June). Influ-
tion of aggregates. The 6 inch specimens were ence of the compaction process on the air void homo-
less affected by degradation as compared to geneity of asphalt mixtures samples. Construction and
4 inch specimens because of availability of more Building Materials 24(6), 885–897.
space for aggregate reorientation. Goetz, W.H. (1989). The evolution of asphalt concrete mix
• The type of binder also affected the aggregate design. In Asphalt Concrete Mix Design: Development of
degradation mainly through OBC. Availabil- More Rational Approaches. ASTM International.
ity of more lubrication through higher binder Hunter, A., G. Airey, & A. Collop (2004, January). Aggregate
content ensures particles are allowed to reori- Orientation and Segregation in Laboratory-Compacted
Asphalt Samples. Transportation Research Record: Jour-
ent themselves during compaction. This in turn nal of the Transportation Research Board 1891, 8–15.
decreases aggregate deterioration mechanism. IRC 111 (2009). Specification for Dense Graded Bituminous
• The degradation of aggregates significantly Mix. Technical report, Indian Road Congress.
affects the VMA of HMA mixes. The aggregate IRC SP53 (2010). Use of Modified Bitumen in Road Con-
gradation changes significantly due to degrada- struction. Technical report, Indian Road Congress. IS 73
tion which influences the VMA. (2006). Paving Bitumen - Specification. Technical report,
BIS.
Khan, Z.A., H.I. Al-Abdul Wahab, I. Asi, & R. Ramadhan
(1998, September). Comparative study of asphalt concrete
REFERENCES laboratory compaction methods to simulate field compac-
tion. Construction and Building Materials 12(67), 373–384.
AASHTO R30 (1990). Standard practice for mixture condi- Linden, F. & J. Van Der Heide (1987). Some Aspects of the
tioning of hot mix asphalt. American Association of State Compaction Of Asphalt Mixes And Its Influence on Mix
Highway and Transportation Officials. Properties (With Discussion). Volume 56.
AASHTO T 245 (2013). Method of Test for Resistance to Linden, R.N., J.P. Mahoney, & N.C. Jackson (1989). Effect
Plastic Flow of Bituminous Mixtures Using Marshall of compaction on asphalt concrete performance. Trans-
Apparatus. American Association of State Highway and portation Research Record (1217).
Transportation Officials. Masad, E., B. Muhunthan, N. Shashidhar, & T. Harman
Abdi, H. (2009). Experimental design and analysis for psy- (1999, January). Quantifying Laboratory Compaction
chology. Oxford University Press. Effects on the Internal Structure of Asphalt Concrete.
Airey, G.D. & A.C. Collop (2016, January). Mechanical and Transportation Research Record: Journal of the Transpor-
structural assessment of laboratory- and field-compacted tation Research Board 1681, 179–185.
asphalt mixtures. International Journal of Pavement Engi- Tashman, L., E. Masad, J. D’Angelo, J. Bukowski, &
neering 17(1), 50–63. T. Harman (2002, January). X-ray Tomography to Char-
Airey, G.D., A.E. Hunter, & A.C. Collop (2008, May). The acterize Air Void Distribution in Superpave Gyratory
effect of asphalt mixture gradation and compaction Compacted Specimens. International Journal of Pave-
energy on aggregate degradation. Construction and Build- ment Engineering 3(1), 19–28.
ing Materials 22(5), 972–980. Williams, L.J. & H. Abdi (2010). Fishers least significant dif-
Asphalt Institute (1997). Mix Design Methods for Asphalt ference (LSD) test. Encyclopedia of research design, 1–5.
Concrete and other Hot-Mix Types. Manual Series No. 2. Yue, Z., W. Bekking, & I. Morin (1995). Application of dig-
Brown, E., P.S. Kandhal, F.L. Roberts, Y.R. Kim, D.-Y. ital image processing to quantitative study of. Transporta-
Lee, & T.W. Kennedy (2009). Hot mix asphalt materials, tion Research Record 1492, 53–60.

249
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Evaluation of mixing temperature impact on warm mix

asphalt performance

A. Abed, N. Thom & J.R.A. Grenfell

The University of Nottingham, Nottingham, UK

ABSTRACT: This study presents a new approach for determining an optimum mixing temperature for
a chemical and an organic WMA additive based on correlating the mixing temperatures with performance
measures including the Indirect Tensile Stiffness Modulus (ITSM) and the Indirect Tensile Strength (ITS).
WMA was produced at 145, 130 and 115°C and compacted at 110°C, while the control HMA was mixed
at 160 and compacted at 150°C. The results revealed that the WMA produced using the chemical additive
did not have any dependency on mixing temperature, and the ITSM and ITS results were comparable to
the performance of the control HMA at all mixing temperatures with insignificant reduction (10%) in the
ITSM, as confirmed by statistical analysis. Therefore, its mixing temperature can be determined based on
other factors such as the distance between the construction site and the asphalt plant. However, the WMA
produced using the organic additive showed a significant dependency on mixing temperature; the higher
the mixing temperature, the higher the ITSM and ITS values. Furthermore, the ITSM results were compa-
rable or improved in comparison with the control HMA, but the ITS results were significantly lower than
those of the control HMA when the mixing temperature was 115°C based on statistical analysis. Based
on the relationship between the ITS and the mixing temperature, the same ITS as the control HMA was
achieved at 135°C;at lower temperatures, the ITS started dropping. Therefore, this temperature is sug-
gested as an optimum mixing temperature for the wax additive in this case.

Keywords: WMA, mixing temperatures, performance measures

1 INTRODUCTION Generally, these methods can be classified into

three broad categories (EAPA, 2010, Zaumanis,
Warm Mix Asphalt (WMA) refers to asphalt mix- 2010). The first is by using organic additives such
tures that are produced at temperatures lower than as wax to reduce bitumen viscosity. The second
conventional Hot Mix Asphalt (HMA) in order technique is by using chemical additives which
to achieve economic and environmental benefits. work as a surfactant to reduce the friction between
The temperature reductions for most WMA tech- the aggregate and the bitumen which eventually
nologies are between 10–40°C (National Asphalt improves aggregate coating at lower temperatures.
Pavement Association, 2016). This type of mixture The third method is by temporarily foaming the
was developed in Europe at the end of the 1990s bitumen during the mixing process by applying
as a response to signing the Kyoto protocol which either a small amount of water (water based) or
requires developed countries to reduce greenhouse using chemical foaming additives (water contain-
gas emissions in order to decrease CO2 emissions ing); the foamed bitumen can then be mixed at
during asphalt production (Prowell et al., 2012). lower temperatures.
Since then these mixtures have gained wide imple- All of the above techniques can offer a reduction
mentation in Europeand later in the United States in asphalt production temperatures, but the crite-
due to the expected benefits of reducing produc- ria to determine WMA production temperatures
tion temperatures. These benefits include reducing are still uncertain. Moreover, it is not clear whether
greenhouse gas emissions, lowering fuel consump- there are any effects of mixing/compaction tem-
tion, increasing asphalt haul distance, enabling the peratures on asphalt performance. To determine
asphalt to be compacted at cooler temperatures appropriate WMA production temperatures, cer-
and providing safer environments for workers in tain methods have been proposed. One is to deter-
asphalt plants (D’Angelo et al., 2008). mine mixing temperature by conducting dynamic
WMA can be produced by applying a number of viscosity tests (Silva et al., 2010, Zeng et al.,
additives or following several available techniques. 1997). Theoretically, this method assumes that the

251
mixing viscosity (i.e. 0.2 Pa.s) for modified bitu- 145, 130 and 115°C and control HMA samples at
men can be achieved at temperatures lower than 160°C; then the performance of WMA samples
for the base bitumen. Therefore, the temperature at has been correlated with mixing temperature and
which the modified bitumen exhibits this viscosity compared with the control HMA performance.
can be considered as the mixing temperature. How- The mixing temperature at which WMA shows a
ever, this approach is valid only for additives that comparable performance to the HMA is selected
alter bitumen viscosity such as organic additives. as the optimum mixing temperature.
Other studies (Wang et al., 2013, Zhao and Guo,
2012) used mixture workability as the criterion to
determine the production temperature of WMA. 2 MATERIAL DESCRIPTION
The logic for using workability in determining the
production temperature is based on achieving the 2.1 Aggregate and bitumen
same workability as HMA at its normal produc-
The aggregate used in this study was granite aggre-
tion temperature with WMA at lower tempera-
gate supplied from Bardon Hill Quarry in the UK.
tures. These studies have demonstrated that WMA
The aggregate was supplied in different separated
can be produced 15–30°C lower than HMA based
sizes (i.e. 14 mm, 10 mm, etc.), and by mixing dif-
on the mixture workability criterion. However,
ferent proportions from every size, the final grad-
depending only on the workability in determin-
ing matched the grading limits of 0/14 mm dense
ing production temperatures can be risky, because
asphalt concrete as specified by BS EN 13108-
the mixing temperatures may cause other impacts
1:2006 (British Standards Institution, 2006), as
on asphalt performance. Another study (Hurley
shown in Figure 2. The bitumen used in this study
and Prowell, 2005) suggested using the results of
was 40/60 grade with 49 being the actual penetra-
compaction (density) and rutting comparable to
tion. This grade can be used for surface and base
a control HMA in determining WMA produc-
course layers in the UK (BITUMINA, 2016).
tion temperatures. This study recommended 129
The bitumen content was determined to be 5.1%
and 110°C mixing and compaction temperatures
according to British Standard BS 4987-1:2005
respectively, to produce WMA using Sasobit. The
(British Standards Institution, 2005).
German Asphalt Paving Association (2009) sug-
gested using the density of a reference HMA to
determine the compaction temperature of WMA, 2.2 WMA additives description
as illustrated in Figure 1. Basically, the HMA den-
Two of the most commonly used WMA additives
sity is determined at 135–145°C; then the WMA
namely Sasobit and Cecabasewere selected in this
samples are compacted at a temperature range of
study. The first is an organic additive that consists
110–150°C; the temperature at which the WMA
of long chains of hydrocarbons with a carbon
exhibits the same density as the HMA can be
chain length between C45–C100. It can be made from
selected as the optimum compaction temperature.
natural gas by following the Fischer-Tropsch proc-
The aim of the study is to propose a rational
ess (Sasol, 2016).
method to determine WMA production temperatures
based on performance analysis. Basically, WMA
samples have been produced at temperatures of - - - Upper limit - - - Lower limit - - - Combined Agg.
100

90

2.420
80

..
c
E
-
0 Asphalt (reference mix)
Asp hi! It {with vim xity m odified hindrr) 70

.,
·;::; 2.410
60
~
c
·r;."'

~
iii 2,400 ~ 50
.s:; 0..
~
"'
::;; 2,390
"" 40

(;
!!' 2,380
·;;;
I vr,A= VT-6YT
30

20
.,
c vr,. VT
0 10
2,370
100 110 120 130 140 150 160 170
0
Compaction t emperature ['C] 0.01 0.1 10 100
Sieve Size (mm)
Figure 1. WMA compaction temperature determina-
tion (German Asphalt Paving Association, 2009). Figure 2. Aggregate gradation.

252
 Its principal properties include high melting 70
because of the long carbon chain, low melt viscos- 61.7
ity and high crystallinity. Therefore, it offers lower 60
mixing viscosity when applied to the asphalt, and 50.7
better rutting resistance due to the hardening and 50
crystallinity effects. Hurley and Prowell (2005)
evaluated the use of Sasobit in producing WMA. u
0 40
They stated that Sasobit could reduce bitumen vis- .s
0
cosity at high temperatures and may not affect the 0.. 30
viscosity at normal pavement temperatures. They .s
bJ)

also observed that Sasobit retarded the aging proc- Q)

~
20
q::
ess which may increase rutting potential. However, 0
Vl
according to Jamshidi et al. (2013), Sasobit can 10
improve rutting resistance at intermediate temper-
atures, but may make the pavement susceptible to 0
low-temperature cracking. BB BB+S BB+C
On the other hand, Cecabase is a liquid sur-
factant that allows the asphalt to be produced and Figure 3. Softening point test results.
compacted about 40°C lower than in traditional
hot mix asphalt while the mixture workability and 60
compaction are improved even with the incorpora- 50
tion of high percentages of RAP (Arkema Group, 50 48
2016). Goh et al. (2013) studied the impacts of
applying 0.2, 0.35 and 0.5% of this additive on the s
performance of WMA with respect to dynamic .§ 40 34
modulus, moisture resistance, fatigue cracking and 0
~
rutting. Based on their results, they concluded that .8 30
the application of Cecabase: reduced the dynamic ~
.!:l
modulus regardless of the additive dose, allowed Q)
~
WMA samples to pass the 0.8 tensile strength Q) 20
0...
ratio, significantly improved the fatigue life regard-
less of the dosage, but increased rutting potential, 10
the increase having a direct relationship with addi-
tive quantity. It can be concluded that the two
0
additives work in different ways, and may improve
BB BB+S BB+C
certain measures whilst negatively impacting oth-
ers. Therefore, a careful evaluation when selecting
an additive is indeed required. Figure 4. Penetration test results.

460
3 ADDITIVE EFFECT ON BITUMEN 451
CHARACTERISTICS AND RHEOLOGY 450 443
440
In order to understand and the behavior of the
430
selected additives and their influence on bitu-
men characteristics, softening point, penetration fr 420
and rotational viscosity tests were conducted. c
"§ 410
Figures 3, 4 and 5 show the test results for the base (.)

bitumen (denoted by BB), the Cecabase modified ~ 400 391 -

bitumen (BB+C) and the Sasobit modified bitu- 390
men (BB+S). Based on the penetration results, it

•
can be seen that the Sasobit significantly reduced 380
the penetration while the Cecabase did not alter 370
this property. This result is also consistent with the
360
softening point results which show that the Sasobit
increased the softening point while the Cecabase BB BB+S BB+C
hada negligible effect. This effect of the Sasobit
can be attributed to the crystallisation or stiffen- Figure 5. Viscosity test results.

253
ing effect of this additive which is believed to form Again, this is consistent with the nature of this
a lattice structure that prevents movement of the additive which does not alter or modify bitumen
base binder molecules when the Sasobit is at tem- properties but is said to reduce the frictional forces
peratures lower than its melting point (100°C) with the aggregate.
(Jamshidi et al., 2013). Meanwhile, the results for
Cecabaseare expected because of the role of this
additive, which is meant to reduce the friction 4 SAMPLE FABRICATION
between the bitumen and the aggregate and not
to alter bitumen rheology. The viscosity results are To prepare the necessary testing samples, a
also consistent with the previous results; Cecabase 0/14 asphalt concrete mix with 5% air voids was
hada very minor effect on the viscosity while Saso- designed. The asphalt was mixed in a bowl mixer
bit reduced the viscosity at 135°C to about 87% that contained heaters and thermo-sensors to con-
that of the base bitumen. trol mixing temperatures; it was then compacted
Furthermore, to study the effect of the additives using a roller compactor in accordance with British
on bitumen rheology, a frequency sweep test using Standard BS EN 12697-33. The samples produced
a DSR over a temperature range of 0–80°C was in this way were slabs with dimensions 305 by
conducted. Figure 6 shows the complex modulus 305 mm. After demolding the slabs, five cylindrical
and the phase angle master curves. The Sasobit samples with 100 mm diameter were cored from
modified bitumen clearly shows an improvement each slab. The maximum density of the mix was
in the complex modulus and the phase angle in calculated by following procedure C of the British
comparison with the base bitumen at all tem- standard BS EN 12697-5-2009 using foil wrapping.
peratures. The complex modulus for the modified By this procedure, control HMA samples mixed at
bitumen is higher than that of the base bitumen 160 and compacted at 150°C were produced, while
while the phase angle is lower, which means that non-standard HMA (Hot Mix Asphalt produced
the modified bitumen became more elastic and at lower temperatures than standard) and WMA
stiffer as the lag between the stress and the strain samples were mixed at 145, 130 and 115°C and
was reduced. This indicates that the Sasobit can compacted at 110°C. The reason for compact-
improve rutting resistance at high temperatures, ing the samples at the same temperature was that
but may make the pavement more susceptible to the focus of this study is to evaluate and analyse
low-temperature cracking. On the other hand, the impacts of mixing temperature on asphalt per-
Cecabase modified bitumen did not show any sig- formance. Moreover, Jalali (2016) found that com-
nificant modification to the rheological properties. pacting asphalt with or without WMA additives
at lower temperatures (e.g. 100°C) does not have
a significant impact on performance. Therefore, it
• BBG* • BB+S G* • BB+C G* was believed that mixing temperature has amore
.6 BBc5 .. BB+S c5 .. BB+CcS important impact than compaction temperature.
I 00000 ,--,,...,....._ _-,-------,-----,--------,-- 90 Table 1 shows some of the volumetric properties
of the samples. It can be seen that the air voids
80 are generally within ±1.7% of the target air voids.
Furthermore, the non-standard HMA samples
70 achieved the required compaction level without
using any additives, which is consistent with the
60 ~
Q) findings of Jalali (2016). However, the compac-
Eo
Q) tion effort was less with WMA additives. The com-
50::::::,
paction process using roller compactor consisted
Q)
of four levels of compaction pressure, 25, 40, 55
40 ]l
and 70 psi. All asphalt samples required 30 passes
"'
Q)
for the first three compaction levels (10 passes at
30 .2
~ each level), the difference was only in the number
0.1 of passes at 70 psi compaction pressure. Therefore,
20
only this value was used to assess the workability
0.01 10
of the samples. Table 1 indicates that samples con-
taining Cecabase required only 2.6 passes (aver-
0.001 +---~----~----~---+----+ 0 age of three samples), and samples containing
0.0001 O.ol 100 10000 1000000 Sasobit took 3.6 passes, while the non-standard
Reduced frequency (Hz) HMA samples all required 8 passes to achieve
the required density. This clearly shows the posi-
Figure 6. Frequency sweep test results. tive impact of the additives on asphalt workability.

254
Table 1. Volumetric properties of the samples. using Linear Variable Displacement Transducers
(LVDTs) that are fixed at the sides of the speci-
Mixing Air Compaction men, as shown in Figure 7. After determining the
Bitumen temp. Compaction voids effort, No. of stress and converting displacement to strain, the
type °C temp.°C % passes at 70 psi
stiffness modulus can then be calculated using the
BB 160 150 5.0 1 dividing the stress by the strain.
BB 145 110 5.4 8 Figure 8 shows the ITSM test results; every
BB 130 110 4.3 8 point on this figureis, in fact, an average of fifty
BB 115 110 4.0 8 readings (five cores times 10 readings for each),
BB+C 145 110 6.3 3 so the results are statically robust. They show
BB+C 130 110 4.9 3 that the non-standard HMA samples do not have
BB+C 115 110 5.7 2 a dependency on mixing temperature since the
BB+S 145 110 6.0 2 ITSM values are quite comparable to that of the
BB+S 130 110 6.7 6
BB+S 115 110 6.1 3

Nevertheless, the samples containing Sasobit had

a slightly higher air voids content than other sam-
ples. This canbe attributed to the melting point of
this additive which is between 80–115°C (Gandhi,
2008). Therefore, compacting WMA containing
Sasobit at temperatures lower than the Sasobit
melting temperature can have a negative impac-
ton the compaction process due to the hardening
effects of this additive at these temperatures.

5 TESTING PROGRAMME AND

RESULTS ANALYSIS

5.1 Indirect Tensile Stiffness Modulus (ITSM)

The stiffness modulus is an important and funda-
mental characteristic that is used to analyse pave-
ment response under traffic loading by applying
the theory of elasticity. Furthermore, traffic load
distribution over underlying pavement layers Figure 7. Nottingham asphalt tester.
depends on this modulus; the higher the modulus,
the better the load distribution, and vice versa. 5,000
Moreover, this modulus can be determined for - BB
4,800
cylindrical samples which can be cored from a road - BB+S
or a laboratory-manufactured slab. Additionally, 4,600 - BB+C
this is a non-destructive and quick test, so the same 4,400 - . - Control HMA
asphalt samples can be tested repeatedly which is
~ 4,200
economical and saves time. Therefore, this test was ::E
implemented in this study to evaluate asphalt sam- ::E 4,000
{/]
ple performance. The test was conducted accord- t:: 3,800
ing to British Standard BS EN 12697-26:2004
3,600
(British Standards Institution, 2004) using the
..
-
NAT (Nottingham Asphalt Tester) machine, as 3,400
shown in Figure 7. Prior to the test, the specimens
3,200
•
were stored for 24 hours in a cabinet at the test-
ing temperature 20°C to assure that had reached 3,000
the required temperature. The test applies a hori- 100 llO 120 130 140 150 160
zontal displacement within the range 5–7 microns Mixing temp. °C.
and measures the stress required to provide that
displacement. The displacement was measured Figure 8. ITSM test results.

255
control HMA. Therefore, it can be concluded that had the required testing temperature and to allow
for the unmodified bitumen, the ITSM does not for any stresses from the ITSM test to relax. The
depend on the mixing temperature even down to test was conducted using an Instron Servo hydrau-
115°C. Moreover, the WMA produced using the lic loading frame as shown in Figure 8. The test
Cecabase had almost the same behavior, within sig- was performed by applying a monotonic load that
nificant dependence of the ITSM on mixing tem- vertically deforms the sample at rate of 50 mm/min
perature. However, Cecabase reduced the ITSM by as recommended by the test standard. The load at
about 10% regardless of the mixing temperature, failure was recorded, and the strength calculated
and it is believed that the thin film created by this using the following equation:
additive to reduce the friction with the aggregate is
the cause of this reduction in the stiffness modulus. 2P
In contrast, Sasobit has a significant dependence T =
ITS (1)
πD
DH
on ITSM and a direct relationship with mixing
temperature; the higher the mixing temperature, where P is the load at failure, d is the diameter and
the higher the stiffness modulus and vice versa. h is the height of the specimen.
Furthermore, at all mixing temperatures, Sasobit Figure 9 presents the ITS test results. It is clear
increased the ITSM. It is believed that the stiff- that the non-standard HMA samples show a slight
ening effect of the Sasobit is the reason for this dependence on mixing temperature, and reducing
increase in the modulus. Also, it is believed that the the mixing temperature may very slightly reduce
dependency is related to the mixing of the Sasobit the ITS. The samebehavior can be seen for the
in the base bitumen; the higher the mixing tem- Cecabase, with a negligible dependence of the
perature, the better the mixing, eventually leading strength on the mixing temperature and with ITS
to higher stiffness modulus. To validate this inter- values comparable to the control HMA at all mix-
pretation, a T-test statistical analysis between the ing temperatures. However, the strength of the
control HMA and the other asphalt types was per- samples containing Sasobit shows a high depend-
formed, as shown in Table 2. It can be seen that ency on mixing temperatures and mixing lower
all ITSM results are comparable or insignificantly than 130°C can cause a significant reduction in
different from the control HMA, except the Saso- the tensile strength. Mixing at this temperature or
bit samples produced at 130 and 145°C which were higher gives an acceptable or possibly improved
significantly stiffer than the control HMA. performance in comparison with the control speci-
mens. To validate these results, a T-test analysis was
5.2 Indirect Tensile Strength (ITS) conducted between the control HMA and all other
asphalt types, as shown in Table 3. The statistical
In order to understand the effect of the mixing analysis shows that all ITS results except one are
temperatures and WMA additives on asphalt ten- insignificantly different from the control HMA.
sile strength, the ITS test was performed according
to British Standard BS EN 12697-23-2003 (British
Standards Institution, 2003). Since this testis con-
ducted on cylindrical samples, it can be conducted
after ITSM testing.
However, all samples were stored at 20°C for
24 hours in a cabinet to assure that the samples

Table 2. Statistical analysis of the ITSM results.

Asphalt Mixing P
type temp. °C value Significant

BB 145 0.34 NO
BB 130 0.39 NO
BB 115 0.41 NO
BB+C 145 0.34 NO
BB+C 130 0.09 NO
BB+C 115 0.08 NO
BB+S 145 0.005 YES
BB+S 130 0.04 YES
BB+S 115 0.42 NO
Figure 9. The test set-up for ITS test.

256
Table 3. Statistical analysis of the ITS results. 6 SUMMARY AND CONCLUSIONS
Asphalt Mixing P This study presents a new approach to determine
type temp. °C value Significant appropriate WMA mixing temperatures by cor-
BB 145 0.46 NO relating the mixing temperature with performance.
BB 130 0.49 NO Two WMA additives were selected to achieve the
BB 115 0.26 NO aim of this study, Cecabase and Sasobit. WMA and
BB+C 145 0.44 NO non-standard HMA cylindrical specimens were pro-
BB+C 130 0.36 NO duced at 145, 130 and 115°C, and ITSM and ITS
BB+C 115 0.14 NO tests were then performed to analyse the mixture’s
BB+S 145 0.34 NO performance. Based on the test results and the sta-
BB+S 130 0.4 NO tistical analysis, the following conclusions are drawn:
BB+S 115 0.03 YES
1. Based on the bitumen’s basic characteristics,
Sasobit decreased the penetration and increased
1,480 the softening point, and this clearly shows the
1,460 crystallisation effect of this additive, whereas
1,440
the Cecabase did not change these characteris-
tics. Sasobit reduced bitumen viscosity by about
1,420
15% (at 135°C) while Cecabase did not alter the
1,400 binder viscosity.
"' 1,3XO
0..
:.::
2. The complex modulus and phase angle mas-
if> ter curves showed that Sasobit increased the
!::: 1,360 - BB
1,340
modulus and decreased the lag at temperatures
- BB+S from 0–80°C, which indicates that Sasobit may
1,320
- BB+C improve rutting resistance at high temperatures
1,300 but may make the pavement more prone to low
1,280 - + - Control HMA temperature cracking.
1,260 3. Non-standard HMA was produced at low tem-
100 110 120 130 140 ISO 160 peratures, 130 and 115°C, and compacted to the
Mixing temp. °C. required level of compaction at 110°C without
significantly compromising mixture performance.
Figure 10. ITS results. This suggests the ability to produce WMA with-
out applying WMA additives. However, compac-
In other words, these results belong to the same tion effort in the presence of WMA additives
distribution and vary about the mean. The only was lower, which shows the necessity of applying
result that showed a significant difference is that additives to improve compactibity and increase
of the Sasobit sample mixed at 115°C, which is sig- workability at low compaction temperatures.
nificantly lower than the control HMA strength. 4. The WMA samples produced with Sasobit
Therefore, mixing Sasobit modified bitumen at showed a significant dependence of the ITSM
lower than 130°C should be avoided because it and the ITS on mixing temperature, which sug-
causes a significant reduction in the ITS values in gests the need correlate those measures with
comparison with the control HMA. Moreover, the mixing temperature in order to determine an
WMA specimens produced using Sasobit show appropriate mixing temperature.
that at 135°C mixing temperature these samples 5. Based on the ITSM and ITS results, WMA using
achieved the same ITS as the control samples, so Sasobit can be produced at 135°C without nega-
this temperature may be considered as the opti- tively influencing those measures. However, at
mum mixing temperature for WMA production lower than this temperature, the ITS may be sig-
using Sasobit with this particular base binder. nificantly decreased. Therefore this temperature
However, it was observed during the asphalt mixing is proposed as the optimum mixing temperature
process that a longer mixing time may be required for Sasobit in this case.
at temperatures such as 130 or 115°C to provide 6. The WMA samples produced using Cecabase
complete coating; this can be therefore a further were successfully produced as low as 115°C
possible criterion to determine a suitable mixing without altering the ITSM or the ITS. There-
temperature, especially with WMA chemical addi- fore, the mixing temperature for this additive
tives, and the authors are currently developing a can be determined based on other factors such
new approach using image processing to analyse as the distance between the asphalt plant and
aggregate coating during the mixing process. the construction site.

257
REFERENCES Asphalt Technology. Procedia—Social and Behavioral
Sciences, 96, 2782–2790.
Arkema Group. 2016. Cecabase RT [Online]. Available: Hurley, G.C. & Prowell, B.D. 2005. Evaluation of Saso-
http://www.arkema.com/en/products/product-finder/ bit® For Use in Warm Mix Asphalt NCAT, National
range-viewer/Cecabase-RT/ [Accessed 15/03 2016]. Ceneter of Asphalt Technology, Auburn University,
Bitumina. 2016. Hi-Tech Pavement Binders [Online]. Alabama, U.S.
Available: http://www.bitumina.co.uk/product-sheet- Jalali, V. 2016. Performance of Warm Mix Asphalt: Com-
data.php [Accessed 04/08/2016 2016]. pacted at Reduced Temperatures.Philosophy of Doc-
British Standards Institution 2003. BS EN 12697-23- torate The University of Nottingham.
2003, Bituminous mixtures—Test Methods for Hot Jamshidi, A., Hamzah, M.O. & You, Z. 2013. Perform-
Mix Asphalt — Part 23: Determination of The Indi- ance of Warm Mix Asphalt containing Sasobit®:
rect Tensile Strength of Bituminous Specimens. State-of-the-Art. Construction and Building Materials,
British Standards Institution 2004. BS EN 12697- 38, 530–553.
26:2004, Bituminous mixtures—Test Methods for Hot National Asphalt Pavement Association. 2016. Warm
Mix Asphalt—Part 26: Stiffness. Mix Asphalt [Online]. Available: http://www.warm-
British Standards Institution 2005. BS 4987–1:2005, mixasphalt.com [Accessed 09 March 2016].
Coated Macadam (asphalt concrete) for Roads and Prowell, B.D., Hurley, G.C. & Frank, B. 2012. Warm Mix
Other Paved Areas -Part 1: Specification for Constitu- Asphalt: Best Practices. National Asphalt Pavement
ent Materials and for Mixtures. Association, Quality Improvement Publication 125,
British Standards Institution 2006. BS EN 13108–1:2006, 3rd Edition.
Bituminous Mixtures -Material Specifications-Part Sasol. 2016. Sasobit Asphalt Technology [Online].
One: Asphalt Concrete. Available: http://www.sasolwax.us.com/sasobit.html
D’angelo, J., Harm, E., Bartoszek, J., Baumgardner, G., [Accessed 01 March 2016].
Corrigan, M., Cowsert, J., Harman, T., Jamshidi, Silva, H.M.R.D., Oliveira, J.R.M., Peralta, J. & Zoorob,
M., Jones, W., Newcomb, D., Prowell, B., Sines, R. & S.E. 2010. Optimization of Warm Mix Asphalts Using
Yeaton, B. 2008. Warm Mix Asphalt: European Prac- Different Blends of Binders and Synthetic Paraffin
tice. American Trade Initiatives, Washinton D.C., Wax Contents. Construction and Building Materials,
United States. 24, 1621–1631.
EAPA 2010. The Use of Warm Mix Asphalt. EAPA posi- Wang, C., Hao, P., Ruan, F., Zhang, X. & Adhikari, S.
tion paper, European Asphalt Pavement Association, 2013. Determination of The Production Temperature
Brussels, Belgium. of Warm Mix Asphalt by Workability Test. Construc-
Gandhi, T. 2008. Effects of Warm Asphalt Additives on tion and Building Materials, 48, 1165–1170.
Asphalt Binder and Mixture Properties Doctor of Zaumanis, M. 2010. Warm Mix Asphalt Investigation.
Philosophy, Clemson University. Master of Science, Riga Technical University.
German Asphalt Paving Association 2009. Warm Mix Zeng, E., Wu, C. & Zhang, J. 1997. Determining Mixing
Asphalts: Tips and Tricks Developed by Profession- and Compaction Temperatures of Evotherm® Warm
als For Professionals, Deutscher Asphaltverband e.V. Mix Asphalt Using 100% Reclaimed Asphalt Pave-
Germany. ment. Chinese Society of Pavement Engineering.
Goh, S.W., Hasan, M.R.M. & You, Z. 2013. Perform- Zhao, G.-J. & Guo, P. 2012. Workability of Sasobit Warm
ances Evaluation of Cecabase® RT in Warm Mix Mixture Asphalt. Energy Procedia, 16, 1230–1236.

258
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Effect of compaction temperatures on the warm mix asphalt

volumetrics and stability

H.I. Ozturk
Middle East Technical University, Ankara, Turkey

O.C. Pamuk
Ted University, Ankara, Turkey

ABSTRACT: Warm Mix Asphalt (WMA), the technology that allows significant reduction in the mix-
ing and compaction temperatures, has been commonly used in USA and Europe since the early 2000’s.
Although WMA additives decrease these temperatures, in some practices production is maintained with-
out lowering temperatures to increase the compactibility in the field. Therefore, this paper studies the effect
of Compaction Temperature (CT) on the mixture stability and volumetric properties. Since Superpave
gyratory compactor is not sensitive to temperature changes, Marshall hammer is used for the sample prep-
aration. The test matrix includes a control mixture (HMA) and two different WMA mixtures; prepared
with a non-foaming additive (Sasobit) and a foaming additive (Advera). Additionally, two different dos-
ages for each additive are used in WMA samples preparation to observe the effect on the compaction. The
mixture properties are limited with one dense gradation and one penetration grade binder. Four replicates
of each mixture are prefabricated at four different CT`s. The temperatures are selected as the typical CT
of HMA, 15, 30, and 45°C lower than HMA. Two of the replicates are used to determine the stability of
the mixtures and the rest of the samples are used to analyze the gradation changes during the compaction
process. Accordingly, these samples are burned in the ignition oven and re-sieved to quantify the degrada-
tion of the aggregates during the compaction and mixing. It was observed that WMA is less susceptible
to CT changes than HMA and the degradation pattern for these mixtures are similar at different CT’s.

1 INTRODUCTION total asphalt mixture production (Hansen & Cope-

land 2013). Although these WMA technologies take
1.1 WMA additives significant attraction, there are still many concerns
in these applications since there are no complete
Warm Mix Asphalt (WMA), which is the general
specifications and standards. The significant prop-
term used for the asphalt pavements produced and
erties like the mixing, compaction temperatures and
placed at lower temperatures, is introduced to the
dosages are determined based on the suggestion of
pavement industry to overcome the various chal-
the additive industry according to their limited expe-
lenges of Hot Mix Asphalt (HMA) (Ozturk 2013).
riences. However, a more solid approach is needed as
WMA usage is increasing worldwide majorly
the usage is rapidly increasing globally.
because of its engineering, environmental, and
This study focuses on the compaction behavior
economical benefits (De Groot et al. 2001, Larsen
of additive-based WMA mixtures in the laboratory
et al. 2004, Cervarich 2003, Prowell et al. 2007).
under different compaction temperatures. The need
There are various WMA technologies based on
for this study is emerged because of the new trends in
foaming, chemical additives/surfactants, and non-
the construction industry. Although WMA technol-
foaming additives in practice, and also new tech-
ogy is essentially introduced for lowering the produc-
nologies have been developing continuously.
tion temperatures, these technologies are started to
According to European Asphalt Pavement Asso-
be used as compaction aid by keeping the produc-
ciation (EAPA) (2014), countries like Germany, Den-
tion temperatures at high level. This study presents
mark, Turkey, Sweden, Norway, France, and Czech
the change in volumetrics and stability of the mix-
Republic initiated researches on WMA and also
tures compacted at different temperatures. In addi-
constructed test sections. On the other hand, USA
tion, it analyzes the change in gradation/degradation
significantly increased the WMA usage in the recent
during compaction due to breaking of aggregates.
years. In 2012, WMA share was about 24% of the

259
1.2 Laboratory compactor selection Table 1. Asphalt mixture properties (S.G. = specific
gravity, Theo. = Theoretical).
Laboratory compactor selection in this study is one
of the major components of the research to ana- Sieve Sizes % Passing Sieve Sizes % Passing
lyze the effect of Compaction Temperature (CT) on
volumetrics, stability and degradation. In this study, ¾″ 100 No.16 41.9
Marshall Hammer is selected upon Superpave Gyra- ½″ 98.3 No.30 29.3
tory Compactor. Although Superpave Gyratory 3/8″ 88.7 No.50 13
Compactor was developed to better represent the No.4 63.8 No.100 6.8
field compaction accounting the traffic loading and No.8 52.8 No.200 4.4
environmental conditions by the Strategic Highway Aggregate type Basalt
Research Program, it was proven by many reseach- Aggregate S.G. (Gsb) 2.484
ers that this type of compactor is not sensitive to Binder grade Pen 50/70
compaction temperature (Roberts et al. 1996). Binder S.G. (Gb) 1.02
Initially, Hunter & Brown (2001) investigated the Binder content 7.5%
effect of reheating and compaction temperature on Mixing temperature 150°C
volumetrics of HMA Mixtures. Due to the analy- Compaction temperature 140°C
sis on eight different mixtures, it is concluded that Target air void content 4.00%
in the range of ±14°C of compaction temperature, Theo. Max. S.G. (Gmm) 2.310
gyratory compaction has no effect on volumetric Theo. Bulk S.G. (Gmb) 2.218
properties of HMA mixtures. Then, Bahia et al.
(2001) stated similar findings on HMA mixtures
prepared with modified binders. However, their 1200
study was limited with four different mixtures and ,_,_1 000
no field validation was performed. Furthermore, rr.J
~ 800
~
Delgadillo & Bahia (2008) studied the HMA com-
~ \
paction under different temperatures and pressures -~ 600
in the laboratory and at the field. It was revealed ~
that the temperatures used in Gyratory compactors "'
~ 400
are higher than it needs to be in the field. In another Ji'
study, two different mixtures prepared with three 200 ... [ll r-
different WMA additives (Sasobit, Evotherm and 000 ~-~
Zeolite) were compacted at four different tempera- 100 120 140 160 180 200
tures. It was also proved that addition of WMA Temperature (OC)
additive had no effect on temperature sensitivity
of Superpave compactor (Hurley et al. 2006a). Figure 1. Viscosity of the binder.
However, Marshall Hammer recognized to be sus-
ceptible to compaction temperature (Hurley et al.
according to ASTM D 5: Penetration of Bituminous
2006a, Prowell et al. 2011). Therefore, instead of
Materials, as shown in Table 1. The binder viscosity
Superpave gyratory compactor, Marshall Hammer
is determined according to ASTM D 4402: Viscosity
is used for sample compaction in this study. Addi-
Determination of Asphalt Binder Using Rotational
tionally, Marshall Design procedure is still used by
Viscometer, as shown in Figure 1. Accordingly, the
many agencies and contractors in the world.
optimum mixing and compaction temperatures for
the control mixtures are estimated as 150°C and
140°C, respectively. The number of blows is deter-
2 MATERIALS AND TEST MATRIX mined as 75 blows. Many studies stated that 75 blows
is necessary for a density equal to Superpave gyration
The job mix formula in this study is limited by one (Rushing et al. 2010, Christensen & Bonaquist 2006).
aggregate gradation, one binder type and one aggre- At that compaction level, the optimum binder con-
gate source. Accordingly, an E1 fine gradation mix tent is determined to be 7.5%, according to Marshall
is selected as the control mixture (HMA), which Mix design procedure. There are two reasons for the
belongs to a previously used Superpave design. The high optimum content: (i) high fine percentage in the
mix gradation consists of 36.2% of coarse aggre- mix, (ii) high absorption capacity of aggregates (2.1%
gates, 59.4% of fine aggregates and 4.4% of filler, for coarse aggregates and 3.3% for fine aggregates).
as detailed in Table 1. Aggregate type is selected as The optimum compaction temperature for the
Basalt, obtained from a local quarry. The binder type control mix is assumed to be the highest compac-
is determined to be a virgin binder Pen 50–70. The tion temperature for the overall study. The other
penetration grade of the binder is measured to be 64 compaction temperatures are determined to be

260
15°C, 30°C, and 45°C lower than the optimum analyze the degradation of the aggregates during
temperature, in order to analyze the effect of the the compaction process. As a summary, overall 80
CT drop on the compaction behavior, and vice samples (5 mixtures × 4 CT × 4 replicates) are pre-
versa for WMA. pared for four replicates of five different mixtures
In this study, WMA mixtures are prefabricated at four different compaction temperatures.
either using a non-foaming additive (Sasobit) or a
foaming based additive (Advera). These additives
2.1 Sample preparation procedure
are selected based on the ease of supplier acces-
sibility. In addition, both additives are commonly The samples are prepared according the Marshall
used in WMA construction in USA. 2.5–3 millions mix design procedure.
of ton of WMA mixtures prepared with Sasobit
i. The preparation procedure for mixtures with
is placed between 2005–2012 (Prowell et al. 2012).
foaming additives is as follows: Initially, pre-
Besides, over 1 millions of ton WMA mixtures pre-
heated aggregates are placed in pre-heated mix-
pared with Advera is used in USA between 2006–
ing bowl and a crater is formed at the mid of
2012 (Prowell et al. 2012).
the sample. The binder is poured into crater and
Sasobit is a synthetic paraffin wax, which com-
then the pre-weighed Advera is added to binder.
bines hot coal and natural gas with steam with the aid
With the help of a glass rod, the additive is rap-
of a catalyst. Sasobit has higher viscosity than binder
idly mixed with binder. Then, the bowl is placed
below its melting point and lower viscosity than
in the mechanical mixer, equipped with a heater.
binder above its melting point. They harden in asphalt
The mixing temperature is kept constant dur-
from 65 to 155°C into regularly distributed, micro-
ing the mixing procedure and continues for two
scopic stick-shaped particles (Prowell et al. 2012).
minutes.
Advera is a foaming additives that is a kind of
ii. The preparation procedure for mixtures with
synthetic zeolite, which is composed of alumino-
non-foaming additives is as follows: Initially,
silicates and alkalimetals that contain about 20%
pre-heated aggregates are placed in pre-heated
crystallized water. The water is released by increas-
mixing bowl and pre-weighed additives are
ing the temperature above the boiling point of
directly added and mixed with the aggregate
water, creating a controlled foaming effect. This
batch. In the following, a crater is formed in
leads to an increase in binder volume and a reduc-
the middle of the additive and aggregate mix.
tion in binder viscosity (Prowell et al. 2012).
Rapidly, the binder is added and the bowl is
Two different dosages from both additives are used
moved to the heater equipped mechanical mixer.
in this study to further analyze the additive amount
Similarly, the mixing temperature kept constant
on the mixture properties. The additive dosages are
as 2 minutes.
determined according to the suppliers’ suggestions.
For Sasobit, the typical additive dosage is 1.5% by The mixes are quickly spread on trays before
weight of the binders, whereas the dosage can vary compaction for cooling. The samples are homog-
from 0.8%-4% depending on the application (Prow- enously cooled through mixing with a spatula. Dur-
ell et al. 2012). For this study, two different dosages ing this process, the temperature is continuously
around the typical rate are selected as 2% and 3% by controlled with the help of an infrared temperature
weight of the binder, in order to limit the perform- gun and a temperature prob. These samples were
ance grade (PG) change in the binder. It is known compacted in the compaction molds, heated to
that addition of Sasobit directly affects the PG of the the desired compaction temperature. As an important
binder. Shaw (2007) stated that if the Sasobit dosage remark, the optimum binder content and the mixing
is increased from 1.5% to 2%, the high PG increases temperatures were kept constant through the study.
4–6°F and the low PG increases by 0–3°F.
For Advera, typical dosage is suggested as 0.20–
0.25% by weight of the mix. The minimum and max-
3 DISCUSSION AND RESULTS
imum dosages are determined to be 0.1% and 0.3%,
respectively. Additionally, higher additive dosage is
3.1 Volumetrics
suggested for high binder content mixtures (Prowell
et al. 2012). Therefore, the dosages are determined as The average volumetric properties for each test set
0.25%, the highest of the typical dosage, and 0.3%, are only listed as percent of air voids (AV%), Voids
the highest suggested dosage, by weight of the mix. in the Mineral Aggregate (VMA), and Voids Filled
Four replicates are prepared for each different with Asphalt (VFA) in Table 2. The averages are
mixture type and compaction temperatures. As it only tabulated because the repeatability in sample
will be discussed further in the following parts of preparation is very high. As shown in Figure 2a,
this paper, two of the replicates are used to per- the maximum standard deviation of air voids is
form Marshall Stability and the rest is kept to calculated as 0.20% for each set of mixes.

261
Table 2. Volumetric properties of mixtures. As shown in Figure 2a and 2b, there are clear
trends in between AV% and CT for all the mix-
Mix. Additive CT AV VMA VFA tures, which is impossible to catch if the samples
Type Content (°C) (%) (%) (%) are prepared using Superpave Gyratory Compactor
HMA NONE 140 4.00 17.42 77.04 (Hurley & Prowell 2006b). As the CT decreases, air
125 4.57 17.91 74.49 voids significantly increases, as shown in Figure 2b.
110 5.30 18.54 71.43 Percent of air void reduces by 2% from 95°C to
95 5.69 18.87 69.88 140°C for HMA, whereas this change is less in
Sasobit 2% by weight 140 3.40 16.90 79.90 WMA mixtures in between 1–1.5%. WMA addi-
of binder 125 3.79 17.24 78.04 tives are clearly less susceptible to temperature
110 4.31 17.69 75.62 change. The increase in the Sasobit dosage from 2%
95 4.99 18.27 72.70 to 3% does not significantly affect the compact abil-
Sasobit 3% by weight 140 3.58 17.06 79.02 ity. However, it is clear that 3% Sasobit containing
of binder 125 3.98 17.40 77.13 samples compacted better than 2% Sasobit contain-
110 4.23 17.62 75.98 ing samples at 110°C and 95°C. It can be concluded
95 4.87 18.17 73.18 that the dosage increase improves the compact abil-
Advera 0.25% by weight 140 3.58 17.06 79.01 ity since the binder is softer at lower temperatures.
of mix 125 3.79 17.24 78.02 On the other hand, less additive amount improves
110 3.85 17.29 77.75 the compact ability at higher temperatures. Simi-
95 4.81 18.12 73.45 larly, the increase in the Advera dosage from 0.25%
Advera 0.30% by weight 140 3.63 17.10 78.80 to 0.3% improved the compact ability in a small
of mix 125 3.73 17.18 78.32 amount. According to Figure 2b, the change in
110 4.09 17.49 76.64 AV between 110°C to 125°C is very slight, almost
95 4.51 17.86 74.75 ignorable.
According to MDOT specifications, E1 mix-
tures should met minimum 14% for VMA and
65–78% for VFA criteria (MDOT 2007). The tar-
get 4% air void, VMA and VFA are achieved at
140°C in HMA mixes. Similarly, 4% air void with
± 0.5% accuracy, VMA and VFA are met as low as
110°C in all the WMA mixtures (highlighted with
~5.00 +-----------l;]----~>1-h=---
~
gray in Table 2). According to the volumetrics, the
~ 4. 50 +-- - - - -III· - - - --1 1..-- - -+ J production temperatures for WMA mixtures are
:> 30°C lower than HMA.
~ 4.00 +-rl-r-- - -h t,..-- -+ :1

3.50 3.2 Marshall stability

The stability and flow parameters are not adequate
140 125 110 95
Compaction Temperatures (DC)
to understand the performance of mixtures, though
DHMA 0 Sasobit (2%) they reflect the performance. Thus, the stability
1!1 Sasobit (3%) I!I Advera (0.25%) should be high enough to handle the traffic loads.
~ Ad vera (0.30 %)
In this study, the stability of mixtures are deter-
mined according to AASHTO T 245: Resistance to
Plastic Flow of Bituminous Mixtures Using Mar-
5.50 t--=- ......::::------------ shall Apparatus. According to Asphalt Institute
~.00 ~~------~~~--------­ (1997), necessary stability and allowable range for
flow are minimum 817 kgf (8010 N) and 2–3.5 mm,
~4 50 &--':-r:-"'""'::s::-:-------'- -:::-- - - - respectively for heavy traffic (75 blows).
>
~4 00 +-------''""-;<'=!""--..:-~'--. ...,- - -"""""'• The stability tests are limited with two repli-
cates for each sample set. According to Table 3,
it is proven than the stability of asphalt mix-
3.00 +------,--------...-------, tures decreases as CT decreases. This outcome
95 110 125 140 is expected since AV increases by decreasing CT.
Compaction Temperature (°C)
~ Sasobit(2%) - HMA
Therefore, the stability of HMA mixtures com-
- • - Sasobit (3%) • ··• · • Ad vera (0.25%)
pacted at 140°C is approximately equal to the sta-
-A - Ad vera (0.30%) bility of WMA mixtures compacted at 110°C. In
addition, the flow value is within the specified lim-
Figure 2. Change in air void with respect to CT. its for all the mixes. It can be concluded that WMA

262
Table 3. Marshall stability and flow of mixtures. Resistance to Degradation of Small-Size Coarse
Aggregate by Abrasion and Impact in the Los
CT (°C) Angeles (L.A.) Machine. L.A. abrasion loss is
measured as 17.28%. Although there is no specifi-
Mix. Type 140 125 110 95
cation for L.A. abrasion, most states in USA gives
Stability HMA 1020 1026 906 822 a maximum range of 40–45% (Wu et al. 1998).
(kgf) Sasobit (2%) 1105 1014 949 872 Therefore, the aggregate selected in this study is
Sasobit (3%) 1230 1112 1001 989 relative stiff. The aggregate selection is expected to
Advera (0.25%) 1044 1028 995 796 have no impact in the degradation level in this study.
Advera (0.30%) 1100 1043 976 873 In order to determine the gradation after compac-
Flow HMA 3.40 3.21 3.14 3.66 tion, the mixture samples are burned according to
(mm) Sasobit (2%) 3.15 2.97 3.08 2.95 AASHTO T 308: Determining the Asphalt Binder
Sasobit (3%) 2.77 2.90 2.67 2.80 Content of HMA by the Ignition Oven. For each
Advera (0.25%) 2.96 2.76 3.10 3.14 sample set, two replicates are selected according to
Advera (0.30%) 2.85 3.00 2.73 3.24 their AV. Although the AV change within the sam-
ple sets are in a narrow range, the minimum and
maximum AV containing samples of each set is
chosen to observe the critical gradation change.
performance is as good as HMA. It is identified The effect of CT temperatures on the degra-
that at the same compaction levels, the stability of dation is studies as well as the mixture type. The
the HMA and WMA mixtures are approximately results indicated that although there is differences
same. Hence, at high temperatures if the samples in the percentage change in the retaining aggregate
are compacted at less effort, the mixtures are going for 3/8”, #4, #30, #50, #100, #200 and pan, the
to simulate the same amount of stability. percentage change for all sample sets and different
At higher temperatures, the stability of WMA CT are oddly similar, as shown in Figure 3. As it is
mixtures are higher than HMA. Besides, the expected, there is decrease in the coarse materials
increase in Sasobit dosage form 2% to 3% increase and increase in the fine materials in all the sample
in the stability of mixtures. The increase for all sets. In this study, percentage change as a function
temperatures is about 10% by the increase of 1% in of Total Aggregate (TA) is calculated according to
dosage. However, the increase in the Advera dos- Equation 1 (Airey et al. 2008):
age does not affect the stability significantly, but
still the increase is visible. Percent Change (TA) = (RETbefore − RTEafter)/100 (1)

where RETbefore is the original retained percent-

3.3 Change in gradation during compaction
age for a specific sieve fraction, and RETafter is the
There are various factors affecting the compaction retained percentage for that specific sieve fraction
such as environmental, mix property and construc- after the compaction and the mixture burning in
tion factors (Kassem 2008, US Corps of Engineers the ignition oven. A plus sign in the percentage
2000). One of the construction factors, CT at con- change refers aggregate broke up and less materi-
stant densification effort, is studied in this paper. als retaining on the sieve. On the contrary, a minus
Previous researchers enlightened the follow- sign in the percentage change indicates the increase
ing. Firstly, the degree of degradation is majorly in the retaining material on the sieve.
affected from aggregate gradation. If densely As shown in Figure 3, the change in 1/2”, No.16
graded, less degradation is expected. Secondly, Los and No.8 sieve fraction is less than 0.5% for all
Angles abrasion resistance of aggregates affects the the sample sets at all the CT’s. However, it is clear
degradation. Lastly, among the laboratory compac- that the aggregates retaining on 3/8” sieve fraction
tors, Marshall Hammer causes more degradation is degrading during the compaction. Thus, there is
since it is impact based (Moavenzadeh et al 1963, approximately 2.5% decrease as a function of total
Amirkhanian et al. 1991, Button et al. 1994, Col- aggregates in all samples. As expected, there is clear
lins et al. 1997, Airey et al. 2008). Airey et al. (2008) increase in the retaining #4 sieve fraction. However,
studied the relation between mixture gradation and this increase is approximately 2% as a function of
compaction energy on aggregate degradation in the total aggregates. This means that retaining 3/8”
laboratory. According to that study, it is concluded aggregates degrading into even smaller sizes than
that the increase in the compaction energy effort #4, and most probably there is also a degradation
did not affect the pattern of degradation. However, of aggregates retaining on #4 sieve. There is a simi-
it slightly increased the level of degradation. lar change also in smaller sieve fractions. There is
In this study, the aggregates resistance to the approximately 1% decrease in No. 30 and slightly
degradation is tested according to ASTM C 131: more than 1% decrease in No. 60 sieve fractions as a

263
 "
are distributed in a narrow range. Although it is

I ~
~
Pl'rccntagc Change a~ a fu nction
a weak conclusion, the increase in Sasobit dos-
1~fTo~l A gg~cgatc ("/o) ""'
0
age leads to a more workable and compactable

I ~ ~ ;~ I
0 mixture. Therefore, coarse aggregates are broken
more as the dosage increases. In Sasobit dos-
0

age increase, an approximately 1% difference in

0

degradation is observed from Figure 3b to 3c.

0

Similarly, a parallel conclusion can be driven for

"
"
.,
A

s:
~
0
n

n
00o000060

mixtures prepared with Advera. The increase in

the Advera dosage also improves the workability

" '-
::.
Percentage Change as a function

and compact ability of mixtures. As expected,

~
of Total Aggregate(%)

more degradation is observed in the mixtures with

more additives.
As stated before, the degradation pattern of
o

HMA and WMA mixtures are significantly simi-

!..

lar. Therefore, it is proven that even if WMA mixes

N

are used at slightly high temperatures to gain

•
A

:5:
D
~

~
~
0
~

advantage in the compaction, it is not expected to

affect the mixture gradation.
"
.:;.

'-
-;;;
.~ ~
Percentage Change as a function
~fTo~IA gg~cgate(%) ~
b
0

4 CONCLUSIONS
b

The paper summaries a sequence of laboratory

b

#8

experiment that performed to understand the vari-

b

ations in volumetrics, stability and degradation of

"; · 0
..,;I N_
~ D
~~ D
" "

aggregates on WMA mixtures prepared at high

o

temperatures. Although these additives are intro-

.
0.
o-
~~
~~

duced to the construction industry to lower the

Pe,rccntage Change as a function
~fTo~IAgg~egate~"'·) ""'
0

production temperatures, the end product users

have the tendency to use them at slightly higher
0

temperatures to provide better compaction at the

0

field. Since many agencies entail the construction

0

companies to pay penalties if the target air void

cannot be achieved at the field.
0

The following conclusions are drawn from this

"

study:
~
6

• As the CT temperature increases, the easier the

Percentage C hange as a function

N ~ o - ~ w
6

asphalt mixtures are compacted. Consequently,

ofTotai Aggregate (%)
6

the AV content decreases as CT increases.

0

• WMA mixtures are less susceptible to CT tem-

0

perature changes than HMA mixtures.

6

• The dosage increase in the WMA additives

6

improved the compaction.

~
6

•
Ill

"i
:5:
~
~

~
(')
D

• WMA can be laid and compacted 30 °C lower

0
~
6

than HMA.
Figure 3. Average percentage change of individual • Stability of the mixtures are significantly
aggregate sizes as a function of total aggregate percentage. affected from the CT change. Stability increases
with the increase of CT, since the AV content
decreases.
function of total aggregates. According to that, there
• The degradation pattern for WMA and HMA
is approximately 1% increase as a function of total
mixtures are oddly similar in terms of magnitude.
aggregates in No.100, No.200 and pan materials.
• During the compaction process, there is a
The change in the gradation is clear during
decrease in the coarse aggregates and an increase
the compaction, though it is significantly hard to
in the fine aggregates.
build a quantitative relationship between the CT
and the degradation. As shown from Figure 3a In the further steps of this study, the effect of
to 3e, the percentage changes of total aggregates compaction temperature on the mixture perform-

264
ance is going to be studied, since Marshall Stability Hurley, G.C. & Prowell, B.D. 2006a. Evaluation of
and flow are not sufficient to analyze the rutting potential processes for use in warm mix asphalt. Jour-
and fatigue performance of mixtures. Addition- nal of the Association of Asphalt Paving Technologists
ally, in the future, a similar study can be performed 75: 41–90.
Hurley, G.C. & Prowell, B.D. 2006b. Evaluation of Saso-
on a coarser mixture to analyze the adverse affects bit for use in Warm Mix Asphalt. NCAT Report 05–06,
of compaction. Moreover, this study may be vali- Auburn, AL: National Center for Asphalt Technology.
dated with field trials. Kassem, E.A.R. 2008. Compaction Effects on Uniform-
ity, Moisture Diffusion, and Mechanical Properties of
Asphalt Pavements. Dissertation, Texas A&M Univer-
REFRENCES sity, College Station, TX.
Larsen, O., Moen, Ø., Robertus, C., and Koenders, B.
Airey, G.D., Hunter A.E, Collop, A.C. 2008. The effect 2004. WAM Foam Asphalt Production at Lower
of asphalt mixture gradation and compaction energy Operating Temperatures as an Environmental Friendly
on aggregate gradation. Construction and Building Alternative to HMA. In Proceedings of the 3rd Euras-
Materials 22(5): 972–980. phalt and Eurobitume Conference Book 1: 641–650,
Amirkhanian, S.N., Kaczmarek, D., Burati, JL. 1991. Breukelen, The Neterhlands: Foundation Eurasphalt.
Effect of Los Angeles abrasion test values on the MDOT. 2007. Special Provision for Superpave HMA
strengths of laboratory-prepared marshall specimens. Mixtures. 03SP501(F), Michigan: Michigan Depart-
Transportation Research Record 1301: 77–86. ment of Transportation.
Bahia, H.U., Hanson, D.I., Zeng, M., Zhai, H., Khatri, Moavenzadeh F., Goetz W.H. 1963. Aggregate degrada-
M.A., Anderson, R.M. 2001. Characterization of tion in bituminous mixtures. Highway Research Record
Modified Asphalt Binders in Superpave Mix Design. 24, Washington, DC: National Research Council.
NCHRP Report 405, Washington, D.C: NCHRP. Ozturk, H.I. 2013. Quantification of Quality of Foamed
Button J.W., Little D.N., Jagadam, V., Pendleton, O.J. Warm Mix Asphalt Binders and Mixtures. Phd. Dis-
1994. Correlation of select laboratory compac- sertation, Lansing, MI: Michigan State University.
tion methods with field compaction. Transportation Prowell, B.D., Hurley, G.C., Frank, B. 2012. Warm Mix
Research Record 1454: 193–201. Asphalt: Best Practices, 3rd Edition, National Asphalt
Cervarich, M., “Cooling Down the Mix—New “Warm Pavement Association, Lanham: MD.
Mix Asphalt”, Technologies Developed in Europe” Hot Prowell, B.D., Schreck, R.J., Sasaki, S. 2011. Evaluation
Mix Asphalt Technology, National Asphalt Pavement of compaction benefits of foamed asphalt mixtures at
Association, Lanham, MD, March/April.13–16, 2003. varying production temperatures. www.asphaltpave-
Christensen, D.W. & Bonaquist, R.F. 2006. Volumetric ment.org.
Requirements for Superpave Mix Design, National Prowell, B.D. 2007. Warm Mix Asphalt: The interna-
Cooperative Highway Research Program (NCHRP) tional technology scanning program summary report.
Report 567, Transportation Reseach Board, Washing- U.S. Department of Transportation, Federal Highway
ton: D.C. Administration American Association of State High-
Collins R, Watson D, Johnson A, Wu Y. Effect of aggre- way and Transportation Officials, National Coopera-
gate degradation on specimens compacted by super- tive Highway Research Program.
pave gyratory compactor. Transportation Research Roberts, F.L., Kandhal, P.S., Brown, E.R., Lee, D.Y.,
Record 1590, TRB, National Research Council, Kennedy, T.W. 1996. Hot Mix Asphalt Materials, Mix-
Washington, DC; 1997. p. 1–9. ture Design, and Construction. Lanham, MD: National
De Groot, P., Bowen, C., Koenders, B., Stoker, D., Asphalt Pavement Association Education Foundation.
Larsen, O., Johansen, J. 2001. A Comparison of Emis- Rushing, J,F., Mejias, M., Brown, E.R. 2010. Develop-
sions from Hot Mixture and Warm Asphalt Mixture ment of criteria for using the superpave gyratory com-
Production. IRF World Meeting, Paris: International pactor to design airport pavement mixtures, 2010 FAA
Road Federation. Worldwide Airport Technology Transfer Conference,
Delgadillo, R. & Bahia, H.U. 2008. Effects of Tempera- Atlantic City, NJ.
ture and Pressure on Hot Mixed Asphalt Compac- Shaw, J. 2007. Sasobit—Sasol Wax Americas, 2007
tion. Field and Laboratory Study 20 (6): 440–448. WMA Technical Working Group Meeting, Austin, TX.
EAPA. 2014. The use of Warm Mix Asphalt. EAPA The Asphalt Institute. 1997. Mix Design Methods for
Position Paper, Brussels, Belgium: European Asphalt Asphalt Concrete and Other Hot-Mix Types (MS-2),
Pavement Association. 6th Edition, Lexington, KY: Asphalt Institute.
Hansen, K.R. & Copeland, A. 2013. Annual Asphalt U.S. Army Corps of Engineers. 2000. Hot-Mix Asphalt
Pavement Industry Survey on Recycled Materials and Paving Handbook, Washington, DC: U.S. Army
Warm-Mix Asphalt Usage: 2009–2012. Information Corps of Engineers.
Series 138, Lanham, MD: National Asphalt Pavement Wu, Y., Parker, F., Kandhal, K. 1998. Aggregate Tough-
Association. ness/Abrasion Resistace and Durability/Soundeness
Huner, M.H. & Brown, E.R. 2001. Effects of Re-Heating tests relaed to asphalt concrete performance in pave-
and Compaction Temperature on Hot Mix Asphalt ments. NCAT Report 98-04, Auburn, AL: National
Volumetrics. NCAT Report 01–04, Auburn, AL: Center for Asphalt Technology.
National Center for Asphalt Technology.

265
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Comparison of DSR and BBR tests for determining the Performance

Grade (PG) of asphalt binder at low temperature

C. Riccardi, A. Cannone Falchetto & M.P. Wistuba

Technische Universität Braunschweig, Braunschweig, Germany

M. Losa
University of Pisa, Pisa, Italy

ABSTRACT: The current specification requires the use of two different testing devices for determin-
ing the Performance Grade (PG) of asphalt binder. The Dynamic Shear Rheometer (DSR) is used for
estimating the high PG while the Bending Beam Rheometer (BBR) is the reference method for obtaining
the low PG. This is associated to a number of drawbacks both from a practical testing viewpoint (differ-
ent specimen preparation) and because of the costs required to purchase two devices. Therefore, if the
DSR can be used to characterize asphalt binder at low, intermediate and high temperature, this would be
beneficial for practitioners on a testing routine basis. In this paper, the possibility of using the DSR for
determining the binder low PG, avoiding BBR testing is experimentally investigated. For this purpose, a
plate-plate geometry with a diameter of 4 mm and a gap of 1.75 mm was selected to estimate the material
rheological properties and, hence, the low critical temperature. In particular, the slope and the magnitude
of the shear stress relaxation modulus master curve, determined with the DSR, are correlated with the
corresponding stiffness and m-value from the BBR measurements. The temperature corresponding to the
PG grade is determined and compared with the one obtained in the BBR. The experimental results indi-
cate that the low PG obtained from DSR and BBR are equivalent.

1 INTRODUCTION Farrar et al. 2015). For this purpose, the DSR lim-
iting testing temperature of 5°C was addressed by
According to US standard, two testing devices measuring the instrument compliance and apply-
are currently used to determine the Performance ing appropriate correction to the data. The binder
Grade (PG) of asphalt binder (AASHTO M320 low PG was obtained based on a simple experimen-
2010). The Dynamic Shear Rheometer (DSR) tal correlation between the creep stiffness at 60s,
(AASHTO T315 2012) is used for obtaining the S(60s), measured from BBR tests and the complex
PG at high and intermediate temperatures while modulus determined with the DSR expressed in
Bending Beam Rheometer (BBR) creep tests are the time domain at 60s, G(60s).
performed to estimate the low temperature PG In this paper BBR measurements and DSR tests
(AASHTO T313 2012). at low temperature (starting from –30°C) were
This two-instruments approach was chosen performed on a set of five different asphalt bind-
during the Strategic Highway Research Program ers. Then, a relationship that links the BBR tests
(SHRP) (Anderson & Kennedy 1993) since the results in ethanol with the DSR tests was derived.
DSR measurements at temperature lower than 5°C
are significantly affected by substantial compliance
errors in the absolute values of the dynamic mod- 2 MATERIALS AND TESTS
uli. In order to overcome this problem, the BBR
method was developed. This consists of simple low Five different asphalt binders having the PG
temperature creep tests on small asphalt binder (AASHTO M320 2010) reported in Table 1 were
beams immersed in a conditioning bath of ethanol. used: two 50/70 Pen Grade binders from differ-
More recently, the possibility of using the DSR ent sources, a 70/100 Pen grade binder, a Polymer
(AASHTO T315 2012) with 4 mm parallel-plate Modified binder (PmB) with thermoplastic elas-
geometry for determining the creep stiffness and tomer Styrene Butadiene Styrene (SBS), and an
the low PG of asphalt binder, without relying on extracted RAP binder (AASHTO T49 2014). It is
BBR, was proposed. (Kim 2007, Sui et al. 2011, interesting to observe that in spite of the different

267
Table 1. Performance Grade of the asphalt binders
used.

D Asphalt Binder PG

A 50/70 FA 70–22
B 50/70 64–22
C 70/100 64–22
D PmB 82–22
E RAP binder 82–22

Figure 2. DSR 4 mm geometry.

dLogS (t s)
m(t s) =
s) (2)
dLog(t s)

BBR tests were performed on binder sub-

jected to Rolling Thin-Film Oven Test (RTFOT)
(AASHTO T240 2013) and Pressurized Aging Ves-
sel (PAV) (AASHTO R028 2012) aging. This pro-
cedure was selected for the first four binders (A to
D), while for extracted RAP binder (binder E), the
recondition proposed by the National Cooperative
Highway Research program (NCHRP) 452, which
requires the determination of the low temperature
PG on the RTFO residue, was followed (McDaniel
& Anderson, 2001).
Figure 1. BBR-Pro testing device. All the five asphalt binders were also tested
using a DSR (AASHTO T315 2012) (Fig. 2).
Temperature (from −30°C to +10°C) and fre-
quency sweep tests (from 0.1 to 10 Hz) in stress
binder origin and the different binder classification
controlled mode were conducted in the classical
obtained from the conventional Pen Grade system
parallel-plate configuration using the diameter of 4 mm
(AASHTO T49 2014), the entire set of five binders
with a gap of 1.75 mm (Farrar et al. 2015). For all the
present the same low PG equal to –22.
binders a stress level of 100 kPa was chosen in order to
Three point bending creep tests were performed
remain in the Linear Viscoelastic Range (LVE).
on all asphalt binders using the BBR-Pro device
In order to use the DSR for temperature lower
shown in Figure 1.
than 5°C, the procedure proposed by the Western
A constant load was applied to the middle point
Research Institute (Farrar et al. 2015) was adopted.
of a small binder beam for 240s and its mid span
It is well known that, at very low temperature, the
deflection, δ(t) was recorded for the entire dura-
effect of instrument compliance is significant.
tion of the test. Then, based on the Euler-Bernoulli
Therefore, in order to avoid significant errors in
beam theory, creep stiffness, S(t), and creep compli-
measuring the rheological properties at very low
ance, D(t), were computed according to Equation 1:
temperature the compliance should be taken into
account. A simple procedure was applied to meas-
σ P l3 1
S (t ) = = = (1) ure it. An instantaneous glue was used to glue the
ε (t
() ⋅ b h ⋅ δ (t ) D(t ) plates together and then a stress sweep test at 1 rad/s
was performed. Plotting the angle displacement
where P is the applied constant load equals to versus torque, the instrument compliance can be
980 mN; l is the specimen length (102 mm); b is obtained as the slope of the linear fit of the plot.
the specimen width (12.7 mm); h is the specimen Then, both the storage modulus (G’) and the loss
thickness (6.25 mm); and t is the time. In addi- modulus (G’’) were corrected taking into account
tion, the relaxation parameter m-value was com- the instrument compliance as reported in Equation
puted as: 3 and 4 respectively.

268
 Table 2. BBR S(60s) experimental results.
⎛ J ⎞ J
Gm′ ⎜1 − tool Gm′ ⎟ − tool Gm′′2
⎝ kg ⎠ kg ID Binder T [°C] S(60s) [MPa] CoV [%]
Gs′ =
⎛ Jtool
2
⎞ ⎛ Jtool ⎞
2 (3) A 50/70 FA −12 132 2.7
⎜1 − k Gm′ ⎟ + ⎜ k G′′
Gm ⎟ −18 301 4.9
⎝ g ⎠ ⎝ g ⎠
−24 710 5.2
Gm′′ B 50/70 −12 197 2.2
Gs′′=
⎛ Jtool
2
⎞ ⎛ Jtool ⎞
2 (4) −18 404 4.5
⎜1 − k Gm′ ⎟ + ⎜ k Gm′′ ⎟ −24 721 4.5
⎝ g ⎠ ⎝ g ⎠ C 70/100 −12 128 0.4
−18 287 6.2
where Gs is the sample modulus; Gm is the meas- −24 495 14.1
ured modulus, Jtool is the machine compliance, kg is D PmB −12 181 0.4
the geometry conversion factor expressed by Equa- −18 353 3.5
tion 5. −24 615 4.6
E RAP binder −12 206 3.2
2h −18 423 4.7
kg = (5)
π R4 −24 829 7.9

where h is the gap; and R is the radius of the

plates. Table 3. BBR m(60s) experimental results.
From the measurements of the complex shear
modulus and of the phase angle, the relaxation ID Binder T [°C] m(60s) [MPa] CoV [%]
modulus, G(t), was obtained using the interconver- A 50/70 FA −12 0.359 0.6
sion method by the approximate expression devel-
−18 0.276 1.5
oped by Christensen (1982) reported in Equation 6.
−24 0.212 3.6
B 50/70 −12 0.321 0.2
(t ) ≈ G ′( ) |ω =2 /π t (6)
−18 0.229 0.7
−24 0.203 13.3
Plotting the relaxation modulus, G(t), master C 70/100 −12 0.358 0.8
curve versus time in a bi-logarithmic plane, and −18 0.278 2.7
fitting the curve with a second order polynomial
−24 0.223 10.5
expression (Equation 7) the relaxation modulus cor- D PmB −12 0.325 3.8
responding to 60s at different temperatures can be
−18 0.243 4.1
found.
−24 0.198 11.1
E RAP binder −12 0.338 6.5
LogG (t s ) = −aa( Log (t 60 s ))2 − b og(t = 60 s ) + c
−18 0.254 7.0
(7)
−24 0.181 11.3

Similarly to the BBR (AASHTO T313 2012)

analysis procedure, the corresponding relaxation
parameter, mr-value, that represents the slope of (COV). In Table 2 the measured Stiffness at three
the relaxation modulus at 60 s, can be determined different temperatures are reported, while in
by taking the derivative of the second order poly- Table 3 the m-value can be found.
nomial in Equation 7 as:
3.2 DSR test results
dLogG(t s)
mr (t s) =
s) = − a 2 Logt − b (8) Based the DSR experimental data, master curves
dLog(t s) of the complex modulus and of the phase angle at
different reference temperatures (–12, –18, –24°C)
were first plotted. Then, the relaxation modulus
was calculated using Equation 6 and plotted versus
3 RESULTS
time in a bi-logarithmic plane. The obtained curves
were fitted using a second order polynomial and
3.1 BBR test results
the relaxation modulus at 60 s was determined.
In Table 2 and 3, the BBR tests’ results in ethanol Finally, by taking the derivative of the polynomial
are summarized with their Coefficient of Variation function the mr-value at 60 s was obtained. Table 4

269
Table 4. DSR G(60) and mr-value (60). 1000 ,------------------------------------,
900
ID Binder T [°C] G(60s) [MPa] mr(60) 800
700
600
A 50/70 FA −12 101 −0,298
;_" 500
−18 193 −0,247 "" 400
−24 416 −0,167 g 300
B 50/70 −12 118 −0,260 ~ 200
100
−18 200 −0,217
−24 375 −0,173 100 200 300 400 500 600
C 70/100 −12 802 −0,272 G(60s) [MPa]
• 50170 FA A. 50/70 0 701100 • PmB + RAP binder
−18 135 −0,219
−24 199 −0,167 Figure 3. Correlation between BBR creep stiffness,
D PmB −12 153 −0,222 S(60s), and DSR relaxation modulus, G(60s).
−18 236 −0,191
−24 346 −0,155
0,4 .------------------------------------------,
E RAP binder −12 121 −0,297
.... . . .
y = 1.102x - 0.027 0 . .....
−18 243 −0,231 0,35

−24 550 −0,139

~0,25
0,3
R' = 0.86 •
.•. ... ~

~ 0,2
7
report the calculated G(60) and mr-value (60) at dif- Ia 0, 15
ferent temperatures for the different binders. ~ 1 +------.-------.------.------.----~
0, 1 0, 15 0,2 m, (6 0s) 0,25 0,3 0,35

e 50170 FA &50/70 0 70/ 100 • PmB + RAP binder

4 COMPARISON BETWEEN DSR AND
BBR RESULTS Figure 4. Correlation between m-value measured with
the BBR and mr-value measured with DSR.
Low temperature rheological parameters such
as creep stiffness, S(t), and the m-value, deter-
From this latter plot, the mr-value corresponding
mined with the BBR can be estimated through a
to m-value of 0.300 is equal to 0.250. As shown a
correlation with 4 mm DSR following the proce-
simple linear relationship between BBR and DSR
dure developed by Sui et al. (2011). Based on this
data was obtained. Therefore, using this simple
method, the slope and the magnitude of the shear
relationship it is easy to estimate stiffness, S(t), and
stress relaxation modulus G(t) master curve at
m-value from DSR measurements carried out using
2 hours and at actual low PG grading temperature
the 4 mm plate. In addition, using these correlations
are correlated with the corresponding S(t) and
at 60s, it is possible to find the low PG of the binder,
m-value at 60 seconds and 10°C above the actual
based on DSR test results. In fact, the temperature
low PG grade temperature from BBR tests.
at which G(60s) is 172 MPa or mr-value is 0.250 cor-
In order to reduce the test time and increase
responds to the low critical temperature.
the test temperature, the procedure proposed by
Therefore, based on Figures 3 and 4, the follow-
Sui and co-workers (Sui et al., 2011) was modified
ing simple, linear experimental relationships were
measuring the G(t) slope and the magnitude at
obtained for S(t) and m-value, respectively:
60 seconds at different low temperatures. In Fig-
ure 3 the stiffness S(t) at 60 seconds measured at (t ) = 1.. G (t ) + 2E + 07 (9)
three different temperatures (−12, −18 and −24°C)
for the five binders in ethanol are plotted versus the m(t ) = 1. mr (t ) .027 (10)
relaxation modulus G(t) in the same conditions. These relationships are very useful for deter-
From the plot shown in Figure 3, the value of mining the low temperature properties, performing
G(60s) corresponding to S(60s) = 300 MPa (as pre- DSR tests instead of BBR tests.
scribed by AASHTO M320) is determined to be
equal to 172 MPa. Therefore, this value can be poten-
tially used as limit to indirectly determine the low PG 5 DETERMINATION OF THE LOW PG
of asphalt binder on a stiffness criterion base. GRADE
In Figure 4 the m-value of the BBR in ethanol is
plotted against the slope of G(t), previously name Using the two limits found in the previous section,
mr, at 60 second for the DSR, in absolute value. 172 MPa for the relaxation modulus and 0.250 for

270
Table 5. Comparison between low critical temperature REFERENCES
determined with BBR and DSR.
AASHTO M320 2010. Standard method of test for per-
Actual MP1 Actual MP1 formance graded asphalt binder. American Associa-
Low Low Low Low tion of State Highway and Transportation Officials.
PG PG PG PG Difference AASHTO MP1 1997. Standard Specification for Per-
ID Binder BBR BBR DSR DSR °C formance Graded Asphalt Binder. American Associa-
tion of State Highway and Transportation Officials.
A 50/70 FA −14 −22 −16 −22 −2 AASHTO R028 2012. Standard practice for accelerated
B 50/70 −14 −22 −13 −22 +1 aging of asphalt binder using a pressurized aging ves-
C 70/100 −16 −22 −15 −22 +1 sel (PAV). American Association of State Highway and
D PmB −14 −22 −16 −22 −2 Transportation Officials.
E RAP binder −13 −22 −14 −22 −1 AASHTO T240 2013. Standard method of test for effect
of heat and air on a moving film of asphalt binder
(Rolling Thin-Film Oven Test), American Association
of State Highway and Transportation Officials.
AASHTO T313 2012. Standard method of test for deter-
the mr-value, the low critical temperatures for all the mining the flexural creep stiffness of asphalt binder
different binders were determined and compared using the bending beam rheometer (BBR). Ameri-
to those obtained with BBR measurements. The can Association of State Highway and Transportation
actual PG and the PG grade determined in accord- Officials.
ance to AASHTO-MP1 (MP1 PG) from DSR and AASHTO T315 2012. Standard method of test for deter-
BBR measurements are summarized in Table 5. mining the rheological properties of asphalt binder
As shown, the MP1 PG is always the same of using a dynamic shear rheometer (DSR). American
Association of State Highway and Transportation
the PG obtained through conventional BBR tests.
Officials.
This is confirmed by the very limited difference in AASHTO T49 2014. Standard method of test for pene-
the actual PG temperature which is in the range of tration of bituminous materials. American Association
1 to 2°C. of State Highway and Transportation Officials.
Anderson, D. A. & Kennedy, T., “Development of SHRP
Binder Specification”, Journal of the Association of
6 CONCLUSIONS Asphalt Pavement Technologies, Vol. 62, 1993, pp.
481–507.
In the present work Bending Beam Rheometer tests Christensen, R.M. 1982. Theory of viscoelasticity. An
introduction. New York: Academic Press.
in ethanol and Dynamic Shear Rheometer tests with
Farrar, M., Sui, C., Salmans, S. & Qin, Q. 2015. Deter-
4 mm plate-plate configuration were used to deter- mining the low-temperature rheological properties
mine the low PG of 5 different asphalt binders. Rela- of asphalt binder using a dynamic shear rheometer
tionship linking the stiffness S(60) and the m-value, (DSR). Report FP 08, Laramie: Western Research
determined with the BBR, and the relaxation modu- Institute.
lus G(60) and mr-value, determined with DSR, were Kim, S-S. 2007. Development of an asphalt binder
found. In addition, limits for the relaxation modulus cracking device. NCHRP IDEA-99, Transportation
and for the mr-value, determined with the DSR, corre- Research Board, Washington D.C.
sponding to S(60) = 300 MPa and m-value = 0.30 were Mcdaniel, R. & Anderson, R.M. 2001. Recommended
use of reclaimed asphalt pavement in the superpave
obtained. Based on the experimental analysis per-
mix design method: technician’s manual. National
formed, it can be concluded that the low temperature Cooperative Highway Research Program, NCHRP
PG grade can be determined as the highest tempera- Report 452, 2001.
ture between that obtained for a relaxation modulus Sui, C., Farrar, M.J., Harnsberger, P.M., Tuminello, W.H.
of 172 MPa and that corresponding to mr-value of & Turner, T.F. 2011. New low-temperature perform-
0.25. When comparing the PG grades obtained from ance-grading method using 4-mm parallel plates on
BBR and from DSR result, it can be observed that a dynamic shear rheometer. Transportation Research
this results into equivalent MP1 PG grade while very Record, 2207: 43–48.
limited differences can be detected for the actual PG.

271
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Aging of bituminous binders in asphalt pavements and laboratory tests

X. Lu
Nynas AB, Nynäshamn, Sweden

H. Soenen
Nynas NV, Antwerp, Belgium

O.-V. Laukkanen
University of Massachusetts, Amherst, MA, USA
Aalto University, Aalto, Finland

ABSTRACT: Aging of bituminous binders is one of the key factors affecting the performance and dura-
bility of asphalt pavements. To simulate binder aging in laboratory, a number of methods are available.
In this paper, RTFOT (Rolling Thin-Film Oven Test), PAV (Pressure Aging Vessel) and RCAT (Rotating
Cylinder Aging Test) using different aging times and temperatures were employed to age two straight-run
bitumens and a styrene-butadiene-butadiene (SBS) polymer modified binder. For field aging, a number
of asphalt pavements of different years in service were investigated. The binders (virgin, laboratory aged,
and extracted from asphalt pavements) were evaluated by penetration and softening point tests, rheologi-
cal measurements with a Dynamic Shear Rheometer (DSR), as well as chemical analyses using Fourier
Transform Infrared Spectroscopy (FTIR) and Gel Permeation Chromatography (GPC). It was confirmed
that the rheological changes upon laboratory aging and the formation of chemical functionalities were
strongly temperature dependent. Great differences were found between the unmodified and SBS polymer
modified binders in the rheological response upon aging. For the modified bitumen, different chemi-
cal reactions of the two components (bitumen and polymer) may compensate each other in some ways,
making the binder less age-hardening and more durable. Apparently, the standardized PAV and RCAT
simulate about 10 years of field aging for the unmodified bitumens when used in a dense asphalt surface
layer, but for open graded mixes a longer PAV or RCAT aging time is necessary. However, for polymer
modified bitumen the relationship between laboratory and field ageing when studying both mechanical
and chemical compositional changes is less trivial.

1 INTRODUCTION production and road construction (short-term

aging) and at ambient temperature during the serv-
As one of the key factors affecting the perform- ice life of an asphalt pavement (long-term aging).
ance and durability of asphalt pavements, the The main mechanism of bitumen aging is oxida-
aging behavior of bitumen has been investigated tion, or called oxidative aging. Oxidative aging is
intensively for a long time. Usually, aging makes highly temperature dependent, and also largely
the binder harder and more brittle, thus increasing affected by the chemical nature of the bitumen
risk of pavement failure, such as surface cracking (Tuffour et al 1989, Branthaver et al. 1993, Petersen
and ravelling (Francken 1997, Leech & Nunn 1997, 2009). For certain type of bitumen, high aging
King et al. 2012). In reducing the potential for low resistance has been observed (Soenen et al. 2016).
temperature cracking, less aging of bitumen will In the field, important factor affecting bitumen
be beneficial (Page et al. 1985). It has also been aging also include the air voids content of asphalt
shown that on the long lasting asphalt pavements, mixtures. Much work has indicated that asphalt
the slow aging of the binder keeps asphalt layers mixtures of low voids show a low degree of bitumen
flexible enough to resist cracking (Lu et al. 2011). aging while higher void content facilitates the aging
A low degree of bitumen aging also minimizes the process (Dickinson 1980, Kemp & Predoehl 1981,
effects of moisture, thus retaining a high strength Kemp & Sherman 1984, Leech & Numn 1997,
for the pavement structures (Thomas et al. 2006). Oliver 1992, Lu et al. 2011). It is believed that air
For a Hot Mix Asphalt (HMA) pavement, aging voids content determines the rate of aging by con-
takes place at high temperatures during asphalt trolling oxygen access to the bitumen. Consequently,

273
the oxidative aging of bitumen as function of depth Table 1. Properties of bituminous binders.
in the pavement is closely linked to the air voids
content of the mixture. Regardless of pavement Penetration, Softening Viscosity at
structure, asphalt surface or top thin layer is aged Binders 1/10 mm point, °C 135°C, mm2/s
more than the asphalt layers at lower depths (Coons B1 76 45.5 381
& Wright 1968, Mirza & Witczak 1995). B2 220 35.3 170
To simulate bitumen aging in laboratory, differ- PMB 98 95.0 1700
ent types of test methods may be used, including
conduction of accelerated aging on bituminous
binders, on loose asphalt mixtures, or on com-
pacted asphalt specimens. For bituminous binders, According to the current European binder speci-
there are three European standardized tests for fications, the polymer modified binder is classified
the short-term aging at high temperatures, namely as 90/150-75, while unmodified bitumen B1 and B2
Rolling Thin-Film Oven Test (RTFOT, EN 12607- are classified as 70/100 and 160/220, respectively.
1), Thin Film Oven Test (TFOT, EN 12607-2), and
Rotating Flask Test (RFT, EN 12607-3). These 2.2 Aging test methods
tests reasonably simulate aging particularly during
mixing process in an asphalt mixing plant. As for Aging of the binders was performed by different
the long-term aging during in-service, laboratory methods under different conditions. Those include
simulation is rather difficult. Ideally, such a test RTFOT, followed by PAV at 100°C (PAV 100) for
should be able to predict the chemical and physical 8 h, 20 h, and 48 h; RTFOT, followed by PAV at
property changes of the binder after certain years 75°C (PAV 75) for 48 h, 120 h, and 220 h; and
in asphalt pavement. This may be achieved by con- RCAT at 163°C for 4 h, then at 90°C for 17 h, 65 h,
ducting a test under artificially severe conditions, and 140 h. The use of different temperatures and
for example, at a temperature higher than pave- aging times in PAV is to study aging kinetics, which
ment service temperature and at a pressure higher is believed to be binder specific. RCAT is another
than ambient pressure. Two European standard- standard long-term aging test method. It was also
ized long-term aging tests are Pressure Aging Ves- claimed that RCAT at 163ºC for 4 h (used as the
sel (PAV, EN 14769) and Rotating Cylinder Aging first step of RCAT) can simulate short-term aging
Test (RCAT, EN 15323). Although numerous (Verhasselt 2003).
investigations have been carried out on these test
methods, field data for different types of binders 2.3 Evaluation of aging by conventional and
under different climatic conditions are still needed rheological measurements
to further demonstrate if these test methods are
relevant enough or if natural aging occurred in the Various conventional, rheological and chemical
pavement can be properly predicted. property measurements were carried out to evalu-
The main objectives of this paper are to study ate bitumen aging. Those are softening point,
how different binders behave under different aging complex viscosity by a Dynamic Shear Rheometer
tests and to assess how laboratory aging tests are (DSR), functional groups (sulfoxides and carbon-
related to field aging. Two straight-run and one yls) by Fourier Transform Infrared Spectroscopy
polymer modified bitumens were investigated. (FTIR), and molecular weight distributions by
Characterization of the binders (virgin, labora- Gel Permeation Chromatography (GPC). Detailed
tory aged, and extracted from asphalt pavements) procedures for these tests can be found elsewhere
was performed using conventional test methods, (Lu et al. 2011).
fundamental rheological measurements, as well as In Figure 1, the complex viscosities measured at
chemical analyses. 60ºC and 10 rad/s are plotted versus the aging time
in PAV and RCAT. The starting points at zero hour
are corresponding to those after the short-term
aging, i.e. RTFOT or RCAT at 163ºC. According
2 LABORATORY AGING to literature (Verhasselt 2003), the short-term aging
tests by RTFOT and RCAT at 163°C are quite
2.1 Bituminous binders similar. This is confirmed in this study both for the
Three types of binders were selected for this study, unmodified and SBS polymer modified binders. In
including a penetration-grade bitumen B1, a the long-term aging, as expected, the temperature
viscosity-grade bitumen B2, and a highly polymer is shown as a factor significantly affecting aging
modified bitumen PMB with styrene-butadiene- kinetics, and the temperature effect is strongly
styrene (SBS) block copolymer. Conventional binder-specific. For the unmodified bitumen B1
properties of the binders are shown in Table 1. and B2, under the tested durations, the complex

274
 UJOE+04
v; 80
e PAV 7 5 e RCAT 90
g_ Bl • PAV 100
E
..
.
II Ill
E

.
60
~ 0
"' /
··.:.:·:::::•···-~· · · ·· · ···· · ·· · ·····
...-<
-;::;
"'
-~ LOOE+03
40
f.~::::::e::·····
c:
0 0

>
~

1:i
Ci.
E
0
-~

~
c:
"'
c..
20

u II Bl B2 PMB
u 1.00[+02 • PAV 100"C, 20h • PAV 75"C, 120h • RCAT 90"C, l~Oh
0 50 100 150 200 250
Lab Aging Time (h) 100
l.OOE+03 RO
........-~

I I Ill
60
~:· ..... ................. ::·-• ···············• 40
"'
-~ 1.00E+02 20
.12:! 0
> B2 e PAVloo • PAV75 e RACT90
~ Bl B2 PMB
Ci.
E
• PAV lOO"C, 20h • PAV 7~"C, 120h • fKAT 90"C, 140h
8 1 .00F+01
0 100 200 300
Lab Aging Time (h)
Figure 2. Comparison of the long-term aging methods
-;;;- l.OOE+04 by penetration and softening point.
..;
-'u=- PM B e PAV 100 e PAV75 e RCAT 90

'b
w 1000
ro e Bl, PAV 100 e 81, PAV 75
·f l.OOE+03 ~.-.:;:·:·:·;,:,.:eo,,,,, ......-.• :::·····---·-··········• _. B2, PAV 100_. 82, PAV 75
+ PMB, PAV 100 + PMB, PAV 75
>
J1c._
eE
E
0
"'....
.....
u 1.00E 102 d
c 100
0 so 100 1 50 200 250 0

Lab Aging Time (h) -~

c
...
~

Figure 1. Evolution of complex viscosity during PAV

and RCAT under different conditions.
10
20 40 60 80 100
viscosity gradually increases with the aging time, Softening Po int R&B (oc)
both in PAV and RCAT. The general rule that
increasing temperature by 10°C doubles the rate of Figure 3. Penetration vs softening point after different
aging is also likely followed in the PAV. For exam- long-term aging tests.
ple, to reach the same level of complex viscosity of
1000 Pa.s at 60ºC, for B1, aging time in the PAV at
100ºC is estimated to be 12 h, which is about 1/8 of the complex viscosity of the modified binder
the time (90 h) required in PAV at 75ºC. However, increases gradually with the aging time. However,
the rule is not valid when PAV and RCAT are com- when the temperature is raised from 75ºC to 100°C
pared to each other. To obtain a similar level of in PAV or 90°C in RCAT, such trends in the viscos-
aging, RCAT will take much longer time than PAV ity change are not seen. The different behaviors are
if the temperature is the same. The DSR data, as also reflected in softening points (Figure 3). Unlike
well as softening point and penetration (Figure 2), the unmodified bitumens, the softening point of
show that for the unmodified bitumens, aging of the modified binder does not increase with the
140 h in RCAT at 90°C is more or less equivalent to aging time. This is attributed to a combined effect
20h PAV at 100°C, or 120 h PAV at 75°C. of bitumen oxidation and polymer degradation,
On the other hand, for the PMB, aging kinetics and to changes in the compatibility or polymer
is very different from those of unmodified bitu- networks of the system. The degradation of the
mens. As shown in Figure 1, using PAV at 75°C, polymer apparently compensates for bitumen

275
 8.0 aging are those after RTFOT. As indicated, at both
e ~l e ~:l e PM~
low (75°C) and high (100°C) aging temperatures,
81 :1< 82 • PMB the formation of sulfoxides is significant. With
6.0
0 increasing aging time, the rate of sulfoxides for-
-~
mation tends to decrease. Unlike sulfoxides, car-
0II 4.0 PAVlOO bonyl compounds are formed at a more constant
!::!. rate, which is much higher at 100ºC than at 75ºC.
.~:; ;::::::;:::::·::·::·:·:·:·:·:·::·:·:·:·:·;:·:·:·:·:·:·:·::·:·:·:·~; :.;:·:.;:;;:.:.:i AV 75
~
..c The observations agree well with those findings
..:
:l.U reported in the literature (Branthaver et al. 1993).
Figure 4 also indicates that the formation of
sulfoxides in the modified binder is lower than that
n.o
in the unmodified ones, while an opposite trend is
20.0 seen for carbonyl compounds. Notice that the base
--e-· 81 --e-- 62 --e-· PMB bitumen used for the modified binder is very simi-
- - Bl ......,_ 62 - PMB lar to B2. The inhibiting effect of the polymer on
15.0
0
sulfoxide formation has also been observed earlier
·~
!!! (Lu & Isacsson 2000). It is suspected that the SBS
?," 100 polymers may compete with sulfur compounds
Vl in the bitumen for oxidants. As for carbonyls, the
~
..c higher increase is contributed by oxidation of the
<(
J.O polymer (or polymer degradation).
For a given bitumen, correlation may be expected
between rheological and compositional changes,
0.0 and this was the case for the three binders inves-
0 50 100 150 200 250 tigated in this paper. The correlation coefficients
Lab Aging 1 ime {h) were found between 0.68 and 0.91 for the complex
viscosity versus carbonyls, and between 0.61 and
Figure 4. FTIR analysis of bitumen aging—effect of 0.96 for the complex viscosity versus sulfoxides.
temperature. However, such correlations did not exist when dif-
ferent binders were examined together because of
differences in the temperature dependence of oxi-
oxidative hardening. As a consequence, the equiva- dation and other contributing factors.
lences between the long-term aging tests observed
for the unmodified bitumens are not valid for the
3 FIELD AGING IN ASPHALT
PMB (Figure 2). The strong temperature depend-
PAVEMENTS
ence of aging mechanisms and kinetics also means
that it will be difficult to predict PMB aging prop-
3.1 Road sections
erly in the field by a laboratory simulation at a high
temperature. Three pavement sections were selected for field
aging study (Table 2). They are coded as B1-E6,
B2-RV53 and PMB-E4, representing the differ-
2.4 Evaluation of aging by FTIR
ent binders (B1, B2 and PMB) used. The section
Chemical changes during aging have been studied B1-E6 was selected from a test road on the high-
extensively in the past years. It is well-known that way E6, Geddekippel—Kalsås, in Sweden. It was
oxidation of bitumen produces carbonyls (C = O) a reference section of the test road constructed in
and sulfoxides (S = O) and increases aromaticity, 2006, where B1 was used in a 40 mm stone mastic
causing increases in bitumen viscosity and elas- asphalt ABS16 (SMA16) as the wearing course.
ticity. The chemical changes may differ largely Asphalt cores were taken from the road section
between different bituminous binders, especially after nine years of service in October 2015.
between polymer modified and unmodified. From For bitumen B2, a pavement section was selected
infrared (IR) spectrograms, absorbance bands from the national road RV53 based on the Swedish
(peak areas) at about 1705 and 1030 cm–1 were LTPP (Long Term Pavement Performance) data-
measured for carbonyl compounds (e.g. ketones, base. The section (B2-RV53) was located in Kvick-
carboxylic acids and anhydrides) and sulfoxides, sund, and was built in 1977 with a 25 mm MAB16T
respectively. (AC16). According to the technical requirements
In Figure 4, the IR absorbance ratios of aged of that time, MAB16T was a dense graded asphalt
to unaged bitumen are plotted against the aging mixture with maximum aggregate size of 16 mm
time in PAV. Again, the samples of zero-hour and with about 6.2% (by weight) binder and 3–5%

276
Table 2. List of pavement sections. Table 4. IR absorbance ratio (field aged/unaged).

Road Mix Binder Air Years in the field Recovered binders Carbonyls Sulfoxides
sections types content,% voids,% (and as top layer)
B1-E6 0.56 3.32
B1-E6 SMA16 6.0 3.0 9 (9) B2-RV53 1.57 26.3
B2-RV53 AC16 6.2 3.5 26 (5) PMB-E4 1.27 2.73
PMB-E4 SMA11 6.5 3.5 15 (15)

Table 3. Conventional and viscosity measurements of

Results are shown in Table 4. Unexpectedly, for
the field aged binders. B1-E6, the IR absorbance ratio of carbonyls is
less than 1. This might indicate that carbonyl com-
Complex pounds probably were not completely extracted
Recovered Penetration, Softening viscosity at from the asphalt material due to e g. certain active
binders 1/10 mm point, °C 60(C, Pa.s aggregates.
B1-E6 NM* 56.4 1091
B2-RV53 101 44.0 114 4 FIELD AGING VERSUS LABORATORY
PMB-E4 54 81.0 686 AGING TESTS
*Not measured due to insufficient amount of sample.
To compare with field aging, laboratory aging tests
with PAV at 100°C and RCAT at 90°C are used.
air voids. The layer was on the top of the pavement Figure 5 shows the viscosity changes of the dif-
for about five years, and a surface treatment was ferent binders during PAV and RCAT, as well as
then carried out in 1982. In total the asphalt had the viscosities of the field aged samples. As can be
been in the field for 26 years when samples were seen from Figure 5, the standardized PAV condi-
taken. tions (100°C and 20 h) predicts about 10-year field
The field object PMB-E4 was a bridge deck aging of the unmodified bitumen (B1) in the sur-
pavement on the High Coast Bridge on the high- face layer. However, for the binder extracted from
way E4. The bridge was constructed between 1993 RV53, very low aging is observed, and to produce
and 1997, and the surface layer was 35 mm ABS11 the same level of aging as in the field, only about
(SMA11) produced with the studied PMB. The 8 hours are required for the PAV. It should be
asphalt mixture contained about 6.5% binder (by noticed that this binder had been in the pavement
weight) and 3.5% air voids. After nearly 15 years for more than 25 years, but only 5 years as top
of service, asphalt samples were collected. For this layer. The results imply that oxidation of bitumen
PMB, field data from literature were also used for may be slowed down when the asphalt layer was
laboratory aging comparison. overlaid, due to a limited access of oxygen (and
For binder extraction and recovery, the Euro- UV) and lower temperature as compared to that in
pean standards EN 12697-1 and EN 12697-3 were the top layer. The asphalt mixture itself also had a
followed. The solvent used was dichloromethane. low air voids content (1.9%) when field cores were
taken. Taking these into account, for this binder,
PAV seems also give a quite good prediction of
3.2 Conventional and rheological tests
field aging (around 10 years).
The results of penetration and softening point For polymer modified binders, the prediction of
tests for the field aged binders are summarized in field aging by PAV generally becomes difficult due
Table 3. The binders were also characterized using to more complicated chemical reactions. As shown
DSR at different temperatures and frequencies. in Figure 5, a very low degree of aging is observed
For simplicity, only the complex viscosities meas- for the PMB studied even after a long time (15
ured at 60°C and 10 rad/s are shown in Table 3. years) in the bridge deck pavement. The equivalent
These data will be used for comparing with labora- aging time in PAV 100°C is estimated to about 9 h,
tory aging tests later. which is considerably lower than the standardized
aging time of 20 h. The smaller viscosity change
or less age-hardening of the field aged PMB could
3.3 FTIR analysis
be due to more degraded polymers that act as a
The chemical compositions (functional groups) kind of softening agents, which then compensate
of the field aged binders were analyzed by FTIR. for the oxidative hardening of the bitumen. This is
By comparing with un aged samples, absorbance illustrated by the GPC analysis of the laboratory
ratios were calculated for carbonyls and sulfoxides. and field aged samples in Figure 6.

277
 l.OOE I 04 100
- -Bl - - B2 - - PMB
Vi' 90
ni
~ ~ 80
!;' c:
0 1.00E+03 0
<J) ·.;::;
70
1;i ~
>
.t: ~
VI c: 60
0
.Q:!
1.00E+02
"'
'C
"-
> 50
a')
a.
B2-RV53 "'c:
·c;;
E ~ 40
0 c:: PA16
u
30 (15% air voids, 10 years) +
l.OOE+Ol
0 10 20 30 40 :;o 20
Aging Time in PAV at 1oo• c {h) 0 10 20 30 40 50

Aging Time in PAV at 1oo•c

1.00E+04
- -Bl - - B2 - -PMB
Vi' Figure 7. Retained penetration of PMB—comparison
..;
C>. of laboratory aging and field aging.
u
B1-E6
0 1.00E+03
"'10
for PMB-E6; the last two are considerably shorter
.;> than 140 h.
8
~ 1.00E+02 To further examine the relevance of PAV for the
..
X

a. B2-RV53 modified binders, the same PMB used in other mix

E types and in other road sections are compared. In
0
u this case, the binder is assessed by age-hardening
l.OOE+Ol through penetration measurements. In Figure 7,
0 50 100 150 the percent retained penetrations, i.e. (penetration
Aging Time in RCAT at go•c (h) of aged binder/penetration of unaged binder) ×
100, are calculated partly based on literature data
Figure 5. Comparison of laboratory aging and field (Jacobson & Horn wall, 2000). It is evident that
aging by the complex viscosity. the prediction of the PMB field aging is strongly
dependent on air voids content of the asphalt
mixtures. For the Porous Asphalt (PA) of 15% air
voids, the age-hardening of 10-year in the field is
not predictable by the PAV. On the other hand, in
the SMA containing 3% or 5% air voids, the extent
"'c:
V> of binder age-hardening after 8 to 15 years in serv-
0 ice is rather low, for which a corresponding aging
5';
time in PAV will be much shorter than that being
"'
c::
standardized.
0
t)
"'
~
Cl
5 DISCUSSION AND CONCLUSIONS

During the production and service life of an

asphalt pavement, the rheological changes of bitu-
10 12 14 16
minous binders are mainly due to oxidative aging.
Retention Time (min)
As already mentioned, oxidative aging has been
studied extensively for a long time by monitoring
Figure 6. GPC chromatograms of the laboratory aged changes both in the chemistry and in the mechani-
PMB and that extracted from the pavement section. cal properties of bitumen. It is well-known that
the oxidation of bitumen increases its viscosity
and elasticity, and this is normally attributed to
Similar observations can also be made with the formation of polar, oxygen-containing chemi-
RCAT. For these field aged samples, the equivalent cal functionalities and/or aromaticity on bitumen
aging times with this test method are estimated to molecules, resulting in stronger molecular interac-
be 107 h for B1-E6, 37 h for B2-RV53, and 56 h tions (Petersen 2009). The importance of chemical

278
functionality to bitumen rheological properties was the standardized PAV and RCAT simulate about
also very much emphasized in the American SHRP 10 years of field aging for the unmodified straight-
investigations. The present study confirms that, run bitumens when used in a dense asphalt surface
for a given bitumen and during aging, correlation layer. However, for polymer modified bitumen,
exists between changes in viscosity and in chemical the relationship between laboratory and field age-
functionalities. However, such relationship has not ing when studying both mechanical and chemi-
been seen when different binders were examined cal compositional changes is less trivial. It should
together, indicating other chemical parameters also also be noticed that, for more open asphalt mixes,
contributing to viscosity. longer PAV or RCAT aging times will be necessary
Regarding chemical reactions of bitumen upon in order to predict the service life time.
aging, Petersen and Harnsberger (1998) proposed
a two-step mechanism; in a first rapid step oxygen
reacts with a limited amount of highly reactive REFERENCES
hydrocarbons, most likely polycyclic hydroaromat-
ics which form hydro peroxides and which in turn Branthaver, J.F., Petersen, J.C., Robertson, R.E., Duvall,
react with sulfides to form sulfoxides and aroma- J.J., Kim, S.S., Harnsberger, P.M., Mill, T., Ensley,
tized rings. During this first reaction, a much slower E.K., Barbour, F.A. & Schabron, J.F. 1993. Binder
oxidation of benzylic carbons involving formation characterization and evaluation—Vol. 2: Chemistry.
SHRP-A−368.
of benzyl free radicals is initiated, which produces Coons, R.F. & Wright, P.H. 1968. An investigation of
both ketones and sulfoxides. This dual oxidation the hardening of asphalt recovered from pavements of
mechanism explains well the viscosity changes dur- various ages. AAPT 37: 510–528.
ing the aging, which is normally a larger increase Dickinson, E.J. 1980. The hardening of Middle East
during RTFOT followed by a slower increase dur- petroleum asphalts in pavement surfacings. AAPT 49:
ing PAV. The magnitude and duration of the dif- 30–57.
ferent steps in the viscosity change are bitumen Francken, L. 1997. Long term ageing of pure and modi-
specific, and these characteristics are believed to fied bitumen: Influence on the rheological properties
be critical to asphalt durability in the field. Con- and relation with the mechanical performance of
asphalt mixtures. Proc. 8th International Conference
sidering polymer modified binders, aging-induced on Asphalt Pavements, August 10–14, Seattle, USA.
mechanical property changes will be complicated Jacobson, T. & Hornwall, F. 2000. VTI Notat 53-2000
also by the chemical changes of the polymer. (in Swedish).
It can be speculated that chemical reactions Kemp, G.R. & Predoehl, N.H. 1981. A comparison of
during asphalt production and in-service are not field and laboratory environments on asphalt durabil-
necessarily the same as those occurring during ity. AAPT 50: 492–537.
laboratory aging tests, because of differences in Kemp, G.R. & Sherman, G.B. 1984. Significant studies
temperature and in reaction environments (e.g. on asphalt durability: California experience. Trans-
presence/absence of aggregates, fillers, water, and portation Research Record 999: 36–41.
King, G, Anderson, M, Hanson, D & Blankenship, P.
UV etc.). The oxidative aging of bitumen in the 2012. Using Black space diagrams to predict age-
field is also affected by asphalt air voids and layer induced cracking. Proceedings of 7th RILEM Int
thickness and position. And as already mentioned, Conf on Cracking in Pavements 453–463.
the surface or top thin layer of asphalt pavement is Leech, D. & Nunn, M.E. 1997. Deterioration mecha-
aged more than that at lower depths. All these fac- nisms in flexible roads. Proceedings of 2nd European
tors make it rather difficult to predict field aging by Symposium on Performance and Durability of Bitu-
a laboratory method, especially for polymer modi- minous Materials, Leeds, April 1997.
fied binders. In spite of this, selection of an aging- Lu, X. & Isacsson, U. 2000. Artificial aging of polymer
resistant binder based on these laboratory tests is modified bitumens. Journal of Applied Polymer Sci-
ence 76: 1811–1824.
still of great importance to ensure long-term dura- Lu, X. Redelius, P., Soenen, H. & Thau, M. 2011. Mate-
bility of asphalt pavements. rial characteristics of long lasting asphalt pavements.
From the results obtained in this study, the fol- Road Materials and Pavement Design 12(3): 567–585.
lowing conclusions can be drawn. The rheological Mirza M.W. & Witczak, M.W. 1995. Development of a
changes upon laboratory aging and the forma- global aging system for short and long term aging of
tion of functionalities are strongly temperature asphalt cements. AAPT 64: 393–418.
dependent. There are great differences between Oliver, J.W.H. 1992. A long life asphalt mix for lightly
unmodified and SBS polymer modified binders in trafficked streets: results after 10 years. Australian
the rheological response upon aging. For modified Research Board, ARR 228, July 1992.
Page, G.C., Murphy, K.H., Ruth, B.E. & Roque, R. 1985.
bitumen, different chemical reactions of the two Asphalt binder hardening—Causes and effects. AAPT
components (bitumen and polymer) may compen- 54:140–167.
sate each other in some ways, making the binder Petersen, J.C. 2009. A Review of the fundamentals
less age-hardening and more durable. Apparently, of asphalt oxidation—Chemical, physicochemical,

279
 physical properties and durability relationships. Trans- Thomas, K., McKay, J.F. & Branthaver, J.F. 2006. Sur-
portation Research Circular E-C140, TRB: 1–36. factants in aged asphalt and impact on moisture
Petersen, J.C. & Harnsberger, P.M. 1998. Asphalt aging— susceptibility of laboratory-prepared mixes. Road
Dual oxidation mechanism and its interrelationships Materials and Pavement Design 7(4):477–490.
with asphalt composition and oxidative age harden- Tuffour, Y.A., Ishai, I. & Craus, J. 1989. Relating asphalt
ing. Transportation Research Record 1638: 47–55. aging and durability to its compositional changes.
Soenen, H., Lu, X., Muehlich, U. & Laukkanen O.-V. AAPT 58: 163–181.
2016. The influence of aging on binder fatigue and Verhasselt, A. 2003. Short- and long term ageing of bitu-
other fracture related binders test results. ISAP Sym- minous binders—Simulation with the RCAT method.
posium 2016, Jackson Hole, USA. 6th RILEM Symposium PTEBM’03, Zurich, 2003.

280
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Effects of mineral fillers and bitumen on ageing of asphalt

mastics properties

R. Alfaqawi, G.D. Airey & J.R.A. Grenfell

Nottingham Transportation Engineering Centre, University of Nottingham, Nottingham, UK

ABSTRACT: Bituminous binder is always in contact with mineral fillers in asphalt mixtures, thus the
mechanisms of oxidative aging of binders are influenced by the physical and chemical interaction between
fillers and the bitumen. This study attempts to investigate into the factors affect the ageing process of
binders. The effect of granite fillers and hydrated lime with two different bitumen binders on mastic
ageing were studied. The Thin Film Oven Test (TFOT) was used for short-term aging and the Pressure
Aging Vessel (PAV) used for long-term aging. Different ageing indices were used to measure the changes in
physical and rheological properties such as changes in the viscosity, softening point, and complex modu-
lus (G*) of mastics. The chemical changes of the binders were investigated by the means of Fourier
Transform Infrared Spectroscopy (FTIR). Results indicate that hydrated lime slows down bitumen ageing
more than granite fillers. This effect depends mainly on the bitumen type and filler concentration. Also it
showed that hydrated lime interacts differently with the two binders used in this study in terms of stiffen-
ing and ageing effects. In addition results show that different ageing indices were correlated well to each
other and also with the chemical changes measured by FTIR.

1 INTRODUCTION Therefore, the mechanisms of asphalt mastic

ageing are influenced by the characteristics of the
Ageing of bitumen has been known as one of the asphalt binder and the mineral filler properties,
main reasons for the deterioration of asphalt pave- in addition to the molecular interaction between
ments over time. The ageing process can cause stiff- them. The chemical and rheological changes asso-
ening and hardening of asphalt mixture, which may ciated with ageing are well understood for neat
result in increased cracking and reduced pavement bitumen as the result of much research on this sub-
life (Petersen 2009). Among the different mecha- ject (Read and Whiteoak 2003). However, even if
nisms causing bitumen ageing and hardening, the bitumen ageing has been studied for long time, the
most important are oxidation and loss of volatile exact chemical mechanisms that control this proc-
components, which lead to chemical changes in the ess remain debated (Petersen 2009), (Petersen and
binder. The other mechanism is a reversible process Glaser 2011), (Greenfield, Byrne et al. 2015). In
called physical hardening that can be attributed to addition, recent works on the impact of additives,
molecular restructuring. Generally, bitumen age- polymers and nanomaterials maintain bitumen age-
ing takes place in two stages, short-term ageing at ing as an exciting research field (Lesueur, Teixeira
high temperatures during asphalt mixing and pav- et al. 2016). However, for the ageing of asphalt mas-
ing, and long-term ageing at ambient temperatures tics, many topics in this research area still need to be
during in-service life. (Moraes and Bahia 2015), investigated (Moraes and Bahia 2015).
showed that, even though binder ageing in pave- (Little and Petersen 2005) investigated the effect
ment always occurs when the binder is in contact of using hydrated lime to improve the properties
with aggregates and mineral filler, in most ageing of asphalt pavements. They compared the effect of
studies, asphalt binders are individually aged with- hydrated lime to a similar size filler comprised of
out taking into account the effect of mineral fillers. limestone. They performed extensive laboratory test-
(Anderson, Bahia et al. 1992) stated that the prop- ing, on mastics and mixtures and the results showed
erties of the fine fraction dominate in terms of that when hydrated lime is added to bitumen, its
physicochemical interactions between the bitumen effect will be equal to other mineral fillers or be con-
and the mineral surface, as the fines are embedded siderably greater. This effect depends mainly on the
in the bitumen and, thus, the majority of the sur- physicochemical interaction developed between the
face area is generated by the fine minerals. bitumen and hydrated lime particles.

281
 In order to develop a better understanding of mastics, the main properties of the fillers were
the way by which the mineral fillers affect bitu- characterized using the standard tests as follows:
men ageing, an ageing study on the bitumen-filler
1. Particle size distributions were assessed by a
mastics involving DSR and FTIR testing after
grading test. A Beckman Particle Size Ana-
different ageing stages was carried out. Mastic
lyzer (LS200) was used to measure the parti-
ageing has its advantages compared to asphalt
cle size of Granite filler. In this device, a flux
ageing. As mentioned above, the surface area of
curve is generated which is a composite of
fillers is much larger than the coarse aggregate,
the flux curves for all the filler particles that
which will make the adsorbing and catalysing
passes through a laser beam during a sample
phenomenon more obvious. In addition, as the
run. The LS 200 systems convert the compos-
fillers in mastics can be viewed as being embed-
ite flux curve into a particle size distribution.
ded in the binder, the effects of ‘void content’ and
This device uses water as liquid dispersant,
‘binder film thickness’ which were concerns dur-
so it was not suitable for testing the Hydrated
ing the mixture ageing studies, could be avoided
lime fillers as there is a reaction between
in mastic ageing.
water and the hydrated lime Ca(OH)2. So,
the particle size of hydrated lime fillers was
tested by the supplier as using a Malvern
2 MATERIALS AND EXPERIMENTS
Mastersizer.
2. Particle density, the particle density (SG) was
In this study mastics were produced using two
measured by the pycnometer method in accord-
types of bitumen and two different fillers. The
ance to BS EN 1097-7:2008.
binders used in this study was a 40/60 penetration
3. Specific Surface Area (SSA), the BET method is
grade bitumen with initial penetration of 45 dmm
used for the calculation of SSA.
and softening point of 52°C, designated B1. The
4. Voids in dry compacted filler-Rigden Voids
other bitumen is a soft binder, a 100/150 penetra-
(RV) were determined with the Rigden equip-
tion grade bitumen with penetration of 119 dmm
ment in accordance with BS EN 1097-4:2008.
and softening of 44.8°C, designated B2. The two
5. Particle shape was determined by Scanning
fillers used in this study were Granite filler (G) and
Electron Microscopy (SEM) imaging.
Hydrated Lime (HL). All fillers used in this study
were substantially finer than 63 μm in particle size. Table 1 shows the results of the characteriza-
Mastic samples have been made by mixing the tion properties of the used fillers in this study. The
Granite filler with bitumen B1 and B2 to produce SEM microscopic morphology images of the filler
mastics of 50% filler by mass. Then the Granite particles are presented in Figure 1.
was replaced by hydrated lime at percentages of
10% and 20% by mass to produce a total of six
2.2 Mastic testing
different mastics. Thus the mastics produced
with the (40/60) bitumen B1, were labelled as 2.2.1 Mastic ageing procedures
B1+50G, B1+40G+10HL and B1+30G+20HL for The Thin Film Oven Test (TFOT) was used to
mastics with only 50% granite by mass, 10% HL simulate the effect of short-term ageing on mas-
replacement and 20% HL replacement respec- tics. The test was carried out according to BS-EN
tively. Similarly, the mastics produced with the 12607-2:2000. In this test, mastic samples with the
(100/150) bitumen B2 were labelled as B2+50G, same volumes are placed in a TFOT pan to form a
B2+40G+10HL and B2+30G+20HL. thin layer of the same thickness, which is then held
All mastic mixing was done manually. The mix- in the TFOT oven at 163ºC for 5 hours.
ing temperature was 150°C. The bitumen was The Pressure Ageing Vessel (PAV) ageing
stirred by hand and the accurate mass of filler was method is used to simulate long-term field oxida-
added slowly while the stirring was continued. The tive ageing of mastics in accordance with BS-EN
mastic was continuously stirred as it was cooling
down until the temperature became low enough to
prevent the filler from settling. The mastics were Table 1. Filler properties.
poured into small vessels to be saved for later test-
ing. Before each test the mastics were heated to be Rigden Surface Grading
liquid and stirred to ensure homogeneity. Specific voids area
Filler type gravity (%) (m2/g) D10 D50 D90

2.1 Filler tests and properties Granite 2.66 47.15 1.26 1.075 14.58 55.10
Hydrated 2.22 61.62 2.24 1.509 7.137 20.27
In order to investigate the effect of the morphology Lime (HL)
and physical properties of the fillers on bitumen

282
Figure 1. SEM images of fillers a) Granite, b) Hydrated lime.

14769:2012. The mastics in this study are aged for 2.2.3 Chemical analysis of bitumen and mastics
20 hours at 90°C in the PAV after the short-term Fourier transform infrared FTIR spectroscopy is
ageing in the TFOT oven. the most powerful means of detecting and iden-
In addition to the standard ageing procedures, tifying chemical bonds (functional groups) in
an ageing time study using the TFOT oven was either organic or inorganic materials. The asphalt
carried out on the mastics. The mastics were aged binder’s main functional groups can be quantified
in TFOT oven for different ageing times 2,5,10 by using the FTIR technique. Therefore FTIR is
and 20 hours. The behaviour of mastics then were widely used to characterise the oxygen-containing
measured at different ageing stages by means of functionalities found in bitumen, particularly
ageing indices as will be discussed in the following those produced during oxidation (Wu 2009). Typi-
section. cal wave numbers for bitumen ageing products are
sulfoxides S = O at 1030 cm−1 and carbonyls C = O
2.2.2 Ageing index calculation at 1700 cm−1. The FTIR test was performed on the
In order to measure the effects of ageing on the recovered bitumen from mastics before and after
mechanical and rheological properties of different ageing. Carbonyl C = O ageing index was used to
mastics, an ageing index was calculated after each evaluate the effect of ageing on different mastics
stage of ageing. The Ageing index was calculated and it was calculated as follows:
using different tests performed on the mastics: sof-
tening point, rotational viscosity, and the dynamic Carbonyl index (C = O): A1700/ΣA
shear complex modulus G*.
The softening point was determined according The sum of the area ΣA represents: A1700 +
to BS EN 1427:2007 and the rotational viscos- A1600 + A1460 + A1376 + A1030 + A864 + A814 + A743 +
ity was performed using a Brookfield device at a A724 + A(2953, 2923, 2862).
temperature of 135°C. The rheological properties
of the mastics were measured in this study with a
Bohlin Gemini model DSR. A frequency sweep 3 RESULT AND DISCUSSION
is performed with a 0.3% strain amplitude over a
range of 0.1 to 10 Hz and temperatures from 0 to 3.1 Ageing index after short and long term ageing
50°C. The resulting complex shear modulus |G*|
at 20°C and 0.4 Hz, were selected to calculate the The ageing index after the TFOT short-term age-
ageing index from the rheological changes due ing and the long-term ageing after the PAV ageing
to ageing. The aging index was calculated using procedures where calculated. The ageing index was
equation 1 calculated from the results of viscosity, softening
point, and the complex modulus of the mastics
Material
i Propertiesaged mastic and is presented in Table 2.
I x=
Aging Inde (1) It can be noticed from Table 2, that the mastics
Material
i Propertiesunage
n d mastic containing hydrated lime are initially stiffer than

283
Table 2. Properties of unaged, TFOT and PAV aged mastics.

Viscosity G* S.P Viscosity G* S.P

Mastic ID (Pa s) (Pa) (°C) AI AI AI

B1 + 50G 2090 5.03 × 106 60.6

B1 + 50G TFOT 3506 1.14 × 107 66.8 1.68 2.27 6.2
B1 + 50G PAV 8314 2.29 × 107 77.4 3.98 4.56 16.8
B1 + 40G + 10HL 2970 6.63 × 106 62.3
B1 + 40G + 10HL TFOT 4739 1.38 × 107 70.2 1.60 2.08 7.9
B1 + 40G + 10HL PAV 6803 2.76 × 107 76 2.29 4.17 13.7
B1 + 30G + 20HL 3465 6.76 × 106 64
B1 + 30G + 20HL TFOT 6143 1.41 × 107 70.9 1.77 2.08 6.9
B1 + 30G + 20HL PAV 8614 2.70 × 107 78 2.49 4.00 14
B2 + 50G 853 1.74 × 106 51.8
B2 + 50G TFOT 1842 2.89 × 106 57.4 2.16 1.66 5.6
B2 + 50G PAV 2185 6.48 × 106 63.8 2.56 3.72 12
B2 + 40G + 10HL 1294 1.38 × 106 53.2
B2 + 40G + 10HL TFOT 1540 2.79 × 106 58.4 1.19 2.03 5.2
B2 + 40G + 10HL PAV 2394 6.51 × 106 62.8 1.85 4.73 9.6
B2 + 30G + 20HL 1644 1.61 × 106 55.8
B2 + 30G + 20HL TFOT 2573 3.00 × 106 59.2 1.57 1.86 3.4
B2 + 30G + 20HL PAV 2924 6.12 × 106 63.8 1.78 3.81 8

the mastics without hydrated lime as can be seen 12.00

from the initial viscosity, softening points and
complex modulus values before the TFOT and 10.00

PAV ageing. This higher initial stiffening effect of

8.00
the hydrated lime on the mastic was expected as
the hydrated lime filler has more surface area and 6.00
Rigden voids, which is well known in the rheology
of suspensions, and arises from an increased effec- 4.00
-+-81 (40/60)
tive volume fraction of the particles as captured by 2.00
-+- 82 (1001150)
a low maximum packing fraction. The stiffening
ratio of the fillers was calculated to compare the 0.00

effect of fillers on the mastics prepared with differ- (Bl/B21•SOG (B1/B2)+40G+10Hl (Bl/B21•30G•20Hl

ent binders and to show the effect of the hydrated 6.00

lime on mastics stiffening. The stiffening ratio from
viscosity and complex modulus was calculated by 5.00

dividing the values of viscosity or complex modu-

lus of mastics with the values of the base bitumen 4.00

used for preparing the mastic. For the softening 3.00

point, the stiffening ratio was calculated as the
difference in softening point values between the 2.00

mastic and the base bitumen. The stiffening ratio -+-81 (40/60)
1.00
results are shown in Figure 2. It is obvious that the -+- 82 (1001150)
mastics prepared with the 40/60 (B2) bitumen have 0.00
a higher stiffening ratio at all filler compositions. It (Bl/B2)•SOG (Bl/B2)•40G•10Hl (Bl/B2)•30G•20Hl

also can be noticed that the effect of the hydrated

lime replacement on the mastic stiffening is dif- Figure 2. The stiffening ratio of fillers on mastics a)
ferent between the two used binders in this study. Softening points b) Complex modulus.
These results indicate that the hydrated lime inter-
act with the bitumen and this interaction depends
on the bitumen composition and source. granular type particles of hydrated lime with very
The higher stiffening ratio of mastics contain- rough texture as can be seen from the microscopic
ing hydrated lime can be related to the higher spe- morphology by SEM imaging. However Granite fill-
cific surface area of hydrated lime filler also to the ers have mostly angular type particles with smooth

284
to slightly rough texture. The SEM imaging results the mastics were aged in the TFOT oven for differ-
can be correlated with the other filler tests as well. ent ageing times up to 20 hours. The ageing indices
Regarding the ageing effect, it is clear that all were calculated as mentioned earlier. The results
different ageing indices for the tested mastics are are presented in Figure 3. The results show that as
increasing with the ageing process. That means the mentioned earlier, the ageing indices of the mas-
mastics become stiffer with ageing. In addition, tics with soft binder 100/150 (B2) are lower than
the effect of hydrated lime on the ageing behav- the mastics prepared with the 40/60 bitumen (B1).
iour of the mastics is clear. The mastics contain- This is more obvious with the mastics containing
ing hydrated lime have lower ageing indices for the hydrated lime. These results indicate that hydrated
most samples tested. It could also be noticed, that lime changes the ageing susceptibly of the mas-
by replacing granite filler by 20% hydrated lime tics. It also indicates that an interaction between
does not make significant additional effect com- hydrated lime and bitumen could happen which is
pared to the 10% hydrated lime replacement. not the case with granite.
Also, from Table 2, the effect of the hydrated Also the effect of hydrated lime on slowing
lime on the ageing indices of mastics prepared down the ageing rate is clear, the ageing indices of
with the two different binders can be compared. the mastics containing hydrated lime are always
It is clear that the mastics prepared with the soft lower than mastics with only granite fillers with
bitumen (B2) have lower ageing indices in the both bitumen binders used in this study.
most tested mastics. However it is important to In order to compare the different ageing indices
mention that the bitumen B2 itself has a higher a correlation has been made and the results show
ageing index compared to bitumen B1 in all tests that there is a good correlation between these age-
used in this study. This indicates that an interac- ing indices. The correlation between the different
tion between the fillers and bitumen could happen ageing indices after the TFOT aging for different
and this depends on the chemical composition of hours is shown in Figure 4.
the bitumen and the filler type and concentration. A statistical analysis has been carried out and
Also it can be seen that the ageing indices calcu- the correlation coefficients were calculated and
lated from different tests are ranking the mastics checked, the results indicated that the correlation
ageing behaviour differently, however in a general is significant at the of 0.05 level.
trend the mastics containing hydrated lime always
have a lower ageing index. A correlation between
3.3 FTIR results
these ageing indices has been made and the results
show that these test results correlated well. The The infrared absorption spectrum between
results of the viscosity and softening points were 500 cm−1 and 4000 cm−1 has been recorded for bitu-
better correlated to each other than they are cor- men recovered from granite mastics. The absorp-
related with the complex modulus. tion bands belonging to asphalt functional groups
that are affected by ageing, such as carboxylic
acids (C = O) were used to evaluate the increase
3.2 Ageing time study of mastics
in oxidative ageing products, which can be deter-
In order to compare the effect of hydrated lime mined through FTIR, utilizing a very strong
and the bitumen type on the rate of mastic ageing, band at 1700 cm−1. The carbonyl aging index was

G.oo "F"•'=
s==
I +""
s=o=o:===~~~~~~~~~~~~~~~~~­ 25 n-•~B~l +~5~
0G~~.-~~~~~~~~~~~~~~-­

c B2 +soo ~ B 2 +50G
5.00 • BI +40G+ IOHL f-~~~~~~~~~~~~j-~- • B I +40G+ IOHL f-~~~~~~~~~-­
20
i!l B2 +40G+IOHL ~ B2 +40G+ IOHL

~ 4.00 • Bt+30G+20Hl f-~~~ • B1 +30G+20HL

13 B2 +30G+20HL = 15 rJ B2+ 30G+20HL

i
>-
§ 3.00 +--~~~~~~~~~~~~~~~-
~
> ~1o +-~~~~~~~~~~
< 2.00 +-~~~~~--t~~~-
<
1.00

0.00
5 10 20 5 . 10 20
Ageing Time (Hours) Age1ng Time (Hours)

Figure 3. Ageing indices of the mastics after TFOT ageing a) Viscosity ageing index b) Softening point ageing
index.

285
 5.5 .,....--

R' =0.872 1
.··
4.5
• ..·· 4.5
•
<(
~ 3.5 ..··
<(
-~ 3.5
.. ··
• ..·· •
• 0

•
. ,····· •
:;: 3

1.5
• . •..
.•.
!I• •
.. ··
..
> 2.5

1.5
.,

1 .. ..
• ······~
...
.•...·
..•···
•
II.'• 1 ll··· .. •
11 16 21
complex modulus G* AI softening point AI

Figure 4. Correlation between viscosity and softening point after TFOT ageing.

prepared with two different bitumen binders were

0.0090
• Bl+50%G investigated. The bitumen mastics were aged for
0.0080
short- and long-term ageing periods in the labo-
0.0070 ratory and the change in rheological and chemical
properties of the mastics and the recovered bind-
..
)(
0.0060
ers of the mastics were evaluated by means of the
"'.5 0.0050
different ageing indices and FTIR. The following
Ia 0.0040 conclusions could be drawn:
0.0030
1. The ageing of bitumen mastics is different from
0.0020 the ageing of pure bitumen.
0.0010 2. The mineralogy and the properties of the fillers
0.0000 can significantly affect the stiffening ratio of the
Un-aged TFOT PAV bitumen mastics and that depends mainly on
the filler type and concentration in the mastic.
Figure 5. Change of carbonyl index after different 3. The effect of the hydrated lime replacement on
aging stages. the mastics is clear and it can slow the rate of
ageing more than the granite filler.
4. The stiffening and ageing effects of hydrated
calculated and compared for the different mastics lime on the mastics can be different with mas-
at different ageing stages and the results are pre- tics of different bitumen binders.
sented in Figure 5. 5. The viscosity, softening point and complex
Figure 5 shows the carbonyl ageing index cal- modulus ageing indices for the binders and
culated for the recovered binder from the mastics fillers combinations used in this study are well
prepared with the 40/60 bitumen (B1) before and correlated with each other and also correlate
after different ageing stages. It can be seen that the with the chemical changes as indicated by the
binder recovered from the mastic with hydrated increase of the carbonyl index.
lime contains less oxygenated products after both
TFOT and PAV ageing. These results can also
explain the results shown in Table 2, where the REFERENCES
samples containing hydrated lime are less aged
after short and long-term ageing. So the increase “British Standard (2000) Bitumen and bituminous bind-
of carbonyl originated products can be correlated ers — Accelerated long-term ageing conditioning by
with the increase of the stiffness of the mastics as a Pressure Ageing Vessel (PAV), BS EN 14769:2012.”
measured by the viscosity, softening points and “British Standard (2007) Determination of the Soften-
also the complex modulus. ing Point, Ring and Ball Method, BS EN 1427: 2007.”
“Methods of test for petroleum and its products: Deter-
mination of the resistance to hardening under the
influence of heat and air Part 2: TFOT method, BS
4 CONCLUSIONS EN 12607-2:2000.”
Anderson, D., et al. (1992). Rheological properties of
In this study the effect of granite fillers and mineral filler-asphalt mastics and its importance to
hydrated lime on the ageing of bitumen mastics pavement performance. Effects of aggregates and

286
 mineral fillers on asphalt mixture performance, ratory Study with Mastics.” Transportation Research
ASTM International. Record: Journal of the Transportation Research
British Standard (2000) Determination of Particle Board (2506): 19–31.
Density and Water Absorption: Aggregates between Petersen, J.C. (2009). “A review of the fundamentals of
31.5 mm and 4 mm, BS EN 1097-6: 2000. asphalt oxidation: chemical, physicochemical, physical
Greenfield, M.L., et al. (2015). “XANES measurements property, and durability relationships.” Transporta-
of sulfur chemistry during asphalt oxidation.” Fuel tion Research E-Circular (E-C140).
162: 179–185. Petersen, J.C. and R. Glaser (2011). “Asphalt oxidation
Lesueur, D., et al. (2016). “A simple test method in order to mechanisms and the role of oxidation products on age
assess the effect of mineral fillers on bitumen ageing.” hardening revisited.” Road Materials and Pavement
Construction and Building Materials 117: 182–189. Design 12(4): 795–819.
Little, D.N. and J.C. Petersen (2005). “Unique effects Read, J. and D. Whiteoak (2003). The shell bitumen
of hydrated lime filler on the performance-related handbook, Thomas Telford.
properties of asphalt cements: physical and chemical Wu, J. (2009). “The influence of mineral aggregates and
interactions revisited.” Journal of Materials in Civil binder volumetrics on bitumen ageing.” Nottingham:
Engineering 17(2): 207–218. University of Nottingham.
Moraes, R. and H. Bahia (2015). “Effect of Mineral Fill-
ers on the Oxidative Aging of Asphalt Binders: Labo-

287
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Development of empirical models for the estimation of the rheological

properties of asphalt binders

Fernando Martinez, Marina Cauhape Casaux & Silvia Angelone

Road Laboratory, Institute of Applied Mechanics and Structures, University of Rosario, Argentina

ABSTRACT: The dynamic modulus of asphalt mixtures is the fundamental characteristic used in the
pavement design procedures based on mechanistic principles. As this property is strongly dependent of
the rheological properties of the bitumen, it is important to have a better understanding of the rheo-
logical behavior of asphalt binders. Recently, the Dynamic Shear Rheometer was adopted as a powerful
instrument to characterize the dynamic properties of bituminous materials. However, this is a high cost
equipment that it is not available worldwide and conventional tests like Penetration, Softening Point and
viscosity at different temperatures, are still used in many countries for the evaluation of the consistency
of these binders. The main purpose of this paper is the development of empirical models for the predic-
tion of these rheological properties using routine conventional testing results. It is concluded that reliable
estimations for paving bitumens could be obtained using the procedures developed in this study.

1 INTRODUCTION characterization tests like the Penetration at 25ºC,

Pen25, and the Ring and Ball Softening Point
Nowadays, pavement design procedures are mov- TR&B.
ing towards a more intensive application of mech- According to the conclusions stated by Van
anistic principles. In this case, the characterization der Poel, the accuracy of the nomograph is amply
of the mechanical properties of the different com- sufficient for practical applications in pavement
ponent materials is very important. engineering where Sbit could be estimated at any
For the asphalt mixtures, the main property of temperature condition and time of loading within
interest is the dynamic modulus E* which depends a factor of two (Bonnaure et al., 1977).
among other factors, of the stiffness of the asphalt Minor corrections to the original Van der Poel
binder. nomograph were introduced later. In 1966 and
Based on the simple concept of Young’s modu- 1973, Heukelom reshaped Van der Poel’s nomo-
lus adapted to viscoelastic materials, Van der Poel graph with slight corrections using the results
introduced the concept of Stiffness Modulus of obtained for hundreds of bitumens represent-
asphalt binders as a function of temperature and ing a variety of grades from different sources
loading time in 1954 (Van der Poel, 1954). (Heukelom, 1966, Heukelom 1973).
This Stiffness Modulus of the bitumen, Sbit, According to the own words of Van der Poel,
is simply defined as the ratio between stress and “there is nothing theoretical behind the nomo-
strain for a given bitumen temperature and time graph and all correlations have been empirically
of loading as: deduced” (Van der Poel, 1954). Thus, the math-
ematical functions used by Van der Poel in devel-
⎛σ ⎞ oping the multi variable nomograph were never
Sbit = ⎜ ⎟ (1)
⎝ ε ⎠ t ,T published. An approximation formula has been
proposed by Ullidtz for a limited portion of the
where σ = applied stress; ε = resulting strain; nomograph (Ullidtz, 1979).
t = time of loading; T = bitumen temperature.
Van der Poel developed a nomograph for the Sbit = ( .157 × 10 −7 )).((tt 0 368
).(e − PI ).(TR & B T )5 (2)
estimation of the bitumen stiffness for a given con-
dition of temperature and time of loading using where t = loading time in seconds; PI = Penetra-
the experimental results obtained during 20 years tion Index; T = bitumen temperature in ºC. The
of laboratory work for 47 different conventional formula is valid for the range of t between 0.01
bitumens. The input parameters for this nomo- and 0.1 seconds, PI between −1.0 and +1.0 and
graph are results obtained from normal routine (TR&B−T) between 10ºC and 70ºC.

289
 Several researchers have found several short- was adopted as a powerful instrument to charac-
comings when using the nomograph because it terize the full rheological properties of bituminous
was developed for unmodified bitumens and not materials over a wide range of temperatures and
suitable to be used for polymer-modified bitumens frequencies (Anderson et al., 1994).
(Read and Whiteoak, 2003; Anderson et al., 1994). In this equipment a sinusoidal oscillatory shear
Also, for a complete description of the rheological stress τ is applied to a thin disc of bitumen sand-
behavior of the bitumen, the phase angle between wiched between two parallel plates and the result-
stress and strain is required and this value cannot ing shear strain γ is measured.
be obtained from the Van der Poel nomograph The shear modulus G* is calculated as:
(Yussof et al., 2011).
In 1998 the Shell International Oil Products τ
developed a computerized version of the nomo- G* = (5)
γ
graph available as a commercial software named
Bands 2.0 (Shell, 1998). The phase angle δ between stress and strains is
Viscosity is a measure of the resistance to flow also determined.
of a liquid and is a fundamental rheological prop- The main advantage of this equipment is that
erty of this kind of materials. The viscosity is it allows the determination of these rheological
defined as the ratio between an applied shear stress parameters in a wide range of testing frequencies
and the resulting rate of shear strain as: and temperatures.
However, the DSR is a high cost equipment that
τ
η= is not available in many laboratories and thus it is
( )
(3)
γ necessary to have models or equations that could
estimate the fundamental rheological properties
of bituminous materials from more conventional
with η = viscosity; τ = shear stress; (dγ/dt) = rate of testing results. In many countries, empirical tests
shear strain γ. like Penetration and Softening Point or even vis-
The rotational viscometer test using the Brook- cosity tests are still used for the characterization
field rotational viscometer and the Thermosel and specification of asphalt binders.
system is presently considered the most practical Bonaquist et al. have proposed a model for the
test to determining the viscosity of bitumen. This estimation of G* an δ from viscosity measure-
device allows the testing of bitumen over a wide ments in the form (Bonaquist et al., 1998):
range of temperatures and rates of shear strain.
The relative low cost of this equipment is its a0 + a1ω + 2ω 2
⎛ G* ⎞ ⎛ 1 ⎞
main advantage allowing the determination of the η = ⎜ ⎟ ⋅⎜ (6)
viscosity as a conventional routine test in many ⎝ ω ⎠ ⎝ δ ⎟⎠
laboratories.
If the viscosity of the bitumen is measured at δ = 90 + ( + ) ⋅ l g( )
(7)
different temperatures, the ASTM A-VTS equa- +( ) ⋅ ⎡⎣ l g ( )⎤⎦
2
+
tion (ASTM, 2009) has been proposed as a rela-
tionship between testing temperature and viscosity
in the form of: where ai and bi = fitting parameters and ω = angu-
lar frequency used to measure G* and δ in rad/s.
log ⎡⎣ log (ViscT )⎤⎦ = A VTS.log (T ) (4) For ω = 10 rad/s, which is the specified test fre-
quency in the Superpave PG system, Equation (6)
results:
where ViscT = viscosity in cPoises; T = bitumen
temperature in ºR; A = viscosity intercept param- a3
eter; VTS = Viscosity Temperature Susceptibility ⎛ G* ⎞ ⎛ 1 ⎞
η = ⎜ ⎟ ⋅⎜ (8)
parameter. ⎝ 10 ⎠ ⎝ sin δ ⎟⎠
McLeod has also proposed an estimation pro-
cedure of the stiffness modulus of asphalt binders Unfortunately, the model is not able to proceed
using the Penetration at 25ºC and the viscosity at with the inverse estimation of G* and δ from vis-
135ºC. The model is presented in a nomograph that cosity measurements.
is more or less similar to Van der Poel’s nomograph In order to solve this difficult, Bari and Witczak
and with the same disadvantages (McLeod, 1972). have proposed a predictive model of the phase
After the Strategic Highway Research Program angle δ and the shear modulus G* using viscosity
SHRP was carried out in the United States of results of 33 different conventional and modified
America, the Dynamic Shear Rheometer (DSR) asphalt binders (Bari & Witczak, 2007).

290
 First, a modified A-VTS relationship reflecting Thesis (Morea, 2011) and others obtained by the
the effect of loading frequency on the binder vis- authors were considered totalizing 24 different
cosity is formulated as follows: asphalt binders, conventional (C) and modified
(M). The modified binders are a variety of bitu-
( )
log ⎡⎣ log Viscc f ,T ⎤⎦ = A′ VTS ′.log (T ) (9) men modified with very different modifiers like
SBS, rubber, recycled polyethylene and other non
conventional binder modifiers.
where Viscf,T = viscosity of the bitumen as a func- Table 1 presents the results of the conventional
tion of loading frequency f and temperature T tests of Penetration at 25°C (Pen25), Softening
in cPoises; A′ and VTS′ = modified A and VTS Point (TR&B) and the A and VTS parameters of the
parameters. A-VTS equation for these asphalt binders in their
original condition (unaged). All the Penetration
A′ = f 0 0527
A (10) and Softening Point results were obtained using
the conventional EN Standards (UNE-EN, 2007).
VTS
T f 0 0575
× VTS
T (11) All of these binders were also tested after dif-
ferent aging conditions (Rolling Thin Film Oven
Second, the phase angle δ is obtained from the Test, Pressure Aging Vessel Test at different tem-
following equation: peratures and recovered from pavements after they
were placed). Then, the results of Pen25, TR&B, A
δ = 90 + ( −7.3146 − 2.6162.VTS (
T ′ ) × log f ⋅Viscc f ,T ) and VTS parameters, and shear modulus G* and
phase angle δ at different temperatures and fre-
( )
2
T ′ ) × ⎡⎣ log f V
+ (0.1124 + 0.2029.VTS Visc f T ⎤⎦ quencies for all of them (original and aged), were
compiled in a large database containing more than
(12)
5200 rows. The Dynamic shear modulus G* and
Finally, the complex shear modulus is predicted phase angle δ results were obtained following a
as:
Table 1. Conventional properties of the considered
G* f .Viscc f T . asphalt binders in their original condition.
(13)
.(sin )7.1542 − 0.4929. f f2
Type Pen25 TR&B
Name (*) 10−1 mm ºC A VTS
This model was evaluated in a previous paper
(Martinez et al., 2015) and it was concluded that C11 C 89 47.4 12.8360 −4.3639
the model fail in predicting the phase angle accu- C21 C 60 54.2 12.0921 −4.0890
rately. Also, the main disadvantage of this model C31 C 58 51.8 11.8954 −4.0164
is the interrelationship between the estimates of C41 C 17 65.7 9.3312 −3.0688
δ and G* where the errors in the first prediction C51 C 54 50.5 11.5101 −3.8766
could affect the second one. C61 C 25 63.3 10.8691 −3.6295
Looking for avoiding the disadvantages found N11 C 26 58.0 11.4461 −3.8418
for the different analyzed estimation procedures, P11 C 112 47.0 10.8751 −3.6588
the objective of this paper is the development of P21 C 39 52.1 11.9631 −4.0379
new empirical models for the prediction of the rhe- PG11 C 69 48.5 10.9118 −3.6629
ological properties of asphalt binders using routine CA1 C 55 49.0 10.3591 −3.4645
conventional testing results like the Penetration CT3 C 45 53.0 10.8384 −3.6281
at 25ºC, the Softening Point or the viscosity. The M11 M 60 58.3 10.1551 −3.3733
models were calibrated with the information of MP11 M 64 69.2 12.0203 −4.0625
a large database specially compiled containing N21 M 46 84.5 8.4973 −2.7675
more than 6200 data points for conventional and AM3 M 62 65.0 7.7247 −2.5091
modified asphalt binders with different aging con- 1SB M 12 65.0 11.2311 −3.7502
ditions. The goodness-of-fit of these models was 2SB M 18 63.0 10.4793 −3.484
evaluated using graphical and statistical analysis 3SB M 17 68.0 10.2854 −3.4047
methods as it is shown in the following sections. 4SB M 9 70.0 10.0935 −3.3365
E11 M 45 59.0 9.8767 −3.2744
NV11 M 45 96.0 9.7413 −3.2158
2 MATERIALS AND METHODOLOGY S11 M 45 64.0 9.6177 −3.1753
TL11 M 36 58.0 10.5506 −3.5165
In this paper, experimental results reported by Bari
in his MSc. Thesis (Bari, 2001), Morea in his PhD. (*) C: Conventional; M: Modified

291
similar procedure as in the EN Standard at dif- The Penetration Index PI (Pfeiffer & Van Door-
ferent temperatures and frequencies (UNE-EN, maal, 1936) can be determined from the following
2012). equation:

20 − PI ⎛ log 800 − log Pen25 ⎞

3 DEVELOPMENT OF THE ESTIMATION = 50 ⋅ ⎜ ⎟⎠ (20)
MODELS 10 + PI ⎝ TR & B − 25

Four different empirical estimation models were The fitting parameters a1 to a5 have been
developed using the results compiled in the adjusted for all the results included in the database
database: with PI between −1.4 and 6.0, (TR&B−T) from −68
to 81ºC and frequency f between 0.16 and 16 Hz.
− Model of G* from the conventional test results
of Penetration at 25ºC (Pen25) and Softening
Point (TR&B) 3.2 Model δ - Pen25, TR&B
− Model of δ from the conventional test results of
The proposed model is:
Penetration at 25ºC (Pen25) and Softening Point
(TR&B)
δ = b1 ⋅ ( )
2
− Model of G* using the viscosity properties as −
input parameter. +[ ⋅ + ]⋅ ( − ) (21)
− Model of δ using the viscosity properties as +[ ⋅ + ]
input parameter.
These empirical estimation models, developed where δ is the phase angle in degrees and b1 to b5
by a trial and error procedure, have a polynomial are fitting parameters in the form of:
form and they were optimized by a nonlinear opti-
mization methodology using the Solver function b1 = −0.000106 (22)
included in the Microsoft Excel software minimiz-
ing the sum of the square errors between measured b2 . . PI − 0.000978 (23)
and estimated values in logarithmic space for G*
and in arithmetic space for δ. b3 .000308 . PI 2 + 0.000740 . PI − 0.007808 (24)

b4 .00 780 . PI 2 0.005437 . PI − 0.067407

3.1 Model G* - Pen25, TR&B (25)
The proposed model is: b5 . . PI 2 0.0187785 . PI + 1.385008
(26)
log(G *) a1 ⋅ (TR & B T ) +
2

+ [a log( f ) + a ] (TR & B − T ) (14) Also, the fitting parameters b1 to b5 have been
+ [a4 log( f ) + a5 ] adjusted for to the same range of values as for the
model G* - Pen25, TR&B.
where G* is the dynamic shear modulus in Pa, f is
the testing frequency in Hz, T is the testing tem- 3.3 Model G* - ViscT
perature inºC and a1 to a5 are fitting parameters in
the form of: The proposed predictive model using the viscosity
as input parameter is:
a1 = 0.000232 (15)
log(G *) c1 ⋅ ⎡⎣ log (ViscT )⎤⎦
2

a2 0.000446 . PI − 0.004097 (16)

+ ⎡c2 ⋅ log(T f ) 3 ⎤ log l g (Vis
V cT ) (27)
a3 0.000227 . PI 2 0.006002 . PI + 0.059462 (17) ⎣ ⎦
+ ⎡c4 ⋅ l (T f ) + c5 ⎤
a4 0.00 89 . PI 2
002189 0.027486 . PI + 0.791450 (18) ⎢⎣ ⎥⎦
a5 003872 . PI 2
0.00387 0.178917 . PI + 3.825399 where ViscT is the viscosity at the testing tempera-
(19) ture T in cPoises and c1 to c5, fitting parameters as:

The temperature T and frequency f are those of c1 0 00141 VTS 2 + 0.00465 . VTS
T − 0.03635 (28)
interest, required for the estimation of the corre-
sponding dynamic shear modulus G*. c2 0 01433 VTS 2 + 0.13124 . VTS
T + 0.36764 (29)

292
 1.E+09 . . . . - - - - - - - - - - - - - - - - - - - - - - - . , . ,
c3 0 04005 VTS 2 − 0.12529 . VTS
T + 1.19337 Model G* - Pen2s, TR&B
(30)
1.E+07
c4 0 09266 VTS 2 − 0.75342 . VTS
T − 2.83358
co
(31) a..
;;- 1.E+05
c5 0 01087 VTS 2 − 1.02485 . VTS
T − 3.81335 ~
co
E
(32) ~ 1.E+03
Q)

&
with VTS obtained from Eq. 4. 1.E+01
In this case, the fitting parameters were adjusted
o Conventional Asphalt Binders
for binders with A values between 7.5 and 14.3 and
1.E-01 -IL-------.----"T""""------r----"T""""-----l
VTS values from −4.9 to −2.4. The range of test-
1.E-01 1.E+01 1.E+03 1.E+05 1.E+07 1.E+09
ing temperatures was between 10 and 115ºC and
frequencies from 0.16 to 16 Hz. G* measured (Pa)

3.4 Model δ - ViscT Figure 1. Comparison of measured and estimated G*

values using the model G* - Pen25, TR&B for the conven-
This proposed predictive model results: tional asphalt binders.

δ = d1 ⋅ ⎡⎣ g ( )⎤⎦
2

modified asphalt binders are in well agreement for

+ ⎡d2 ⋅ log(T ⎤
f ) + d3 ⎥⎦ ⋅ g ( ) (33) the two developed models.
⎣
According to Van der Poel, a model with an
+ ⎡d 4 ⋅ log((T ⎤
⎣ f ) + d5 ⎥⎦ accuracy of a factor of 2 in the estimation of the
stiffness of an asphalt binder is amply sufficient
for engineering purposes (Van der Poel, 1954). For
with d1 to d5:
the model G* - Pen25, TR&B, 88% of the values are
within this accuracy range for all the asphalt bind-
d1 VTS + 0.01052 (34)
ers in the database. For the model G* - ViscT, 76%
d2 VTS − 0.07568 . VTS
2
T − 0.12880 of the estimated values are within the same accu-
(35) racy range.
To quantitatively evaluate the performance of
d3 VTS 2 − 0.04363 . VTS
T − 0.31727 (36) the proposed models, the quality of the compari-
d4 VTS 2 + 0.44782 . VTS
T + 0.81228 (37) sons between measured and estimated values was
assessed using goodness-of-fit statistics according
d5 VTS 2 + 0.29045 . VTS
T + 2.99781 to subjective criteria proposed by Witczak et al.
(38) (2002), and shown in Table 2. The statistics include
correlation coefficient, R2 and Se/Sy (standard
The fitting parameters d1 to d5 have been error of estimate values/standard deviation of
adjusted for to the same range of data values as for measured values).
the model G* - ViscT. Table 3 presents the evaluation of the quality of
the comparisons according to these criteria for the
two estimation models of G*. This evaluation was
4 EVALUATION OF THE PROPOSED done for the conventional (C) and modified (M)
ESTIMATION MODELS asphalt binders separately and also for both data
sets assembled.
Figures 1 and 2 show the comparison between The goodness-of-fit statistics show that the
measured and estimated G* values using the comparisons between estimated and measured val-
G* - Pen25,TR&B model in log-log space for the ues are excellent for both estimation models.
conventional and modified asphalt binders Figures 5 and 6 show the comparison of esti-
respectively. mated and measured phase angle δ for the model
Figures 3 and 4 show the comparison between δ - Pen25, TR&B and for the conventional and modi-
measured and estimated G* values using the G* - fied asphalt binders respectively.
ViscT model in log-log space for the conventional For the estimation of the phase angle δ, it
and modified asphalt binders respectively. was considered that an accuracy of 10 degrees
The comparison between measured and is enough sufficient for pavement engineering
estimated G* values for the conventional and purposes.

293
 1.E+09 , -- - - - - - - - - - - - - - - - - ,
Model G* - Pen25• TR&B Model G*- Viser

1.E+07
1.E+07
...
"-
~ ;;- 1.E+05
"-
:;- 1.E+05
s... ~
~ 1.E+03
E
ti
Q)

b
1.E+03
" 1.E+01
~> Modified Asphalt Binders

1.E+01 1.E.01 ¥---~--~--~--~-----j

1.E.01 1.E+01 1.E+03 1.E+05 1.E+07 1.E+09

• Modified Asphd Binders
G• measured (Pa)
1.E-01 ¥---~--~---~--~-----j

1.E-01 1.E+01 1.E+03 1.E+05 1.E+07 1.E+09

Figure 4. Comparison of measured and estimated
G* values using the model G* - ViscT for the modified
G* measured (Pa)
asphalt binders.
Figure 2. Comparison of measured and estimated G*
values using the model G* - Pen25, TR&B for the modified
asphalt binders. Table 2. Criteria for Goodness-of-Fit statistical
parameters.

Criteria R2 Se/Sy
Model G'- Vise,
Excellent ≥ 0.90 ≤ 0.35
1.E+07 Good 0.70–0.89 0.36–0.55
~
Fair 0.40–0.69 0.56–0.75
!!:. 1.E+05 Poor 0.20–0.39 0.76–0.89
"C
s... Very Poor ≤ 0.19 ≥ 0.90
.5
u;
Q)
1.E+03

b
1.E+01 Table 3. Goodness-of-Fit statistics for the estimation
o Conventional Asphalt Binders models of G*.
1.E-01 -J<---~--~---~--~-----j
Model Set R2 (%) Se/Sy Evaluation
1.E-01 1.E+01 1.E+03 1.E+05 1.E+07 1.E+09

G* measured (Pa) G* - Pen25, TR&B C 99.1 0.09 Excellent/Excellent

M 98.2 0.13 Excellent/Excellent
Figure 3. Comparison of measured and estimated G* All 99.0 0.10 Excellent/Excellent
values using the model G* - ViscT for the conventional G* - ViscT C 98.9 0.10 Excellent/Excellent
asphalt binders. M 94.5 0.24 Excellent/Excellent
All 97.4 0.16 Excellent/Excellent

For the model δ - Pen25, TR&B, 96% of the esti-

mated values are within this range while for the 00 ,------------~
model δ - ViscT, 90% of the estimated values are in 00
the same range.
70
Also, Figures 7 and 8 show the same compari-
sons for the model δ - ViscT.
Table 4 presents the statistics for these models
and for the same data sets considered previously.
The quality of the estimations is Excellent for the
conventional asphalt binders and it is between
10
Good and Fair for the modified ones. ., Conventional Asphalt Binders

Based on the obtained results it could be stated

10 20 30 40 50 60 70 80 90
that the four models presented in this paper are & measured (')
capable to predict the rheological properties of
conventional and modified asphalt binders with Figure 5. Comparison of measured and estimated δ
enough accuracy to be applied for practical pur- values using the model δ - Pen25, TR&B for the conven-
poses in pavement engineering. tional asphalt binders.

294
 90 Table 4. Goodness-of-Fit statistics for the estimation
80
Modell>- Pen2s,TR&s models of δ.

70 Model Set R2 (%) Se/Sy Evaluation

e::: 60
"'C δ - Pen25, TR&B C 93.8 0.22 Excellent/Excellent
2 50 M 83.9 0.45 Good/Good
ns
E
~
40 All 92.2 0.26 Excellent/Excellent
Q)
I¢ 30
δ - ViscT C 93.8 0.20 Excellent/Excellent
M 71.5 0.66 Good/Fair
20
All 87.5 0.31 Good/Excellent
10 6 Modified Asphalt Binders

10 20 30 40 50 60 70 80 90 5 CONCLUSIONS

This paper has reviewed some models and procedures

Figure 6. Comparison of measured and estimated δ used to estimate the rheological properties of asphalt
values using the model δ - Pen25, TR&B for the modified binders from routine conventional testing results.
asphalt binders. Experimental results reported in the analyzed lit-
erature and others obtained by the authors describ-
ing the rheological properties of 24 different asphalt
90 binders were compiled in a large database.
Modell) - Viscr
80 Four different empirical estimation models were
70
proposed for the prediction of G* and δ values
using conventional test results like Penetration
e::: 60 at 25ºC, Softening Point or Viscosity properties.
"'C
2 50 These estimation models, developed by a trial and
ns
E 40 error procedure, have a polynomial form and they
~
Q) were optimized by a nonlinear optimization meth-
I¢ 30
odology using the Solver function included in the
20 Microsoft Excel software.
10 o Conventional Asphalt Binders
It could be concluded that the developed models
can adequately estimate the rheological properties
of conventional and modified asphalt binders to
10 20 30 40 50 60 70 80 90
be applied with practical purposes in pavement
engineering.
However these models are merely predictions in
Figure 7. Comparison of measured and estimated
δ values using the model δ - ViscT for the conventional
order to have a first order estimation and they do
asphalt binders. not replace the experimental determination of the
rheological properties of the bituminous binders.
They should be used with caution within the range
90 of input values considered for the calibration of
Modell> - Viscr the different models.
80
Finally, it should be pointed out that these mod-
70
els were evaluated on an asphalt binder database
e::: 60 available at the time this paper was prepared. As
"'C
2ns 50 this is one of the objectives of an ongoing research
E 40 project at the University of Rosario, other experi-
ti
Q) 64f'
mental results are being added to the database in
I¢ 30 order to verify or recalibrate the considered models
20 to be included in a practical design procedure of
asphalt pavements based on mechanistic principles.
10 6 Modified Asphalt Binders

10 20 30 40 50 60 70 80 90 ACKNOWLEDGEMENT

The authors would like to express their gratitude

Figure 8. Comparison of measured and estimated δ values to Dr. Francisco Morea for sharing rheological
using the model δ - ViscT for the modified asphalt binders. data of those binders tested during his PhD Thesis.

295
 The contents of this paper reflect the views of logical properties of paving asphalt binders. Bitumi-
the authors who are solely responsible for the facts nous Mixtures and Pavements VI, London, Taylor and
and accuracy of the data presented herein. Francis Group.
McLeod, N.W. 1972. A four year survey of low tempera-
ture transverse pavement cracking on three Ontario
test roads. Proc. Association of Asphalt Paving Tech-
REFERENCES nologists, Vol. 41: 424–493.
Morea F. 2011. Deformaciones permanentes en mezclas
American Society of Testing and Materials. 2009. asfalticas: efecto de la reología de los asfaltos, la tem-
D2493-01 Standard Viscosity-Temperature chart for peratura y condiciones de carga. Tesis Doctoral. Uni-
asphalts”. Volume 04.03. versidad de La Plata, Argentina (In Spanish).
Anderson D.A, Christensen D.W., Bahia H.U., Dongré Pfeiffer, J.P. and van Doormaal. 1936. The rheological
R., Sharma M.G. and Antle C.E. 1994. Binder char- properties of asphaltic bitumen. Journal of the Insti-
acterization and evaluation, vol. 3: physical charac- tute of Petroleum Technologists. 22: 414.
terization. Report SHRP-A-369. Read J, Whiteoak D. 2003. The Shell bitumen handbook.
Bari, J. 2001. Investigation of the rheological properties 5th ed. London. Thomas Telford Publishing.
of typical asphalt binders used in Arizona. MSc. The- Shell International Oil Products. 1998. Bands2.
sis. Arizona State University. Ullidtz, P. 1979. A fundamental method for the predic-
Bari, J. and Witczak, M.W. 2007. New predictive models tion of roughness, rutting and cracking in asphalt
for viscosity and complex shear modulus of asphalt pavements. Proc Association of Asphalt Paving Tech-
binders: for use with mechanistic–empirical pavement nologists. 48: 557–586.
design guide. Transportation Research Record: Journal UNE-EN 1426 Standard, 2007: Determination of needle
of the Transportation Research Board, 2001: 9–19. penetration,
Bonaquist R, Pellinen T and Witczak M. 1998. Devel- UNE-EN 1427 Standard, 2007: Determination of the
opment of relationship between binder viscosity and softening point–Ring and ball method.
stiffness. Department of Civil Engineering, University UNE-EN 14770 Standard, 2012: Determination of com-
of Maryland, College Park. plex shear modulus and phase angle - Dynamic Shear
Bonnaure F, Gest G, Gravois A and Uge P. 1977. A new Rheometer (DSR).
method of predicting the stiffness of asphalt paving Van der Poel C. 1954. A general system describing the
mixtures. Proc Association of Asphalt Paving Tech- viscoelastic properties of bitumens and its relation to
nologists, 46:64–104. routine test data. J. Appl. Chem.; 4:231–6.
Heukelom W. 1966. Observations on the rheology and Witczak, M., Pellinen T. and El-Basyouny M. 2002. Pur-
fracture of bitumens and asphalt mixes. Proc Associa- suit of the simple performance test for asphalt con-
tion of Asphalt Paving Technologists. 36:359–97. crete fracture/cracking. Proc Association of Asphalt
Heukelom, W. 1973. An improved method of character- Paving Technologists. 71: 767–778.
izing asphaltic bitumens with the aid of their mechan- Yusoff, N.I., Shaw, M.T. and Airey, G.D. 2011. Model-
ical properties. Proc Association of Asphalt Paving ling the linear viscoelastic rheological properties of
Technologists. 42: 67–98. bituminous binders. Construction and Building Mate-
Martinez, F, Angelone, S. and Cauhape Casaux, M. rials 25, 2171–2189.
2015. Evaluation of predictive models for the rheo-

296
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

An alternative method for determining thermal stress in asphalt binder

based on Laplace transform

A. Cannone Falchetto, C. Riccardi, D. Wang & M.P. Wistuba

Technische Universität Braunschweig, Braunschweig, Germany

K.H. Moon
Korea Expressway Corporation, Gimcheon, South Korea

ABSTRACT: Low temperature cracking is one of the most severe distresses for asphalt pavement
experiencing severely cold weather conditions. As temperature drops, significant tensile stress develops
in the restrained asphalt layers, and, when the material strength is overcome, cracking occurs. Many road
authorities recognize thermal stress as a crucial parameter for evaluating the low temperature perform-
ance of asphalt pavement. Thermal stress is conventionally computed with a two-step approach where
the relaxation modulus is derived from the experimental creep compliance after which the convolution
integral is numerically solved. In this paper, a one-step computation solution based on Laplace transfor-
mation is proposed. Thermal stress and corresponding critical cracking temperature of asphalt binder are
computed and graphically and statistically compared to the values obtained with traditional approach. It
is observed that use of Laplace transformation provides reasonably close results to those obtained with
the conventional two-step solution.

1 INTRODUCTION determining the relaxation modulus and the choice

of the numerical solution for obtaining thermal
Low temperature cracking is the dominant failure stress and critical cracking temperature.
mode in asphalt pavements built in cold regions. Asphalt binder presents Linear Viscoelastic
As temperature rapidly decreases, thermal stresses (LVE) characteristics when the level of the imposed
develop in the restrained asphalt surface layer and, stress is sufficiently low. Therefore, the functions
beyond a threshold limit, known as critical crack- used for describing the linear viscoelastic material
ing temperature Tcr, cracking occurs potentially behavior are interrelated (Ebrahimi et al. 2014).
leading to premature failure (Marasteanu et al. Hopkins & Hammings’ algorithm (1967) is com-
2012; Cannone Falchetto & Moon 2015). monly used for deriving E(t), from D(t) experimental
Road authorities and pavement agencies have data, while Christensen, Anderson and Marasteanu
allocated significant resources for large mainte- (CAM) model (Marasteanu & Anderson 1999) is
nance plans over the years. At the same time, a applied, in a second step, for calculating σ(T) and Tcr
number of experimental and numerical proce- (Moon et al. 2013). This procedure provides a numer-
dures were introduced and developed to explain ical technique for solving the convolution integral
and describe the low temperature behavior of expressing the exact interrelationship between E(t)
asphalt pavement materials (AASHTO T313 2012, to D(t). However, the solution proposed Hopkins
Marasteanu et al. 2012, Moon et al. 2013, Can- & Hammings’ (1967) is rather complex and several
none Falchetto et al. 2014, Farrar et al. 2015). calculation steps are required for obtaining thermal
Creep testing (AASHTO T313 2012) is conven- stress and corresponding critical cracking tempera-
tionally used to obtain asphalt binder and mixture ture. Alternatively, Laplace transformation can be
creep compliance, D(t), from which relaxation used for solving the convolution integral and for
modulus, E(t), can be derived through specific determining σ(T) and Tcr. The Laplace transform is
inter-conversion procedures. Finally, thermal stress, a convenient integral transform and presents a very
σ(T), and critical cracking temperature, Tcr, which simple solution for converting differential equations
can be used for estimating the stress at failure, can into algebraic equations and convolution into mul-
be calculated (Moon et al. 2013). Two fundamental tiplication. The use of such a simpler mathematical
steps can be identified in the calculation process of tool would make the calculation of low temperature
thermal stress and critical cracking temperature: cracking parameters more accessible to practition-
the selection of the inter-conversion method for ers on routine basis.

297
 In this paper, the low temperature properties Table 1. BBR experimental results.
of asphalt binder (σ(T) and Tcr) obtained from
Hopkins & Hammings’ algorithm (1967) in combi- Binder S(60 s) CoV m(60 s) CoV Aging (1R 2P)
nation with CAM model (Marasteanu & Anderson PG [MPa] [%] [–] [%] & Temp
1999) are graphically and statistically compared to 58-28 74.3 3.9 0.545 1.8 R, −12ºC
the corresponding values determined with Laplace 179.6 1.4 0.429 1.6 R, −18ºC
transformation. The results are, then, analyzed and 448.2 1.0 0.315 1.6 R, −24ºC
discussed. Finally, the major findings for thermal 112.8 0.9 0.425 3.2 P, −12ºC
stress and Tcr computations using the two different
242.3 0.5 0.348 1.4 P, −18ºC
calculation methods are reported and a practical
537.7 1.1 0.274 2.1 P, −24ºC
recommendation is proposed.
58-34 71.1 4.7 0.622 1.9 R, −18ºC
178.7 2.8 0.423 0.8 R, −24ºC
449.6 0.9 0.319 1.3 R, −30ºC
2 EXPERIMENTATION 108.1 6.7 0.505 2.5 P, −18ºC
238.7 2.7 0.352 1.1 P, −24ºC
A set of two asphalt binders for pavement appli- 529.2 2.3 0.289 1.2 P, −30ºC
cations, were used in the experimental phase of
1
this research. The binders presented Perform- R: RTFOT aged 2P: PAV aged
ance Grade (AASHTO M320 2010) PG 58-28
(unmodified binder) and PG 58-34 (SBS (Styrene Three asphalt binder replicates were prepared
Butadiene Styrene modified binder). Both asphalt for testing at each temperature-aging condition
binders were short and long term aged accord- corresponding 18 measurements per asphalt binder
ing to the Rolling Thin Film Oven Test (RTFOT) type. Table 1 summarizes the mean experimental
(AASHTO T240 2013) and Pressurized Aging results for creep stiffness and relaxation parameter
Vessel (PAV) (AASHTO R028 2012) procedures, measured at 60 s: S(60 s) and m(60 s). CoV stands
respectively. for coefficient of variation.
Bending Beam Rheometer (BBR) creep tests The good repeatability of BBR tests is confirmed
(AASHTO T 313 2012) were performed on thin by the low CoV shown by both creep stiffness and
asphalt binder beams (102.0 ± 5 mm × 12.7 ± 0.5 m-value.
mm × 6.25 ± 0.5 mm). In this tests, the mid-span
deflection, δ(t), of an asphalt binder beam is meas-
ured for the entire duration of the test (240 s), and
then used to calculate the creep stiffness, S(t) (and 3 COMPUTATION OF THERMAL
its inverse creep compliance, D(t)), together with STRESS AND CRITICAL CRACKING
the corresponding relaxation parameter, m-value TEMPERATURE
(AASHTO T 313 2012).
In order to generate relaxation modulus, E(t), 3.1 Hopkins and Hamming algorithm
master curves and, hence, to calculate thermal The convolution integral is commonly used to link
stress, σ(T), and critical cracking temperature, Tcr, the values of D(t), and relaxation modulus, E(t), of
BBR creep tests were performed at three different a viscoelastic material (Ebrahimi et al. 2014):
temperatures: low(PG+10)°C, low(PG+10°C)-6ºC
and low(PG+10°C)+6ºC. t

Generally, only two different temperature con- ∫ E (t ) ⋅ D(t )dτ t (1)

ditions are needed to compute thermal stress 0

and Tcr (Moon et al., 2013); however, three dif-

ferent testing temperatures were considered in where t > 0 is time.
this paper. This was done to obtain a sufficiently In the past, Hopkins and Hammings proposed
large data set for providing extensive compari- a numerical solution (1967) to solve Equation 1,
son and validation of the numerical computation which allows to determine E(t), from D(t). First a
techniques (Hopkins & Hammings’ algorithm time interval is selected as t0 = 0, t1 = 1, … t240 = 240 s.
and Laplace transformation) and corresponding Then, the integral form of D(t) is expressed as:
results. Therefore, the following two cases were
t
considered:
f t ) = ∫ D(t )dt (2)
• Case 1: σ(T) and Tcr between temperature 0
low(PG+10)ºC and low(PG+10)-6ºC
• Case 2: σ(T) and Tcr between temperature Equation 2 is then rewritten in a discretized
low(PG+10)ºC and low(PG+10)+6ºC form as:

298
 1 thermal stress. For this purpose, the shift factor, aT,
f tn ) = f tn ) + ⋅ (D
( D(tn + ) D(tn )) ⋅ (t
(tn tn ) (3) and parameters C1 and C2 at the reference tempera-
2
ture Ti = 22ºC were first determined based on the
where: experimental BBR creep stiffness curve:
tn + 1 n ti + 1
LogaT ( reff = C1 C2 ⋅ T (9)
(4) C)
tn + 1 = ∫ E tt′′ ⋅ D tn + − t ′ dt ′ ∫ E (t′′ ) D(tn +1 t ′ )dt ′
0 i =0 ti
aT = 10C C ⋅T
= 10 C1 C2 ⋅(Ti C0 ⋅t )
(10)
aT = 10(C C ⋅Ti ) C2 ⋅C t
= 10C3 +C4 ⋅t = A0 ⋅10C4 ⋅t
Based on this set of equations, the relaxation
modulus can be obtained as: Then the master curves of D(ξ) in the reduced
n −1 time domain were generated at temperature equal
tn ∑ EE(t(tt i +1/ 2 f (ti f (ti )] to low (PG + 10)ºC and low (PG + 10) − 6ºC for
Case 1 and to low (PG + 10)ºC and low (PG + 10)
E (tn +1/ 2 ) = i =0
(5)
f tn +1 ) − f tn ) + 6ºC for Case 2:

where f(t0) = 0, E(t0) = 0 and E(t1) = t1/f(t1). D(ξ ) a ξ b + c ⋅ ξ d (11)

The CAM model (Marasteanu & Anderson
1999) can be used to generate E(t) master curves where a, b, c and d are fitting parameters.
based on BBR experimental data obtained at two In order to identify the most suitable thermal stress
different testing temperatures (Case 1: low (PG + computation methods, two different approaches
10)°C and low (PG + 10) − 6ºC or Case 2: low (PG were selected for generating master curves:
+ 10)°C and low (PG + 10) + 6ºC) as: a. Use the two data sets D (ξ, at low (PG + 10)ºC)
w and D (ξ, at low(PG + 10) − 6ºC for Case 1 or
E g − ⋅ Log ⎡1+ ( ) ⎤⎦
v
LogE Log t Log aT Log tc
(6) low (PG + 10) + 6ºC for Case 2) distinctly to
v ⎣
obtain a wider master curve;
b. Use the average D(ξ): i.e. average between D(ξ, at
where Eg is the glassy modulus, 3GPa for asphalt low (PG + 10)ºC) and D(ξ, at low (PG + 10) − 6ºC
binders according to Moon et al. (2013), tc, v and w for Case 1 or low (PG + 10) + 6ºC for Case 2).
are fitting parameters, aT is the horizontal shift factor.
Thermal stress can be then calculated by solving Strain and thermal stress were then related
the following convolution integral: under the assumption of an idealized one dimen-
sional scheme:
t

σ( ) ∫ ε( ) ⋅ E(( )dt ′ (7) ξ

∂σ
ξ
∂( ΔT )
−∞ εt ∫ ξ ξ′
∂ξ ′
dξ ′ ∫ α (ξ ξ ′)
∂ξ ′
dξ ′ = 0 (12)
0 0

where σ(t) is the time dependent stress, ε ( ) αΔT

is the strain rate, α is the coefficient of thermal and the Laplace transform applied as:
expansion assumed equal to 0.00017°C (Moon et al.
2013), and ΔT is the temperature cooling rate. The L (ε t ) D( ) ⋅ σ ( s ) + s α ( s ) ⋅ ΔT ( s ) = 0 (13)
numerical solution of Equation 7 can be obtained
by estimating thermal stress, σ(ξ), in the reduced Based on Equation 13, the following expression
time domain, ξ = t/aT, within a specified temperature of thermal stress can be written:
range: for example, 22ºC to −40ºC (Moon et al. 2013)
as selected for this study. In the present research, the
α ( ) ΔT ( )
value of cooling rate for asphalt binder was set to σ( ) = − (14)
2ºC/hour for all the computation cases and applied D(( )
to the following expression of thermal stress:
Equation 14 needs to be first back-transformed
d((α
t
) to the reduced time domain, (ξ) using Stehfest
σ ξ) = ∫ ⋅ E((ξ ξ (t ))dt ′ (8) algorithm (Stehfest 1970) and then fitted with the
−∞
dt ′ following simple power function:

3.2 Laplace transformation σ ξ ) = a + b ⋅ξ c (15)

In the present research, Laplace transformation was Finally, the value of thermal stress in the
selected as an alternative solution for computing reduced time domain, σ(ξ), is converted to the

299
 cr(T"C) comparison (PG 58-28-R-Cl)
- - - Thermal stress 0.9
----------- SAP method -o- L.T.-~~:~:~~d I
___________! I
0.8
~ 0.7
!\
I
-+- H&H Algorithm
G o.6
!~
~ 0.5
l 0.4
\
~ 0.3
\
c; 0.2
\
5 0.1 ~
!~
~ 0.0
-50 -40 -30 -20 -10 10 20 30
Temperature,("C)

Temperatme
cr(T"C) comparison (PG 58-28-P-Cl)
Figure 1. Schematic of the Shenoy (2002) SAP method 1.4
for computing Tcr. 1.2
l J -D-L.T.-Combined I
~
l I ··*·· L.T.-Averaged
-+- H&H Algorithm
1.0

actual time domain, σ(T,t) and compared with the

values obtained from the inter-conversion solution
G 0.8
~
~ 0.6
~
· ~\
proposed in the previous section in the same tem- ~ 0.4 \
perature interval: from 22ºC to −40ºC with 2ºC/ ~
~ 0.2
:' -.
hour of cooling rate. ~ 0.0
The critical cracking temperature, Tcr, was -50 -40 -30 -20 -10 10 20 30
finally estimated with the Single Asymptote Proce- Temperature,("C)

dure (SAP) method (Shenoy 2002). This approach

was selected since, in the present study, no strength cr(T"C) comparison (PG 58-28-R-C2)
tests were performed on asphalt binder at low tem- 0.9
perature. Figure 1 presents the schematic of the 0.8 -o- L.T.-Combined 1:

SAP procedure to estimate Tcr. ~ 0.7

\
: ~1~A;~;~~~:m
G o.6 -~
~ 0.5
~
\
4 THERMAL STRESS COMPARISON 0.4
\
~ 0.3 -~
~ 0.2
Two different computation approaches for the ~
~ 0.1
solving convolution integral and computing σ(T)
and Tcr, were selected for this study:
0.0
-50 -40 -30·~ -20 -10
Temperature,("C)
10 20 30

1. Hopkins and Hammings’ algorithm in combi-

nation with CAM model (Case 1: D(t) results at
low(PG + 10)ºC & low(PG + 10) − 6ºC, Case 2: cr(T"C) comparison (PG 58-28-P-C2)
D(t) results at low(PG + 10)ºC & low(PG + 10) + 1.2
-o- L.T. -Combined

:;;:::~:.~~~:m
1-
6ºC); ~
2. Laplace transformation using the averaged
1.0
I
value of D(t) obtained at different temperatures G
e_
0.8
-\
or the distinct experimental D(t) curves com- ~ 0.6
bined into a master curve using the experimen- ~ 0.4
\

'
tal results at two temperature conditions (Case
1: low(PG + 10)ºC & low(PG + 10) − 6ºC and ~ 0.2 ~-
Case 2: low(PG + 10)ºC & low(PG + 10) + 6ºC). ~ 0.0
-50 -40 -30 -20 -10 10 20 30
Based on the procedures described in the pre- Temperature,("C)
vious sections, a set of thermal stress curves were
obtained for the two asphalt binders investigated. Figure 2. Thermal stress curves for PG 58-28 binder.
Figure 2 presents an example for binder PG 58-28.
R stands for RTFOT aged and P for PAV aged. binder is used, higher thermal stress at T = −40ºC
The calculation indicates that the two computa- can be observed when applying Laplace transfor-
tion processes provide comparable results in both mation. On the other hand, in the low tempera-
for Case 1 and 2. When the stiffer PG 58-28 asphalt ture range between −10ºC and −30ºC, Hopkins

300
and Hammings’ algorithm (1967) combined with 2. Method_1 (H&H) versus Method_3 (Laplace
CAM model (Marasteanu and Anderson, 1999) Combined: LC)
provides higher thermal stress compared to the 3. Method_2 (Laplace Average: LA) versus
Laplace transform. In most of the cases, no signifi- Method_3 (Laplace Combined: LC)
cant differences were found between the two differ-
ent Laplace approaches (although some moderate
differences can be seen from the statistical analysis
performed below). However, the transformation t-test results (PG 58-28-R-Cl)
1.0 ..........., ........... ,............ ,............ , -)!<
method based on the master curve obtained from 0.9 --- -----i-~- -o- M.I(H.H.) VS . M.2(L.A.) LL
the distinct D(t) experimental data showed slightly
higher values of σ(T) than when using the average ~:~ ::::::::::::: ----------i-1+::::::::·: __ ;_~:~~-:;:{;;:_~}~:2.-? ~-1- -------~
0 0.6 ::::::::::jf:\ ----------,1
value of D(t). It is interesting to observe that for the
modified PG 58-34 asphalt binder similar thermal l 0.5
::l 0.4
stress was calculated with all the three different com- ~ 0.3
putation methods. <>, 0.2
0.1
A simple statistical analysis based on hypothesis
0.0
test, t-test, was used to quantitatively evaluate the -50 -40 -30 -20 -10 10 20 30
results with 5% significant level (α = 0.05) (Cook Temperature,(•C)

and Weisberg, 1999). The t-test provides the numer-

ical significance of the statistical comparison which t-test results (PG 58-28-P-Cl)
measure the difference of mean between two testing 1.0

groups. For this purpose, the assumptions of data 0.9

-D-
I (H H M.2(L.A.) I
0.8 '''*'"
normality and constant variance were imposed and !\ )1:
'J(T vs. M.3 :L.C.)
0.7
the test hypotheses were set as follows: 0 0.6 !
0 ! 1\
Null hypothesis ~
II 0.5
I :o
0.4
::l
~ T /_

H o : μσ (T ) _ method μσ (T ) _ method (16) ~<>, 0.3

1 I !IT'
th d A: thh d A:2 0.2
0.1 -----------· -~--\~ ~ ~I !U
0.0
Alternative hypothesis: -50 -40 -30 -20 -10 10 20 30
Temperature,("C)

H a : μσ (T ) _ method
th d A: ≠ μσ (T ) _ method
thh d A:2 (17)
t-test results (PG 58-28-R-C2)
where μ is the mean of σ(T) obtained either from 1.0

Hopkins and Hammings’ algorithm or Laplace

transformation methods identified by symbols A:1,
0.9
0.8 + ----------i------------>--- H---- .. i------ .. ----f-..
0.7
- - - - -i- -,1~· .... .

and A:2. In addition, since Laplace transformation 0 0.6

---------H-t+-----·----j--'r- - ->~
- ---------·-J.JI::::J::~::::::::::!
0
II 0.5 + ----------,0: .. ..
was applied both to the mean value of D(t), as well ~
Q) 0.4
as to the D(t) master curves, the two approaches ~
::l
0.3
I I -o- M.1(H.H.) vs. M.2(L.A.~ l.i

were separated only for statistical analysis purposes: <,_ 0.2 ----------, \-----11-- ~~ J I * M.1(H.H.) vs. M.3(L.c.) 1-!
o.1 ---------- ~V--J ~L.-;-~?.<~:~:>,~s-._ ~:3".c~.c.) i
• Method_1: Hopkins and Hammings’ algorithm 0.0 ""'~'* nc ..
-50 -40 -30 -20 -10 10 20 30
(1967) with CAM model (1999) (H&H); Temperature,("C)
• Method_2: Laplace transformation with average
value of D(t) (Laplace Average: LA);
• Method_3: Laplace transformation with D(t) t-test results (PG 58-28-P-C2)
1.0
master curve (Laplace Combined: LC). I
0.9
Ti :I; ~ bI i~
Finally, the p-value, the output of the t-test, 0.8
i\ !A I V
0.7
which provides the level significance of the tested 0 0.6 \ I j 'Q. I
hypothesis, is computed. If the p-value is less than 'it \ ~ ! ?1
0.05 there is a significant difference in σ(T) between
~
0.5
t I v :o

two compared groups (calculation methods) other-

Q)
::l
~
0.4
0.3
'~ -D- M.1(H.H.) vs. M.2(L.A.) I
T
wise, the two tested groups are comparable. In this
<,_ 0.2 l:i
f, :~· ;rr· : · : ~:.- ~--:~_-~.-?
0.1 f---------- -k\( ~I
l
.M!"J .I'-
paper, the following three statistical comparisons 0.0
-50 -40 -30 -20 -10 10 20 30
were performed for two different computation Temperature,("C)
cases (i.e. Cases 1 and 2):
1. Method_1 (H&H) versus Method_2 (Laplace Figure 3. Results of t-test on thermal stress computa-
Average: LA) tion for PG 58-28 binder.

301
 No statistical differences were found between
t-test results (PG 58-34-R-Cl)
1.0 the two different Laplace transformation solutions
til (Method 2 and 3), except in one case: PG 58-28
0.9
0.8
vn ~ ....... binder in PAV aging condition down to T = −28ºC.
0.7 ........... :/~
V)
0 0.6 I When comparing Model 1 (H&H) to Model 2
6
II 0.5 6 (LA) and to Model 3 (LC) some significant trends
~
"::l 0.4
I
I -D- M.l(H.H.) VS. M.2(L.A.) 1::::::
• could be observed.
Above the critical temperature limit (T = −28ºC
~ 0.3
i\*
f ·········
I- )1( M.l(H.H.) VS. M.3(L.C.)
.::... 0.2
and −34ºC for the two binders, respectively), sta-
··········
~~ I ..... M ?{[ A) vs. M.3(L.C.)
0.1
0.0 tistically significant differences in σ(T) were found
-50 -40 -30 -20 -1 0 10 20 30 except for PG 58-34 binder in PAV aged condition.
Temperature,(°C)
However, these findings do not prevent the use of
Laplace transformation for computing thermal
t-test results (PG 58-34-P-Cl) stress for two reasons: first, in case of σ(T), tem-
1.0 peratures higher than the critical temperature limit
0.9 ··· .;,.
0.8
are not significant; second, thermal stresses in this
0.7 temperature range were rather small (less than
V)
0 0.6
~ ~
0.1∼0.2 MPa).
1 0.5
~--······ For softer PG 58-34 asphalt binder no differ-
"::l 0.4
~ 0.3
•···········
f\l ~t······
\
-D- M. 1(H.H.) VS. M.2(L.A.) I

: ~:~~~-·:} ~:.· ~--:~_·g-~ f ·········

ences in thermal stresses calculation were observed
.::... 0.2 at temperature lower than the specific low temper-
\ 7 ~'"\-,
0.1
-. ----- -- -------
ature limit (−34ºC) for all aging conditions. How-
0.0
-50 -40 -30 -20 -1 0 10 20 30
ever, in case of stiffer PG 58-28 binder comparison
Temperature,(°C) between Method 1 (H&H) and Method 2 (LA)
showed statistical differences in thermal stress cal-
culations. Based on these results, it appears that
t-test results (PG 58-34-R-C2)
1.0 the average values of D(t) obtained at two differ-
0.9 ~ ~ "1flDD{ 19. ent low temperature conditions does not represents
.in!\ \ .,oD'-"
0.8 ----------- ...
y\ \ a valid option for computing thermal stress when
0.7
V)
0 0.6
\{ ~ \7 using Laplace transformation on stiffer binder
~\ :H\
~ 0.5
~
"::l 0.4
~ 0.3
11
II
1\
' ... ....... \ ~

. -D- M. 1(H.H.) vs. M.2(L.A.) I

data. More reliable thermal stress values can be
obtained when adopting D(t) master curves as
input for Laplace transformation.
"- 0.2 ···)I(·- M.1(H.H.) VS. M.3(L.C.)
0.1 ~~ ....p,)f .. ..... M.2(L.A.) VS . M.3(L.C.) Overall, almost identical thermal stress results
0.0 were graphically observed when using different
-50 -40 -30 -20 -10 10 20 30 testing temperatures. The visual impression was
Temperature,(0 C)
confirmed by the performed statistical analysis.

t-test results (PG 58-34-P-C2)

1.0 ......... 5 CRITICAL CRACKING TEMPERATURE
0.9 .... 1-D- M.l(H.H.) vs . M.2(L.A.; I

V)
0.8
0.7
j
b
- ~

9:\~-~~~-:-~ ~:.· ~--:~.-~.-~ COMPARISON

0 0.6 Critical cracking temperature was obtained from

~ 0.5 rl. \\ !\
t the thermal stress curves in combination with the
~
"::l 0.4
~ 0.3
----------
(\j ~
\

\\
\..
~ SAP (Shenoy 2002) method. Along with simple
.::... 0.2 !!; ~ ..;1('' 0 graphical comparison (Fig. 5), statistical analysis
0.1 ~ t \'-'r!Q .~
(Tables 2 and 3) was performed with a procedure
:
0.0 similar to the one used in the previous section.
-50 -40 -30 -20 -1 0 10 20 30
Temperature,(0 C) Bold numbers identified statistically significant
differences and were rounded to the second signifi-
Figure 4. Results of t-test on thermal stress computa-
cant digit.
tion for PG 58-34 binder. Clear differences were observed in critical crack-
ing temperature among the computation methods
from which thermal stress was obtained except for
Figures 3 and 4 present the results of the statis- PG58-34 binder in RTFOT aging condition. Lower
tical analysis for both asphalt binders; the red line values of Tcr were found when Laplace transforma-
represents the significance level, α = 0.05, while M tion was used. This is most likely influenced by the
indicate the specific computation method. results of thermal stress previously obtained.

302
 Table 3. Results of t-test: Case 2 (Tcr between lowPG +
Results of Tcr ec), PG 58-28-Cl
-35 10ºC and lowPG + 10 + 6ºC).
DM.l(H.H.) I!:IM.2(L.A.) DM.3(L.C.)

-34 p-value (RTFOT) p-value (PAV)

-+- Binder
E -33
~
- PG M1v2 M1v3 M2v3 M1v2 M1v3 M23
~
~ -32 58-28 0.00 0.01 0.57 0.00 0.00 0.00
~
-31
I
I
58-34 0.29 0.23 0.80 0.03 0.01 0.99

-30~~....____,____,

RTFOT PAY

It should be remarked that, in this paper SAP

Results of Tcr (°C), PG 58-34-Cl
-35 method was used for Tcr calculation. Therefore,
DM.1(H.H.) 1BM.2(L.A.) DM.3(L.C .)
although similar results could be graphically
-34 observed for σ(T), both when using Hopkins &
G Hammings’ algorithm and Laplace transforma-
~ -33
;:I tion, different values of Tcr were found due to the
1<!
~ -32 sensitivity of the SAP method (Moon et al. 2013).
~
Nevertheless, despite different values of critical
-31
cracking temperature, were obtained from the
input thermal stress derived from Hopkins and
RTFOT PAY Hammings algorithm (1967) and Laplace trans-
formation, these two approaches can coexist and
Results of Tcr ec), PG 58-28-C2 can be potentially used for setting upper and lower
-35
DM.l(H.H.) ITJM.2(L.A.) DM.3(L.C .) limits for both σ(T) and Tcr.
-34
G -+-
0
-33 - 6 SUMMARY AND CONCLUSIONS
~
1<!
~ -32
In this paper, one numerical and one analytical
~
-31
I
I method were investigated to obtain thermal stress
and critical cracking temperature of asphalt binder,
-30~~....____,____,
RTFOT PAY starting from experimental creep compliance data.
Thermal stress and critical cracking temperature
Results of Tcr ec), PG 58-34-C2 were obtained on two asphalt binders having dif-
-35
DM.1(H.H.) 111M.2(L.A .) DM.3(L.C.) ferent performance grade and aging conditions.
-34 Based on the graphical and statistical analysis the
G following conclusions can be drawn:
~ -33
• Comparable thermal stress results were observed
~
~ -32 from the two calculation methods.
~
• In most case, the computations performed using
-31
Hopkins and Hammings’ algorithm in combina-
tion with CAM model showed higher and lower
RTFOT PAY values of thermal stress for PG 58-28 and PG
58-34 binders than the results based on Laplace
Figure 5. Comparison of Tcr (PG 58-28 & PG 58-34 transformation approaches, respectively.
binders, Case 1&2). • Statistically higher critical cracking temperature
was found when using Hopkins and Hammings’
algorithm.
Table 2. Results of t-test: Case 1 (Tcr between lowPG +
10ºC and lowPG + 10 − 6ºC). Thermal stress takes an important role in man-
aging asphalt pavement conditions especially during
p-value (RTFOT) p-value (PAV) cold weather season. Therefore, reliable estimation
Binder of thermal stress and critical cracking temperature
PG M1v2 M1v3 M2v3 M1v2 M1v3 M23 are of paramount importance for pavement agen-
cies. Hopkins and Hammings’ algorithm together
58-28 0.01 0.01 0.61 0.00 0.00 0.00
58-34 0.37 0.32 0.89 0.03 0.01 0.99
with CAM model are commonly used for research
purposes; however, this approach is rather complex

303
and may not be of convenient use for practitioners Cannone Falchetto, A. & Moon, K.H. 2015. Compari-
and road authorities on a routine basis. A method sons of analytical and approximate interconversion
based on Laplace transformation provides a sim- methods for thermal stress computation. Canadian
ple and solid alternative, since comparable thermal Journal of Civil Engineering 42(10): 705–719.
Cook, R.D. & Weisberg, S. 1999. Applied regression
stresses were obtained when using conventional including computing and graphics. St. Paul, MN: Wiley
and Laplace transformation approaches. In case of Series in Probability and Statistics.
critical cracking temperature, the results based on Ebrahimi, M.G., Saleh, M. & Gonzalez, M.A.M. 2014.
Hopkins and Hammings’ algorithm and Laplace Interconversion between viscoelastic functions using
transformation can be assumed as upper and lower the Tikhonov regularization method and its compari-
limits of the specific parameter due to the statisti- son with approximate techniques. Road Materials and
cally significant differences observed in some of Pavement Design, 15(4): 820–840.
the aging conditions for the material investigated. Farrar, M., Sui, C., Salmans, S. & Qin, Q. 2015. Determin-
Therefore, Laplace transformation, can be success- ing the low-temperature rheological properties of asphalt
binder using a Dynamic Shear Rheometer (DSR).
fully used as an alternative approach for thermal Report FP 08, Laramie: Western Research Institute.
stress computation compared to the conventional Hopkins, I.L. & Hamming, R.W. 1967. On creep and
approach. relaxation. Journal of Applied Physics, 28(8): 906–909.
Marasteanu, M.O. & Anderson, D.A. 1999. Improved
model for bitumen rheological characterization. Pre-
REFERENCES sented at European Workshop on Performance-Related
Properties for Bituminous Binders, Luxembourg,
AASHTO M320-10-UL. 2010. Standard specification Belgium.
for Performance-Graded asphalt binder. Marasteanu, M. et al. (2012). Investigation of low tem-
AASHTO R028-12-UL. 2012. Standard practice for perature cracking in asphalt pavements. National
accelerated aging of asphalt binder using a Pressu- Pooled Fund Study-Phase II 776. Minnesota Depart-
rized Aging Vessel (PAV). ment of Transportation (MN/DOT).
AASHTO T240-13-UL. 2013. Standard method of test Moon, K.H., Marasteanu, M. & Turos, M. 2013. Com-
for effect of heat and air on a moving film of asphalt parison of thermal stresses calculated from asphalt
binder (Rolling Thin-Film Oven Test). binder and asphalt mixture creep tests. Journal of
AASHTO T313. 2012. Standard method of test for Materials in Civil Engineering, 25(8): 1059–1067.
determining the flexural creep stiffness of asphalt Shenoy, A. 2002. Single-event cracking temperature of
binder using the bending beam rheometer (BBR). asphalt pavements directly from bending beam rhe-
Cannone Falchetto, A., Marasteanu, M., Balmurugan, ometer data. Journal of Transportation Engineering,
S. & Negulescu, I.I. 2014. Investigation of asphalt 128(5): 465–471.
mixture strength at low temperatures with the bend- Stehfest, H. 1970. Algorithm 368: Numerical inversion
ing beam rheometer. Road Materials and Pavement of Laplace transform. Communication of the ACM,
Design, 15(sup1): 28–44. 13(1): 47–49.

304
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Laboratory investigation of physical performance and FT-IR analysis

of warm mix asphalt binders

Jingyun Chen & Weiying Wang

School of Transportation and Logistics, Dalian University of Technology, Dalian, China

Bin Xu
Research Institute of Highway Ministry of Transport, Haidian District, Beijing, P.R. China

ABSTRACT: Warm Mix Asphalt (WMA) technologies have become increasingly popular in recent
years around the world due to their environmental benefits and advantages in energy conservation. In the
present study, a laboratory investigation of the performance and chemical compositions was conducted
on three WMA binders mixed with M-1, WB-2 and WB-3, respectively. The testing procedures including
penetration tests, softening point tests, Rotary Viscosity (RV) tests and Fourier Transform Infrared Spec-
troscopy (FT-IR) analysis were performed to determine the influences of the three additives on asphalt
binders. The test results indicated that the WMA additives had a significant influence on the physical per-
formance of asphalt, such as penetration, softening point and rotary viscosity. Also, the rotary viscosity
was not the single index to determine the mixing and compaction temperature of WMA binders since the
WB-3 increased the viscosity. The chemical investigation conducted using FT-IR analysis showed that the
effect of M-1, WB-2 and WB-3 on the asphalt may be mainly physical not chemical modification, there
was no chemical bond cracking or recombination found in the FT-IR spectra. The change of the density
of asphalt component may be the most effect of WMA additives on the base asphalt. By analyzing the
integral spectral peak area calculation, the change trend between the area ratio of the aromatics base
absorption peak to rotary viscosity was found the same, which could be due to the fact that the aromatics
has impact on the rotary viscosity directly in the three WMAs.

Keywords: warm mix asphalt, warm mix additive, physical performance tests, Fourier transform infrared
spectroscopy

1 INTRODUCTION viscometer, the Dynamic Rheometer (DSR) and

the Bending Beam Rheometer (BBR). The results
Warm Mix Asphalt (WMA) technologies have indicated that the additives and aged binders
become more and more popular in recent years played an important role in determination of the
around the world due to their advantages in the binder properties (Lee et al., 2009).
energy conservation and environmental protec- Zahid et al. used spectroscopy techniques to
tion. WMA can lower the mixing and compac- evaluate the effects the asphalt (WMA) additives
tion temperature at approximately 28°C or more on the chemical compositions of a performance
in comparison of Hot Mix Asphalt (HMA). grade binder (PG 64-22) including the FT-IR.
Lower production temperatures can potentially Through the analysis of the FT-IR data, a relation-
improve pavement performance by reducing ship was built between the peaks and stiffness of
binder aging, providing additional time for mix- WMA binders (Hossain et al., 2013).
ture compaction, and allowing improved com- Xiao et al. conducted a laboratory investiga-
paction during cold weather paving (Ghabchi tion of effects of non-foaming WMA additives
et al., 2013). on asphalt binders at intermediate, low and high
Many methods have been applied to investigate performance temperatures using Fourier Trans-
the mechanics performance and chemical charac- form Infrared Spectroscopy (FT-IR). The FT-IR
teristics of WMA. tests illustrated that the binder type and source
Soon-Jan et al. addressed the laboratory investi- play important roles in determining the rheologi-
gation of performance properties of WMA bind- cal properties of WMA binders (Xiao et al., 2012,
ers containing aged binders using the rotational Xiao et al., 2013).

305
 However, there were little studies to explore the and WB-2 were deep brown sticky liquid surfactants
mechanism of WMA binders using FT-IR and while the WB-3 was a kind of white solid powder,
build a relationship between the peaks of FT-IR commonly called manufactured synthetic zeolite. All
and physical performance. Thus, conducting a pri- the three WMA additives were respectively blended
mary research of WMA binders using some fun- with the control asphalt binder. The additive con-
damental testing methods and a complete FT-IR tents by weight of binder were listed in Table 3.
test to study the mechanism of WMA binders are
necessary and beneficial for further research.
The main objectives of this study are to character- 2.2 Penetration test
ize the performance and chemical compositions of According to ASTM D5, a penetrometer is used
WMA binders based on three WMA additives, M-1, to conduct the penetration test, which a standard
WB-2 and WB-3. The testing procedures including needle is injected into the asphalt sample under
penetration tests, softening point tests, Rotary Vis- specific conditions. It is used as a measure of con-
cosity (RV) tests and Fourier Transform Infrared sistency. Higher values of penetration indicate
Spectroscopy (FT-IR) are adopted to determine the softer consistency (ASTM, 2013). It provides an
influences of the three additives on asphalt binders. index that divide the asphalt number in China. The
test uses a penetration temperature of 25°C under
a load of 100 g applied for 5 s.
2 MATERIALS AND EXPERIMENTS

2.1 Materials 2.3 Softening point test

One type of asphalt binder commonly used in As a kind of viscoelastic material, asphalt
northeast China, whose penetration value was 91.3 gradually become softer and less viscous as the
(0.1 mm) at 25°C, was considered as base asphalt. temperature rises. This test method uses a ring-
Three WMA additives, M-1, WB-2 and WB-3, and-ball apparatus immersed in distilled water
which can lower the mixing and compaction tem- (30 to 80°C) or USP glycerin (above 80 to 157°C)
perature by 20 to 50°C, were utilized in this study. to determine the softening point of asphalt in
The chemical compositions and physical properties accordance with the ASTM D36 (ASTM, 2014).
of the base asphalt binder were respectively shown Increase the temperature by 5°C/min, until the
on the Tables 1 and 2. The former two additives, M-1 ball touches the bottom as the asphalt is being
softened, then the recorded temperature is sof-
tening point.
Table 1. Chemical composition of base asphalt (%).

Saturates Aromatics Resin Asphaltene

2.4 Rotary Viscosity (RV) test
28.64 57.74 11.36 2.26 The viscosity is considered to be an important
property since it is associated with the mixing and
Table 2. Physical properties of base asphalt. compaction temperature of asphalt mixture. Fur-
thermore, the viscosity relates to the workability of
Penetration Softening Ductility RV asphalt binders. In this study, the RV is performed
(25°)/0.1 mm point/°C (15°C)/cm (135(C)/Pa⋅s to describe the influence of the WMA additives
on asphalt binders in accordance with the ASTM
91.3 48.34 >100 0.256 D4402 test method (ASTM, 2015).

Table 3. Physical properties of WMA binders.

Type & Percentage (%) Binder code Penetration (25ºC)/0.1 mm Softening point/ºC RV (135ºC)/Pa⋅s

M-1 0.1 A 88.0 43.82 0.248

0.3 B 85.7 44.65 0.235
0.5 C 81.0 46.60 0.231
WB-2 1 D 93.0 43.60 0.223
3 E 118.0 41.93 0.190
5 F 166.0 38.28 0.158
WB-3 5 G 85.0 42.52 0.261
7 H 79.0 43.53 0.266
9 I 77.3 45.83 0.275

306
2.5 FT-IR test results of the viscosities that the addition of M-1
and WB-2 exhibits a decreasing tendency com-
Fourier Transform Infrared Spectroscopy (FT-IR)
pared to the base asphalt, and the WB-2 binders
is a test to determine the functional groups of mate-
show lower viscosity than M-1 binders. However,
rials on the basis of the principle that the covalent
the addition of WB-3 shows an increase of the vis-
bonds in molecules absorbing the infrared spec-
cosity, which is higher than the base asphalt.
tra at specific frequencies is distinctive (Xiaoyang
The results of the physical properties of WMA
et al., 2014). It is commonly used to study the
binders indicate that the change of the penetration
molecular structure of asphalt binder and addi-
and the softening point is completely opposite. But
tives by interpreting the infrared absorption spec-
neither the penetration nor the softening point
trum, the covalent bonds in molecules (Gao et al.,
does have an obviously relation with the viscosity
2013, Hossain et al., 2013). The asphalt materials
to the three kinds of WMA binders. The M-1 and
can use the FT-IR because the FT-IR spectra of
WB-2 are effective in reducing the viscosity values,
pure compounds are generally different from the
and the WB-2 is more effective. And the rotary vis-
others (Xiao et al., 2013). In this study, the FT-IR
cosity is not the single index to determine the mix-
is performed to explain the mechanism of WMA
ing and compaction temperature of WMA binders
as a microcosmic test method, utilizing Nicolet
since the WB-3 increases the viscosity. Thus it is
6700 Flex, whose resolution is 4 cm–1, scanning fre-
necessary to use some other approaches to deter-
quency is 32 Hz and test range is 350 to 7800 cm–1.
mine the temperature, such as the volume param-
The samples were prepared by casting a film onto
eters of the mixtures.
a potassium bromide (KBr) window from a 5% by
weight solution in carbon tetrachloride (CCl4).
3.2 FT-IR results
The FT-IR analysis of WMA binders is shown
3 RESULTS AND DISCUSSIONS
in Figure 2 (a–c). It could be observed that along
the spectral bands, the peaks around 2923 cm-1,
3.1 Performance test of WMA binders
2853 cm-1, 1604 cm-1, 1460 cm-1, 1376 cm-1,
Table 3 shows the performance test results of all 872 cm-1 and 787 cm-1 are obvious. Peaks at
the asphalt binders. 2923 cm-1 and 2853 cm-1 represent the aliphatic
The results clearly demonstrate that compared CH2 stretching of alkane with strong intensity.
to the base asphalt, the addition of M-1 and WB-3 And both the two peaks have a shoulder at 2955
decreases the penetration of the binders, and with and 2870 cm-1, which are CH3 stretching. The
the increase of dosage, the value decreases. On shoulders appear because of CH2 stretching and
the contrary, the value of the binders with WB-2 CH3 stretching overlapping with each other, which
increases. All results of the softening point decrease indicates the content of CH2 was more than that
comparing to the base asphalt, and are opposite to of CH3. The peak at 1604 cm–1 is the absorbance
the change of penetration. of C = C stretching in benzene ring and at around
The rotary viscosity at high temperature is con- 1460 cm–1 could indicate bending of CH2 and CH3.
sidered to be one of important properties as it is The CH3 symmetric stretching is represented at
generally related to the mixing and compaction 1376 cm-1 (Chen, 2012). The “fingerprint” region
temperature. Figure 1 presents the experimental (wave number ≤ 1,300 cm-1) is useful to iden-
tify the type of compound, but the bands in this
region commence in interacting vibrational modes
300
resulting in a complex absorption pattern, which
is quite complex and often difficult to interpret
250
(Hossain et al., 2013). In the “fingerprint” region,
substitutes on Benzene ranged from 650 to 910 cm-1.
200
All the peaks indicate that asphalt is a kind
of mixture, which is composed by saturated
150
long-chain alkanes, aromatic hydrocarbon and
long-chain hydrocarbon, whose structure is very
100
complex.
50
It could be observed that all the test results of
WMA binders are so similar to the base asphalt
0
that there is no new functional group produced
A B c D E F G H I Base with the addition of the three kinds of WMA
agent. Hence, the effect of M-1, WB-2 and WB-3
Figure 1. Rotary viscosity of WMA binders. on the asphalt may be mainly physical modified not

307
 100 100

80
ao
*1l
l\i
ii!l 60
60
E
!ilc ~
[3. 40

~ 40 I M-1 1
20

20
0
4500 4000 3500 3000 2500 2000 1500 1000 500
Wavenumber/em·'
4000 3000 2000 1000
Wavenumber/em·'
(a)
(a)
100
100

60

*1i
60
*"g
it "'
.!:l
60
60
~
t ,_~ 40
,_~ 40

20
20

0
4500 4000 3500 3000 2500 2000 1500 1000 500
4000 3000 2000 1000 Vl/avenumber/cm·'
\Navenumber/cm 1

(b)
(b)
100

100
80
*
*
ao
"~ 60
"~
E
60
E

~
~ ~ 40

,_I" 40
1-'Ml-31
20

20

4500 4000 3500 3000 2500 2000 1500 1000 500

4000 3000 2000 1000

wavenumber/em·,
wavenumber/em·, (c)
(c)
Figure 3. FT-IR analysis of the WMA additives:
Figure 2. FT-IR analysis of the WMA binders: (a) M-1, (b) WB-2, and (c) WB-3.
(a) M-1, (b) WB-2, and (c) WB-3.

chemical modified. Although there is no new func- infrared spectra of aromatic hydrocarbon con-
tional group produced, some changes appeared at tains C = C and C-H that is respectively at 1,380
the peaks absorbance. So the FT-IR spectra of the to 1600 cm−1 and 650 to 900 cm−1. So M-1 and
three WMA additives is shown in the Figure 3 (a-c) WB-2 could be composed of saturates and aro-
to explain the reason. matics, yet WB-3 contains aromatics most. The
It can be observed that the infrared spectra of absorbance range of hydroxide radical (OH-) is
M-1 and WB-2 is similar, and is very different from 3,200 to 3,650 cm−1. It can form the hydrogen bond
that of WB-3. The broad peaks between 2,850 and between molecules or in a molecule. When in the
3000 cm−1 relate to the aliphatic C-H stretching case of between molecules, there would be crystal
(CH3, CH2, and CH) of alkane with strong inten- water because of O-H stretching in the three kinds
sity. The peaks of C-C stretching are around 800 of WMA agents. Therefore, with the addition of
to 1,250 cm−1, and C = C is 900 to 1000 cm−1. The the three kinds of WMA additives, there was no

308
chemical bonds cracking or recombination. Chang- 60
ing the density of asphalt component may be the
50
t • Base asphalt • D • E • F
most effect of WMA agents on the base asphalt.
Then, something extra, but necessary analysis is 40
done to describe the absorbance change of differ-
ent chemical group. 30

20
3.3 The area analysis of absorption peaks
10
As the results of absorbance or transmittance will
be effected by the variation of asphalt film thick- 0
ness between samples (Chunfa et al., 2006), it needs Al/W A2/W A3/ W
to calculate the area ratio of intensity of the peaks
of functional groups to the whole FT-IR spectra. Figure 5. The result of area analysis for WB-2
The calculation is accordance with the following additive.
equations:
60
A
Ratio2848,2930 = 1 (1) 50
whole
l
W
40
A
Ratio 1604 = 2 (2) 30
whole
l
W
20
A3
Ratio690,810 = (3)
whole
l
W 10

0
where A1 = the Area of methylene stretching absorp- A l/W A2/ W A3/ W
tion peak around 2930 cm–1 and 2848 cm–1; A2 = the
Area of aromatics base absorption peak around Figure 6. The result of area analysis for WB-3
1604 cm−1; A3 = the Area of benzene substituents additive.
absorption peak around 810 cm−1 to 690 cm−1; and
W = the Area of the whole FT-IR spectra. the area ratio of benzene substituents absorption
The results were as following Figure 4 to 6. peak (A3/W) is the biggest when the middle density.
As shown in the Figure 4 to 6, it cannot be It is notable that the change of the area ratio of the
found a regular trend between the base asphalt and aromatics base absorption peak (A2/W) is related
WMA binders. It could be explained that some- with the change of rotary viscosity. The change
thing changed of density of the chemical com- trend between the area ratio of the aromatics base
ponent with the addition of WMA additives, but absorption peak (A2/W) and rotary viscosity is the
there is no new functional group produced. Except same. So it can be explained that the aromatics has
the base asphalt, similar trend is shown in the influence on the rotary viscosity directly.
Figure 4 to 6. It can be noticed that the area ratio of However, because of the limit of the test data, it is
methylene stretching absorption peak (A1/W) is the necessary to conduct more tests to prove the relation-
smallest when the middle density. On the contrary, ship between the aromatics and the rotary viscosity.
And continuous tests need to be done to research
60 that what chemical reactions had been produced.
• Base asphalt • A • B •C
50

40
4 CONCLUSIONS

30 Based on the test results from the three types of

WMA and one base asphalt through a series of
20 performance tests and FT-IR, the following con-
10 clusions can be drawn:

0
1. The WMA additives had different effect on
Al!W A2/W A3/W the penetration, softening point, and rotary
viscosity. Like base asphalt, the penetration
Figure 4. The result of area analysis for M-1 additive. and softening point of WMA binders were

309
 opposite. But either the penetration or the sof- ASTM. (2015) Standard Test Method for Viscosity
tening point did not have an obviously related Determination of Asphalt at Elevated Temperatures
to the viscosity. Using a Rotational Viscometer. D4402, West Consho-
2. The rotary viscosity was not the single index to hocken, PA.
Chen, J. & Qiu, L. (2012) Analysis of the Mechanism of
determine the mixing and compaction tempera- Aging and Regeneration of SBS Modified Asphalt
ture of WMA binders since the WB-3 increases Based on the Infrared Spectrum. Journal of Shenyang
the viscosity. Thus it was necessary to use some Jianzhu University (Natural Science), 28, 859–864.
other approaches to determine the temperature, Chunfa, O. Y., Wang, S.F., Zhang, Y. & Zhang, Y.X.
such as the volume parameter of the mixtures. (2006) Improving the aging resistance of asphalt by
3. The effect of M-1, WB-2 and WB-3 on the addition of Zinc dialkyldithiophosphate. Fuel, 85,
asphalt could be mainly physical modified not 1060–1066.
chemical modified. But some changes appeared Gao, Y., Gu, F. & Zhao, Y. (2013) Thermal oxidative
at the intensity of the peaks absorbance. aging characterization of SBS modified asphalt. Jour-
nal of Wuhan University of Technology-Materials Sci-
4. Changing the density of asphalt component ence Edition, 28, 88–91.
may be the most effect of WMA additives on Ghabchi, R., Singh, D., Zaman, M. & Tian, Q. (2013)
the base asphalt. Application of asphalt-aggregates interfacial energies
5. By analyzing the integral spectral peak area cal- to evaluate moisture-induced damage of warm mix
culation, it showed that except the base asphalt, asphalt. In Chakroborty, P., Reddy, H.K., Amekudzi,
there was similar trend produced among the A., Das, A., Seyfried, A., Maitra, B., Teodorovic,
three WMAs. It could be noticed that the area D., Srinivasan, K.K., Devi, L., Pulugurtha, S.S. &
ratio of methylene stretching absorption peak Mathew, T.V. (Eds.) 2nd Conference of Transportation
(A1/W) was the smallest when the middle den- Research Group of India.
Hossain, Z., Lewis, S., Zaman, M., Buddhala, A. &
sity. On the contrary, the area ratio of benzene O’rear, E. (2013) Evaluation for Warm-Mix Additive-
substituents absorption peak (A3/W) was the Modified Asphalt Binders Using Spectroscopy Tech-
biggest when the middle density. niques. Journal of Materials in Civil Engineering, 25,
6. The change of the area ratio of the aromatics 149–159.
base absorption peak (A2/W) was related with Lee, S.-J., Amirkhanian, S.N., Park, N.-W. & Kim, K.W.
the change of rotary viscosity. The change trend (2009) Characterization of warm mix asphalt binders
between the area ratio of the aromatics base containing artificially long-term aged binders. Construc-
absorption peak (A2/W) and rotary viscosity tion and Building Materials, 23, 2371–2379.
was the same. So it could be explained that the Xiao, F., Punith, V.S. & Amirkhanian, S.N. (2012) Effects
of non-foaming WMA additives on asphalt binders at
aromatics has influence on the rotary viscosity high performance temperatures. Fuel, 94, 144–155.
directly in the three WMAs. Xiao, F., Punith, V.S., Amirkhanian, S.N. & Putman,
B.J. (2013) Rheological and chemical characteristics
of warm mix asphalt binders at intermediate and low
REFERENCES performance temperatures. Canadian Journal of Civil
Engineering, 40, 861–868.
ASTM. (2013) Standard Test Method for Penetration of Xiaoyang, J., Baoshan, H., Bowers, B.F. & Sheng, Z. (2014)
Bituminous Materials. D5, West Conshohocken, PA. Infrared spectra and rheological properties of asphalt
ASTM. (2014) Standard Test Method for Softening cement containing waste engine oil residues. Construction
Point of Bitumen (Ring-and-Ball Apparatus). D36, and Building Materials, 50, 683–691.
West Conshohocken, PA.

310
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Study of asphalt binder diffusion using atomic force microscopy

Abhishek Savarnya, Ambika Kuity, Shivam Gupta, Akash Chandra & Animesh Das
Department of Civil Engineering, IIT Kanpur, Kanpur, India

ABSTRACT: During asphalt recycling process, diffusion of asphalt binder takes place between the aged
binder adhered to the RAP aggregates and the virgin binder. Complete mixing between the two binders
may not necessarily happen during the recycling process. Study of diffusion is, therefore, an important
consideration in asphalt recycling. Researchers have experimentally studied the phenomenon of diffu-
sion typically by using Dynamic Shear Rheometer (DSR) or Fourier Transform Infrared Spectroscopy
(FTIR). In the present work, possibility of using Atomic Force Microscopy (AFM) to study binder diffu-
sion is explored. In this study, two different binders is used in a circular geometry, so as to closely resemble
an aggregate (from RAP) surrounded by a pool of virgin binder. AFM images are taken along the radial
direction of the sample to study the progression of diffusion at different locations. The relative concentra-
tion of binders at these locations are estimated from the AFM images. From the present study, based on
limited data, it appears that AFM can be used for studying diffusion between asphalt binders.

1 INTRODUCTION • Virgin binder

lh] Aged binder
Asphalt recycling is done by mixing suitable propor-
tions of virgin binder, Reclaimed Asphalt Pavement
(RAP) materials and virgin aggregates. For mix (a) Parallel geometry
design purpose, it is generally assumed that perfect
mixing (Carpenter & Wolosick 1980, Kennedy et al.
1993, Kandhal & Mallick 1997, Austoroads 2000,
McDaniel & Anderson 2001, ASTM D4887 2003,
Aravind & Das 2007) between the binders (that is,
RAP binder and virgin binder) takes place during
the recycling process but it may not necessarily be
(b) Circular geometry (c) RAP aghrregate in a pool
the case (Zhao et al. 2014, Karlsson & Isacsson
of virgin binder
2003a, Chaffin et al. 1995). In such a case, one
needs to study the diffusion between RAP binder
and virgin binder so as to have a better estimate of Figure 1. Parallel and circular geometry scenarios.
the extent of mixing at any given point of time dur- shows characteristic ‘bee-structures’, using
ing the recycling process. which different types of asphalt binders and their
In most of the previous diffusion studies on compositions can be identified (Aguiar-Moya
asphalt binder, parallel geometry (Carpenter & et al. 2015, Karlsson & Isacsson 2003b, Jäger
Wolosick 1980, Karlsson et al. 2007) has been con- et al. 2004). These ‘bee-structures’ also has been
sidered (refer Figure 1(a)). In the present study, successfully used in the past to study ageing (Wu
a concentric circular geometry is chosen (refer et al. 2009, Zhang et al. 2013) and moisture dam-
Figure 1(b)). age (Tarefder & Arifuzzaman 2010, Santos et al.
It may be argued that such a geometry closely 2014) in asphalt binder. Since, AFM can iden-
resembles the arrangement of RAP and virgin tify different grades of asphalt binders (Jäger
binder in the recycled mix (refer Figure 1(c)). Thus, et al. 2004) and diffusion is a process where mix-
using this geometry for diffusion between two dif- ing of the binders take place, it may be argued
ferent asphalt binders, the present paper attempts that AFM may as well be used for studying the
to explore the following two aspects: progression of the diffusion process. This consti-
1. Use of Atomic Force Microscopy (AFM) tutes the first objective of the present study.
to study the asphalt binder diffusion process: 2. Estimation of coefficient of diffusion from
Asphalt binder sample studied under AFM AFM study: If the first objective of the study

311
 (as above) is fulfilled, it may further be possible to
obtain a quantitative information on the extent Bi1U!III'B
of diffusion at different position (and time) of
the sample under study. These results can then
be matched with the theoretical model of the
diffusion process to estimate the coefficient of
diffusion. Thus, the steps involved in the second
part of the study can be envisaged as follows:
a. Concentric circles are made with two differ-
ent asphalt binders (say, binder A and binder
B) and is allowed to diffuse into each another
at a fixed temperature.
b. After a pre-specified time, the temperature is
Figure 2. Diffusion in circular geometry.
brought down to a low level. Diffusion at this
temperature is assumed to be slow or virtu-
ally stopped. (binder A and binder B, in this case), is shown in
c. The AFM study is conducted on these sam- Figure 2. To study diffusion it is assumed that ini-
ples to estimate the relative concentration of tially the concentration of the binder A is A0 up
asphalt binder along radial direction. to radius R1 from center, while concentration of
d. Count of ‘bee structure’ as an estimate of the binder B in that portion is zero (refer Figure 2).
relative concentration is observed through The initial concentration of binder A beyond R1 is
AFM. The relative concentration due to the also assumed to be zero. Hence, the initial condi-
diffusion process is calculated from the cali- tions can be written as (Caretto 2009):
bration developed a-priori with AFM study
on separate samples prepared with known ⎧ A0 , 0 R1
relative concentration. C (r,0) = ⎨ (4)
e. The experimental and theoretical variation ⎩0, R1 (R2 − R1 )
of concentration with radial distance can
be compared to estimate the co-efficient of Assuming no change in concentration rate at
diffusion. the periphery, it can be written (Caretto 2009) as:
These are discussed in the subsequent
∂C
paragraphs. =0 (5)
∂t ( )

2 THEORETICAL BACKGROUND Hence (Caretto 2009),

Fick’s second law on diffusion (Crank 1980, ∞
2 A 0 J 0 ( m r ) − λ 2Dt
Caretto 2009) is given as: C ( r, t ) = ∑ e m (6)
m =1 m R J1 ( m R )
λ
∂C
= ∇ 2C (1) The Equation 2 is the solution for the diffusion
∂t
equation for circular geometry (Caretto 2009). It
where C is the concentration and D is the coeffi- gives the concentration of binder A, at any time at
cient of diffusion. In cylindrical co-ordinate sys- a point along the radial distance r varying between
tem, the equation can be written as (Crank 1980): 0 to R1. This formulation subsequently is used to
estimate the coefficient of diffusion presented in
∂ C D ∂ ⎡ r∂ C ⎤ the section titled ‘Analysis of results’.
= (2)
∂t r ∂r ⎢⎣ ∂r ⎥⎦
3 EXPERIMENTAL STUDY
Solving the above equation using separation of
variable, Experiments are carried out as per the steps inden-
tified in the section titled ‘Introduction’. The
C (r, t ) Ce λ2D
J0 ( r) k (3) physical properties of binder A and binder B are
mentioned in Table 1. Sample preparation com-
where J0 is the Bessel function of zeroth order prised of making a concentric ring of binder (say,
while λ and k are constants. The idealized circular binder B) around a drop of the other binder (say,
geometry involving two different diffusing liquids binder A) on a glass slide as shown in Figure 3(a).

312
Table 1. Physical properties of the binders.

Binder A Binder B Specifications

Penetration 55 89 ASTM D5
(25°C, × 0.1 mm)
Softening point 49.8 46.4 ASTM D36
(°C)
Ductility (cm) > 100 82 ASTM D113
Absolute viscosity 262.7 73.8 ASTM D2171
(60°C, Pa s) Figure 5. AFM phase images along radial line shows
variation of ‘bee structures’.

Table 2. Observed number of “bee structures” for

known proportion of binders.

Percentage
heated for I 0 minutes No. of replicate Average no. of
(b) Diffusion starts when
Binder A Binder B samples tested ‘bee structures’
(a) Binder A and Binder B
not in contact Binder A and Binder B
100 0 5 262
come in contact
90 10 5 204
80 20 4 147
50 50 4 131
10 90 3 68
stored at low 0 100 4 62
rempernntre

(c) Sample is taken out (d) Diffusion is assumed

from oven to have stopped
probe tip of the AFM (refer Figure 4) intermit-
tently touches the specimen. This mode is pre-
Figure 3. Sample preparation for AFM study. ferred for soft and sticky material, as asphalt
binder (Gracia 2002). The probe tip used in the
present study has a force constant of 5 N/m. Scan
BinderB area of 40 μm × 40 μm and scan rate of 0.5 Hz are
Binder A
selected.
AFM study is conducted at various loca-
Rt ----- --- tions along the radial direction (after diffusion is
~--
d .... ...... ___ ~M specimen assumed to have stopped), as schematically rep-
resented by Figure 4. As already mentioned, the
number of ‘bee structures’ in the phase images
captured by AFM, is considered to be the mor-
phological feature indicative of progression of
Figure 4. AFM study on the sample. diffusion. This is illustrated in Figure 5 through
a sample image where one can see the change
The outer diameter of the binder ring is taken as in number of ‘bee structures’ along the radial
approximately 14 mm in the present study. distance.
A petri-dish containing the slide is kept in To obtain the relative concentration of binder A
the convection oven at 110°C allowing the bind- and binder B, another set of samples are prepared
ers to flow until they come in contact with one where binders are blended in different known pro-
another. Diffusion is said to have started from portions. The average number of ‘bee structures’
the time when the binders come in contact (refer for these samples (that is, blended samples with
Figure 3(b)). The sample is taken out of the known proportions) are presented in Table 2. The
oven after 10 minutes of diffusion process (refer AFM test data on these blended binder samples (of
Figure 3(c)). Thereafter, the samples are allowed known proportions) is used to develop the calibra-
to cool down for 5 minutes before storing them tion chart for estimation of the binder concentra-
at low temperature, assuming diffusion to have tion (which is unknown) during binder diffusion
stopped (refer Figure 3(d)). test. The calibration chart is presented as Figure 6
In the present study, tapping mode is chosen (the average values of which are already presented
for the AFM study. In this mode, the cantilever in Table 2).

313
4 ANALYSIS OF RESULTS et al. 2007, Zhang & Greenfield 2007). Consider-
ing the fact that rejuvenators are softer than the
The number of ‘bee structures’ are estimated penetration grade binder, the values obtained in
from the images acquired by the AFM at various the present study (from limited experimental data),
radial distances (refer Table 3). Data could not be seem to be significantly high. Thus, further studies
acquired at certain locations due to high curvature are needed to gain confidence on the data and the
of the sample. value of the coefficient of diffusion estimated.
Using the calibration chart developed a-priori
(refer Figure 6) proportions between the binders
at different radial distances are estimated. It may 5 CONCLUSIONS AND DISCUSSIONS
be noted that the radial distance mentioned here
is somewhat approximate, because it is found to be Use of Atomic Force Microscopy (AFM) to study
difficult to identify the positions under the micro- the diffusion of asphalt binder has been explored
scope. Further studies are ongoing to improvise in the present study, using a concentric circle
the experimental process. geometry. It is argued that such concentric circle
The experimentally observed curve is superim- geometry, resembles the RAP aggregate in a pool
posed on the theoretical curves (refer Equation 6) of virgin binder better than conventional paral-
to iteratively estimate the coefficient of diffusion lel geometry used in the past studies. Preliminary
between binder A and binder B. The value of coef- study, based on the limited data, suggests that
ficient of diffusion for the present case is obtained AFM can be used to observe the progression of
as 5.7 × 10−3 mm2/s for the temperature (110°C in the diffusion and to estimate coefficient of diffusion
present case). Although research results are availa- between asphalt binder samples. Information
ble on estimation of coefficient of diffusion for reju- regarding topography, phase shift, amplitude error
venators (for example, softer binder, components of of the oscillating cantilever etc. could possibly be
binder etc.), results on bitumen seems to be scanty used to interpret the diffusion process.
(Cong et al. 2016, Ding et al. 2016, Ekblad & Lund-
ström 2017, Karlsson & Isacsson 2003b, Karlsson
REFERENCES

Table 3. Number of ‘bee structures’ along radial Aguiar-Moya, J. P, Salazar-Delgado, J. Bonilla-Mora, V.,
distance$. Rodríguez-Castro, E., Leiva-Villacorta, F., and Loría-
Salazar, L. 2015. Morphological analysis of bitumen
0 mm 2 mm 3 mm 4 mm 6 mm phases using atomic force microscopy, Road Materials
and Pavement Design, 16:sup1, 138–152.
388 – 295 – 115 Aravind, K. and Das, A. 2007. Preliminary constituent
167 169 182 – – proportioning for central plant hotmix asphalt recy-
241 – – 205 – cling. Journal of Materials in Civil Engineering, 19(9):
247 146 143 80 101 740.
89 – 80 95 85 ASTM-D4887-99 2003. Standard practice for prepara-
tion of viscosity blends for hot recycled bituminous
100 – 110 99 99
materials. ASTM International, West Conshohocken.
130 – 94 80 96
ASTM D113-99 1999. Standard test method for ductility
of bituminous materials, ASTM International, West
Note:
Conshohocken.
– Data could not be acquired.
$ ASTM D2171/D2171M-10 2010. Standard test method
The measure of radial distance is approximate.
for viscosity of asphalts by vacuum capillary visco-
meter, ASTM International, West Conshohocken.
400 . - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - , ASTM D36/D36M-14e1 2014. Standard test method for
350 softening point of bitumen (Ring-and-Ball appara-
~ 300 tus), ASTM International, West Conshohocken.
~ 250 ASTM D5/D5M-13 2013. Standard test method for
~ 200 ....... • .... ····· penetration of bituminous materials, ASTM Interna-
e
0
150
.:............... ····· ·······I • tional, West Conshohocken.
Austroads 2000. Framework specifications for asphalt recy-
~ 100
50
.... ....... cling. Publication No. AP-T02/00, Sydney, Australia.
Caretto, L. 2009. Notes on engineering analysis—solution
of diffusion equation. www.csun.edu/~lcaretto/
20 40 60 80 100
me501b/laplace.doc, California State University,
Binder A(%)
Northridge, Last accessed, February 17, 2017.
Carpenter, S. H. and Wolosick, J. R. 1980. Modifier
Figure 6. Calibration chart for estimation of concen- influence in the characterization of hot-mix recycled
tration of binder A during binder diffusion test. material. Transportation Research Record, 777: 15–22.

314
Chaffin, J. M., Davison, R. R., Glover, C. J., and Bullin, Kennedy, T. W., Moulthrop, J. S., and Huber, G. A. 1993.
A. J. 1995. Viscosity mixing rules for asphalt recycling. Development of SHRP specification and design and
Transportation Research Record, 1501: 75–87. analysis system. Journal of the Association of. Asphalt
Cong, P., Hao, H., Zhang, Y., Luo, W., and Yao, D. Paving Technologists, (62): 508–528.
2016. Investigation of diffusion of rejuvenator in aged McDaniel, R., and Anderson, R.M. 2001. Recom-
asphalt. International Journal of Pavement Research mended use of reclaimed asphalt pavement in the
and Technology, 9(4): 282–288. Superpave mix design method: Technician’s Manual.
Crank, J. 1980. The mathematics of diffusion. Oxford NCHRP Rep. No. 452, TRB.
University Press. Santos, S., Partl, M. N. and Poulikakos, L. D. 2014.
Ding, Y., Huang, B., Shu, X., Zhang, Y., and Woods, Newly observed effects of water on the microstruc-
M. E. 2016. Use of molecular dynamics to investigate tures of bitumen surface. Construction and Building
diffusion between virgin and aged asphalt binders. Materials, 71: 618–627.
Fuel, 174: 267–273. Tarefder, R.A. and Zaman, A. M. 2010. Nanoscale
Ekblad, J., and Lundström, R. 2017, Soft bitumen asphalt evaluation of moisture damage in polymer modified
produced using RAP, Materials and Structures, 50:13. asphalts. Journal of Materials in Civil Engineering,
García R, and Rubén, P. 2002. Dynamic atomic force 22(7): 714–725.
microscopy methods. Surface Science Reports, 47(6– Wu, S., Pang, L., Mo, L., Chen, Y. and Zhu, G. 2009.
7): 197–301. Influence of aging on the evolution of structure,
Jäger, A., Lackner, R., Eisenmenger-Sittner, C. and Blab, morphology and rheology of base and SBS modified
R. 2004. Identification of four material phases in bitu- bitumen. Construction and Building Materials, 23(2):
men by atomic force microscopy. Road Materials and 1005–1010.
Pavement Design, 5(1): 9–24. Zhang, H., Xu, H., Wang, X. and Yu, J. 2013. Micro-
Kandhal, P.S., and Mallick, R.B. 1997. Pavement recy- structures and thermal aging mechanism of expanded
cling guidelines for state and local governments. Pub- vermiculite modified bitumen. Construction and
lication No. FHWA-SA-98-042, FHWA, National Building Materials, 47: 919–926.
Center for Asphalt Technology. Zhang, L., and Greenfield, M. L. 2007. Relaxation time,
Karlsson, R., Isacsson, U. and Ekblad, J. J. 2007. Rheo- diffusion and viscosity analysis of model asphalt
logical characteristation of bitumen diffusion. Journal systems using molecular simulation, The Journal of
of Material Science, 42(1): 101–108. Chemical Physics, 127.
Karlsson, R. and Isacsson, U. 2003a. Application of Zhao, S., Bowers, B., Huang, B., and Shu, X. 2014. Char-
FTIR-ATR to characterization of bitumen rejuvena- acterizing rheological properties of binder and blend-
tor diffusion. Journal of Materials in Civil Engineer- ing efficiency of asphalt paving mixtures containing
ing, 15(2): 157–165. RAS through GPC. Journal of Materials in Civil Engi-
Karlsson, R. and Isacsson, U. 2003b. Laboratory stud- neering, 26(5): 941–946.
ies of diffusion in bitumen using markers. Journal of
Materials Science, 38(13): 2835–2844.

315
Performance of asphalt mixes with additives
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Performance properties of asphalt mixture containing Linz-Donawitz

(LD) steel slag

J. Groenniger & M.P. Wistuba

Department of Civil Engineering, Braunschweig Pavement Engineering Centre (ISBS), Technische Universität
Braunschweig, Braunschweig, Germany

ABSTRACT: Standard asphalt mixtures for road infrastructures consist of natural aggregate and bitu-
men. A number of research efforts have successfully investigated the possibility of replacing the con-
ventional aggregate skeleton with industrial by-products such as slag originating from steel production
process. However, little is known on the effect of steel slag on the mixtures performance properties such
as resistance to low temperature cracking and to permanent deformation, stiffness and fatigue. In the
paper the results of a comprehensive investigation on the fundamental performance properties of differ-
ent types of asphalt mixtures prepared with 100% LD slag aggregate and a conventional asphalt mixture
containing natural Gabbro aggregate are presented. Sophisticated testing methods were used to evaluate
the key performance parameters for the set of asphalt mixtures investigated. In this study, low tempera-
ture cracking was addressed through Thermal Stress Restrained Specimen Tests. Penetration Tests and
Cyclic Compression Tests were used to evaluate the response of asphalt binder and asphalt mixture to
permanent deformation due repeated loading, respectively. The Cyclic Indirect Tensile Test was selected
for investigating both stiffness properties and fatigue resistance. For this purpose the complex stiffness
modulus was measured to quantify material stiffness under different temperature and loading conditions
providing information on the visco-elasto-plastic material behavior. Fatigue tests were used to determine
the progressive and localized material damage caused by cyclic loading. The experimental results indicate
that asphalt mixtures prepared with LD slag are suitable for asphalt pavement construction and that in
most cases they perform better than conventional asphalt mixtures prepared with Gabbro aggregate.

1 INTRODUCTION X-Ray Diffraction (XRD) analysis, Scanning Elec-

tron Microscopy (SEM), Thermogravimetry Analy-
1.1 Use of slag in asphalt mixtures sis (TGA) and mercury porosimeter tests. Based on
The production process of steel results in a sig- the experimental results, they concluded that steel
nificant amount of slag by-product: approximately slags can be used as aggregate since they represent
1 ton of slag per 3 tons of stainless steel (Proctor an alternative, cost effective and environmentally
et al. 2000). For Europe every year nearly 12 mil- friendly aggregate source. The use of steel slag in
lion tons of steel slag is produced. The European asphalt mixture was also addressed by Hassan &
Union has among its targets an efficient use of nat- Khabiri (2007) and Norman et al. (1992) based on
ural resources; within this framework, the asphalt Marshall and stability tests according to European
industry needs to recycle industrial wastes such as Standard EN 12697-34 (2012). Kara et al. (2004)
steel slags. used steel slag to prepare asphalt pavement mixtures
In Germany, the use of slags in asphalt mixtures for base, binder and wearing courses. According to
has a long tradition. A number of research efforts the authors, the physical properties of steel slag satis-
were performed by the German Research Associa- fied the standards requirements for asphalt mixture
tion on blast furnace and steel slags, so that the mate- production. Regarding skid resistance of asphalt
rial characteristics of slags are well known (Motz mixture containing steel slag, Asi (2007) stated
et al. 2001). At international level, research activities that mixtures containing 30% slag have the highest
centered their effort on addressing the chemical and skid number compared to Superpave, SMA, and
physical properties of slags used in asphalt mixtures. Marshall mixtures. In addition, a positive effect of
(Wu et al. 2006, 2007) evaluated the possibility of steel slag on rutting resistance and fatigue perform-
using steel slag as aggregate in Stone Mastic Asphalt ance was identified (Asi et al. 2007). One of the most
(SMA) mixtures; mechanochemistry and physical critical possible drawbacks associated to the use of
changes of the steel slag were investigated through steel slag is volume expansion caused by hydration

319
of lime or magnesia components. Hence, high lev- Asphalt mixtures were chosen to ensure that all
els of free lime or magnesia may cause cracking of layers of an asphalt pavement structure (surface
asphalt pavement. The simplest solution to limit this layer, binder layer and base layer) are taken into
deleterious effect is obtained by aging slags or by account. Furthermore, conventional asphalt mix-
accelerating the hydration reaction with water. Alter- tures with different filler content were used. The
natively, slag washing can be used, so that minimal corresponding mix design was chosen in accord-
volume expansion occurs (Kneller et al. 1994). ance with the German Technical Standard: TL
Asphalt—StB 07 (2007). Table 1 shows the aggre-
gate gradation and binder content for the asphalt
1.2 Objective
mixtures investigated.
The objective of the present research is to deter-
mine how the use of LD slag in asphalt mixtures 2.3 Test methods
affects the functional performance of pavement
construction, and specifically, in comparison to 2.3.1 Resistance to permanent deformation
conventional mixture prepared with natural aggre- Deformation resistance was addressed by
gate such as Gabbro. The results of this study and • Cyclic Compression Test in accordance with
the recommendation for construction process are European Standard EN 12697-25 (2005) for
discussed in this paper. rolled asphalt SMA 11 S, AC 16 B S and AC 22
T S and
• Penetration Test in accordance with European
2 INVESTIGATIONS Standard EN 12697-25 (2005) for mastic asphalt
MA 11 S.
2.1 Research approach
In Cyclic Compression Test (Figure 1, left) a
The research approach used in this study is based cylindrical specimen is subjected to repeated pulsed
on an extensive experimental program which compressive load for 0.2 s, followed by a 1.5 s rest
includes different test methods to assess the
mechanical response and the performance of the
Table 1. Aggregate gradation and binder content of the
asphalt mixture composite prepared with differ- asphalt mixtures.
ent aggregates: steel slag and natural Gabbro. The
following properties were evaluated on a set of 8 MA 11 S SMA 11 S AC 16 B S AC 22 T S
asphalt mixture variants used in the different struc-
tural layers which an asphalt pavement is made of: Residue
surface, binder and base layers (MS-4, 1989): on sieve
[M.-%]
• Resistance to permanent deformation; 0.063 mm 27.0 6.5 6.2 6.6
• Stiffness; 1 mm 26.0 15.9 22.4 28.8
• Fatigue resistance, 2 mm 13.0 7.1 13.6 17.7
• Resistance to low temperature cracking; 5.6 mm 14.0 27.4 16.4 9.9
8 mm 20.0 42.8 14.2 9.7
2.2 Materials 11.2 mm – – 25.1 10.6
16 mm – – 2.1 15.8
A total of 8 asphalt mixtures were prepared in the 22.4 mm – – – 0.9
asphalt pavement laboratory. Four different types Binder 7.4 6.5 4.6 4.0
of mixtures, each with natural Gabbro aggregate content
and LD slag, were designed using a standard 50/70 [M.-%]
asphalt binder and a SBS modified binder with
45/80-65 penetration grade according to European
Standard EN 1426 (2013a) and ring and ball tem-
perature according to European Standard prEN
t a(t)
T =+50 \C

stamp
1427 (2013b): 1-...

• MA 11 S (surface layer, mastics asphalt, high

filler content),
specimen

.....
1
60mm

• SMA 11 S (surface layer, stone-mastic-asphalt, ......

low filler content),
• AC 16 B S (binder layer, asphalt concrete with a 100 mm
0 150mm

maximum aggregate size of 16 mm),

• AC 22 T S (base layer, asphalt concrete with a Figure 1. Cyclic Compression Test principle (left), Pen-
maximum aggregate size of 22 mm). etration Test principle (right).

320
period at a test temperature of 50°C. During the test Table 2. Experimental parameters for the determi-
irreversible deformations along the loading direc- nation of stiffness based Cyclic Indirect Tensile Test
tion are recorded for each load cycle. The specimen according to EN 12697-24.
is positioned centrally under the loading frame to
Parameters Values/Response
ensure most homogeneous load distribution.
The load curve is characterized by the lower Test Temperature [°C] +20; +10; 0; −10
stress level σu and the upper stress level σo. σu Loading frequency [Hz] 0.1, 1, 5 und 10
ensures contact between specimen and load stamp. Results Stiffness modulus |E| in
The test starts with the application of the first load [MPa] and Master-Curve
cycle. A constant upper stress of σo = 0.35 MPa
was chosen for testing.
Test ends as soon as 10,000 cycles are reached or Table 3. Experimental parameters for the determina-
when a deformation of 40 ‰ is exceeded. The basis tion of determination of fatigue performance by Cyclic
of evaluation is the creep curve, which shows the Indirect Tensile Test according to EN 12697-24.
evolution of the irreversible permanent deforma- Parameters Values/Response
tions in the specimen.
For the evaluation of deformation resist- Test Temperature [°C] +20
ance, the following parameters are used for Loading frequency [Hz] 0.1, 1, 5 und 10
characterization: Results Number of load cycles until
makro crack
• No. of load cycles at the turning point nw [–],
Fatigue function
• Strain at the turning point εw [‰]. NMakro C1 ⋅ ε eCl2,anf with
In the Penetration test (Figure 1, right) the material specific parameters
dynamic penetration (ETdyn) is determined. The C1 und C2
specimen is subject to a sinusoidal load for 0.2 s
followed by a rest period of 1.5 s. The test ends
when 2,500 load cycles ( = load pulse + rest period) 2.5 Resistance to low temperature cracking
are reached. Penetration depth at end of test is In order to determine the resistance to low temper-
documented. The test temperature was set to 50°C. ature cracking, prismatic asphalt specimens were
subjected to Thermal Stress Restrained Specimen
2.4 Stiffness and fatigue resistance Tests in accordance with European Standard EN
12697-46 (2012b).
Stiffness and fatigue resistance (cracking resist- Prior to the test, prismatic specimen having
ance to repeated loading) is addressed by Cyclic dimensions 40 × 40 × 160 mm3 were glued to steel
Indirect Tensile Test according to the European adapters. After a minimum of 48 h glue curing, spec-
Standard EN 12697-24 (2012a). A sinusoidal load imens are installed in the test device. During Ther-
is applied to a cylindrical asphalt specimen via mal Stress Restrained Specimen Test, the length of
two diametrically opposite load transfer rails. To the specimen is held constant through a set of Lin-
stabilize the position of the asphalt specimen in ear Variable Displacement Transformer (LVDT),
the loading frame, a seating load equivalent to a while its temperature is decreased from an initial
stress of σu = 0.035 MPa is applied, while the maxi- value of T = +20°C with a constant cooling rate of
mum stress σo is chosen based on elastic horizontal ΔT = −10°C/h. A close-loop control system keeps the
strain εel between 50 and 100 μm/m. Tests were car- specimen at constant length. Due to the prohibited
ried out in a temperature-frequency sweep mode as thermal shrinkage, the specimen is subjected to an
shown in Table 2. Resulting parameter is stiffness increasing (cryogenic) tensile stress. The test ends
modulus in function of frequency, represented by at a minimum test temperature of T = – 40°C or at
isotherm plot and by master curve plot. failure, when the cryogenic stress exceeds the tensile
Fatigue performance is obtained based on a strength of the asphalt mixture, respectively.
continuous cyclic loading procedure in the non- The test returns a temperature-dependent func-
linear domain. Stress is varied in a way that initial tion of cryogenic stress σcry (T) [MPa], a failure
strains are in a range of 0.05 to 0.30 ‰ and the stress σF [MPa] and a failure temperature TF [°C].
number of load cycles until macro cracking (drop
in stiffness modulus by 50%) is in the range of
10.000 to 1.000.000 load cycles. on Cyclic Indirect 3 RESULTS AND DISCUSSION
Tensile Test were carried out under the conditions
specified in Table 3. Resulting parameters are the 3.1 Resistance to permanent deformation
number of load cycles until macro cracking, and Penetration depths observed for mastic asphalt
the material-specific fatigue function. (MA 11 S with LD slag and with natural Gabbro

321
aggregate) are shown in Figure 2, (top). After 2,500 0,9
load cycles MA 11 S with LD slag showed much t:: 'E 0,8
~E
lower penetration depth compared to the MA with 1- ';' 0,7
natural Gabbro aggregate. ~~ 0,6
The use of LD slag in mastic asphalt leads to an ->-
~~ 0,5
advantageous deformation resistance compared to t::(C
o.2 0,4
mastic asphalt with natural Gabbro aggregate. ~g 0,3
Deformation resistance of stone mastic asphalt -II(')
~N
t:: ... 0,2
SMA 11 S (Figure 2, bottom) shows the same ten- ~~
C.¢:: 0,1
(C
dency as for the mastic asphalt variants. The number 0
of tolerable load cycles until the turning point of MA 11 S + Gabbro MA 11 S+ LD
the SMA with LD slag is significantly higher than
for the SMA 11 S with natural aggregate Gabbro. 250
Gl
The resulting strains are comparable in magnitude. £....
200
Hence, a trend similar to those observed for MA is lV

shown by the SMA when using LD-slag. : :!:

"'[~ 150
The deformation resistance in terms of number u 0
-ca.
of load cycles at the turning point for the asphalt lV C)
~.= 10 0
mixture for binder and base courses (AC 16 B S and 'bE
... ::I
AC 22 T S with LD slag and with natural Gabbro Gl ....
.0 50
aggregate, respectively) is shown in Figure 3. The E
::I
s::::
results indicate an opposite trend compared to the
asphalt mixtures for surface layer (MA and SMA). SMA 11 S + Gabbro Sr:vtA 11 S + LD
The number of tolerable load cycles until the turn-
Figure 2. Deformation resistance of mastic asphalt MA
ing point is much lower for mixtures prepared with 11 S with LD slag and natural Gabbro aggregate (top);
LD slag; the resulting strains are comparable in deformation resistance of stone mastic asphalt SMA 11 S
magnitude. with LD slag and natural Gabbro aggregate (bottom).
Concerning the disadvantageous deformation
behavior of the asphalt binder and asphalt base mix- 12000
tures prepared with LD slag it has to be noted that "'
..r::
10000
similar mix design (same gradation and bitumen
_;"' ::!::
content) was used for mixtures containing natural ......
>or::
8000

and recycled aggregates. Consequently, LD slag .. ·c; 6000

"CC..

mixtures showed higher air voids contents (up to "' Cl

.2r::::
4000
~·E
3.0 vol.-%). This is probably due to a higher asphalt :Ui! 2000
binder demand of LD slag associated to their porous .0

~
structure. High air voids contents negatively affect r::::
AC16BS + AC16BS + AC22TS + AC22TS+LD
stiffness and resistance to permanent deformation. Gabbro LD Gabbro

Figure 3. Deformation resistance of asphalt binder AC

3.2 Stiffness and fatigue resistance 16 B S with LD slag and natural Gabbro aggregate (left);
deformation resistance of asphalt base course AC 22 T S
The stiffness of the asphalt surface mixtures (mas- with LD slag and natural Gabbro aggregate (right).
tic asphalt MA 11 S and stone mastic asphalt
SMA 11 S, each with LD slag and natural Gabbro 30 .000
aggregate) is shown in Figure 4 in terms of mod- - SMA 11 S + Gabbro
25 .000 - - - ---- SMA 11 S + LD
ulus E (complex modulus) in function of tem- - MA 11 S + Gabbro
perature. The resulting stiffness is for all asphalt D..
~
20 .000 - - - --- -MA 11 S + LD

surface mixtures at a comparable level, regardless ::l

:g 15.000

of whether LD slag or natural Gabbro aggregate ~ 10.000

ai.J
was used. 5.000
The stiffness of the asphalt binder and asphalt
0
base course mixtures (AC 16 B S and AC 22 T S, -40 -30 -20 -10 10 20 30 40 50
Temperature T rcJ
each with LD slag and natural Gabbro aggregate)
are comparable too (not shown). Overall, a minimal Figure 4. Stiffness of mastic asphalt MA 11 S and stone
decrease in stiffness can be observed for mixture mastic asphalt SMA 11 S: E-Modulus [MPa] in function
prepared with LD slag. Nevertheless, this is negli- of temperature, observed on asphalt mixtures with LD
gible and does not significantly affect the perform- slag (dashed lines) and with natural Gabbro aggregate
ance of asphalt pavement under traffic loading. (full lines).

322
 Fatigue resistance of the asphalt surface mixtures 11 S, each with LD slag and natural Gabbro aggre-
(mastic asphalt MA 11 S and stone mastic asphalt gate) were determined based on Thermal Stress
SMA 11 S, each with LD slag and natural Gabbro Restrained Specimen Tests. Figure 7 shows the
aggregate) are shown as Wöhler curves in Figure 5. cryogenic failure stresses (maximum stress until
Similar fatigue resistance is shown for mixtures pre- failure) while Figure 8 presents the corresponding
pared from natural and recycled aggregates (LD slag). temperatures at failure.
Fatigue resistance of the asphalt binder and The parameters describing the resistance to
asphalt base course variants (AC 16 B S and AC 22 low temperature cracking are comparable for all
T S, each with LD slag and natural Gabbro aggre- asphalt surface mixtures prepared with LD slag
gate) are shown as Wöhler curves in Figure 6. or natural aggregate. The failure stresses observed
The Wöhler curves of asphalt binder and asphalt are at a high (good) level. This level is typical for
base mixtures with LD slag are shifted upwards in an commonly used asphalt pavement surface layers
almost parallel manner with respect to those of the in Europe. The failure temperatures are below
mixtures containing natural aggregate. Thus, for the T = −20°C (marked in Figure 8), and thus on a
same elastic strain, mixtures with LD slag can sus- non-critical level for thermal cracking.
tain a higher number of load cycles before a macro The resistance to thermal cracking of the asphalt
crack (material failure) appears. Therefore, the binder and asphalt base course mixtures (AC 16 B S
experimental results indicate that the use of LD slag and AC 22 T S, each with LD slag and natural Gabbro
in asphalt binder and asphalt base course mixtures aggregate) is shown in Figure 9 and Figure 10.
leads to a higher fatigue resistance in comparison to The parameters describing the resistance to
mixture designed with natural Gabbro aggregate. thermal cracking indicate that the low-temperature
behavior of mixtures with LD slag tends to provide
relatively poor performance compared to the mix-
3.3 Resistance to low temperature cracking
ture with natural Gabbro aggregate. The failure
The resistance to low temperature cracking of
the asphalt surface mixtures (MA 11 S and SMA 7,000

5,000
g 100.000 ftl
c.. 5,000
~
E "' 4,000
:g
::I
Ill
Cll""i"'

~ ~ 10.000 +----+--1--+--+~f-H+----+1~1--+--+-+-+-++-t
i
-=~
3,000

2,000
~u
~ 1,000
Q)
.c 0,000
E MA 11 S + MA 11 S + LD SMA 11 S + SMA 11 S + LD
~ 1.000 ...___ __.__.__............................._.___ __.__.__........._._.........u
Gabbro Gabbro
0,01 0,10 1,00
strains [%o]
Figure 7. Resistance to low temperature cracking of
Figure 5. Fatigue resistance of mastic asphalt MA mastic asphalt MA 11 S and stone mastic asphalt SMA
11 S and stone mastic asphalt SMA 11 S: Wöhler curves 11 S: failure stress [MPa] for asphalt mixtures with LD
observed for asphalt variants with LD slag (dashed lines) slag and with natural Gabbro aggregate.
and with natural Gabbro aggregate (full lines).
MA11 S + SMA11 S +
Gabbro MA 11 S + LD Gabbro SMA 11 S + LD
0,0
E -5,0
Cll

~
~
-10,0

_; :::!::10.000 +:-:-:==±:-:=+=:±-::-::-:ef:¥-::''t±:H\'=:-::::-:-:-:!=:±-::-::,-:±-,-,±-::-:±~±:l
u...,
! -15,0
~~
"C .. ] -20,0
_2u
0
:! -25,0
.a 1.000 -L-----'---'--..:.......;'--'--'--'--'-'--'------'---'---'-...:....:.--'--'-'-l -30 ,0
~ 0,01 0,10 1,00
strainsr/oo] -35,0 ..1...--------------------

Figure 6. Fatigue resistance of asphalt binder AC 16 Figure 8. Resistance to thermal cracking of mastic
B S and asphalt base course AC 22 T S: Wöhler curves asphalt MA 11 S and stone mastic asphalt SMA 11 S:
observed for asphalt variants with LD slag (dashed lines) failure temperature [°C] for asphalt mixtures with LD
and with natural Gabbro aggregate (full lines). slag and with natural Gabbro aggregate.

323
 4,500 .....-----------1~--------- with LD slag showed slightly higher air voids
4,000 + - - - = = - - - - - -- i f -- - - - - - - - - contents resulting in a disadvantageous resist-
'l 3,500 ance to deformation.
!.. 3,000 • The stiffnesses of the asphalt surface, asphalt
~ 2,500
binder and asphalt base mixtures are at a com-
"' 2,000
parable level, regardless of whether LD slag or
:I
J! 1,500
natural Gabbro aggregate was used.
1,000
• Fatigue resistances for the asphalt surface mix-
0,500
tures are at a comparable level, regardless of
whether LD slag or natural Gabbro aggregate
0,000
AC 16 B S + AC 16 B S + LD AC 22 T S + AC 22 T S + LD was used. The use of LD slag in asphalt binder
Gabbro Gabbro and asphalt base course leads to a higher fatigue
resistance in comparison to mixtures with natu-
Figure 9. Resistance to low temperature cracking of
ral Gabbro aggregate.
asphalt binder AC 16 B S and asphalt base course AC
22 T S: failure stress [MPa] for asphalt mixtures with LD • The level of resistance to thermal cracking are
slag and with natural Gabbro aggregate. comparable for all asphalt mixtures, regardless
of whether LD slag or natural aggregate Gabbro
AC16BS+ AC22TS+
was used. The level is evaluated as "adequate"
Gabbro AC 16BS+ LD Gabbro AC22TS+LD for European weather conditions.
0 ,0

~"' -5,0
Overall, the experimental results and the analysis
!
c.
-10,0
conducted in the present investigation indicate that
asphalt mixtures prepared with LD slag are suit-
! -15,0 able for asphalt pavement construction and that in
"'
~ -20 ,0
most cases they perform as good as or even better
;j!! than conventional asphalt mixtures prepared with
-25,0
natural aggregate.

REFERENCES
Figure 10. Resistance to thermal cracking of asphalt
binder AC 16 B S and asphalt base course AC 22 T S:
failure temperature [°C] for asphalt mixtures with LD Asi, I.M., Qasrawi, H.Y., Shalabi, F.I. 2007. Use of steel
slag and with natural Gabbro aggregate. slag aggregate in asphalt concrete mixes, Can. J. Civil
Eng., 34 (8), 902–911.
Asi, I.M. 2007. Evaluating skid resistance of different
stresses are lower and the failure temperatures are asphalt concrete mixes, Build. Environ., 42, 325–329.
higher. However, the level is evaluated as “ade- European Committee for Standardization (CEN). 2005.
quate” for European weather conditions. Bituminous mixtures—Test methods for hot mix
asphalt—Part 25: Cyclic compression test, Brussels,
Belgium.
European Committee for Standardization (CEN). 2012a.
4 SUMMARY AND CONCLUSIONS European Standard EN 12697-12. Bituminous mix-
tures—Test methods for hot mix asphalt—Part 24:
In this paper the fundamental performance prop- Resistance to fatigue, Brussels, Belgium.
erties of different types of asphalt mixtures pre- European Committee for Standardization (CEN). 2013a.
pared with 100% LD slag and natural aggregate European Standard EN 1426. Bitumen and bitumi-
were investigated. Resistance to permanent defor- nous binders—Determination of needle penetration,
mation, stiffness, fatigue and low temperature Brussels, Belgium.
cracking were evaluated based on a comprehensive European Committee for Standardization (CEN) 2012b:
set of experimental methods. The following con- European Standard EN 12697-46. Bituminous mix-
tures—Test methods for hot mix asphalt—Part 46:
clusions can be drawn: Low temperature cracking and properties by uniaxial
• The use of LD slag in mastic asphalt leads to an tension tests, Brussels, Belgium.
advantageous deformation resistance compared European Committee for Standardization (CEN) 2013b.
to mastic asphalt with natural Gabbro aggre- European Standard prEN 1427. Bitumen and bitu-
minous binders—Determination of the softening
gate. In contrast asphalt binder and asphalt point—Ring and Ball method, Brussels, Belgium.
base course mixtures with LD slag showed a Forschungsgesellschaft für Straßen- und Verkehrswesen
lower deformation resistance compared to the (FGSV). 2007. Technische Lieferbedingungen für
mixtures with natural Gabbro aggregate. Due to Asphaltmischgut für den Bau von Verkehrsflächen-
a higher binder need of LD slag, the mixtures befestigungen (TL Asphalt-StB 07), Köln, Germany.

324
Hassan, Z., Khabiri, M. 2007. Preventive maintenance of MS-4 1989. The Asphalt Handbook, Asphalt Institute,
flexible pavement and mechanical properties of steel College Park Medison, Vol. 4.
slag in asphalt J. Environ. Eng. Landscape Manage., Norman, A., Joseph, C., Papagiannakis, T., Theriault, E.G.,
15 (3), 188b–192b. Bergan, A.T. 1992. Use of Steel Slag in Asphaltic Con-
Kandhal, P. S., Hoffman, G. L. 1997. Evaluation of steel crete, 1147 ASTM Special Technical Publication, 3–18.
slag fine aggregate in hot-mix asphalt mixtures, Trans- Proctor, D.M., K.A. Fehling, E.C. Shay. 2000. Physical
portation. Research. Record 1583, 28–36. and chemical characteristics of blast furnace, basic
Kara, M., Gunay, E., Kavakli, B., Tayfur, S., Eren, K., oxygen furnace, and electric arc furnace steel industry
Karadag, G. 2004. The use of steel slag in asphaltic slags, Environ. Sci. Technol., 34, 1576–1582.
mixture, Key Eng. Mater., 264–268 (III), 2493–2496. Wu, S., Xue, J., Ye, Q., Chen,Y. 2007. Utilization of steel
Kneller, W. A., Gupta, J., Borkowski, M. L., Dollimore, slag as aggregates for stone mastic asphalt (SMA)
D. 1994. Determination of original free lime content mixtures, Build. Environ., 42 (7), 2580–2585.
of weathered iron and steel slags by thermogravimet- Xue, Y., Wu, S., Hou, H., Zha, J. 2006. Experimental
ric analysis, Transportation Research Record 1434, investigation of basic oxygen furnace slag used as
National Research Council, Washington, DC, USA. aggregate in asphalt mixture, J. Hazard. Mater., 138
Motz, H., Geiseler, J. 2001. Products of steel slags an (16), 261–268.
opportunity to save natural resources, Waste Manage.,
21, 285–293.

325
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Feasibility study on a thermoset polymer-coated emulsified warm-mix

asphalt mixture

P.J. Yoo, B.S. Ohm & K.S. Park

Highway Research Institute, Korea Institute of Civil Engineering and Building Technology, Goyang-Si,
South Korea

I.L. Al-Qadi
Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign,
Urbana, IL, USA

ABSTRACT: A feasibility study was performed on a warm-mix asphalt mixture including aggregates
coated by a chemically modified thermoset polymer. The thermoset polymer is a chemically synthe-
sized with a carboxylic butadiene rubber and a silanol coupling agent. The adhesive strength was veri-
fied through the Dolly test and the Boiling test. The adhesive strength of the synthesized epoxy was at
least 2-fold greater than that of a general epoxy. The boiling test verified that a superior coating effect
maintaining greater than 90% of the epoxy-coated binder film. The mixture including the synthesized
epoxy-coated aggregates resulted in 36% higher average indirect tensile strength compared to the general
PG64–22 mixture. The rut deformation of the epoxy-coated mixture resulted in 2.1 times lower than the
rut of the standard PG64-22 section through an accelerated field test.

1 INTRODUCTION double bonds (c=c) in the styrene or polybutadi-

ene segment. The unexpected degradations of the
Asphalt is composed of various polar and non- polymeric networks cause the loss of cross-links
polar components containing nitrogen, oxygen, of the SBS modifier and the oxidation of bitumen
sulfur, and metal atoms. Although asphalt has itself so that the softening point of the modified
been widely used as a binding material for flexible binder results in an almost equal temperature to
pavements, due to the chemical complexity of the that of the virgin bitumen with less than 500 hours
asphalt, it is impossible to characterize its chemi- of aging time in a pressure aging vessel (Shao-peng
cal composition exactly. On the other hand, it is et al., 2009, Zaho et al., 2016).
true that the chemical variety of the asphalt makes A polymer modified asphalt binder is supposed
possible to induce a promising chemical bond with to be used in a region where the average 7-day max-
an external agent or additive (Peiliang et al., 2011). imum temperature exceeds 70°C or the absolute
The chemical composition of asphalt is highly value of temperature difference between high and
dependent on the source of crude oil, and the low temperature in a performance grade system
components of asphalt are usually grouped into is greater than 90°C. SBS has been widely used to
saturates, aromatics, resin and asphaltenes. The modify a bitumen because of its excellent engineer-
cohesive and adhesive strength of asphalt are ing properties, such as superior flexibility at a low
strongly affected by the contents of the resin and temperature and higher viscosity at a high tem-
asphaltene, i.e., polar groups. The non-polar com- perature than a unmodified binder. However, the
ponents in asphalt, such as saturates and aromatic chemical role and mechanical effect of SBS in the
groups hardly contributes to enhance an adhe- bitumen are still questionable in view of the rela-
sive or cohesive strength due to the disassociation tively fast aging or oxidation due to the easy cou-
property of them (Yao et al., 2015). pling of oxygen molecules in the air or water with
The cohesive strength of a polymer-modified the unsaturated carbon double bonds in the SBS.
asphalt binder is enhanced by the cross-linking This may lead to the disappearance of cross-linked
molecular networks polymerized by the tri-block network structures (Riande et al., 2000, Sirin et al.,
copolymer such as Styrene-Butadiene-Styrene 2008, Xiaohu & Ulf 1998, Hanyu et al., 2005).
(SBS); however, some unsaturated mid-blocks may To obtain an enhanced cross-linking network in a
cause undesirable oxidative reactions of the carbon molecular structure, there must be a polymerization

327
process, including relatively high heat energy and subjected to an impact loading. Therefore, the
pressure in the process of thermal copolymeriza- flexibility of the epoxy must need to be improved
tion or polycondensation (Riande et al., 2000). by adding a rubber additive, e.g., a Carboxyl-Ter-
Besides the additional cost for a polymerization minated-Butadiene acrylonitrile rubber (CTBN),
or polycondensation process, relatively high heat- that is designed to react with the epoxide group of
energy consumption in an asphalt plant and the the epoxy resin before the irreversible hardening
phase separation due to long-term storage may be reaction has proceeded (Riande et al., 2000, Yee &
practical drawbacks in using the polymer-modified Pearson 1986, Hajime et al., 2009).
asphalt binder. Figure 2 shows the molecular structure of CTBN,
An alternative way to enhance the strength or which can be used as the rubber modifier of 10 to
toughness of asphalt mixtures without engaging 16 weight % usage to the total amount of DGEBA.
the polymerization or polycondensation of asphalt Figure 3 shows the synthetic mechanism of the
binder is addressed in this study. This study pro- rubber-modified epoxy resin used to enhance the
poses an engineered thermoset epoxy for the coat- flexibility of the epoxy resin. The enhanced tensile
ing agent of aggregates in view of the post-coating property due to the rubber was verified by an exten-
process instead of using the pre-modification of sion test.
asphalt binder. The rubber modification of epoxy resin was per-
A thermoset epoxy resin was chemically modi- formed in a 5-neck round-bottom flask containing
fied to be used as the coating agent of the surface the DGEBA and the rubber compound. The rub-
of aggregates. In addition, a silanol coupling agent ber and the DGEBA were mechanically stirred first
as a bonding strength promoter at the surface of for 30 minutes. Then, 2 ppm (Part-Per-Million) of
aggregates was utilized instead of engaging the ETPPI (Ethyl TriPhenyl Phosphonium Iodide) was
pre-polymerization in this study. added into the stirred liquid to initiate the rubber
To overcome the chemical and practical short- modification. The temperature of the stirred liquid
comings in the polymer pre-modification while in the rounded flask was gradually increased from
maintaining equivalent mechanical properties, 100°C to 170°C at the rate of 1°C per minute and
such as adhesive strength and indirect tensile maintained the reaction time for 3 hours at 170°C.
strength, this study utilizes an asphalt mixture After done the rubber modification and cooled
with an emulsified asphalt binder and rubber- and down to the temperature to 120°C, then the silanol
silanol-treated epoxy-coated aggregates under the coupling agent in 1.0 weight% to the total amount
low coating temperature in the range of 80–100°C of the rubber modified-DGEBA was added into
and the mixing temperature in between 120–130°C. the flask and stirred for 2 hours.
The relative enhancement in laboratory and
field performances on the warm-mix asphalt con-
crete including the aggregates coated by the synthe-
sized polymer was verified through indirect tensile
strengths and accelerating pavement tests.
Figure 1. DGEBA epoxy resin.

2 RUBBER AND SILANOL MODIFIED 0 0

EPOXY RESIN HO-~vvvR-4~~c=c-~~~~~~tR-~-OH
[. 'x
H \ HI) m

The epoxy resin and hardener are supplied in a C_N

two-part formulation, which is in a liquid state in

a room temperature. The brittleness of the epoxy Figure 2. Molecular structure of CTBN.
and the difficulty in curing time-control make hard
to use as a paving binder despite of its superior 0
II[(H,
0
H\(H, H\ 111
mechanical property such as an adhesive strength HO-Cl\C -~=~-Cr\C -~nC-OH
c==N
and the chemical stability at a room temperature.
The ductility and toughness of epoxy can be
improved by adding a rubber, which is a carboxyl-
terminated copolymer of butadiene designed to
react with the epoxide group.
The general molecular structure of a liquid Dig-
lycidyl-Ether-Bispenol-A (DGEBA) epoxy resin is
shown in Figure 1 before a chemical modification.
The curing makes the molecular structure of Figure 3. Synthetic mechanism of rubber-modified
the epoxy rigid; a brittle failure can occur when epoxy resin.

328
 After the rubber modification of the epoxy resin Blocking mechanism
for its flexibility, for the enhancement of adhesive 1
R 'C=O + R3-NH2
interfacial strength and cohesive bond strength R~
at the interface between the aggregates and the
asphalt binder, a silanol coupling agent, which Unblocking mechanism
includes two different reactant groups, such as the R,,
X-site and the OR-site in Figure 4, is coupled to + /C=N-R3
the rubber-modified epoxy resin. R2

The silanol additive is usually added to a tack

coat emulsion as a small amount of additive in the Figure 6. Moisture-hardening mechanism: blocking
and unblocking.
water used to dilute the emulsion. The enhanced
bond strength and anti-stripping capacity were
contacts with water molecules, which are blocked by
reported as the main advantage of the silanol addi-
the ketone group (R1(CO)R2), as shown in Figure 6.
tive (Hajime et al., 2009).
The moisture-curable hardener engages a block-
The mechanical effect of the silanol coupling
ing and unblocking mechanism in the hardening
agent was verified by the Dolly test in this study
process.
(Cynthia et al., 2002).
When the blocked ketone group comes into
Besides the polar-polar reaction at the surface of
contact with water, the amine groups (−NH2) are
the aggregates, which is an organic and inorganic
dissociated by the unblocking reaction, as shown
reaction (OR-site), to introduce additional nonpo-
in Figure 6.
lar-nonpolar reactions between the asphalt binder
The unblocked amine group undergoes a hard-
and the rubber-modified epoxy, which is an organic
ening reaction with the rubber and silanol modi-
and organic reaction (X-site) in Figure 4, this study
fied epoxy resin. This reaction requires water;
introduces an additional modification of the rubber-
therefore, an asphalt emulsion containing 30
modified epoxy using the silanol coupling agent, as
weight % of water is utilized in this study.
shown in the molecular structure of silanol-rubber-
General asphalt emulsions are normally con-
modified epoxy resin, Figure 5. Figure 5 shows the
sidered to be oil-in-water type and contain
conceptual diagram of the hydrogen bond between
approximately 60 weight % asphalt, 0.1% to 2.0%
the hydroxyl group of the silanol and the hydroxyl
emulsifier, and approximately 40 weight % water
group at the surface of the inorganic aggregate.
plus some minor components. However, for the
The hydroxyl group (−OH) at the surface of the
moisture-hardening mechanism, the modified
inorganic aggregate exists when the aggregate con-
emulsion is manufactured to maintain the weight
tains some amount of water or moisture.
% ratio between the water and solid asphalt at
The relative humidity of hot-mix aggregate
30:70 to ensure that there is no surplus water after
should be less than 2% after the dry process is per-
completing the hardening reaction (Delmar 2006).
formed at an asphalt plant. However, the hardener
for the modified epoxy resin was chemically designed
to initiate the hardening reaction when the hardener
3 LABORATORY TESTS

OR
OCH, 0 3.1 Extension and adhesion test
I I I\ To verify the enhanced flexibility of the rubberized
c:::::::> H,co-i;-c-c-c-o-c~
H, H, H, H,
X"""""""Si-OR
IOR and silanol treated epoxy, an extension test at 25°C
OCH, was performed on the dog-bone type specimens
of 5 mm thick, 25 mm width, and 55 mm length,
Figure 4. Molecular structure of the silanol coupling Figure 7a.
agent. While the extension of the general epoxy speci-
men was about 10 mm at the fracture point, the

0 p;-
,j.,\1.
I
0<,
C"
I""=T
A I
CH 1Di""
Q-CH,-Ot-CH,-
-
I·oF"\
I~
I
CH 0I
Q-CH,-Ot-CH,-Q-C-""1""'0<,
I
/ ••···•' \
.., :' l eo.., \
rubberized and silanol treated epoxy was extended
to 38 mm before the fracture point met as shown
in Figure 7b. The rubberized and silanol treated
y 0<, y y y 0\ y y /
>+1'-C~~~:;; ••. /
epoxy may result in enhanced ductility at the sur-
HO-s~-o-sl-o-s~-oH Ho-s1-o-s~-o-s~-OH face of the aggregate after setting.
H J....._, J....._ H J....._ - - J J H J....._ "
' o "-o 'o H& ~ ' o H H H
In addition, the effect of the silanol coupling
_ l _ l _ _ l_ _l _ l_ _ l _
Inorganic Aggregate Surface Inorganic Aggregate Surface agent was verified through the portable pull-off
adhesion Dolly test, which can measure the force
Figure 5. DGEBA+CTBN+Silanol-modified epoxy required to pull a specified diameter of coating
resin. away from its substrate, as shown in Figure 8a.

329
 (a)

(b)
Figure 9. Boiling test for 6 hours: (a) general PG64-
Figure 7. (a) Extension test machine and (b) test 22 coating specimen, (b) 4.0% SBS modified PG64-22
result. binder coating specimen and (c) modified epoxy and
emulsion coating specimen.

This result represents the superior surface bond

strength at the surface of aggregate, which is
coated by the modified epoxy.

3.2 Boiling and indirect tensile strength test

The boiling test, which analyzes the effect of water
Figure 8. (a) Dolly test kit (b) steel, (c) concrete, and
on a bituminous-coated aggregate using boiling
(d) granite substrates.
water according to ASTM D3625, was performed to
check the peeling or unpeeling extent on the aggre-
Table 1. Portable pull-off adhesion test. gate surface coated by the general PG64-22 asphalt
binder, the 4% SBS modified asphalt binder, and the
Rubber-Silanol modified epoxy and emulsion (Waite et al., 2012).
modified Unmodified The aggregates remaining portion greater than
Test protocol epoxy (MPa) epoxy (MPa) Substrates the sieve #8 (2.5 mm) up to 13 mm was prepared at
160°C. The general specimen, including the coated
Dolly 1.40 0.64 Steel
aggregates with the PG64-22 asphalt binder, was
ISO 4624 3.19 2.95 Concrete
boiled for 6 hours continuously, and almost all
at 25°C 8.36 6.13 Granite
coating films were peeled out of the aggregate sur-
face above 90% of its initial coating, as shown in
Figure 9a.
The maximum pressure indicator of the pres- The 4% SBS pre-modification of the PG64-22
sure gauge provides a direct reading of the pressure asphalt was done at 185°C for 2 hours and the aggre-
at which the pull-off occurred. Various substrates, gates were coated completely. After 6 hours boiling
such as steel, concrete, and granite, were used to was done, the coating film was remained above 90%
measure the relative bond strength between the of its initial coating, as shown in Figure 9b.
unmodified epoxy and the rubber-silanol modified The modified epoxy specimen was mixed at
epoxy, Figure 8b, 8c, and 8d respectively. 80–100°C for the effective hardening reaction
The bond strength of the rubber-silanol-modi- between the modified epoxy resin, the hardener,
fied epoxy at the surface of steel is approximately and the water in the emulsified asphalt. The coat-
two-fold higher than the unmodified epoxy. The ing film of the modified epoxy specimen main-
superior adhesion strength for the modified epoxy tained more than 90% of the initial film during the
on inorganic substrates, such concrete and granite, 6 hours of boiling, as shown in Figure 9c.
indicates that the silanol underwent polar-nonpo- The relatively higher adhesive strength of the
lar reactions at the surface of each substrates, as modified epoxy mixture resulted in the supe-
shown in Table 1. rior resistance to adhesive failure at the interface

330
between the aggregate and binder at the lowest mix- average indirect tensile strength of the modified
ing temperature. epoxy cores was 35% greater than the general
A 13 mm aggregate gradation as shown in mixtures.
Table 2 was selected to make the gradation coar- The relatively higher bond strength at the inter-
serincluding at least 60 remaining % of the aggre- face between the aggregate and the modified epoxy
gates greater than 4.75 mm sieve for the indirect binder contributed to the enhanced indirect tensile
tensile strength test and the test-bed construction strength in the lab and field specimens.
as following section. The optimum asphalt content
of 5.5% was decided through the Marshall Mix
3.3 Accelerated field test
Design process. The average air voids resulted in
6.0% for the general mixtures and 6.3% for the A test-bed was built for evaluating the field perform-
modified epoxy specimens. ance of the warm-mix asphalt mixture treated by the
The indirect tensile strength tests were done on modified epoxy and emulsion. The test-bed began
the specimens of the general PG64-22 specimens with the aggregate pouring to the mobile mix truck in
and the modified epoxy coating specimens, such as Figure 11. For the pre-coating of aggregates, the liq-
Figure 10a for the lab specimens and Figure 10b uid compound of the modified epoxy resin and emul-
for the field core specimens. sion were prepared around 100°C according to a mix
The average indirect tensile strength of the design table in the Table 3. The PG64-22 binder, the
modified epoxy specimens resulted in approxi- epoxy harder, and the coated aggregates were finally
mately 37% higher than the value of the general mixed in the temperature range of 120–130°C.
PG64-22 specimens for the lab specimens as shown To investigate the relative field performance
in Figure 10a. between the unmodified asphalt mixture and the
The indirect tensile strength tests for field speci- modified epoxy mixture, the accelerated field test
mens cored from the accelerated pavement test was performed using the heavy vehicle simulator
section were done as shown in Figure 10b. The equipped with a dual truck tire, 275/80R22.5 and
720 kPa of tyre pressure.
Table 2. Percent passing of 13 mm coarse aggregate. The wheel loadings were continuously applied
to the sections with 4.1 metric tons, 6.1 metric tons,
Sieve (mm) 20 13 10 5 2.5 0.6 0.3 0.15 0.08 and 8.1 metric tons, respectively, as shown in Fig-
ure 12a. The overlay thickness of the whole section
% passing 100 92.8 57.6 36.4 23.2 17.0 11.8 6.6 2.3
was 5 cm with the 13 mm coarse aggregate grada-

0 .9

0.8 0 .75 0 .73

&
6 0 .7 0 .68
0 .59
~ 0 .6
0 .50 0 .49
~ 0 .5 r
Q) .........
-~
0.4
~ 0 .3 li;
~
~
0 .2

0 .1
1.,f>
....
•.. ·•·
----r-- '----
General PG64-22 Mix Modified Epoxy Mix

(a)
Figure 11. Mix preparation using mobile mix-truck.
0 .7 , - - - - - - - - - - - -0-.6-3- - - :0;:-;.6::-;:5- - - - - ,

Table 3. Epoxy modification table.

~ 0 .51
;;;. 0 .5
~ 0 .42 Section Section
~ 04
Components Weight percent Type A Type B
-~ 0 .3
,!!!
0
~ 0.2 Modified Epoxy wt.% of 0.8% 1.3%
~ Resin Coat Aggregates
0 .1
Asphalt Emulsion wt.% of 1.8% 1.3%
Coat Aggregates
General PG64-22 Mix: Field Cores Modified Epoxy Mix: Field Cores
Hardener wt.% of Epoxy 35% 35%
(b) Resin
Asphalt Binder wt.% of 4% 4%
Figure 10. Indirect tensile strengths: (a) Laboratory Aggregates
specimens and (b) field core specimens.

331
 EpoxymoflfiedSeetions li»..Generai~C<meteSeciion

;4-f ~+~" : ""

il l)!>tA ! T>1" B IC...
..
2.511 1

(a) (b)

HorlzontaiDistance(cm)

(a)
(c) (d) (e)

Figure 12. (a) Heavy vehicle simulator, (b) test-bedlay-

out, (c) laying-out, (d) compaction, and (e) final sections.

tion, which is the same gradation for the indirect

tensile strength test aforementioned, for the type A
and B of the epoxy modified sections and the PG64-
22 control section, Figure 12b. The laying-out and
compaction were performed in the temperature Horlz:ontaiDistance(cm)

(b)
range of 120–130°C, Figure 12c, 12d, and 12e.
The water content of the asphalt emulsion was
approximately 30%. The mixing temperature was
in the range of 120°C–130°C, which is less than the
general warm-mix asphalt concrete, such as 140°C.
The rut measurement was performed by the
straight-edge method every 5,000 loading repeti-
tions up to 60,000 loading repetitions. The mag-
nitude of wheel loading was varied at three load
levels, 4.1, 6.1, and 8.2 metric tons for 20,000 repe-
titions, respectively. All the axle load was converted (c)
to the Equivalent Single Axle Load (ESAL) based
on a forth power law such as ESAL = ((Wheel
Load)/(4.1 metric ton))^4 × Loading Repetitions Modified EpoxySecliJn A

with the 4.1 metric ton as a reference axle load. .,I\ -+-ModlfiedEpoxySecliJnB

The absolute value of the min-max rut in verti- ~ Conlrc1_13mm_Rut area A

cal direction were 9.2 mm for the epoxy modified

section A, 6.9 mm for the modified section B, and
14.8 mm for the control section, as shown in Fig-
:l
.s ) \

X -- -- X,'' · , , , '
----.,

'
-><-Conlrol_13mm_Rut Area B

X · ---

ure 13a, 13b, and 13c respectively.

All the modified epoxy sections had greater
resistance to rut deformation than the control sec- (d)
tion. While the accumulated rut-depth versus ESAL
of the epoxy modified section A and B shows a pla- Figure 13. Rut measurement: (a) modified epoxy sec-
teau region after initial densification as shown in tion type A, (b) modified epoxy section type B, (c) con-
Figure 13d, the general sections were continuously trol section, and (d) accumulated rut.
deformed till the end of loading repetitions.
During the loading repetitions up to 5,000 rep- aggregates at the mixing temperature in the range
etitions as an initial densification stage, the rut of 120°C–130°C may result in lower horizontal
profile fluctuated due to maintain its initial surface rut-profile despite of using the 3.3 metric ton of a
texture of the large aggregate. As increasing the light weight roller compactor.
loading repetitions, the profile was getting distin-
guished to be treated as the rut deformation. The
irregular deformations under the left and right 4 CONCLUSIONS
wheels were caused by the non-uniform tire pres-
sures of them. This study proposed a warm-mix asphalt mixture
The enhanced adhesive strength and indirect including aggregates coated by chemically modi-
tensile strength by the modified epoxy coating fied epoxy.

332
 The CTBN rubber and the silanol coupling REFERENCES
agent were used to modify the epoxy resin to
enhance the flexibility of the epoxy and the adhe- Cynthia, A.G., Fletcher, J. & Peters, J. 2002. Standard
sive strength at the surface of the aggregate. test method for pull-off strength of coatings using
The emulsified asphalt binder, which is an oil- portable adhesion testers. ASTM International Com-
in-water emulsion including less than 30 weight % mittee D01, 100 Barr Harbor Drive, West Consho-
hocken: PA, USA.
water, was utilized to initiate the moisture-harden- Delmar, R.S. 2006. Asphalt Emulsion Technology, Trans-
ing reaction at the surface of the aggregate. portation Research Circular E-C102, TRB.
The bonding effect at the aggregate surface was Hajime, K., Kentaro, I., Atsushi, N., Satoshi, M. &
verified by the boiling test for 6 hours. The coating Sigeki, H. 2009. CTBN-epoxy Carbon Black Nano-
film of the general PG64-22 binder peeled out more composite with Impact-energy Absorbability. The
than 90% of the film, but more than 90% of the Seventeenth International Conference on Composite
coating film of the synthesized epoxy-coated aggre- Materials, Edinburgh, UK.
gate remained after the boiling was done for 6 hours. Hanyu, A., Sadaharu, U., Atsushi, K. & Kazuo, S. 2005.
The superior flexibility of epoxy was verified Effect of the Morphology of SBS Modified Asphalt on
Mechanical Properties of Binder and Mixture. Journal
by the extension test, which resulted in three-fold of Eastern Asia Society for Trans Studies6:1153–1167.
higher extension than that of the unmodified epoxy. Peiliang, C., Shanfa, C., Jianying, Y. & Huaxin, C. 2011.
The enhanced adhesive strength of the rubber- Compatibility and mechanical properties of epoxy
silanol-modified epoxy at the surface of a steel resin modified asphalt binders. Journal of Pavement
substrate was two-fold higher than the strength Research and Technology 4(2): 118–123.
from the unmodified epoxy in the Dolly test. Riande, E., Ricardo, D.C., Marfarita, G.P., Rosa, M.
The relatively higher adhesive strength of the & Catalina, S. 2000. Polymer viscoelasticity. Marcel
modified epoxy mixture resulted in superior resist- Dekker, Inc., New York.
ance to adhesive failure at the interface between Shao-peng, W., Ling, F., Lian-tong, M., Yong-chun, C.
& Gou-jun, Z. 2009. Influence of aging on the evolu-
the aggregate and the binder. tion of structure, morphology and rheology of base
From the indirect tensile loading test, the aver- and SBS modified bitumen. Construction and Building
age indirect tensile strength of the modified epoxy Materials 23:1005–1010.
coating specimens was 36% higher than the general Sirin, O., Hong-Joong, K., Mang, T. & Bouzid. C. 2008.
PG64-22 specimens. Comparison of rutting resistance of unmodified and
The enhanced adhesive strength and indirect SBS-modified Superpave mixtures by accelerated
tensile strength due to the epoxy coating at the pavement testing. Construction and Building Materials
mixing temperature in the range of 120–130°C 22: 286–294.
resulted in lower rut deformation. From the accel- Waite, G.N., McDaniel, R.S., & Knake, M. 2012. Stand-
ard Practice for Effect of Water on Bituminous-
erated pavement test, the absolute value of the Coated Aggregate Using Boiling Water, ASTM
min-max rut in the vertical direction was 2.1 times International Committee D04 (ASTM D3625), 100
lower than the control section. Barr Harbor Drive, West Conshohocken: PA, USA.
The aggregate coating mechanism through the Xiaohu, L. & Ulf, I. 1998. Chemical and rheological eval-
modified epoxy was verified as a promising meth- uation of aging properties of SBS polymer modified
odology with the enhanced adhesive and indirect bitumens. Fuel 77(10): 961–972.
tensile strength. Yao, H., Qingli, D. & Zhanping, Y. 2015. Fourier trans-
The superior performance of the emulsified form infrared spectroscopy characterization of aging-
warm-mix asphalt mixture when using the rubber- related properties of original and nano-modified
asphalt binders. Construction and Building Materials
and silanol-treated asphalt mixture can contribute 101(1): 1078–1087.
to extend the field performance of asphalt concrete Yee, A.F., & Pearson, R.A. 1986. Toughening mecha-
pavement. nisms in elastomer-modified epoxies. Journal of Mate-
rials Science 21(7): 2462–2474.
Zaho, X.,Wang, S., Wang, Q. & Yao, H. 2016. Rheologi-
ACKNOWLEDGEMENTS cal and structural evolution of SBS modified asphalts
under natural weathering, Fuel 184: 242–247.
This work was supported by the Internal Research
Program (Pothole-Free Pavement System) of
Korea Institute of Civil Engineering and Building
Technology funded by the Ministry of Knowledge
and Economy, Rep. of Korea.

333
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Effect of new types of synthetic waxes designed for low-temperature

asphalt mixtures

L. Benešová, J. Mastný & J. Valentin

Department of Road Structures, Faculty of Civil Engineering, CTU in Prague, Prague, Czech Republic

ABSTRACT: The main advantage of using synthetic waxes for bituminous binders and asphalt mixes is
reducing of the viscosity and better workability at lower paving temperature. It makes possible to reduce
compaction temperature of asphalt mixtures which are related to energy saving in the production and lay-
ing of asphalt mixtures in comparison with conventional HMA. Another advantage is the possibility to
increase technological time interval for the transport and processing. There are many additives to reduce
working temperatures for asphalt mixtures and this area noted considerable progress in the past 10–15
years. Currently there are developed and tested a new generation of waxes, with using various bases—
traditional FT waxes, fatty acid amides, as well as polyethylene waxes and waxes derived from extracts of
sugar cane. The article deals with comparison of the selected types of synthetic waxes mixed with pure
bituminous binder gradation 50/70. The properties of bituminous binders were evaluated using the basic
empirical tests, rotational viscosity, MSCR test and oscillatory test using the Dynamic Shear Rheometer
(DSR). Subsequently, the individual binders were used for the production of asphalt mixture ACsurf 11
at reduced temperature. The asphalt mixtures were tested for porosity, water sensitivity (ITSR), stiffness
modules, resistance to permanent deformation and low temperature characteristics. The results show
that the use of synthetic waxes may reduce the operating temperature of production and compaction of
asphalt mixtures. However great attention should be given to the potential negative effects that can be
caused by both chemical composition of the wax and more often to so improved workability at reduced
temperature, that there is a significant decrease in porosity. Attention is also focused on the connection
between performance based characteristics of asphalt mixtures and bituminous binders.

1 INTRODUCTION US research program, SHRP (Strategic Highway

Research Program) from the 1990’s. The research
The use of synthetic waxes constitutes one of the focused on a higher level of connecting laboratory
possibilities of producing warm asphalt mixtures. testing with real-life pavement behaviour and per-
Reduction of production temperature is becoming formance which was intended to ultimately result
the focal point for both investors and contractors in the choice of such input materials that would
in road sector. The advantages include not only improve asphalt mix quality and reduce the neces-
the total energy demand related to asphalt pro- sary investments in repair and maintenance. From
duction but also reduced greenhouse gas (CO2 in an engineering perspective, the importance of
particular) generation and improved health condi- functional (performance-based) testing is tied to
tions of workers. According to D’Angelo (2008), the viscoelastic behaviour of bituminous binders
typical expected emission reductions are 30–40% as their properties depend on the temperature and
for CO2 and Sulphur Dioxide (SO2), Volatile traffic load.
Organic Compounds (VOC) can be reduced by up
to 50%, Carbon Monoxide (CO) 10–30%, Nitrous
Oxides (NOX) 60–70% and 20–25% for dust. How- 2 SCOPE OF LABORATORY
ever, what remains a necessary precondition is the ASSESSMENTS
requirement for the mixtures to achieve character-
istics stipulated by the individual product standards Laboratory comparisons were made for the selected
even under the decreased production temperatures. ACsurf 11 mix, with known and optimised grading
Nowadays, attention is paid to functional testing curve, using basalt aggregate from the Měrunice
(Kriz et al. 2016; project Fun DBitS) which allows quarry. The mix was designed with 5.8 M% of
not only improved and more complex predictions paving grade bitumen 50/70, which was selected
of bitumen or asphalt mix behaviour. Its asser- as the basic binder to which the chosen synthetic
tion has been greatly assisted since the extensive wax-based additives were added to reduce the

335
production temperature of the control mix (150°C). Table 1. Tested values of empirical bitumen tests
For standard cylindrical Marshall test specimens
the asphalt mixing occurred at same temperature Softening Pen.
and number of blows. The mixing time for each of Penetr. point (R&B) index
Binder type [0.1 mm] [°C] [–]
the tested laboratory variants was the same as well.
Sets of basic empirical bitumen tests (softening 50/70 63.8 47.1 −1.4
point, penetration) and selected functional tests 3% A20K 46.9 79.4 4.1
(rotational viscosity, multiple stress creep recov- 3% E10K 47.5 50.9 −1.1
ery—MSCR test and complex shear modulus as 1.5% E10K + 1.5% RH 66.5 44.9 −1.9
well as phase angle determination) were performed 2% BIT 64.4 49.6 −0.7
for all bituminous binders. 3% BIT 61.5 51.9 −0.2
Selected binders were used to prepare ACsurf 11 2% T39A 42.3 66.3 1.8
mixture always with the same grading curve. The 3% T39A 42.9 75.5 3.3
version with bitumen 50/70 (no additives) mixed 2% T39A + 1% RH 49.1 64.9 1.9
at 150°C is marked as control/reference mix. The
test specimens were subsequently also compacted
using the same temperature. Asphalt mixtures with
added synthetic waxes were mixed under slightly point slightly when compared to bitumen 50/70.
lower temperatures (140°C) and compacted at The versions combining two waxes have a lower
125–135°C. softening point and increased penetration mainly
The standard empirical properties as well as due to the RH-wax effect.
functional mix characteristics focusing mainly on
3.1.2 Determination of rotational viscosity
durability, resistance to permanent deformation
Volumetric rotational viscosity is an important
and behaviour in the low temperature range were
characteristic for the description of bituminous
assessed for the asphalt mix versions designed.
binder workability. The characteristic is mainly
important from the point of view of bitumen
transport and pumping e.g. in the mixing plant.
3 EXPERIMENTAL TESTING AND
The rotational viscosity test was carried out
RESULTS
according to the standard EN 13302 using a spin-
dle rotational viscometer. The values determined
3.1 Assessed bituminous binders
with shear rate of 6.8 s-1 (the shear rate considered
3.1.1 Basic empirical characteristics the indicative parameter for bituminous binder
Basic paving grade bitumen 50/70 which met both workability assessment in the U.S.) at different
threshold values for penetration and softening temperatures were measured as expected, lower for
point according to the relevant standard EN 12591 all versions of binders with wax-based additives
was chosen for the assessment of synthetic wax on the test than for the reference bitumen. The
behaviour. Additives in a content of 2–3 M% (even viscosity reductions are more distinctive in most
for some combinations of two synthetic waxes) variants under lower temperatures; with respect to
were added to the bituminous binder and mixed the option of reducing compaction temperature of
for at least 15 minutes at 150°C. asphalt mixtures, this is a more important value.
Table 1 indicates the summary of empirical The lowest viscosity was achieved for binders with
test results. The most distinctive penetration drop 3% T39A and if a combination of waxes E10K
(by over 20d mm) was recorded for wax additive and RH was prepared, while similar results were
marked T39A (FT paraffin-based wax). In the case recorded by a version with a higher quantity of BIT
of using waxes A20K (amide based wax) and E10K as well. The effect of synthetic waxes on rotational
(polyethylene wax), penetration also dropped quite viscosity is shown in Figure 1. Measurements were
considerably (by over 15d mm). Contrastingly, the taken for other shear rates and for a temperature
versions containing RH wax (a Chinese product on range of 100–150°C, which allows plotting the so-
a polyethylene wax basis) recorded increased pen- called flow curves (Fig. 2). The curves show that
etration. The addition of the synthetic wax marked at the lowest temperature (100°C), not all versions
BIT does not affect penetration significantly. The have viscosities lower than the reference bitumen.
comparison of softening points shows that this A sharp decline of viscosity is noticeable in the
characteristic improved greatly for additive A20K temperature range of 100–120°C. An exception is
(by more than 30°C), which consequently deliv- the A20K wax where the viscosity only becomes
ers a high penetration index. Also additive T39A lower when compared to the reference binder from
increases this characteristic significantly, particu- 130°C; however, this is a finding already verified
larly with a greater content of wax in the bitumen. for other types of fatty acid amides, too (Butz
The remaining waxes only increase the softening 2008, Valentin 2014).

336
 0 ,5 0,48--------------------1~ Assessments of fatigue behaviour may use param-
~ eter G*x sin(δ).
0 ,4 l - - -0 , 3 6 - 0 3 8 - - - - " ' - - - - " 8 - - - - 0 ,36
Table 2 summarises selected values measured
during the oscillation test for the frequency of
1.59 Hz and temperature of 60°C. The highest
G* and the lowest δ values at the same time are
achieved by the binder with additive A20K which,
therefore, should be theoretically slightly more
50/70 3%A20K 3% EUJ< 1,5% El (){ 2%BIT 3%BIT 2% T 3gt>, 3% T 39A 2%T39A+ elastic and more resistant to permanent deforma-
·~~ mu
tion under higher temperatures. All three variants
with additive T39A have the complex shear modu-
Figure 1. Rotational viscosity of tested bituminous lus value higher than the 50/70 reference binder.
binders. The remaining variants have lower G* value when
compared to paving grade bitumen 50/70 which
indicates a potentially lower resistance to perma-
50{70 nent deformation. This corresponds with the qual-
1- - - l ,S" ElCI<.t 1,5% RH
ity indicators depicted in Figure 3; the minimum
2%BIT
3% 8 1T
value of 1.0 kPa for G*/sin(δ) as used for bitumen
2% T39A
performance grading assessment was met by all

-
3% T39A
2% T39A + 1% R.H
binders tested. The minimum limit was in pre-
~
1-- sented study used just as a base for comparing all
0,1

F" tested binders between each other, it is not a regu-

1--
0,01 larly required parameter. If focusing on the second
calculated indicator G*.sin(δ) it should be noted
T~mper11ture rCJ
that this fatigue measure should be determined at
intermediate performance temperature and pref-
Figure 2. Flow curves of tested bituminous binders in
semi-log scale (shear rate 1.0 s–1). erably for aged bitumen test samples. At 60°C

Table 2. Characteristics of complex shear modulus at

3.1.3 Determination of complex shear modulus 60°C.
and phase angle
The determination of the complex shear modulus G* δ G*/sin(δ) G*x sin(δ)
G* and phase angle δ of bituminous binders in the
medium and higher operation temperature range Binder type [kPa] [°] [kPa] [kPa]
uses the Dynamic Shear Rheometer (DSR) and
the measurements are governed by the standard 50/70 2.4 87.1 2.4 2.4
3% A20K 7.0 76.4 7.2 6.8
EN 14770. The usual oscillation test is taken under
3% E10K 2.0 84.7 2.0 1.9
constant shear stress within the selected load fre-
1.5% E10K + 1.5% RH 1.5 87.7 1.5 1.5
quency range and is termed the “frequency sweep”
2% BIT 2.2 84.8 2.2 2.2
(FS). The test is designed to verify the bituminous
3% BIT 2.0 84.9 2.0 2.0
binder behaviour under various temperatures and
2% T39A 3.1 82.0 3.1 3.0
frequencies which might simulate different com-
3% T39A 3.4 81.6 3.4 3.4
positions and rates of the repetitive stress which
2% T39A + 1% RH 3.1 82.0 3.1 3.0
might be encountered in pavements.
The main characteristic obtained, G*, particu-
larly describes material resistance to permanent 8,0 • s0/70
deformation and can define the fatigue behaviour 7,0
7,2
5,8
. 3% A20K
. 3%[10:.

of the material (Pavement Interactive 2011). The • t ,S%E1ct:t 1,5%R.H

5,0
deformation behaviour was in the past expressed
5,0 • 3%T39A
by G*/sin(δ) parameter. If the bituminous binder . 2%T3'lA t l%RH

has a high resistance to deformation, the character- •~ <,0

3,0
istic (G*) should have a high value while the phase
angle δ should has as low value as possible. The 2,0

1,0
ratio of G*/sin(δ) should then be as high as pos-
sible. The minimum value required for G*/sin(δ) 0,0
G 0 1tsi n(6j60"C

in case of bituminous binders was determined

within SHRP in the past as 1.0 kPa for unaged Figure 3. Quality performance characteristics accord-
bituminous binders and the PG grading system. ing to SHRP recommendations.

337
this measure is of course meaningless for fatigue elastic recovery and the irrecoverable shear compli-
and it was just added because 60°C is a standard ance Jnr in the bituminous binder. Elastic recovery
temperature at which testing on DSR should be (relaxing) is an indicator of the degree to which the
done in the Czech Republic for un aged binders. It sample recovers in the original shape, having been
was again used just to compare the tested binders stressed and relieved several times.
between each other. This test method is suitable mainly for the assess-
The transformation of the data measured to ment of behaviour under high temperatures as it
a selected specific temperature allows obtaining, is the best prediction of asphalt mix resistance to
with the application of time and temperature rutting. Great elastic recovery of modified binders
superposition principle, a complex curve known as results in smaller permanent deformation. Simul-
the master curve. This curve allows a good assess- taneously, such binder is expected to achieve the
ment of the deformation effect of individual bind- smallest value of irrecoverable shear compliance
ers within a very broad frequency range which are possible. The lower Jnr, the higher proportion of
even outside the functional range of the testing stress the binder is capable to absorb, which makes
device (rheometer). This allows to interpret vari- it less susceptible to plastic deformation. Based on
ous impact of traffic loads and intensities affecting the findings so far, stressed pavements are required,
the material in the pavement structure. if the bitumen is loaded by strain of 3.2 kPa and
The master curve in Figure 4 clearly shows temperature of 60°C, to have Jnr of the final value <
that higher values of the complex shear modu- 1 kPa–1. The test was taken in accordance with EN
lus (G*) are achieved by the binder with additive 16659 with shear stress of 0.1 kPa and 3.2 kPa and
A20K again, along with all versions containing the results presented are the average from measure-
T39A—this points to a higher binder stiffness ments taken on two samples at least.
throughout the entire frequency range. With high The aforementioned findings basically cor-
stress frequencies, the stiffness of the binders on respond with the results of the oscillation test.
test equal out; this means that the bitumen variant Again, the distinctively best values were achieved
with wax A20K has a lower thermal susceptibility by the bitumen containing A20K wax, which has
and, therefore, broader plasticity range than the
reference binder. This potential of the binder was
Table 3. Results of MSCR test for tested bituminous
obvious from the previous results already, and the binders.
master curve reconfirms its validity. Improvements
in the high temperature range should be achieved 0.1 kPa 3.2 kPa
by all binders tested, with the exception of the ver-
sion combining waxes E10K and RH, where no El. rec. Jnr El. Jnr
improvement is noticed, rather the contrary was Used binder [%] [kPa–1] rec. [%] [kPa–1]
shown.
50/70 1.16 5.59 0.05 6.58
3% A20K 78.38 0.11 3.16 2.12
3.1.4 Multi-Stress Creep Recovery test (MSCR)
3% E10K 5.05 5.19 0.37 7.54
The currently in Europe widely discussed multi-
1.5% E10K + 1.5% RH 1.20 10.46 0.02 11.22
stress creep recovery test of bituminous binder
2% BIT 1.61 7.50 0.08 7.37
samples is considered the most suitable method of
3% BIT 5.18 4.94 0.17 6.84
testing deformation characteristics of binders in
2% T39A 52.96 0.91 0.53 5.44
the high temperature range. The above applies par-
3% T39A 59.31 0.10 0.12 4.16
ticularly for polymer-modified bituminous bind-
2% T39A + 1% RH 62.75 0.54 0.74 4.66
ers. The test method determines the percentage of

12,0

10,0

~1,E+06
i 8,0
1-
H
• 3%A20K 6,0
• 3%E10K
• l.S% El OK + l ,S% RH
• 2%BIT _; ~
4,0
• 3%BIT ~

1,E+()2 u .l!till:]L_l_ill_t:!ll_
• 2%T39A
• 3%T39A
_l__Lll'!L_LWUll_l_lj_b;;'%~T~39A~
>l~%~R;;_H _jl
I 2,0

l,E-QS 1,E-Q4 1,E..03 1,E..02 l ,E..Ol l ,E..OO l ,E+Ql 0,0

Frequencyf[Hz]

Figure 4. Master curves of complex shear modulus of Figure 5. Non-recoverable creep (shear) compliance of
tested bituminous binders. tested bituminous binders at stress level of 3.2 kPa.

338
the highest elastic recovery values at the same time 3.2.2 Water susceptibility
as having the lowest non-recoverable shear com- Resistance to the effects of water immersion was
pliance. The version combining two waxes scored tested by the indirect tensile strength test and the
worst in both the characteristics; this suggests lim- calculation of ITS ratio in compliance with the
ited resistance to permanent deformation. method indicated in EN 12697-12. The method
The aforementioned very good correlation of was extended by another set of test specimens
results between the elastic recovery and complex which were exposed to a single freezing cycle in
shear modulus (G*) is depicted by Figure 6 with compliance with the conditions stipulated by the
determination coefficient R2 = 0.89. US standard AASHTO T283-3. This allows us
not only to take into consideration the effects of
water but also the combined effect of water and
3.2 Assessment of asphalt mixtures
frost and, therefore, the subsequent resistance of
3.2.1 Basic volumetric characteristics the specimens to such effects. As generally known,
The voids content in the individual asphalt mix- the test result is a ratio of indirect tensile strength
tures was determined on the basis of compacted of a group of saturated specimens to the indirect
specimen bulk density and maximum bulk density tensile strength of a group of dry specimens (ITSR
of loose asphalt mix according to EN 12697-8. The value).
individual values are listed in Table 4. The voids Figure 7 summarises the final values of both
content of ACsurf 11 asphalt mixtures should range methods. According to the EN method, the high-
from 2.5 to 4.5% in compliance with the require- est ITSR value was achieved by the reference mix
ments of the relevant product standard. This range while all versions with waxes scored less. The
was met only by the control mix which was com- required indirect tensile strength ratio according to
pacted in the laboratory at 150°C. The remaining the Czech national appendix of EN 13108-1 stand-
variants have lower voids content than the required ard is at least 70% for an ACsurf 11 mix; however,
value despite the fact that they were compacted by the ongoing review of harmonised standards could
temperatures of 15°C to 25°C lower. This suggest make this requirement more stringent at 80%.
that the compaction temperature can be theoreti- All of the variants assessed exceeded even this
cally even more reduced or that there is a larger threshold which means that there is no tendency
compaction window. of decreasing indirect tensile strength due to the
effects water (frost).

3,5 3.2.3 Determination of stiffness modulus

3,0 • Stiffness modules of tested asphalt mixtures were
determined on cylindrical test specimens, apply-
-2,5
/
ing the test method according to EN 12697-26
!!!.
~ 2,0 y =0,5772x - 1,1262 / (Annex C)—non-destructive repetitive indirect
~
~ 1,5
R2 = 0,8 922
/ tensile stress test on cylindrical test specimens (IT-
/ CY). The stiffness modulus describes the ability of
i:U 1,0

0,5
v.
r a material to resist stress. The test was taken for
three selected temperatures: 0°C; 15°C and 27°C.
0 ,0
0,0 1,0 2,0 3,0
• 4,0 5,0 6,0 7,0 8 ,0 The values of determined stiffness modules are
G• [kPa] indicated in Table 5. The assessment of the defor-
mation behavior in the moderate temperature
Figure 6. Relationship between complex shear modulus
and elastic recovery.
1,4

1,2

Table 4. Basic volumetric characteristics of tested

asphalt mixtures. zo,s
~ 0,6
Bulk Max. Comp.
0,4
Used density density Voids temp.
0,2
binder [g. cm–3] [g. cm–3] [%] [°C]
50/70 3% A20K 3%E10K 1,5%El 0K + 1,5% RH
50/70 2.614 2.699 3.15 150 - EN ~'-~ AASHTO - Minimum valueiTSR(according toEN)
3% A20K 2.652 2.712 2.18 135
3% E10K 2.629 2.694 2.44 125
1.5% E10K + 1.5% RH 2.648 2.713 2.41 125 Figure 7. Results of ITSR tests according to two differ-
ent approaches.

339
Table 5. Determined stiffness modules for selected vari- correlation must be approached with a certain level
ants of ACsurf 11 mixture. of caution.
Stiffness @ [MPa] Thermal 3.2.4 Bending beam (flexural) strength in the low
Used susceptibility
binder 15°C 0°C 27°C S0/S15
temperature range
The flexural strength test was conducted in compli-
50/70 6.468 15.111 – 2,34 ance with the procedure stipulated by Czech techni-
3% A20K 8.649 23.158 2.549 2,68 cal specifications TP 151 for asphalt concrete with
3% E10K 9.858 23.735 2.820 2,41 a high stiffness modulus. The test can be charac-
1,5% E10K 7.192 22.092 1.961 3,07 terized as a simple 3-point bending beam test with
+ 1,5% RH controlled loading rate. The levels of specimen
stress rates were selected according to the specifica-
tions (50 mm/min and 1.25 mm/min) while the test
itself was taken at 0°C. Test specimens obtained
from the slabs intended for the test of asphalt mix
resistance to permanent deformation were used for
determining the flexural strength.
ACsurf 11 type of mix does not generally require
the characteristic of flexural strength. Therefore,
no threshold parameter to be achieved by a mix
is stipulated. In compliance with TP 151, a mini-
mum value is only required for asphalt mixtures
of HMAC type. In this case, for the stressing rate
of 1.25 mm/min, a minimum value of 6 MPa is
required for such mixtures. As is shown by the
results presented below, this requirement is met
Figure 8. Found relationship of asphalt stiffness and by both the reference asphalt mix and the variants
voids content. with a combination of E10K and RH synthetic
waxes. The remaining two variants have very simi-
range is given by the values measured at 15°C; this lar flexural strength values although the ratio of
temperature is also considered as the crucial from stressing rates for 50 mm/min and 1.25 mm/min is
the point of view of pavement structure design significantly higher than that of the control mix. In
calculation according to the technical specifica- the case of the lower stressing rate, both variants
tions of the Czech Ministry of Transportation, demonstrate a slight deterioration which reflects,
Design Manual TP 170. From the performed tests to a certain degree, the occasional fears that syn-
it is obvious that the highest stiffness modules for thetic waxes in asphalt mixtures could eventually
all test temperatures were achieved by the asphalt result in slightly increased brittleness which is
mix with additive E10K (more than 50% increase driven by higher stiffness, therefore causing a par-
if compared with control mix). However, even the tial deterioration of its characteristics in the low
remaining mixtures demonstrated increased stiff- temperature range.
ness modules when compared to the reference mix.
The ratio of stiffness modules at different tem- 3.2.5 Resistance to permanent deformation
peratures can be understood as a thermal suscep- (rutting)
tibility indicator of the asphalt mix. The lower the Resistance to permanent deformation was deter-
S0/S15 (or S0/S27) ratio, the less susceptible the mix is mined by the wheel tracking test applying the pro-
to temperature changes. The least susceptibility is cedure stipulated by the standard EN 12697-22.
demonstrated by the reference mix, while the high- The test utilized a small test device with an air bath
est susceptibility is shown by the mix variant com- and test temperature of 50°C during the application
bining two different waxes. of 10,000 loading cycles. The WTSAIR characteris-
The stiffness modulus value is significantly tics (average value of rut depth increment—wheel
affected by the voids content, as is obvious from tracking slope), PRDAIR (proportional rut depth
Figure 8 which depicts the dependence of the for the material on test after 5,000 cycles) and the
two characteristics. Their interdependence is lin- average rut depth after 10,000 loading cycles were
ear with a very strong correlation. On the other determined. For the assessed ACsurf 11 asphalt
hand, it must be emphasized that this statisti- mix, the national annex of the EN 13108-1 prod-
cal evaluation has been made with a very limited uct standard requires WTSAIR and PRDAIR values
data set (only four asphalt mixtures and maximum to be declared only. If we would apply the criteria
6 test specimens per mixture) and, therefore, the used presently for a superior mixture ACsurf 11S,

340
WTSAIR is defined as 0.07 mm/1000 cycles and 3,5 . - - - - - - - - - - - - - - - - - - - - -

PRDAIR has the limit value of 5.0%. The results 3 ,0 +------*--

~--=---------
c-------­
obtained for the assessed mixtures are summarized
in Table 6 below. ~ 2,5 ......._____.
=-----
+ - - - - - - - - - - - - - --=..,....
It is obvious that according to the criteria for 12,0 v = -Q, 1 7 93x + 3,8492
ACsurf 11 (superior) mix, none of the tested mix- ~ 1,5 +-------- Rz = 0,9931

tures with added wax would pass such criteria. ~ 1 ,0 +---------------------

With respect to previous test results and the deci- 0 ,5 +---------------------
sion made during asphalt mix sample preparation 0 ,0 +----~---~---~---~---,
(using same compaction temperatures) it is highly 0,0 2,0 4,0 6,0 8,0 10 ,0

questionable if this test is suitable to describe PRD.4m [%)

the deformation behavior of the assessed type of

asphalt mixtures. Figure 10. Found relationship between the PRDAIR and
If we compared the individual variants to one the mixture voids.
another, the worst resistance was recorded by the
asphalt mix with wax A20K, which has ruts of
double the depth of the reference mix. Such result material in the gaps when ruts are formed. It is
are not consistent with previous results gained by pressed outwards and plastic deformation occurs.
CTU in Prague for asphalt mixtures where the The very strong dependence of permanent defor-
bitumen was improved by amide waxes. Usually mation on the asphalt mix voids content is depicted
the application of this type of wax leads to a very in Figure 10 (R2 = 0.99). As already mentioned this
low rutting. does not agree with stiffness results which raises the
Resistance to permanent deformation should question if one of the tests is suitable for describ-
depend on the properties of the bituminous bind- ing the behaving of this type of asphalt mixture.
ers used; however, when compared to the DSR test
results, it is evident that the opposite is the case.
In our opinion, one of the reasons is the lower 4 CONCLUSION
voids content of these mixtures. Asphalt mixtures
with synthetic waxes in the binder show a very low For more than fifteen years, warm asphalt mixtures
voids content which rules out any capacity to press and the related low-viscosity bituminous bind-
ers present developing road construction trends
driven by health and safety issues as well as the
Table 6. Results of wheel tracking test of assessed potentials of reducing energy demands. Neverthe-
asphalt mixtures. less, attention must be paid both to the asphalt mix
properties and, even more, to the characteristics
Ø rut depth of bituminous binders used and modified by the
WTSAIR PRDAIR after 10,000 effects of additives applied to improve binder vis-
Used binder [mm/1000] [%] cycles [mm]
cosity. Results in this paper focused on new types
50/70 0.062 4.0 2.31 of such additives.
3% A20K 0.166 9.5 4.61 The experimental study assessed firstly the
3% E10K 0.119 7.6 3.62 impact of selected synthetic waxes on the prop-
1.5% E10K 0.126 8.0 3.84 erties of bituminous binders. To analyze properly
+ 1.5% RH the effects of these additives on bitumen and pos-
sible alteration of the bitumen properties basic
as well as selected functional tests have been run.
To get a complex understanding it would be even
necessary to determine the bitumen behavior in
low temperature range, preferably using bending
beam reometer. This was not assessed so far. In
the second step selected characteristics of asphalt
mixtures of same grading were always determined.
The measurements taken clearly show that the
performance-based properties of asphalt mixtures
50/70 3%A20K 3% E10K 1,5% El0K + l ,S%RH strongly depend on the voids content a mix reach,
• 50 mm.min~l ~ 1.25 mm.min·1 while being significantly less determined by the
properties of the bituminous binders per se. How-
Figure 9. Flexural strength of assessed asphalt ever, this in turn means that the bituminous binder
mixtures. viscosity plays an important role again. It would

341
be therefore crucial to set for each mix variant Cápayová, S. et al. 2016. Road Construction Materi-
(using different wax) the mixing and compaction als from the Perspective of Energy Savings. SGEM
temperature according to the viscosity. Results 2016. 16th International Multidisciplinary Scientific
presented in this paper considered the same com- GeoConference. Book 4. Energy and Clean Tech-
nologies: Conference proceedings. Albena, Bulgaria.
paction temperature for all mixture and the results pp. 127–132.
are driven not by the same viscosity level but by the D’Angelo, J. et al. 2008. Warm-Mix Asphalt: European
same compaction temperature. This influences of Practise. American Trade Iniatives. Washington, DC:
course the results and their comparability. All vari- U.S. Department of Transportation.
ants of tested mixtures with synthetic waxes had Dynamic Shear Rheometer. Pavement Interactive
lower voids content than the value required by the [online]. 2011 Available from:http://www.pavementin-
product standard for ACsurf 11 mixture type; due to teractive.org/article/dynamic-shear-rheometer/
that, it is more than likely that the remaining prop- Hamzah, M.O. et al. 2016. Recycled Asphalt Pavement
erties were significantly affected by the lower voids with Warm Mix Additive for Sustainable Road Con-
struction [online]. International Journal of Civil, Envi-
content. The question remains whether the indi- ronmental, Structural, Construction and Architectural
vidual characteristics would be more influenced by Engineering. 10(3), pp. 296–299.
the bituminous binder properties or not in the case Hýzl, P. & Varaus, M. 2007. Funkční zkoušky asfaltových
of the correct level of voids content. směsí. Stavebnictví 06-7 (2007). Available from:
The application of synthetic waxes proved the http://www.casopisstavebnictvi.cz/funkcni-zkousky-
potential of reducing bituminous binder viscos- asfaltovych-smesi_N278
ity and the related lower temperature necessary Kakar, M.R. et al. 2015. A review on moisture damages
for correct asphalt mix compaction. Although the of hot and warm mix asphalt and related investiga-
durability of asphalt mixtures determined by the tions [online]. Journal of Cleaner Production. 99pp.
39–58.
ITSR is slightly lower than that of the control mix, Kriz, P. et al. 2016. Assessment of performance specifica-
it still meets the requirements of the relevant prod- tion tests for rheologically simple bitumens. 6th Euras-
uct standard. The stiffness of asphalt mixtures phalt & Eurobitume Congress, Prague 1.-3.6.2016
with waxes was higher than that of the control mix Soukupová, L. & Valentová, T. 2016. Porovnání vlast-
in all cases presented, what is an expected result. ností asfaltových směsí ACO11+ s 30% R-materiálu a
On the other hand, the mixes with synthetic waxes užitím vybraných typů rejuvenátorů, Juniorstav, VUT,
demonstrated significantly lower resistance to per- Brno.
manent deformation, most probably because of Valentin, J. & Soukupová, L. 2014. Posuzované varianty
too low voids content. asfaltových pojiv vhodných pro nové generace asfal-
tových směsí typu VMT. CESTI competence centre –
Partial Research Report, CTU Prague.
Valentin, J. et al. 2015. Odborná zpráva o postupu
ACKNOWLEDGEMENTS prací a dosažených výsledcích za rok 2015—projekt
WMARec, Výzkumná zpráva, ČVUT, CIIRC, Praha,
This paper was supported by the project No. 85 s.
TA04031255 (Development of functional bitumi- Valentin, J. et al. 2015. Vliv vybraných přísad pro níz-
nous binders war warm mix asphalts with increased koteplotní směsi na běžná PMB pojiva používaná u
content of reclaimed asphalt) which is financed by SMA směsi. Konference Asfaltové vozovky, Sdružení
the Technology Agency of the Czech Republic. pro výstavbu silnic, České Budějovice.

REFERENCES

Butz, T. 2008. Warm Mix Asphalt Pavements based on

the Fischer—Tropsch Wax Sasobit. Pavement perform-
ance prediction symposium. Wyoming 16.-18.7.2008.

342
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

An investigation of the mechanical properties of rubber modified

asphalt mixtures using a modified dry process

Ayad Subhy, Gordon D. Airey & Davide Lo Presti

Nottingham Transportation Engineering Centre, University of Nottingham, Nottingham, UK

ABSTRACT: Incorporating recycling tyres into asphalt mixtures contributes to solving a serious envi-
ronmental problem as well as the potential of improving the mechanical performance of asphalt mixtures.
The recycled tyre rubber is added by two technologies known as the dry process (when rubber particles are
treated as an aggregate modifier) and the wet process (when rubber particles are treated as a binder modifier).
Although, the dry process is easier and requires minimal or no modification in the asphalt plant, the incon-
sistency in field performance makes the dry process not widely used. The main reason for this is believed to
be the lack of interaction between the rubber particles and binder. In order to overcome this issue fine rubber
particles were used to enhance the interaction with the binder and reduce the springy nature of rubber par-
ticles that poses difficulties during the compaction process of mixtures. The mechanical performance-based
properties for the control mixture (without adding rubber) and other different rubber asphalt mixtures were
evaluated and characterised in terms of their fatigue resistance using the Indirect Tensile Fatigue Test (ITFT),
rutting resistance using the Repeated Load Axial Test (RLAT). The stiffness modulus for the different mix-
tures was also measured by the means of Indirect Tensile Stiffness Test (ITSM). The results have shown that
asphalt mixtures produced with the addition of crumb rubber using fine crumb rubber are able to reduce the
permanent deformation rate as well as produce improvements in the fatigue cracking at lower strains levels.

Keywords: asphalt mixture, dry process, wet process, rutting, fatigue

1 INTRODUCTION two technologies known as the Wet process and

Dry process. In the wet process, ground rubber
Millions of scrap tyres are generated each year is blended with the bitumen at high temperatures
across the world. In the UK alone, around 30 mil- before introducing to the aggregate; while in the
lion are produced each year (Rahman, 2004, Lee dry process, the recycled tyres rubbers are added
et al., 2007, Huang et al., 2007). These scrap tyres by replacing a small part of the aggregate in the
pose a serious landfill problem because of their large asphalt mixture. Fine rubber particles (0.075 to
size and the fact that they are a non-biodegradable 1.2 mm) are normally used with the wet process
waste. The environmental impacts of the used tyres while coarser particles (0.4 to 10 mm) are used
have driven research towards recycling them into the with the dry process (Rahman et al., 2010).
different construction applications as a sustainable Although larger quantities of recycled tyre
practice for protecting the environment and also for rubber are consumed in mixtures produced using
enhancing the mechanical properties of materials. the dry process and with minimal or no modifica-
Bituminous materials constitute the main ele- tion required in asphalt plant, the inconsistency
ments of a flexible pavement. Thus, incorporating in field performance makes the dry process not
recycled tyre rubber into bituminous materials would widely used and increasingly being abandoned
open a large market for the use and recycling of (Rahman, 2004). The rubber particles, in terms of
scrap tyres in the pavement industry (Lo Presti, 2013, the dry process, would keep their physical shape
Subhy et al., 2015a). Additionally, this approach has and properties within the asphalt mixture due to
been broadly recognised to improve the perform- poor interaction and lack of adhesion with the
ance characteristics and increase the service life of base bitumen. This, in turn, causes instability to
pavements in comparison to conventional materi- the volumetric properties of the matrix because
als (Ruth and Roque, 1995, Wang et al., 2012, Paje of the continuous swelling of rubber particles
et al., 2010, Cao, 2007, Chiu and Lu, 2007, Subhy after compaction leading to stripping and prema-
et al., 2015b). ture cracking on the asphalt pavement surface.
The introduction of rubber particles into road There are two common procedures used in
asphalt mixtures are generally accomplished by the industry for the dry process; PlusRide and

343
Generic dry process. PlusRide was developed in 2 MATERIALS, MIXTURE PRODUCTION
the United States in the late 1960’s by modify- AND EXPERIMENTAL PROGRAMME
ing the gap-graded mixture, the modification is
achieved by replacement of up to 3% of aggregate A straight-run bituminous bitumen has a penetra-
by a coarse granulated crumb rubber. The interac- tion of 40 dmm and a softening point of 51.4°C
tion between crumb rubber and bitumen is kept was used for manufacturing the asphalt mixtures.
at a minimum level so the rubber particles main- The coarse and fine aggregate fractions used in
tain their physical and elastic properties within the this study consisted of Granite aggregate obtained
asphalt mixture which are important to modify from Bardon Hill, UK. The filler was a lime-
the stability of a gap-graded mixture (Rahman, stone obtained from Foster Yeoman (Torr Works
2004). Generic dry process or also known as the Quarry, Somerset, UK). The two different crumb
“TAK” system, on the other hand, was developed rubbers that were used to modify the asphalt mix-
in the late 1980’s by modifying dense-graded and tures are, ambient ground rubber, labelled ‘A’,
gap-graded mixtures (Takallou and Hicks, 1988). (0–0.6 mm) obtained through mechanical grind-
Both coarse and fine rubber are added to the ing, and another ambient, labelled ‘B’, (0–0.8 mm)
dense graded mixture by a percentage up to 3% of obtained also through mechanical grinding but
total mixture mass. with chemical modification of the surface. A
Hernández-Olivares et al. (2009) were probably Stone Mastic Asphalt (SMA) gradation (10 mm)
the first and only to modify asphalt mixtures, by was used to manufacture the different mixtures,
the means of dry process, using only fine crumb the gradation of the aggregate skeleton is shown
rubber (less than 1 mm). In their research, they in Figure 1.
characterised and evaluated, in the laboratory as For mixtures modified with crumb rubber
well as in field sections, mixtures containing differ- using the dry process, the designed rubber quan-
ent increasing percentages of crumb rubber which tity was added to the aggregate before adding the
were fabricated using different methods. The main designed bitumen. The rubber particles were stored
conclusion of this study was that higher rubber at room temperature before added to the aggre-
quantities could produce detrimental effects in the gate. The mixture was mixed using the mechanical
agglomerate: higher air voids percentage, compac- mixer and then placed in a preheated slab mould
tion difficulties in the field and rubber segregation. (306 × 306 mm) and compacted by a smooth steel
Furthermore, using finer crumb rubber at a small roller, as shown in Figure 2, according to BS EN
content is recommended to produce asphalt mix- 12697-33:2003, until the desired final height of the
tures that can be easily handled (mixed and com- slab (∼60 mm) that gives a 4% targeted air voids, is
pacted) in similar way to a conventional asphalt achieved.
mixture. The mixing and compaction temperatures of
In this study, a typical stone mastic gradation bituminous mixtures are normally specified based
(10 mm) suitable for surface courses was selected on their binder viscosities. For example, the selected
from the British specification BS EN 13108-5/PD mixing and compaction temperatures should cor-
6691:2007 for designing the conventional and rubber respond to binder viscosity of 170 ± 20 mPa.s and
modified mixtures. The conventional asphalt mix- 280 ± 30 mPa.s, respectively (Asphalt Institute,
tures were modified by a replacement of only 10% 2003). However, this criterion is not always possi-
of bitumen with different tyre rubbers. The replace- ble to apply to rubber modified mixtures using the
ment of rubber was considered as a replacement of wet process because unreasonably high mixing and
bitumen and not as a replacement of aggregate to compaction temperatures are predicted with this
maintain the cost effectiveness of rubber modified method. For instance, the mixing and compaction
mixtures. Two different recycled tyre rubbers were
used in the current experiment; one is a straight
ambient produced crumb rubber, and the second is 100

also obtained through ambient mechanical grind- 90

80
ing but their surfaces were chemically treated to
70
promote more interaction with bitumen. With the
~ 60 ''
I
aim of investigating the performance of fine rubber ~ 50
I
,' : Combined
modified asphalt mixture using the dry process, the ~ , '/ ; ~rtf,!alr used
40
rutting and fatigue cracking resistance in addition
'~!- ,'Vf
::::::::-::::::::~---~''-~---~
30
to the stiffness were characterised in the laboratory. 20
This paper also presents a comparison between the 10
0
results of conventional asphalt mixtures and rubber 0.01 0.1 10 100
modified mixtures using the dry process, on the one Sieve Size (mm)
hand, and rubber modified bitumen using the wet
process on the other. Figure 1. The 10 mm SMA gradations.

344
 -- ~ -- ~ ---- ~ --- ~ -

i 1---- --
- lSOO

24SO
i.f
•.1
J i

Figure 3. Gravimetric proportion and maximum densi-

Figure 2. The mechanical mixer (left) and steel roller ties of the different mixtures.
(right) used for asphalt mixture production.

temperatures associated with the mentioned vis-

cosities would be higher than 220°C. This higher
temperature is not acceptable, it raises concerns
about workers’ health due to possible hazardous
fumes in addition to possible thermal separation
and oxidation. Therefore, the mixing and com-
paction temperatures were selected so that the
aggregate fractions can be sufficiently coated by
the binder and practically compacted to the pre-
scribed voids content. The mixing and compaction
temperatures were specified for wet process rubber
modified mixtures as 190 ± 5°C and 170 ± 5°C, Figure 4. ITSM testing configuration and test
respectively. For control mixtures, and dry proc- conditions.
ess modified mixtures, the mixing and compac-
tion temperatures were specified as 170 ± 5°C and tensile mode, for stiffness determination. The test
150 ± 5°C, respectively. parameters were applied according to BS EN 12697-
The binder content of the control mixture was kept 26:2004. The ITSM is calculated from the following
the same as recommended in the specification which equation by applying an impulse loading to induce
is 6.2%. For rubber asphalt mixtures (dry process), small horizontal strains of 5 ± 2 μm.
10% of rubber by the mass of bitumen was added to
the mixtures, this means that about 0.6% of rubber by P( + )
the mass of total mixtures was added. Thus, the bitu- ITSM = (1)
men content for rubber asphalt mixtures (dry proc- δt
ess) was 5.6% instead of 6.2%. For mixtures modified
with crumb rubber using the wet process, the crumb where; P = applied load, t = specimen thickness,
rubber ‘A’ at concentration of 15.25% by weight, were δ = horizontal deformation and υ = Poisson’s ratio.
blended with the same straight-run bitumen at tem-
perature 180°C for 90 min using the Silverson L4RT 2.2 Rutting-related properties
High Shear Mixer. A stabiliser is normally added to
the SMA mixtures to prevent drain down of the bitu- The RLAT is performed according to BS DD
men. Thus, cellulose fibres at 0.3% of the bitumen 226 using the Nottingham Asphalt Tester (NAT)
were included in the control mixtures. Cellulose fibres machine. In this test, a load pulse consists of a
were not included into the rubber asphalt mixtures. square wave with a frequency of 0.5 Hz (one sec-
Five specimens with 100 mm diameter were cored ond loading followed by one second rest period),
from each slab. The cores were then trimmed from is applied by an actuator. Figure 5 shows the con-
each end to produce cylindrical samples of 100 mm figuration and test parameters of the RLAT.
diameter and 40 mm thickness suitable for indirect
tensile tests and RLAT. The theoretical mass propor- 2.3 Fatigue-related properties
tions of the compositions of asphalt mixtures are
presented in Figure 3. The ITFT is the most widely test used in the United
Kingdom to evaluate the fatigue properties of bitu-
minous mixtures. The test is implemented according
2.1 Stiffness modulus (ITSM) to EN 12697-24:2012 (E) using the test parameters
The Nottingham Asphalt Tester (NAT), as shown in presented in Figure 6. The tensile strain at the cen-
Figure 4, was used for testing samples in the indirect tre of the specimen is calculated as following:

345
 service temperatures due to the traffic load).
Furthermore, another important parameter that
Test dur.-tlon: 7200 seconcb (3600 cydn)
allows characterisation of the mixtures for design
purposes is the Stiffness Modulus.

3.1 Indirect Tensile Stiffness Modulus (ITSM)

The stiffness modulus is an important indicator for
asphalt mixtures, and it is considered the main input
property to determine the required layer thickness
in the mechanistic pavement design. According to
Figure 5. RLAT testing configuration and test
conditions.
BS EN 12697-26:2004, the ITSM is taken as the
mean of two measurements on one specimen by
rotating 90°± 10° about its horizontal axis. Figure 7
shows the ITSM results for the Control and rubber
modified mixtures. The average value of 10 sample
for each mixture is presented in Figure 7, and the
range bars represent the maximum and minimum
of values of ITSM. It can be seen that there is not
a significant difference among the ITSM values of
asphalt mixtures A-dry and B-dry. This indicates
that pre-treatment of rubber surface did not affect
the stiffness of mixtures while the size and grada-
tion of the mineral aggregate skeleton in addition
Figure 6. ITFT testing configuration and test to the binder content had a dominant effect on the
conditions. ITSM. Because of the mixtures A-dry and B-dry
shared lower binder content and consequently thin-
2δ ( ν) ner binder film compared to the Control and A-wet
εο = (2) mixtures, there is a clear increase in the ITSM
d ( + πν − π ) values that can be seen for those mixtures. This is
expected as part of the binder which is highly flex-
where; ε ο = the tensile strain, δ = the horizontal ible material, is replaced by rubber particles.
deformation (mm), ν = the Poisson’s ratio and
d = the specimen diameter (mm). If ν = 0.35, then
3.2 Repeated Load Axial Test (RLAT)
2.1δ
εο = (3) The typical results obtained from the RLAT for the
d different mixtures are shown in Figure 8, where per-
The initial tensile strain is an important param- manent axial strain is plotted against load cycles.
eter for determining the fatigue life of different This kind of plot generally gives three distinctive
mixtures. Therefore, the fatigue life of bitumi- phases of the material. In the first or primary phase,
nous mixtures can be represented by the following the accumulation of vertical strain increases rapidly
equation: due to a combination of adjustment of loading
platens and material densification. The second-
n ary phase takes place when the strain rate reaches
⎡1⎤
Nf k⎢ ⎥ (4)
⎣ εο ⎦

where; N f = the fatigue life, k and n = material

constants and ε ο = the initial tensile strain at the
centre of the specimen.

3 RESULTS AND ANALYSIS

Performances of asphalt mixtures for road pave-

ments are generally associated with two main
damage phenomenon, namely rutting (permanent
deformation at high-service temperatures) and
fatigue cracking (cracks developing at medium Figure 7. ITSM results for the different mixtures.

346
 surface only marginally promoted the rutting resist-
ance as can be seen in Figures 9 and 10.

3.3 Indirect Tensile Fatigue Test (ITFT)

The fatigue data for asphalt mixtures is generally rec-
ommended to be plotted against the strain criterion
(Airey et al., 2006). Results of ITFT for each mixture
o +---~--~--~------~--~--~--~ are shown in Figures 11 in terms of the relationship
0 500 1000 1500 2000 2500 lOCO )SO() 4000
Numb~• of pulses
between the resilient strain at the 100th load applica-
- conuol - A-dry - B·dry - A·wet tion and fatigue life (Nf) in logarithm scale. The equa-
tions of fatigue curves shown in Figure 11, are used
Figure 8. RLAT results of different mixtures. to predict the fatigue lives for 50 and 100 με, as pre-
sented in Figure 12. The fatigue results demonstrate
that the addition of rubber using the dry process has
approximately a steady state response, and the
strain increases approximately in proportion to the
load cycles. The tertiary phase starts when the rate
of deformation increases rapidly indicating speci-
men failure; however, all the mixtures considered in
this study have not reached the tertiary phase.
The cumulative axial strain at the end of the 3600
load pulses or at the initiation of tertiary phase, and/
or, the slope of the steady state phase, have been used
to distinguish between better performing materials.
In this study, the slope of the steady state phase
is determined from a segment between 1500 to
Control A-dry B·dry A· wet
3000 pulses as follows;
Figure 9. RLAT results in terms of total strain.
i rate ⎡ με cycle ⎤
l ⎥⎦ =
Minimum strain
⎢⎣
ε 3000 ε1500
× 10 −6 (5)
1500
where; ε3000 = accumulated strain at 3000 pulses
ε1500 = accumulated strain at 1500 pulses
Figure 9 shows the permanent deformation
results in terms of the total accumulated strain at
the end of 3600 pulses, and Figure 10 shows the rut-
ting results in terms of the minimum strain rate. The
range bars represent the maximum and minimum
values for replicates. The results clearly confirm the
enhanced rutting performance of rubber modified
Control A·dry ..... A·we t

mixtures in comparison to the Control mixture. The

Figure 10. RLAT results in terms of the minimum
results also show that using the wet process can sig- strain rate.
nificantly improve the rutting resistance of mixtures
where the mixture A-wet clearly outperformed all
1000 .-------------------------------,
other mixtures. The effect of adding the rubber using & Control
the dry process was clearly revealed as an enhance- • A·dry
• B-dry
ment modifier in the minimum strain rate while this
effect vanished in the total strain. However, the mini- 0

mum strain rate is considered more fundamental to ~ 100

reliably characterise the rutting resistance of mix-

1
tures than the total strain. The total strain is largely
affected by the initial strain at the beginning of the
test and it is more sensitive to the initial conditioning 10 L - - - - - - - - - - - - - - - - - - - !
1.E+03 1.E+04 1.E+05 1.E+06
of the test apparatus as well as the different orienta- Cycles to failure
tion of the aggregate within the mixtures for the dif-
ferent specimens. Again, the pre-treatment of rubber Figure 11. ITFT fatigue lives versus strain.

347
 l .E+07
tion and Scientific Research of Iraq and the
Iraqi Cultural Attaché in London for the PhD
scholarship.

IIII
REFERENCES

l .E+04
Airey, G., Liao, M. & Thom, N. Fatigue behaviour of
bitumen-filler mastics. 10th International Conference
on Asphalt Pavements-August 12 to 17, 2006, Quebec
l .E+Ol
Control A·dry B·dry
City, Canada, 2006.
ASPHALT Institute 2003. Superpave Mix Design Super-
• 100 mlcrostraln • 50 mlcrostraln
pave Series No. 2 (SP-2), Lexington, KY.
Cao, W. 2007. Study on properties of recycled tire rub-
Figure 12. Fatigue lives at 100 με and 50 με. ber modified asphalt mixtures using dry process. Con-
struction and Building Materials, 21, 1011–1015.
somewhat decreased the fatigue lives of mixtures in Chiu, C.-T. & Lu, L.-C. 2007. A laboratory study on
stone matrix asphalt using ground tire rubber. Con-
comparison to the Control at strains higher than
struction and Building Materials, 21, 1027–1033.
100 με. However, the dry process mixtures have bet- Hernández-Olivares, F., Witoszek-Schultz, B., Alonso-Fern-
ter fatigue lives than the Control mixture at 50 με and ández, M. & Benito-Moro, C. 2009. Rubber-modified
lower. On the other hand, the wet process modifica- hot-mix asphalt pavement by dry process. International
tion has resulted in higher fatigue lives and that is Journal of Pavement Engineering, 10, 277–288.
particularly evident at low strain levels. Huang, Y., Bird, R.N. & Heidrich, O. 2007. A review of
the use of recycled solid waste materials in asphalt
pavements. Resources, Conservation and Recycling, 52,
4 CONCLUSIONS AND 58–73.
RECOMMENDATIONS Lee, S.-J., Amirkhanian, S.N., Putman, B.J. & Kim, K.W.
2007. Laboratory study of the effects of compaction
A stone mastic gradation (10 mm) was selected on the volumetric and rutting properties of CRM
for manufacturing the rubber modified mixtures asphalt mixtures. Journal of Materials in Civil Engi-
using the dry and wet process. Replacing only 10% neering, 19, 1079–1089.
of bitumen by fine rubber particles by the means Lo Presti, D. 2013. Recycled tyre rubber modified bitu-
mens for road asphalt mixtures: a literature review.
of dry process allows mixing and compacting the
Construction and Building Materials, 49, 863–881.
asphalt mixtures akin to conventional mixtures. On Paje, S., Bueno, M., Terán, F., Miró, R., Pérez-Jiménez,
the other hand, the mixing and compaction proc- F. & Martínez, A. 2010. Acoustic field evaluation of
ess of wet process rubber modified mixtures were asphalt mixtures with crumb rubber. Applied Acous-
reasonably accomplished by considering higher tics, 71, 578–582.
mixing and compaction temperatures. Rahman, M., Airey, G. & Collop, A. 2010. Moisture sus-
In general, the addition of rubber can produce ceptibility of high and low compaction dry process
bituminous materials with enhanced rutting char- crumb rubber-modified asphalt mixtures. Transporta-
acteristics. This is especially marked for rubber tion Research Record: Journal of the Transportation
Research Board, 121–129.
modified mixtures using the wet process.
Rahman, M. 2004. Characterisation of dry process
The fatigue cracking resistance of rubber modi- crumb rubber modified asphalt mixtures. University of
fied mixtures was significantly enhanced at lower Nottingham.
strain levels which makes them more suitable for Ruth, B.E. & Roque, R. Crumb rubber modifier (CRM)
perpetual pavement. Although the rubber modi- in asphalt pavements. Transportation Congress, Vol-
fied mixtures were produced with lower bitumen umes 1 and 2: Civil Engineers—Key to the World’s
content in comparison to the control mixtures, the Infrastructure, 1995. ASCE, 768–785.
results have shown improvements in the perform- Subhy, A., Lo Presti, D. & Airey, G. 2015a. An investiga-
ance of those mixtures. Thus, the addition of fine tion on using pre-treated tyre rubber as a replacement
of synthetic polymers for bitumen modification. Road
rubber can be considered a cost effective option.
Materials and Pavement Design, 16, 245–264.
Finally, the mechanical properties and perform- Subhy, A., Lo Presti, D. & Airey, G. 2015b. Rubberised
ance of mixtures produced using the pre-treated bitumen manufacturing assisted by rheological meas-
rubber ‘A’ have not shown any tangible differences urements. Road Materials and Pavement Design, 1–21.
to the mixtures that were produced using the non- Takallou, H. & Hicks, R.G. 1988. Development of
pre-treated rubber ‘B’. improved mix and construction guidelines for rubber-
modified asphalt pavements.
Wang, H., You, Z., Mills-Beale, J. & Hao, P. 2012. Labo-
ACKNOWLEDGEMENTS ratory evaluation on high temperature viscosity and
low temperature stiffness of asphalt binder with high
The principal author would like to acknowledge percent scrap tire rubber. Construction and building
the support of the Ministry of Higher Educa- Materials, 26, 583–590.

348
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Rutting performance of bituminous mixtures composed with red mud

M.S.S. Lima & L.P. Thives

Federal University of Santa Catarina, Brazil

V. Haritonovs
Riga Technical University, Latvia

ABSTRACT: Red mud is a solid waste generated at aluminum production. Currently in Brazil this
sub-product does not receive appropriate destination and is stored in deforested areas. Annually, approxi-
mately 11 million tons of red mud are generated just in Brazil and represent a great imminent risk of pol-
lution and contamination. The large amount of red mud generated by aluminum industry has motivated
researchers to develop methods for using this residue. Studies show the possibility of using red mud com-
pound ceramic materials, however, there is a lack of research in other areas, such as applied to pavements.
On the other hand, more than 50% of Brazilian highways are in poor condition, especially in the north,
where high temperatures favor permanent deformation in pavement. In this scenario, this paper compares
permanent deformation results of asphalt mixtures composed by 7% filler, varying the type of filler (red
mus and stone powder). For this, red mud characterization tests were performed. The results showed that
red mud can improve the resistence of asphalt mixtures.

Keywords: Red mud, bitumen, permanent deformation

1 INTRODUCTION alkalinity [10, 16] assign some hazardous proper-

ties to this residue, that even stored efficiently, it
The red mud is a waste originated in the benefi- can be a threat to biodiversity and might cause
ciation process of bauxite to aluminum. For each serious environmental damages such as contami-
ton of aluminum produced are generated 0.3 to 1.5 nation of surface effluent and underground, affect-
tons of red mud [1, 2, 3]. ing the health of closest populations [10].
The amounts related to disposal of red mud Despite being developed researchs aiming the
changes according the production methods and baux- use of red mud at bricks, tiles and cement [10, 17,
ite components. Thus, those with low aluminum con- 18], there are few references using this solid waste
tent are able to produce up to 2.5 tons of red mud [4]. applied in other contexts.
In northern Brazil, are produced between 4.40 In Brazil, most of the roads evaluated are clas-
million and 6.26 million tons of red mud [2]. It is sified as poor or very poor considering the traf-
estimated that about 10.6 million tons of waste ficability conditions. The most common defect
are discarded annually in the country, while in the found on Brazilian highways is the permanent
world, reaches more than 117 million tons/year [1]. deformation, caused not only high volume of traf-
In the State of Paráis located the largest bauxite fic, as well high temperatures [34]. In this context,
waste deposit of Brazil, considered the biggest gener- this work aims to contribute to find the best way
ator in Latin America [5, 6]. Only in Barcarena (State for using the red mud, evaluating the properties
of Pará, Brazil), the red mud generation is around and analyzing the possibility to apply it on asphalt
4.5 million tons/year, occupying around 4 km2 [7]. mixtures.
The red mud is produced in large quantities
and, even though there are studies that prove the
feasibility of using this residue for ceramic mate- 2 MATERIALS AND METHODS
rials [8, 9, 10], and also technologies capable of
neutralizing it [11, 12], in Brazil, reuse policies In this study, it is compared the permanent defor-
are not adopted and there is no prospect of doing mation performance of two types of bituminous
it in a large scale. Thus, the red mud is cumula- mixtures, differentiated only by the type of filler
tively stored, occupying increasingly larger areas. used: red mud and stone powder. Under Brazilian
Some researches classify the red mud as a legislation, filler corresponds to the material with
non-inert waste [13, 14, 15]. However, the high 65% or more passing the 0.075 mm sieve.

349
 Thus, bituminous mixtures were composed by 1000 Σ = 0,07P4 + 0,14P3 + 0,33P2 +
bitumen CAP-50/70, granite aggregates and filler. 0,81P1 + 2,7S3 + 9,15S2 + 21,9S1 + 135F (1)
As the quantity and type of filler used can affect
the performance, it was fixed for both mixtures 7% At where:
filler, varying the type used. It was used 3% red P4 – fraction retained between the sieves 50–25 mm;
mud + 4% stone powder to evaluate the red mud P3 – fraction retained between the sieves 25 and
influence. The results were compared with a refer- 12.5 mm;
ence mixture composed of 7% of stone powder. P2 – fraction retained between the sieves 12.5 to
Both mixtures adopted a theoretical grad- 4.76 mm;
ing curve corresponding to the average of the P1 – fraction retained between the sieves 4.76 to
range “C” of the National Department Brazilian 2.00 mm;
Transport Infrastructure (DNIT). S3 – fraction retained between the sieves 2.00 to
The dosage of the bitumen content was per- 0.42 mm;
formed using Servopac gyratory compactor. The S2 – fraction retained between the sieves 0.42 to
Number of turns (N) design used to establish com- 0.177 mm;
paction efforts during the dosage process was 125, S1 – fraction retained between the sieves 0.177
corresponding to a high traffic volume. to 0.075 mm;
Prior to carrying out the tests, the characterization F–fraction passing the sieve 0.075 mm;
of materials was carried out. Red mud was evaluated
by testing X-Ray Diffraction (XRD) and lasergranu- Prior to the tests, the red mud was dried at
lometry. The bitumen by softening point test, pen- 100°C and crushed in a mortar, which provided to
etration and viscosity, using Brookfield equipment. the residue granulometric regularity.
The specific surface of red mud and stone pow-
der filler were calculated by Vogt formula. 2.2 Bitumen
The performance to permanent deformation was
evaluated with French traffic simulator Orniéreur. The bitumen used was the CAP-50/70, provided by
the company CBB Bitumen, based in the city of
Curitiba, Parana.
2.1 Red mud The bitumen characterization was performed
The red mud used in this work is originally from by softening point test—ring and ball method,
Hydro Alunortecompany, located in Barcarena, regulated by DNIT ME 131/10 [35], penetration,
Pará, Brazil. regulated by DNIT ME-155/10 [36] that allows
The mineralogical analysis of the red mud was measuring consistency of the bitumen through the
performed using X-ray diffractometer RIGAKU, depth to which a standard needle (100 g) penetrates
model Miniflex II with copper Ka radiation from into the sample during the time of 5 seconds at
the tube and wavelength of 1.540562, parafocal 25°C, and viscosity held in a Brookfield viscometer.
geometry endowed with Bragg-Brentano system. The appropriate mixing temperature for unmodi-
The test was performed between 2θ angles of fied binders according SUPERPAVE methodology
5° to 90°, with step of 0.05 and 1 second. The is when the binder has a viscosity corresponding to
phases were identificated using a database of the 170 ± 20 cP. While adequate compaction tempera-
International Center for Diffraction Data-Powder ture corresponds to viscosity of 280 ± 30 cP. These
Diffraction File (ICDD-PDF) and Crystallogra- values are traditionally applied to asphalt mixtures
phy Open Database (COD). dosed using the Marshall methodology with pure
The particle size of the red mud was observed binders, and have also been used Gyratory compac-
in laser particle analyzer Microtrac, model S3500, tion. According to the DNER ME 043/95, binder
with a detection range of 0.02 to 2,800 micrometers temperature in both procedures should not be less
(μm) using Microtrac Flex software version 10.5.4. than 107°C or exceed 177°C.
This methodology was chosen for convenience and With the viscosity results could be set tempera-
precision of results compared to other forms of tures 148°C and 137°C for mixing and compaction
particle size measurement. procedures, respectively. The heating temperature
With the particle size was possible to calculate of aggregates was 15°C above the mixing tempera-
the specific surface area of the red mud. The same ture of the bitumen, which is 160°C.
was done for the stone powder filler used for the
composition of the reference mixture.
2.3 Dosage
In Brazil, the specific surface area has been
measured [19, 20, 21] using an adaptation of Duriez Dosage procedures performed followed the recom-
formula, known as formula Vogt, according to mendations of the standards AASHTO M 323 -
Equation 1, the result being expressed in m2 /kg. Superpave Volumetric Mix Design and AASHTO

350
R 35 - Superpave Volumetric Design for Hot-Mix Readings were taken at 0, 100, 300, 1,000, 3,000,
Bitumen (HMA) [29,30]. 10,000 and 30,000 cycles at 15 different points of
Prior to the compaction process, the bitumen the slabs, using a digital caliper.
mixtures were conditioned for 2 hours in an oven at The depth of permanent deformation in each
a temperature of 137°C (compaction temperature), cycle was calculated by averaging the depths meas-
in order to simulate the aging of the bitumen [27]. ured in 15 points against the benchmark of zero
Compaction was performed in Servopac gyra- cycle.
tory compactor, using rotation angle of 1.25°
± 0.02, a rate of 30 revolutions per minute and
600 kPa vertical stress. 3 RESULTS AND DISCUSSION
After compaction, it was determined the bulk
specific gravity (Gmb) of compacted samples, con- 3.1 Red mud characterization
ducting weighing the specimens under the condi-
tions of dry surface, saturated dry and submerged, The X-ray diffraction detected phase of hematite
according to ASTM D 2726 standard—Method (Fe2O3), anatase (TiO2), quartz (SiO2), karelianita
for Bulk Specific Gravity and Density of Non- (V2O3), gibbsite (Al(OH)3) and sodalite (Na8Mg3-
Absorptive Compacted Bituminous Mixtures [28]. Si9O24 (OH)2), formed during the Bayer process.
The mixtures were compressed to 125 rotations Hematite, the anatase, quartz, gibbsite and
to obtain bitumen content for 4% voids. sodalite are substances that do not pose health
To perform the mixtures used a theoretical curve risks, except as noted in concentrations higher
corresponding to the average range C DNIT [31]. than those allowed by standard (Appendix ‘G’)
The content design adopted for the mixes with NBR 10.004/2004 [37].
7% stone powder and 3% red mud + 4% stone pow- The karelianita (V2O3), or vanadium trioxide
der was 4.7% and 4.6% of bitumen, respectively. in the presence of moisture can come to oxidize
and turn into vanadium pentoxide (V2O5), in the
list of ‘substances which give hazardous waste’, on
2.4 Permanent deformation Annex ‘C’ of NBR 10.004/2004.
For doing the permanent deformation test, the mix- Thus, the red mud can be classified as Class I
tures were compacted into slabs form, according with a residue—Dangerous according Brazilian
to the criteria of the standard NFP 98-250-2 [32], standards.
in dimensions of 50 cm length, 18 cm width and At laser granulometric test can be seen that the
5 cm thickness. For this, it was used the compactor red mud is a fine grained material with 100% of the
developed by Institut Francais current des Sciences particles smaller than 290 μm, 80% than 50 μm in
et Technologies des Transports, L’aménagement et diameter and 40% than 5 micrometers, as Figure 1.
des Reseaux (IFSTTAR). Approximately 85% of the red mud particles
The amount of bitumen mixtures to be com- have a diameter in the range of 0.4 μm to 60 μm,
pressed is defined by calculating the Gmb refers and 20% corresponding to the clay fraction (<2 μm)
to 4% voids. In all, was molded four slabs, two and 65% to silt fraction (2 μm to 60 μm), as soil
for each type of bitumen mixture. The slabs were classification based on grain size criteria of the
tested 15 days after compaction. Brazilian Association of Technical Standards [40].
The permanent deformation test is regulated Only 15% of the particles have a size in the range
by the French standard NFP 98-253-1: Essaisre- from 60 μm to 200 μm, characteristic of fine sand.
latifs Aux Chaussées, Permanent Deformation des
Mélanges Hydrocarbonés [33] and performed in 100
French traffic simulator called Orniéreur. 90
The equipment is provided by pneumatic wheels,
80
applying a load of 5.0 kN, and the inflation pres-
70
sure of tires 0.6 MPa. The machine allows test two g60
specimens simultaneously at a temperature 60°C, ~ so
exposed to longitudinal loads up to 30,000 cycles ~
40
with a frequency of 1 Hz. Being a cycle equivalent
30
to go and return movement of the tire.
20
Before starts the test, it was made the accom-
10
modation of the slabs in the mold, subjecting them
to 1,000 cycles. Then the slabs were conditioned at 0,1 10 100 1000
60°C. However, until this temperature was reached Gr11n ulo~n etry(lolm)

and stabilized it was required approximately 8 to

12 hours. Figure 1. Red mud granulometry.

351
 According to granulometric results similar to during the compaction process, obtaining 4.91% of
those obtained in this work, Antunes, Conceição voids for the reference mixture (7% of stone pow-
and Navarro (2011) classified the red mud as a der) and 5.51% for the mixture with 3% red mud.
sandy-clay-silty material, according to the tex- This it happened due the dosage and the slabs
tural classification proposed soil by United States samples was made using different methods of
Department of Agriculture [38] characterizing it as compactation. The gyratory compactor has more
a cohesive material with low porosity and intense accuracy than the method used for doing the slabs,
phenomena of capillarity [39]. which allows small variations in the final thickness,
The particle size of the red mud is an important resulting in large changes in the voids of the com-
factor to be considered when it is used compound pacted mixture.
bituminous mixtures, since particles bigger than It is noteworthy that bituminous mixtures com-
40 μm tend to fill the voids of the aggregates and pacted in the field also present voids variation,
particles smaller than 20 μm are mixed to the bitu- allowed in the service specifications, which pro-
men, modifying the viscosity, softening point and vided a gap between 3% and 6% of voids [31].
thermal susceptibility [41, 42, 43, 44]. Thus, the slabs produced were considered
Therefore, since about 70% of the grains that appropriate for permanent deformation test.
make up the red mud have a particle size less than Table 2 presents the data depth for bituminous
20 μm, it is possible that part of the residue used mixtures produced with 7% stone powder and 3%
in the composition of bitumen mixtures is incor- red mud + 4% stone powder obtained at each cycle.
porated into the CAP, changing its rheological The depth reduction represents the performance
properties. improvement obtained by substituting the stone
Knowing that 85% of red mud grains have powder for 3% of red mud, which means the bitu-
a diameter less than 0,075 mm and the remain- minous mixture with red mud it had 12.82% better
ing 15% are corresponding to the variable range results compared by reference mixture.
S1, and that the specific gravity of the residue is In Figure 2 the graphical aspect of the data is
2.609 g/cm3, the correction factor used was 1,008 presented.
obtained by interpolation. Thus, it obtained a
specific surface value corresponding to red mud
Table 2. Values of sag of bitumen mixtures comprised
119.003 m2/kg.
of stone powder (PP) and Red Mud (RM).
For comparison, it carried out the calculation
of the specific surface of the stone powder filler Depth (%)
used in the composition of the mixtures employ-
ing Vogt formula. Given that 100% of its particle Numbers of cycles 0%RM 3%RM
size is passing the No. 200 sieve and that its density
corresponds to 2.717 g/cm3, used a correction fac- 100 1,09 0,79
tor of 0.98. Thus, the specific surface of the stone 300 1,52 1,14
powder was found to be 132.287 m2/kg. 1000 2,19 1,71
3000 3,05 2,46
10000 4,38 3,68
3.2 Bitumen 30000 6,10 5,31
The results are shown in Table 1. Depth reduction (%) – 12,82

3.3 Permanent deformation 100

The performance of bituminous mixtures as per-
manent deformation was made in French traffic
simulator Orniéreur. b1~~
Although the bituminous mixtures were dosed 5 31-%
for 4% voids, none of the slabs achieved this index

Table 1. Results of characterization of bitumen.

0.1
Test CAP-50/70
100 1000 Cycles 10000 100000
Softening Point (°C) 49,5 ~ O%RM ~ 3%RM
Penetration (0, 1 mm) 64
Viscosity (cP) 135°C, SP 21, 50 rpm 308,67 Figure 2. Performance to permanent deformation of
compacted mixtures.

352
 It is important to say that the percent deforma- With the red mud characterization was possible
tion is equivalent to strains, and that the X-axis of to classify this solid waste as environmental aspects,
Figure 1 begins with 100, because on 0 it doesn’t according Brazilian standards. Thus, the vanadium
have any deformation. pentoxide (V2O5) finded at red mud composition
The reference mixture had 6.1% of deformation allows classifying it as Class I—Dangerous.
and the mixture with 3% of red mud had 5.31%. As permanent deformation behavior: both
The French guidelines limit by 10% the accept- mixtures were classified as acceptable according
able amount of deformation for bituminous mix- French guidelines, with values less than 10%. Nei-
tures used intop layer [45]. There are also other ther of the mixtures achieved the established value
European guidelines that limit the permanent for the european regulation (5%), however the
deformation at 5% for dense bituminous mixtures bituminous mixtures composed for 3% of red mud
used in high traffic roads [46]. had a result 12.82% better than the reference one.
Thus, both mixtures obtained results of per- In general, the performance improvement of
manent deformation lower than 10% specified by permanent deformation can be attributed to the
the French regulation. However, neither of them use of red mud, since the residue has a specific sur-
obeyed the European standards, having deflection face area smaller than stone powder filler. In this
values higher than 5%. case, it reduced the amount of bitumen required,
Even with a higher quantity of voids the mix- restricting the thermal susceptibility of mixture.
ture with 3% red mud obtained better results than It is also possible that part of the red mud with
the reference mixture, which means the permanent particle size smaller than 20 μm it was incorporated
deformation provided by the mixtures decreased into the bitumen, changing the properties and pro-
with the incorporation of the residue. moting resistance of permanent deformation.
Several factors can influence the permanent Although the same amount of filler was used
deformation performance of bituminous mixtures, in both mixtures, the red mud mixture had the
among them are: voids quantity, bitumen content, best results regarding the performance of the per-
type and percentage of filler, graduation and type manent deformation. Therefore, as important as
of aggregate used in the composition of mixtures. determining the amount of filler to be inserted it
In general, as both mixtures in this work had the is knowing the chemical and physical properties of
same graduation and type of aggregate used (gran- the material applied.
ite), the differences obtained in outcomes can be Despite the good results, to recognize the viability
attributed to the volume of voids, the bitumen con- of using red mud into highways, it is necessary doing
tent, the percentage and the type of employed filler. environmental tests to evaluate possible leaching of
Like all specimens obtained voids according the weight metals and consequential soil contamination.
specification established it can not be declared that
this variable has contributed significantly in resist-
ance to permanent deformation. REFERENCES
The stone powder filler had a higher specific
surface area than the red mud. Thus it can be [1] Ribeiro, D.V.; Labrincha, J.A.; Morelli, M.R. Efeito
established a direct equivalence between the bitu- da adição da lama vermelha na corrosibilidade do
men content and the amount of stone powder filler concreto armado avaliado por meio de técnicas elet-
used, that is, higher the percentage of stone pow- roquímicas. RIEM—Revista IBRACON de Estru-
der filler, higher is the bitumen consumption. turas e Materiais, v. 5, n. 4, p. 451–467, Agosto 2012.
The bitumen content has a large influence on ISSN 1983-4195.
thermal susceptibility of bituminous mixtures, [2] Hildebrando, E.A. et al. Application of Bauxite
Waste from Amazon Region in the Heavy Clay
being able to promote lubrication between particles, Industry. Materials Research, v. XVI, n. 6, p. 1418–
and most efficient accommodation of the aggre- 1422, Nov/Dez 2013. ISSN 1516–1439.
gates, increasing the workability and eventually, [3] Hydro. Como a alumina é extraída da bauxita. A
reducing the resistance to permanent deformation. Hydro no Brasil, 2013 B. Disponivel em: <http://
www.hydro.com/pt/A-Hydro-no-Brasil/Sobre-o-
aluminio/Ciclo-de-vida-do-aluminio/Refino-da-alu-
4 CONCLUSION mina/>. Acessoem: 25.abril. 2014.
[4] Komnitsas, K.; Bartzas, G.; Paspaliaris, I. Efficiency
This study evaluated the possibility of using red of limestone and red mud barriers: laboratory col-
umn studies Miner. Minerals Engineering, v. 17, p.
mud in bituminous mixtures. Thus, was performed 183–194, 2004. ISSN 0892-6875.
tests of permanent deformation in bituminous [5] Abal. Fluxo da cadeia de produção do alumínio
mixtures composed by 3% red mud + 4% stone primário. Associação Brasileira do Alumínio, 2013
powder and 7% stone powder, considered as refer- A. Disponivel em: <www.abal.org.br/aluminio/pro-
ence mixture. ducao>. Acessoem: 20 Dez 2013.

353
 [6] ABAL. Associação Brasileira do Alumínio. Asso- [20] Moreira, H.S. Comportamento mecânico de mistu-
ciação Brasileira do Alumínio, 2015. Disponivel ras asfálticas a frio com diferentes teores de agre-
em: <http://www.abal.org.br/estatisticas/nacionais/ gado fresado incorporado e diferentes modos de
bauxita/>. Acessoem: 26 abr. 2015. compactação. Universidade Federal do Ceará. Dis-
[7] Bezerra, W.S.; Macêco, E.N.; Souza, J.A.D.S. Uti- sertação de Mestrado. Fortaleza/CE. 2005.
lização da lama vermelha proveniente do processo [21] Bernucci, L.B. et al. Pavimentação asfáltica: for-
produtivo da alumina como filler em Concreto mação básica para engenheiros. 3ª. ed.Rio de
Betuminoso Usinado a Quente (CBUQ). 40ª Janeiro: PETROBRAS: ABEDA, 2010.
RAPv—Reunião Anual de Pavimentação. Rio de [22] Brasil. Dner ME 078/94: Agregado graúdo—adesiv-
Janeiro/RJ: [s.n.]. 2010. idade a ligante betuminoso. Departamento Nacional
[8] Kavas, T. Use of boron waste as a fluxing agent in de Estradas de Rodagem. [S.l.], p. 3. 1994 A.
production of red mud brick. Building and Environ- [23] ASTM. D2493-09: Standard Viscosity-Temperature
ment, v. 41, p. 1779–1783, 2006. ISSN 0360-1323. Chart for Bitumens. American Society for Testing
[9] Yang, J.; Xiao, B. Development of unsintered con- and Materials. West Conshohocken. 2009.
struction materials red mud wastes produced in the [24] NCHRP. Mixing and Compaction Temperatures
sintering alumina process. Construction and Build- of Bitumens in Hot-Mix Bitumen: REPORT 648.
ing Materials, n. 22, p. 2299–2307, 19 Novembro National Cooperative Highway Research Program.
2008. ISSN 0950-0618. Washington, D.C. 2010.
[10] Antunes, M.L.P.; Conceição, F.T.D.; Navarro, [25] Bitumen Institute. Superpave mix designs, super-
G.R.B. Caracterização da Lama Vermelha Bra- pave series no.2. USA. 2003.
sileira (Resíduo do Refino da Bauxita) e Avaliação [26] Zegarra, J.R.E. Avaliação de misturas produzidas
de suas Propriedades para Futuras Aplicações. 3rd com ligantes asfálticos peruanos convencional PEN
International Workshop Advances in Cleaner Pro- 60/70 e modificados por polímero SBS tipo I 60/60 e
duction. São Paulo, Brazil: 2011. p. 10. PG 76-22. Dissertação. Escola de engenharia de São
[11] Palmer, S.J. et al. Thermally activated seawater neu- Carlos, Universidade de São Paulo. São Carlos, SP.
tralised red mud used for the removal of arsenate, 2007.
vanadate and molybdate from aqueous solution. [27] SHRP-A-698. Development and Use of the
Journal of Colloid and Interface Science, v. 342, n. Repeated Shear Test (Constant Height): An
1, p. 147–154, 2010. ISSN 0021-9797. Optional Superpave Mix Design Tool. Strategic
[12] Cooling, D.J. Improving the sustainability of resi- Highway Research Program. National Research
due management practices-Alcoa World Alumina Council. Washington, D.C. 1994 A.
Australia. Australian Centre of Geomechanics, [28] ASTM. D 2726-00: Standard Test Method for Bulk
Perth, Australia, 2007. Specific Gravity and Density of NonAbsorptive
[13] EPA. Environmental Protection Agency. U.S. Gov- Compacted Bituminous Mixtures. American Soci-
ernment Printing Office. Electronic code of fed- ety for Testing and Materials. USA. 2000.
eral regulations, 2014. Disponivel em: <http://ecfr. [29] AASHTO. M 323 - Standard specification for
gpoaccess.gov.>. Acesso em: 12 abril 2014. Title 40, SUPERPAVE Volumetric mix Design. American
Part 261, Sect 4 (b) (7) (ii) (c). Association of State Highway and Transportation.
[14] Nunn, R.F. Advances in red mud dewatering and Washington, DC. 2011.
disposal technologies. J. Light Metals—The Miner- [30] AASHTO. R 35-12: Standard Practice for SUPER-
als, Metals & Materials Society, Warrendale, 1998. PAVE Volumetric Design for Hot-Mix Bitumen
p. 107–114. (HMA). American Association of State Highway and
[15] Lima, F.S.S. Utilização da Lama Vermelha e do Transportation Officials. Washington, D.C. 2012.
Resíduo Caulinítico na Produção de Pigmento [31] BRASIL. DNIT 031/06-ES: Pavimentos flexíveis—
Pozolânico para Argamassas e Concretos de Cimento Concreto asfáltico—Especificação de serviço.
Portland. Dissertação de Mestrado. Universidade Departamento Nacional de Infra-estrutura de
Federal do Pará—UFPA. Belém/PA. 2006. Transporte. Rio de Janeiro, RJ. 2006 A.
[16] Collazo, A. et al. Evaluation of red mud as sur- [32] AFNOR. NF-P-98-250-2: Essaisrelatifsauxchaussées
face treatment for carbon steel painting. Process —préparation des mélanges hydrocarbonés, Partie2:
in Organic Coating, v. 52, p. 351–358, 2005. ISSN Compactagedes Plaques. Association Française de
0300-9440. Normalisation. France. 1997.
[17] Manfroi, E.P. Avaliação da lama vermelha como [33] AFNOR. NF P 98-253-1: Préparation des Mélanges
material pozolânico em substituição ao cimento Hydrocarbonés, Partie1: Essai d’Orniérage. Associ-
para produção de argamassa. Dissertação de ationFrançaise de Normalisation, AFNOR. France,
Mestrado. Universidade Federal de Santa Cata- p. 11. 1993.
rina—UFSC. Florianópolis/SC, p. 151 p. 2009. [34] CNT. Pesquisa CNT de rodovias 2014: Relatório
[18] Macêdo, A.N. et al. Comportamento de blocos Gerencial. Confederação Nacional do Transporte.
cerâmicos estruturais produzidos a partir da mis- Brasília/DF:CNT, SEST, SENAT, p. 389. 2014.
tura de lama vermelha e argila. Revista Ambiente [35] BRASIL. DNIT 131/2010-ME: Materiais asfál-
Construído, Porto Alegre, v. XI, n. 4, p. 25–36, outu- ticos-Determinação do ponto de amolecimento-
bro/dezembro 2011. ISSN 1678-8621. Método do Anel e Bola. Departamento Nacional de
[19] Silveira, M.A. Estudo sobre adição de polímero em Infraestrutura de Transportes. [S.l.], p. 6. 2010 A.
areia-asfalto a frio. Dissertação de Mestrado. Uni- [36] BRASIL. DNIT 155/2010-ME: Material asfáltico-
versidade de São Paulo. São Carlos/SP. 1999. Determinação da penetração-Método de ensaio.

354
 Departamento Nacional de Infraestrutura de Trans- [43] Santana, H. Considerações sobre os nebulosos con-
portes. [S.l.], p. 7. 2010 B. ceitos e definições de fíler em misturas asfáltica. 29ª
[37] ABNT. NBR 10.004: Resíduos sólidos—Classifi- Reunião Anual de Pavimentação. Associação Bra-
cação. Associação Brasileira de Normas Ténicas. sileira de Pavimentação. Cuiabá. 1995.
Rio de Janeiro, p. 71. 2004 A. [44] Motta, L.M.G.; Leite, L.M.F. Efeito do fíler nas
[38] USDA. Soil survey laboratory methods and pro- características mecânicas das misturas asfálticas. XI
cedures for collecting soil samples. Department of PANAM. Brasil, p. 09–19. 2000.
Agriculture. Washington, USA, p. 50. 1967. [45] LCPC. Manuel LPC d’aide à la formulation des
[39] Camargo, O.A. et al. Métodos de Análise Química, enrobés à chaud. Groupe de travail RST Formula-
Mineralógica e Física de Solos do Instituto tion des enrobés à chaud. Laboratoire Central des
Agronômico de Campinas. Campinas, SP, p. 77. 2009. Pontset Chaussées. France. 2007.
[40] ABNT NBR 6502/95: Rochas e solos. Associação [46] COST 333. Development of New Bituminous Pave-
Brasileira de Normas Técnicas. Brasil, 1995. ment Design Method: Final Report of the Action.
[41] Cavalcante, V.T.F.; Soares, J.B. O efeito do tipo e do European Cooperation in the field of Scientific and
teor de fíler nas propriedades mecânicas das mistu- Technical Research. European Commission Direc-
ras asfálticas. 33a Reunião anual de pavimentação. torate General Transport. Belgium. 1999.
Florianópolis/SC, p. 13. 2001.
[42] Pinilla, A. O sistema fíler-betume, algumas consid-
erações sobre sua importância nas misturas densas.
Instituto de Pesquisas Rodoviárias. 1965.

355
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

An experimental study on waxy bitumens

J. Oner
Department of Civil Engineering, Faculty of Engineering, Usak University, Usak, Turkey

B. Sengoz
Department of Civil Engineering, Faculty of Engineering, Dokuz Eylul University, Izmir, Turkey

G. Malkoc
Turkish Asphalt Contractors Association, Ankara, Turkey

ABSTRACT: Almost all bitumen obtains from crude oil by refining process but only certain crude oils
contain good quality bitumen for asphalt pavement. Naphthenic-base crude oils often give a large amount
of bitumen that may be good quality, while paraffinic crude oils may give bitumen of good quality or yield
bitumen not suitable for asphalt pavement. Wax in bitumen has been referred to as petroleum wax and
is obtained from refining of paraffinic crude oils. In spite of the fact that high wax contents have been
considered as a negative effect on the quality of the bitumen, there is no common agreement among the
scientists regarding the effect on bitumen rheology and asphalt performance.
The scope of this study is to evaluate the rheological properties of waxy bitumens obtained from differ-
ent sources. Following the determination of wax contents by two different methods, various conventional
methods were performed. The high temperature performance levels of bitumen were also determined
according to Superpave system by Dynamic Shear Rheometer (DSR) test on samples before and after
aging processes. The intermediate temperatures performances levels of bitumen were also performed by
DSR. Rutting performance of bitumens has been evaluated using Zero Shear Viscosity (ZSV) and Multi-
ple Stress Creep Recovery (MSCR) tests.

1 INTRODUCTION process and mainly consists of naphthens and iso-

paraffins as well as crystallizes as small microscopic
Bitumen is adhesive material which is used as an needles. A micro crystalline petroleum wax is char-
agent at flexible pavements on roads as well as in acterized also by a less distinct melting area and its
other areas of application, such as water proofing, high average molecular weight giving higher viscos-
flooring and joint materials. Bitumen is very com- ity compared to macro crystalline paraffin wax (Das
plex material including hydrocarbon molecules et al. 2013, Musser & Kilpatrick 1998).
with small amounts of heteroatoms consist of Definitions of wax concepts have varied over the
sulphur, nitrogen, oxygen and gradually liquefies years and sometimes been contradictory (Redelius
when heated (Redelius & Soenen 2015). & Soenen 2015, Edwards 2009). Factors that influ-
The definition of bitumen wax has been formu- ence the effect of waxes are chemical composition
lated to facilitate the distinction between harm- (source of bitumen) and rheological behavior of
ful wax and less harmful or non-harmful (Lyne the bitumen as well as content, and the crystallin-
et al. 2013, Edwards & Redelius 2003, Soenen et al. ity of the wax. Consequently, many bitumen speci-
2013, Canestrari et al. 2013). Waxes in bitumen are fications include requirements concerning wax
divided into two general categories such as paraf- content. It has been reported that wax content in
fin wax (macro-crystalline) and micro-crystalline bitumen should not exceed 3% (Wong & Li 2009).
wax (Edwards 2009). Paraffin wax also known as The wax content is limited to 2.2% in Europe
macro-crystalline wax crystallizes in large flat plates according to European standard EN 12606-1.
or needles. It refers to the group of n-alkanes with few Based on literature, the high content of wax
or no branches (C20–C40) (Soenen et al. 2013). The affects the properties of bitumen in different ways.
melting point of macro-crystalline paraffin waxes The low melting point of wax decreases resistance
are around 50–70°C (Lu et al. 2005, Lu & Redelius of asphalt mixtures against rutting at high temper-
2007). On the other hand, micro-crystalline wax is atures, and the crystallization of wax causes crack-
collected in the bitumen fraction after the distillation ing at low temperatures (Lu & Redelius 2006).

357
Physical hardening, poor ductility and poor bitu- The system includes a distilling system, an elec-
men adhesion can be listed as the results caused tric cooler, a thermo-regulator and compressor
by the high content of wax (Edwards & Redelius refrigeration.
2003). The performance grade of bitumen at high Test procedure is performed on two portions
temperature may decrease with melting of wax. for each of 25 g bitumen samples. The distillate
Somewhat contradictory, also positive effects from bitumen is obtained by a specified distillation
of wax in bitumen have been reported, such as process. The distillation unit is consisted of a dis-
improved low-temperature properties and lower tillation flask, a laboratory burner and erlenmeyer
handling temperatures (for better compaction) flask used as distillation receiver. The distillation
(Redelius et al. 2002). Strategic Highway Research takes place at very high temperatures (up to over
Programme recommends the performance of 500°C) at which cracking of molecules may occur
some additional rheological and chemical tests on (Lu & Redelius 2006). On the second portion of
unaged, short-term aged and long-term aged bitu- the test procedure; the waxy distillate is dissolved
men to evaluate the exact definition of wax con- in ether/ethanol (50/50, V/V) solvent and crystal-
cept (Petersen et al. 1994). lized at −20°C. The crystallized wax is obtained by
This paper evaluates the rheological behav- filtration (EN 12606-1 2007).
ior of waxy bitumens obtained from different Differential Scanning Calorimetry (DSC)
sources. Following the determination of wax con- method has been utilized to register thermal effects
tents by DSC (Differential Scanning Calorimetry) of waxes in bitumen through the cooling-heating
and EN 12606-1, the detailed properties of bitu- cycles according to ASTM D4419-90 (Standard
men samples have been evaluated using Dynamic test method for measurement of transition temper-
Shear Rheometer (DSR) and various conventional atures of petroleum waxes by Differential Scanning
methods. The upper critical temperatures used in Calorimetry (DSC)) (ASTM D4419-90 2015). In
Superpave Performance Grading (PG) system have this study, thermal characterization was performed
been determined for each bitumen sample using using DSC; Perkin Elmer-Diamond. Approxi-
G*⁄sinδ results gained from Dynamic Shear Rhe- mately 15 mg of bitumen samples is placed in an
ometer (DSR) test. The effects of loading and tem- aluminum pan and sealed under nitrogen atmos-
perature on the performance of the waxy bitumen phere. The bitumen sample is heated to +120°C and
samples have been evaluated by low (0.01 Hz) and then cooled at 7°C /min to −50°C, followed by heat-
high frequency (10 Hz) loadings at five different ing to +120°C at the same rate. DSC procedure is
temperatures (40°C–80°C). Besides, the intermedi- also used to determine wax content in bitumen. The
ate temperatures performances levels of bitumen DSC wax content of bitumen is calculated from an
were determined to investigate the fatigue per- endothermic peak during the heating scan. For the
formances of different sources bitumens. Rutting calculation of percentage of wax in the bitumen, a
performance of bitumens has also been evaluated constant melting enthalpy value of 121 J/g is used
using Zero Shear Viscosity (ZSV) and Multiple as a reference (Lu et al. 2008, Edwards et al. 2006).
Stress Creep Recovery (MSCR) tests performed in
creep mode. 2.2.2 Conventional bitumen tests
In order to characterize the properties of the dif-
ferent sources bitumens, conventional bitumen
2 EXPERIMENTAL tests such as: penetration test, softening point test,
ductility test, fraass breaking point test and Rolling
2.1 Materials Thin Film Oven Test (RTFOT) as well as penetra-
Three bitumens from three different sources were tion, softening point, and ductility after RTFOT
used in this study. The bitumen samples are identi- were performed (ASTM D5-06 2006, ASTM D36-
fied as Bit No:1; Bit No:2 and Bit No:3 respectively. 06 2006, ASTM D113-86 1994, EN 12593 2007,
Turkey (Bit No:1) bitumen was 50/70 penetration ASTM D2872-12 2012).
grade bitumen, while the others, from Iraq (Bit In addition, the temperature susceptibility of
No:2) and Russia (Bit No:3) bitumens were 70/100 the bitumen samples has been calculated in terms
penetration grade bitumens. of Penetration Index (PI) using the results obtained
from penetration and softening point tests
(Whiteoak & Read 2003).
2.2 Test methods
2.2.1 Determination of the paraffin wax content 2.2.3 Rheological test methods
The European Standard EN 12606-1 is based on The Dynamic Shear Rheometer (DSR) is used
DIN 52015, a German method that is performed for to describe the viscous and elastic behaviour of
the determining the paraffin wax content of bitu- asphalt bitumen at medium to high temperatures.
men samples (Baldino et al. 2012, EN 12606-1 2007). This characterization is used in the Superpave PG

358
asphalt bitumen specification. The basic DSR test Developments led to the existence of Multiple
uses a thin bitumen sample sandwiched between Stress Creep and Recovery (MSCR) test (Dubois
two circular plates. The lower plate is fixed while et al. 2014). MSCR test was run on RTFOT aged
the upper plate oscillates back and forth across the samples using the dynamic shear rheometer (ASTM
sample at 10 rad/sec (1.59 Hz) (or any other set fre- D7405-08 2008). Specimens were tested using a
quency) to create a shearing action. DSR tests are 25 mm parallel plates and with 1 mm gap setting
conducted on non-aged and RTFOT aged bitumen at temperature of 60°C and at a stress of 100 and
samples. The DSR measures a specimen’s complex 3200 Pa. The tests were performed at the selected
shear modulus (G*) and phase angle (δ). The com- temperature using a constant stress creep of 1 sec-
plex shear modulus (G*) can be considered the sam- ond duration and a relaxation period of 9 seconds,
ple’s total resistance to deformation when repeatedly for ten cycles at each stress level. Percent recoverable
sheared, while the phase angle (δ), is the lag between and non-recoverable components of creep compli-
the applied shear stress and the resulting shear ance were determined at the end of 10 cycles.
strain. G* and δ are used as predictors of rutting
and fatigue behaviours. Early in pavement life rut-
ting is the main concern, while later in pavement 3 RESULTS AND DISCUSSIONS
life fatigue cracking becomes the major concern.
G*⁄sinδ is described as rutting indicating parameter 3.1 Determination of wax content in bitumens
and G*.sinδ as fatigue cracking indicating factor through EN 12606-1 and DSC
(Ruan et al. 2003). The Superpave® specification
Calculation of the wax content was based on the
parameter, G*/sin δ, was identified as the term to be
following Equation (1) according to European
used for high temperature performance grading of
Standard Method EN 12606-1. Wax content is cal-
paving asphalts in rating the bitumens for their rut-
culated as:
ting resistance. Although used for many years as a
rutting parameter, it has been demonstrated that the
m1 ∗ mw
relationship between G*/sin δ and rutting is poor. P= ∗100 (1)
This term was found to be inadequate in describing mb ∗ m2
the rutting performance of bitumens.
Zero Shear Viscosity (ZSV) is the viscosity meas- where P (m%) is wax content, mb (g) is the mass of
ured in shear deformation at a shear rate approach- specimen, m1 (g) is the total mass of distillated oil,
ing to zero. This parameter is an indicator for rutting m2 (g) is the mass of distillated oil for wax extrac-
related bitumen characteristics. It has been observed tion and mw (g) is the mass of wax extracted (EN
by researchers in the recent years, that the rutting 12606-1 2007).
parameter G*/sin δ is very effective in predicting The crystallization or melting of waxes in bitu-
the rutting performance of bitumen. ZSV has been mens involves energy change. This can be easily
evaluated to be a more appropriate indicator in pre- determined by DSC. In Table 1, results of the EN
dicting the rutting behaviour of bitumen (Visscher 12606-1 and DSC measurements are summarized
et al. 2004). The determination of ZSV in this study in terms of wax content. It should be noted that
is made by the application of creep test using DSR. the wax content as calculated by using a constant
Static creep is defined as the slow deformation of enthalpy in this study is not an absolute value. The
a material measured under a constant stress. In DSC measures transformation enthalpy and in
the static creep test, a fixed shear stress is applied fact different waxes can have different enthalpies,
to the sample and the resultant strain is monitored meaning the wax content in bitumen may vary
for a predetermined amount of time. If the stress is with a constant melting enthalpy value.
applied for a sufficiently long duration of time, the As indicated in Table 1, bitumen samples from
deformation in the bitumen reaches a constant value, different sources differ widely in the wax content
which corresponds to the steady state flow of the
bitumen. The viscosity of the bitumen at this stage
is known as the steady state viscosity or the ZSV Table 1. DSC and EN 12606-1 analysis of bitumens.
(Gungor & Saglik 2012). All of the samples were
short-term aged using a RTFOT procedure. The Bitumen Bit No:1 Bit No:2 Bit No:3
ZSV of the samples have been predicted by applica-
tion of creep test at 60°C (CEN TS 15325 2008). Penetration Grade 50/70 70/100 70/100
The test geometry for the creep tests consisted of Sources Turkey Iraq Russia
25 mm parallel plates and the gap between the Wax content by EN 0.80 1.12 2.30
12606-1 (%)
plates was 1 mm. The stress level was 10 Pa for bitu-
Wax content by 2.61 3.27 4.04
men samples and each creep test was continued for DSC (%)
30 minutes.

359
according to DSC and EN 12606-1 methods. It is tent has a higher PI and thus depicts lesser temper-
clearly seen that, EN 12606-1 method gives much ature susceptibility as compared to other samples.
lower values in wax content as compared with DSC. It can be also concluded from the results that bitu-
It is believed that the high temperature (up to over men containing low amount of wax significantly
500°C) distillation step in EN 12606-1 method may lowers temperature susceptibility of samples.
destroy the molecular structure of waxes (thermal With regard to the physical properties, the duc-
cracking). The resulting smaller paraffin molecules tility is sometimes helpful for the evaluation of
could be soluble in ether/ethanol and do not crystal- the low-temperature failure properties of bitumen
lize in the solvent at the specified low temperature (Feng et al. 2012). The ductility is increased by
(−20°C), leading to lower apparent wax content. 45 cm and 61 cm for Bit No:3 in comparison with
These might be reasons for the low wax contents as Bit No:1 and Bit No:2 respectively, indicating an
determined by the EN 12606-1 method. Although improved low temperature failure resistance of Bit
both DSC method and EN 12606-1 methods are No:3 with high content of wax.
basically very different, they exhibit a reasonably The low temperature properties of the bitu-
good correlation. Bitumen sample from Russia con- men may be affected in a negative way with higher
tains large amount of wax compared to other bitu- breaking point according to fraass. As presented
men samples according to EN 12606-1 and DSC in Table 2, fraass breaking point is decreasing with
methods. However, Bit No:1 has the lowest amount increasing of wax content of bitumens. Ductility
of wax. and fraass breaking point tests may indicate that
utilization of Bit No:3 containing high content
of wax provides an advantage for fatigue crack-
3.2 Conventional test results
ing due to its high elasticity and low temperature
The conventional properties of different sources cracking.
bitumens are presented in Table 2. The effects of
wax on different sources bitumen can be explored
3.3 Rheological test results
using Table 2. Results indicated that as expected
from the penetration grades of bitumen; Bit No:1 This section includes the results of rheological tests
with the lowest wax content sample depicted applied on all of the bitumen samples obtained
a significant decrease in penetration value and by the different sources. The results have been
increase in softening point temperature as com- presented into three groups of Dynamic Shear
pared to bitumens including relatively high wax Rheometer (DSR) test results, Zero Shear Viscos-
content. ity (ZSV) test results and Multiple Stress Creep
The softening point temperature can be used Recovery (MSCR) test results.
along with the penetration to determine the tem-
perature susceptibility of bitumen. The Penetration 3.3.1 Dynamic Shear Rheometer (DSR)
Index (PI) is an indicator of temperature suscepti- test results
bility of bitumen. A high PI indicates low tempera- The determination of complex rheological proper-
ture susceptibility (Whiteoak & Read 2003). The ties of bitumen is currently possible with the use of
results show that Bit No:1 with the lowest wax con- the Dynamic Shear Rheometer (DSR). Parameters
most commonly tested in DSR include the bitu-
men complex shear modulus (G*) and phase angle
Table 2. Conventional test results of bitumens. (δ), tested in various temperature and frequency
ranges. In order to determine upper critical temper-
Bitumen Bit No:1 Bit No:2 Bit No:3 ature used in Performance Grading (PG) system,
non-aged and RTFOT aged specimens of bitumens
Penetration Grade 50/70 70/100 70/100 containing different content of wax (with dimen-
Penetration at 25°C 50 75 76 sions of 25 mm in diameter and 1 mm in height)
(0.1 mm) were subjected to oscillating shear in the DSR at
Softening Point (°C) 54 48 48 the frequency of 10 rad/s (1.59 Hz). The tempera-
Penetration Index (PI) −0.25 −0.62 −0.59 ture cycles were set to start at 52°C for non-aged
Retained Penetration after 62 59 55 and 64°C for RTFOT aged samples an run up in
RTFOT (%) 6°C increments. The upper critical temperatures
Softening Point Diff. after 9 17 5 (Tcrit) used in PG system were determined for each
RTFOT (°C)
sample using the obtained G*/sinδ results and
Ductility at 25°C after 14 14 58
RTFOT (cm)
DSR specifications for PG bitumen. In PG system,
Ductility at 25°C (cm) 64 48 109 the upper critical temperature is the temperature
Fraass breaking (°C) −12 −13 −18
at which G*/sinδ = 1.0 kPa for non-aged bitumen,
and G*/sinδ = 2.2 kPa for RTFOT aged bitumen

360
(Ruan et al. 2003). The upper critical temperatures
(Tcrit) for each sample are presented in Table 3.
It can be concluded from the Table 3 that PG
upper critical temperature of Bit No:1 (Tcrit) con- 'ii
taining the lowest wax content is higher than Bit
No:2 and Bit No:3. There is no significant vari-
..e,~ l ,OOE...04 f---.Yr- - - - - - - - - - - - - -
ation for Tcrit values of Bit No:2 and Bit No:3.
Higher Tcrit value is an indicator of higher resist- "
ance to permanent deformation. G*/sinδ values for
non-aged Bit No:2 sample and RTFOT aged Bit
No:2 sample does not fix at the same temperature.
Lower temperature (64°C) was selected to be on "
Temperature(0 C)
70 80

the safe range as PG upper critical temperature. "*"Bit i\o:J

Difference of G*/sinδ values for non-aged and

RTFOT aged sample implies that Bit No:2 hard-
Figure 1. G*⁄sinδ values for different sources bitumens
ens and oxidizes much faster than other bitumen at 0.01 Hz.
samples because of aging process.
All of the RTFOT aged samples including dif-
ferent contents of wax were subjected to oscillating 1,20E+06

shear in the DSR at low (0.01 Hz) and high (10 Hz) l ,OOE+-06 '\

.
"
frequency levels at five different temperature cycles
8.00[ +05
ranging from 40°C to 80°C with 10°C increment.
The variation of G*/sinδ (rutting parameter) val-
!!:-
'l! 6,00[+«!5
~
ues of the samples at low and high frequencies are j
4,00 E +fl5
~
presented in Figure 1 and Figure 2 respectively. "
As presented Figure 1 and Figure 2; regarding all : ,00[+-05 ~ ~
of bitumen samples, G*⁄sinδ values increase with 0,0(1[.;.()(1
~~
decrease in temperature at both frequencies. An
increment in G*⁄sinδ value indicates higher perform- - Bit::\o:J
ance against rutting. Besides, as expected G*⁄sinδ
values increase with an increase in frequency for all
Figure 2. G*⁄sinδ values for different sources bitumens
of the bitumen samples. This is due to the rheologi-
at 10 Hz.
cal behaviour of the bitumen since bitumen under
shorter loading times (high frequency level) exhibit
elastic behaviour (Whiteoak & Read 2003). As can be seen in Figure 1 and Figure 2, Bit
No:3 sample containing the highest wax content
depicted lower G*⁄sinδ value than Bit No:1 and
Table 3. Determination of PG upper critical tempera-
tures for different sources of bitumens. Bit No:2 at both high and low frequencies and
at all temperatures. Among the bitumen samples
DSR, G*/sinδ (Pa) which contains low wax content, the highest value
Temp. Performance in the G*⁄sinδ values can be seen regarding Bit
Bitumen (°C) Non-aged Aged Grades No:1 sample at 10 Hz. frequency and all tempera-
tures. However this case is only valid for Bit No:1
Bit No:1 52 12500 B 70 at low frequency level (0.01 Hz) and at 40°C, 50°C
58 5918 and 60 °C. Under longer loading times (at low fre-
64 2820 5816 quency level), a significant increase in G*⁄sinδ val-
70 1384 2868 ues can be seen at Bit No:2 sample at intermediate
76 702 1462 temperature levels (40°C, 50°C and 60°C).
Bit No:2 52 5160 B 64 Fatigue properties of bitumen can be evaluated
58 2253
by the fatigue factor at intermediate temperatures.
64 1012 11700
The fatigue behaviour was investigated by applying
70 460.9 5837
continuous oscillatory shear loadings with using
76 2791
DSR. Temperature sweep test is performed under
82 1307
the strain controlled mode at a frequency of 10 rad/s
Bit No:3 52 5165 B 64
58 2431
(1.59 Hz.). The temperature cycles were set to start
64 1147 3346
at 40°C within the PAV aged samples a run up in
70 556.7 1648 3°C decrements. During the test procedure, an 8 mm
diameter plate with a 2 mm gap between parallel

361
plates was used for each PAV aged bitumen sample. Besides, the limitation of 5000 kPa value is
There is a limiting maximum stiffness at the inter- reached at Bit No:1, Bit No:2, Bit No:3 which
mediate temperature to mitigate fatigue cracking. occur at the test temperature 19°C, 22°C and 13°C
G*.sinδ values of PAV aged bitumen samples does respectively. It can be seen that Bit No:3 have
not exceed 5000 kPa and low values of these param- improved (lowered) the intermediate temperature
eters are considered good indicators of fatigue compare to Bit No:1 and Bit No:2 which contain
cracking resistance. Variation of the fatigue param- the lowest amount of wax.
eters (G*.sinδ) for PAV aged bitumens containing On the other hand, Figure 4 shows a reasonably
different contents of wax at different intermediate good correlation with a correlation coefficient (R2)
temperatures and these temperatures were plotted of 0.737 between the wax content by EN 12606-1
against the wax content according to EN 12606-1 and the temperatures at critical G*.sinδ values of
are illustrated in Figure 3 and Figure 4 respectively. each bitumen under long term aging. In general,
As shown in Figure 3, regarding all of bitumen the high wax content led to lower G*.sinδ values of
samples, G*.sinδ values increase with decrease in the bitumen samples. The increase in wax content
temperature at 10 rad/s. Owing to the aging, the has an obvious effect on long term aged bitumen
lower G*.sinδ value means the less shearing energy fatigue parameter.
loss and the better ability of the fatigue resistance.
Bit No:1 and Bit No:2 samples containing the low- 3.3.2 Zero Shear Viscosity (ZSV) test results
est wax contents depict the highest G*.sinδ value at All of the samples were short-term aged using a
all intermediate temperatures. On the other hand, RTFOT procedure. The ZSV of the samples have
Figure 3 confirms the superior fatigue perform- been predicted by application of creep test at 60°C
ance of Bit No:3 at all intermediate temperatures (CEN TS 15325 2008). ZSV results of all of the
with the lowest value of G*.sinδ. Higher G*.sinδ bitumen samples including different content of
values is not preferable since the samples with wax are illustrated in Figure 5.
higher G*.sinδ values exhibit fatigue cracking. As depicted in Figure 5, Bit No:2 and Bit
No:1 samples show higher ZSV value than Bit No:3
7,00[ +03
which contains the highest content of wax. Bit
~
No:2 sample yields the peak ZSV value at 60°C com-

~
'·"'"'' f - - - - - - - - - - - - - - - - - - - - - , - - - - - , -'-
1\!a~imumSpeciflcatiomtimit for faugue:panlll!ter
' -
.
..........
.'
.'.

/·"
I pared to Bit No:1 and Bit No:3. A higher ZSV value
indicates higher resistance to permanent deforma-
C , ,OO[ +OJ f - - - - - - - - - - - - - - - - : - "'--r'------:;-'"-- - - tion under long term loading. As can be seen in the
~ /',.J" // figure, the presence of high wax content significantly
f-- - - - - - - - -,-,-c
. ,~....
, :___o'~-~-----
J ,OOE+OJ
decreases the ZSV values of Bit No:3. In spite of the
1.:1 ~,00[+03 f----------:;--"'"'~~-"·4"-·-_ ___.
-:;o-<-
' -·- - - - - - - fact that Bit No:2 contains relatively higher wax con-
.-:t: =- :·-f.:· ~ ·~ · .,..- ·
1,00[ +03 f---.,.,...=-'---~-"'-----------
tent than Bit No:1, there is a slightly augmentation
0,00[ +00 ~.:=.~ - -·-·
L..:_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ __ in terms of ZSV value of Bit No:2 compared to Bit
No:1. The short term aging process is expected to
have a positive effect on the ZSV value of Bit No:2.

Figure 3. G*.sinδ values for different sources

bitumens.

y • -4.982•• 25,00
R.' • 0,737

0,5 1,5 2.5

Waxconlenl h~· E 12606-1 (•f.)
Sit Nl):l Bi t NI):Z Bit lllo: 3

Figure 4. Correlation between the wax content by EN

12606-1 and the temperatures at critical G*.sinδ values of Figure 5. ZSV values for bitumens containing different
each bitumen under long term aging. content of wax.

362
Table 4. Average values of the MSCR test parameters In the light of the findings, the DSR test results
calculated from the test data. show that wax content had an influence on the
bitumen rheology on rutting and fatigue crack-
Bitumen Bit No:1 Bit No:2 Bit No:3 ing. Detailed investigation performed by DSR test
R @100 Pa (%) 35.029 43.155 17.455 indicates that bitumen containing the highest wax
R @3200 Pa (%) 29.862 28.095 5.991 content depicts lower G*⁄sinδ value than other
Jnr @100 Pa (1/kPa) 0.211 0.229 1.164 bitumen samples, thereby possibly decreasing the
Jnr @3200 Pa (1/kPa) 0.231 0.289 1.474 rutting resistance. Under the PAV aging condition,
Jnr diff. (%) 9.522 26.555 26.615 the temperature of bitumens of high wax content is
Stress sensitivity 0.095 0.265 0.266 lower than that for bitumen with low wax content
at the time the stiffness value of G*.sinδ reaches
5000 kPa. It indicates that high wax content posi-
tively affects the intermediate temperatures stability
3.3.3 Multiple Stress Creep and Recovery of bitumens under the long term aging condition.
(MSCR) test results In the light of findings from ZSV values indicate
The percent Recoveries (R), the compliance (Jnr) that the utilization of bitumen containing low
at two different stress level of 100 Pa and 3200 Pa amount of wax improves rutting performance of
and stress sensitivity as well as the percent differ- mixtures. This can be due to high viscosity and low
ences in non recoverable compliances (Jnr-diff) of wax content that give the mixture stability at service
the bitumens are presented in Table 4. temperature. According to MSCR test results, bitu-
Bit No:3 sample containing the highest content men containing the lowest content of wax has the
of wax has the lowest recovery at any of the stress highest recovery and the lowest Jnr values in any of
levels which indicates that Bit No:3 performs less the stress levels. Higher percent recovery, lower non-
rut resistance than other samples. This can be due recoverable compliances may suggest that bitumens
to the low stiffness and the low elastic behaviour are less prone to rutting after the application of
of Bit No:3 sample. While Bit No:2 exhibits more loading–unloading cycles at a common high tem-
recovery at 100 Pa stress levels, Bit No:1 has the perature. In terms of the asphalt mixture, bitumens
highest recovery value at 3200 Pa. Higher percent with lower Jnr values and/or higher recovery values
recoveries mean that the bitumen can recover a will less contribute to the accumulation of unrecov-
higher portion of its total strain at the end of each ered strain in the asphalt layer. Based on the rheo-
loading–unloading cycle, which is favourable to logical test results, G*⁄sinδ and ZSV values have a
the resistance of the material to rutting. good correlation with the MSCR parameters.
In stress levels of both 100 and 3200 Pa, Bit In terms of this study, conventional and rheo-
No:1 with the lowest content of wax has the low- logical bitumen tests have been conducted to eval-
est Jnr value. Lower non recoverable compliances uate the rheological properties of waxy bitumens
indicate a minor contribution of the bitumen to obtained from different sources. It is recommended
the appearance of rutting in the asphalt mixture or, to perform rheological tests on different sources of
in other words, a lower susceptibility of the bitu- bitumen involving different types of polymers in
minous material to rutting. As the results show, Bit order to reduce negative effects of waxes.
No:3 has the highest the compliance (Jnr) value at
100 Pa and 3200 Pa. It can be concluded that Bit
No:3 is the most susceptible sample to rutting.
REFERENCES

ASTM D113-86. 1994. Standard test method for ductility

4 CONCLUSIONS AND of bituminous materials. American Society for Testing
RECOMMENDATIONS and Materials, West Conshohocken (PA): USA.
ASTM D2872-12. 2012. Standard test method for effect
Based on the results presented in this paper, of heat and air on a moving film of asphalt (Rolling
the following conclusions can be drawn. EN Thin-Film Oven Test). American Society for Testing
12606-1 method gives much lower values in the and Materials, West Conshohocken (PA): USA.
wax content as compared to DSC. Although both ASTM D36-06. 2006. Test method for softening point of
the DSC and EN 12606-1 method are basically bitumen (ring-and-ball apparatus). American Society
very different methods, they show a reasonably for Testing and Materials, West Conshohocken (PA):
USA.
good correlation based on the selected bitumen ASTM D4419-90. 2015. Standard test method for meas-
samples in terms of wax content. Conventional urement of transition temperatures of petroleum waxes
bitumen tests indicated that bitumen containing by Differential Scanning Calorimetry (DSC). Ameri-
low amount of wax significantly lowers tempera- can Society for Testing and Materials, West Consho-
ture susceptibility of samples. hocken (PA): USA.

363
ASTM D5-06. 2006. Standard test method for penetration Gungor, A.G. & Saglik, A. 2012. Evaluation of rutting
of bituminous materials. American Society for Testing performance of neat and modified binders using zero
and Materials, West Conshohocken (PA): USA. shear viscosity. 5th Eurasphalt & Eurobitume Con-
ASTM D7405-08. 2008. Standard test method for multiple gress, Istanbul: Turkey.
stress creep and recovery (MSCR) of asphalt binder Lu, X. & Redelius, P. 2006. Compositional and structural
using a dynamic shear rheometer. American Society for characterization of waxes isolated\n from bitumens.
Testing and Materials, West Conshohocken (PA): USA. American Chemical Society: Energy & Fuels 20(2):
Baldino, N., Gabriele, D., Rossi, C.O., Seta, L., Lupi, 653–660.
F. R. & Caputo, P. 2012. Low temperature rheology Lu, X. & Redelius, P. 2007. Effect of bitumen wax on
of polyphosphoric acid (PPA) added bitumen. Con- asphalt mixture performance. Construction and Build-
struction and Building Materials 36: 592–596. doi: ing Materials 21(11): 1961–1970.
10.1016/j.conbuildmat.2012.06.011 Lu, X., Kalman, B. & Redelius, P. 2008. A new test
Canestrari, F., Graziani, A. Pannunzio, V. & Bahia, H.U. method for determination of wax content in crude
2013. Rheological properties of bituminous binders oils, residues and bitumens. Fuel 87(8): 1543–1551.
with synthetic wax. International Journal of Pavement Lu, X., Langton, M., Olofsson, P. & Redelius, P. 2005.
Research and Technology 6(1): 15–21. Wax morphology in bitumen. Journal of Materials
CEN TS 15325. 2008. Bitumen and bituminous binders: Science 40(8): 1893–900.
Determination of zero shear viscosity (ZSV) using a Lyne, A., Wallqvist, V. & Birgisson, B. 2013. Adhesive
shear stress rheometer in creep mode. European Com- surface characteristics of bitumen binders investigated
mittee for Standardization. by Atomic Force Microscopy. Fuel 113: 248–256.
Das, P.K., Kringos, N., Wallqvist, V. & Birgisson, B. Musser, B.J. & Kilpatrick, P.K. 1998. Molecular charac-
2013. Micromechanical investigation of phase separa- terization of wax isolated from a variety of crude oils.
tion in bitumen by combining AFM with DSC results. Energy Fuels 12(4): 715–25.
Road Materials and Pavement Design 14: 25–37. Petersen, J.C., Robertson, R.E., Branthaver, J.F., Harns-
De Visscher, J., Soenen, H., Vanelstraete, A. & Redelius, berger, P.M., Duvall, J.J., Kim, S.S., Anderson, D.A.,
P.A. 2004. Comparison of the zero shear viscos- Christiansen, D.W. & Bahia, H.U. 1994. Binder char-
ity from oscillation tests and the repeated creep test. acterization and evaluation (Report no. SHRP-A-367).
Proceedings of Euroasphalt & Eurobitume Congress, Strategic Highway Research Programme, Washington:
Vienna: Austria. USA.
Dubois, E., Mehta, D. & Nolan, A. 2014. Correlation Redelius, P. & Soenen, H. 2015. Relation between bitu-
between multiple stress creep recovery (MSCR) results men chemistry and performance. Fuel 140: 34–43.
and polymer modification of binder. Construction and Redelius, P., Lu, X. & Isacsson, U. 2002. Non-classical
Building Materials 65: 184–190. wax in bitumen. Road Materials and Pavement Design
Edwards, Y. & Redelius, P. 2003. Rheological effects of 3(1): 7–21.
waxes in bitumen. Energy and Fuels 17(3): 511–520. Ruan, Y., Davison, R.R. & Glover, C.J. 2003. Oxida-
Edwards, Y., Tasdemir, Y. & Isacsson, U. 2006. Effects tion and viscosity hardening of polymer-modified
of commercial waxes on asphalt concrete mixtures asphalts. Energy & Fuels 17(4): 991–998.
performance at low and medium temperatures. Cold Soenen, H., Besamusca, J., Fischer, H.R., Poulikakos,
Regions Science and Technology 45(1): 31–41. L.D., Planche, J., Das, P.K., Kringos, N., Gren-
Edwards, Y. 2009. Influence of waxes on bitumen and fell, J.R., Lu, X. & Chailleux, E. 2013. Laboratory
asphalt concrete mixture performance. Road Materi- investigation of bitumen based on round robin DSC
als and Pavement Design 10(2): 313–335. and AFM tests. Materials and Structures 47(7):
European Standard EN 12593. 2007. Bitumen and bitu- 1205–1220.
minous binders—Determination of the Fraass break- Whiteoak, D. & Read, J.M. 2003. The Shell bitumen
ing point. European Committee for Standardization. handbook. Thomas Telford Services Ltd, London:
European Standard EN 12606-1. 2007. Bitumen and England.
bituminous binders—determination of the paraffin Wong, W. & Li, G. 2009. Analysis of the effect of wax
wax content—part 1: method by distillation. European content on bitumen under performance grade clas-
Committee for Standardization. sification. Construction and Building Materials 23(7):
Feng, Z., Yu, J. & Wu, S. 2012. Rheological evaluation 2504–2510.
of bitumen containing different ultraviolet absorbers.
Construction and Building Materials 29: 591–596.

364
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Comparative investigation on the reinforcing effects of polymer-based

fiber into bituminous mixtures

E. Toraldo, E. Mariani & M. Crispino

Department of Civil and Environmental Engineering, Politecnico di Milano, Milan, Italy

ABSTRACT: This paper reports the results of a laboratory investigation on the reinforcing effects of
fibers into bituminous mixtures, in which neat bitumen is used as binder. The investigated fibers were
selected in the European market and were composed of polymer compounds. For comparative purposes,
a polymer-modified bituminous mixture was included in the study. Laboratory tests investigated compac-
tion characteristics, and mechanical performance of the bituminous mixtures. The results suggest that
mixtures bounded with neat bitumen and reinforced with fibers exhibit same performance, or even better,
than the one in which polymer-modified bitumen is used.

1 INTRODUCTION Two dosages of fibers were considered in the

study (3% and 9% of the weight of the bitumen).
Bituminous mixtures are the materials most used For comparative purposes, a bituminous mixture
for road pavements through the world. They are bounded with polymer-modified bitumen was
composed of aggregates, bitumen and filler; differ- considered.
ent types of additives are also frequently used for The investigation included a number of tests to
both processing and performance purposes. highlight the effects of fibers on the corresponding
Among the wide range of additives available on bituminous mixtures in terms of:
the international market, fibers have been widely
− compaction properties (Workability and Self-
used as additives for bituminous mixtures. Acting
compaction from the Gyratory Shear Compac-
mainly as thickening agents, they increase viscos-
tor) at the same compaction temperature;
ity at high service temperatures, thus minimizing
− mechanical behavior (Stiffness Modulus and
drain-off effects that may occur in the early phases
Indirect Tensile Strength) at different tempera-
of transportation and laying (PIARC, 1998;
tures and at the same content of voids.
Brown, 1992; Mallick et al., 2000). Recently, how-
ever, spurred by the extensive use of fibers in other
fields of engineering and industry, fiber technology
2 EXPERIMENTATION
has rapidly evolved, leading to the production of
high-performance fibers that may play an impor-
2.1 Key materials
tant and innovative role in bituminous mixtures
(Tapkin et al., 2009; Wu et al., 2008; Lee et al., The key materials used during the research activi-
2005; Woodside et al., 2009; Hassan et al., 2005; ties herein described were:
Chen et al., 2005; Mahrez et al., 2003; Mahrez et al.,
− 50–70 neat (standard) bitumen, herein named
2005; Chen et al., 2010; Bullinger, 2004; Xu et al.,
SB;
2010; Chowdhury et al., 2006; Putman et al., 2004;
− 25–55/60 polymer-modified bitumen, contain-
Sanchez-Alonso et al., 2011 (a); Sanchez-Alonso
ing 4% of SBS polymer and named MB;
et al., 2011 (b); Movilla-Quesada et al., 2011).
− limestone aggregates;
Given the literature review briefly mentioned
− limestone filler completely passing through the
above, the research herein described focuses on the
63 μm sieve;
evaluation of the use of polymer-based fibers as
− five polymer-based fibers: polyacrylonitrile
reinforcement of bituminous mixtures in which
(Fiber 1), aramid-polyolefin (Fiber 2), poly-
neat bitumen is used as binder.
propylene added (Fiber 3), polyolefin polymers
For this purpose, a laboratory comparison was
(Fiber 4) and pure polypropylene (Fiber 5).
carried out, considering five types of polymer-
based fibers currently available on the Italian More details of both bitumen and fibers are
market. The fibers were composed of polyacrylo- given in Tables 1 and 2, respectively; Figure 1 shows
nitrile, aramid-polyolefin, polypropylene added, the fibers’ general appearance; Figure 2 reports the
polyolefin polymers and pure polypropylene. aggregates gradation.
365
2.2 Experimental program
100.0
As mentioned above, the research was aimed
90.0
at evaluating the role of fibers in reinforcing
bituminous mixtures bonded by neat bitumen in 80.0

70.0
Table 1. Main characteristics of the binders.
~ 60.0
e...
Polymer- Cl 50.0
c
Neat Modified ·;;;

Test
Standard
specification
Bitumen
[SB]
Bitumen
[MB]
"'"'
D..
40.0

30.0
Needle EN 1426 78.6 35.2 20.0
Penetration
[dmm] 10.0
Ring and Ball EN 1427 45.9 61.7 0.0
[°C] 0.01 0.1 10 100 1000
Rotational ASTM D4402 103 424 Sieve size [mm]
Viscosity
@ 163°C [cP] Figure 2. Aggregates sieve size distribution.

Table 2. Main characteristics of the fibers.

Fiber 1 Fiber 2 Fiber 3 Fiber 4 Fiber 5

Chemical Polyacrylonitrile Aramid-polyolefin Polypropylene Polyolefin Pure polypropylene

composition added polymers
Shape Fibrillated Twisted, fibrillated Mono-filament Mono-filament Fibrillated
and mono-filament
Density [g/cm3] 1.18 0.91 ÷ 1.94 0.93 0.91 0.91
Diameter [μm] 16 na 25 ÷ 75 677 na
Length [mm] 24 19 6/12 /18 19 na
Resistance to tensile >559 480 ÷ 2,800 400 ÷ 500 620 ÷ 758 na
stress [MPa]

na = not available.

Figure 1. General appearance of the fibers.

366
replacement of the polymer-modified one. For 3 RESULTS AND DISCUSSION
this purpose, twelve mixtures were prepared in the
laboratory, considering a bitumen content equal to In Figures 3, 4, and 5 the laboratory results (com-
4.5% by the weight of the aggregates. paction characteristics, and mechanical perform-
Two reference mixtures were prepared using either ance) are shown as a function of both bitumen
neat bitumen or polymer-modified; ten mixtures were type (neat or polymer-modified) and fiber type and
obtained adding 3% or 9% of fibers (by weight of dosage. The average results are presented as histo-
the neat bitumen). The fiber contents were selected grams, including error bars, displaying the maxi-
according to the range (1.5% and 12% by the weight mum and minimum test result of each mixture.
of bitumen) currently used for road applications
(Crispino et al, 2013; Toraldo et al. 2015; Bonica et al.,
3.1 Compaction characteristics
2016), as also suggested by the suppliers of the fibers.
In this paper, the mixtures are identified by an Compaction results, Self-compaction (C1) and
alphanumeric code formed by type of bitumen (SB Workability (k), are given in Figure 3.
for neat bitumen or MB for polymer modification) As far as the Self-compaction (C1), graph (A)
followed by the fiber identification (F1, F2, F3, of the figure shows that all the mixtures present
F4, or F5) and dosage (3 as 3% and 9 as 9% by the the same ability to be compacted under their own
weight of neat bitumen); e.g., the code of a bitu- weight (C1 in the range of 77–79%), except for the
minous mixture containing neat bitumen and fiber mixtures reinforced with fibers F2 and F3 at the
F1 at 3% will be SB-F13. lower dosage, which exhibit a slight increase of the
Mixtures bounded with neat bitumen were investigated parameter (C1 between 80% and 82%).
mixed and compacted at the temperatures of As shown in graph (B) of Figure 3, the other
165°C ± 5°C and 155°C ± 5°C, respectively. Both compaction parameter (workability—k) is not sig-
temperatures were increased of 10°C for the poly- nificantly affected by the use of fibers. Actually,
mer-modified mixture. the average Workability of all the investigated bitu-
Each mixing batch was 20 kg and obtained by minous mixtures was in the range of 7.8 and 8.3.
using a laboratory heated mixer. After mixing, a To sum up, the compaction results demonstrated
3.0 kg sample of each batch was used to check both that fibers not affect the ability of the correspond-
aggregates gradation and bitumen content and for ing mixtures to be compacted by the GSC, even if
measuring the Theoretical Maximum Density (EN the compaction temperature was not adjusted as
12697-5). The other part of the batch was used to a result of the presence of fibers. It is an interest-
compact cylindrical specimens by using the Gyra- ing feature because other studies of the authors
tory Shear Compactor (GSC) (ASTM D4013-09), demonstrated that the compaction temperature
obtaining nine nominally alike specimens (100 mm should be conveniently increased to obtain the
in diameter and 65 mm in height) for each mixture. desired volumetric characteristics of the mixtures
The use of the GSC allowed to record the degree if cellulose-based fibers are used (Crispino et al.,
of compaction of a specimen at each rotation, 2013; Toraldo et al. 2015).
and thus to estimate the fundamental compaction
parameters (Cominsky et al., 1994): Self-compac-
3.2 Mechanical performance
tion (C1), that is able to describe the mixtures’
ability to reduce its internal voids under its own Figures 4 and 5 report the mechanical test results.
weight, and Workability (k), that provides infor- Concerning the Stiffness Modulus, results in
mation about the mixtures’ ability to be compacted Figure 4 indicate, as expected, a decrease of the
under the rollers during construction operations. mixtures stiffness due to the test temperature
Obviously, an increase of both parameters reveals increase.
a better mixtures’ ability to be compacted on site. At low test temperature (5°C) all the mix-
Both the compaction parameters were calcu- tures show the same average stiffness (between
lated at 100 cycles of the GSC. 16000 MPa and 18000 MPa). Even if the mix-
In order to compare the mechanical behavior ture reinforced with fiber F4 at 9% of dosage
of the mixture independently from the volumetric exhibits the lower stiffness, revealing a positive
effects, all the mixtures were compacted at a target role of this kind of fiber to prevent typical dis-
voids content of 3.5% ± 0.5%. tresses in the field due to thermal cracking at low
Mixtures’ mechanical behavior was evaluated temperatures.
by means of Stiffness Modulus (SM) tests (EN All the investigated fibers produce a general
1269726 Annex C) at 2 Hz of load application, and increase of the stiffness on the corresponding
Indirect Tensile Strength (ITS) (EN 12697-23). mixtures at the intermediate test temperature
Both the mechanical tests were performed at 5°C, (20°C) if compared to both reference mixtures,
20°C and 40°C of temperature, in order to cover in which the sole neat bitumen or the polymer-
the typical range of the Mediterranean region. modified bitumen are used. The higher stiffness is

367
 A

~
O eo.o
c
0

~ 1:~~
0
~ 78.0 -rfl -
~ 76.0 -
:::;
-
"' :::1
74.0 - .,
:-:o
·.·1
- I

I .·.•
::.: ~

B
0.00 -
( r-
::;: , .00 -
c
(:::,
I r-
~

?; 0.00 -
• >·i
~ ::_: , ill r-
•:::::,
~: -:· 1

--r't:::::-:-• I r-
"'j •oo --
• .00
-

(
""' r-
,00 - ' .·.-·'.,. r-
,00 - ::o
r-
-:I
0.00 - r-
.::

Figure 3. Compaction characteristics: (A) self-compaction C1 and (B) workability k.

---±-- ,
- 1-- 1-
- 1-- ~ 1-
- 1-- 1-
- 1-- r-
- 1-- 1-
- 1-- 1-
-

. _,
1-- 1-
- ~ -
1--
_,
~· ·~· -~
-
~
-u
-- ·-~03
.,. .
~ ~---------------.--~r-~--~m---~-----------+-----­
~ - - --!-----:;t.-----,;l,,----------,h----111!1- il!l-----,r-----;--ill!-- - -
~ - - -m---{H+;- -ta!l---l!lll-----l!!il---.---t!:lil----'!!11--m--~1!----+­
~ 7~----±:- ~j:jii-:--Jj;~--------ll+f---l!i!!----l!!!I--!G---!i!l--------l I-

1~ =§~,!----1hil---·--· --i't---l\'l!---·--· ---m---l

I-
I-
I-
I-
I-
1-

i=
~- -----------;---------.!.---------,!l:.,---ffi---------------;-----

i= ----+-~[~·----;-~~---1m---------~a-~--
I ·~- - ~·• :;- -1:11---~AI---lil!----1!!1---i\!l---M-----Iil~ I-
j - - - r -{ :-;---1~]1---!fi---WA------i!:t---l!i!!---M---!IIIic----l I-
~ ~- - r--j ! 1-

Figure 4. Stiffness Modulus at 5°C, 20°C and 40°C.

368
 ~
!!. .oo
~ ·~
T -

flil
@
-
~ >.00

~ OM -
;;;
-
-
---+ :
-r:- -
.~
~
- ~"
-
-
~ 0.00
~
- - -jf- - - - I
- - -----j}-- - - -
i
>M

•oo -
~ OM -
- -u:- -
- ----{}---- - I

~ OM
i!.
~ 0.00 + ~ -

~~
@

..• - rt - III
f
~ Iliil
-

~
~
u
>00 - -
----jf- -

•
~ ·~
- -
---l?~ -
•·:;: o

~
i!. ·~
~
~ OAO
rt, !t····· -
H::l
z ·~ - -
- -i}- -
~
i>:
i
~
·~
- -
---++--- -

1·;:-i

Figure 5. Indirect Tensile Strength at 5°C, 20°C and 40°C.

reached by the mixtures in which fibers F2 and F3 A positive role of certain types of fibers is instead
(dosage at 9%) are used. appreciable at high temperature (40°C). Actually, the
The positive role of the fibers is also appreci- mixtures reinforced with the higher dosages of fibers
able at the higher test temperature (40°C), actually F2, F3 and F5, and the lower amount of fiber F4
a stiffness increase of the mixtures is noticeable if exhibit the same performance, or even better, than
compared to both reference mixtures (neat bitu- the one in which polymer-modified bitumen is used.
men and the polymer-modified bitumen). It is a
positive issue because that means fibers are able
to prevent typical distresses in the field due to rut- 4 CONCLUSIONS
ting phenomena at high temperatures: the best per-
formance is guaranteed by higher dosages of fibers The comparative laboratory investigation described
F2, F3 and lower dosage of F4. in this paper focused on the reinforcing effects of
Indirect Tensile Strength results (Figure 5) con- five polymer-based fibers on compaction charac-
firm the positive contribution deriving to the use teristics, and mechanical performance of bitumi-
of fibers when neat bitumen is used. nous mixtures for intermediate (binder) layers for
At 5°C of test temperature, the mixture in which road pavements bounded with neat bitumen. Two
polymer-modified bitumen exhibits the best per- dosages of fibers (3% and 9% by the weight of
formance (4.05 MPa in average). However, mixtures the bitumen) and two reference mixtures bounded
reinforced with F1 at both dosages, and F3 and F5 with the sole neat bitumen and a polymer-modified
(both at the higher dosage) give an average ITS bitumen were considered in the investigation.
value (higher than 3.50 MPa) close to the one of The results reported in this paper can be sum-
the mixture containing polymer-modified bitumen. marized as follows:
Fibers reinforced mixtures show an intermedi-
ate behavior at 20°C between the one with the sole − compaction characteristics of the investigated
neat bitumen and the one in which polymer-modi- bituminous mixtures are not significantly
fied bitumen is used. affected by the use of fibers;

369
− in terms of stiffness and failure properties, fib- EN 12697-26 Bituminous Mixtures – Test Methods for Hot
ers generally increase the mechanical behavior Mix Asphalt – Part 26: Stiffness. European Standards,
of the corresponding bituminous mixtures in European Committee of Standardization. Brussels, 2003.
Hassan, H.F., Al-Oraimi, S., Taha, R. 2005. Evaluation of
comparison with the one in which the sole neat
open-graded friction course mixtures containing cellulose
bitumen is used; moreover the reinforcement fibers and styrene butadiene rubber polymer, Journal of
deriving from the fibers produces the same, or Materials in Civil Engineering ASCE, July-August 2005.
even better, mechanical performance of the mix- Lee, S.J., Rust, J.P., Hamouda, H., Kim, Y.R., Border, R.H.
ture bounded with polymer-modified bitumen. 2005. Fatigue cracking resistance of fiber reinforced
asphalt concrete, Textile Research Journal. 75: 123–128.
Finally, it is important to highlight that the posi- Mahrez, A., Karim, M.R., Katman, H.Y. 2003. Prospect
tive role of fibers was assessed with no optimization of using glass fiber reinforced bituminous mixes, Jour-
of both bitumen and fibers dosage. Thus, Authors nal of The Eastern Asia Society for Transportation Stud-
believe better performance can be reached from ies. Vol. 5.
an optimization process of the above-mentioned Mahrez, A., Karim, M.R., Katman, H.Y. 2005. Fatigue
variables. Moreover, it is necessary to remark that and deformation properties of glass fiber reinforced
some issues remain still open about the contribu- bituminous mixes, Journal of The Eastern Asia Society
tion of the fibers in order to prevent fatigue cracks for Transportation Studies. Vol. 6.
Mallick, R.B., Kandhal, P.S., Cooley Jr., L.A., Watson,
and rutting potential, for this reason a new set of
D.E. 2000. Design, construction and performance of
fatigue and rutting laboratory tests is planned by new-generation open-graded friction course, NCAT
the authors. Report n° 2000–01.
Movilla-Quesada, D., Vega-Zamanillo, A., Caldaza-Pérez,
M.A., Castro-Fresno, D. 2011. Evaluation of water
REFERENCES effect on bituminous mastics with different contribu-
tion fillers and binders, Construction and Building
American Society for Testing and Materials (ASTM). Materials. 29: 339–347.
2008. Construction: roofing, waterproofing and bitu- Putman, B.J., Amirkhanian, S.N. 2004. Utilization of waste
minous materials. Annual Book of ASTM Standards fibers in stone matrix asphalt mixtures. Resources, Con-
Vol. 04.04, Philadelphia, PA, USA. servation and Recycling. 42: 265–274.
Brown, E.R. 1992. Experience with Stone Matrix Asphalt Sanchez-Alonso, E., Vega-Zamanillo, A., Castro-Fresno,
in the United States, NCAT Report n° 93–04. D., DelRio-Prat, M. 2011. Evaluation of compactabil-
Bullinger, L. 2004. The influence to the mechanical proper- ity and mechanical properties of bituminous mixtures
ties of bituminous mastics and, thus, to asphalt mixes with warm additives, Construction and Building Materi-
using special PAN-Fibres, 3rd Eurasphalt & Eurobitume als. 25:2304–2311.
congress. Vienna. pp. 889–902. Sanchez-Alonso, E., Castro-Fresno, D., Vega-Zamanillo,
Bonica, C., Toraldo, E., Andena, Marano, C., Mariani, A., Rodriguez-Hernandez, J. 2011. Sustainable asphalt
E. (2016). “The effects of fibers on the performance of mixes: use of additives and recycled materials, The Bal-
bituminous mastics for road pavements”, Composites tic Journal of Road and Bridge Engineering. VI: 249–257.
Part B: Engineering. 95: 76–81. Tapkin, S., Usar, U., Tuncan, A., Tuncan, M. 2009.
Chen, H., Xu, Q. 2010. Experimental study of fibers in Repeated creep behavior of polypropylene fiber-rein-
stabilizing and reinforcing asphalt binder, FUEL. n. 89; forced bituminous mixture, Journal of Transportation
pp. 1616–1622. Engineering. 135: 240–249.
Chen, J.S., Lin. K.Y. 2005. Mechanism and behavior of Technical Committee 8 Flexible Roads, World Road Asso-
bitumen strength reinforcement using fiber, Journal of ciation (PIARC). 1998. Use of modified bituminous
Material Science. 40:87–95. binders, Special Bitumens and Bitumens with Additives
Chowdhury, A., Button, J.W., Bhasin, A. 2006. Fibers from in Pavement Applications, International Workshop on
recycled tire as reinforcement in hot mix asphalt, Texas Modified Bitumens, Rome, Italy.
Transportation Institute, Report SWUTC/06/167453-1. Toraldo, E., Mariani, E., Malvicini, S. (2015). “Laboratory
Cominsky, R., Leahy, R.B., Harringan, E.T. 1994. Level investigation into the effects of fibers on bituminous
one mix design: material selection, compaction, and mixtures”, Journal of Civil and Environmental Manage-
conditioning, SHRP-A-408 Report, National Research ment. 21: 45–53.
Council, Washington, DC; USA. Woodside, A.R., Woodwart, W.D.H., Akbulut. H. 1999.
Crispino, M., Mariani, E., Toraldo, E. (2013). “Assessment The effect of cellulose fiber addition on bitumen prop-
of fiber-reinforced bituminous mixtures’ compaction erties, 3rd European Symposium Performance and Dura-
temperatures through mastics viscosity tests”, Con- bility of Bituminous Materials and Hydraulic Stabilised
struction and Building Materials. 38: 1031–1039. Composites. Leeds. UK. 17–26.
EN 12697-5 Bituminous Mixtures – Test Methods for Hot Wu, S., Ye, Q., Li, N. 2008. Investigation of rheological
Mix Asphalt – Part 5: Determination of the Maximum and fatigue properties of asphalt mixtures containing
Density. European Standards, European Committee of polyester fibers, Construction and Building Materials.
Standardization. Brussels, 2010. 22: 2111–2115.
EN 12697-23 Bituminous Mixtures – Test Methods for Hot Xu, Q., Chen, H., Prozzi, J.A. 2010. Performance of fiber
Mix Asphalt – Part 26: Indirect Tensile Strength. Euro- reinforced asphalt concrete under environmental tem-
pean Standards, European Committee of Standardiza- perature and water effects, Construction and Building
tion. Brussels, 2003. Materials. 24: 2003–2010.

370
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Fibre behaviour and influence on the properties of asphalt mortar

Monketh Mohammed, Tony Parry & James R.A. Grenfell

Department of Civil Engineering, Nottingham University, Nottingham, UK

ABSTRACT: This research is focused on the effect of fibres (glass, steel and cellulose) on the rheological
properties of asphalt mortar. The motive of this research is to gain a better understanding of the fibre
reinforcement and binding properties of asphalt mortar. Pavement deformation at high temperatures is
a significant problem for asphalt concrete mixtures. It is believed that permanent deformation is related
to residual strains and shear strains. Residual strains are mainly due to the compaction. Shear strains
are linked with the stiffness of the materials and formed by load induced stresses. When the pavement
temperature is high the probability of shear strains in asphalt pavement is increased. In this study the
potential of different fibre types to modify the rheological properties of asphalt mortar has been investi-
gated. An experimental programme was carried out on mortar mixtures with different contents of fibre.
The properties of the base mortar and resulting modified mortars were tested by using scanning electronic
microscopy, penetration, viscosity and rheological testing in the dynamic shear rheometer. The experi-
mental results showed that the fibres can entangle with each other and that adding fibres improves the
rutting parameters of asphalt mortar across a range of high temperatures. Finally adding fibre reduced
the penetration and increased the viscosity of mortar indicating an improvement in rutting resistance of
asphalt mixtures using these mortars.

1 INTRODUCTION Consequently, many additives have been added to

bitumen to increase the life service of asphalt pave-
The use of fibres to modify the performance ments. These additives are mainly organic poly-
of asphalt mixture dates back to many decades mers, that have been extensively studied (Peiwen,
(Putman, 2011). Several reports have shown that 2001, Cao and Ji, 2011).
adding different types of fibre led to improved In this study, fibre modified asphalt mortars
fatigue resistance (Yu and Sun, 2010, Alrajhi, 2012, were prepared with different fibre types (cellu-
Guan et al., 2014), moisture damage resistance (Zhu lose, 6 mm glass, 13 mm glass fibres and steel fibre
et al., 2007) and permanent deformation resistance recovered from tyres). Cellulose fibres are natu-
(Chen et al., 2004), in addition, the main use of ral fibres obtained most commonly from woody
fibres was to prevent the drain-down of binder dur- plants. Although glass fibres have desirable prop-
ing mixing and compaction of stone matrix asphalt erties, such as low thermal expansion coefficient,
and gap graded mixtures (Alvarez et al., 2008). high tensile modulus, high elastic recovery and high
It is significant to improve asphalt mixtures softening point, a literature review shows that their
performance in response to increasing traffic vol- use has not been reported very often (Abtahi et al.,
umes and axle loads and this may have increased 2013). In terms of recyclability and cost effective-
the bearing capacity of the pavement (Mohammed ness this study decided to use steel fibre recovered
and Mittal, 2016). Flexible pavements are sub- from tyres. Experiments were conducted to evaluate
jected to damage like cracking and rutting under the viscoelastic characteristics of the base mortar
the influence of repeated loading and temperature and modified mortar with various fibre combina-
cycling (Xu and Solaimanian, 2008). Generally, it tions. These tests included penetration, viscosity
is well known that pavement temperature may have and Dynamic Shear Rheometery (DSR).
an effect on the bearing capacity (Kristiansen,
2004). In asphalt mixtures, aggregates are the
stiffer and stronger components compared to the
2 EXPERMENTAL PROGRAMME
other materials, the bituminous binder or filler
mastic. Therefore, the possibility of failure is
2.1 Materials
expected to be very small through aggregates, while
there is a high possibility of failure in the bonding The binder used in the experimental programme
material that binds adjacent particles (Mo, 2010). was a 40/60 penetration grade bitumen. Lime-

371
Table 1. The basic properties of fibers.

Modulus of elasticity
Fibre Specific density (g/cm3) Length (μm) Width (μm) at 23°C (GPa)

Glass-l 2.58 13,000 12 to 20 80.3

Glass-s 2.58 6,000 12 to 20 80.3
Steel 7.85 4,000 to 12,000 180 to 300 *210
Cellulose 1.50 20 to 2,500 25 –

*standard steel modulus.

Figure 2. SEM images of cellulose fibre modified

asphalt mortars.
Figure 1. SEM images of 13 mm and 6 mm glass fibre
modified asphalt mortars.
tively. Micrographs of fibre modified mortars are
given in Figure 1 and 2, using a Scanning Electron
stone filler with maximum size less than 0.063 mm Microscope (SEM).
was used. Four fibre types including two types of
glass fibre, steel fibre recovered from tyres and cel-
2.2 Experimental method
lulose fibres were used as the additive to asphalt
mortar. Table 1 lists the basic properties of these Bitumen was placed in an oven at 160°C for one
fibres. The glass fibres had two lengths of 13 and hour to make it fluid enough for mixing. A spe-
6 mm, which are termed glass-l and glass-s respec- cific experimental procedure was used to obtain

372
homogeneous bitumen-fibre mortar. A Stuart network, stronger connections and joints through
mixer with a SS10 stirrer was used to apply a spe- the asphalt mortar as shown in Figure 1.
cific mixing speed (750 rpm) to prevent air bubbles With regard to cellulose fibres the scanning elec-
forming in the mortar-fibre mixture. Thermocou- tronic microscopy did not show entanglement of
ple probes were used to monitor the temperature cellulose fibres through asphalt mortar as shown in
throughout mixing by placing the probe directly Figure 2. However, cellulose fibres showed a good
inside the composite in the mixing vessel. Bitumen- compatibility and interaction with the asphalt
fibre mortars were mixed for one hour at more mortar.
than 160°C to produce homogeneous composites.
The penetration of the bitumen mortar
3.2 Conventional asphalt mortar properties
mixed with different fibres was measured with a
STANHOPE-SETA penetrometer (Europen Com- Penetration (EN, 2007) is a common test used to
moittee for Standardization, 2007). The penetra- measure bitumen hardness. The penetration value
tion test was used to evaluate the hardness of the was taken as the average of three readings. It can
fibre reinforced mortar. Penetration is define as the be observed that the penetration value of fibre
depth of penetration of a standard needle under a modified asphalt mortar decreases significantly
one hundred grams load after a five second load- with 0.5% and 1% addition of glass-l and glass-s
ing time at a specified temperature and the pen- fibres, with slight decreases with 1% addition of
etration value expressed in units of 0.1 mm. cellulose and steel fibres. The increase in hardness
A standard rotational viscometer (Model of modified asphalt mortar indicates a significant
LVDV-II+ PRD, Brookfield Engineering Inc.) was increase in bitumen stiffness with fibre modifica-
used to measure the viscosity of the fibre rein- tion especially at high fibre content.
forced mortar (Europen Commoittee for Stand- Rotational viscosities (EN, 2010) for the modi-
ardization, 2010). A Bohlin Gemini 200 (DSR) fied asphalt mortars made with the four fibre types
was used in this study for measuring the rheologi- (100–180°C) are shown in Figures 3 and 4. The
cal properties of base and fibre reinforced mortar. addition of fibres increased the viscosity of base
It has a torque range between 0.5 (μN⋅m) and 200 mortar and all data from three repeated tests were
(mN⋅m) (Airey, 1997). averaged and the error bars represent the stand-
ard deviation of original data. It is clear that the
viscosity decreased with increasing temperature,
3 RESULT AND DESICCATION while the viscosity increased with the increase of
fibre content. There was a marked increase in mor-
3.1 Microstructure of fibre modified asphalt tar viscosity at 0.5% glass and cellulose fibres con-
mortar tent. At the 1% glass and cellulose fibre content,
the viscosity greatly increased. A network of fibres
Scanning Electronic Microscopy (SEM) images
might provide support to the composite structure
of the glass and cellulose fibre modified asphalt
by holding the components together and resisting
mortars are shown in Figure 1 and 2. These SEM
deformation (see Figure 1 and 2) (Tapkın et al.,
images are taken from Double Edge Notch Ten-
2010). As with the penetration results, the viscosity
sion (DENT) test samples in the notch area. The
microstructure of the glass fibres were different
from that of cellulose fibre since the cross sec-
tions of glass fibres were straight, round, with a 1000000
- -0.5%vol. Glass-[
smooth surface texture and also longer than cellu- ---0.5%vol. Glass-s
lose fibre. This microstructure helps the glass fibres 1"'\
---r-0.5%vol. Celhiose
~ 100000
to entangle with each other to form a continuous u ----t-0.5 o/ovol. Steel
'-' ---<~--Base mortar

Table 2. Penetration test result.

£ 10000
"'u0
Penetration values "'
Mortar type 0.5% Fibre 1% Fibre
> 1000
Base mortar 30 30 100 135 160 180
Steel fibre modified mortar 30 27
Temperature (0 C)
Celluloses fibre modified mortar 29 28
Glass-s fibre modified mortar 25 24
Glass-l fibre modified mortar 23 21 Figure 3. Viscosity for base mortar and 0.5% mortar
fibre.

373
 1000000 - - 1o/ovol. Glass-I
l.E+07
--+-- lo/ovol. Glass-s
---1<--- l o/ovol. Cellulose '? l.E' 06
';'1 00000 ---+- 1% vol. Steel
e:..
Pot
.._..
(j
___.,____ Base mortar •
0.. l.E+05
;>,
:<: 10000 =
"'
0 ~ l.E+()4
(j =
s , 0.5%vol. Cellulose
>"' 1000 ~ l.E+03
, 0.5 %vol. Glass-!
Q. 0.5 %vol. Glass-s
0.5%vol. Steel
1oo L---~----.---~---. g l.E+02
u • Basemortar
100 135 160 180
Temperature (0 C) I.E ' 01
l.E-04 l.E-02 l.E+OO l.E+02
Reduced Frequency (Hz)
Figure 4. Viscosity for base mortar and 1% fibre-
mortar.
Figure 5. Complex modulus master curves for base
mortar and 0.5% fibre modified mortar.

results show an indication of the stiffening effect

and reduction in temperature susceptibility with t.E•07
fibre modification.
~I .E-06
3.3 Rheological complex modulus master curves
•
~ I.ETOS
The complex modulus G* for fibre modified =
asphalt mortar has been evaluated (Figures 5–7) ~ I.ET().l
=
E
, l%vol. Ctlluost
by constructed a complex modulus master curves , l%voi.Gim-l
~ l.E-03
using the Time Temperature Superposition Prin- Q.
l%voi.Giass-s
ciple (TTSP) at a reference temperature of 40ºC gt.E.,-02 • lo/o,'OI.Stt!l
·Base mortar
(Ferry, 1980). The concentration of fibres is fixed u
at 0.5% by bitumen volume. It reveals that the val- l.E-01
ues of complex modulus for mortar containing dif- I.E-04 I .E-O2 I.E+OO I.ET02
ferent types of fibre are slightly higher than that of Reduced Frequency (Hz)
base mortar in the whole range of frequencies and
temperatures except cellulose fibre, which showed Figure 6. Complex modulus master curves for base
a marked effect only at low frequencies and high mortar and 1% fibre modified mortar.
temperatures (60–70°C).
The master curves display a marked effect of
fibres on complex modulus for the 1% and 2% l.E+07
fibre at low frequencies and high temperatures
(50–70°C). The complex modulus master curves i' I.E 06
of fibre modified asphalt mortar exhibit slightly e:.
different behaviour from base mortar as shown Q l.ET0)
in Figure 7. The complex modulus master curves ..
=
of fibre modified asphalt mortar are significantly ~.. l.E+04
increased at low frequencies where the fibre entan- · 2%voi.Steel
glement is predominate with a slight plateau being = • 2%vol. Glass-!
~ l.E+03
formed for the 1 and 2% fibre addition. While, at Q. 2%vol. Glass-s
high frequencies and low temperatures, the high
stiffness of the base mortar control the effect of
~ l.E+02 , 2%vol. Cdlulose
• Base mortar
the fibre, the differences between base and modi- l.E-,.01
fied mortar are not as noticeable. However, the l.E-04 l.E-02 l.E+OO I.E+02
high fibre contents may help to entangle with each Reduced Frequency (Hz)
other to form a continuous network. This network
stabilises the composite by absorption, swelling or Figure 7. Complex modulus master curves for base
adhesion effects (Wu et al., 2014). mortar and 2% fibre modified mortar.

374
 It is interesting to note that adding 2% glass and able decrease in the phase angles (higher elastic
steel fibres showed a significant increase in com- response) of the highly modified mortars com-
plex modulus especially at low frequencies and pared to the base mortar. It is important to note
high temperatures. This result might be explained that there is clear difference in the extent of the 1%
by the fact that at these temperatures the stiffness and 2% fibre addition for the δ values mainly at
of the base mortar significantly decreased, where high temperatures (> 50ºC). While the phase angle
the high modulus of elasticity and continuous net- of base mortar approaches 90º (viscous behav-
work of fibres dominate the rheological properties iour) at high temperatures, the fibres significantly
of asphalt mortar. improve the elastic property. It seems possible that
these results are due to the viscosity of base mortar
became low enough to permit the continuous mat
3.4 Phase angle isochronal plots and master
of the fibre to affect the rheological properties of
curves
the fibre reinforced mortars (Goodrich, 1991) (see
Phase angle δ isochrones at 1.58 Hz for the base Figure 1 and 2).
and fibre modified mortars are shown in Figure 8. The phase angle master curves for all fibre addi-
The phase angle isochrones show a notice- tions show a decrease in phase angle, as shown
in Figure 9. The phase angles of the modified

90 100 • Base m IX1ar

85 • 0.5%,-ol.S teel
90
so •·•• tf~• Ojo/o,·ot Glass-s
• .::•..o~. ... · o:~%,-ol : Glass-1
80
c...75 ·. ...··:~Tf. • 0.5%\-oi.Celh.iose
70 c 70 ·:' . . 'i
Qi;
65
...§
~

~ 60 .:· );
-:,
3 60
0:: 55
50
-----+- 0 _5°ovoL Steel
- - 0 .5°o,-oL Glas~s
~
~

c..
50
40 ••
45 - - 0 .5°o'\"'l. Glass-1 30
---..c.-- 0 .5°o'\"'l. CellU.ose
40 20
25 35 -15 55 65 75
10
Temperature \C)
1.E-04 1£-02 1£ +00 1£+02
Reduced Frequency (Hz)
90
85 · Ba~m CCCU"

100 o. l%voi.Steel
so
........,.....
· lo/ovoi.G lass-1
c... 75 90 •
~
>< 1%'-ot. Ceth.iose
• 1o/ovol. Glass-s
70 80
::·~ ~
"El:
§ 65
... "2 70 ..·./'
'l.r
·.·;:~~~
~
"- 55
60 - Basemonar
- + - - - 1 °ovol. Steel
"§!>60
~ 50 .... -~.
...
50 - 1° ovol. G lass-1
_____,......._ I 0 ovol. Cellulose ~ 40
c..
45 -t~ovol.Glass-s 30
40 20
25 35 -15 55 65 75
10
Temperature ("C) l.E-04 l.E-02 1£+00
R educed Freque ncy (Hz)
90
85 100
80 90
c 75 so
"
Ot
70 €: 70
§ 65 "c
Oil 60

w"
~
60
----+--- Base mortar
"
~
50
c: 55
50
~ 2°·ovoL Celhiose
~ 2°ovol. Glass-s
0: 40
30
- 2°ovol.Glass-1
45
---+- 2°ovol. Steel + 20 '·
40 10
25 35 -1 5 55 65 75 1.E-04 LE-02 1.E+OO 1£ +02
Temperature ("C) R educed Fre quency (Hz)

Figure 8. Isochronal plots of phase angle at 1.58 Hz for Figure 9. Phase angle master curves for base mortar
base and fibre modified mortars. and fibre modified mortar.

375
mortars were significantly lower than the phase swelling and adsorption) between the modifier and
angle of base mortar within the range of tempera- bitumen or filler (Airey, 2003, Asgharzadeh et al.,
tures (50–70°C) and low frequencies, exhibiting a 2015).
higher elastic behaviour than viscous behaviour.
However, the stiffness of base mortar increased at
3.5 Rutting parameters
the range of low temperatures (< 50ºC) and higher
frequencies. This increase in base mortar stiffness The rutting parameter G*/Sinδ was used to evalu-
dominate the influence of the fibre entanglement. ate the effect of different fibre types on rutting
Therefore, the difference between the base and resistance of asphalt mortar according to the
modified mortars are not as marked. SHRP specification (Al-Mansob et al., 2014, Wu
The highly modified mortars exhibit different et al., 2015). It is well known that higher G*/Sinδ
behaviour from the base mortar with the pres- values indicate higher permanent deformation
ence of plateau zone at the middle. Data from resistance of asphalt mortar. The rutting factors
past studies suggested that the plateau zone gives of base and modified mortar with four fibre types
an indication of the existence of elastic networks at different testing temperatures were evaluated in
of modification, in addition the differences in the this investigation. The rutting parameter increased
phase angle master curves might be due to the dif- with the addition of high fibre content, implying
ferences in the molecular interaction (adhesion, a significant improvement in permanent deforma-
tion resistance at moderate and high temperatures
as shown in Figure 10. There is a slight increases
in G*/Sinδ at low fibre contents but there is a sig-
nificant increases in G*/Sinδ at high fibre contents
with the demonstration of a significant improve-
ment, indicative of a dominant fibre entangle-
ment (see Figure 1). The 2% vol. of glass and
steel fibres showed the highest increase in rutting
parameter for fibre modified mortar. This result
may be explained by the fact that glass or steel
fibre can tangle with each other to form a network
throughout the composite, this network reinforces
the matrix components and increases its ability to
withstand repeated applied, impact loading (Wu
et al., 2014).

4 CONCLUSIONS

This study evaluated the effect of fibres on the

mechanical and rheological properties of fibre
reinforced asphalt mortar, based on penetration,
viscosity and dynamic shear rheometer tests. The
following conclusions are offered:

• The rheological properties of mortar (bitumen

and filler) are enhanced by adding fibres as rec-
ognised by both common (penetration and vis-
cosity) and Dynamic Shear Rheometer (DSR)
parameters.
• Fibre properties such as shape, size and micro-
structure were observed using a Scanning Elec-
tron Microscope (SEM). These properties will
influence how fibres can reinforce mortar by
the formation of networks through the asphalt
mortar.
• Generally, the differences in rheological prop-
erties between base and modified mortars are
Figure 10. Rutting parameter for base mortar and fibre more notice at higher temperatures. Under these
modified mortar. circumstances that the viscosity of mortar is low

376
 enough to allow the fibre entanglement to con- EN, B. 2010. 13302: Bitumen and bituminous
trol the rheological properties of fibre reinforced binders—Determination of dynamic viscosity of
mortars. Thus may have led to increase service bituminous binder using a rotating spindle apparatus.
life of pavement or load bearing capacity. British Standards.
Ferry, J.D. 1980. Viscoelastic properties of polymers,
• The addition of fibre significantly improves the John Wiley & Sons.
rutting parameter G*/Sinδ, at moderate and Goodrich, J.L. 1991. Asphaltic binder rheology, asphalt
high temperatures. This implies that the addition concrete rheology and asphalt concrete mix proper-
of fibres may lead to improved rutting resistance ties (with discussion). Journal of the Association of
of asphalt mixtures. Asphalt Paving Technologists, 60.
Guan, B., Xiong, R., He, R., Chen, S.-F. & Ding, D.-H.
2014. Investigation of usability of brucite fber in
REFERENCES asphalt mixture. International Journal of Pavement
Research and Technology, 7, 193–202.
Abtahi, S.M., Esfandiarpour, S., Kunt, M., Hejazi, S.M. & Kristiansen, J. 2004. Bearing capacity development on
Ebrahimi, M.G. 2013. Hybrid reinforcement of low volume roads (a 15-year case study).
asphalt-concrete mixtures using glass and polypropyl- Mo, L. 2010. Damage development in the adhesive zone
ene fibers. Journal of Engineered Fibers and Fabrics, and mortar of porous asphalt concrete. PhD Thesis,
8, 25–35. Delft University of Technology.
Airey, G.D. 1997. Rheological characteristics of polymer Mohammed, B. & Mittal, O.P. 2016. Effect of Fillers on
modified and aged bitumens. PhD Thesis, University Bituminous Mixes. International Journal of Advanced
of Nottingham. Research in Education & Technology (IJARET), 3,
Airey, G.D. 2003. Rheological properties of styrene buta- 178–182.
diene styrene polymer modified road bitumens. Fuel, Peiwen, H. 2001. Study on Compatibility between SBS
82, 1709–1719. Modifier and Asphalt. Petroleum Processing and Pet-
Al-Mansob, R.A., Ismail, A., Alduri, A.N., Azhari, rochemicals, 32, 54–56.
C.H., Karim, M.R. & Yusoff, N.I.M. 2014. Physical Putman, B.J. 2011. Effects of fiber finish on the perform-
and rheological properties of epoxidized natural rub- ance of asphalt binders and mastics. Advances in Civil
ber modified bitumens. Construction and Building Engineering, 2011.
Materials, 63, 242–248. Tapkin, S., Çevik, A. & Usar, U. 2010. Prediction of
Alrajhi, A. 2012. Fiber dosage effects in asphalt binders Marshall test results for polypropylene modified
and hot mix asphalt mixtures. Master Thesis Master dense bituminous mixtures using neural networks.
Thesis, Arizona State University. Expert Systems with Applications, 37, 4660–4670.
Alvarez, A.E., Martin, A.E., Estakhri, C. & Izzo, R. Wu, M., Li, R., Zhang, Y., Wei, J., Lv, Y. & Ding, X. 2014.
2008. Determination of volumetric properties for per- Reinforcement effect of fiber and deoiled asphalt on
meable friction course mixtures. Journal of Testing high viscosity rubber/SBS modified asphalt mortar.
and Evaluation, 37, 1–10. Petroleum Science, 11, 454–459.
Asgharzadeh, S.M., Tabatabaee, N., Naderi, K. & Partl, Wu, M.-M., Li, R., Zhang, Y.-Z., Fan, L., Lv, Y.-C. &
M.N. 2015. Evaluation of rheological master curve Wei, J.-M. 2015. Stabilizing and rein-forcing effects of
models for bituminous binders. Materials and Struc- different fibers on asphalt mortar performance. Petro-
tures, 48, 393–406. leum Science, 12, 189–196.
Cao, D.-W. & JI, J. 2011. Evaluation of the long-term Xu, Q. & Solaimanian, M. 2008. Measurement and
properties of sasobit modified asphalt. International evaluation of asphalt concrete thermal expansion and
Journal of Pavement Research and Technology, 4, contraction. Journal of Testing and Evaluation, 36,
384–391. 1–10.
Chen, H.-X., Li, N.-L., Hu, C.-S. & Zhang, Z. 2004. Yu, X. & Sun, L. Anti-cracking ability of asphalt mixture
Mechanical performance of fibers-reinforced asphalt added with polyester fiber. Traffic and Transportation
mixture. Journal of Chang’an University (Natural Sci- Studies, 2010. ASCE, 1399–1406.
ence Edition) 24, 1–5. Zhu, H., Yang, J., Cong, L., Cui, J. & Wan, J. 2007.
EN, B. 2007. 1426: Bitumen and bituminous binders— Influence of polyester fibre on the performance of
determination of needle penetration. British Stand- Asphalt mixes. Southern African Transport Confer-
ards, 24, 1–5. ence (SATC).

377
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Effect of waste metallic particles on asphalt induction heating

B. Gómez-Meijide, H. Ajam & A. García

Nottingham Transportation Engineering Centre (NTEC), Department of Civil Engineering,
University of Nottingham, Nottingham, UK

H.S. Al Mahely
Palestinian Crossings Authority, Malehy Building, Rafah, Gaza Strip, Palestine

ABSTRACT: Asphalt is a complex visco-elasto-plastic material with very slow self-healing properties
that can be accelerated by means of induction heating. With this purpose, metal particles have to be
added to the mixture and heated once the first microcraks appear in the road. By using this method, it
is possible to extend the service life of roads by at least 30%. One of the main disadvantages of asphalt
induction heating is the increased production costs and environmental impact caused by the metallic
particles. However, this inconvenient might be minimized through the use of waste metallic particles. The
present investigation explores the use of 4 types of metal particles for asphalt induction heating, 2 of them
waste materials. These are (1) steel grit, (2) steel wool, (3) metal fibres from old tires and (4) steel shavings.
Results show that even contents below 0.45% can produce satisfactory heating and healing results.

1 INTRODUCTION The alternating electromagnetic field focus the

heating energy only on the conductive fibres
Conventional asphalt roads are made of a mix of embedded by the bitumen (Garcia et al., 2012),
aggregates and bitumen. Since the bitumen is a differently from conventional heating methods
liquid with very high viscosity (Read & Whiteoak based on heat radiation and conduction (Garcia
2003) it can flow, even at environmental tempera- 2012). As a consequence, the induction heating
ture, and self-heal the cracks that might occur due resulted to be very energy efficient, being able to
to traffic loads (Garcia et al., 2015). However, this increase the temperature of the bitumen above its
natural process is too slow (Qiu 2008) and requires flash point in just 1–2 minutes (Ajam et al., 2016).
the disruption of traffic for long periods of time Nevertheless, previous research showed that it
(Garcia et al., 2013a). In order to accelerate it, the is not necessary to reach so high temperatures,
viscosity of bitumen can be drastically reduced by being enough increases above the so-called New-
increasing the temperature of the material. One of tonian Temperature (normally between 50–70°C)
the most promising methods that has been recently in order to obtain an effective healing of cracks
developed in this regard, is the addition of metal (Sung et al., 2005, Heyes et al., 1994).
fibres to the asphalt mixture that can be heated Previous research has proven the important
by means of external electromagnetic induction role that the morphology and content of metal
(Garcia et al., 2013b). The differential heating fibres plays regarding the healing potential and the
between aggregates and bitumen and the conse- mechanical properties of asphalt mixtures (Garcia
quent differential thermal expansion, produces a et al., 2013c, 2014). That investigation reported
confining effect that forces the flow of bitumen that the fibres tend to produce clusters during the
through the cracks (Gomez-Meijide et al., 2016). first moments of mixing that can affect negatively
When the temperature decreases to ambient tem- the mechanical properties. Although these clusters
perature the cracks remain healed. can be dissolved during prolonged mixing times,
This not-invasive technology can be applied they are more likely to appear as long as the fibres
very easily on-site by means of a coil and a power content is increased and their diameter is reduced.
generator (Garcia et al., 2012) and, as the metal Nevertheless, other publications show that metal
particles remains in the mixture forever, the fibres can be used in asphalt mixtures in order to
induction heating can be applied as many times enhance their strength and fatigue resistance while
as necessary (Norambuena-Contreras & Garcia increasing ductility (Brown et al., 1990, Fitzgerald
2016), prolonging the lifetime of the road around 2000), what contributes to avoid the formation and
30% for each application (Menozzi et al., 2015). propagation of cracks (Maurer & Gerald 1989) and

379
improves the resistance to ageing, moisture damage 1. Steel grit, normally used for basting, is metal
and reflection cracking (Goel & Das 2004). Other granular material with uniform gradation
considered benefits of conductive particles in between 1 and 2 mm.
asphalt mixtures include improving electrical con- 2. Steel wool. This product was supplied as very
ductivity (Wu et al., 2002, 2005, 2006, Garcia et al., fine fibres with diameter ranging from 16 μm
2009), useful in monitoring (Wu et al., 2009) and to 72 μm and length from 0.15 mm to 5 mm.
defrosting applications (Wu et al., 2006). It is used in industrial applications, such as
Despite previous considerations, the method the reinforcement of materials exposed to high
also involves disadvantages, being the main ones, levels of abrasion (for example, car breaks
the increased production costs and environmental components).
impact of these mixtures due to the use of metal 3. Steel fibres from old tyres. This recycled product
particles. This inconvenient might be minimized was selected for its low cost and environmental
through the use of conductive by-products that interest. It was obtained as fibres with maximum
can be obtained in great quantities at very low length 30 mm and average diameter 1.1 mm.
price. At the same time, it would also contribute 4. Steel shavings from metal industry. This by-
to reduce the amount of heavy waste disposed in product of metal industry, such as cars manu-
landfills. facturing, was supplied at zero-cost, being the
The present investigation explores the use of dif- most economical of the studied kinds of parti-
ferent metal products and by-products, such as (1) cles. In this case, the helical shape of the fibres
steel grit, (2) steel wool, (3) metal fibres from old did not allow an accurate measurement of
tires and (4) steel shavings from the metal indus- lengths and diameters.
try and their affection on the healing properties of The natural aggregate was limestone (den-
asphalt mixtures. sity 2.67 g/cm3), whose gradation is graphically
described in Figure 2, and the binder used for all
the samples was 40/60 pen (density 1.03 g/cm3).
2 MATERIALS AND METHODS Due to its rounded shape, the steel grit did not
present any kind of mixing problem, as it behaved
2.1 Description of materials like a conventional aggregate. Additionally, because
The present investigation was carried out with the fine particle size of the steel wool, it also tended
the following 4 different kinds of metal particles to mix well providing it has been sprinkled to the mix
(Fig. 1): gradually. However, the shavings and especially the
fibres from old tyres tended to form clusters during
mixing resulting in mixtures not homogenous and
with high air voids contents (Fig. 3). It was checked
that the most critical kind of metal particle was the
fibres from old tyres (the clusters size for the tyre
fibres in some sections reached 55–60 mm2), not
being possible the addition of more than 0.45%
by volume of dry aggregate (1.2% by mass). Thus,
in order to compare the performance of mixtures
with the different kinds of fibre, all of them were
added in the same proportions (by volume): 0.15%,
0.225%, 0.30%, 0.375% and 0.45%.

100

g 80 I
I
1>0 1/
"'ra
"iii
VI
60
c.
40 I
·~"' I ~
ra 20
:; I
I
E r'"
u
:I 0
Figure 1. Appearance of the metal particles used in this 0.01 0.1 1 10 100
study: steel grit (top left), steel wool (top right), metal Sieve size (mm)
fibres from old tires (bottom left) and steel shavings
(bottom right). Figure 2. Gradation curve of asphalt samples.

380
Figure 3. Clusters and protruding fibres observed in
asphalt samples when old tyre fibres are used in a 0.45%
volumetric content.

Another problem detected when old tyre fibres

are used is that, due to their greater length and stiff-
ness, a series of fibres emerge from the surface even
after compaction (Fig. 3). This might cause safety
issues for road users if this kind of mixture is used
in superficial course layers. For this reason, the use
of this type of fibres needs to be restricted to lower
layers (i.e. base layers) providing that a course layer
with a different type of fibres is placed on top.

2.2 Manufacture of test specimens

First, 101.6 mm diameter × 50 mm height cylin-
drical specimens were manufactured by blending
920 g of natural aggregate with the correspond-
ing amount of metal particles and pre-heated at
160ºC. Then, 5.5% of bitumen (by mass of aggre-
gate) was added and mixed for 3 min. This mixing
time was enough in order to produce the homog-
enous distribution of metal particles, although the Figure 4. CT-Scans of metal particles after compaction
highest levels were achieved by the steel grit and of asphalt specimens. From top to bottom: steel grit, steel
wool thanks to their reduced size. This could be wool, fibres from old tyres and metal shavings.
checked by CT-Scans, as shown in Fig. 4 (samples
dimensions 30 × 30 × 12 mm). The compaction
was carried out by applying 50 blows of Marshall under 3-point bending conditions (Fig. 5) in
Hammer to each end of the specimens. strain control with an increasing load ramp at
The test samples were produced by cutting the a constant deformation rate of 50 mm/min. The
cylindrical specimens through the diametrical ultimate force applied at the moment of break
plane by a radial saw blade suitable for concrete is registered as Fi. Then both halves were stored
and stone materials and a notch was cut at the at 20 ± 2ºC for 4 hours, put together again into
midpoint of the diametrical plane in the direc- a mould and heated by applying an alternating
tion of the loading, with thickness = 3 mm and electromagnetic field. The electromagnetic field
depth = 5 mm (Fig. 5). was obtained by circulating an alternating cur-
rent of 80A through a 15 × 15 cm squared coil
composed by three windings and placed at a dis-
2.3 Testing of asphalt self-healing properties
tance of 1 cm above the samples. For each kind
The samples were cooled down to −20 ± 2ºC (to of the studied conductive particles, 6 samples
ensure a brittle crack) and split in two halves were manufactured, healing each of them for a

381
 F Notch

Figure 6. Temperature changes during the heating and

cooling processes.

the concept of healing energy was defined in (Ajam

et al., 2016) as the area below the temperature-time
Figure 5. Setting for healing test. curve that a certain samples experiments during
both heating and cooling stages. The expressions
that mathematically give such an area, in units of
different heating time (20, 40, 60, 80, 100, 120 s). K⋅s are given by Eq. 2 and 3 for heating and cool-
Finally, the healed samples were cooled again ing phases respectively:
to –20 ± 2ºC and the 3-point bending test was
Tss Taair
repeated under the same conditions explained
above. The new ultimate force obtained at the
τ h (t ) Tss ⋅ t +
kh
(e kh t
− ) (2)
moment of the second brake was registered as Ff
and compared to the initial strength through the
concept of Healing Ratio (HR), giving an idea τ c (t ) Taair ⋅ (t theat ) +
Tmax Taair
kc
( −e kc t theat
heat )
) (3)
about the percentage of initial strength that was
recovered thanks to the healing process:
where τh and τc are the healing energies obtained
Ff during the heating and cooling stages; kh and kc are
HR(%) = (1) heat transfer coefficients (s−1) during the heating or
Fi the cooling phase respectively. Tair is the ambient
temperature (293 ± 2 K); Tss is the steady state tem-
perature that the samples reach after a certain heat-
3 THEORETICAL FRAMEWORK ing time and never exceed, since the heating energy
is balanced with the energy dissipation to the envi-
3.1 Healing energy ronment (in K); and theat is the heating time (s).
The main parameters affecting induction heating Since the samples were cooled at ambient tem-
are heat gains coming from Joule losses in the fibres perature for 4 hours and then stored at −20 ± 2ºC,
and heat losses coming from the thermal radiation, time at which the healing process was considered
convection with air and conduction with other lay- finished, the total energy will be calculated as the
ers of the pavement (Garcia et al., 2012). However, energy applied during the whole heating stage
the healing of a crack in an asphalt material is a (t = theat) plus the energy involved in 4 hours of
process that not only occurs during the heating, cooling stage (t = 14,400 s):
but also during the cooling of samples (Fig. 6), as
long as the temperature remains above a critical τh ( ) + τc ( ) (4)
temperature normally so-called Newtonian tem-
perature (Tnewt) (Ajam et al., 2016). Furthermore,
3.2 Asphalt self-healing theory
the healing will be more effective if higher temper-
ature and/or longer heating time is applied. Hence, In (Garcia et al., 2013a) it is explained that healing
in order to fairly compare the healing results of of cracks in asphalt mixture starts when both faces
samples that were heated for different times and/or of a crack are in contact. Then, bitumen can drain
temperatures (as in the present investigation, since from the mixture into the cracks, healing them.
not all the fibres have the same heating potential), In (Garcia et al., 2013a) the capillary phenomena

382
were studied through a modification of the Lucas– 60%
Washburn equation (Hamraoui & Nylander 2002, ~50%
Washburn 1921, Duarte et al., 1996). This research
will follow this theory. According to the model, the
"'
~40% •
•
0.15
0.22
·~ 30%
healing happens until it is reached an equilibrium bO
• 0.3
• 0.37
between surface tension, gravity and energy dissi- :§ 20%
• 0.45
pation due to friction. According to Garcia et al., "'"' 10%
(2013a), after solving the balance of forces and
:I:
•
0%
integrating, the healing level S(τ) can be predicted 4.22E+06 4.24E+06 4.26E+06 4.28E+06 4.30E+06
using the following equation: Healing energy (K·s)

Dτ 2
C1 − Dτ ⎛ ⎞ Figure 7. Relationship between the healing ratio and
S (τ ) = ⋅e 1+ e 2 ⎟ (5) the healing energy and fitting of the model to the experi-
F0 ⎝ ⎠
mental data obtained by samples with different contents
of steel wool (0.15%, 0.22%, 0.30%, 0.37% and 0.45%).
where S(τ) is the healing rate or percentage of
recovered strength after the healing treatment (%),
F0 is the initial 3-point bending strength of the test 50%
samples (kN), τ is the energy applied during the
~ 40%
healing (K⋅s) and D and C1 are parameters that can e:. • Tyre
be calculated as: ._g 30% • Shavings
:::
ρ ggr .~ 20% Wool
D= (6) -;;
β :::g 10%

0%
σ u ⋅C
C1 = 8 (7) 0.15 0.22 0.3 0.37 0.45
L H Fibre content(%)

where ρ is the density (kg/m3), g is the gravity (m2/s), Figure 8. Healing ratios obtained with different con-
r is the width of the crack (m), β is a dimensionless tents of different types of conductive particles after heat-
parameter that takes into account possible sources ing for 120 s.
of energy losses, σu is the maximum force resisted
by the beam (N), L is the span of the beam (m), H
is its height and C is a material constant with units not so sharp. In order to illustrate it, a reference
(m2). Furthermore, the D value is an indication of healing energy was chosen (in this case 4.3⋅106
the healing rate, when its value increases, the heal- K⋅s, obtained approximately after 120 s of heat-
ing happens faster. ing) and the healing ratios obtained by different
In the present research, this model will be fitted contents of different types of particles were gath-
to the experimental data in order to better clarify ered in Figure 8. Besides the mentioned increas-
the trends followed by the results and compare ing correlation, it is also noticeable that steel grit
them depending on the type of metal particle that tends to produce, in general, the highest healing
was added to the mix. values (average 38.4%), followed by the tyre fibres
(33.4%) and the shavings (18.5%). With steel wool,
it is possible to obtain healing ratios comparable to
4 RESULTS AND DISCUSSION the grit but only for the highest studied contents, as
for the lowest, the performance resulted even close
As an example, the experimental healing ratios to the shavings.
obtained by samples with steel wool at differ- Once explained this, it must be also mentioned
ent healing energies can be seen in Figure 7 for that, as explained above, due to their greater length
asphalt samples containing 5 different metal con- the tyre fibres tended to make clusters during the
tents. Moreover, the model described in Eq. 5–7 mixing when the content exceeded 0.45%. In this
was fitted to these data, unravelling the correla- case, none of the studied samples contained more
tion between all these factors. As expected, a direct than that but it was detected than after splitting
correlation between the healing ratio and both the the samples by the first time, it was very difficult
healing energy and the metal content in the mix to put both sides of the crack together due to the
could be observed. interfering fibres. This fact could have produced
This double increasing trend was also found reductions on the experimental healing ratios,
for the other sorts of metal particles, although especially for the highest contents. Proof of this,

383
are the satisfactory results obtained for low con- coil and metal, these fibres become again very
tents and the exceptional temperature that these recommendable for lower layers (i.e. base layers).
fibres can gain under electromagnetic induction, Finally, and according to Garcia et al., (2013c), it is
compared to the rest (in Fig. 9 it can be seen the expected that the use of this type of fibre but with
evolution of temperatures over heating time for reduced length would provide a better distribution
samples with 0.3% of different types of fibre, while and less clustering, what would minimise the issues
in Fig. 10 it is compared the temperature after 100 commented above.
s for all contents). In a real case, the cracks to be On the other hand, steel grit and steel wool
healed would be product of small deformations at are more advisable for their use in superficial lay-
a micro-scale, remaining both sides of the cracks ers, close to the induction coil. Due to the poorer
very close to each other. Therefore, the perform- results obtained with metal shavings, further inves-
ance of this kind of fibre is expected to be signifi- tigation would be advisable in order to improve its
cantly better than as shown by this simplified test. healing capacity.
As can be seen, the temperatures obtained by Finally, the increasing trends show that in order
these fibres can even reach the bitumen flashpoint. to improve the healing ratios, it would not be nec-
This is obviously counterproductive, not only essary to increase the content of metal particles,
due to the damage that this can produce to the and consequently the price of the asphalt mixture.
binder, but also from the point of view of health For this purpose, it would be enough to increase
and safety of workers implementing the healing the healing energy by increasing the current inten-
technology. Nevertheless, it must be taken into sity through the induction coil (in this case only
account that these results were obtained with a dis- 80 A were used) and/or the heating time (for the
tance between coil and sample of 1 cm. As these present research no times longer than 120 s were
fibres were already rejected to be used in superfi- studied).
cial course layers due to the emerging spikes and
considering that the induction heating potential
decreases exponentially with the distance between 5 CONCLUSIONS

Since the fibres from old tyres tended to produce

250 clusters during the mixing process, the maximum
~ 200 volumetric content was limited to 0.45%, although
the rest of particle types, due to their smaller size,
z
Q)

~
150
could have been added in greater proportions.
Q)
c_ 100 Nevertheless, the healing results were satisfactory
E
Q)
>-- 50 with such a low contents, at low induction inten-
0
sity (80A) and with a maximum testing time of
0 20 40 60 80 100 120 only 2 min, being enough to increase any of these
Heating time (s)
parameters to reach higher healing results. In any
case, the objective of the present paper was not the
Figure 9. Evolution of temperature with the heating optimisation of any of them, but a comparison of
time for samples with different types of metal particles properties obtained with the four types of metal
(content 0.3%). particles. From such a study, the following conclu-
sions could be obtained:
1. For a given heating time, the fibres from old
>2002C >200gC
200 tyres increase the temperature of the samples
significantly more than the rest of the fibres.
~ 150 • Tyre This difference was not observed in the healing
Q) results but probably due the configuration of
£iloo the healing test. It is expected that this type of
~
Q)
c_ fibre produces healing rates significantly higher
E
Q)
so than those observed when only microcracks are
E-
produced in real conditions.
0
0.15 0.22 0.3 0.37 0.45
2. For all the fibres studied, the healing ratio is
Metal particles content(%) proportional to the metal particles content
and to the applied healing energy. Steel grit and
Figure 10. Temperature of samples with different wool produced healing results very similar and
contents of different sorts of metal particles after being satisfactory, while the results obtained by the
heated for 100 s. shavings were considerably lower.

384
 Taking into account the previous conclusions, in REFERENCES
order to increase the healing ratios, the amount of
metal particles (with exception of fibres from old Ajam, H., Lastra-González, P., Gómez-Meijide, B., Gar-
tyres) could be increased, what would also increase cia, A. 2016. Self-Healing of Dense Asphalt Concrete by
the price of the mixture. As an alternative, it would Two Different Approaches: Electromagnetic Induction
be enough to increase the healing energy by increas- and Infrared Radiation. In Proceedings of 8th RILEM
International Conference on Mechanisms of Crack-
ing the current intensity through the induction coil ing and Debonding in Pavements. Springer:241–246.
(in this case only 80A were used) or the heating Brown, S.F., Rowlett, R.D. & Boucher, J.L. 1990.
time (for the present research no times longer than Asphalt modification. In Proceedings of the confer-
120 s were studied). ence on the United States strategic highway research
From the studied types of fibre, it can be recom- program: sharing the benefits. London: Thomas
mended the use of fibres from old tyres for lower Telford:181–203.
layers (i.e base layers). These fibres have great Dean, D.A. & Gerald, M. 1989. Field performance of fab-
heating potential what can be translated to higher rics and fibers to retard reflective cracking. Geotextiles
healing ratios, even when they are placed far from and Geomembranes 1248:239–267.
Duarte, A.A., Strier, D.E. & Zanette, D.H. 1996. The
the surface and the induction coil. In addition, rise of a liquid in a capillary tube revisited: a hydro-
their waste nature make them a cheaper and more dynamical approach. American Journal of Physics
sustainable product whose use in base layers (the 64:413–418.
thickest one) can contribute to reduce the ecologi- Fitzgerald, R.L. 2000. Novel applications of carbon fiber
cal impact of roads without increasing the costs for hot mix asphalt reinforcement and carbon–carbon
and needs of raw materials in a significant way. pre-forms. MSc Thesis. Department of Chemical
For superficial course layers, it is advisable the Engineering, Michigan Technological University.
use of steel wool or grit. They behave very similar Garcia, A. 2012. Self-healing of open cracks in asphalt
and satisfactory when placed close to the surface. mastic. Fuel 93:264–272.
Garcia, A., Schlangen, E. & van de Ven, M. 2010. Two
In addition, they can be mixed very easily, not pro- ways of closing cracks on asphalt concrete pavements:
ducing clusters or emerging spikes. microcapsules and induction heating. Key Engineering
Materials 417–418, 573–576.
Garcia, A., Schlangen, E., van de Ven, M. & Liu, Q.
ACKNOWLEDGEMENTS 2012. A simple model to define induction heating in
asphalt mastic. Construction and Building Materials
The authors would like to acknowledge that the 31:38–46.
research leading to these results has received Garcia, A., Bueno, M., Norambuena-Contreras, J. &
funding from the UK project EPSRC EP/ Partl, M. N. 2013a. Induction healing of dense asphalt
M014134/1, “Induction heating for closing concrete. Construction and Building Materials 49:1–7.
cracks in asphalt concrete” and Infravation (an Garcia, A., Norambuena-Contreras, J. & Partl, M.N.
2013b. Experimental evaluation of dense asphalt con-
ERA-NET Plus on Infrastructure Innovation) crete properties for induction heating purposes. Con-
under the grant agreement no. 31109806.0003 - struction and Building Materials 46:48–54.
HEALROAD. Funding partners of the Infrava- Garcia, A., Norambuena-Contreras, J., Partl, M.N. &
tion 2014 Call are: Ministerie van Infrastructuur Schuetz, P. 2013c. Uniformity and mechanical proper-
en Milieu, Rijkswaterstaat, Bundesministerium ties of dense asphalt concrete with Steel wool fibres.
für Verkehr, Bau und Stadtentwicklung, Danish Construction and Building Materials 43:107–117.
Road Directorate, Statens Vegvesen Vegdirek- Garcia, A., Norambuena-Contreras & J., Partl, M.N.
toratet, Trafickverket-TRV, Vegagerdin, Minis- 2014. A parametric study on the influence of steel wool
tere de l’Ecologie du Developpement Durable fibers in dense asphalt concrete. Materials and Struc-
tures 47:1559–1571.
et de l’Energie, Centro para el Desarrollo Tec- Garcia, A., Bueno, M., Norambuena-Contreras, J. &
nológico Industrial, Anas S.P.S, Netivei Israel— Partl, M.N. 2015. Single and multiple healing of porous
National Transport Infrastructure Company and dense asphalt concrete. Journal of Intelligent
LTD and Federal Highway Administration Material Systems and Structures 26(4):425–433.
USDOT. The HEALROAD project is being car- García, A., Schlangen, E., van de Ven, M. & Liu, Q.
ried out by the University of Cantabria, Univer- 2009. Electrical conductivity of asphalt mortar con-
sity of Nottingham, German Federal Highways taining conductive fibers and fillers. Construction and
Research Institute (BASt), European Union Building Materials 21(10):3175–3181.
Road Federation (ERF), Heijmans Integrale Goel, A. & Das, A. 2004. Emerging road materials and
innovative. In: National conference on materials and
projecten B.V. and SGS INTRON B.V. their application in civil engineering. Hamipur, India.
The authors would also like to acknowledge Gómez-Meijide, B., Ajam, H., Lastra-González &
the financial support provided by the Ministry of P., Garcia, A. 2016. Effect of air voids content on
Higher Education and Scientific Research in Iraq, asphalt self-healing via induction and infrared heating.
which provided funding for the present research. Construction and Building Materials 126:957–966.

385
Hamraoui, A. & Nylander, T. 2002. Analytical approach Sung, Y.T., Kum, C.K., Lee, H.S., Kim, J.S., Yoon,
for the Lucas–Washburn equation. Journal of Colloid H.G. & Kim, W.N. 2005. Effects of crystallinity and
and Interface Science 250:415–421. crosslinking on the thermal and rheological proper-
Heyes, D.M., Mitchell, P.J. & Visscher, P.B. 1994. Viscoe- ties of ethylene vinyl acetate copolymer. Polymer
lasticity and near-newtonian behaviour of concentrated 46(25):11844–11848.
dispersions by Brownian dynamics simulations. Trends Washburn, E.W. 1921. The dynamics of capillary flow.
in Colloid and Interface Science 97:179–182. Phys Rev 13(3):273–283.
Menozzi, A., Garcia, A., Partl, M.N., Tebaldi, G. & Wu, S.P., Mo, L.T. & Shui, Z.H. 2002. An improvement
Schuetz, P. 2015. Induction healing of fatigue damage on electrical properties of asphalt. Journal of Wuhan
in asphalt test samples. Construction and Building University of Technology. Materials Science Edition
Materials 74:162–268. 17(4):63–65.
Norambuena-Contreras, L. & Garci, A. 2016. Self-heal- Wu, S., Liu, X., Ye, Q. & Li, N. 2006. Self-monitoring
ing of asphalt mixture by microwave and induction heat- electrically conductive asphalt-based composite with
ing. Materials and Design 106:404–414. carbon fillers. Transactions of Nonferrous Metals
Qiu, J. Self-healing of asphalt mixes: literature review. Society of China 16: 512–516.
2008. Report 7-08-183-1. Delft University of Technol- Wu, S.P., Mo, L.T., Shui, Z.H. & Chen, Z. 2005. Investi-
ogy, Delft. gation of the conductivity of asphalt concrete with con-
Read, J. & Whiteoak, D. 2003. The shell bitumen hand- ductive fillers. Carbon 43(7):1358–1363.
book. Thomas Telford.

386
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Improving the raveling resistance of porous asphalt with

kraft lignin modified bitumen

N.R.Z. Poeran & B.W. Sluer

Boskalis, Nieuw Vennep, The Netherlands

M.F.C. van de Ven & W.F. Gard

Faculty of Civil Engineering and Geosciences, Delft University of Technology (DUT), Delft, The Netherlands

ABSTRACT: Porous Asphalt (PA) is used extensively as a surface layer on highways in the Netherlands.
PA has a relatively high void content and because of this possesses excellent noise absorbing and drainage
properties. It however brings with it a large disadvantage too; PA surface layers require maintenance quite
often. The main failure mode, in 75% of occasions, is raveling, which is defined as loss of aggregate from the
surface of the asphalt pavement. An experimental program was set up at DUT to explore the effects of lignin
modification on the raveling resistance of PA. A penetration bitumen 70/100 was used as a reference and was
modified with both 5% and 10% kraft lignin. Subsequently, mortars (fresh and aged) and PA specimens were
produced and empirical, rheological and mechanical testing was carried out. From the results it is concluded
that kraft lignin counteracts ageing and drainage but low temperature characteristics are not improving.

1 INTRODUCTION Lastly, especially drainage of the mortar during

transport and laydown causes a thickness reduc-
PA (Porous Asphalt) is a mixture designed accord- tion of the mortar bridges present near the pave-
ing to the gap-graded principle, meaning that the ment surface making them more prone to failure.
aggregate gradation and skeleton are dominated by As all three failure mechanisms seem to con-
the coarser fractions. Accordingly, PA has a mini- cern the (evolution of) the binder/mortar behavior
mum void content of 20% and as a result exhibits it was decided to modify the bitumen binder. The
excellent drainage and noise absorbing properties. modifying agent should be able to reduce or delay
It therefore enlarges road safety during precipita- oxidation, improve bitumen-aggregate adhesion
tion and enables residential areas and highways and increase binder viscosity without compromis-
to exist alongside each other. Because of this the ing the PA mixture’s workability.
use of PA as a surface layer often is obligatory on A modifying agent that could possess all of
national highways in The Netherlands. However, these properties is lignin, a biopolymer that serves
also due to the high void content PA surface layers as an important building block in the cellular
are especially susceptible to raveling. structure of higher order, vascular plants. In plants
Raveling is defined as loss of aggregate from the it provides rigidity of the cell wall and it serves
pavement surface. It is considered the main failure as an anti-oxidant. Furthermore, because of its
mode, ≥75%, for PA surface layers. Aside from the high phenol content lignin is often used as a base
gradual loss of mentioned favorable PA character- compound for adhesives in the chemical industry.
istics raveling can be a precursor for consequential In this paper the effects of a kraft lignin bitumen
damage such as pothole formation. modification on the raveling properties of PA are
In an attempt to improve the raveling resistance reported and discussed.
of PA at material level a research study was set up.
The three factors deemed most influential are: age-
2 LITERATURE REVIEW
ing of the bituminous binder, stripping and mor-
tar drainage in the vertical plain. Due to the high
2.1 Kraft lignin
void content of PA the mixture is highly exposed
to the elements, which causes accelerated ageing of Lignin is an organic compound found in the cell
the mortar bridges. Consequently, the bitumen at the walls of green plants. Contrary to most biopoly-
surface becomes brittle and the chance of crack for- mers lignin is amorphous. It is a three-dimensional
mation increases. Stripping is best explained as the structure in which ordered repeating units cannot
breakdown of the adhesive bond between the bitu- be found. Not much is known about the chemis-
men and the aggregate caused by moisture ingress. try of lignin mainly due to its erratic structure.
387
However, it is known that the biopolymer is built and yields on average 60000 tons of highly pure
from three basic monolignols via oxidative phenolic biopolymer a year. It renders relatively strong paper,
coupling reactions. The heterogeneity of lignin hence the name kraft meaning ‘strength’ in German.
polymers exists in molecular composition and link-
ages types between the phenylpropane monomers:
2.2 Previous experiences with lignin modified
syringyl-, guaiacyl-, and p-hydroxyphenylunits.
bitumen binders
(Gosselink, 2011). These basic building blocks of
lignin are depicted in Figure 1. Three relevant studies were found in which a lignin
Similar to lignin the resin fraction of bitumen or co-biopolymer modification of asphalt was
contains a considerable amount of unsaturated, aro- investigated. It is noted that two of these studies
matic structures joined by alkyl chains, Figure 2. It intended to use lignin as a bitumen extender not as
is believed that the affined ring structures in lignin a modifying additive.
and bitumen make the two materials highly compat- In Iowa State University (Willams & McCready,
ible. Additionally, phenols are the functional groups 2008) the effects of a lignin addition to bituminous
associated with anti-oxidation and adhesive proper- binders was investigated assuming that lignin as an
ties. The phenol content of lignin is however very anti-oxidant in asphalt binder could help to slow
much dependent on the pH of the extraction proc- down oxidative ageing and delay pavement failures
ess (Faustino, 2010). The extraction process can thus and aging problems. Four lingo-cellulosic co-prod-
prove to be defining for the effectiveness of a kraft ucts were mixed with bitumen binders from 3% up
lignin bitumen modification. to 12% by weight. The result was an improvement
To extract lignin from plant tissue requires of the high temperature properties, but a worsen-
chemical procedures. These procedures affect the ing of the low temperature properties. Reference
structure of the lignin and thus also affect the samples without lignin aged significantly more
properties of the material. Since lignin structures than samples with lignin.
coming from an individual plant can vary, no At the University of Connecticut (Sundstrom,
extracted lignin is the same, even when the method 1983) two bitumen types, AC-10 and AC-20, were
of extraction and the tissue of origin are the same. modified with lignin obtained from steam-explod-
In this research study use is made of kraft lignin ing aspen wood chips. They concluded that a mix-
from sapwood of pine wood. Lignin originating ture with 6% total binder content of AC-20 binder
from softwoods commonly exists for ≥95% out of has similar properties as a mix with 30% lignin
guaiacyl units (Gosselink, 2011). Essentially, kraft in AC-10 binder, but the mix with 30% lignin in
lignin is a waste product of the kraft pulping proc- AC-10 should be paved at a 20°C higher tempera-
ess. This process is principal in the paper and pulp- ture than the same mix using unmodified AC-20.
ing industry for the extraction of lignin from wood At the National Center for Asphalt Technology
of the Auburn University (Kandhal, 1992) report
was made that lignin can be used as a substitute for
bitumen or as an extender in HMA. A 30 percent
replacement of bitumen appeared feasible with no
significant effect on physical properties. Based on
limited laboratory experiments it was concluded that
HMA mixtures containing lignin-asphalt binders
can be designed to match the structural strength of
gualacyt unit syringyt unit S·hydroxyguaiacyt unit
HMA mixtures containing conventional materials.
From the literature it is thus suggested that lignin
Figure 1. Phenylpropane monomers—building blocks could potentially reduce ageing of bitumen bind-
of lignin. ers. On the other hand literature also indicates that
detailed attention should be paid to the low tem-
perature characteristics of lignin modified bitumen
and the workability of the lignin modified asphalt

_X_*:~
mixture.

---x=H IO H 3 OUTLINE OF THE RESEARCH

PROGRAM
Alenoli<: (1) 2-Quinolone type (I)

3.1 Experimental program

Figure 2. Some functional chemical groups naturally To map the effect of a kraft lignin bitumen modifi-
present in bitumen (Hagos, 2008). cation on the raveling resistance of PA a four stage

388
experimental program was set up (Figure 3). Stages containing 10% kraft lignin by bitumen mass. Mor-
II, III and IV were dedicated to the binders, the tars were then produced by mixing in filler, fine
mortars (fresh & aged) and PA respectively. The crushed sand (<500 μm) with the binders accord-
tests performed in stage II were mainly to map the ing to the ratio bitumen: filler: sand = 0.34: 0.30:
change in base properties of the reference binder 0.36 by mass. The PA mixture was composed of
through empirical experiments and Dynamic Shear (modified) binder, filler, crushed sand and Bestone
Rheometer measurements. In stage III an assess- aggregate > 2 mm graded according to the Stand-
ment of the impact of kraft lignin on mortar ageing aard RAW Specifications. For both the mortars
was made by means of Fourier Transform Infrared and PA mixtures a limestone filler containing at
Spectroscopy (FTIR), Dynamic Shear Rheometer least 25% calcium hydroxide (Ca(OH)2) was used.
(DSR) measurements and Direct Tensile Tests
(DTT). It was decided to age the mortars, not the
3.3 Bitumen modification procedure
binders, to involve possible chemical influence of
filler. During stage IV the drainage sensitivity of The reference bitumen is modified with kraft lignin
modified PA mixtures was compared to that of the by means of a high shear mixer. The stator fitted
reference mixture. However, before starting stages onto the mixer has square holes. These square holes
II to IV a pre-research stage I was implemented. In strongly increase the shear rate during mixing and
stage I the used base materials were characterized. therefore promote the production of homogeneous
Figure 3 gives an overview of the tests performed binders.
during the four experimental stages. An integral Before mixing, a can filled with 100 grams of
account of the experimental program including reference bitumen was placed in a 160°C oven for
detailed descriptions of testing methods and test half an hour. This mixing temperature was chosen
conditions reference is made to (Poeran, 2014). because the Differential Scanning Calorimetry-
Thermogravimetric Analysis (DSC-TGA) performed
on the kraft lignin showed its softening point to be at
3.2 Materials and mixture compositions
160°C. After the bitumen had softened the can was
It was decided to first modify the binder with kraft placed on a hot plate set at 160°C and remained there
lignin and subsequently produce the mortars and during mixing.
PA with the modified binders. All binders, mortars To start mixing the head of the high shear
and asphalt mixtures are listed in Table 1 accompa- mixer was lowered into the can and the shear rate
nied with their tag. was gradually increased. Throughout the shear
Aside from the reference bitumen, two kraft rate increase the temperature of the bitumen was
lignin modifications have been mixed; one con- monitored because the constant shearing caused
taining 5% kraft lignin by bitumen mass and one the material to heat as well. When the equilibrium
shear rate for the reference bitumen was reached
kraft lignin was added bit by bit. While adding the
STAGE I . CHARACTERIZATION OF BASE MATERIALS kraft lignin the can was moved from time to time to
ensure proper mixing. While adding the lignin the
~MEN70/100 ~FT LIGNIN

PARTICLE SIZE DISTRIBUTION

shear rate had to be increased constantly to keep
PENETRATION TEST
SOFTENING POINT TEST PYCNOMETER DENSITY the mixing temperature at 160°C. Interestingly, the
DSC.TGA ESEM
FTIR Dsc.TGA equilibrium shear rate for the 5% and 10% modifi-
FTIR
cations did not differ and was found to be 4800 rpm.
STAGE II . BITUMEN TESTING STAGE Ill . MORTAR TESTING STAGE IV . PA TESTING

SOFTENING POINT TEST FTIR (EXTRACTED BINDER) SCHELLENBERG

DSR FREQUENCY SWEEP
DSR VISCOSITY
DSR FREQUENCY SWEEP
DIRECT TENSION TEST
3.4 Method of mortar ageing
For the ageing of the mortar the LOT virgin age-
Figure 3. Experimental program (Poeran, 2014). ing method (Khedoe & Moraal, 2007) was chosen.
The LOT virgin ageing method entails the heating
of 2mm thick mortar slabs in an oven for 1.5 hours
Table 1. Overview of mixtures with tags.
at 175°C. It was assumed that this ageing protocol
Tag would produce short term aged mortars only.

Modification Aged Porous

[%m/m] Bitumen Mortar mortar asphalt 4 RESULTS

0 BR MR MRa PAC_R 4.1 Stage I—Characterization of base materials

5 BL_5 ML_5 ML_5a PAC_L_5
10 BL_10 ML_10 ML_10a PAC_L_10
During the first stage of the experimental pro-
gram basic properties of the bitumen 70/100 and

389
the kraft lignin were determined. These properties
have been useful in optimizing the modification
procedure and provide insight into the non-me-
chanical behavior of the modified mixes, such as
segregation sensitivity.

4.1.1 Basic properties of bitumen 70/100

On the reference bitumen four tests were per-
formed; penetration test, ring & ball test, FTIR
and DSC-TGA. The results from the penetration
and ring & ball test are provided in Table 2.
The FTIR absorption spectrum showed a char-
acteristic bitumen profile. This is in line with the
results from the penetration and ring & ball test,
both of which yielded results characteristic for this
type of bitumen. The DSC-TGA measurements
failed unfortunately. Figure 4. ESEM images of kraft lignin.

4.1.2 Basic properties of kraft lignin

The kraft lignin used in this study is lignin extracted
from pine, a softwood. It is a fine, brown powder Table 3. Softening point results for reference and modi-
fied binders.
that looks and feels like filler. DSC-TGA measure-
ments performed on the kraft lignin showed its Mean softening point
softening point to be at 160ºC. Visually no change
takes place during the softening of kraft lignin Binder [ºC]
apart from minor lump formation.
To obtain a visual image of the kraft lignin a BR 48.3
particle size distribution was made by image count BL_5 48.0
and ESEM imaging was made. The particle size BL_10 50.0
distribution showed that the bulk of the kraft lignin
particles are between 0.5 μm to 1.5 μm in length
and in width. The ESEM images are depicted in
Figure 4. From Figure 4 it appears that kraft lignin 1025 kg/m3 this is relatively high. However, from
particles are hollow, porous grains with dimen- the ESEM imaging follows that the grains are
sions ranging from 100–200 μm. The image count porous and hollow. The chance of segregation
and ESEM images implicate these grains break occurring is deemed low.
easily leaving smaller kraft lignin shards.
Subsequently, the density of kraft lignin was
4.2 Stage II—Bitumen testing
determined with a (helium) pycnometer. The den-
sity of kraft lignin was measured at 1305 kg/m3. After the production of modified binders the sof-
Compared to the density of bitumen 70/100 of tening point was determined with the ring & ball
test, the mastercurve of the complex modulus and
Table 2. Penetration and softening point results for BR.
the phase angle and the viscosity at 180ºC were
measured with the DSR set-up.
Penetration Mean Table 3 contains the softening points of the
modified binders. BL_10 shows a slight increase of
Sample no. [0.1 mm] [0.1 mm] about 5% compared to BR. The changes are not
significant though. The DSR viscosity measure-
#1 73 71 75 73.0 ments correspond with the softening point results.
#2 76 78 76 76.7 Figure 5 shows that at low shear rates the viscos-
#3 75 76 78 76.3 ity of BL_5 and BL_10 are approximately 10 times
Softening point Mean higher than that of BR. For the purpose of pre-
venting drainage this is a favorable change in binder
Sample no. [ºC] [ºC] response. At higher shear rates the lignin modified
#1 47 47 47
binders show shear thinning behavior; their viscosi-
#2 50 50 50
ties decrease to a level close to that of BR, which
#3 48 48 48 indicates that no increase of mixing temperature is
needed to produce quality modified PA.

390
 10 , - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Table 4. Ratios of G* with respect to BR.

Frequency
[Hz]
Temperature
G*/G*(BR) [ºC] 0.01 1 10
BL_5 −10 1,01 1,02 1,02
20 1,02 1,03 1,05
50 1,19 1,21 1,19
BL_10 −10 1,22 1,11 1,08
100 200 300 400 500 600
20 1,57 1,40 1,31
SHEAR RATE [1/S]
50 1,75 1,46 1,42
- B R - - BL_5 - - - BL_10

Figure 5. DSR viscosity measurements binders at

180ºC.
Table 5. Ratios of δ with respect to BR.

Frequency
From Figure 5 follows that BL_5 is more vis- [Hz]
Temperature
cous than BL_10 for all shear rates, which was δ/δ(BR) [ºC] 0.01 1 10
unexpected. It is therefore not clear if the kraft
lignin bonds with the bitumen or simply disperses BL_5 −10 0,94 1,00 0,98
throughout, like a filler. It could also point towards 20 1,00 1,01 1,01
the existence of an optimum modification content 50 0,99 1,02 1,00
at which maximum entanglement is achieved.
For all binders DSR frequency sweeps were BL_10 −10 0,81 0,88 0,92
performed from −10ºC to 60ºC and at frequencies 20 0,96 0,97 0,96
50 0,99 1,02 0,99
ranging between 0.1 rad/s and 100 rad/s. The ratios
of the complex modulus G* and the phase angle
δ of the modified binders with respect to BR are
given in Table 4 and Table 5. The mastercurves of
the binders are given in Figure 6.
From Tables 4 and 5 can be concluded that a 5%
kraft lignin addition to bitumen 70/100 does not
affect the general behavior of the bitumen binder.
One exception to this observation is the complex
modulus of BL_5 at 50ºC. At this temperature and
across the entire frequency spectrum the complex
modulus of BL_5 is 20% higher than that of BR.
The phase angle, interestingly enough, does not
show the same trend. REDUCEDFREQUENC Y [Hz]

A 10% kraft lignin addition however results in a • IG"I BR • IG"I BL_5 •IG"I BL_10 O PHASE ANGLEBR O PHASE ANGLEBL_5•PHASE ANGLEBL_10

more pronounced stiffening of the base bitumen.

For all temperature-frequency combinations the Figure 6. Mastercurves of G* and δ of (modified)
complex modulus of BL_10 is higher than that of binders.
BR. This difference ranges from 10% to a maxi-
mum of 75% at 50ºC and 0.01 Hz.
After a short term ageing procedure three tests
were performed on fresh and aged material; FTIR
4.3 Stage III—Mortar testing
on extracted binder, DSR frequency sweeps (on
In stage III of the experimental program tests were mortar columns) and Direct Tensile Test (DTT).
performed on mortar, a mixture of bitumen, filler From the FTIR absorption spectra the peak
and fine sand (<500 μm). The main focus during areas at wave numbers associated with oxidation
this stage was to identify whether a kraft lignin products C = O, ketone, and S = O, sulfoxide, are
modification results in reduced ageing of bitumi- evaluated. The peak areas at 1700 cm−1, C = O, and
nous mortar. And if so, to what extent this higher at 1030 cm−1, S = O, of the fresh reference mortar
ageing resistance benefits the mechanical proper- and all aged mortars along with the ICO and ISO
ties of the mortar. oxidation indices are given in Table 6.

391
Table 6. Peak areas of S = O and C = O and oxidation aged mortars are shown respectively in Figure 8
indices ISO and ICO. and Figure 9. The behavior of the fresh mortars
resembles that of the (modified) bitumen binders.
S=O C=O ICO ISO After ageing on the other hand more distinct dis-
similarities are observed between the behavior of
MR 0.2453 0 0 0.0528
MRa 0.3583 0.0365 0.0076 0.0747
the reference mortar MRa and the two kraft lignin
ML_5a 0.4019 0.0291 0.0059 0.0818
modified mortars. Table 7 contains the ratios of
ML_10a 0.3124 0.0159 0.0035 0.0679 the measured properties (G* and δ) of ML_5a and
ML_10a with respect to MRa.
It appears that the ratios of G* with respect to
the reference mortar generally have decreased after
0.04 O.Q1
ageing. And those of the phase angle have increased.
~ These observations are especially valid for measure-
0.035
0.008
0.03 ments done at higher temperatures. It appears that
0.025

0.02
~
-
0.006

~
ageing hardly has any influence on the modified
mortars compared to MR. The smallest reductions
0.015 - F
0.004 are found at low temperatures. Ratios of ML_5a and
- - - ML_10a even show some increase. ML_10 however
0.002
0.005 - - - seems to have aged less than ML_5. From these
results it follows that kraft lignin has contributed
MR MRa Ml_5a "' ML_10a
to a higher resistance against ageing, provided that
t'lPeak area o iCO
the concentration of kraft lignin is large enough.
Important also is that kraft lignin only seems to be
Figure 7. C = O peak areas and ICO values.
effective at moderate to high temperatures.
To verify the findings described above the
The ICO and ISO values indicate that the ratios of the measured properties for the aged
amount of ketones and sulfoxides has increased
for the aged mortars compared to MR. However, I>IR, ML_S&ML_lO

the ICO oxidation index of the aged reference

mortar is consistently higher than the ICO values
of the modified aged mortars. This would imply
that the addition of lignin contributes to a reduc-
tion of bitumen binder ageing. The ISO values of ~ 1.E>06 t--+ttHt- t+h11---t*
the mortars however do not follow the same trend ~

as described for the ICO values.

The spectrum of the filler Wigro 60 K contains a
large peak in the same region as S = O, vibrations of
the CH in-plane aromatic groups. Because the bind-
ers were extracted from mortar the calculation of the
sulfoxide index is therefore contaminated. A clear • fG1 MR •fG1 UL_5
REDUCED FREQUENCY [Hz]

•IG"]ML_10 • f>HASEIINGLEUR • PHASEANGLEML...5 o F'HASEANGLE ML_10

conclusion regarding the degree of ageing based on

the increase of the S = O peak can therefore not be Figure 8. Mastercurves of all fresh mortars.
made. Because of this reason the C = O peak area
increase and the ICO oxidation index are used to
assess the ageing behavior of the modified bitumen.

~~ ll ~
I I
In Figure 7 the peak area and calculated ICO are
plotted for each mortar. The decrease of the C = O I
"'""
I
I~
peak area after ageing for ML_5a and ML_10a
compared to MRa is clear. The ICO values follow
the same trend. It appears that a higher amount of
I 'I f ···~~
I

added kraft lignin will result in less formation of ~

ketones. However, caution should be taken with con-
clusions drawn from the FTIR results. The different
II
II
·+- ++ ~ ~
1
factors influencing the peak areas of the C = O and
1111

I II Ill
S = O peaks should be carefully mapped in order to
1111

confirm the anti-ageing properties of kraft lignin. • JG"I MRa •IG"IML...~ •fG"] ML_10. • F'HASE~LEMR~ oPHASE A,_NGLEUL_!I.> •PHASEA,NGLEML_10.

The mastercurves constructed from the fre-

quency sweep data obtained on all fresh and Figure 9. Mastercurves of all aged mortars.

392
Table 7. Ratios of G* and δ of aged modified mortars Table 8. Ratios of G* of aged mortars with respect to
with respect to aged reference mortar. their fresh equivalents.

Frequency Frequency
[Hz] [Hz]
G*AGED/ Temperature G*AGED/ Temperature
G*(MRa) [ºC] 0.01 1 10 G*FRESH [ºC] 0.01 1 10

ML_5a −10 0,81 0,96 0,98 MR −10 1,08 0,91 0,87

20 0,48 0,63 0,75 20 2,67 1,91 1,54
50 0,24 0,15 0,15 50 3,87 5,24 4,75
ML_10a −10 0,96 1,07 1,09 ML_5 −10 1,44 1,20 1,15
20 0,53 0,72 0,86 20 1,95 1,54 1,36
50 0,53 0,42 0,44 50 0,99 0,88 0,78
ML_10 −10 1,00 1,00 1,00
Frequency
[Hz] 20 1,00 1,00 1,00
δAGED/ Temperature 50 1,00 1,00 1,00
δ(MRa) [ºC] 0.01 1 10
ML_5a −10 1,07 1,14 1,08 Table 9. Ratios of δ of aged mortars with respect to
20 1,06 1,10 1,13 their fresh equivalents.
50 0,94 0,99 1,10
Frequency
ML_10a −10 1,29 1,11 1,12 [Hz]
20 1,09 1,13 1,14 δAGED / Temperature
50 0,97 1,02 1,07 δFRESH [ºC] 0.01
MR −10 0,78 MR −10
mortars with regards to their own fresh equiva- 20 0,91 20
lents are given in Table 8 and Table 9. This data 50 1,06 50
shows that MR has aged significantly more than ML_5 −10 0,73 ML_5 −10
the modified mortars and that ML_10 has aged 20 0,92 20
less than ML_5. A distinct difference is observed 50 0,98 50
in the ageing behavior of both mortars though; ML_10 −10 1,00 ML_10 −10
A 10% kraft lignin modification appears to effec- 20 1,00 20
tuate a rather constant change in rheological 50 1,00 50
properties across the entire considered tempera-
ture and frequency regions. On the contrary, a
5% lignin addition does not instigate a constant 1.00 mm/s. Results at −5.1ºC and 0.51 mm/s for the
change in properties. Actually, the data shows fresh and aged mortars are given in Figure 10 and
that at low temperatures the behavior of ML_5 is Figure 11.
adversely affected. Whereas at high temperatures From all DTT tests it is concluded that a kraft
its response can be considered expedient com- lignin modification of 10% can contribute, albeit
pared to that of ML_10. This could mean that minimally, to more favorable behavior in the low
a too low amounts of kraft lignin can adversely temperature region of fresh bituminous mortars.
affect the ageing resistance of bituminous mor- Compared to MR, ML_5 has not improved in duc-
tars at low temperatures. tility and in strength.
A critical note regarding the method of ageing For the aged lignin modified mortars different
should be kept in mind considering these results. behavior is observed. Force-displacement curves
The ageing consists of relatively short term, ther- for ML_5a imply that the addition of 5% kraft
mal ageing performed protocol at temperatures lignin does not contribute to more favorable mor-
far above the service temperature of pavements. tar behavior after ageing. In fact, the response of
The effects of moisture and UV radiation are not ML_5a and ML_5 AT −5.1ºC & 0.51 mm/s are the
taken into account. Expected is that with this age- same. This confirms the observations from the
ing protocol only short term ageing as a result of frequency sweep results of the aged mortars. This
mixing is simulated. observation corresponds with the finding from the
Direct tensile tests were performed on the frequency sweep results that kraft lignin is more
fresh and aged mortars at −5.1ºC with displace- effective at moderate to high temperatures (30°C
ment speeds of 0.02 mm/s and 0.51 mm/s and at to 170°C) and will not significantly improve low
−10ºC with displacement speeds of 0.10 mm/s and temperature properties.

393
 not be recorded. After turning the beakers upside
l down the percentages of residue left in the glass
beakers presented a distinct trend. As the per-
z t-1k-Cc-- - - - - - - - - - - - - - - - - - - -
centage of added lignin increases the amount of
ll
-0.15

~
residue decreases. Kraft lignin can thus be consid-

l: ered to improve the drainage resistance of PAC+

i

~ ~· I '
mixtures.
-0.05 :

If I

oL~--~---~---~--~---~--~
0 5 DISCUSSION OF RESULTS
DISPLACEMENT [MM]

--- ML_5 -5.1C&0.51MM/S - ML_ 10 -5.1C&0.51MM/S- ·- MR -5.1C&0.51MM/S

Results from tests performed in this research indi-
Figure 10. Direct tensile test results at −5.1ºC at cate that a kraft lignin addition to bituminous mix-
0.51 mm/s for fresh mortars. tures gives beneficiary results at moderate to high
temperatures (30°C to 170°C). The magnitude of
this beneficiary effect is very much dependent on the
added amount of kraft lignin. In addition, it appears
that a suboptimal amount of kraft lignin modifica-
tion can adversely affect low temperature properties.
i~ . .,.r
;[ ~2 t------+7-
;; - - - - - - - - - /o¥'~---+----
= Measured rheological properties show that kraft
lignin has a positive effect on the binder response
~ -0.15 t--'-i-;-:--------=el-/"--------h-"'·_':__r-_··-:-'i_ _ __
~ / ,·' i
at high in-service temperatures. Specifically within
this temperature range a higher kraft lignin con-
/ centration seems to benefit the properties of the
bituminous mixtures. At low temperatures a kraft
lignin modification of fresh bituminous mix-
DISPLACEMENT [MM]
tures can improve ductility but does not improve
-- - ML_ 5a -5.1 C&0.51 MM/S - ML_ 1 Oa -5.1 C&0.51 MM/S - ·- MRa -5.1 C&0.51 MM/S
strength. After ageing the contribution of kraft
lignin is noticeably decreased. Even at low tem-
Figure 11. Direct tensile test results at −5.1ºC at
0.51 mm/s for aged mortars. peratures though, a dependence on kraft lignin
concentration is found.
The kraft pulping process has a lot of influence
Table 10. Results Schellenberg test on PA mixtures.
on the molecular weight of the kraft lignin mol-
W1 W2 T W3 Residue D ecules (Li, 2011). In particular it instigates a weight
gain (Lin & Dence, 1992). It is interesting to see
# [g] [g] [°C] [g] [%] that at very high temperatures like 180°C viscos-
ity measurements give lower results than at 160°C.
PAC_R 1 165.0 A kraft lignin addition can thus result in lower vis-
2 214.2 1197.7 165.0 222.4 0.83 cosity values. This could be the result of the lack of
3 215.3 1223.6 165.0 223.2 0.78 lubrication between kraft lignin matrix and the bitu-
PAC_L_5 1 168.7 men. It could also indicate possible lignin damage.
2 215.1 1207.5 168.7 221.2 0.61 In addition to this it is noted that the modified
3 237.2 1241.7 168.7 244.6 0.74 mortars and PA mixtures effectively contained less
PAC_L_10 1 169.8 pure bitumen than the reference mixtures. Also,
2 237.2 1237.1 169.8 242.3 0.51 the addition of filler increased the specific area tre-
3 224.1 1228.7 169.8 230.9 0.68 mendously affecting the available amount of bitu-
men for lubrication of the lignin polymer.

4.4 Stage IV—PA testing

6 CONCLUSIONS
To determine the impact of a kraft lignin addition
on the drainage behavior of the bituminous mor-
The purpose of this research was to find out if
tar the Schellenberg test was performed at 165ºC.
kraft lignin modified bitumen could potentially
To this end PA mixtures were produced with the
improve the raveling resistance of PA.
reference and modified mortars.
For each PA mixture three similar batches − The mortar testing strongly indicates that kraft
were mixed, one of which served as a ‘calibra- lignin can counteract the effects of ageing on the
tion’ batch. Of this calibration batch results need rheological properties of bituminous mixtures.

394
− For kraft lignin modified mortars the degree of Hagos, E., 2008, The effect of ageing on binder proper-
anti-ageing appears to be highly dependent on ties of Porous asphalt concrete, Delft University of
the in-service temperature and the kraft lignin to Technology, The Netherlands.
Kaplan, D., 1998, Biopolymers from renewable resources,
bitumen ratio.
United States of America, Medford, Springer-Verlach.
− The addition of kraft lignin has no negative Khedoe, R. & Moraal, J., 2007, Lifetime Optimization
effects for the workability of mixtures at the Tool; Sample preparation and laboratory testing for
high mixing temperatures. the LOT research program, Delft University of Tech-
− Viscosity measurements at high production tem- nology, The Netherlands.
peratures show that the lignin modified binders Li, J., 2011, Isolation of lignin from wood, Saimaa Uni-
are of the shear-thinning type and they show versity. of Applied Sciences, Finland.
that at mixing shear rates the viscosity is simi- Lin, S. & Dence, C., 1992, Methods in lignin chemistry,
lar to that of unmodified bitumen. At low strain Springer.
Poeran, N., 2014, The effects of a kraft lignin and wood
rates the viscosity strongly increases.
addition on the raveling resistance of porous asphalt
− Accordingly, the Schellenberg test showed that concrete, MSc thesis, Delft University of Technology,
kraft lignin has a very positive influence the drain- The Netherlands.
age resistance of PA. Here as well, a strong positive Pouteau, C., 2003, Antioxidant properties of lignin in
dependence on kraft lignin concentration is found. polypropylene, Polymer Degradation and Stability
− The behavior at low temperatures needs more vol. 81 Issue 1 p.9–18, 2003.
investigation. Preferably this research is per- Sundstrom, D. et al, 1983, Use of by product lignins
formed on lignin modified PA mixtures in the as extenders in asphalt, University of Connecticut,
form of specific mechanical tests. United States of America, 1983.
Williams, C. & McCready, N., 2008, The utilization of
agriculturally derived lignin as an antioxidant in
asphalt binder, Iowa State University, United States
REFERENCES of America Standard RAW Specifications 2010,
CROW, Ede.
Gosselink, R.J.A., 2011, Lignin as a renewable aromatic
resource for the chemical industry, Wageningen Uni-
versity, The Netherlands.

395
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Usage of ion-irradiated recycled polypropylene as modifier

in bituminous binder

P. Ahmedzade, T. Günay, S. Altun & B. Kultayev

Ege University, Izmir, Turkey

A. Fainleib, O. Grigoryeva & O. Starostenko

National Academy of Sciences of Ukraine (NASU), Kyiv, Ukraine

ABSTRACT: The increase in the number of trucks and other heavy vehicles in highways lead to crack-
ing and deteriorations in flexible highway. Polymer additives such as Styrene-Butadiene-Styrene (SBS)
and Ethylene Vinyl Acetate (EVA) are used in the bituminous binder satisfactorily to extend service life
of pavements. However, polymer additives increase considerably construction cost of flexible pavement.
Hence, using recycled polymers as a modifier in bitumen has been gain traction in recent years consider-
ing environmental protection, lower energy consumption, and their affordable cost. The aim of this paper
is to study the usage of recycled PP (PPR) as modifier on bitumen However, normally there is no chemical
reaction between recycled polymers and bitumen. To this end, surface of the PPR was exposed to ion Irra-
diation (I-PPR) which creates double bonds (>C = C<) on recycled polymer that can interact with bitumen.
Five samples having different I-PPR modifier amounts of 1%, 3%, 5%, 7% and 9% total weight of mixture.
The effects of I-PPR on the bitumen were investigated by means of chemical and physical testing program;
including Fourier Transform Infrared spectroscopy (FT-IR) spectroscopy, conventional tests, Rotational
Viscosity (RV) and Dynamic Shear Rheometer (DSR) test. Tests results reveals that the using waste poly-
mers in bitumen by means of ion irradiation process, since increased I-PPR modifier content in bitumen
leads to an increased rutting parameter which means permanent deformation in flexible pavement can be
restrained by I-PPR modification.

1 INTRODUCTION interaction, ion beam irradiation was applied to PPR

so as to create double bonds (>C = C<) on the poly-
Polymer modification of bitumen is the most conven- mers that can interact with bitumen. After this proc-
tional method in flexible pavement in order to reduce ess, ion Irradiated PPR (I-PPR) modifier was added
deteriorations such as rutting. Styrene Butadiene to bitumen having a penetration of 160/200 with
Styrene (SBS), Ethylene Vinyl Acetate (EVA) are the amounts of 1%, 3%, 5%, 7% and 9% by total weight
polymers mainly employed to enhance temperature of mixture. The effects of I-PPR on the bitumen
susceptibility of bitumen by increasing stiffness at were investigated by means of chemical and physical
high temperature, decreasing stiffness at low tem- testing program. Fourier Transform Infrared spec-
perature (Read & Whiteoak 2003, Airey 2002, 2003). troscopy (FT-IR) work was done on PPR polymers
In addition to SBS and EVA, various polymers have before and after ion beam irradiation to analyze the
been applied as modifier in bitumen, namely, chemical changes. Conventional tests such as pene-
Polyolefins; Low Density Polyethylene (LDPE), tration, softening point and ductility were conducted
High Density Polyethylene (HDPE) and Polypro- on base and I-PPR modified binders. Rotational Vis-
Pylene (PP) (Giavarini et al., 1996, González et al., cosity (RV) and Dynamic Shear Rheometer (DSR)
2005, Hınıslıoğlu & Ağar 2004). test were employed to investigate rheological behav-
Recently, employing recycled polymers so as to ior of the bitumen after I-PPR modification.
reduce both the construction cost of the pavements
and amount of waste has been gained traction for
2 EXPERIMENTAL INVESTIGATION
researchers (Morales et al., 2004, Lei et al., 2007,
Naskar et al., 2010). In this context, in this study, it
2.1 Materials
was aimed to investigate the effect of recycled Poly-
propylene (PPR) as modifier in bitumen. Neverthe- The bitumen having 160/220 penetration grade
less, recycled polymers have no chemical interaction was used throughout the testing program. Physi-
with bitumen. In order to provide a polymer/bitumen cal properties of the bitumen were given in Table 1.

397
Table 1. Physical properties of the base bitumen. Base bitumen – “B”;
Base bitumen + 1% I-PPR – “B-1-I-PPR”;
Properties Standard B160/220 Base bitumen + 3% I-PPR – “B-3-I-PPR”;
Base bitumen + 5% I-PPR – “B-5-I-PPR”;
Penetration ASTM D5 195.5
Base bitumen + 7% I-PPR – “B-7-I-PPR”;
(25°C; 0.1 mm)
Softening point (°C) ASTM D36 38.7
Base bitumen + 9% I-PPR – “B-9-I-PPR”;
Ductility (25°C; cm) ASTM D113 103
Penetration Index (PI) – −0.73
Specific gravity ASTM D70 1.03 3 TESTING PROGRAM
(25°C; gr/cm3)
3.1 Fourier transform infrared spectroscopy
Fourier Transform Infrared (FTIR) spectroscopy
PPR polymer was obtained from the company of was used to characterize various functional groups
Roksana, Ukraine where waste polypropylene in the PPR polymer and to detect the chemical
sacks were recycled and produced as pallet shaped changes before and after ion beam irradiation.
recycled PP. FTIR spectra were recorded with a Bruker Ten-
PPR employed in this study has a melting point sor 27 DTGS spectrometer between 4000 and
of 165°C, a flow rate of 4.0–6.0 g/10 min, a tensile 450 cm−1 using the Attenuated Total Reflection
strength of 365 kg/cm2 and fisheye defects of 6 in (ATR) mode. For each spectrum, 32 consecutive
every gram. scans with a resolution of 4 cm−1 were averaged.

2.2 Ion beam irradiation process 3.2 Conventional tests

Ion irradiation is a process that polymers are In order to investigate the effects of the modifier
exposed to charged particle beam consists of ions on physical properties of bitumen, conventional
and promising method to enhance the chemical tests such as, penetration (ASTM D5), softening
structure and the physical properties of polymers point (ASTM D36) and ductility (ASTM D 113)
(Pignataro et al., 1997). were conducted accordance with ASTM standards
It has many uses in various industries and can to base and PPR modified bitumens. Penetration
be used for sputtering or etching and for ion beam index was also calculated in order to investigate the
analysis. This method enables a certain change in temperature susceptibility of the binders.
chemical structure that leads to create stronger bonds
between polymer and bitumen by emerging new C-C
3.3 Rotational viscosity test
bonds. In this research, the air ion irradiation process
was done at 0.5 m3 area surrounding by protection A Brookfield viscometer (DVRV-II Pro Extra) was
walls at 25°C, in normal air pressure (760 mm). PPR employed in this study for the viscosity tests of
particles exposed to ion with a current density rang- the base and the I-PPR modified bitumens. Rota-
ing from 3 × 107 to 8 × 107/cm3 for ten hours by means tional viscosity was measured at 135°C and 165°C,
of an automatic controlled system. respectively according to ASTM standard. After
the measurements, ideal mixing and compaction
temperature ranges of the binders in flexible pave-
2.3 Preparation of samples
ment application were calculated by means of vis-
Before the modification process, I-PPR polymers cosity test results.
were first milled. After, the crumbled polymers were
sieved with a sieve of 0.6 mm to obtain homog-
3.4 Frequency sweep test
enous particles. 160/220 penetration grade bitu-
men was heated for 90 minutes at 170°C, and then The Dynamic Shear Rheometer (DSR), was used
poured into the flask of the high shear mixer (Sil- to determine the viscoelastic behavior of modified
verson LSM) adjusted to 1000 rpm. Subsequently, bitumen binders especially at intermediate to high
I-PPR was added to bitumen at intervals at same service temperatures (Kennedy et al., 1994). The
mixing rate. After the addition of the polymer, the frequency sweep test was applied by means of a
mixing rate was increased to 3200 rpm and mixing Anton Paar Physica SmartPave Plus dynamic shear
was continued for 45 minutes. End of the mixing rheometer to base and modified bitumen to analyze
process, the samples were removed from the flask the changes of the fundamental rheological param-
and divided into small containers, covered with eters of the binders at different temperatures and
aluminum foil and stored for various testing. The frequencies. Several temperatures ranging from 25
different binders were coded as follows: to 75°C and three different frequencies of 0.1, 0.72

398
and, 1.94 Hz were selected to reflect the free flow, in Table 2. As can be seen, there is a consistent
decreasing free flow and heavy traffic. decrease in penetration and increase in soften-
ing point with increasing I-PPR modifier content.
Accordingly, B-9-I-PPR binder has the lowest pen-
4 RESULTS AND DISCUSSIONS etration and highest softening point values whilst
it was vice versa for base bitumen. Penetration
4.1 Fourier transform infrared spectroscopy and softening point results show that I-PPR modi-
results fier provides increased stiffness to bitumen as each
In Figure 1, the normalized FT-IR spectra (the conventional test results confirm each other.
band at ν ≈ 715 cm−1 has been used as internal As can be seen from Table 2, there is a slightly
standard) of initial (non-irradiated) PPR and decrease in ductility after I-PPR. Thus, it can be
I-PPR (before and after ion irradiation) were pre- said that I-PPR has a slightly adverse effect on elon-
sented. As can be seen from the Figure 1, there is gation ability of binder. Higher values of PI sig-
no significant difference in position and intensity nifies that lower temperature susceptibility which
of the main absorption peaks at ν ≈ 721, 1165, means I-PPR modified binder could be used a
1302, 1377, 1459 cm−1, characteristic for vibrations larger temperature range. (Read & Whiteoak 2003,
of C-C and C-H bonds and at 1740 cm−1 of C = O ASTM D5-97 1998).
group in the spectra of PPR and I-PPR.
However, changes in peaks were observed at
ν ≈ 1022, 1099 and 1647 cm−1 in the spectrum of 4.3 Rotational Viscosity (RV) test results
I-PPR. The decreasing intensity at 1099 cm−1 after The rotational viscosity of base and I-PPR binders
ion irradiation evidences of formation of unsatu- were given in and Table 3. In Table 3, modification
rated bonds in macromolecular chains in I-PPR indexes and ideal compaction and mixing tempera-
which occurred as a result of high energy action tures of the binders were also included.
of the ion beam on PPR polymer. In the scheme As can be seen, the viscosity of the binders
free radicals appear at breakage of polymer chains, tends to increase with increasing I-PPR content at
which can then transform to different forms both temperatures of 135°C and 165°C. Accord-
including formation of double bonds (>C = C<). ingly, B-9-I-PPR has 9 times higher viscosity than
Simultaneously, some changes in the intensities of that of the base bitumen. This remarkable incre-
the existent bands appear. The unsaturated bonds ment in viscosity is a clear evidence of stiffening
and new appeared C = O groups formed in I-PPR effect of the modifier (Wen et al., 2002). This result
are reactive to the unsaturated bonds of binder also confirms the findings in conventional tests.
components that could provide better mixing and Modification index (η) of the binders signifies that
more effective reinforcement of the bitumen by the increment in viscosity is very consistent at both
means of ion irradiation.
Table 2. Conventional test results.
4.2 Conventional test results
Binder type
Conventional test results were conducted on
base and the I-PPR modified bitumen were given B-1-I- B-5-I- B-9-I-
Properties Base PPR PPR PPR

1459
1458
1377 Penetration (0.1 mm) 195.5 151 115 40
Softening point (°C) 38.7 38.5 52.7 110.0
Ductility (cm) at 25°C 103 97 92 86
Penetration Index (PI ) −0.73 −1.87 1.90 7.00

1165
Table 3. Rotational viscosity test results.

Binder RV (cP) RV (cP) Mixing Compacting

types @ 135°C @ 165°C Temp. °C Temp. °C

B 202.5 65.5 135–142 122–128

B-1-I-PPR 280 97.5 144–149 131–137
1800 1600 1400 1200 1000 800 B-3-I-PPR 440 145 152–157 142–147
Wave Number, em·'
B-5-I-PPR 677.5 222.5 159–154 143–148
B-7-I-PPR 1117.5 362.5 152–158 140–146
Figure 1. Normalized FT-IR spectra of PPR and I-PPR B-9-I-PPR 1857.5 602.5 159–163 151–154
polymers.

399
temperatures of 135°C and 165°C. Similar modifi- I,EiOJ

cation indexes at different temperatures could be a I,Ei06

sign of reduced temperature susceptibility. I,EiOl

I,Ei04
Calculated mixing and compaction tempera-
tures for the base and the modified binders tend
to increase with increasing I-PPR modifier con-
tent. Excessive increment in mixing and compac-
so
tion temperatures means that the workability of lO 30 40
Tempenture('C)
lO 60 70

bitumen considerably decreases. However, as can

be seen from Table 3, there is no mixing tempera- Figure 2. Curves of complex modulus of base and
ture exceeds 165–170°C which normally used in I-PPR Modified binders at 0.1 Hz.
applications.
1,E+07

4.4 Frequency sweep test results ~ 1, E+ 06

(9 1,E+05

Isochronal plots of log-reduced complex modulus :g~ 1,E+04

(G*) of the original binders versus temperature ~ 1,E+03

iii 1,E+02
at frequencies of 0.1 Hz, 0.72 Hz, and 1.36 Hz, ~
~1 , E+0 1

which equivalent to low, medium and high traf- 0 1,E+00 +---~-~-~--~-~--~-~-~

0 40
fic speeds respectively were given in Figures 2–5. Temperature (0 C)

At the middle temperature range of 10–25°C

all binders have high values of G*, which signi- Figure 3. Curves of complex modulus of base and
I-PPR Modified binders at 0.72 Hz.
fies that the binders are resistant against shear
stress at that temperature range. However, as
temperature increases, three different rheological 1,E+OB
behaviors of bitumens were detected. The first 1,E+07

behavior was belong to the G* of B and B-1- 1.E+06

1,E+05
I-PPR binders which reduce progressively and 1.E+04

means that after a certain temperature (around 1,E+03

50°C), regardless of traffic speed, permanent 1,E+02

1,E+01
deformation can be occurred rapidly with these 40
binders. The second was the behavior of G* val- Temperature (OC)

ues belong to B-3-I-PPR and B-5-I-PPR binders.

G* of these binders also reduce with increasing Figure 4. Curves of complex modulus of base and
temperature, however, after a consistent decreas- I-PPR Modified binders at 1.94 Hz.
ing, G* values remain almost same and binder
provide enough resistance against shear stress
at higher temperature ranges of 60–85°C. This
result indicates that these binders can be used at
high temperature ranges but traffic speed is also
another important factor in terms of permanent
deformations. The third type of the behavior of
G* is belong to B-7-I-PPR and B-9-I-PPR which
are almost not effected at all testing temperatures
and can be used regardless of traffic speed and
temperature. Figure 5. Curves of phase angle of base and I-PPR
Phase angle vs temperature graphs were given modified binders at 0.1 Hz.
in Figure 5–7. Phase angle, the values of δ, ratio
of permanent deformation to elastic deforma-
tion, is another fundamental parameter in terms
of bitumen rheological behavior (Lu & Isacs-
son 2000). δ generally tend to decrease with
increasing I-PPR polymer content which signi-
fies that bitumen recover more after shear stress
:
20
~
-+- B-1-1-PP.
----*- B-3-I-PP"
--.- B-5-I-PP"
10 --+--B-7-I-PP•
removes. Thus, it can be said that I-PPR poly- O ~ B-9-I-PP

40
mer modifier can reduce permanent deforma- T""P"'NR ('C)

tion of flexible pavement by means of reducing

temperature susceptibility and increasing elastic Figure 6. Curves of phase angle of base and I-PPR
deformation. modified binders at 0.72 Hz.

400
 Airey, G. D. 2002. Rheological evaluation of ethy-lene
vinyl acetate polymer modified bitumens, Construc-
tion and Building Materials. 16(8), 473–487.
ASTM D113-99, Standard Test Method for Ductility
of Bituminous Materials, ASTM International, West
Conshohocken, PA, 1999, 3 pages.
ASTM D36/D36M-14e1. “Standard Test Method for
Softening Point of Bitumen (Ring-and-Ball Appara-
tus), ASTM International, West Con-shohocken, PA,
Figure 7. Curves of phase angle of base and I-PPR 2014, 3 pages.
modified binders at 1.94 Hz. ASTM D4402-06. “Standard Test Method for Vis-cosity
Determination of Asphalt at Elevated Temperatures
Using a Rotational Viscometer.” ASTM International,
West Conshohocken, PA, 2006, 3 pages.
5 CONCLUSION ASTM D5/D5M-13. “Standard Test Method for Pen-
etration of Bituminous Materials.” ASTM In-terna-
tional, West Conshohocken, PA, 2013, 4 pages.
The following conclusions can be reported based ASTM D7175-15. “Standard Test Method for De-
on the results of the study: termining the Rheological Properties of Asphalt
FT-IR findings signifies that unsaturated bonds Binder Using a Dynamic Shear Rheometer.”
in macromolecular chains in I-PPR was appeared ASTM International, West Conshohocken, PA,
after ion irradiation I-PPR and these bonds reac- 2015, 16 pages.
tive to the unsaturated bonds of binder compo- Gad, Y. H., Magida, M. M. & El-Nahas, H. H. 2010.
nents that could provide better mixing and more Effect of ionizing irradiation on the thermal blend
effective reinforcement of the binder. of waste low density polyethylene/ethylene vinyl ace-
tate/bitumen for some industrial applications. Jour-
The decrement in penetration and increase in
nal of Industrial and Engineering Chemistry. 16(6),
softening point as modifier content increase clearly 1019–1024.
indicate that I-PPR polymer has stiffening effect on Garcıía-Morales, M., Partal, P., Navarro, F. J.,
bitumen. PI index calculated by means of conven- Mar-tínez-Boza, F., Gallegos, C., González, N.,
tional tests reveal that enhanced temperature suscep- González, O. & Muñoz, M. E. 2004. Viscous pro-
tibility of the binders with I-PPR polymer. According perties and microstructure of recycled eva modified
to ductility test modifier has a slightly adverse effect bitumen. Fuel. 83(1), 31–38.
on elongation ability without corruption of binder. Giavarini, C., De Filippis, P., Santarelli, M. L. &
A significant increment was determined in vis- Scarsella, M. 1996. Production of stable polyp-ropyl-
ene-modified bitumens. Fuel. 75(6), 681–686.
cosity by means of RV test which is an evidence
González, O., Muñoz, M. E. & Santamaría, A. 2005.
of stiffening effect of the modifier. Modification Bitumen/polyethylene blends: using m-LLDPEs to
indexes signify that the increment in viscosity is improve stability and viscoelastic properties. Rheo-
very consistent at both temperatures of 135°C and logica Acta. 45(5), 603–610.
165°C which could be a sign of a reduced tempera- Hınıslıoğlu, S. & Ağar, E. 2004. Use of waste high den-
ture susceptibility effect of I-PPR modifier. sity polyethylene as bitumen modifier in asp-halt con-
The frequency sweep was conducted by means crete mix. Materials Letters. 58(3–4), 267–271.
of DSR reveals that usage of I-PPR more than 3% Lei, Y., Wu, Q., Yao, F. & Xu, Y. 2007 Preparation and
can significantly reduce permanent deformation by properties of recycled HDPE/natural fiber compos-
ites, Composites Part A: Applied Science and Manu-
decreasing temperature susceptibility and increas-
facturing. 38(7), 1664–1674.
ing elastic response of bitumen against shear stress. Lu, X. & Isacsson, U. 2000. Modification of road bi-
However, above this level of modification the binder tumens with thermoplastic polymers. Polymer testing.
could become too stiff at both low and high tem- 20(1), 77–86.
peratures zone. Thus, for optimal use, the amount of Naskar, M., Chaki, T. K. & Reddy, K. S. 2010. Effect
modifier in bitumen should not exceed to 3%. of waste plastic as modifier on thermal stability and
Based on the results, it can be noted that I-PPR degradation kinetics of bitumen/waste plastics blend.
modifier are compatible with bitumen in contrast Thermochimica Acta. 509(1–2), 128–134.
to other waste materials and, I-PPR modified bitu- Pignataro, B., Conte, E., Scandurra, A. & Marletta,
G. 1997. Improved cell adhesion to ion beam-
men are more resistant deformations such as rut-
irradiated polymer surfaces. Biomaterials. 18(22),
ting especially at high temperatures. 1461–1470.
Read, J. and Whiteoak, D. 2003. The Shell Bitumen
Handbook, Thomas Telford.
REFERENCES Wen, G., Zhang, Y., Zhang, Y., Sun, K. & Fan, Y.
2002. Rheological characterization of storage-sta-
Airey, G. 2003. Rheological properties of styrene bu- ble SBS-modified asphalts. Polymer testing. 21(3),
tadiene styrene polymer modified road bitumens. 295–302.
Fuel. 82(14), 1709–1719.

401
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Chemical and physical properties of an asphalt binder modified by the

sap of Euphorbia Tirucalli plant: Application in bituminous prime coat

L.M. Gondim
Science and Technology Center, Federal University of Cariri, Juazeiro do Norte-CE, Brazil

S.A. Soares
Department of Organic and Inorganic Chemistry, Federal University of Ceará, Fortaleza-CE, Brazil

S.H.A. Barroso & C.M.C. Alecrin

Departament of Transportation Engineering, Federal University of Ceará, Fortaleza-CE, Brazil

ABSTRACT: The present work aims to produce a bio-binder from the modification of a petroleum
based asphalt by the sap of Euphorbia Tirucalli, investigating the physical and rheological properties
of the modified asphalts and testing its application on prime coat services. This plant is also known as
petroleum plant due to the similarity of hydrocarbons presented in these materials. The contents of 3%,
5% and 10% of sap (by weight) were added to the asphalt binder. The mixtures were subjected to chemical
characterization tests (FTIR and Termogravimetric Analysis) and physical characterization test (penetra-
tion, softening point, Brookfield viscosity, and DSR), before and after short-term aging, through RTFOT.
The asphalt modified by 10% of sap was used to produce 2 cutbacks: one diluted on kerosene and another
diluted on d-limonene. Both cutbacks were tested for application on prime coat services. Although the
results of FTIR have shown an increase in the carbonyl and sulfoxyde indexes, the results of physical char-
acterization showed that the addition of the sap turned the binder less susceptible to hardening effects of
aging, since lower viscosity values (at high temperatures) and phase angles (at high in-service tempera-
tures) were observed. The alternative cutbacks showed potential for application at prime coats.

1 INTRODUCTION emphasize that it is a typical vegetable of desert

environments and that they are naturally adapted
World’s crude oil reserves are becoming depleted, to the salinity and dry conditions.
while its costs are usually high and its markets are This plant was studied by the chemist Melvin
unpredictable. A worldwide increase of environ- Calvin in the 70’s, when he named it “Petroleum
mental awareness led to a search for sustainable Plant”, as he found in its latex the presence of
practices in asphalt industry, such as the use of hydrocarbons of molecular weight similar to those
renewable sources materials. found in petroleum products such as gasoline
In this context, many researches are being con- (Calvin, 1980). Calvin (1980) further commented
ducted all over the world, in order to achieve alter- that this hydrocarbon rich material resembles
native materials capable of replacing, totally or petroleum and may even be processed to produce
partially, the petroleum based asphalt binders. similar fractions to all commercial petrochemical
Many materials are been experienced, as bio- products.
mass from different sources (Mohamed Metwally, The presence of these hydrocarbons allied to the
2010; Ecopave, 2011; Peralta et al, 2013; Çelik & adhesive and viscous properties of this sap led to
Atasağun, 2012); microalgae lipids (Aldo et al., believe that this material has potential for generate
2012); acidulated soybean soapstock (Seidel & a bio-binder that could be applied in the paving
Haddock, 2012), vegetal oils (Leite et al., 2012), industry.
wasted cooking oil (Wen, Bhusal & Wen, 2012) Regarding productivity, in Calvin’s (1980)
among others. experiment without selected seeds or any agricul-
The Euphorbia Tirucalli is a plant native from tural improvement, only 1-year-old plants weigh-
Madagascar and very commonly found among ing about 2 kg were able to produce 15 barrels
the Brazilian Flora, that presents a sap with vis- of oil per acre (1 acre = 4046, 86 m2) in one year.
cous and adhesive properties. Eshel et al. (2010) This author also commented on an experiment in

403
Okinawa, Japan, where Euphorbia Tirucalli’s pro-
duction was estimated to be between 10 and 20
barrels of oil per acre in a year.
The formulation of a green binder from a plant
so well adapted to semi-arid region would allow the
commercial exploitation of this product, develop-
ing productive arrangements in arid and semi-arid
climates, thus promoting sustainable develop-
ment in poor regions. Thus, the development of
a bio-binder based on Euphorbia Tirucalli sap
has relevance for subsidizing the development of
technologies that can reduce costs in road projects,
as well as promoting sustainable development in
unproductive areas.
This research work aims to produce a bio-
binder from the modification of a petroleum based Coagulation
container
asphalt by 3%, 5% and 10% of sap of Euphorbia
Tirucalli, investigating the physical and rheological Figure 2. Extraction process of the sap.
properties of the modified asphalts and testing its
application on prime coat services.
scratcher with a short thin point, in order to not
injure the tree;
2. Cutting the trunk of the tree and opening of the
2 MATERIALS AND TESTING METHODS
channel using the V-shaped knife, taking care
not to deepen the channel excessively;
2.1 The sap of Euphorbia Tirucalli
3. Positioning the spout and immediate position-
ing of the container;
2.1.1 Sap extraction 4. Observing the bleeding, until there are signs of
The Euphorbia Tirucalli sap was extracted fol- coagulation of the sap in the channel;
lowing the procedures used for the extraction of 5. Collecting of material at the end of coagula-
the rubber tree, as described by Teixeira (2013). tion or the appearance of more visible signs of
For that, were employed a marking template, a contamination.
scratcher, a v-shaped knife, a spout and a con-
tainer, presented on Figure 1.
2.1.2 Sap preparation
The procedure consisted on the following steps,
The sap in natura presented very fluid aspect, due
illustrated at Figure 2:
to the large amount of water in its composition,
1. Marking the line where the cut should be made, which makes it incompatible with the asphalt
using the marking template as a reference of binder modification. Thus, it was necessary to
minimum height and cutting angulations, and a dehydrate the sap before its application.
The sap showed a behavior similar to an aque-
ous emulsion when exposed to air: it remained
stable for a few days, but soon began to dehydrate
naturally by evaporation. During the dehydration
process, the material resembles a gum, with the
formation of an elastic film on the surface (see
Figure 3). The formation of this film indicates
the possibility of an existing natural polymer on
the sap constitution. With the extended exposure
time, the sap became stiffer and brittle, taking the
form of lumps, leaving a minimum of moisture
(Figure 3).
The sap was oven dried at 60°C, during 60 hours.
After this period, the lumps became united, as if
coalescence had occurred. The material was then
fragmented by mechanical action.
It has also been observed that while the sap is
not dehydrated, it can be diluted in water without
Figure 1. Instruments for the sap extraction. separation of phases. However, once the sample is

404
 • Softening Point (ASTM D36);
• Rotational Viscosity in Brookfield® viscometer,
(ASTM D 4402), using spindle 21, at 135ºC,
150ºC e 177ºC and shear rate of 20 rpm.
• Frequency Sweep Test in a Dynamic Shear Rhe-
Gum aspect Lumps aspect ometer (DSR), under stress-controlled mode
(120 Pa), over a range of frequencies (0.01–
Figure 3. Sap aspects during dehydration process. 100Hz) (ASTM D7175). The samples were
tested at high temperatures only (45ºC –85ºC).
Afterwards, the asphalt modified by the biggest
dried, it is no longer possible to dissolve it in water, content of sap (10%) was used to produce 2 types
only in organic solvents. of cutbacks: one was diluted in kerosene (denoted
as CAK) and another was diluted in D-limonene
(denoted as CADL).
2.2 Materials
The composition of these cutbacks was 52% of
The materials applied in this research were: modified asphalt and 48% of solvent, in weight.
They were produced on a low shear mixer reactor,
• Asphalt binder classified as a PG 64–28, pen-
at 70°C, 500 rpm, for 15 minutes. The densities of
etration grade 50/70;
the cutbacks were determined following the rec-
• The sap of the Euphorbia Tirucalli prepared as
ommendations of ASTM D70-09.
described on the item 2.1.2;
To evaluate the potential of these materials for
• Kerosene, as an organic solvent;
application in prime coat services, the methodol-
• D-limonene, a biodegradable organic solvent
ogy of the capsule developed by Almeida (2013)
obtained from the waste of the citrus processing
was employed. This test methodology consists on
industry;
measuring the penetration of an applied prime coat
• An A-2-4 soil sample (AASHTO
in a soil sample, previously compacted into small
Classification).
capsules (43 mm of diameter, and 22 mm in height).
The asphalt binder was modified by the addi- According to this procedure, the soil is com-
tion of 3%, 5% and 10% of the dehydrated sap, on pacted onto the capsules using a cohesimeter,
weight, on a low shear mixing reactor at 160ºC, led to rest in order to lose 50% of its humidity,
1500 rpm, for one hour. is pre-hydrated, receive the prime coat, cures for
The asphalt binder and the modified samples 24 hours, and then is divided in half to permit the
were submitted to the hardening and oxidative measurement of the prime coat penetration on
short term aging on rolling thin film oven (RTFO) the soil sample. The whole process is described in
as described on ASTM D2872. Almeida (2013), and is illustrated at Figure 4.

2.3 Testing methods

l~ . .
.. -~
In order to characterize the sap of the Euphorbia
Tirucalli were performed:
c.,
~
• Fourier Transform Infrared Spectroscopy
(FTIR), absorbance spectrum, in the region of
,:~
4000–400 cm−1 wave numbers, in transmission
module, using K-Br pellets;
• Thermo Gravimetric Analysis (TGA), on a
temperature range from 20°C to 700°C, with a
heating rate of 10ºC/min, on both inert (N2) and
oxidative (O2) atmospheres;
The neat and modified asphalt binder were sub-
mitted to the following tests
• Fourier Transform Infrared Spectroscopy
(FTIR), absorbance spectrum, in the region of
4000–400 cm−1 wave numbers, in transmission Sample divided in half Penetration measuments
module, using liquid cell and solutions on Car-
bon Disulfide; Figure 4. Methodology of the Capsule for the measu-
• Penetration at 25ºC (ASTM D5); ment of prime coat penetration.

405
 It was also carried out a test to measure de abra-
sion caused by the cohesimeter on the surface of the
primed samples. For that, it was measured the torque
necessary to make the cohesimeter spin 90° over the
sample.

3 RESULTS AND DISCUSSION

3.1 Euphorbia Tirucalli’s sap characteristics

The FTIR spectrum shown in Figure 5 is regarded
to sample of the Euphorbia Tirucalli sap.
Temperature(•C)
The main functional groups identified were
Hydroxyl groups (OH) from water and different Figure 6. Thermogravimetric curve of the sap in inert
natures; CH3 and CH2 of aliphatic chains; satu- (N2) atmosphere.
rated chains; C-H from aldehyde, C-O from sterols
and alcohols, Carbonyl group (C = O) and Sulfox-
yde group (S = O). A deeper analysis of this FTIR
spectrum, as well as the physical and rheological
properties of the neat and modified asphalt bind-
ers can be found at Gondim et al. (2016).
The thermogravimetric curves of the sap in
both inert (N2) and oxidative (O2) atmosphere are
shown in Figure 6 and Figure 7, respectively.
Both thermograms showed an initial mass loss
from the start of the test to 150°C, which may be
associated with loss of volatiles, loss of free water by
evaporation, and loss of boiling water. This initial
degradation was more pronounced in the oxidative Temperature(•C)

atmosphere, where three events (at temperatures of

40ºC, 60ºC and 100ºC) were observed, resulting in a Figure 7. Thermogravimetric curve of the sap in oxida-
loss of mass of 12.5% against a loss of 5% in inert tive (air) atmosphere.
atmosphere.
The decomposition of the material in N2 occurred
degradation of 77%) and several smaller ones, with
in three events, one with undefined maximum tem-
minor mass reductions. This peculiarity of the curve
perature, referring to the loss of water and volatile,
in oxidative atmosphere indicates complexity in the
and two other well defined events at 280ºC and
chemical structure of the sap under study, and may
370ºC. The decomposition in air occurred in sev-
suggest the presence of unsaturated compounds.
eral events, with two main ones (responsible for a
The maximum decomposition temperatures
(Td) of each event were determined by means of
the first derivative of the thermogravimetric curves
0 ,8 (given by the blue curves), and are presented in
0 .7
Table 1, together with the residual values at the end
of the test (700°C).
0 .&
It is noteworthy that, although the presence of
~ 0 ,5 oxygen resulted in a more complex thermogram
~
-e o.•
and a greater initial loss of water and volatiles,
@
.0
no large variations in the maximum temperatures
< 0 .3
in the main events (at 280ºC and 370ºC) were
0 ,2 observed, suggesting that oxygen has not acceler-
0 .1
ated the main process of degradation.
0 .0
•ooo 3500 3000 2500 2000 1500 1000 500
3.2 Physical and rheological properties of neat
Wave Number (cm-1)
and modified asphalt binders
Figure 5. FTIR spectrum of the sap, from 4000 cm−1 a The results of the penetration, softening point,
400 cm−1 wave numbers. rotational viscosity and FTIR (carbonyl and sul-

406
Table 1. Maximum Temperature and percent of degrada- 800
tion of each decomposing event. ~ ~ PG 64-28 Unaged

O2 N2 6:'
700
' ~ 3%Unaged
......,._ 5% Unaged
~ 10%Unaged
atmosphere atmosphere ~600
- ~ • PG 64-28 RTFOT
£ - .. · 3%RTFOT
Event 1º 2º 3º 4º 5º 6º 1º 2º 3º
Td (ºC) 40 60 100 280 370 510 – 280 370
8 500 - l r · 5%RTFOT
Degradation (%) 12.5 77.0 6.0 5 75.3 14.0 ~ -* · 10% RTFOT
-; 400
Residual at 700ºC (%) 4.24 5.24
.s=
E3oo
~
200
Table 2. Properties of the neat and modified asphalt
100
binders.
130 140 150 160 170 180
Test PG 64 3% 5% 10% Temperature (C)

Penetration (1/10 mm) 59 59 57 55 Figure 8. Viscosity-temperature charts of unaged and

Softening Point (°C) 48 50 49 48 RTFOT aged binders.
Carbonyl Index 1.7350 1.8194 3.205 –
Sulfoxyde Index 0.9106 1.8023 2.1340 –
Rotational 135 592 510 505 478
Viscosity 150 325 257 277 240 90
(cP) 177 120 98 115 105
After RTFOT
Retained Penetration (%) 72 70 81 87 (:85
<!) - PG 64-28 Unaged
Softening Point Increase 5 5 2 5 "50
(°C) - - 3%Unaged
§
Carbonyl Index 3.4032 4.8978 – – <!) 80 - - 5%Unaged
Sulfoxyde Index 1.9560 2.2873 – – "'
~
...s:: - - 10% Unaged
A.. • • • • • • PG 64-28 RTFOT
Rotational 135 753 781 722 699
Viscosity 150 357 375 348 323 75 • • • • • • 3% RTFOT
(cP) 177 123 134 125 125 • • • • • • 5% RTFOT
•••••• 10% RTFOT
70
0.1 10 1000 100000
Complex Modulus (Pa)
foxyde indexes) tests on the neat and modified
asphalt binders, before and after RTFOT aging, Figure 9. Black spaces of neat and modified asphalt
are presented at Table 2. For didactical purpose binders.
the results of the viscosity test are also presented at
Figure 8. Figure 9 presents the Black Space of neat
and modified asphalt binders. The effects of the sap at high in-service tempera-
The penetration test results indicated that, at tures (45ºC to 85ºC) can be analyzed from the black
25ºC the addition of the sap turned the binder spaces presented at Figure 9. One of the expected
stiffer before aging. On the other hand, after effects of the short term aging at RTFOT is the
RTFOT the modified samples showed lower pen- increase at Complex Modulus and the decrease
etration values than the neat one, expressed by the of phase angle of the samples. The Figure shows
increase on the retained penetration, showing that that as higher contents of sap are applied, higher
the modified binders were less susceptible to short are the phase angles for this range of temperature,
term aging (by the penetration point of view). indicating a more viscous (less elastic) behavior.
Around 50ºC, from the results of softening point After RTFOT, the modified asphalt binders pre-
tests, no particular changes were observed on both sented higher phase angles than the neat binder,
unaged and RTFOT samples. The softening points evidencing the solvency effect that the sap has on
before aging are almost the same for all samples, as the asphalt, leading to less stiffer materials.
well as the softening point increase after RTFOT. At higher temperatures (above 135ºC) both
The only exception is the 5% sample, and it can be unaged and RTFOT aged samples indicated that
due to little variations on the time and temperature the addition of sap tends to reduce the viscosity
during the mixing process. of the base binder, showing once more that the sap

407
prevented the excessive hardening of the samples Table 4. Penetration and Abrasion Values of the prime
during RTFOT procedures. coats.
Nevertheless, the carbonyl and sulfoxyde indexes
(indicative of hardening and oxidative aging) 50ºC 60ºC
do not agree with all these results: both indexes Penetration Abrasion Penetration
increased with the simple addition of the sap, and Sample (mm) (kgf/cm2) (mm)
rose even more after RTFOT. This can be explained
by the fact that the sap already contained carbonyl CM-30 8,30 11,2 –
and sulfoxyde groups on its composition, as it was CAK 8,04 8,0 13,6
showed on the item 3.1. Therefore, the FTIR anal- CADL 8,7 8,5 13,3
ysis for the evaluation of the aging process is not
conclusive for the material studied on this research.
the conventional CM-30 binder has already been
approved by Rabêlo methodology (Rabêlo, 2006)
3.3 Prime coat study
according to Almeida (2013).
The soil samples were compacted on a humidity of The results of the tests carried out at 50°C
two percent points below the optimum humidity showed that the penetrations obtained by the new
because this is a common reality in tropical areas, binders at a rate of 0.8 l/m2 (8.04 mm and 8.70 mm)
due to the ease of moisture loss during the com- were compatible with the values observed for the
pacting process of granular layers of pavements. conventional binder (8.30) applied at a higher rate.
The cutback rate of 0.8 l/m2 is considered a This indicates that addition of the sap did not
low rate, and was chosen because it represents an avoid the binder from penetrating satisfactorily in
economic option that has been proven sufficient this type of soil. It is likely that if the new bind-
in previous research, using the traditional cut- ers were applied at the rate of 1.0 l/m2 greater pen-
back (CM-30) (Rabêlo, 2006) and using a soybean etrations would be observed, or there would be a
biodiesel as solvent for the asphalt binder (Vascon- higher concentration of binder at the top of the
celos, 2009). specimens.
The temperature of the cutback application in It was also observed that the penetrations
prime coat services is normally around 50°C when obtained with the CAK and CADL were compat-
CM-30 is applied. On this research, the tempera- ible with each other, being a little larger with the
ture of application were 50°C and 60°C, in case use of CADL, indicating that d-limonene was suit-
of the recurring temperature was not sufficient to able for the task of diluting the modified asphalt.
fluidize the sample. The increase in primer temperature was unnec-
The densities of the kerosene-diluted modified essary, since the penetrations at 50°C were similar
binder (CAK) and d-limonene-diluted modified to that observed for CM-30, even at the lower con-
binder (CADL) are shown in Table 3. The densities tent. As the temperature increased, considerably
are important data since the control of the applica- higher penetrations were obtained, which could be
tion rate of the binder was performed by measur- adverse to the durability and functionality of the
ing the weight. prime coat. Due to excessive penetrations, no abra-
The penetration and the abrasion values of the sion tests were performed at 60°C.
samples primed with the CAK and the CADL Regarding the abrasion tests, the values found
cutbacks are presented at Table 4. For compari- for the prime coat with the CADL were similar
son purposes, this table also contains the penetra- but slightly larger than those found with the CAK,
tion and abrasion values of the same soil sample, indicating that the d-limonene added a little more
compacted at the same humidity, using CM-30 friction to the surface of the specimen.
at 50ºC, but at a rate of 1.0 l/m2, performed by The abrasion observed with the application
another operator. These data are important, even of CM-30 at a rate of 1.0 l/ m2 was considerably
if the test conditions are not exactly the same, higher than that found with alternative binders.
because the prime coat of this soil sample with This increase in strength may be due to the greater
amount of binder at the top of the specimen, by
the nature of the asphalt binder, or simply for the
change on the operator.
Table 3. Densities of the CAK and CADL cutbacks.

Sample Density (g/cm3) 4 CONCLUSIONS

CAK 0,931
CADL 0,900 The FTIR spectrum of the sap of Euphorbia
Tirucalli as well as the termogravimetric curves

408
evidenced the complexity of the chemical structure ASTM D36/D36M-12, Standard Test Method for Sof-
of this material. tening Point of Bitumen (Ring-and-Ball Apparatus),
Carbonyl and sulfoxyde groups were identified American Society for Testing and Materials, 2012.
at the sap spectrum, turning unfeasible the aging ASTM D4402-06. Standard Test Method for Viscos-
ity Determinations of Unfilled Asphalts Using the
evaluation of the modified samples by the FTIR Brookfield Thermosel Apparatus. American Society
analysis. for Testing and Materials, 2006.
Both termogravimetric curves showed degra- ASTM D5-06e1, Standard Test Method for Penetration
dation process from the begging of the test until of Bituminous Materials, American Society for Test-
150ºC, which can be associated to the loss of vola- ing and Materials, 2006.
tiles and water, being that loss more pronounced at ASTM D7175-15. Standard Test Method for Determin-
oxidative atmosphere. ing the Rheological Properties of Asphalt Binder
An event at 510ºC only observed on oxidative Using a Dynamic Shear Rheometer. American Soci-
atmosphere indicates the presence of compounds ety for Testing and Materials, 2015.
Audo, M.; Chailleux, E.; Bujoli, B.; Queffélec, C.;
that suffered oxidation. Legrand, J.; Lépine, O. (2012) Alternative Binder
The sap of Euphorbia Tirucalli promoted a from Microalgae. Alternative Binders for Sustain-
hardennig on the unaged asphalt binder at 25ºC, able Asphalt Pavements—Papers from a Workshop,
but at higher temperatures the presence of the Transportation Research Circular, Number E-C165,
sap had a solvency effect, increasing the phase August, 2012, Washington D.C.
angle at high in service temperature and dimin- Calvin, M. (1980) Hydrocarbons from Plants: Analytical
ishing the viscosity on mixing and compaction Methods and Observations. Die Naturwissenschaften
temperatures. 67:525 (1980).
The solvency effect contributes to reduce the Çelik, O.N.; Atasağun, N. (2012) Rheological Properties
of Bituminous Binder Modified with Nigella Pulp
effects of oxidative aging, once the modified Liquefied by Means of Pyrolysis Method In: 2ND
asphalts were softer than the neat binder after INTERNATIONAL SYMPOSIUM ON ASPHALT
RTFOT, even though the carbonyl and sulfoxyde PAVEMENTS & ENVIRONMENT, 2012, Fortaleza.
index were higher on modified samples. International Society for Asphalt Pavement—ISAP.
The cutbacks produced from the binder modi- ECOPAVE (2011) Available at: http://www.ecopave.com.
fied with 10% sap, using kerosene and d-limonene au/index.htm. Access on 17/11/2011
as diluents, provided primers with similar penetra- Eshel, A.; Zilberstein, A.; Alekparov, C.; Eilam, I.; Oren,
tions to those observed with conventional cutback I.; Sasson, Y.; Valentini, R.; Waisel, Y. (2010) Biomass
(CM-30), even though a lower rate was applied. production by Desert Halophytes: Alleviating the
Pressure on Food Production. Proceedings of the 5th
The modified binder diluted in d-limonene IASME/WSEAS Steven Points, Wisconsin. USA.
showed penetrations and adhesions similar to Gondim, L. M.; Soares, S. A.; Barroso, S. H. A. (2016)
those obtained by the modified binder diluted in Pretoleum Plant Sap as an Asphalt Modified for Pave-
kerosene, indicating the potential use of this prod- ment Applications. International Journal of Civil &
uct in the manufacture of alternative cutbacks. Environmental Engineering IJCEE-IJENS, Vol 16
Issue 06 - http://ijens.org/Vol_16_I_06/162006-4848-
IJCEE-IJENS.pdf
ACKNOWLEDGMENTS Leite, L.; Chacur, M.; Nascimento. L.A.; Cravo,
M.C.; Martins, A.T. (2012) The Use of Vegetal
Products as Asphalt Cement Modifiers In: 5TH
The authors acknowledge FUNCAP for financial EUROASPHALT & EUROBITUME CONGRESS,
support and LUBNOR/Petrobras for the donation 2012, Istanbul.
of the asphaltic binder samples. Mohamed Metwally, M. A. R. (2010) Development of
Non-Petroleum Binders Derived from Fast Pyrolysis
Bio-Oils for Use in Flexible Pavement. Theses (Doctor
REFERENCES of Philosophy) Civil Engineering Materials. Iowa State
University. Ames, Iowa, United States of America.
Almeida, F. D. de. (2013) Desenvolvimento de um Peralta, J.; Silva, H.M.R.D; Williams, R.C.; Rover, M.;
método expedido para a avaliação da imprimação Machado, A.V.A. (2013) Development of an Inno-
betuminosa de rodovias. Monografia, Curso de Tec- vative Bio-Binder Using Asphalt-Rubber Technol-
nologia em Estradas, Instituto Federal de Educação, ogy. International Journal of Pavement Research and
Ciência e Tecnologia do Ceará, Fortaleza, CE, 70f. Technology 10.6135/ijprt.org.tw/2013.6(4).
ASTM D2493-09. Standard Viscosity-Temperature Rabêlo, A. N. (2006) Contribuição ao Estudo da Impri-
Chart for Asphalts. American Society for Testing and mação Betuminosa das Bases Rodoviárias do Estado
Materials, 2009. do Ceará. Dissertação de Mestrado, Programa de
ASTM D2872-12e1. Standard Test Method for Effect of Mestrado em Engenharia do Transportes, Universi-
Heat and Air on a Moving Film of Asphalt (Rolling dade Federal do Ceará, Fortaleza, CE, 204 fl.
Thin-Film Oven Test). American Society for Testing Seidel, J. C.; Haddock, J. E. (2012) Soy Fatty Acid as
and Materials, 2012. Sustainable Modifier for Asphalt Binders. Alternative

409
 Binders for Sustainable Asphalt Pavements—Papers Imprimação Betuminosa de Pavimentos Rodoviários.
from a Workshop, Transportation Research Circular, Monografia de Graduação em Engenharia Civil,
Number E-C165, August, 2012, Washington D.C. Curso de Engenharia Civil, Universidade Federal do
Teixeira, S. M. (2013) Estudo do Asfalto Modificado Ceará, Fortaleza, CE, 81 fl.
com a Seiva do AVELOZ (Euphorbia Tirucalli): Uma Wen, H.; Bhusal, S.; Wen, B. (2012) Laboratory Evalua-
Revisão. Monografia de Conclusão do Curso. Curso tion of Waste Cooking Oil-Based Bioasphalt as Sus-
de Engenharia Civil, Universidade Federal do Ceará— tainable Binder for Hot-Mix Asphalt. Alternative
Campus Cariri, Juazeiro do Norte, CE, 58fl. Binders for Sustainable Asphalt Pavements—Papers
Vasconcelos, M. A. G. (2009) Primeiro Estudo da Apli- from a Workshop, Transportation Research Circular,
cação de Biodiesel como Material Alternativo para Number E-C165, August, 2012, Washington D.C.

410
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Physical properties of sepiolite clay nanofibers modified bitumen

T. Günay, P. Ahmedzade & S. Altun

Ege University, Bornova, Turkey

ABSTRACT: Bitumen is a material especially used in flexible pavements, runaway in airfields as binder.
Several types of modifiers are employed in order to enhance properties of bitumen against permanent
deformation and to avoid cracking and deteriorations of pavement. Recently, nanomaterials gain traction
in many fields including engineering materials by presenting entirely new properties of materials such as
nanoscale particle size, very high surface area, and good suspension ability. This paper describes the physi-
cal effects of nanofibers on bitumen as modifier. Sepiolite clay nanofibers having of particle diameter of
less than 100 nm was employed in bitumen with amounts of 2%, 3%, and 4% by total weight of mixture.
The physical effects of the sepiolite clay on bitumen were investigated by conventional tests such as pene-
tration, softening point and ductility. Storage stability test was performed on modified bitumen to observe
the settling condition of modified bitumen during storage. Rolling Thin Film Oven Test (RTFOT) was
applied to modified bitumen to investigate effect of the nano modifier on short term aging properties
of bitumen. Rotational Ciscometer (RV) and Dynamic Shear Rheometer (DSR) was done to investigate
fundamental rheological properties such as viscosity, complex shear modulus and phase angle of sepiolite
clay modified bitumen. Nanoclay modification leads to a decrease in penetration, an increase in soften-
ing point, and which indicate an increasing stiffness of bitumen. RTFOT results shows that Sepiolite clay
nanofibers has a positive effect on short term aging of bitumen while good storage ability of modified
bitumen was determined by storage stability test. Increased rutting parameter after modification obtained
by DSR test is a clear indication of better rheological properties of bitumen which means permanent
deformation occurred in flexible pavement can be restrained by sepiolite clay nanofibers modification.

1 INTRODUCTION Thus, considerable research have been done so far

which investigate effects of nanomaterials as modi-
Bitumen, a complex material basically composed fier in bitumen and improved rutting performance,
of hydrocarbons and used in many engineering good adhesive interaction, better short term aging
applications especially in flexible highway pave- properties were reported (Yao et al., 2013; Nazzal
ments and runaways, is a residue from crude oil et al., 2012; Ghile 2006; Ye & Chen 2009).
distillation. Due to its viscoelastic and thermo- Within this scope, in this paper, Sepiolite Clay
plastic properties, bitumen behaves like an elastic Nanofiber (SCNF) was used as a modifier and
solid at low temperatures or under rapid load- bitumen was investigated in terms of physical and
ing and like a viscous fluid at high temperatures rheological properties. SCNF having of particle
or under slow loading (Behzadfar & Hatzikiria- diameter of less than 100 nm was employed as a
kos 2013). In order to minimalize the permeant modifier in bitumen. SCNF was added to bitumen
deformation caused by these two properties vari- with amounts of %2, %3 and % 4 by total weight
ous additives have been used in bitumen. Poly- of mixture. The physical effects of the sepiolite clay
mers; such as Styrene-Butadiene-Styrene (SBS), on bitumen were investigated by conventional tests
Ethylene-Vinyl-Acetate (EVA), rubbers, fibers; such as penetration, softening point and ductility.
polypropylene, fiber glasses (Ferotti et al. 20129, Storage stability test was performed on modified
micro and nanoscale fillers; Portland cement, lime bitumen to observe the settling condition of modi-
(Niazi & Jalili 2009) carbon black (Chaala et al. fied bitumen during storage. Rolling Thin Film
1996), nanoclays (Jahromi & Khodaii 2009) are Oven Test (RTFOT) was applied to modified bitu-
some of the additives used for modification of men to investigate effect of the nano modifier on
bitumen and obtained some promising results in short term aging properties of bitumen. Rotational
terms of rheological improvement of bitumen. Viscometer (RV) and Dynamic Shear Rheometer
Nano scale materials gain traction in all fields of (DSR) was done to investigate fundamental rheo-
science recently due to its colloidal structure, high logical properties such as viscosity, complex shear
surface area and ability to improve physical prop- modulus and phase angle of sepiolite clay modi-
erties of polymers by forming nanocomposites. fied bitumen.

411
 applied to base and SCNF modified bitumens to
investigate the effects of the modifier on physical
properties of bitumen. Conventional tests were all
conducted in accordance with ASTM given in the
parenthesis. Penetration index was also calculated
in order to investigate the temperature susceptibil-
ity of the binders.

3.2 Storage stability test

Since sepiolite clays provide a stable suspension
with increased viscosity and thixotropic behavior
the settling of the SCNF modified bitumen during
Figure 1. Sepiolite clay nanofiber. storage at high temperature was investigated. Stor-
age stability test was conducted by means of a tube
with a height of 160 mm and a diameter of 30 mm,
Table 1. Properties of sepiolite nano fiber clay. filled with 50 g of modified bitumen. According
to EN 13399 standards, the tube filled with modi-
Chemical formula Mg4Si6O15(OH)2.6H2O fied bitumen is closed tightly and placed in an oven
adjusted to 180°C for 72 hours. After, simulation
Diameter < 100 nm
of storage conditions, the tube is left in room tem-
Type Phyllosilicate
peratures to cool down. The tube is cut into 3 equal
Appearance White, powder
parts, the bitumens remained in the top and bot-
tom parts of the tube are reserved for penetration
and softening point tests to observe whether phase
2 EXPERIMENTAL INVESTIGATION separation is occurred during storage.

2.1 Materials
3.3 Rotational viscosity test
Bitumen having penetration of B 50/70 was
Rotational viscosity of Base and SCNF modi-
used throughout testing program. Sepiolite Clay
fied binders were measured at 135°C and 165°C,
Nanofiber (SCNF), is a hydrated magnesium sili-
respectively according to ASTM standard of 4402.
cate component (Fig. 1.) (Fernández-Saavedra
After the measurements, ideal mixing and compac-
et al. 2004). Properties of SCNF were given in
tion temperature ranges of the binders in flexible
Table 1. Silverstone L5M high shear mixer was
pavement application were calculated by means of
used for preparation of modified bitumen. Bitumen
viscosity test results (ASTM D4402).
was heated to 180°C for 90 min then poured into
flasks of high shear mixer fixed to 800 rpm. SCNF
(2%, 3%, and 4% by total weight of mixture) was 3.4 Short term aging of bitumen
added to bitumen in 15 minutes. The rate was rose
Rolling Thin Film Oven Test (RTFOT) was con-
to 2800 rpm and mixing process continues for 60
ducted for short term aging of the base and modi-
minutes. After preparation process SCNF modified
fied binders according to ASTM 2872. Eight
bitumen removed from flask and cooled in room
cylindrical glass containers each having 35 g of
temperature.
binder are used for ageing at 163°C for 75–85 min-
The sample used in this work was coded as below;
utes. Cylindrical containers are fixed in a vertically
carousel shelf of the oven and it revolves at rate
Base 50/70 Bitumen – “B”
of 15 rpm.
Base 50/70 Bitumen + 2% SCNF – “B-SCNF2”
Base 50/70 Bitumen + 3% SCNF – “B-SCNF3”
Base 50/70 Bitumen + 4% SCNF – “B-SCNF4” 3.5 Dynamic mechanical test
Dynamic shear rheometer test (Anton Paar Smart-
pave Plus) was used to investigate rheology of base
and SCNF modified bitumen by determining high
3 TESTING PROGRAM
performance grade (PG X-Y) of bitumens. After
the DSR tests the fundamental rheological param-
3.1 Conventional tests
eters such as G* and δ were determined for base
Penetration (ASTM D5), softening point (ASTM and SCNF modified bitumens (ASTM D7175). Vis-
D36) and ductility (ASTM D 113) tests were coelastic range (LVE) of binders was determined in

412
order to eliminate occurrence of permanent defor- However, it should be noted that an increment
mations on the samples. End of the LVE tests, the in modifier content leads to an increase in phase
amplitude of the strain was limited of 5% for all type separation since total amount of the modifier in
of binders. PG Measurements were taken at 58, 64, bitumen increases.
70, 76, 82°C. One type of plate with a diameter of
25 mm and 1 mm gap between the plates was selected
4.3 Rotational viscosity test
as test geometry throughout the testing program.
Viscosity values of base and SCNF modified bitu-
mens were given in Table 4. The mixing and compact-
4 RESULTS AND DISCUSSIONS ing temperatures calculated by viscosities of binders
were also represented in Table 4. As can be seen, vis-
4.1 Conventional tests cosity of base bitumen was found as 412,5 cP at and
110 cP at 165°C. The results of viscosity belong to
Penetration, softening point, ductility tests result
SCNF modified bitumen show a consistent increase
and Penetration Indexes (PI) were given in Table 2.
at all test temperatures compared to base bitumen
Penetration is in a decreasing trend after SCNF
which indicates bitumen became stiffer after SCNF
modification as modifier content increase whereas
modification. The increment in viscosity rises as
softening point values generally tend to increase.
modifier content increases. Mixing and compacting
Accordingly, base bitumen has the highest penetra-
temperatures should not be elevated considerably
tion and lowest softening value than that of any
after polymer modification due to excessive energy
binder in this testing program. The trends in pen-
consumption. It should be noted that while SCNF
etration and softening point clearly point out that
modification has a stiffening effect on bitumen at
SCNF modifier increases stiffness of bitumen. PI
medium and high temperature zones, this stiffening
which signifies thermoplastic behavior of bitumen is
effect tend to diminish at elevated temperatures such
−2.13 for base bitumen which is the lowest between
140–160°C where mixing and compacting processes
the all binders. Penetration index are clearly higher
are applied.
for modified bitumen that indicate temperature sus-
ceptibility is reduced after modification.
Ductility test results reveal that there is a slight 4.4 Short term aging of binders
increase in ductility which means elongation abil-
Loss of weight of the binders after short term
ity of bitumen without corruption is slightly
aging simulating by means of RTFO were given
enhanced after the modification. This result can be
in Table 5. As can be seen, the losses of weight of
attributing to fiber structure of SCNF.

4.2 Storage stability Table 3. Storage stability results of SCNF modified

binders.
Penetration and softening point results taken
at top and bottom of the storage stability tube Properties B-SCNF2 B-SCNF3 B-SCNF4
were given in Table 3. There is a slight difference
in softening point and penetration results taken Penetration at 39 37 37
from lower and upper sections of SCNF modified bottom (dmm)
binder. This result indicates that the phase separa- Penetration at 41 39 40
top (dmm)
tion between bitumen and SCNF modifier during
Softening point 54,5 56,5 55,5
storage is acceptable. Low phase separation during at bottom (oC)
hot storage condition can attribute to good sus- Softening point 53,5 56 55
pension ability of nanoparticles (Che et al. 2007). at top (oC)
Table 2. Physical test results of base and SCNF Modi-
fied binders.
Table 4. Rotational viscosity test results.
Properties B B-SCNF2 B-SCNF3 B-SCNF4
Temperature range
Penetration 53 43 41 40
(dmm) Binder type 135°C 165°C Mixing Compacting
Softening 46 53 54 53.5
point (oC) B 412,5 110 135–142 122–128
Penetration −2.13 −0.82 −0.70 −0.87 B-SCNF2 447,5 127,5 144–149 139–143
index B-SCNF3 450 130 147–153 143–148
Ductility (cm) 79 83 86 82 B-SCNF4 455 127,5 149–155 150–153

413
Table 5. Loss of weight of the binders after RTFO. Table 6. DSR test result and performance grade of the
binders.
Properties B B-SCNF2 B-SCNF3 B-SCNF4
Temp. G* Δ G*/sinδ
Original (g) 34.23 35.20 35,10 35,00 Binders (°C) (KPa) (°) (KPa) PG
Aged (g) 34.00 34.98 34,89 34,79
Loss (%) 0.66 0.62 0.61 0.61 B Original 52 9.7 84.2 9.77
58 3.8 86.1 3.85
64 1.5 87.6 1.59

PG 64-Y
binders are all below the 1% percent which is gener- 70 0.7 88.9 0.72
ally considered as a limitation for short term aging Aged 52 20.0 76.7 20.6
[Corbett and Merz 1975]. This result means that all 58 9.0 77.6 9.24
binders are suitable in terms of short term aging. 64 4.5 78.9 4.62
In addition to this result, weight of loss in short 70 1.9 79.7 2.01
term aging, which occurs due to volatilization of
Original 52 11.4 82.1 11.5
aromatics in bitumen, diminishes after SCNF
58 5.6 83.5 5.61
modification. The reason behind the this phe- 64 2.1 85.5 2.11
nomenon might be explained by nanoparticles

B-SCNF2
70 1.0 86.7 1.00

PG 64-Y
that found in bitumen as exfoliated which blocks 76 0.5 88.0 0.49
vaporization of gas molecules in bitumen by act- Aged 52 25.1 72.2 26.4
ing as barrier, as reported former study (Zare- 58 11.6 73.6 12.3
Shahabadi et al. 2010). 64 4.8 76.1 4.91
70 2.2 78.2 2.27
4.5 Dynamic shear rheometer 76 1.2 79.9 1.23

The rheological parameters such as complex shear Original 52 12.1 82.2 12.2
modulus (G*) phase angle (δ) and rutting param- 58 6.0 83.8 6.01
64 2.3 84.9 2.26
eter (G*/sinδ) obtained by means of DSR device
B-SCNF3

70 1.1 85.5 1.08

PG 64-Y
were given in Table 6.
76 0.5 86.4 0.51
Table 6 is also included high temperature Per-
Aged 52 24.1 71.4 25.5
formance Grade (PG X-Y) of binders which deter-
58 11.9 72.9 12.5
mined by criteria that the highest temperature is
64 5.9 73.8 6.12
assigned as PG at which rutting parameter both 70 2.3 75.2 2.40
provide higher value than 1.0 kPa for original 76 1.2 77.5 1.27
binder and 2.2 kPa for aged. It can be seen, base
bitumen has a grade of PG 64-Y whist all SCNF Original 52 13.5 81.4 13.6
modified binders has a grade of PG 70-Y which is 58 6.4 83.0 6.45
one level higher than PG 64-Y. 64 2.4 84.6 2.42
B-SCNF4

Better high temperature PG means that SCNF 70 1.2 85.9 1.21

PG 64-Y
modified bitumens can be employed at higher tem- 76 0.6 86.9 0.60
perature region than base bitumen without any per- Aged 52 25.8 72.5 27.0
manent deformation. This result is also confirmed 58 12.3 73.5 12.9
64 6.0 74.9 6.21
penetration index result that was found higher for
70 2.4 77.2 2.47
modified bitumen than that of the base bitumen.
76 1.3 79.0 1.35
Thus, it can be said that, SCNF modification lead
to diminish temperature susceptibility of bitumen.
Moreover, G* values tend to increase for almost
whole testing temperatures after SCNF modifi- 5 CONCLUSION
cation that confirms stiffening effect of SCNF.
However, the increment in G*, which indicates the The effects of the SCNF modifier on physical prop-
resistance against shear force, remains almost same erties of bitumen were evaluated by means multiple
as SNCF content in bitumen increases. Hence, it tests such as penetration, softening point, ductility,
was found that increasing SCNF modifier content RTFO, storage stability, RV and DSR. Conven-
does not effective on G* and high temperature PG. tional test results provide clear evidence of stiffening
The values of δ, ratio of permanent deformation effect of SCNF modifier on bitumen as penetration
to elastic deformation, was found generally lower decreased and softening point generally increased
for modified binders. High values of G* and low after SNCF modifier addition. Short term aging
values of δ attribute to lesser rutting occurrence properties of all binders are in acceptable limit as
[Lavin 2003]. none of their weight loss higher than 1%. RV test

414
results were another indication for stiffening effect Corbett, L. and Merz, R. 1975. Asphalt binder harden-
of SCNF modifier as rotational viscosity values ing in the Michigan test road after 18 years of service.
consistently increases. Penetration indexes belong Transportation Research Record. 544.
to the binders indicate that SNCF modified bind- Chaala, A., Roy, C.& Ait-Kadi, A. 1996 Rheological
properties of bitumen modified with pyrolytic carbon
ers have low temperature susceptibility than that black, Fuel. 75(13), 1575–1583.
of base bitumen. This result was also confirmed by Che, J., Xiao, Y., Wang, X., Pan, A., Yuan, W. & Wu,
DSR test where high temperature PG of binders X. 2007. Grafting polymerization of polyacetal onto
were found elevated after the modification. nano-silica surface via bridging isocyanate. Surface
As a result of the testing program, it should and Coatings Technology. 201(8), 4578–4584.
be noted that SCNF modifier has multiple posi- Fernández-Saavedra, R., Aranda, P. & Ruiz-Hitzky, E.
tive effects on bitumen especially in term of short 2004. Templated synthesis of carbon nanofibers from
term aging, stiffening, elastic behaviour and tem- polyacrylonitrile using sepiolite. Advanced Func-
perature susceptibility. Rheological findings were tional Materials. 14(1), 77–82.
Ferrotti, G., Canestrari, F., Pasquini, E. & Virgili, A. 2012.
also proved that SCNF modified bitumen is more Experimental evaluation of the influence of surface
resistant against permanent deformation and thus, coating on fiberglass geogrid performance in asphalt
it can be employed in wider temperature ranges. pavements. Geotextiles and Geomembranes. 34, 11–18.
Ghile, D. B. 2006. Effects of nanoclay modification on
rheology of bitumen and on performance of asphalt
REFERENCES mixtures. Delft, The Netherlands: Delft University of
Technology.
ASTM D113-99, Standard Test Method for Ductility Jahromi, S. G. & Khodaii, A. 2009. Effects of nanoclay
of Bituminous Materials, ASTM International, West on rheological properties of bitumen binder. Con-
Conshohocken, PA, 1999, 3 pages. struction and Building Materials. 23(8), 2894–2904.
ASTM D2872. “Standard Test Method for Effect of Lavin, P. (2003). Asphalt Pavements: A Practical Guide
Heat and Air on a Moving Film of Asphalt (Rolling to Design, Production and Maintenance for Engi-
Thin-Film Oven Test).” ASTM International, West neers and Architects, Taylor & Francis.
Conshohocken, PA, 2003, 6 pages. Nazzal, M. D., Kaya, S., Gunay, T. & Ahmedzade, P.
ASTM D36/D36M-14e1. “Standard Test Method for 2012 Fundamental characterization of asphalt clay
Softening Point of Bitumen (Ring-and-Ball Appa- nanocomposites. Journal of Nanomechanics and
ratus), ASTM International, West Con-shohocken, Micromechanics. 3(1), 1–8.
PA, 2014, 3 pages. Niazi, Y. & Jalili, M. 2009. Effect of Portland cement and
ASTM D4402-06. “Standard Test Method for Viscosity lime additives on properties of cold inplace recycled
Determination of Asphalt at Elevated Temperatures mixtures with asphalt emulsion, Construction and
Using a Rotational Viscometer.” ASTM Interna- Building Materials. 23(3), 1338–1343.
tional, West Conshohocken, PA, 2006, 3 pages. Yao, H., You, Z., Li, L., Goh, S. W., Lee, C. H., Yap,
ASTM D5/D5M-13. “Standard Test Method for Pen- Y. K. & Shi, X. 2013. Rheological properties and
etration of Bituminous Materials.” ASTM Interna- chemical analysis of nanoclay and carbon microfiber
tional, West Conshohocken, PA, 2013, 4 pages. modified asphalt with Fourier transform infrared
ASTM D7175-15. “Standard Test Method for Deter- spectroscopy. Construction and Building Materials.
mining the Rheological Properties of Asphalt Binder 38, 327–337.
Using a Dynamic Shear Rheometer.” ASTM Interna- Ye, C. & Chen, H. 2009. Study on road performance of
tional, West Conshohocken, PA, 2015, 16 pages. nano SiO−2 and nano TiO−2 modified asphalt. New
Behzadfar, E. & Hatzikiriakos, S. G. 2013. Viscoelastic Building Materials, 6, 027.
properties and constitutive modelling of bitumen, Zare-Shahabadi, A., Shokuhfar, A. & Ebrahimi-Nejad,
Fuel. 108(0), 391–399. S. 2010. Preparation and rheological characteriza-
BS EN 13399:2010. “Bitumen and Bituminous Binders. tion of asphalt binders reinforced with layered silicate
Determination of Storage Stability of Modified Bitu- nanoparticles. Construction and Building Materials,
men.” British Standards Institution, 2010, 12 pages. 24(7), 1239–1244.

415
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

High temperatures performance investigation of geopolymer modified

bitumen binders

S.I.A. Ali
Near East University, Lefkoşa, North Cyprus

H.A.M. Yahia
Middle East College, Knowledge Oasis Muscat, Oman

A.N.H. Ibrahim & R.A. Al Mansob

Universiti Kebangsaan Malaysia, Bangi, Malaysia

ABSTRACT: Bitumen binder mixing techniques attempt to balance the composition of bitumen and
the modifier to achieve the long-lasting performance of modified bitumen binder, which led to enhance
the pavement performance in roads construction. The main purpose of modification of bitumen binder
is to produce mixtures with high resistance to deformation and cracking. The properties of the produced
blends are depending on the physical and chemical properties of the used materials. The concentrations
that used to produce modified bitumen binders were 0, 3, 5 and 7%.wt from the weight of bitumen,
and various tests were used to evaluate the performance of base and modified asphalt binders including
softening point, viscosity and Dynamic Shear Rheometer (DSR). The results of the best performance
of modified bitumen binders 5% GPMB shows that the stiffness at high temperatures was improved
as observed from the softening point and the viscosity, the enhancement was up to 20.21 and 31.33%,
respectively. Also, the permanent deformation reduced nearly 3 times compared the base asphalt binder.

1 INTRODUCTION come out with a significant modification of the

theological reaction (Fuentes-Audén et al. 2008; Per-
The bitumen binder and mixture need to be alta 2010). Geopolymer Modified Bitumen (GPMB)
improved to reduce and prevent pavement distresses. is a new method to capitalize waste materials and
As a result, polymers became a part of the materi- use those materials in road paving. According to
als used in the large quantity of applications than Ibrahim et al 2016, the geopolymer has constituents
any other categories of materials available. There- two main components namely; the alkaline liquid
fore to improve the properties of asphalt binder in and the source of materials. The alkaline liquids
the present, the polymers were used as additives to ordinarily used in geo-polymerization are blends
improve bitumen binder and mixture properties of sodium silicate (Na2SiO3) and sodium hydroxide
because of their excellent physical and chemical (NaOH) or potassium silicate (K2SiO3) and potas-
properties. Polymers can be defined as large mol- sium hydroxide (KOH), while the source of materials
ecules composed of a large number of correlations is based on aluminum and silicon (Ibrahim, Ahmad,
of small molecules with each other and this small et al. 2016). In this study, the rheological properties
molecule is called monomers (Saoula et al. 2008). of geopolymer modified bitumen binders with the
Polymer modified bitumen binder shows significant concentrations of 3, 5 and 7% at elevated tempera-
improvement of mechanical and rheological prop- tures were instigated using several testings.
erties at low, intermediate and high temperatures
(Airey 2003). The combination of polymers provides
an opportunity to be reused and recycled at the same
2 EXPERIMENTAL METHODS
time allows the possibility of maintaining sources
as well as reduce the costs. On the other side, the
2.1 Materials
combination of polymers successfully improved the
performance and properties of bitumen (Liu et al. The base bitumen used was 60/70 penetration grade,
2009). The effectiveness of polymer concentration and the geopolymer was the combination of fly
on the microstructure, thermal and rheological prop- ash and the alkali liquid. The fly ash class F with a
erties using recycled polyethylene modified bitumen specific gravity of 2.26 was obtained from local

417
company in Malaysia, while the alkali liquid was 3 EXPERIMENTAL TEST PROCEDURES
sodium silicate solution (Na2SiO3) and sodium
hydroxide (NaOH) pallet diluted in water to produce 3.1 The physical properties tests
8 Molar (8M) NaOH solution. A combination of
The physical properties tests were used to investi-
sodium silicate solution and sodium hydroxide solu-
gate the physical properties of GBMB including
tion was prepared to activate the alumino-silicate
the penetration, softening point, and the ductility
precursors in fly ash through a series of dissolution-
(25°C) tests which were conducted based on ASTM
hydrolysis-polycondensation (Ibrahim, Yusoff, et al.
specification namely; D5, ASTM D36, and ASTM
2016). The physical properties of the base bitumen
D113 respectively. According to the research aims,
are shown in Table 1, while the chemical oxide com-
the softening point test was only test performed on
positions of the fly ash are listed in Table 2.
the modified bitumen binders, while the penetra-
tion and ductility were conducted for base bitumen
2.2 Preparation of Geopolymer Modified binder to ensure that the bitumen was within the
Bitumen (GPMB) requirement of ASTM specification.
The base bitumen was heated until it becomes fluid,
then the geopolymer gel was added into the base 3.2 Viscosity
asphalt blend with different concentrations namely;
The Brookfield rotational Viscometer was used for
3, 5 and 7% by weight of the bitumen and the blends
the assessment of the viscosity of base bitumen
was mixed using mechanical shear mixer for 90 min-
binder and GPMB samples. Three readings were
utes under a speed of 1000 rpm with a temperature
noted for each test temperature, and the average was
of 150 C°±5 to produce a homogenous blends. The
recognized as the test result. In addition, the test
softening point test was used to estimate the homo-
temperatures were 135°C and 165°C respectively.
geneity of the blends. A sample from each test was
taken every 30 minutes during the 2 hours mixing
period until the value for the softening point was 3.3 The rheological properties test
stabilized. Moreover, the melt blending method was
The Dynamic Shear Rheometer (DSR) is used to
used to prepare the GPMB, and each concentration
characterize the viscous and elastic behaviour of
of the geopolymer in bitumen blends was coded in
asphalt cement at intermediate and high in-service
this study such as 3% = 3% GPMB.
temperatures. Moreover, DSR measures the rheo-
logical properties factors such as complex shear
Table 1. Physical properties of the base binder. modulus (G*) and the phase angle (δ) of bitumen
binder at the desired temperature and frequency of
Material Properties Test method Value loading, as per AASHTO T315. Frequency sweeps
were conducted for all binders using a 25 mm
Asphalt 60/70 Specific Gravity ASTM D70 1.03 diameter plate and a 1 mm gap for intermediate
Penetration @ ASTM D5 82 to high temperatures 25–75°C. The tests were per-
25°C formed at nine frequencies ranging from 1 rad/s
Softening point ASTM D36 46.0
(nearly 0.159 Hz) to 100 rad/s (nearly 15 Hz) while
(°C)
the test temperatures were from 46+6 to 82°C. The
Viscosity @ ASTM D4402 0.24
135°C (Pa.s) failure temperatures are considered for all bitumen
blends according to the Superpave specification
for G*/sin δ values of less than 1.0 kPa in the case
of unaged bitumen. In addition, the results of the
Table 2. Chemical Composition (%) of Oxide Fly ash complex shear modulus G* and phase angle δ were
(Chien et al. 2012).
used for the construction of master curves. The
Oxide Fly ash master curve of the asphalt cement characteristic
at a reference temperature Tref is defined as the
SiO2 52.50 correlation between the stiffness and the reduced
Al2O3 22.82 loading time or frequency (Walubita et al. 2011).
Fe2O3 5.34
CaO 7.16
MgO 2.56 4 RESULTS AND DISCUSSION
SO3 0.20
K2O 0.99 4.1 The influences of GPMB on softening point
Na2O 0.48
LOI 3.35
The modifier effects on softening point test of
base bitumen binder using geopolymer are shown

418
in Figure 1. The consequences show that the modi- :::: 135'(
fied binders have the highest softening point values
0.5
compared with the base binder. This indicates that
the addition increase of geopolymer in bitumen 0.45
blends able to increase the hardness of modified 0.4
binders. Moreover, the significant enhancement ';;)'

was observed with the concentration of 5% ""

c., 0.35
GPMB. It was noted that 7% GPMB for the sof- 5
0 0.3
tening point slightly decreased, but it’s better than ·;;;
0
the base binder in terms of performance. This is u 0.25
en
reduction it might be due to the ununiformed dis- > 0.2
persion or agglomeration of GPMB.
0.15
0. 1
4.2 Effects of GPMB on viscosity
0.05
Figure 2 shows the results of the viscosity of base
binder and GPMB binders. It can be seen that values 0
were decreased instantaneously with increasing the 0% GPMB 3% GPMB 5% GPMB 7% GPMB
temperature, regardless of the geopolymer concen- Modifier Content (%)
tration. Moreover, it was found that the base binder
has the lowest viscosity, whereas the modified sam-
ple with 5% GPMB has the highest viscosity. This Figures 2. Viscosity of base binder and GPMB binders.
indicated that 5% GPMB has better performance in
terms of rutting resistance. In addition, the viscosity
test results of base binder and modified binders with
4.3 Effects of GPMB on rheological properties
geopolymer remains within the Superpave specifica-
tions at the temperature of 135°C, were the viscos- 4.3.1 Isochronal plot
ity should be less than the maximum limit of 3 Pa.s. Always The isochronal plots of the G* contrasted
The results obtained, were similar to the results that with temperature (°C) at the frequency of 10 rad
found by Al-Mansob (2014) when Epoxidized Nat- (1.59 Hz) are presented in Figure 3. It was found
ural Rubber (ENR) polymer with concentrations of that the G* increased with addition increase
3, 6, 9 and 12% used to modified bitumen binders. of modifier up to 5% then decreased slightly
It noted that the addition of polymer increases the in 7%, but it’s still enhanced compared with the
viscosity values (Al-Mansob et al. 2014). base bitumen in terms of performance. In addi-
tion, the most obvious improvement in G* was
for 5% GPMB, thus leading to the most strongly
improved of temperature susceptibility among all
investigated modified binders. Several studies were
56.5
60 53 conducted to evaluate the viscoelastic functions
!-"
on ....., using different polymers such as Styrene–Butadi-
ene–Styrene (SBS), Natural Rubber latex (NR),
and Epoxidized Natural Rubber (ENR). All these
G0 polymers show similar an increase in G* values at
~ 40 high temperatures which represent an improve-
~ ment in terms of the temperature susceptibility
~ 30 (Airey 2003).
E
QJ
f-
20
4.3.2 Rheological master curves
In The master curves reflect the time dependency
of bitumens over a wide range of loading times a
10 reference temperature was selected first, and then
the data at all other temperatures are shifted hori-
0 zontally to construct a single smooth curve. The
O%GPMB 3%GPM B 5%GPM B 7%GPMB master curves of the complex modulus and phase
angle for base bitumen binder and modified bind-
Modifier Content (%)
ers with geopolymer are shown in Figures 4 and
5. It was observed that an increase in the com-
Figure 1. Softening point of base binder and GPMB. plex modulus upon increasing the geopolymer

419
 100000
The phase angle master curves for the base
- - O%GPMB
binder and modified binders which shown in
- - 3%GPMB
Figure 5, indicate that a reduction in phase angle
with increasing the modifier content at tempera-
~ 10000 tures between 46–64°C, while it was increased
.-- above mentioned temperatures due to modifier
reacted as filler as the modified binders reach the
8
"' failure temperatures. Furthermore, it was shown
.@ 1000 that the 5% has a lower phase angle than the base
0
E binders and other modified binders, which means
] it has better viscoelastic recovery performance.
"'-
§ 100
u 4.3.3 Effects of GPMB on failure temperature
Figure 6 shows the results of failure temperature
obtained from frequency sweep tests. As expected,
10
95
e O%GPMB

Temperature (°C) e 3%GPMB

e S%GPMB
90

Figure 3. Isochronal plots of the G* for base binder and

GPMB binders. s 85
€
"<=
on
1000000
..
• ..t.
1;l 80

~ 100000
c ••
•• ••
...- • 75
!2, 10000
"'
-g 70
E 1000 -3.5 ·2.5 -1.5 -0.5 0.5 1.5
to
Log reduced frequency (Hz)
-iii"
~ 100
Q.
~ Figure 5. Phase angle master curves for base binder
i5 and modified GPMB binders.
u
10

80
71 69
-3.5 -2.5 -1.5 -0.5 0.5 1.5 70 v 67
m
Log reduce frequency (Hz) G
~ v
~q
_ ;'::-
60 ~
e
Figure 4. Complex modulus master curves for base ~ so v ·;:::::::
::::::::::
- .,,; -

binder and modified GPMB. B

e
Q.

40 v :::::::::
- ;; -

_::;
30 v ::::::::::
- ::: -
:'<·

concentration in bitumen matrix. This is due to

the increase in stiffness of modified binders which
'"
"-
20 v t:{ - !-
lead to mitigated and prevented the high-temper- 10 v .@:~ - :1 -
ature pavement distresses such as rutting. On the
other hand, the concentration of 7% shows a dif- 0 v ·........
::· - ::: -----;?
0% GPMB 3 % GPMB 5 % GPMB 7% GPMB
ferent behaviour, as the complex modulus reduced
slightly. This reduction, it might be due to chemical Modifier Content (%)
reactions with addition increase of modifier con-
tent, also it could be due the agglomeration occurs Figure 6. Failure temperature of base binder and
among the geopolymer components. GPMB binders.

420
it was found that the base binder has the lowest of bitumen at high temperatures. Nonetheless, the
failure temperature nearly 59°C, while the modi- 7% display different behavior as the resistance to
fied binders have better failure temperatures com- permanent deformation slightly declined. Accord-
pared with base asphalt binder. This indicates that ing to SPS and ASP parameter, the evaluation
the modifier able to enhance the high-temperature and comparison of rutting parameter of the base
properties of bitumen, the highest improvement of binder and modified binders using both SPS and
failure temperature as shown if Figure 5 was for 5% the ASP methods revealed that there were no much
GPMB with an enhancement of 39.22% followed differences between the both techniques, this indi-
by 7% GPMB and 3% GPMB with an improve- cates that both methods are able to use to estimate
ment of 35.19 and 31.37%, respectively. However, and determine the rutting parameters of base
previous studies conducted on various types of modified bitumens.
polymers modified bitumen such as crumb rubber,
polyphosphoric acid, and propylene maleic anhy-
dride show similar performance that the modified 5 CONCLUSION
binders have slightly higher fail temperatures than
the base binder. Number In this paper, high-temperature rheo-
logical properties of bitumen modified with
4.3.4 Examples effects of GPMB on rutting geopolymer were investigated using various
parameter tastings and techniques. The amounts of 3, 5,
Figure 7 shows the graphs of rutting for base and 7% of the modifier were added to the base
binder and geopolymer modified binders using asphalt cement. Based on the test outcomes of
both techniques Superpave Specification (SPS) this chapter, the softening point and viscosity
G*/sinδ and Aroon Shenoy suggested parameter were approved that the hardness of modified
G*/(1−(1/tanδ sinδ)) when δ>52 (Shenoy 2001). binders increased, thus resulting in improved
From Figure 7, it was observed that the similar temperature susceptibility for both modifiers.
improvement to resist the permanent deformation Moreover, the master curves results indicated an
obtained from the previous testings, the lowest rut- increase in elastic behavior compared with the
ting resistance was for the base binder, while modi- base binder and from the failure temperature
fied binders show an enhancement and the 5% tests, it can be observed that all modified blends
GPMB has the highest rutting resistance among had slightly greater failure temperatures in com-
the modified binders, which performs much better parison to the base binder. In addition, accord-
than other modified binders. The results approve ing to the results of the rutting parameters, it can
that the modifier capable to increases the hardiness be concluded that the uses geopolymer as bitu-
men modifier were able to improve the resistance
against rutting at high temperatures. Generally,
:!:I 70"C SPS • · 70 "C ASP • 76 "C SPS • 76"C ASP the best result noted when the addition of modi-
fiers was the 5%.
1.2
1.09d "12
• REFERENCES
~
1 0.91
0.8

..••
~ ~; ~~ Airey, G. D. 2003. Rheological Properties of Styrene
60.8 0.73~0.76 ~; - Butadiene Styrene Polymer Modified Road Bitumens.
~;
••
Fuel 82(14): 1709–1719.
Oil !
. ' S1 • Al-Mansob, R. A., Ismail, A., Alduri, A. N., Azhari,
·:B 0.6 ~
.•
~•0.4'3.4
.'::!..;_:JI
C. H., Karim, M. R. & Yusoff, N. I. M. 2014. Physi-
•• 410.41
:::l

"' •
: . ~
6 cal and Rheological Properties of Epoxidized Natural
fi.
-
0.4 0.3 0.31 •• - 1-
H
Rubber Modified Bitumens. Construction and Build-
:-:
• ••
:.1 ••• -
..••• • -
ing Materials 63: 242–248.
Chien Yet, T., R. Hamid, and M. Kasmuri. 2012.

...:••
f-
H
0.2

0
fin •
••t -:
1---, 1- ••~ 7
Dynamic Stress-Strain Behaviour of Steel Fiber Rein-
forced High Performance Concrete with Fly Ash.
Adv. Civ. Eng. 2012: 1–6.
Fuentes-Audén, C., Sandoval, J. A., Jerez, A., Navarro,
0% GPMB 3% GPMB S% GPMB 7% GPMB F. J., Martínez-Boza, F. J., Partal, P. & Gallegos, C.
Modifier Content(%) 2008. Evaluation of Thermal and Mechanical Prop-
erties of Recycled Polyethylene Modified Bitumen.
Polymer Testing 27(8): 1005–1012.
Figure 7. Rutting parameter for base binder and Ibrahim, A. N. H., Ahmad, A. S., Akhir, N. M. D. &
GPMB binders. Borhan, M. N. 2016. Performance Evaluation of

421
 Stone Mastic Asphalt (Sma) Using Geopolymer as an Saoula, S., Mokhtar, K. A., Haddadi, H. & Ghorbel,
Asphalt Modifier Jordan Journal of Civil Engineering E. 2008. Improvement of the Stability of a Modified
10(4): 442–452. Bituminous Binders within Eva. International Journal
Ibrahim, A. N. H., Yusoff, N. I. M., Mohd Akhir, N. & of Applied Engineering Research 3(4): 575–584.
Borhan, M. N. 2016. Physical Properties and Storage Shenoy, A. 2001. Refinement of the Superpave Speci-
Stability of Geopolymer Modified Asphalt Binder. fication Parameter for Performance Grading of
Jurnal Teknologi 78(7–2): 133–138. Asphalt. Journal of transportation engineering
Liu, S., Cao, W., Fang, J. & Shang, S. 2009. Variance 127(5): 357–362.
Analysis and Performance Evaluation of Different Walubita, L. F., Alvarez, A. E. & Simate, G. S. 2011.
Crumb Rubber Modified (Crm) Asphalt. Construc- Evaluating and Comparing Different Methods and
tion and Building Materials 23(7): 2701–2708. Models for Generating Relaxation Modulus Master-
Peralta, J., Hugo Mrd Silva, and Ana V. Machado. 2010. Curves for Asphalt Mixes. Construction and Building
High Performance Asphalt Pavements Incorporating Materials 25(5): 2619–2626.
Recycled Polymers as Aggregates and Binder Modi-
fier.” Technology 77: 192–200.

422
Effect of traffic loading
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Effects of super heavy trucks on the condition of road infrastructure

P. Kolisoja
Tampere University of Technology, Tampere, Finland

T. Saarenketo & P. Varin

Road scanners Ltd, Rovaniemi, Finland

ABSTRACT: A new legislation increasing the allowable maximum total truck weights from 600 to
760 kN was launched in Finland on October 1st 2013. On certain important transportation routes maxi-
mum truck weights even exceeding 1 MN (100 tons) are allowed with special permission. While this has
enabled positive development in transportation efficiency and the amount of traffic induced emissions,
questions have also arisen regarding traffic safety and especially the loading effect of super heavy trucks
on the road infrastructure. This paper gives a brief overview of the research projects that have been car-
ried out on the latter topic in Finland from summer 2013 to summer 2016.

1 INTRODUCTION previously to move the same amount of goods

from one place to another, a lesser amount of
Finland is a sparsely populated country located far emissions would be produced. At the same time,
from the international markets of the goods pro- the main concerns related to the introduction of
duced by the main fields of the export industries, the new super heavy trucks have been their poten-
especially forestry. Therefore, the international tially adverse effects on traffic safety and especially
competitiveness of the whole economy is strongly on the condition of road infrastructure. Regarding
dependent on the efficiency of heavy transports the latter issue it is important to note that currently
moving the raw-materials and components to pro- the allowable truck weights in Finland are consid-
duction facilities and then the final products to erably higher than in the rest of the Europe while
export harbors. This is the background why new the thicknesses of asphalt pavements are among the
legislation increasing the allowable maximum total thinnest. Furthermore, Finnish roads are exposed
truck weights from 600 up to 760 kN on all roads annually to the detrimental effects of seasonal frost
and with special permission maximum weights which occurs on most of the roads and is known to
exceeding even 1 MN on certain important trans- have a dominant role in the deterioration of road
portation routes was launched in Finland on Octo- structures (Saarenkero & Aho 2005). This is the
ber 1st, 2013. In a heavy truck weighing 760 kN problem setting behind a series of research projects
there must be nine axles while the maximum weight that are briefly summarized in this paper.
of an eight axle truck is 680 kN. For a transition
period of five years the allowable total weight
of existing trucks was increased from 600 kN to 2 POTENTIAL DETERIORATION
640 kN (independent of the tire type they use) MECHANISMS RELATED TO SUPER
which meant also an increase in axle loads. Oth- HEAVY TRUCKS
erwise the allowable axle loads were not increased,
but the maximum load applied on the road struc- In the Northern climatic conditions of Finland
ture via e.g. a group of three axles (triple bogie) typical road structures have a thin Asphalt Con-
was rising from 240 kN to 270 kN. Nevertheless, crete (AC) surfacing with a thickness ranging from
the main difference to the earlier practice is that about 50 mm on roads with low traffic volumes
the number of axles on a heavy truck unit is now up to 150…200 mm on the most heavily trafficked
higher than before. main roads. At the same time, relatively thick lay-
In addition to the positive effect on transport ers of coarse grained Unbound Granular Materials
efficiency an additional benefit was also expected (UGM) are used below the asphalt concrete both
with regard to the amount of traffic induced emis- to increase the bearing capacity of the road struc-
sions into the atmosphere since, a smaller amount ture and to make it more resistant to the harmful
of truck kilometers would need to be driven than effects of seasonal frost. The loading effects of the

425
super heavy trucks potentially resulting in acceler- of permanent deformations occurring in the sub
ated deterioration of this type of road structures grade (deterioration mechanism 4 above) also
can be listed as follows: increases. In the Nordic climatic conditions, by far
the most critical period from this point of view is
1. Permanent deformations in the asphalt concrete
late thawing period of seasonal frost, when there is
layer
still essentially impermeable frozen soil left deeper
2. Fatigue of the asphalt concrete layer
below the sub grade surface efficiently preventing
3. Permanent deformations in the unbound base
the upper layers to draining from the melting water
course layer, so-called Mode 1 rutting (Dawson
of snow and ice. According to many studies (e.g.
& Kolisoja 2004, Kolisoja et al. 2013)
Saarenketo and Aho 2005), a major part of the
4. Permanent deformations in the subgrade below
structural damages developing in a road structure
a too thin road structure, the so-called Mode
throughout the whole year may take place during
2 rutting (Dawson & Kolisoja 2004, Kolisoja
those few weeks in the spring.
et al. 2013)
The precise separation of rutting caused by the
5. Pumping of water from wet sub grade soil into
super heavy trucks from that caused by other vehi-
the road structure resulting in wetting of the
cles is practically impossible, but when heavier axle
unbound layers and consequent development
groups are used the combined loading effect caused
of excess pore water pressure under cyclic traf-
by them on the sub grade level is inevitably higher
fic loading
and therefore also accelerating the development of
6. Excessive loading on various types of spe-
sub grade rutting. In connection with the studies
cial structures (e.g. bridges and culverts) and
carried out in Finland since the introduction of the
increased dynamic loads in connection with
new legislation, a number of roads were selected
other types of local discontinuities along the
for frequent monitoring of rut depths and in time
road (bumps etc.)
more information concerning the rut development
Concerning the permanent deformations in the rates will be obtained. Even though the monitoring
AC layer (deterioration mechanism 1 above) the period is still too short for making any final con-
potentially adverse effect of super heavy trucks is clusions some examples of the monitoring results
that as a single heavy truck has more axles, the wheel will be given in Chapter 5.
path wander will decrease and also the rest periods in A further sub grade related problem is faced
between the consecutive axles will be shorter. Until when the road structure rests on a very wet sub-
now this effect has not been studied in Finland. grade soil with low stiffness, e.g. peat, in which
Regarding the heavy truck loads the most impor- consecutive heavy truck axle loads passing over
tant effect on the fatigue of the AC layer and rutting the site cause large deflections in the sub grade soil
of the unbound base course layer (deterioration and squeeze excessive water upwards into the road
mechanisms 2 and 3 above) is related to the types of structure, i.e. pumping (deterioration mechanism
tire configurations used on the super heavy trucks. 5 above). Just recently there has been some field
The relevance of this issue has increased during measurements carried out on this topic in Finland
recent years since, on the trucks operating for long and the results indicate the importance of that phe-
distances, there has been a rapid transition from nomenon at least on peat sub grade, but the results
dual tires to single tires. From a road distress point have not yet been analyzed thoroughly. Analogous
of view this development has clearly been negative risk assessments have been made earlier in Sweden
since it is well-known since the AASHO road tests where especially the transition zones from peat sub
performed in the USA already 50 years ago that the grades to stiffer subgrade soils have been located to
more concentrated loading effect the tire configu- be the spots going to be damaged first on the routes
ration has, the more detrimental it is to the upper utilized by heavy trucks transporting iron ore
part of road structure. Since the new legislation, (Saarenketo at. al. 2012, Saarenketo et. al. 2013).
launched on October 1, 2013, included a require- One more completed study related to the super
ment that at least 65% of the super heavy truck heavy trucks was focused on the distresses expe-
trailer mass must be transmitted onto the road rienced by steel and plastic culverts installed at a
surface via dual wheels, this issue became an actual shallow depth below the road surface. The results
topic of research in Finland as well. The results of of this study are briefly summarized in Chapter 4.
those studies are summarized in Chapter 3.
Interaction between the sub grade soil and the
road structure on top of it becomes more critical 3 EFFECTS OF HEAVY TRUCK TIRE
when the road embankment gets thinner, i.e. the CONFIGURATION
stresses to which the sub grade soil is exposed to
are increasing, and more so when the sub grade The effect of tire configuration on the distresses
soil is weak and wet. This means that the likelihood heavy trucks impose on road structures was studied

426
on two test sites. One of the sites was selected to rep- the high quality road structure the loading equiva-
resent a high quality road consisting of sections with lency factor calculated for the narrowest type of
an AC layer thicknesses of 120 mm and 200 mm single tire (385/65R22.5) was 1.8 for the 200 mm
both underlain by a 150 mm thick unbound base AC layer thickness and 2.7 for the 120 mm AC layer
layer and 400 mm thick sub-base layer resting on thickness (Pienimäki & Pihlajamäki, unpubl.).
top a well-drained sandy subgrade. The other test On the test site representing low volume road
site, selected to represent a road with low traffic vol- conditions the differences between tire types were
umes had a 40 mm soft asphalt concrete surfacing observed to be even greater. In Table 1 there is a
on top of granular base and sub-base layers having summary comparing the results of response meas-
a combined thickness of about 450 mm. That site urements carried out and interpreted as normal-
was located in a silty sub grade soil area (Figure 1). ized values with respect to those obtained under
The truck loading was accomplished on both of the normal type of dual wheel configuration. As
the test sites by using a combination of a three- Table 1 indicates the responses measured under the
axle-truck and a two-axle-trailer. The axle loads of narrow single wheels are up to 30% higher than
the truck were 74 kN on the front axle and 152 kN those detected under a dual wheel configuration,
on the double bogie while the front axle of the but as single tire width increases the differences in
trailer had an axle load of 100 kN and the rear axle the measured responses are diminishing. At this
80 kN. The axle loads were kept constant through- point it is, however, important to note that the rec-
out the measurements but the tire types used in the ommended tire inflation pressures in the modern
trailer were varied. In addition to a normal type wide-base single tires are clearly lower than those
of dual wheel configuration (2 × 275/70R22.5) four of the old generation narrow single tires, so the
different single tire types with nominal widths var- observed differences between the various single tire
ying from 385 mm to 495 mm were used. types are not only due to tire width but very much
On the test site representing a high quality road also due to the different inflation pressures.
the main focus of the measurements was on fatigue When the observed differences in the measured
of the AC layer and that is why the site was instru- responses of Table 1 are interpreted as respective
mented heavily with strain transducers installed differences in the loading equivalency factor accord-
in the AC layer. In addition to these, there were ing the existing damage model with regard to base
pressure cells at three levels below the AC layer course rutting given in the existing Finnish design
(unbound base, sub-base and sub grade levels). guidelines (Tiehallinto 2004), the result is as shown
Correspondingly, on the low volume road test site
pressure cells had been installed at three different Table 1. A summary of the response measurements
levels (120 mm, 230 mm and 390 mm below the carried out on the low volume road test site in Vesilahti.
road surface) and deflections of the road surface The results are indicated as relative values with respect
were measured using LVDTs installed below the to those obtained when 100 kN standard axle on dual
road surface and anchored with streel rods into a wheels was used.
stiff stratum a few meters below the surface.
The conclusions that could be drawn from the Vertical
measurements conducted at both of the test sites Axle pressure at the
load Surface depth of (mm)
were very much in line with the results obtained ear- deflect-
lier by others (e.g. Al-Kadi & Wang 2009, Greene at Tire type used kN tion 120 230 390
al. 2010). Regarding the fatigue of the AC layer on
D 2×275/70R22.5 100 ±0 ±0 ±0 ±0
S 385/65R22.5 100 +13 +20 +20 +18
S 425/65R22.5 100 +26 +25 +33 +28
S 455/45R22.5 100 +13 +11 +15 +16
S 495/45R22.5 100 +2 +2 +3 +4
D 2×275/70R22.5 80 −22 −6 −13 −17
S 385/65R22.5 80 +15 +19 +20 +6
S 425/65R22.5 80 +8 +9 +11 +3
S 455/45R22.5 80 −7 −4 −9 −9
S 495/45R22.5 80 −17 −17 −18 −17
D 2×315/60R22.51 76 −12 −5 −12 −20
S 385/55R22.52 74 +7 +12 +8 +2

D = dual wheel, S = single wheel.

1
Figure 1. A view of the low volume road test site located First axle of the truck triple bogie.
2
in Vesilahti, about 30 km South of Tampere. Front wheel of the truck.

427
 4, 5
Dua l Single t ires
4,0 Truck

~
tires f ront
3,5
.!! i- axle
,. 3,0
f-
~ 2, 5
f-
~ 2,0
-~ 1,5 i---

]
r- I .I
!-.
1,0
r-L I II
•
0,5
0,0 I I

Tire type

• 100 kN axle load • 80 kN axle load

Figure 3. Loading vehicle used in the culvert response
Figure 2. Load equivalency factors of the different tire measurements.
types and loads based on the response measurements
performed at the Vesilahti test site.
One concern the owner of the Finnish road net-
work, the Finnish Transport Agency, had regard-
in Figure 2. Based on the results it seems that the ing the super heavy trucks was the loading effects
damaging effect of narrow single tires can be up to they have on culverts installed at a shallow depth
fourfold with respect to dual tire configuration when below the road surface. The studies carried out
the same axle load is considered. In more detail, since then were focused on steel and plastic cul-
the results of in-situ measurements concerning the verts with varying age and condition since concrete
loading effects of different tire configurations and culverts were considered to be in any case rigid
respective numerical modelling are presented else- enough to withstand the applied truck loads.
where (Kolisoja et al 2015, Kalliainen & Kolisoja In summer 2014 measurements were done on
2015). four steel culvert and four plastic culvert instal-
An evident conclusion from both of the meas- lation sites in which the culvert diameters varied
urement series briefly summarized in this chapter from 600 mm to 1800 mm and the installation
was that the requirement enforcing that at least depths from 200 mm to 900 mm. Since all the
65% of the super heavy truck trailer weigh must be culverts were newish, they were considered to be
resting on dual wheels is well justified and thus it in good condition. The installed instrumenta-
was kept in force even though there was quite a lot tion enabled monitoring of the changes in culvert
of lobbying against it right after the introduction diameter both in horizontal and vertical directions
of the new legislation. as well as strains of the culvert inner wall while a
truck with different axle loads and tire configura-
tions was passing over the culvert installation site
4 EFFECTS ON BRIDGE AND CULVERT (Figure 3). In summer 2015 the measurements were
DISTRESSES continued on three more culvert sites but now the
focus was on older steel culverts that had already
With the launch of new regulations for allowable suffered from damages caused by corrosion. In
truck weights in 2013 one of the first consequences addition to the measurement of culvert distresses
was that temporary weight restrictions were set on as such the study was at that stage aiming to test
a number of existing bridges along the national and verify functionality of different types of repair
road network as the knowledge regarding their techniques applicable for corroded steel culverts.
bearing capacity was not adequate. In the very Regarding the loading effects of heavy trucks
beginning the number of weight restricted bridges on steel and plastic culverts a fairly unambiguous
counted above 300, but as more detailed analysis result from both of the culvers studies was that,
on the bearing capacities have been carried out as far as the culverts are installed at a minimum
the number has been decreasing. Even so, a large depth of culvert diameter or 500 mm whichever
number of existing bridges need to be strengthened is greater, the total truck weights or the tire con-
and the cost of these works on the public road net- figuration do not have any marked effect on the
work alone was estimated to exceed 200 MEUR distresses the culverts are experiencing. Especially
(Ministry of Transport and Communications at shallow installation depths and with soft sub-
2013). In addition to that, the costs for strengthen- grade soil conditions special attention should,
ing a number of bridges located on private roads however, be given to the quality and compaction
and city streets must be considered. degree of the granular materials surrounding the

428
culvert. For more details, the studies conducted Central Finland. The lower part of Figure 4 indi-
regarding the loading effects on steel and plastic cates a plane view of the measured rut depths on
culverts have been reported elsewhere (Kolisoja & both of the lanes (Dir1 and Dir2) in color scale
Kalliainen 2016 and Haakana et al. 2015). ranging from 0 mm (white) to 28 mm (black).
At the same time, a vertical Ground Penetrating
Radar (GPR) profile of the same road section
5 EFFECTS ON RUT DEVELOPMENT indicated in the upper part of figure clearly reveals
the places where steel net reinforcement has been
As already pointed out in Chapter 2 it is not easy installed in the unbound base layer of the road.
to separate the effect of super heavy trucks from These sections have been surrounded with rectan-
the total amount of permanent deformations, i.e. gles in Figure 4.
rutting, of a certain road section. This is especially One very interesting remark concerning
the case over the short term. Over a longer times- Figure 4 is that in sections were steel net reinforce-
cale at least qualitative information can, however, ments have been installed into the unbound base
be obtained when the rut development rates of course layer of the road, the rut development rate
roads now being exposed to frequent super heavy has been very much slower than in the sections
truck loads can be compared to the respective rut without steel net reinforcement. Even though the
development rates of the same roads prior to the steel net reinforcements have in this case prima-
year 2013 and to those not on the transportation rily been installed for preventing the longitudinal
routes frequently used by the super heavy trucks. cracking of AC layer as a result of uneven frost
Even though definitive conclusions concern- heave, the reinforcement seems to also have been
ing the effects on rut development rate cannot be very efficient in slowing down the development of
drawn yet, some fairly interesting observations permanent deformations. Especially so because
can already be made based on the results obtained the steel nets have most likely been installed in
from road sections that are frequently monitored places were frost related problems have been the
e.g. in connection with the PEHKO-project in two severest and thus also the potential for thawing
maintenance contract areas located in Central and time rut development greater than on the neigh
Northern Finland (Tapio et al. 2016). In those boring sections without reinforcement. Physically
project areas, modern laser scanning technology steel nets act in a road structure in principle in the
has been utilized by Road scanners Ltd in making same way as reinforcements in a concrete slab as
frequent surveys of the road surface profile. they increase the bending stiffness of the upper
In Figure 4 there is an example of the point part of road structure and thus reduce deflections
cloud model derived from a mobile laser scanning both on the road surface and at sub grade soil level
survey of a section of road number 58 in Multia, (Kolisoja et al. 2002). As the deflections decrease,

Figure 4. Point cloud model of a section of road number 58.

429
the risk for permanent deformation development visually from normal longitudinal cracks caused
and pumping of water from the sub grade soil both by uneven frost heave both by their location and
decrease. especially by their straight appearance.
Another important observation made in connec-
tion with the rut development monitoring of the
PEHKO maintenance contract areas is frequent 6 ECONOMIC CONSIDERATIONS
appearance of a new type of cracking mechanism
called here ‘Mode 2 cracking’. It seems to appear At the time of introduction of the new legislation
typically on roads that have a fairly thin AC surfac- the economic benefits to be obtained in the form
ing and that suffer from severe sub grade rutting of reduced transportation costs were expected
(Mode 2 rutting). The longitudinal crack develops to be up to 160 MEUR by the Finnish Ministry
in between the wheel paths in the middle of a lane of Transport and Communications (2013) and
due to too high tensional forces straining the AC by the Confederation of Finnish Industries even
layer as schematically illustrated in Figure 5 and higher than that. A very recent study conducted
exemplified in Figure 6. This cracks differentiate by Liimatainen & Nykänen (2016) has shown that
these numbers were, however, too optimistic and
the actual savings in transportation costs seem to
be stabilizing at about 60 MER per year or slightly
less. As an additional benefit on that comes of
course the reduction in emissions of CO2 which
according to the same study are calculated to be
about 3.7% of the total emissions caused by trucks.
Due to the slow nature of road deterioration
Figure 5. A schematic picture of the ‘Mode 2 cracking’ mechanisms it is at least for the time being not pos-
mechanism. sible to make a definite estimate on the additional
annual road maintenance costs caused be the intro-
duction of the new super heavy trucks. One very
rough calculation exercise can, however, be made
for instance by comparing the achieved savings
in the transportation costs to the typical costs of
road rehabilitation even though the beneficiaries
and cost payers are, in this case, not the same par-
ties. If the costs including drainage improvement,
strengthening measures targeted on the weakest
spots of the road and putting a new overlay are
assumed to be of the order of 0.1 MEUR/km, the
achieved cost savings in transportation costs cor-
respond to the rehabilitation cost of about 500 to
600 km of paved road. When this figure is com-
pared to the total length of the Finnish public road
network, about 80 000 km, of which roughly two
thirds are paved roads, it can be observed that it
corresponds to about 1% of the total length of
paved roads. In other words, it means that rehabili-
tation works needed to compensate for the acceler-
ated road deterioration caused by the super heavy
trucks, if used directly to road maintenance, could
be made once in a hundred years for a certain road
section. Since in addition to maintenance costs
of paved roads comes, of course, the additional
costs caused by the need to strengthen bridges
and increased maintenance costs of public gravel
roads, private roads and city streets, it is highly
questionable whether the introduction of the new
super heavy truck weights was a profitable decision
from a national economy point of view. More light
Figure 6. Two examples of ‘Mode 2 cracking’ appear- on that question will, however, be gained when
ing on roads with thin AC surface layer. actual observations on the possible changes in the

430
deterioration rate of the road infrastructure have under heavy truck loads), Finnish Transport Agency,
been collected on a longer timescale. Report 18/2015. 72 p. (In Finnish)
Kalliainen, A. & Kolisoja, P. 2015. Effect of tire configu-
ration on the performance of a Low Volume Road
7 CONCLUSIONS exposed to heavy axle loads—mechanical modelling,
Transportation Research Record 2474, pp. 174–184.
The main conclusions that can be drawn based on Kolisoja, P. & Kalliainen, A. 2016. Modelling of Plas-
the studies carried out so far regarding the loading tic Culvert and Road Embankment Interaction in
effects of the super heavy trucks introduced by the 3D. Proceeding of the 3rd International Conference on
Transportation Geo-technics, Guimaraes. 8 p.
new legislation that came into force in 2013 are as
Kolisoja, P., Kalliainen, A. & Haakana, V. 2015. Effect
follows: of tire configuration on the performance of a Low
− The rule enforcing that at least 65% of the total Volume Road exposed to heavy axle loads—response
mass of a super heavy truck trailer must be rest- measurements, Transportation Research Record 2474,
pp. 166–173.
ing on dual wheels is definitely appropriate, since
Kolisoja, P., Kalliainen, A. & Vuorimies, N. 2013. Mecha-
the loading effect of old generation single tires nistic Design of Low Volume Road Structures, 9th
on road infrastructure can be up to fourfold in International Conference on the Bearing Capacity of
comparison to the loading effect of dual wheels. Roads, Railways and Airfields, Trondheim. pp. 331–340.
− The loading effect of super heavy trucks is not a Kolisoja, P., Zarghampour, H. & Salmenkaita, S. 2002.
critical issue regarding steel and plastic culverts Use of steel net reinforcement in unbound pavement
since at shallow installation depths the culverts layers. Proceedings of the 6th International Confer-
are experiencing the truck loads basically one ence on the Bearing Capacity of Roads and Airfields,
axle at a time. Lisbon. Vol. 2, pp. 1031–1040.
Liimatainen, H. & Nykänen, L. 2016. Impacts of
− Steel net reinforcement installed in the unbound
increasing maximum truck weight—Case Finland,
base layer seems to have a marked effect in slow- International Colloquium on Green logistics manage-
ing down the rut development rate. ment: balancing environmental and shareholder priori-
− Because the achieved cost savings in transpor- ties. Naples.
tation costs seem to remain quite a lot smaller Ministry of Transportation and Communications. 2013.
than what was expected at the time of launching Memorandum related to the introduction of new
the new legislation, it is highly questionable if legislation concerning allowable trucks weights (In
the obtained benefits will compensate the addi- Finnish).
tional maintenance cost of road infrastructure Pienimäki, J. & Pihlajamäki, J. Virttaa 2013 mittauk-
from a national economy point of view. set, Paripyörien ja leveiden yksittäispyörien vaikutuk-
set tierakenteeseen. (Measurements in Virttaa 2013,
Effects of dual wheels and wide-base single tires on
ACKNOWLDGEMENTS road structure) Unpublished draft raport. (In Finnish)
Saarenketo, T., Matintupa, A., Varin, P., Herronen, T. and
Hiekkalahti, A. Pajala mine road consequence analysis.
Finnish Transport Agency is gratefully acknowl- Unpublished report, Trafikverket, Sweden. 15p.
edged for financing most of the research work pre- Saarenketo, T., Matintupa. A., Varin, P., Kolisoja, P.,
sented in this paper. Herronen, T. & Hiekkalahti, A. 2012. Pajala road
impact analysis. ROADEX IV Project report. 55 p.
Available at: http://www.trafikverket.se/contentassets/
REFERENCES ff1e307734044347afb4ca46d99810ef/pajala_road_
impact_analysis_final_02022012.pdf
Al-Qadi, I.L. & Wang, H. 2009. Evaluation of Pavement Saarenketo, T. and Aho, S. 2005. Managing spring thaw
Damage Due to New Tire Designs, Illinois Department weakening on low volume roads. Problem description,
of Transportation, FHWA-ICT-09-048. 66 p. load restriction policies, monitoring and rehabilitation.
Dawson, A. & Kolisoja, P. 2004. Permanent Deformation, ROADEX II Project report, 127 p. Available at: http://
ROADEX II Project report. 47 p. Available at: http:// www.roadex.org/wp-content/uploads/2014/01/2_3-
www.roadex.org/wp-content/uploads/2014/01/2_1- Spring_Thaw_Weakening_l.pdf
Permanent-Deformation_l.pdf Tapio, R., Lehtinen, J., Ylinampa, J. and Saarenketo, T.
Greene, J., Toros, U., Kim, S., Byron, T. & Choubane, B. 2016. PEHKO project 2015–2025, Increasing the pro-
2010. Impact of Wide-Base Single Tires on Pavement ductivity of paved road management in Finland. Pro-
Damage. Transportation Research Record 2155, TRB, ceeding of the EAPA Conference, Prage. DOI: dx.doi.
Washington, D.C. pp. 82–90. org/10.14311/EE.2016.144
Haakana, V., Kalliainen, A. & Kolisoja, P. 2015. Raskai- Tiehallinto. 2004. Tierakenteen suunnittelu (Structural
sta ajoneuvoista tierumpuihin kohdistuvat rasitukset Design of Road Structures), Finnish Road Adminis-
(Deformation behaviour of steel and plastic culvers tration, Helsinki. (In Finnish). 69 p.

431
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Dynamic pavement response coefficient to estimate the impact of

variation in dynamic vehicle load

Seyed-Farzan Kazemi, Peter E. Sebaaly, Raj V. Siddharthan, Elie Y. Hajj,

Adam J.T. Hand & Md. Ahsanuzzaman
University of Nevada Reno, Reno, USA

ABSTRACT: This study evaluated the influence of roughness-induced dynamic truck axle loading on
the variation in pavement responses. Measured dynamic tire load data at 40, 60, and 80 km/h (24.9,
37.3, and 49.7 mph) for air and rubber truck suspensions from five sites (smooth to severely rough) were
investigated. A total of 840 individual 3D-Move analyses were conducted to estimate responses of two
different pavement structures under a moving dual tire tandem axle at multiple locations within pavement
structures. A new parameter called Dynamic Response Coefficient (DRC) was introduced to represent the
variation in pavement responses due to the variation in dynamic truck axle loading. DRC is the coefficient
of variation for a specific response at a specific point within the pavement structure. The results showed
that DRC differs from the Dynamic Load Coefficient (DLC) and is a function of location within pave-
ment, pavement structure and temperature, suspension type, vehicle speed, and road roughness.

1 INTRODUCTION sider the influence of the additional dynamic load

component at the tire-pavement interface. Thus, it
Tire-pavement interaction under a heavy vehicle is essential to account for the variation in pavement
load is quite complex where the loaded area is responses due to the variation in dynamic loading
typically noncircular, with non-uniform normal for realistic prediction of pavement performance.
stress distribution. In addition to vertical stresses, Past studies (LeBlanc and Woodrooffe, 1995;
the tire loading may be accompanied with a con- Siddharthan et al., 1998; Sweatman, 1983;
tact shear stress component that might be sub- Woodrooffe et al., 1986) have quantified tire load
stantial especially under vehicle braking conditions variation (or perturbation) caused by pavement
or while traveling on a steep grade. The applied roughness using field measurements and analytical
vertical force at the tire-pavement interface can be dynamic vehicle-road interaction models. Several
divided into two components: the static load and researches have used the Dynamic Load Coeffi-
a fluctuating component known as the dynamic cient (DLC), which is defined as the coefficient of
vertical force. The applied static load depends on variation of dynamic load (i.e., standard deviation
the geometry and mass distribution of the vehicle divided by the mean tire load) (Gillespie, 1993; Sid-
as well as the static load sharing characteristics of dharthan et al., 1998; Streit et al., 1994). The DLC
the suspension system. The dynamic load, on the is strongly dependent on pavement roughness, vehi-
other hand, is caused by vibration of the vehicle cle speed, truck axle types, and suspension system
when it is excited by the roughness of the pavement (LeBlanc and Woodrooffe, 1995; Sweatman, 1983;
surface. The dynamic loads generate additional Woodrooffe et al., 1986). For instance, regression
pavement dynamic responses (stresses, strains, and equations were suggested to estimate DLC for
displacements) which are thought to influence the tandem and tridem axles for different suspension
pavement deterioration rate. Therefore, analyzing types and road roughness levels (Sweatman, 1983).
pavements under dynamic loads is important in Siddharthan et al. (1998) suggested use of DLC
order to properly estimate load-induced damage to estimate the variation in pavement responses
over a pavement life. histories due to variation in tire dynamic loading
Mechanistic-Empirical (ME) pavement design traversing a pavement. The approach consisted of
and analysis procedures rely on proper estimation first estimating critical pavement responses histo-
of pavement responses to predict Asphalt Concrete ries for a constant mean tire load (i.e., no tire load
(AC) pavement performance and/or distresses. The variability) and subsequently using Equation 1 to
critical pavement responses might be underesti- estimate variation of the response time-histories
mated in a mechanistic analysis that does not con- that account for tire load variability:

433
 (t0 ) + R (t0 ) × DLC
R(t0 ) = R(t (1) pavement structure, pavement temperature, road
roughness, vehicle speed, and suspension type.
where R(t0) is the desired pavement response at time To account for the influence of moving loads on
t0 for the point of interest; R(t0 ) and R( t ) are the viscoelastic behavior of asphalt mixture, the
0
responses under the mean tire load and under the 3D-Move Analysis software was used in this study.
perturbation load, respectively. This procedure is 3D-Move runs were conducted for multiple com-
based on the assembly of a population of responses bination of vehicle speed, pavement structure, sur-
at each instant of time with a normal distribution face dynamic load, and pavement temperature in
about those responses computed with mean tire order to obtain pavement responses for representa-
loads. The variability of pavement response, which tive loads. Then, the coefficient of variation of
predominantly depends on road roughness and the computed responses at the multiple load levels
vehicle speed, is then assumed to be accounted for were represented by DRC.
through the DLC. This approach assumes that the Figure 1 illustrates the approach that was used
DLC can be directly translated to a variation in in this study to estimate DRC values which can be
pavement responses irrespective of pavement struc- implemented in an M-E pavement analysis and
ture, material properties, and pavement temperature. design procedure. As shown in this figure, previ-
Traditional pavement analysis models are not ously measured dynamic loading by Woodrooffe
capable of considering the influence of the tire et al. (1986) for a dual tire tandem axle was used
dynamic loading. Three-dimensional models based at various roughness levels and speeds and for dif-
on finite-element or finite-difference approaches ferent suspension types, thus resulted in a total of
have also been used by other researchers for pave- 30 different cases. The dynamic load frequency
ment analysis (Wang and Al-Qadi, 2009). However, distribution was then derived for each case as out-
these models demand substantial computational put of the first step. In the second step, multiple
effort which makes them impractical for integra- 3D-Move Analysis runs were conducted for the
tion with mechanistic-empirical pavement design various loading cases combining selected dynamic
and analysis procedures. Thus, there is a signifi- load levels with pavement structures and tempera-
cant demand for a reliable and simple method to tures. These load levels were selected at different
quantify the influence of variation in dynamic tire reliability levels of the dynamic load Cumulative
loading on the variation in pavement responses that Distribution Function (CDF). Output of the sec-
can be implemented with mechanistic-empirical ond step of the approach was the maximum tensile
methods.
Siddharthan et al. (1998) developed a continuum- r--------------------------------------1
1 step 1. Dynamic Load Data Collection 1
based finite-layer approach to evaluate pavement L--------------------------------------1
responses under moving traffic loads. The numeri-
Experimental Data for Dynamic Load of a Dual Tire Tandem Axle:
cal model employed two-dimensional Fourier
• Suspension types (Air, Rubber)
transform to solve equilibrium equations in fre- • Roughness (Smooth, Rough, Medium-Rough, Very-Rough, Severely
quency domain. As a result, a stress distribution of Rough)
·Speed (40, 60, and 80 kmlh)
any shape can be represented by this model. Layer
material properties can be characterized as either
elastic or viscoelastic. The 3D-Move Analysis Dynamic Load Frequency Distribution and Normality Test
computer software was developed based upon the
model. The software is capable of applying DLC
to pavement static responses in order to account r---------------~----------------1
1 step 2. Mechanistic Analysis Using 3D-Move 1
for variation in dynamic loading due to pavement ~--------------------------------------
roughness. However, as mentioned earlier, the 840 Individual 3D-Move Analysis Runs for a Dual Tire Tandem Axle:
direct application of DLC to pavement responses ·Two Suspension Types (A'r, R>ltlbool
remains questionable. ·Five Roughness Levels (SmooCIIR<>Jih, l·lodwn&o;h.~'!lf\S..-'Y~)
• Two Pavement Structure · ~
·Three Speed Levels (~. eo. M~801<m<ll)
• Two Pavement Temperature Levels (2t.1'CcrJO'F, 4/J'Cot 1o<'F)
2 RESEARCH APPROACH • Seven Load Levels for Each Dynamic Load Profile
METHODOLOGY

In this study, a new parameter called Dynamic

•
Pavement Responses:
-Tensile Strain (e.} and Vertical Strain (en) at Center, Middle, Outer,
Response Coefficient (DRC) was introduced to and Inner Points of a Dual Tire Tandem Axle and for Various Depths
identify variation in pavement responses due to
the variation in pavement dynamic loading. DRC
takes several factors into account for estimation Figure 1. Illustration of the approach used to estimate
of the variation in pavement responses including DRC.

434
strains (εt) and vertical compressive strains (εzz) at
various locations within the pavement structure
for each of the cases. A response distribution was
~0.9 85% Reliability
()
I
assigned to every case separately for εt and εzz.

3 DYNAMIC LOAD DATA COLLECTION

~0.8
0
'ic1
::J
u. 0.6
c
0.7 65% Reliability I
I
0 50% Reliability
Dynamic tire loads were obtained from an experi- ; 0.5
.D
mental research conducted by Woodrooffe and his 'Een 0.4 35% Reliability J
team in 1986 (Woodrooffe et al., 1986). An instru-
0
mented heavy vehicle developed by the National Gl 0.3
Research Council of Canada (NRCC) had a 5-axle .2::
~ 0.2 15% Reliabilit/
semi-trailer configuration equipped with air suspen- ::J

~
sion on the drive axles and a rubber suspension on ~ 0.1
the trailer axles. Truck instrumentation consisted ()
(jlf! ~,,

of strain gauges installed on the axles between the ~0 40 50 60 70

suspension brackets and wheels responding to axle Dynam ic Load (kN)
bending. Dynamic axle load measurements were
calculated as the product of bending strains mul- Figure 2. Cumulative Distribution Function (CDF) for
tiplied by a calibration factor (Woodrooffe et al., dynamic load of the severely rough site (IRI = 3.17 m/
1986). Dynamic load data were obtained by driving km = 200.85 in./mile), air suspension, and speed of
the same truck on five pavement sites with Inter- 80 km/h (49.7 mph).
national Roughness Index (IRI) ranging from 0.88
to 3.17 m/km (55.8 to 200.9 in./mile) measured by
an inertial profilometer at distance increments of for the case of air suspension and severely rough
15.24 cm (6 in.). Dynamic load data were surveyed at pavement (IRI = 3.17 m/km = 200.85 in./mile) at
three vehicle operating speeds, specifically 40, 60 and a speed of 80 km/h (49.7 mph). The concept of
80 km/h (24.9, 37.3, and 49.7 mph). Table 1 summa- reliability was employed to select the appropri-
rizes International Roughness Index (IRI) and DLC ate load levels for analysis in 3D-Move in order
values for these five sites at different speeds and two to develop CDF for pavement responses. Load
suspension types. The DLC of the rubber are higher levels corresponding to reliability levels of 0%,
than the air suspension at the same speed and IRI 15%, 35%, 50%, 65%, 85%, and 100% were
value. included in 3D-Move analysis. As a result, each
As an example, Figure 2 shows a dynamic dynamic load scenario is represented by seven
load Cumulative Distribution Function (CDF) load levels which will be later used for the input
to 3D-Move Analysis in order to calculate mecha-
nistic pavement responses. For the case presented
Table 1. IRI and DLC values for different sites, suspen- in Figure 2, the load levels corresponding to the
sion types, and operating speeds. selected reliability levels were 33.4, 46.7, 49.7,
51.4, 53.2, 56.4, and 69.8 kN (10.50, 11.17, 11.56,
DLC (%) 11.96, 12.68, and 15.69 kip), respectively.
Air Suspension Rubber Suspension
4 ANALYSIS
Speed (km/h) 40 60 80 40 60 80

Smooth 3.73 3.38 3.52 4.59 5.24 8.00 4.1 Mechanistic analysis using 3D-Move
(0.88 m/km*) 3D-Move Analysis software was used to estimate
MediumRough 4.55 3.92 4.97 6.73 9.40 14.72 critical pavement responses for a variety of cases
(1.37 m/km*) under a dual tire tandem axle load configuration.
Rough 4.98 5.55 6.38 6.88 8.20 15.97 Wheel spacing, axle spacing, and inflation pres-
(1.52 m/km*) sure were 35.56 cm (14 in.), 121.9 cm (48 in.), and
Very Rough 5.72 8.77 9.01 10.40 10.68 22.20
827 kPa (120 psi), respectively. There were a total
(1.82 m/km*)
Severely Rough 7.28 10.16 9.59 15.58 14.36 26.72
of 210 different combinations for dynamic load as
(3.17 m/km*) presented in Table 1. The mechanistic responses
were obtained for two representative temperatures
*International Roughness Index (IRI) for the corre- for AC bottom-up fatigue cracking and rutting,
sponding site in m/km (1 m/km = 63.36 in./mile). as well as two different pavement structures. As

435
a result, a total of 840 individual 3D-Move runs perpendicular to the traffic direction. Since the
were conducted for this study. focus of this study was more on responses within
the HMA layer, more response points were located
4.2 Pavement structure and material in the HMA layer compared to the base and sub
characterization grade layers. The HMA layer was divided into
sub-layers according to the Mechanistic-Empirical
Two pavement structures were evaluated in this Pavement Design Guide (MEPDG) (OLIDID
study: thin and thick. Details about layer type, thick- and Hein, 2004) recommendations. Responses
nesses, and mechanical properties are presented in were found at the middle of each sub-layer. The
Table 2. A typical polymer-modified (PG64-28 NV) top portion of the asphalt layer was divided into
dense-graded Hot-Mix Asphalt (HMA) mixture was two 1.27 cm (0.5 in.) sub-layers. Each subsequent
selected for this evaluation. The viscoelastic property portion was divided into three 2.54 cm (1 in.) sub-
of the HMA were evaluated in terms of dynamic layers. Responses were also obtained at the bottom
modulus and was used as an input in the 3D-Move of the HMA layer for bottom-up fatigue cracking
analyses. The pavement was analyzed at two temper- analysis. Both vertical compressive strain (εzz) and
atures; 21.1°C (70°F) and 40°C (104°F) correspond-
maximum tensile strain (εt) at different locations
ing to representative temperatures for AC fatigue
were determined which are typically used for rut-
cracking and rutting, respectively. Viscous damp-
ting and fatigue cracking analysis, respectively.
ing was incorporated for the HMA layer which was
determined from the measured dynamic modulus
data. A 5% geometric damping was considered for 5 DYNAMIC RESPONSE COEFFICIENT
the crushed aggregate base and subgrade materials. (DRC)

4.3 Pavement response locations Dynamic load as a function of tire location on

pavement P(x) is represented by two components
Pavement responses were evaluated at a total of
as shown in Figure 3. Mathematically,
44 locations for the thin pavement structure, and
a total of 52 locations for the thick pavement. P ( x ) = P ( x ) + P ( x ) (2)
These locations were distributed within and out-
side the loaded areas, forming a grid in the plane  x ) are the average tire load
In which P( x ) and P(
and perturbation load, respectively. These param-
eters have also been used to calculate DLC as fol-
Table 2. Pavement structure and material properties. lows (Siddharthan et al., 1998):
Pavement Structure P ( x )
DLC = (3)
Thin Thick P(x )

Asphalt Concrete :g:

c... P(x)
Thickness 10.16 cm (4 in.) 15.24 cm (6 in.) -g"
Perturbation Load
Binder Type PG64-28 NV PG64-28 NV
s
(.)

Modulus Variable Variable -~ r-~~~~-+--~--~~------~~~~~~~-+

Poisson’s Ratio 0.35 0.35 ~
~
Damping Ratio Variable Variable F

Base
Thickness 15.24 m (6 in.) 25.40 m (10 in.)
Type CAB CAB Tire Location on Pavement, x
Modulus 207 MPa (30 ksi) 207 MPa (30 ksi)
Poisson’s Ratio 0.40 0.40
Damping Ratio 5% 5%

Subgrade Time, t

Thickness infinite infinite

Type – –
Modulus 103 MPa (15 ksi) 103 MPa (15 ksi)
Poisson’s Ratio 0.35 0.35
Damping Ratio 5% 5% Figure 3. Dynamic tire load and dynamic pavement
strain response.

436
 The strain at a specific time step, t0, is also pre- ε ( 0 )
sented in Figure 3. The same approach used in ε( 0 ) ε ( 0 ) + ε ( 0 ) × = ε ( 0 ) ε ( 0 ) × DRC
ε ( 0) (5)
Equation 2 can be used for strain (Siddharthan
et al., 1998):
ε ( 0 ) ε ( 0 ) + ε ( 0 ) (4) Previous studies showed that DLC varies as a
function of road roughness, vehicle speed, and vehi-
where ε ( 0 ) and ε ( 0 ) are the average and pertur- cle suspension system (LeBlanc and Woodrooffe,
bation strain components. A new parameter called 1995; Sweatman, 1983; Woodrooffe et al., 1986).
Dynamic Response Coefficient (DRC) was intro- Additionally, the DRC was found to be a function
duced to explain the perturbation as a function of of pavement structure, pavement temperature, and
the average response as shown in Equation 5. DRC location within the pavement.
of a response is defined as the coefficient of varia- Figures 4 and 5 present response locations and
tion in that pavement response. example DRC values for thin and thick pavements,

I I I I
5.3} - - - - - - - -6~ - - - -- -- - ~-6:9% - -- - ~ 7 .7"10 - - - - - - - - - - - - - - - - - - - - - - - - 1.oo·
I I
I I I
I I I I

6.0~o 1.\~o :F /.4Lfo :F t.s% 10.001"

I

I I I
6.3~ ?Jet· ,. 73% ;. 7.1%
16.00"
Outer Center Inner Middle

Figure 4. Dynamic Response Coefficient (DRC) of a thin pavement for maximum tensile strength at site 1 (smooth),
80 km/h (49.7 mph), 40˚C (104˚F), and air suspension along with pavement response locations.

!I====~~~'=====~~!!!===!!=~=\===~~========~====~ ~l
o·

2.50"
3.50"
5:~ =:::::: f.~ :::::::: ~:g~~ =::: ~ 1[1f.= ::::::::::::::::::::::: 4.00"
I I I ' I

4.8lo - - - - - - - '6t'lr. - - - - - - - - '*-6]% - - - - h .9"10 - - - - - - - - - - - - - - - - - - - - - - - - 6.00"

I I I I

s~f 6.6, :!. 1.o% :1• 7.6%

8.00"

13.00"
5.91Yo 6.8% 1 7.1% 7.0%

5.9~I sJ¥. 5
li 6.89o
I
•I• 6.6% 18.001 "
I

I I I
I
s.<~;dw 59.~ ,;. 8.3%
iii
8.4%
24.00"
Outer Center Inner Middle

Figure 5. Dynamic Response Coefficient (DRC) of a thick pavement for maximum tensile strength at site 1 (smooth),
80 km/h (49.7 mph), 40˚C (104˚F), and air suspension along with pavement response locations.

437
respectively. These figures are based on εt at dif- considered in this study. Specifically, five sites,
ferent response points for air suspension, smooth two suspension types, two pavement structures,
pavement, 80 km/h (49.7 mph) speed, and pavement three speed levels, and two pavement temperatures.
temperature of 40°C (104°F). As shown in these fig- These figures reveal that DRC increased with
ures, DRC changes with depth and transverse direc- increase in roughness level for both εt and εzz.
tion. While DLC is 3.52% for this case (see Table 1), Figure 8 presents εzz values versus εt for the
DRC varied between 0.3–9.5% and 0.5–20.1% in aggregted DRC values under the same conditions.
the HMA layer for thin and thick pavement, respec- This figure reveals that DRC for εzz had greater
tively. In this case, the maximum DRC value was variation compared to DRC for εt.
estimated near the middle of the HMA layer. This
clearly reveals that DRC is a better parameter than 6.2 Variation of DRC with speed
DLC to reflect the variation in a pavement response
due to variation in tire dynamic loading. Figure 9 illustrates DRC values for εt at 60 km/h
(37.3 mph) and 80 km/h (49.7 mph) as a function
of DRC values for εt at 40 km/h (24.9 mph) while
6 RESULTS AND DISCUSSION other conditions (i.e., pavement structure, pavement
temperature, roughness level, and suspension type)
6.1 Variation of DRC with roughness level remained the same. This figure shows that DRC
increased with increase in speed, particularly at
Figures 6 and 7 present the variation of aggregated higher DRC levels. Because higher DRC values are
DRC values for εtand εzz, respectively, as a func- associated with higher road roughness, the effect
tion of road roughness for all the different cases of speed on DRC is more significant for pavements
with greater roughness.
n-~~e
: : ~-gJ~ :_::j::: r
50% -------------- -----+---- -------T------ -------) _________ _____ l____ _
6.3 Effect of suspension type
Figures 10 and 11 compare DRC values for air and
rubber suspension at various levels of roughness
~ I I I I
~
(.)
40%

I I I R
--- - - --- -- -------r---- ----- --- -t----- --- -----1 ----- --- -----f -- ---
225%

~ ::: ::+::! :f: : :[::i : $: :r::r:::§?

0% ____ _L ___________L_ _________ _ -.1_______ ____ ~ ------- - -- ~----
N
N
w
(.)
~
c
150%

75%
"'
,. ,. ,. Jb DRC Values
IRI=0.88 m/km 1.37 m/km 1.52 m/km 1.82 m/km 3.17 m/km
Roughness Level - Line of Equity
0%
Figure 6. Box plot of DRC for εt and different rough- 0% 75% 150% 225%
ness levels. DRC·d

Figure 8. DRC for εzz versus DRC for εt (aggregated

200% ------------ - --------- - -------------- - - - --------- - -- - - - --------f - -- - - values).
T I
I I 80%
.s::::::

!::l
1so% --- --- - -- - ---- ---- -,-- - -- - - - ---T-- - -- - - - ---- - -- - - -- --- - ---~ - -- - -
T I I I I
e 70%
~

w I I I I I ~ 60%
u 1oo% __ ____ l_____ _____ ___j ______ _______ __ ___________ J __ _________ __r-----
1 -gco 50%
~ I I I I I ~ 40%
50%
I
- --: - - - - -r -- -- T
I I I
-- -- ~- --
I
g· ~ 30%
w

0%-4-~ ~ $
0 20%
_j_ _ ~ 10%
c 0%
IRI=0.88 m/km 1.37 m/km 1.52 m/km 1.82 m/km 3.17 m/km
Roughness Level 0% 10% 20% 30% 40% 50% 60% 70% 80%
DRC of Et at 40 km/h
Figure 7. Box plot of DRC for εzz and different rough-
ness levels. Figure 9. Effect of vehicle speed on DRC.

438
 80%
:u
..c 70%
-§ t: 60%
'7 -~ 50%
'!i3 ~ 40%
.....
0 :::l
~ 30%
~en 20%
2i 10%
0%
0% 10% 20% 30% 40% 50% 60% 70% 80%
DRC for et • Air Suspension
Smooth Medium Rough Rough
Very Rough Severely Rough - - - - Line Of Equity

Figure 10. Comparison between DRC of εt for rubber and air suspension.

250%
:u
..c
..c 200%
:::l t: + 0

'7 -~
N
N
t:
Q)
;:: ~ 100%
150% + +

......
+
...... -
0 :::l
-en
(.) 50%
o=::
c
0%
0% 50% 100% 150% 200% 250%
DRC for ezz • Air Suspension
Smooth Medium Rough Rough
Very Rough Severely Rough - - - - Line of Equity

Figure 11. Comparison between DRC of εzz for rubber and air suspension.

for εt and εzz, respectively. There is scatter in these • DRC increased with increase in temperature.
figures because there are several factors other than The DRC increased up to 12 times and 8 times
roughness level contributing to DRC values. The when temperature increased from 21.1°C (70°F)
factors include pavement structure, speed, pave- to 40°C (104°F) for thin and thick pavements,
ment temperature, and location of interest within respectively.
the pavement structure. The figures clearly show • The tensile strain DRC for severely rough pave-
that DRC is greater for rubber suspension than ment was 30 times and 60 times greater than that
air suspension as the majority of data points are observed for smooth pavement at 21.1°C (70°F)
above the line of equity. and 40°C (104°F), respectively. In most cases, the
maximum DRC of tensile strain was observed
6.4 Variation of DRC with DLC within the center to inner locations of the dual
tires.
Figures 12 and 13 compare DLC and DRC results • Rubber suspension inducesgreater DRC values
over the pavement responses corresponding to εt and than air suspension for both vertical compres-
εzz responses, respectively. DLC values were presented sive strain and tensile strain. The maximum
in Table 1. These figures show that DRC values are DRC was 3 and 6 times larger for vertical com-
generally greater than DLC values;thus bringing pressive strain and tensile strain, respectively,
into question of the traditional approach for directly when comparing rubber suspension to air
applying DLC values to pavement responses. suspension.
• The maximum DRC − ε zz and DRC − ε t for
rubber suspension were observed at the center
6.5 Other observations
location at a depth of 1.27 cm (0.5 in.) from the
Based on review of the entire data set, the follow- surface. It was also noticed that DRC for both
ing observations can be made: responses increased more in the rubber suspen-

439
 Cl
~

"#.

~ ~ --itD-
Et

14.72 I-- ~ ----1

14.36 I - - -t-ill---!- ~

15.97 f - - ~ -I
10.68 I-- {[]---- ;.........
9.59 1 - - ill- ~
10.16 ~-- q:J- ~

15.58 1-- ~ 1
9.01 f - - [];}--- ~
DRC -

8.77 f - - IT}-- -Ti

Cl
_,.

9.40 ~-- o;r- ~

#

8.20 1--- Cif- -:

-o 8.00 1---[ll-- -l

10.40 1 - - ap- ~:
7.28 1--rri--- -1
0 6.88 <---{[]-t- ....
6.38 ~ -I
6.73 1---{[}+--- I
Cl
N
#
Cl
"#

22.20
26.72
4.59

5.24
3.38

3.73

5.72
3.52

3.92
4.55

4.97
4.98

5.55

a::
5
Figure 12. Comparison between DRC of εt and DLC.
'#. ';fl.
Cl
(]'j
N

i
zz
Cl
Cl
N
E
....>.

Cl
(]'j

~
DRC -
Cl
Cl
....>.

';#.
Cl
(]'j

'*'
Cl
0~

10.16
10.40
10.68
14.36

22.20
14.72
15.58
15.97

26.72
5. 24

7.28
-o 8.00
3.38
3.52
3.73
3.92
4.55
4.59
4.97
4.98

5. 55
5.72
6.38
6.73
0 6.88

8.77

9.40
9.59
~ 8 . 20

9.01
5

Figure 13. Comparison between DRC of εzz and DLC.

sion than in the air suspension with higher tem- A new parameter called Dynamic Response
perature. In all cases, higher DRC values were Coefficient (DRC) was introduced to account for
observed for thin pavements compared to the the variation in pavement responses under variation
thick pavements. in dynamic loading. The developed DRC values can
• In the case of a thin asphalt pavement structure, be superimposed through Equation 5 on responses
DRC for rutting ( DRC − zz ) increased up to 10 calculated using traditional static-based pavement
times for severely rough pavement when compared response models (εo) to account for the effect of vari-
to smooth pavement. The maximum increase in ation in dynamic loading. The major findings from
DRC − ε zz was observed at the center location this study are presented below.
between the dual tires and 1.27 cm (0.5 in.) from
the surface. On the other hand, lower DRC val- • For a constant DLC value, a wide range of DRC
ues were observed for thick pavement with the values were calculated within the pavement for
maximum being at a depth of 3.81 cm (1.5 in.) a specific dynamic loading. This shows the
from surface between the dual tires. dependency of DRC on dynamic load, which is
• In a summary, it was observed that DRC val- a function of road roughness, tire suspension,
ues for both responses vary significantly within and truck speed. Furthermore, DRC depends
the HMA layer and as a function of pavement on pavement structure, pavement temperature,
structure, temperature, suspension type, speed, and more importantly location of the point of
depth, and road roughness. interest within the asphalt layer.
• The highest DRC values for rutting and fatigue
cracking were observed at a depth of 1.27 cm
(0.5 in.) and 3.81 cm (1.5 in.) for thin and thick
7 SUMMARY AND CONCLUSION pavement, respectively, with both occurred
under the center of the dual tires.
This study investigated the impact of variation in vehi- • DRC increased with increase in pavement tem-
cle dynamic loading on asphalt pavement responses. perature for both rutting and fatigue cracking.

440
• In general, an increase in DRC was observed Olidid, C. & Hein, D. Guide for the Mechanistic-Empiri-
with an increase in pavement roughness, with cal Design of New and Rehabilitated Pavement Struc-
the increase in DRC for fatigue being greater tures. 2004 Annual Conference and Exhibition of the
Transportation Association of Canada-Transporta-
than that for rutting.
tion Innovation-Accelerating The Pace, 2004.
• DRC values are generally larger for rubber sus- Siddharthan, R.V., Yao, J. & Sebaaly, P.E. 1998. Pave-
pension than for air suspension. ment strain from moving dynamic 3D load distribu-
• Using the DRC data generated in this study, the tion. Journal of Transportation Engineering, 124(6),
effect of road roughness and truck speed on per- pp. 557–566.
formance of flexible pavements can be evaluated Streit, D.A., Kulakowski, B.T. & Wollyung, R.J. 1994.
by using simple static-based pavement analysis Dynamic Vehicle Forces on Pavements: Second Interim
models similar to the one included in the current Report: Pennsylvania Transportation Institute, The
M-E Design. Pennsylvania State University.
Sweatman, P. 1983. A study of dynamic wheel forces in
axle group suspensions of heavy vehicles.
Wang, H. & Al-Qadi, I. 2009. Combined effect of mov-
REFERENCES ing wheel loading and three-dimensional contact
stresses on perpetual pavement responses. Transpor-
Gillespie, T.D. 1993. Effects of heavy-vehicle characteris- tation Research Record: Journal of the Transportation
tics on pavement response and performance: Transpor- Research Board, 2095, pp. 53–61.
tation Research Board. Woodrooffe, J.H., LeBlanc, P. & LePiane, K. 1986.
LeBlanc, P. & Woodrooffe, J. Spatial correlation of Effects of suspension variations on the dynamic wheel
dynamic wheel loads. Proceedings of the fourth Inter- loads of a heavy articulated highway vehicle.
national Symposium on Heavy Vehicle Weights and
Dimensions. UMTRI. Ann Arbor, 1995. 281.

441
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Dynamic behavior of flexible inelastic pavements under moving loads

N.D. Beskou, A.P. Chassiakos & D.D. Theodorakopoulos

Department of Civil Engineering, University of Patras, Patras, Greece

ABSTRACT: A finite element modeling of flexible pavements under moving vehicle loading is presented
in the framework of the ANSYS computer program. The pavement consists of a top asphalt layer with
a linear viscoelastic material behavior and two other layers (base and subgrade) with a linear elastic or
Drucker-Prager elastoplastic material behavior. The dynamic response of this pavement to a moving with
constant speed distributed load is numerically obtained in the time domain. This response involves the
histories of the vertical displacement at the surface, the horizontal strains at the bottom of the asphalt
layer and the vertical stress at the top of the subgrade, which are needed for design purposes. It is observed
that, in general, the maximum values of the response decrease with increasing values of speed in contrast
to the case of all the layers being linear elastic for which the opposite is true. Furthermore, it is observed
that the maximum values of the response of the models with the top layer being viscoelastic and the other
layers elastic or elastoplastic, remain always higher than the ones corresponding to the model with all
layers being elastic.

1 INTRODUCTION conjunction with the Fourier transform. The

moving distributed loads of the vehicle were
Analytical and experimental research on the analy- assumed to be either real dynamic or quasi-static
sis and design of flexible pavements has attracted ones with only the former case taking into account
the attention of many investigators during the last the effect of the pavement inertia. However, the
35 years or so (Huang, 2004; Beskou & Theod- latter case is associated with a small domain size.
orakopoulos, 2011). Originally, flexible pavements A rational way to describe non-linear material
were modeled as layered systems exhibiting linear properties is by using theories of plasticity. Thus,
elastic material behavior and subjected to static, Zaghloul & White (1993), Shen & Kirkner (2001),
stationary, distributed load. Then the load was Saad et al. (2005), Ali et al. (2008) and Huang et al.
assumed stationary but dynamic (impact load) to (2011) used theories of plasticity or viscoplasticity
simulate the test process of deflectometry. Even to describe the nonlinear behavior of the layers and
though elastic flexible pavements under static load the FEM in the time domain to solve the problem
can be analyzed approximately by hand, for real- under 2-D or 3-D conditions and dynamic or qua-
istic boundary conditions and dynamic and/or si-static moving load modeling. However, in all the
moving loads, use of numerical methods, such as cases where inelastic material behavior was consid-
the Finite Element Method (FEM) is imperative ered, only specific problems were solved without
(Beskou et al. 2016a). parametric studies taking into account the effects
Linear elastic material behavior is obviously an of inelasticity and speed on the dynamic response
approximation and if one requires analytic results of the pavement.
close to experimental ones, he has to adopt more In this work, using the FEM in the time domain
realistic material behavior. Use of linear viscoelas- developed in Beskou et al. (2016b), the dynamic
tic material behavior for the top asphalt layer, while response of a 3-D three-layers flexible pavement to
retaining all the other layers linear elastic, was moving with constant speed distributed loading is
found to produce improved response results, i.e. determined for various combinations of material
results with a maximum error of about 15% with behavior in the layers including elastic, linear vis-
respect to experimental field results (Elseifi et al. coelastic and Drucker-Prager elastoplastic behavior.
2006, Chabot et al. 2009, Liao &Sargand 2010).
In these cases except in Chabot et al. (2009), use
was made of the FEM in the time domain under 2 FEM MODELING IN SPACE AND TIME
three-dimensional (3-D) conditions. In Chabot
et al. (2009) use was made of the method of the This section deals with the finite element mod-
3-D layered system in the frequency domain in eling of the pavement structure, the modeling

443
of the moving vehicle loads and the finite ele- 2.3 Finite element solution in time domain
ment solution in the time domain, as described in
Following standard finite element procedures
Beskou et al. (2016b).
(Bathe, 1996), one can formulate the equation of
motion of the pavement structure of Figures 1 and
2.1 Finite element modeling of pavement 2 as:
structure
The typical three dimensional (3-D), three layer [ M] { } + [C]{u} + {R(
{R(u, u, u)} = {F} (1)
flexible road pavement structure model of Figure 1
is considered here. It consists of three layers fully where [M] = mass matrix; [C] = damping matrix;
bonded to each other and to the supporting soil {u} = nodal displacement vector; {F} = nodal exter-
with the top one being the asphalt concrete layer of nal force vector; {R} = vector of the nodal internal
thickness 0.15 m, the intermediate one the granu- resistance forces, which are nonlinear functions
lar base layer of thickness 0.30 m and the bottom of {u}, { } and { } and overdots denote differ-
one the subgrade layer of thickness 29.00 m. It entiation with respect to time t. Internal resistance
should be noticed that the lateral face of the model forces {R} for linear elastic or viscoelastic material
designated by the zx plane is a plane of symmetry behavior, reduce to [K]{u}, where [K] is the stiff-
and that the x axis represents the axis of the road ness matrix. External forces in {F} refer to nodes
pavement (Figs 1 and 2). at the surface z = 0 of the pavement structure and
The above model is supported at its bottom come from the moving loads. It should be noted
and its lateral faces (its fourth face is the plane of that viscous damping is not considered in the
symmetry zx) by rollers. Adoption of rollers at the present studies as its contribution to the energy of
boundaries has been found in (Beskou et al. 2016a) dissipation of the system can be thought of as neg-
to provide almost the same response results with ligible in comparison to that coming from inelastic
those obtained by using viscous absorbers at the deformation.
boundaries for the selected domain dimensions and The step-by-step time integration algorithm
linear elastic material behavior. The above pave- of Newmark of the constant acceleration type
ment structure is discretized into a finite number under zero initial conditions is used in this work
of 8-noded 3-D solid elements (bricks) with 24 in to solve Eq. (1), as for the case of linear elastic
total degrees of freedom (SOLID 185 type in material behavior. However, because of the non-
ANSYS program, 2010). The finite element mesh is linear character of Eq. (1) due to the presence
shown in Figure 2 and consists of 82,080 elements of {R}, iterations are needed at every time step,
or 88,257 nodes (3*88,257 = 264,771 degrees of which are performed with the aid of the modified
freedom).

2.2 Modeling of the moving vehicle loads

The typical heavy vehicle of 80 kN is used here
in its distributed load form, which is actually a
superposition of closely spaced point or concen-
trated loads. The motion of a point load with
constant speed along the x axis from point A to
point B (Fig.1) is simulated by assigning time
dependent load values at all the nodes of the
line segment AB, which are activated at the time
it takes for the load to travel the distance from
A to every other node along AB. The point load
P = 80 kN acts over an area of length 0.45 m and
width 0.30 m and the resulting distributed load
is converted into an equivalent system of point
loads placed symmetrically on either side of the
x axis. These loads are treated as series of point
loads parallel to the x axis for which their effect
can be taken as a superposition of single point Figure 1. General geometry of half of pavement
loads. Use is made of 28 axles of loads moving structure domain (symmetry with respect to the X-Z
along the x axis for an almost perfect simulation. plane): (OA = BR = 13.20 m, AQ = QB = 1.80 m,
One should notice that because of symmetry, only OR = 30.00 m, OC = 15.00 m, CD = 0.15 m, DF = 0.30 m,
half of those loads are utilized. FG = 29.00 m).

444
 character, respectively. Linear viscoelasticity in
the time domain is described by the constitutive
equation:

dΔ
σ( ) t
∫ 0 G( ) de d t
∫ 0K( ) dτ (2)
dτ dτ

where σ = stress tensor; e = deviatoric part of the

strain tensor ε; Δ = volumetric part of the strain
tensor ε; t and τ denote current and past times;
I = unit tensor; G(t) = shear relaxation function;
and K(t) = bulk (volumetric) relaxation function.
In asphalts, viscoelastic effects are negligible in a
volumetric stress state and hence all the viscoelas-
tic deformation is of the shearing type. Thus, func-
tion K is assumed to be a constant. Function G(t)
can be expressed on the basis of the generalized
Maxwell model in the form of Prony series as:
n
G (t ) = G + ∑ G i e −t / τ i
G
Figure 2. Discretization of domain of Figure 1 into (3)
i =1
82,080 finite elements with 88,257 nodes.
where G∞ and Gi = material constants in Pa; and
Newton-Raphson scheme. The above described τ iG = material constant in sec. Temperature plays
solution procedure is done in the framework of the a very important role in viscoelasticity. However,
computer program ANSYS (2010). in this work, the above material constants are
The selection of an appropriate value for the assumed to be independent of temperature as cor-
time step Δt to be used for the above time solution responding to a fixed temperature value (usually
procedure in order to achieve stability and accept- 20˚C).
able accuracy, is based on the rules established
for the linear elastic case (Bathe, 1996). Thus, 3.2 Elastoplasticity
on the basis of extensive studies done in Beskou
et al. (2016a) for the linear elastic case, a value of Elastoplastic material behavior is usually assigned
Δt = 0.3 × 10−3 sec is selected here as providing con- to the base and subgrade layers. This behavior can
vergent results of acceptable accuracy. be thought of as the one produced by a model con-
sisting of springs and friction sliders representing
its elastic and plastic character, respectively.
3 MODELING THE MATERIAL The yield surface (or function) F is the locus of
BEHAVIOR OF LAYERS the stress combination points in the principal stress
space (σ1, σ2, σ3) for which yielding occurs. In plas-
This section briefly describes the two types of ticity one works with increments or rates of stress
inelastic material behavior assumed for the three σ and strain ε tensors. Assuming non-associated
layers of the pavement structure of Figure 1. These plasticity, one can express the plastic strain rate
are linear viscoelastic and elastoplastic. Strictly ε p as:
speaking, linear viscoelasticity is not included in
⎛ ∂Q ⎞
inelasticity, which implies nonlinearity. However, ε p = λ ⎜ (4)
the term inelasticity is used here in the sense of ⎝ ∂σ ⎟⎠
non-elasticity and in that sense includes linear
viscoelasticity. The following brief descriptions of where Q = plastic potential (function of stress);
material behavior come from ANSYS (2010). and λ = constant to be determined. Further more,
one assumes that:
3.1 Linear viscoelasticity
ε = ε e + ε p (5)
Linear viscoelastic material behavior is usually
assigned to the top asphalt layer. This behavior where ε = total strain rate; and ε e = elastic strain
can be simulated by a model consisting of springs rate. Invoking Hooke’s law for the elastic strain
and dashpots representing its elastic and viscous rate and Eq. (5) one finally has:

445
 ⎛ ∂Q ⎞ 35, 50 m/s (18, 36, 72, 126, 180 km/h). Dynamic
σ   λ ⎜ (6) responses have been obtained for all the cases
⎝ ∂σ ⎟⎠
considered here by the FEM in the time domain
described in section 2 with Δt = 0.3 × 10−3 sec.
where D = elastic matrix; and λ = a function of F, Consider the three layers domain of Fig-
Q, σ and ε. Thus, σ finally takes the form: ure 1 with the following material properties:
E1 = 1000 MPa, ν1 = 0.35, ρ1 = 2500 kg/m3 for the top
σ = Depε (7) layer, E2 = 400 MPa, ν2 = 0.35, ρ2 = 2100 kg/m3 for
the intermediate layer and E3 = 80 MPa, ν3 = 0.40,
where Dep = elastoplastic matrix. When plasticity ρ3 = 2000 kg/m3 for the last (third) layer. On the
is associated, then Q = F and relation (4) is the basis of this linear elastic three layer model (Model
normality condition. For most soils, the ε p is not 0 or elastic), the following two inelastic pavement
normal to F and the most appropriate theory of models are considered here: i) Model 1 or viscoe-
plasticity is the non-associated one. The function lastic with the top layer being linear viscoelastic
Q has usually the form of F with different value and the other two linear elastic, ii) Model 2 or VE/
for a material constant as it will be shown in the DP with the top layer being linear viscoelastic and
following. the other two Drucker-Prager (DP) elastoplastic.
Soils are characterized by a stress state, which is In the following subsections, response results of
dependent on the hydrostatic stress state (σ1 = σ2 these two inelastic models are presented and com-
= σ3) and the dilatancy. For this case, a suitable pared against those of the linear elastic model.
yield surface is that of Drucker-Prager (DP) in Due to space limitations, only response results for
the form of a right cone with axis the line σ1 = σ2 the vertical deflection uz at z = 0.00 m (surface of
= σ3. The material parameters for the elastoplastic pavement), strain εy at z = −0.15 m (bottom surface
model are the modulus of elasticity E, the Pois- of top layer) and stress σz at z = −0.45 m (top of
son’s ratio ν, the angle of friction between the the subgrade layer) are provided here. These are
grains θ, the dilatancy angle θf and the cohesion usually needed for design purposes (Huang, 2004).
between the grains c, which is related to θ and the Additional response results can be found elsewhere
yield stress σy by: (Beskou et al. 2016b).

c = σ y 3(3 − si θ ) / 6 cosθ (8) 4.1 Model 1 and its response results

In this model, the top layer has a linear viscoe-
Since F can be expressed in terms of θ, Q can be
lastic behavior described by Eqs (2) and (3). The
obtained from F by simply replacing θ by θf. Thus,
required expression of G(t) in terms of Prony-
for θf = θ one has Q = F and the non-associated
series of the form of Eq.(3) is obtained on the
plasticity becomes an associated one. Use of θf≠θ
basis of experimental results in Berthelot et al.
(non-associated plasticity) results in non-symmet-
(2003) appropriately scaled in order to have
ric stiffness matrices and dramatic increase of the
G(0) = E/2(1+ν), where E and v are the elastic val-
computational time. For this reason, in the present
ues of the top layer. Figure 3 depicts the variation
applications the assumption is made that θf = θ
of G(t) versus time for two cases. The curve with
(associated plasticity). Typical values of c and θ for
the broken line corresponds to the asphaltic mate-
granular bases and subgrades range between 1 and
rial 900901 of Berthelot et al. (2003) scaled down
690 kPa and 10˚ and 45˚, respectively.
from G(0) = 817 MPa to G(0) = 1000/2(1+0.35) =
384.60 MPa in order to be compatible with the
4 COMPARISONS BETWEEN INELASTIC elastic properties of the top layer. The curve with a
AND ELASTIC MODELS solid line is a curve artificially constructed in order
to have a more abrupt reduction of G(t) with time
This section reports the response results of two ine- and hence more pronounced viscoelastic results in
lastic pavement model structures to moving vehicle comparison with the elastic ones. This solid line
loads, as obtained by the FEM described previ- is assumed to be representable by a Prony series
ously, and compares them against those of the cor- with 3 terms and τ 1G 2 0 τ 2G 0 2,τ 3G = 0.2 sec in
responding linear elastic pavement structure. The Eq. (3). Use of nonlinear regression analysis ena-
moving load path is the distance (AB) = 3.60 m of bles one to determine the values:
Figure 1 and the response is recorded at point Q, G1 = 19.36, G2 = 68.66, G3 = 161.86,
which is in the middle of that distance. The distrib- G∞ = 134.74 MPa (9)
uted load p = 0.5926 MPa (due to a load P = 80 kN
acting on a surface 0.45*0.30 m2) moves along the Figure 4 provides the time histories of the ver-
distance (AB) with a constant speed V = 5, 10, 20, tical deflection uz at z = 0.00 m, the strain εy at

446
 400
Go

~
~ ', I Adopte d c ase
3 50 - - Original case
\ ' -2
3 00
\

~
m
(l_
250 \ \

I -4
\ ' \

"'
~

-g
E
2 00

'
"c
~
-6
150 ~-
m ' G oo
="
V> 100

50
""" -8

-1 0
0.0 0 .5 1.0 1.5 2.0
Time [Sec]
T ime ( s)
3.0

Figure 3. Shear modulus of asphaltic material versus

time for examples of subsections 4.1 and 4.2.
2.5

2.0
H
r- r---
I
+---
Elastic c~
- Viscoelastic

z = −0.15 m and the stress σz at z = −0.45 m for 1.5

V = 5 m/s and the two pavement models considered

here: Model 0 (elastic) and Model 1 (viscoelastic).
".
0
X

w
1 .0 r-- -
I Depth
- -
= 0.15 m I

One can observe that the viscoelastic model pro- 0 .5 r--- - r----
'
vides higher maximum values than the elastic one
for all these response quantities in agreement with
Elseifi et al. (2006), Chabot et al. (2009), and Liao
0.0

-0.5
DO
~ Lf 0.5
--------
I
1.0 1 .5 2.0
& Sargand (2010). Time [Sec]
Figure 5 shows the maximum vertical deflection
uz at z = 0.00 m, strain εy at z = −0.15 m and stress
σz at z = −0.45 m versus speed for both the viscoe-
lastic and elastic models. One can observe that (i) -20

the maximum response of the viscoelastic model

is higher than that of the elastic one for all speeds,
(ii) the maximum response of the elastic model "'
"-
~
increases with speed for all response quantities, (iii) t/" -60

the maximum response of the viscoelastic model

decreases with speed for εy and σz for all speeds, (iv) -80

the maximum response uzof the viscoelastic model

decreases with speed up to the speed of 20 m/s and -100-l-- - - - , - - - - - + - - - - - - - , f - - - - -- l
0.0 0.5 1.0 1.5 2 .0
increases for higher speeds. These phenomena can
Time [Sec]
be physically explained as follows: Elastic materi-
als convert all the imparted to them kinetic energy Figure 4. Time histories of uz (z = 0.0 m), εy (z = −0.15 m)
of the system into deformation and since kinetic and σz (z = −0.45 m) for V = 5.0 m/s and Models 0 (elastic)
energy increases with speed, so does the deforma- and 1 (viscoelastic).
tion. In contrast, viscoelastic materials are dissipa-
tive and their dissipation increases with increasing
speed. Thus, only the remaining part of the total described by Eq. (7). The material parameters con-
kinetic energy of the system is converted to defor- cerning the viscoelastic behavior are the same as
mation and hence that deformation decreases with in Model 1. The material parameters concerning
increasing speed. All these results are in qualitative the DP elastoplastic behavior are θ = θf = 40˚ and
agreement with the test results of AASHO Road σy = 100 kPa for the second layer and θ = θf = 30˚
Test (1962) and the numerical and test results of and σy = 30 kPa for the third layer. Of course E,
Liao and Sargand (2010), which are restricted to ν and ρof the layers are the same as in the elastic
speeds up to about 27 m/s and 20 m/s, respectively. model (Model 0).
Figure 6 provides the time histories of the verti-
cal deflection uzat z = 0.00, strain εy at z = −0.15 m
4.2 Model 2 and its response results
and stress σz at z = −0.45 m for V = 5 m/s and Model
In this model, the top layer has a viscoelastic 0 (elastic) and 2 (VE/DP). Figure 7 shows the max-
behavior described by Eqs. (2) and (3), while imum vertical deflection uzat z = 0.00 m, strain εy at
the other two have a DP elastoplastic behavior z = −0.15 m and stress σzat z = −0.45 m versus speed

447
 9.05
I
9.00
8.95 I
8.90 L'
-2
8.85
--E>- ------- -.p--- -- --- - - -€ ( /

I
10
8.80
8.75
8.70
8.65
I
L
I
~
~
I Depth
/
-

-Om
r-

J-
..
I
0

~
-4

-6
8.60 ~ "
E. 8.55
/ -r- ---+ _±
8.50
8.45
y I - e - Elastic
-- 8 - - Viscoelastic
J=
8.40 I
8.35 I
r{ 0.0 0.5 1.0 1.5 2.0
8.30 +-- --+- -I--
8.25
Time ]Sec]
0 10 20 30 40 50
4.0
Speed [m/sec]
3.5 - ,, f - -t- I
3.0 ~-----~~----------f=~E31a0.s~tic~~
2.60

2.55
":
_[ - e -
- - 8--
Elastic
Viscoelastic
J- 2.5

2.0
', --- -- VEJDP]

Depth -0.15 m J
'. I b _________ _]__ _______ [--
2.50
...:..- __ I Depth = -0.15 m I
X 1.5
'-- -----

~
.?
2.45 - - 1.0

10
- .. 0.5

E.
"'
2.40

2.35
- - -~
t t I I
0.0

----+--- -0.5

2.30

2.25 =:::::::1I
--8-

T~
-- -r 0.0 0.5 1.0
Time [Sec]
1.5 2.0 2.5

,___- r- I I 20 T"-------~----r---,

2.20
10 20 30 40 50
Speed [m/sec]
-20

88

87
J -e- Elastic
Viscoelastic
J_ "'
a_
~
-40

-- E>- - ~- -60

86
I Depth - -0.45 m

85 -- - -- -- - - - - -E>-- - -
tO
0..
0 .0 0.5 1.0 1.5 2.0
~ 84
Time [Sec]
b
83
Figure 6. Time histories of uz (z = 0.0 m), εy (z = −0.15 m)
and σz (z = −0.45 m) for V = 5.0 m/s and Models 0 (elastic)
82
and 2 (VE/DP).
81
10 20 30 40 50 It has been found that the dynamic behavior of
Speed [m/sec] the two inelastic models considered here are simi-
lar and that their response results (with the excep-
Figure 5. Maximum values of uz (z = 0.0 m), εy tion of those for εy showing increasing values from
(z = −0.15 m) and σz (z = −0.45 m) versus speed for Mod- model 1 to 2) are also similar. This suggests that
els 0 (elastic) and 1 (viscoelastic). one could successfully use in practice the viscoe-
lastic model, which is the simplest and requires
the lowest computational cost. Of course, more
for both the VE/DP and elastic models. One can detailed comparison studies involving test results
observe from both figures that all results here con- are needed in order to reach definite conclusions
cerning the VE/DP model are almost in qualitative on this matter.
agreement with those of the viscoelastic model It has been found in Beskou et al. (2016b) and
considered previously. The physical explanation of references therein that measured response val-
it is the same as in the viscoelastic case. ues are always higher than but close to dynamic

448
 9.4
5 CONCLUSIONS
1. . . - · · · · · ·
9.2
.. --8···' 1 A general 3-D FEM model for the in time simu-
0-.
- - ---- -- --- -- lation of the dynamic response of flexible road
9.0 - - -t - -
pavements to vehicles moving with constant
I
8.8
~ speed under inelastic material behavior has been
"
0
~
~ t-- I
~
I
constructed in the framework of the ANSYS
computer program. Two inelastic material mod-
8.6

8.4
I--

v 1
Depth
--&-
=0m
Elastic
-- <>-- VE/DP -
els for flexible pavements have been considered:
The viscoelastic (top layer viscoelastic and
other layers elastic) and the VE/DP (toplayer
/
8.2
viscoelastic and other layers Drucker-Prager
0 10 20 30 40 50 elastoplastic). It was found that inelastic mod-
Speed [m/sec] els produce response results (deflections, strains
and stresses at various depths) always higher
3.6 than the corresponding elastic ones and that,
',_
in general, those results decrease for increasing
3.4 speed (at least for speeds up to about 20 m/s or
72 km/h) in agreement with experiments and in
3.2
--- ------- ---€)-- ------------ contrast with elastic model results exhibiting
an increase with increasing values of speed. It
3.0 --
• 1 was also found that the two inelastic models, at
0
~

"'
2.8
t- I-&- Elasti~ I
--<>-- VE/DP
least for the cases considered here, exhibit simi-
lar dynamic behavior and response results and
2.6 hence one could use the viscoelastic model for
2.4 I l I Depth = -0.15 m I practical applications as being the simplest and
I the one requiring the lowest computational cost.
2.2 Besides, viscoelastic models produce results
10 20 30 40 50 close to those coming out of tests.
Speed [m/sec]

93
ACKNOWLEDGEMENT
~- - .
92
91
The first author (Niki D. Beskou) acknowledges
-- ---- -- - --- ----- with thanks the support provided to her by the
90 r- ~----&

‘IKY fellowships of excellence for postgradu-

-ro
a.
89
88 -
- r-----
1-- - - _ I-&- Elastic
-- <>-- VE/DP
t-
ate studies in Greece-Siemens program’. All
authors acknowledge with thanks helpful discus-
~ 87 1-- - - - --t- r- sions with Professors S.V. Tsinopoulos and G.D.
b 86 Depth -0.45 m
Hatzigeorgiou.
85
84
__.;,
83
,.----------- REFERENCES
82 ~__..-< 1- ~ American Association of State Highway Officials—
10 20 30 40 50
Speed [m/sec]
AASHO. 1962. AASHO Road Test. Special Reports
61 A-61G, Highway Research Board.
Ali, B., Sadek M. & Shahrour I. 2008. Elasto-viscoplas-
Figure 7. Maximum values of uz (z = 0.0 m), εy tic finite element analysis of the long-term behavior
(z = −0.15 m) and σz (z = −0.45 m) versus speed for Mod- of flexible pavements. Road Materials and Pavement
els 0 (elastic) and 2 (VE/DP). Design 9: 463–479.
ANSYS Inc. 2010. ANSYS User’s Guide, Release 13.
viscoelastic ones. Furthermore, it has been found Southpointe, Canonsburg, PA, USA.
Bathe, K.J. 1996. Finite Element Procedures. Englewood
here that response values from dynamic viscoelas-
Cliffs, New Jersey: Prentice-Hall.
tic models are always higher than dynamic elastic Berthelot, C.F., Allen, D.H. & Searcy C.R. 2003. Method
ones. Thus, elastic models underestimate the true for performing accelerated characterization of vis-
pavement response and should not be used in coelastic constitutive behaviour of asphaltic con-
design. On the contrary, viscoelastic models pro- crete. Journal of Materials in Civil Engineering 15(5):
duce response results close to the real ones. 496–505.

449
Beskou, N.D. & Theodorakopoulos, D.D. 2011. Huang, Y.H. 2004. Pavement Analysis and Design. Upper
Dynamic effects of moving load son road pavements: Saddle River, New Jersey: Pearson Prentice-Hall.
A review. Soil Dynamics and Earthquake Engineering Liao, J. & Sargand S. 2010. Viscoelastic FE modeling
31: 547–567. and verification of a U.S. 30perpetual pavement test
Beskou, N.D., Tsinopoulos S.V. & Theodorakopoulos, section. Road Materials and Pavement Design 11:
D.D. 2016a. Dynamic elastic analysis of 3-D flexible 993–1008.
pavements under moving vehicles: A unified FEM Saad, B., Mitri, H. & Poorooshasb H. 2005. Three-
treatment. Soil Dynamics and Earthquake Engineering dimensional dynamic analysis of flexible conventional
82: 63–72. pavement foundation. Journal of Transportation Engi-
Beskou, N.D., Hatzigeorgiou, G.D. & Theodorakopou- neering of 131(6): 460–469.
los D.D. 2016b. Dynamic inelastic analysis of 3-D Shen, W. & Kirkner, D.J. 2001. Non-linear finite element
flexible pavements under moving vehicles: A unified analysis to predict permanent deformations in pave-
FEM treatment. Soil Dynamics and Earthquake Engi- ment structures under moving loads. The International
neering 90: 420–431. Journal of Pavement Engineering 2: 187–199.
Chabot, A., Chupin, O., Deloffre, L. & Duhamel D. Zaghloul, S. & White T. 1993. Use of a three-dimensional,
2009. Viscoroute 2.0: a tool for the simulation of mov- dynamic finite element program for analysis of flex-
ing load effects on asphalt pavement. Road Materials ible pavements. Transportation Research Record 1388:
and Pavement Design 10: 1–27. 60–69.
Elseifi. M.A., Al-Qadi I.L. & Yoo P.J. 2006. Viscoelastic-
modeling and field validation of flexible pavements.
Journal of Engineering Mechanics 132(2): 172–178.
Huang, C.W., Abu Al-Rub, R.K., Masad E.A. & Little
D.N. 2011. Three-dimensional simulations of asphalt
pavement permanent deformation using a nonlinear
viscoelastic and viscoplastic model. Journal of Materi-
als in Civil Engineering 23: 56–68.

450
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Responses and performance of flexible pavements in cold climate due to

heavy vehicle loading

S. Erlingsson
Pavement Technology, Swedish National Road and Transport Research Institute (VTI), Linköping, Sweden
Faculty of Civil and Environmental Engineering, University of Iceland, Reykjavik, Iceland

J. Ullberg
Swedish Transport Administration, Luleå, Sweden

ABSTRACT: Four instrumented test road structures have been built in northern Sweden about 100 km
north of the Arctic Circle. The climate consists of cold winters and mild summers with a long spring thaw
period. The pavement structures differs in material quality ranging from conventional to polymer modi-
fied structures. The objective of the test structures is to see the difference in their behaviour and perform-
ance due to the heavy loading in the ambient artic climate. Each test section is instrumented that includes
pressure cells and strain sensors that measure the pavement response to external loading. Climatic varia-
bles are further monitored through a weather station as well as temperature and moisture sensors installed
in the road structure. Response measurements have been carried out using 90 tons vehicles consisting of
a single wheel steering axle, and triple dual wheel tridem axles. Rutting development has been monitored
through four years showing some difference between the structures rutting development.

1 INTRODUCTION binder course in the summer of 2012. The wear-

ing course was thereafter placed in the beginning
Four new Long Term Pavement Performance (LTPP) of July 2013.
test structures were built in Norrbotten County in
Northern Sweden in 2012. The structures are situated
on the road E45 about 100 km north of the Arctic
Circle in a climate that is characterized by long cold
winters with short mild summers where the dura-
tion of the thawing period exceeds two months. The
structures are instrumented with road response and
climate sensors. A local iron ore operator received a
special permission to transport ore along the road
using vehicles that are 25 m long with a 90 tonnes
gross weight, instead of the permissible 60 tonnes.
Due to this new ore transport it was deemed nec-
essary to strengthen the local road network. As a
part of evaluating pavement performances for very
heavy vehicles, four test structures were selected
and built with the main objectives to increase the
knowledge about pavement response and perform-
ance under heavy loading in cold climate with sea-
sonal variation.

2 THE TEST STRUCTURES

The four test sections are located on the road E45

close to the intersection with E10 in the vicinity
of the Svappavaara village (see Figure 1). The Figure 1. Overview of the four test sections in northern
test structures were built up to the bituminous Sweden.

451
 All the test sections were built in a conventional of a large aggregate skeleton layers mixed with cold
matter on top of an existing road. The top 15 cm emulsion with 160/220 penetration grade bitumen.
of the old road were milled and mixed with its sub- The Unbound Base Course (UBC) consist of well
base layers and thereafter widened according to graded crushed rock aggregates with a maximum
the new geometrical design. The remains of the aggregate size 31.5 mm. The Subbase (Sb) consist also
old structure can therefore be expected to consist a crushed rock aggregates from the same quarry as
of about 70 cm of sandy gravel resting on top of the base course but has a maximum aggregate size of
the native silty sand subgrade. 90 mm. These new structures are resting on a 70 cm
The cross sections of the four test structures thick natural sandy gravel that is the old existing road
are shown in Figure 2. A more detailed descrip- structure. The native subgrade (Sg) is gravelly till (silty
tion of the layer composition is further provided in sand). The groundwater table is unknown but as the
Table 1. The layer thicknesses of structures S01–03 structures are resting on relatively high embankments
are identical but structure S04 has a thicker road it is expected to be a great depth.
base course, consisting of a cold asphalt concrete
mix, and a reduced subbase thickness in order to s 01-03 0.0
s 04
0.0
let all the structures have the same total thickness.
2.0
8.0
~~J!~~~~~~e 2.0
8.0
~~~!~~~~~~e
Road Base Road Base
20.0
Further details of the pavement structures can be 30.0
Unbound Base 26.0 Unbound Base
found in Erlingsson and Carlsson (2014). Subbase 36.0
Subbase
The bitumen bound part of all the four struc-
60.0 60.0
tures consists of four layers; a thin surface Wearing Subbase Subbase
Course (WC) layer, a Binder Course (BC) layer and Natural sandy grav- Natural sandy grav-
el
el
two Road Base (RB) layers. All structures have the (Old existing road) (Old ex isting road)
same wearing course TSK 16 with 160/220 penetra-
tion grade standard bitumen. Structure S01 has a
polymer modified binder course as well as the two 130.0 130.0
Subgrade Subgrade
road base layers whilst structure S02 has the same Gravelly till/si lty Gravelly till/si lty
polymer modified binder course as structure S01
Depth [em] Depth [em]
but a conventional bitumen in the road base lay-
ers. Structure S03 consists of conventional bound Figure 2. Cross section of the test structures. Structure
materials, with a 70/100 penetration grade bitu- S01–03 have same layer thicknesses. Structure S04 has a
men for the binder course as well as for the two thicker road base and therefore the subbase thickness is
road base layers. Structure S04 has a conventional reduced in order to let all the structures have the same
binder course but the two road base layers consists total thicknesses.

Table 1. Short description of the layers and their thicknesses of all the four different test structures.

Structure

01 02 03 04

WC 20 mm 20 mm 20 mm 20 mm
Surfacing Surfacing Surfacing Surfacing
BC 60 mm 60 mm 60 mm 60 mm
ABb22 with ABb22 with ABb22 70/100 ABb22 70/100
Nypol 64–34 Nypol 64–34
RB 60 mm 2 × 60 mm 2 × 60 mm 2 × 90 mm
AG22 40/100–75 + AG22 160/220 AG22 70/100 Large aggregate AC
60 mm (Viacomac 32) 160/220
AG22 90/150–75
UBC 100 mm 100 mm 100 mm 100 mm
Crushed rock 0/31.5 Crushed rock Crushed rock Crushed rock 0/31.5
0/31.5 0/31.5
Sb 300 mm 300 mm 300 mm 300 mm
Crushed rock Crushed rock Crushed rock Crushed rock
0/90 + 0/90 + 0/90 + 0/90 +
700 mm 700 mm 700 mm 700 mm
Sandy gravel Sandy gravel Sandy gravel Sandy gravel
Sg Gravelly till/ Gravelly till/ Gravelly till/Sandy silt Gravelly till/Sandy silt
Sandy silt Sandy silt

452
3 TRAFFIC DATA • Asphalt Strain Gauges (ASG) for measuring the
transversal tensile strain at the bottom of the
As stated before the four test sections are located in bounded layers.
sparsely populated area in Northern Sweden. The • Soil Pressure Cells (SPC) for measuring the
traffic amount consisted of 580 vehicles per day induced vertical stress.
and lane with 16% consisting of heavy vehicles. In
The placement of the instrumentation of section
combination with a new ore mining that opened in
S01–03 is shown in Figure 4. The instrumentation
2012, 90 ton vehicles started to drive along the route
changing the loading situation. The ore operation for structure S04 is similar but adapted to the slightly
was gradually increasing during the first year but in different layer thicknesses. For further description of
full operation it was expected that 7–8 passages per the road sensors, see Saevarsdottir et al. (2016).
hour or about seven and a half minutes between the
passes resulting in 66.000 journeys per year. During Wearing Course
the year 2014 in average additional seventy 90 ton Binder Course
vehicles per day and lane were trafficking the road.
Road Base
However in October 2014 the operation was stopped
due to economic problems of the ore operator and
Unbound Base Course
later the company became bankrupted. All transport
30.0
with the 90 ton vehicles was therefore terminated and
the traffic amount went back to its initial values.

4 ROAD RESPONSE MONITORING

60.0
Subbase
The instrumentation of the test sections can be Natural sandy gravel
(Old existing road)
divided into road sensor instrumentation and cli-
mate sensors, see Figure 3 for an overview of the Depth [em]
I Emu- Vertical strain
sensors locations. The road sensors instrumenta- ...... ASG - Horizontal tensile strain
tion are aligned in the centre of the outer wheel !- SPC- Vertical stress
path of the road. The sensors consist of:
• Emu coils (Emu) for measuring the vertical Figure 4. Schematic overview of the instrumentation of
strains. test sections S01–S03.

lOOm lOOm lOOm lOOm

0 Data logger 1, 2, 3 & 4

c::::::::::J Road sensors instrumentation 1, 2, 3 & 4
0 Moisture rod 1, 2 & 3

• Temperature & frost rod 1

• Temperature sensors in AC

Figure 3. Overview of the location of instrumentation for the four test section S01–S04.

453
5 CLIMATIC MONITORING In the beginning of April the thawing period starts
(see Figure 6a). This can also been seen as a rapid
The climate sensors consist of a weather station, frost increase in vol. moisture content starting at the TDR
road, moisture rods and temperature sensors. The sensor that is closest to the surface (the sensor in the
climate has been monitored since the structures were subbase at D = 0.33 m in Fig. 6c), and continues few
built in 2012. Besides a general weather station, a 2 m days later at the sensor at D = 0.685 m and then fol-
long frost rod is installed in structure 3 that consist lows in a timely manner for the other sensors as the
of temperature sensors spaced every 5 cm that moni- thawing front penetrates deeper into the ground.
tor the frost depth. Further, there are moisture rods After the sensors are thawed their vol. moisture con-
installed in structure 1, 2 and 3. Each rod consist of tent remains high with a mild reduction rate. This is
five Time Domain Reflectometer (TDR) sensors due to that the water is trapped in the structure as
with the sensor centre depth at 33.0, 68.5, 104, 139.5 there is a frozen layer underneath and only a small
and 175 cm depth. Each TDR sensor is 200 mm long amount of water migrates laterally towards the edges.
and the registration value is the average volumetric As the deepest located sensors thaws at D = 1.75 m in
moisture value over the entire sensor length. June (Fig. 6d) the TDR sensors registrations shows
Registrations of the air temperature, daily pre- rapid decrease in moisture content at all depths as
cipitation and moisture content for one year (2015) the water can now easily drain downwards.
are shown in Figure 5. The following can be seen
on the figure. 30
In the beginning of the year (first half of January) 20
the structure is frozen down to more than 1.4 m u
2..... 10
depth. This can be seen as the air temperature (Fig. ~ 0

5a) ranges mainly between 0 and −30°C and the l-10

~ -20
moisture content at all sensors (Fig. 5c and d) except -30
the sensor at D = 1.75 m show low values that are -40
affected by the free hydrogen molecules of the frozen Jan Apr Ju l
Date
Oct

material (Erlingsson et al., 2009). In mid-January the 30

sensor at D = 1.75 m depth starts to freeze and the 25
sensor is fully frozen along its entire length before E
..s 20
the end of the month. This can be seen as a drop 0
.E 15
in the volumetric moisture registration in Figure 6d. ·c. 10
·u
.t 5

Jan Apr Jul Oct

Date
30

'* 25

~ 20
t;
·c; 15
20.0 E
Unbound Base Course
30.0 ·;:::
10
Subbase
~
0 5
>
60.0
Subbase Jan Apr Jul Oct
Date
Natural sandy gravel
35
(Old existing road)
30
~
25

130.0
~ 20
~
t; 15
-~
Subgrade
·;::: 10
Gravelly t ill I si lty sand
~
0
5

> 0
Jan Apr Ju l Oct
Date

Frost rO:d ·: Moist"ure rod: Figure 6. Environmental monitoring at the site during
year 2015. a) Hourly values of air temperature, b) daily
Depth [em] precipitations, c) volumetric moisture registration in the
subbase and d) volumetric moisture registration at four
Figure 5. Schematic overview of the climatic sensors. depth in the subgrade.

454
 During the summer and autumn months (second not known but might be the results of the fact that
half of June to October) some peaks in moisture the field experiment is carried out in the Autumn
content can be seen. Further can their intensity been when the layer is at around 5°C and the BC layer
seen decrease with depth. They are quite intense at is polymer modified (ABb22 with Nypol 64–34).
D = 0.33 m, can easily been seen at D = 0.685 m,
1.04 m and 1.35 m but hardly at 1.75 m depth. This
peaks coincides with intense rainy periods, as can be
seen in the precipitation diagram in Fig. 6b.
In October the freezing period starts again
and that can be seen in the moisture registration
as a rapid drop in vol. moisture starting first at
D = 0.33 m and then migrating downward.
Figure 7. The vehicles consist of a steering axle and
three tridem axles. The first and the fourth axle have sin-
gle mounted wheels. All other axles are equipped with
6 HEAVY VEHICLE FIELD TEST dual mounted wheels.

The vehicles used to transport the ore material includ-

120 . - - - - - - - - - . - - - - - - - ; - - - - - - - - - - - - - ,
ing the trailers are 25 in length with a 90 tonnes gross 100 +----~---1--+----­

weight. The vehicles consist of a single wheel steer- ~80+----~-.---n~-+----,------1

ing axle and three tridem axels. The axels weights are ·~ 60 +-- - - + -11
1 × 9 + 3 × 27 = 90 tonnes. All tyres pressures are ~ 40 +----~+1
900 kPa. Figure 7 gives an overview of the vehicles. ·~ 20
~
Field measurements have been carried out
where the fully loaded 90 tones vehicles passes over 1000 2000 3000 4000
the instrumentation keeping a constant speed of Time[ms]

70 km/h. Typical examples of the sensors registra- 60 . - - - - - - - - - - - ; - - - - - - - - - - .

tion are shown in Figure 8. ~

- SPC23

Figure 8a shows a registration of one of the ASG :; 40 +-- - - - - t - - -I H -----111-1-1

~
tensile strain gages. One can see easily see the steer- t;

ing axle as well as the three triple axles. The two ·~ 20

single tyres, the steering axle and the last sub-axle >

in the first tandem axle group (the driving axle)

1000 2000 3000 4000
give the highest registrations. All other sub-axles Time [ms]
consists of dual tyres that gives lower registrations. 350 . - - - - - - - , - - - - - - - , - - - - - - - - - - - ,
Some minor accumulation of tensile strain within 300 +----4---11---~--

the axle group can further be seen. The measure- I 2so

0 200 +-- - - + - - -1
ments were carried out in late September where the ·;:; 150 +-- - - 4 - - -H
pavement surface temperature was only 5–6°C, thus E 100
~Cl
exhibiting mainly as an elastic material. so

Figure 8b shows the registration from one of the

SPC at the boundary between the UBC and Sb lay- 1000 2000 3000 4000
Time[ms]
ers at 30 cm depth. Again all the axle groups can
easily been seen. Here, there is no distinction in the
response from single vs. dual tyres. The middle sub- I 40

axle in the tandem axle groups give higher registra- 0

·;:;
20 +-------1----fl

tion that the other two sub-axles. This is attributed

to the fact of larger superposition impact from the
E
~0
a••••••
-20 +----~-~il--+----+-----1
two sub-axles in front and behind the mid-axles
compared to the other sub-axles. 1000 2000 3000 4000
Figure 8c shows the vertical deformation (Emu- Time [ms]

coils) of the BC between 2 and 8 cm depth as

the heavy vehicle passes over. Here the two axles Figure 8. Registrations from the road sensors as the
with single tyres shows clearly higher peak values heavy vehicle passes over at 70 km/h. a) horizontal lon-
gitudinal tensile strain at the bottom of the asphalt layer,
compared to the sub-axles with dual tyres. It fur- b) induced vertical stress at the bottom of the unbound
ther seems that the peaks intensity decreases with base course (d = 30 cm), c) vertical deformation over the
increased number of sub-axles until a certain equi- bituminous binder course and d) the vertical deformation
librium is reached. The reason for this behavior is over the unbound base course.

455
 Finally Figure 8d shows the vertical deformation test in late September 2013 or almost 3 months
(Emu-coils) over the UBC between 20 and 30 cm after the wearing course was placed. The FWD
depth as the heavy vehicle passes over. The signal were carried out using three load levels; that is 30,
is much noisier than the other registrations but still 50 and 65 kN. The air temperature varied between
the individual axle groups can be seen. An interest- 1 and 7°C during the testing. The results are shown
ing effect shown here is that a negative deformation in Figure 9.
(volume expansion) is registered in front and after As seen in Figure 10 the deflections increase as
the steering axle has passed the sensor. This is prob- the load level increase. Similar deflections results
ably the results of the flexural phenomenon of flexi- are obtained for structure 1, 2 and 3 indicating
ble pavements that under the wheel load the material that their layer stiffness’s are similar but struc-
compresses downwards and to the sides that pusses ture 4 gives higher deflections at d0 and d200 indi-
the material upwards in front and behind the loading cating a softer upper structure. This is attributed
area. The reason that this is not seen for the tridem to the cold emulsion large aggregate skeleton that
axles groups is probably due to their high weight as is used as a road base layer in structure 4 (see
well as their superposition, resulting in much larger Table 1).
deflection bowl under the entire truck.

8 PERFORMANCE DEVELOPMENTS
7 STRUCTURAL BEHAVIOUR
The performance of all the four structures has
The structures behavior of the sections was inves-
been monitored mainly by measuring the rut
tigated in a Falling Weight Deflectometer (FWD)
profile by a profilometer and by visual inspec-
tions for crack severity. Figure 11 shows two
Distance [mm]
rut profiles measurements out of ten for sec-
0 200 400 600 800 1000 1200
tion 01 (the test section is 100 m long and rut
0
profile is taken with 10 m interval) as measured in
September 2015. Still no cracks can be found in
the structures, except that some irregular subtle
50 thin cracks can been seen on structure 04. Their
origin is not known.
E' Figure 11 shows the average rut measured
~100
c: for all the four structures. The measurement in
0
·.;::::;
u
aJ
July 2013 is done immediately after the wearing
't 150 course was placed and can be seen as a reference
Cl - - section1 measurement. During the first year of operation
---•--- Section 2 (July 2013 until August 2014) the increase in rut
200 - -...- - Section 3 depth for the different structures varies between
- ..- - Section 4 2.8 and 3.4 mm. Thus similar rut rates takes place
250 in all the structures during the first year. During
the second year in operation (August 2014 – Sept.
Figure 9. FWD results for all test sections for three load 2015) the rut increase is between 0.2–0.4 mm per
levels. Only average values for section is shown. structure.

Section no.
4
E 0
..§. 0-f-.!...l-~--;M::...._--;:,.,..--+-,----:---.;.._,.I;fj~'YB<G........i

-2
E
..§. -4
.<:
c.
~ -6
• 03/07/2013

1000 2000 3000 4000 -8 • 20/ 08/ 2014

Lane width [m m ]
• 10/09/ 2015
-10
Figure 10. Profile measurements for section 1. Two
measurements are shown out of ten. Figure 11. Average rut depth for all four test sections.

456
9 CONCLUSIONS single vs. dual tyres. This is probably associ-
ated to a superposition of stresses from adja-
Four instrumented test road structures have been cent sub-axels in front and behind the middle
built in northern Sweden in an artic climate. The sub-axle.
structures differs in material quality ranging from • Structures S04 with the cold emulsion large
conventional to polymer modified structures. Each aggregate skeleton as a road base showed much
test section is instrumented with road sensors that higher deflections at d0 and d200 in the FWD
measure the pavement response to external loading. measurements indicating a softer upper struc-
Climatic variables are further monitored through a ture compared to structures S01–S03.
weather station as well as temperature and moisture • The average surface rut depth developments for
sensors installed in the road structure. Response the four structure is still similar. Longer monitor-
measurements have been carried out from 90 tons ing time period is needed to reveal the difference
vehicle consisting of a single wheel steering axle, in the rutting performance of the structures.
and triple dual wheel tridem axles. Further has
FWD measurements been carried out to evaluate
the structural behavior of the pavements. REFERENCES
The main conclusions from the projects are:
Erlingsson, S., Baltzer, S., Baena J. & Bjarnason, G. 2009.
• Large moisture increase takes place in all Measurements Techniques for Water Flow, In: A.
unbound layers during spring thaw that last Dawson (Ed.) Water in Road Structures—Movement,
about two months before it starts gradually Drainage & Effects, Springer, pp. 45–67.
to decrease. During the summer and autumn Erlingsson, S. & Carlsson, H. 2014. The Svappavaara
months rainy periods results in increase in mois- road test sections—instrumentation. VTI notat
ture in the unbound layers. 12 A-2014, 24 p + app.
• The response measurements show that a single Erlingsson, S. & Ahmed, A. 2015. The Svappavaara road
tyre in the tandem axel give higher horizontal test sections – Field coring and laboratory test. VTI
notat 30 A-2015, 33 p + app.
tensile strain at the bottom of asphalt and ver- Saevarsdottir, Th., Erlingsson, S. & Carlsson, H. 2016.
tical strain (deformation) in the unbound layers Instrumentation and performance modelling of
compared to a dual tyres. Vertical stress is on Heavy Vehicle Simulator tests. International Jour-
the other hand highest of the middle sub-axel nal of Pavement Engineering, 17:2, 148–165, DOI:
in the tandem axle group and not affected by 10.1080/10298436.2014.972957.

457
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Impact of overloaded vehicles on load equivalency factors and service

period of flexible pavements

D. Rys, J. Judycki & J. Jaskula

Faculty of Civil and Environmental Engineering, Gdansk University of Technology, Gdańsk, Poland

ABSTRACT: Overloaded vehicles occur less frequently in comparison to properly loaded vehicles but
due to their greater potential to cause damage they significantly contribute to the distress of pavement
structure. As the studies show, the phenomena of vehicles overloading intensifies when the control of
traffic is poor and the average percentage of overloaded vehicles on analyzed roads varies from 6% to
23%. The main goal of the article is to assess the impact of vehicles overloading on the service period of
flexible pavements. In this purpose the data from 10 weigh in motion stations on Polish national roads and
motorways, which included up to 11 million records of heavy vehicles, were analyzed. Mixed axle loads
in particular vehicles were transformed into number of equivalent standard axle loads and further into
the load equivalency factors for particular trucks (truck factors). Each vehicle was also checked for over-
loading. It was revealed that average monthly truck factors are well correlated with the average monthly
percentage of overloaded vehicles. This relationships were used to assess the impact of vehicle overload-
ing on service period of flexible pavements. The approximate method of calculation of extended fatigue
life of a pavement structure and its service period is presented and allows easy evaluation of the effect of
minimizing of percentage of overloaded vehicles.

1 INTRODUCTION Overloaded vehicles have much greater poten-

tial to cause pavement damage in comparison to
1.1 Background the properly loaded vehicles (Jeongho Oh et al.
2007). The damaging effect of overloaded vehi-
In Poland from the last two decades the traffic of
cles depends not only on their percentage in total
heavy vehicles on the main roads network have
number of trucks, but also on the probability dis-
increased rapidly. Especially in the vehicle class of five-
tribution of vehicle loads (Mohammadi and Shah,
axle trucks with semi-trailer it was noted an average
1992, Rys et al. 2016). The phenomena of vehicle
increase by five times (GDDKiA, 2010). After acces-
overloading contribute to reduction in the service
sion of Poland to the European Union in 2005 the
lives of pavements and bridges and the increase in
maximum legal weight have increased up to 115 kN
the cost associated with maintaining, upgrading
per single drive axle as a consequence of the Euro-
and replacing the highway infrastructure. Accord-
pean Union Council Directive 96/53/EC. It resulted
ing to Pais et al. (2013) the maintenance cost of
in increase of real axle loads of vehicles. Moreover
road calculated per one vehicle is higher by 100%
a significant part of vehicles exceed the legal limits
for overloaded vehicles compared to the cost of the
of gross weight and axle loads (Szydlo and Wardega,
same vehicle with legal loads. The infrastructure
2003, Rys et al., 2016, Zofka et. al., 2014).
damage costs should be recover by proper fee for
The simplest statistic to describe the problem
overloaded vehicles (Dey et. al, 2015). However the
of vehicle overloading is the percentage of over-
level of damage-cost recovery fee depend on the
loaded vehicles in the total number of trucks. It is
effectiveness in detection of overloaded vehicles
observed, that the percentage of overloaded vehi-
and it can be improve by regular vehicles control.
cles is much higher in developing countries. For
What is more, regular control reduces the percent-
example in China or Indonesia it was revealed in
age of overloaded vehicles. According to studies
studies of Zhao et al. (2012) and Mulyono and
carried by Tailor et al. (2000) in the state New York
Antameng (2010) that the percentage of over-
(USA), the percentage of overloaded vehicles was
loaded vehicles can reach extremely high level of
reduced from 30% to 2% when the enforcement
80% and for comparison in the USA and UE coun-
level was increased. On the other hand, higher
tries it is in the range from 10% to 30% (Moham-
enforcement level results in shifting of the freight
madi and Shah, 1992; Pais et al., 2013, Fiorillo and
that was being moved on overloaded vehicles
Ghosn, 2014).

459
onto vehicles loaded with accordance to regula- Table 1. Time of WIM measurement and number of
tions. However, according to the thesis stated by vehicles records used in analysis.
Stephens et al. (2003), much less pavement damage
Total number of records
is incurred in carrying freight on more vehicles that
operate legally in comparison with carrying the Raw WIM Records of
same freight on fewer vehicles than operate over records trucks (after
weight. This statement results from the fact that WIM Years of of all filtering
the damaging effect of axle increases with relative station measurements vehicles types process)
increase of its load approximately to the power
four. For pavement structure analysis and design A2 2011–2012 8,411,233 2,580,957
the damaging effect of vehicles is expressed by S7 2013 3,661,002 417,185
truck factors, which determine the average number DK4 2011–2013 17,030,795 3,340,424
of equivalent standard axle loads per one vehicle. DK7 2012–2013 11,239,857 1,089,331
DK11 2010–2011 2,968,179 811,792
DK22 2013 3,969,641 638,915
1.2 Objectives DK46 2011–2013 6,585,462 1,249,711
The objectives of this paper are to determine the DK75 2012–2013 8,873,963 861,494
relationship between the truck factor and the per- DK79 2013 3,012,197 254,237
centage of overloaded vehicles to reveal: (1) how DK94 2013 3,607,302 472,534
Total 60,948,398 11,716,580
much the pavement fatigue life decreases as a result
of overloading and (2) how much the overloaded
vehicles influence on pavement service period.
because same of them have been started after 2010
(see Table 1). In all of cases period of data col-
lecting covers whole years. The WIM stations are
2 ANALYSIS OF TRAFFIC OF
installed for each of traffic direction separately but
HEAVY VEHICLES
for further analysis data from two directions from
the same road section were analysed together as for
2.1 Collecting and verification of data from weigh
one measuring point.
in motion
The raw WIM data were verified using series of
The Weigh in Motion (WIM) system allows to filters of vehicles parameters (e.g. axle loads, total
improve vehicle control and it is developing inten- length, axle configurations etc.). The filters were set
sively in Europe (Jacob and Loo, 2008). The WIM in accordance with the WIM Data Analyst Manual
stations are installed in order to preselect over- (FHWA, 2010) and NCHRP report 538 (NCHRP,
loaded vehicles and to collect statistical data. Cur- 2005) and vehicle technical parameters review. The
rently in the UE countries overloaded vehicles must filtering process was focused on identifying and
be weighing again on the static, legalized weights to removing invalid records from the database and
impose the fine. However there are some trials to choosing only trucks to further analysis. To sum
improve WIM systems measurement precision and up more than 60 million of all types of vehicles
to solve the legislation problems to enable usage were recorded, including cars, vans etc., out of that
of WIM systems for automatic identifying and more than 11.7 million records of trucks were used
imposing the fine for overloaded vehicles (Burnos in the analysis.
and Gajda 2016, Oskarbski and Kaszubowski
2016, Doupal et al., 2008, Burnos et.al, 2007).
2.2 Scale of the problem of vehicle overloading
In the following analyses presented in this paper
in Poland
data from ten WIM stations on Polish national
roads and motorways were used. The WIM sta- The basis to classify a vehicle as overloaded was
tions are equipped in the bending plate sensors the European Union Council Directive 96/53/
PAT DAW 100® (A2 and DK11) or in the piezo- EC which specify the legal limits of vehicles gross
electric quartz sensors Kistler Lineas®. The sys- weights and axle loads. The values of legal limit
tems of automatic vehicle classification are also of vehicle gross weight depends on the class of
installed in all of the stations. The WIM stations the vehicle, e.g. for 2-axles single truck unit it is
can be classified as a class B7 according to COST equal to 18 000 kg and for 5-axles truck with semi-
323 WIM classification. The information about trailer it is equal to 40 000 kg. The legal limit of
WIM data and measurement period are given in axle loads depends on the type of the axle (steer-
the Table 1. The data considered in this study were ing, drive etc.), the distance to neighbouring axles
collected since 2010 till 2013 but the period of (single, tandem, tridem) and the suspension type.
measuring can differ for particular WIM stations For example, for single drive axles the maximum

460
 To calculate TF the damaging effects of each axles
in a given vehicle are summed up according to the
formula (1). The damaging effects of each axle in
the vehicle are expressed by equivalent axle load
factors Fj.
n
Fv = ∑ F j
TF (1)
j =1

WIM station
c Anua average percentage of overloaded vehides where: TFv – truck factor (number of equivalent
I Range of monthly GVcragc percentage of overloaded 'Jchiclc~ axles per one vehicle), j – axle number, n – total
number of axles in a given vehicle, Fj – equivalent
axle load factor for a given j axle.
Figure 1. Annual average percentage of overloaded The term of equivalent axle load factor was
vehicles in selected WIM stations in Poland. firstly developed as a result of AASHO Road Test
and it was adopted for pavement design meth-
ods in many countries. In this analysis the origi-
axle load equals 115 kN and for the rest types of nal AASHO formulas were simplified to the wide
single axles it is equal to 100 kN. known fourth power formula (2):
In the analysis, each vehicle was checked and
marked if overloaded by using the algorithms ⎛ Qj ⎞
4

developed specifically for this study at the Gdansk Fj = ⎜ ⎟ (2)

University of Technology. A vehicle was treated ⎝ Qs ⎠
as overloaded according to EU Council Directive
96/53/EC in the following cases: where: Fj – equivalent axle load factor, Qj – actual
• The vehicle gross weight was greater than the axle load (in the case of multiple axles sum of
legal limit; component axle loads), Qs – standard axle load
• The load of a single or tandem or tridem axle Qs = 100 kN for single axles, Qs = 184 kN for tan-
was greater than legal limit; dem, Qs = 263 kN for tridem. The loads of 184 kN
• Both the gross weight and the axle load were and 263 kN, used in the denominators of formula
greater than legal limits. (1), was calculated from the AASHTO (1993)
equations for terminal level of serviceability
The percentage of overloaded vehicles (abbrevi- pt = 2.5 and structure number SN = 5.15 as loads
ated further as OV) was calculated for each station. for tandem or tridem axles which are equivalent to
The annual average percentage of overloaded vehi- standard axle load 100 kN (Judycki, 2006).
cles and the range of the monthly average percent- Truck factors TF were calculated for each prop-
age of overloaded vehicles are presented in Figure 1. erly weighted vehicles and each of this vehicles
As it can be seen in the Figure 1 the annual aver- were recognized in terms of overloading. For fur-
age percentage of overloaded vehicles ranges from ther analysis, the average monthly truck factors
6% to 22%. However the more detailed analysis were calculated for each of WIM stations.
shows, that the percentage of overloaded vehicles
varies in following days, weeks and months. The
level of enforcement has a significant impact on 3.2 Relationship between truck factors and the
this statistic. It is known from the interview that percentage of overloaded vehicles
control on DK11 is performed almost every day,
The percentage of overloaded vehicles OV and
otherwise on S7 and A2 the control is performed
average truck factors TF can be treated as parame-
hardly ever. The frequency of control of over-
ters, which characterize the traffic condition, and it
loaded vehicles also differs in particular months.
was observed that this parameters varies in follow-
ing days, months and years. The variations of TF
and OV can be caused by several factors includ-
3 CALCULATIONS OF THE IMPACT OF
ing frequency of traffic controls, goods transport
OVERLOADED VEHICLES ON FATIGUE
demand, economic condition or seasonal traffic
LIFE OF FLEXIBLE PAVEMENTS
variations. The analysis have shown that the per-
centage of overloaded vehicles correlate with the
3.1 Truck factors and equivalent axle load factors
average truck factors and following linear regres-
Truck Factors (TF) characterises the damaging sion model can be determine for each WIM
effect of a given truck to pavement structures. stations:

461
TF = a OV b (3) It can be concluded that the relationship
between the truck factor TF and the percentage of
where: TF – average truck factor, OV – percentage overloaded vehicles OV is strong with the coeffi-
of overloaded vehicle, a, b – regression model fac- cient of determination R2 between 0,76 and 0,99.
tors, specified for a given WIM station. Additional analyses revealed that OV do not corre-
The relationships between the average truck fac- late with average daily traffic nor with the percent-
tor TF and percentage of overloaded vehicle OV age of trucks in the whole traffic flow (Rys, 2015).
are given in the Figure 2 for each considered WIM
station. The points marked in each of graphs in the 3.3 Decrease of fatigue life of pavement
Figure 2 represent average monthly values of TF structures caused by overloaded vehicles
and OV so the total number of this points equals to
the number of months at measurement period and To calculate the Decrease of Fatigue Life (DFL) of a
it can vary for particular WIM stations (see Table 1). pavement structure due to overloaded vehicles it was

A2 57
1,5 1,5

~ :+--
1,0 1,0

Tt
....
-u ~ !-!+-
y=2,3656x +0,5142
f+ -~
fl-1 ;!,;; pu t:
0,5 0,5
R2 = 0 9905 1-J. J Ll-1-. f-1- y = 2,2529x
0,0
itttt 11 · .rr 0,0
I R2
+
= 0,7611
0,3537

0% 10% 20% 30% 0% 10% 20% 30%

ov ov
DK4 DK7
1,5 1,5

n
1,0 ~ 1,0

1!: -!+-+-
f~ rr ~
1-
0,5 ttl 0,5

:ct rn y = R 2,9375x + 0,2672

2 0,9314
0,0
0% 10% 20% 30%
ov

1,5

1,0

1!:
0.5

ov ov
DK46
1,5

1,0
u
0,5 ~n ~

0,0
-'_.):!:_

0%
.ta 10%
y =2 , 1914x + 0 ,3747
R2 = 0,9261
20% 30%
ov ov
DK94
1,5

1,0
+R- .W+ fLLll
~I ±
~--
0,5
y = 1,5388x + 0,4027
R2 = 0 ,9458
0,0
0% 10% 20% 30%
ov ov

Figure 2. Relationships between average monthly truck factor TF and average monthly percentage of overloaded
vehicles OV.

462
assumed that the fatigue life of this structure is equal 4 EXTENSION OF RESIDUAL FATIGUE
to N100 equivalent standard 100 kN axle loads and LIFE AND SERVICE PERIOD OF
the given pavement can carry in one case NT0 trucks FLEXIBLE PAVEMENT STRUCTURES
with 0% of overloaded vehicles and in the second case DUE TO REDUCTION OF NUMBER OF
NTOV trucks with OV percent of overloaded vehicles OVERLOADED VEHICLES
till failure. The Truck Factors are equal respectively
TF0 and TFOV. The following equation is valid: On the base of equation (4) and Figure 3 the
following simple approach is proposed to esti-
N100 NT
T0 T
TFF0 NT
NTOV T
TFFOV (4) mate the effect of reduction in percentage of
overloaded vehicles on the extension of residual
The relative Decrease of Fatigue Life DFL due fatigue life and service period of flexible pave-
to OV percent of overloaded vehicles is equal to: ment structure. The fatigue life depends on the
truck factor TF and total number of trucks NT
T0 − NT
NT TOV in the whole service period. Better traffic con-
F =
DFL (5) trol may cause reduction in percentage of over-
NTT0
loaded vehicles thus reducing the value of the
truck factor TF. If the truck factor TF decrease
thus: as a consequence of decrease of the percentage
of overloaded vehicles OV, the total number of
TFF0 trucks NT, which the pavement can carry till
F = 1−
DFL (6)
TF
FOV fatigue, will increase and consequently, the resid-
ual fatigue life and service period will extend. To
where: DFL – Decrease of Fatigue Life caused by evaluate the extension of residual fatigue life and
overloaded vehicles, NT0 – number of trucks with service period of flexible pavement structures
0% of overloaded vehicles till failure of the pave- due to decrease of number of overloaded vehi-
ment, NTOV – number trucks with OV percent of cles let’s assume that the residual fatigue life of a
overloaded vehicles till failure of the pavement, TF0 given pavement was assessed on the base of meas-
– Truck Factor at 0% of overloaded vehicles, TFOV – urement obtained from the FWD test. The per-
Truck Factors at any percent of overloaded vehicles centage of overloaded vehicles OV for that road
OV. The relationship between DFL and percentage is known. To calculate the residual fatigue life
of overloaded vehicles OV is shown in Figure 3. and service period of the pavement in case when
Figure 3 shows that in average the decrease in number of overloaded vehicles will be reduced by
fatigue life of a pavement structure will be equal to ΔOV the following formula can be used:
65% if percentage of overloaded vehicles increases
from 0% till 25%. The Figure 3 can be used to eval- 1 − DFL
FLOV − ΔOV
RFL
FLOV − ΔOV LOOV −
RFL
RFL (7)
uate the detrimental effect of overloaded vehicles if 1 − DFL
FLOV
their percentage is known. However the exact dam-
aging effect of overloaded vehicles is dependent where: RFLOV-ΔOV – Residual Fatigue Life,
upon the axle load distribution, and somehow on expressed as number of Equivalent Single Axle
the type and parameters of a pavement structure. Loads, after reduction of overloaded vehicles
from OV to (OV - ΔOV) percent, RFLOV – Resid-
100% ual Fatigue Life, expressed as number of Equiva-
I I
J_-- lent Single Axle Loads, at the existing number of
80%
- - I
overloaded vehicles OV, DFLOV-ΔOV – Decrease of
~ BOo/o
e...
--'
u.
Cl 40'Al ,
,
!"'
,

-I
-- -.
.
I
.
f!
I
+- Fatigue Life at reduced percentage of overloaded
vehicle OV-ΔOV, DFLOV – Decrease of Fatigue
Life at existing percentage of overload vehicles OV.
The DFL value (Decrease of Fatigue Life) is given
20%
I
II ..... II in Figure 3, as a function of percentage of over-
loaded vehicles OV.
0% r I I
Similarly the extended residual service period
0% 5% 10% 15 % 20% 25%
OV [%] of a pavement after reduction of number of over-
----- maxiumum (DK79) - - averaqe loaded vehicles by ΔOV can be calculated from the
- •- minimum (nKQ4)
formula:
Figure 3. Relationships between the decrease of fatigue 1 − DFL
FLOV − ΔOV
life of pavement structure DFL and the percentage of RSP
SPOV − ΔOV POOV −
RSP
RSP (8)
overloaded vehicles OV. 1 − DFL
FLOV

463
where: RSPOV-ΔOV—Residual Service Period, percentage of overloaded vehicles, the degree of
expressed in years, after reduction of overloaded overloading, pavement structure and materials
vehicles from OV to (OV—ΔOV) percent, RFLOV— used for pavement layers, climatic conditions, etc.
Residual Service Period, expressed in years, at the The traffic composition and characterization as
existing number of overloaded vehicles OV. Both measured in WIM stations is presented in Table 2.
values Residual Fatigue Life and Residual Service
Period will increase by the Increase Factor equal to:
5 CONCLUSIONS
1 − DFL
FLOV − ΔOV
IF = (9)
1 − DFL
FLOV 1. Data delivered from ten weigh in motion stations
in Poland indicated that the annual average per-
where IF is the Increase Factor, other abbrevia- centage of overloaded vehicles ranges from 6%
tions as above. to 22%. The more detailed analysis have shown,
The Decrease of Fatigue Life varies in relation that the percentage of overloaded vehicles var-
to Truck Factor according to equation (6) which ies in following days, weeks and months.
in turn is different for every road (Table 2). For 2. The level of vehicle control has an impact on
approximate calculations, the average value from the percentage of overloaded vehicles. The low-
Fig. 3 may be taken. As an example assume that est percentage was on a road where continuous
for a given road the FWD testing indicated Resid- control of heavy traffic was performed, the
ual Service Period equal to 10 years. The percent- highest was on roads where the control was rare.
age of overloaded vehicles for the road is equal to 3. It was found that the Truck Factor which char-
OV = 20%. Due to improved traffic control this acterizes the aggressiveness of an average vehi-
percentage will be reduced to OV – ΔOV = 5%. cle to pavements structure is very well correlated
From Fig. 3 we read that at OV = 20% the with percentage of overloaded vehicles. This
Decrease of Fatigue Life is DFL20% = 0,60, and for statistical relationships have been determined
OV−ΔOV = 5% the DFL5% = 0,29. Thus Residual and used for analysis of the impact of over-
Service Period will be equal to: loaded vehicles on the fatigue life of pavement
structure and pavement
1 − 0, 29 4. The overloaded vehicles affect significantly
RSP
S 10 18 [ years
y ] (10) the fatigue life of pavement structures. It was
1 − 0, 60
revealed that in average the decrease in fatigue
life of a pavement structure will be equal to 65%
The calculated Residual Service Period will be
if percentage of overloaded vehicles increases
extended from 10 to 18 years.
from 0% till 25%.
The presented method is approximate but
5. The approximate method of calculation of
may well suit a purpose to evaluate the effective-
extended fatigue life of a pavement structure
ness of traffic control measures. The exact value
and its service period is presented and allows
of increase of fatigue life and service period due
easy evaluation of the effect of minimizing of
to decrease of percentage of overloaded vehicles
percentage of overloaded vehicles.
is dependent on several factors which vary from
one road to another. These factors are: composi-
tion and characterization of heavy traffic, actual
REFERENCES
Table 2. Summary of the data from WIM stations.
American Association of State Highway and Transporta-
AADT0 TFactual tion Officials (AASHTO), 1993. The AASHTO Guide
WIM [trucks per OVactual [100 kN axles for Design of Pavement Structures. Washington.
Station day] [%] per truck] Burnos P., Gajda J., Piwowar P., Sroka R., Stencel M.,
Zeglen T., Accurate, 2007. Weighing of Moving Vehi-
A2 2262 16 0.898 cles, Metrology and Measurement Systems, vol.14,
S7 1332 22 0.857 no.4, pp.507–516.
DK4 2557 15 0.718 Burnos P., Gajda J., 2016. Thermal Property Analysis of
DK7 895 16 0.703 Axle Load Sensors for Weighing Vehicles in Weigh-
DK11 1194 6 0.612 in-Motion System, Sensors, vol.16, issue 12. (DOI:
DK22 934 15 0.695 10.3390/s16122143).
DK46 748 14 0.684 COST 323, 2002: Weigh-in-Motion of Road Vehicles.
DK75 790 11 0.548 Final Report. Appendix 1. European WIM Specifica-
DK79 745 17 0.557 tions. LCPC, Paris.
DK94 731 11 0.570 Dey, K., Chowdhury, M., Wiecek, M., and Dunning,
A., 2015. Infrastructure Damage-Cost-Recovery Fee

464
 for Overweight Trucks: Tradeoff Analysis Frame- ment in urban areas. 2nd International Conference
work. Journal of Transportation Engineering. (DOI: Green Cities—Green Logistics For Greener Cities,
10.1061/(ASCE)TE.1943–5436.0000769,04015008). Transportation Research Procedia, vol.16, 449–463
Doupal E., Calderara R., 2008. Combined LS & HS WIM (DOI: 10.1016/j.trpro.2016.11.042).
system for law enforcement and toll road applications. Pais J.C., Amorim S.I.R., Minhoto M.J.C., 2013. Impact
Proceedings of the 10th International Conference on of Traffic Overload on Road Pavement Performanc.
Heavy Vehicle, France. Journal of transportation Engineering, 139(9), 873–
Federal Highway Administration (FHWA), 2010. WIM 879. (DOI: 10.1061/(ASCE)TE.1943-5436.0000571).
Data Analyst’s Manual. Publication No. FHWA-37 Rys D., Judycki J., Jaskula P., 2016. Determination of
IF-10-018. Washington. vehicles load equivalency factors for polish catalogue
Fiorillo G., Ghosn M., 2014. Procedure for statistical cat- of typical flexible and semi-rigid pavement structures.
egorization of overweight vehicles in a WIM database. Transport Research Arena TRA2016. Transportation
Journal of Transportation Engineering, 140(5). (DOI: Research Procedia, vol. 14, 2382–2391.
10.1061/(ASCE)TE.1943-5436.0000655). Rys. D., Judycki J., Jaskula P. 2016. Analysis of effect of
General Directorate for National Roads and Motorways overloaded vehicles on fatigue life of flexible pavements
(GDDKiA), 2010. The synthesis of the general traffic based on weigh in motion (WIM) data. International
measurement. Warsaw. (Available in the Internet: Journal of Pavement Engineering. vol.17, Issue 8. doi:
www.gddkia.gov.pl/pl/987/gpr-2010). 10.1080/10298436.2015.1019493.
Jacob B. Loo H., 2008. Weight in motion for enforcement Rys D., 2015. Loading of roads by heavy vehicles and their
in Europe. 10th International Symposium on Heavy impact on fatigue life of flexible and semi-rigid pave-
Vehicle Transportation Technology HVTT10, France, ment structures. PhD Thesis. Gdansk University of
2008. Technology, 2015.
Jeongho Oh A.M, Fernando E.G., Lytton R.L Szydlo A., Wardega R., 2003. The comparison of the vehi-
2007. Evaluation of damage potential for pave- cles aggressiveness factors in Poland and in the Euro-
ments due to overweight truck traffic. Journal of pean Union. Drogownictwo no 5/2003 (in Polish).
Transportation Engineering, 133, (DOI: 10.1061/ Stephens J., Carson J., Hult D.A., Bisom D., 2003. Pres-
(ASCE)0733-947X(2007)133:5(308)). ervation of infrastructure by using weight-in-motion
Judycki J. 2006. Bases for determination of the equivalent coordinated weight enforcement. Transportation
axle load factors for design of road design of road pave- Research Record 1855, 143–150. TRB, 2003. (DOI:
ments. Drogi i Mosty (Road and Bridges) no 2/2006. 10.3141/1855–18).
(in Polish). Tailor B., Bergan A., Lindgren N., Berthelot C.,2000.
Mulyono A.T., Antameng M. 2010. Analysis of loss cost The Importance of commercial vehicle weight enforce-
of road pavement distress due to overloading freight ment in safety and road asset management. Traffic
transportation. Journal of the Eastern Asia Society for Technology International.
Transportation Studies, Vol. 8. Zofka A., Urbaniak A., Maliszewski M., Bankowski
Mohammadi J., Shah N., 1992. Statistical evalua- W., Sybilski D., 2014. Site specific traffic inputs for
tion of truck overloads. Journal of Transportation mechanistic-empirical pavement design guide in Poland.
Engineering, 118, 651–665. ASCE. (DOI: 10.1061/ Transportation Research Board, 93rd Annual Meeting,
(ASCE)0733-947X(1992)118:5(651)). Washington., USA.
National Cooperative Highway Research Program Zhao Y., Tan Y., Zhou C., 2012. Determination of
NCHRP, 2005. Traffic Data Collection, Analysis, and axle load spectra based on percentage of over-
Forecasting for Mechanistic Pavement Design. Report loaded trucks for mechanistic-empirical pavement
538. design. Journal of Road Materials and Pavement
Oskarbski J., Kaszubowski D., 2016. Implementation Design, Vol. 13, nr 4, 850–863, 2012. (DOI:
of Weigh-in-Motion system in freight traffic manage- 10.1080/14680629.2012.735796).

465
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Implications of changing the maximum legal truck load

for the pavement service life

Carl A. Lenngren
Sweco Civil, Sweden

Reus Salini
Neogennium Technologies, Brazil

ABSTRACT: Some major roads in Sweden are being designed to carry a higher vehicle load of 74 met-
ric tons. This is almost 25% higher than the previous maximum legal load. There is no change of the
maximum 10 ton axle load, and the 11.5 ton driving axle. Thus, in the elastic world, the change would not
incur any changes in the pavement design, as the number of equivalent axle loads will remain the same,
provided the freight levels are the same. However, the experience from a previous change of the legal loads
showed clearly that sub base layers tended to deform at a faster rate. Hence, the specifications for sub base
layer materials were changed with tougher criteria for the gradation curve. The present paper compiles
experience from a failed mining truck road to address issues that may affect specifications for unbound
materials.

1 INTRODUCTION For a while, there were also strengthening projects

going on, particularly on the smaller roads. More
The legal limits e.g. truck axle loads is governed importantly, the rather generous maximum gross
by the vehicle design, which is in turn decided on a weight was increased from about 51 tons to 60 tons
common standard. Thus, bridges and roads can be for truck and trailer combinations. The maximum
optimized to such norms. axle load remained at ten tons for non-driving axles.
Every so often, there is a demand from the indus- The change went rather smoothly, and the trucking
try to raise the standard to a higher level. The rea- industry quickly phased out older three-axle trailers
son is often from an economic aspect, or it may be to four axle ones, which was one of the tax incentives.
driven simply by bigger trucks, being available on However, some newer roads suffered from exces-
foreign markets. One major concern is the bridge sive rutting. It turned out the most of this rutting
design, which is optimized for a certain maximum occurred rather deep in the sub base layer, and that
load limit that cannot be changed without major those layers consisted of sand. Thus, the pavement
investments. For pavement design, there is a general specifications were changed to favor crushed mate-
rule of the number of equivalent axle loads, which rials, which also became mandatory in the higher
is widely used. Thus, an increase of loads could be traffic classes. Even if scarcity of materials, which
calculated as accelerated wear and tear per vehicle. increased costs, the stricter specifications solved
In the 1980:ies, with an effort to harmonize the deterioration problem.
European transportation standards, there was Now, about thirty years later there is a proposal
a change of the truck axle loads and wheel con- to increase the maximum gross weight to 74 tons,
figurations. In Sweden, the gross weight was not without increasing the axle load. The rational being
affected by this. However, there were other changes more payload in less number of vehicles. Trucks
like allowing 11.5 tons on the driving axle, and would be longer, but highways have improved, so
using wide tires, and three axle combinations with the need for passenger cars to pass in a lane with
higher loads than previously. Within a few years, opposing traffic is much less frequent today. A
all bridges were either replaced or rebuilt on all check with proposed axle spacing on bridges has
the national roads. There was an overview of the been approved. A typical 74-ton combination
design classes for pavements. In essence, there was would consist of nine axles. 8 + 11.5T + 3x18T per-
a shift to a higher class for each traffic category, haps. A four-axle trailer 2x18T could be replaced
and some of the lower volume classes were aban- by a five-axle ditto carrying 18 tons on a tandem
doned altogether. and 22 tons on a triple combination.

467
 In a linear elastic model, there is no difference in A hint may be to examine a mining truck road
pavement wear from the proposal, as the number that was exposed to exempted heavy loads, before
of axle passes would remain constant. In reality, construction was completed. It failed only a few
the rest period between loads is shorter, which weeks after it was first used, and the data from
would likely increase fatigue damage, and rutting FWD tests and sampling are used to find a solu-
as well. In the Swedish Pavement Design Code tion to the failure.
(2011), there is uncertainty risk factor, ranging
from 1.0 to 1.15 depending on the road category.
With the proposed new gross weight, a factor of 3 OBJECTIVE
1.25 should be used instead, to account for the
extra wear and tear. However, with the previous The objective with the present paper is to find out
experience in mind it may not be so easy as to fac- if there are other concerns of pavement design cri-
tor in a number. Rather, there is an obvious risk for teria than the traditional top of the subgrade strain
exceeding the strength of some unbound materials. and the maximum strain in the bound layers. In
this context, the concern is the unbound base and
sub base layers only. Typically, yield criteria for soil
2 INITIATIVES FROM OTHER can be used as Mohr-Coulomb, shake down limits
COUNTRIES et cetera.

According to McKinnon (2005) and based on UK

data, there are advantages of rising the maximum 4 HIGH LOAD PAVEMENT DAMAGE
truck weight on traffic levels, road haulage costs
and emissions. The study took account of three The change of the maximum gross weight is being
key factors: the migration of loads to heavier vehi- introduced without any thorough full scale testing.
cles, a traffic generation effect and the diversion of It is assumed that the elastic model is appropri-
freight from the rail network. The net reduction in ate for this purpose, and as an extra precaution, a
trucking kilo meters by 2003 was at the upper end fudge factor is introduced in the design. However,
of the forecast range. A comparison with the issue one can get experience from similar situations with
in the United States was also provided. high loads on industrial pavements, harbors and
In 2013 Finland allowed the maximum weight similar facilities.
limit of trucks increased up to 76 tons, although Fin- One striking example of load related pavement
land, like Sweden, already differed from most EU damage is a 20 km long section of Highway 99 in
countries by using EU’s exemption to allow longer Northern Sweden. It was being reconstructed in
and heavier trucks to operate in national roads. The 2015 to accommodate mining truck trailer com-
Finnish government hopes to achieve both eco- binations weighing as much as 90 metric tons. In
nomic and environmental benefits in the road freight this remote area, traffic is sparse, so the new design
sector by raising the total load, but at the same time, called for a substantial increase in bearing capacity.
Finnish road infrastructure has been reported to be It is in a cold region and the frost index amounts to
in poor condition. As a result, the Finnish haulers 1500 degree-days Celsius, according to 1960–1990
adopted higher maximum payloads mostly over a climate data.
period of one year. In 2015, the cost savings of HCT The design called for six years of 20-ton axle
trucks were about 58 million euro and CO2 savings loads, corresponding to six million 10-ton stand-
about 0.07 Mt or 3.7% of total truck CO2 emis- ard axle loads. The design bound layer thickness
sions in Finland. These savings are significant, but design was 150 mm, but traffic was permitted with
if the savings will not increase in the future, they will only 50 mm of asphalt concrete in place. Soon
remain significantly lower than was estimated prior enough, the deterioration became evident, and
to the change (Liimatainen & Nykänen 2014). an independent investigation was initiated. It was
Increasing the maximum truck-trailer load has divided into two parts being 8.4 and 11 km long
also implications for bridges. A formula for limit- respectively.
ing truck and combination vehicle weights is pro-
posed by James and others (James et al, 1986). The
proposed formula gives the maximum allowable 5 FIELD TESTING
gross vehicle weight as a function of extreme axle
spacing. It is intended to replace the existing for- 5.1 Testing
mula, which depends on the number of axles in the
During 2015 several tests were done:
string as well as the extreme axle spacing. Accord-
ing to the authors, there are several benefits of the • Ground Penetrating Radar (GPR)
proposed formula for the bridges protection. • Visual Survey with Frost Damage Inspection

468
• Falling Weight Deflectometer Testing (FWD) Table 2. Sections ranked after low base stiffness (part 1).
• 300 mm Bore Coring
• Two Full Width Forensic Cuts Section Modulus Section Modulus

7452 51 3652 86
5.1.1 Ground Penetrating Radar 7024 71 6974 87
The GPR tests were used to confirm the asphalt 3375 78 7375 94
layer thicknesses. At the time of measurement, only 3528 80 3652 94
one 50 mm thick layer was placed. It was found to 3428 83 3326 97
vary from 30 to 80 mm. About 10% of the entire 7515 84 1075 99
length was recorded as less than 50 mm. 7425 84 802 99

5.1.2 Visual survey with frost damage

inspection Table 3. Sections ranked after low base stiffness (part 2).
The visual survey noted four potholes, and some
10–30 mm wide cracks. There were five incidences Section Modulus Section Modulus
of fatigue cracking and one settlement. Severe
rutting was found at three locations. This inspec- 11103 53 13672 76
tion did not find any damage from frost actions. 14896 56 7615 81
17923 64 13623 85
5.1.3 FWD testing 12572 71 17222 86
A standard test method using the nominal load 12672 72 14847 87
of 50 kN was used. In addition, time histories 14896 73 17572 97
were saved to disc. The testing took place in June 17522 76 13173 97
of 2015 during two days, and the pavement tem-
perature was around 15 degrees Celsius. A total of
845 sections were tested. Table 4. Sections ranked after low subbase stiffness
(part 1).
5.1.4 FWD back calculation and analysis Section Modulus Section Modulus
A back calculation analysis was done with four lay-
ers. A thin 50 mm pavement resting on a 100 mm 7616 84 7565 105
unbound base course and a 600 mm thick sub 7641 90 7591 105
base. The GPR indicated some variability of the 7606 98 6325 106
pavement layers. In order to compare strains at 1550 100 8094 109
certain depths, the layer thicknesses were kept con- 7743 100 7765 112
stant nonetheless. 1600 102 6600 113
The asphalt layer modulus was found to vary. 6325 103 6248 114
From experience, this is quite common as a small 7692 103 8170 116
variation of thickness affects the stress situation in
the material. The focus here is on the unbound lay-
ers and the median moduli are shown in Table 1. As can be seen the median values provide a
As mentioned previously, the test is divided into rather high bearing capacity. However, there are
two parts. E(2) refers to modulus of layer two quite a few exceptions. Table 2 and 3 show the base
counted from the surface, i.e. the unbound base. and Table 4 and 5 show the sub base layer modulus
E(3) is the modulus of the sub base layer. E(4) is sections sorted from lowest ones.
the sub grade modulus. Evidently, the median stiffness-values are quite
good, likely a result of compaction from the heavy
Table 1. Unbound Layer Moduli. trucks. Nevertheless, a large portion of the length
of the road lacks bearing capacity. It seems the
Section materials lost their stability somehow. It could
be that the materials failed once a certain level of
Modulus 1 2 stress is obtained.
Figure 1 shows the vertical strains in the middle
E(2) median 502 464
of the base layer sorted from worst to best. Note
E(3) median 199 190
that some stations are in tension. The correspond-
E(4) median 209 185
ing sub base layer strain at the same testing points
E(2) 10%-ile 170 256
E(3) 10%-ile 190 149
is shown as Strain(3). There is no good correlation
E(4) 10%-ile 185 144
between the two. The coefficient of determina-
tion R2 is about 0.2. It is interesting to see that the

469
Table 5. Sections ranked after low subbase stiffness 5.2 Destructive testing
(part 2).
Over 100 hundred bore core samples were made
Section Modulus Section Modulus and the material was brought to the laboratory for
sieving. As it turned out most of the samples had
14972 83 16597 109 too much fines in them and the gradation curves
16822 92 14822 109 were outside the permitted area. This did not com-
16796 93 16847 110 pare with the constructer’s self-quality control
16796 95 16572 110 tests, which were on target. Assuming, same source
14772 98 20033 111 material for both tests, the grains must have been
14872 100 10981 112 subjected to mechanical wear. Micro-Deval and
14920 105 17396 114 Los Angeles Drum Tests (LA) were taken from
10981 105 15372 118 materials in the Base and Subbase, and there was
quite some variability as shown on Table 6.
1000
1000
0
0
c:
c:
-~ -1000 ·e -10oo
eu -2000 ~
.!;i -2000
~ :E
- -Strain(2)
-3000 -3000
- - Strain(3)
-4000 -4000

Figure 2. Strains in base (2) and subbase (3) layers,

Figure 1. Strains in base (2) and subbase (3) layers, sorted after (3).
sorted after (2).

distribution is not linear. The shift into tension

occurs where the pavement failed and the hori-
zontal stress is larger than the vertical stress. More
interesting is that the strains higher than 1000 micro
strain in compression seem to have a wider spread,
maybe because the materials start to deform.
Figure 2 shows the same plot, but now sorted
after the sub base vertical strain. The confining
pressure is much higher than for the base, so the Displacement [mu)
“bend” at the left side of the curve occurs at a
higher strain of about 1700 micro strain. At the
right side the lower strains occur when the strain is Figure 3. Load-deflection diagram indicates dissipa-
higher in the base layer. This could mean that there tion in the sub base.
is more dissipation higher up; energy that is used Table 6. LA and micro-deval samples.
for post compaction.
From load-deflection charts, one may estimate Base layer Subbase
the internal losses from the FWD and corre-
sponding truckload. On the highway, it is usually Section Micro- LA Modulus LA Modulus
in the range of 3–5 Nm on asphalt concrete roads. (m) Deval % % (MPa) % (MPa)
In this case, it is about 16 Nm at section 14972,
where the sub base stiffness was only 83 MPa. 2129 7 33 193
The graphs from three sensors at this section is 3296 – – – 6 251
5762 15 23 178
shown in Figure 3. Note that the area from the
5826 – – 6 188
D30 sensor is almost as large as the D0 sensor,
6733 6 – 168
which indicates that most dissipation occurs in
8279 – – – 6 251
the sub base layer here. The highest dissipation
11066 10 18 298 15 170
found was a whopping 44 Nm, where all layers
13830 13 21 250 16 175
deformed for a total surface displacement of 14210 11 20 214 14 255
almost three millimeters.

470
 The Micro-Deval values are required to be less loads, in the Table 8. The two simulated axles have
than 20% and the LA values less than 40%. Most a separation of 3 meters and the configuration of
of these samples were taken on sites that had bet- the four loads (tires) can be seen in Figure 4 and
ter than average values for the E-modulus. As the Figure 5. Figure 5 also shows the rotation of grains
materials evidently had been worn down to finer under the load, something that could have contrib-
gradation, one should reconsider the LA and uted to the wear of the individual particles.
Micro Deval limits for super heavy loads. The Figure 6 through 17 are showing the results
of the FEM for stress, shear stress and Y displace-
ment. It is possible to see that there is no relevant
6 FINITE ELEMENT ANALYSIS overlap among the stress and shear-stress for all
cases of structures and loads for the considered
From the FWD testing we saw a great variability of axles separation (3.00 m), however, there is some
the stresses in the unbound layers. We have to deal overlap for Y displacement, i.e., in the horizontal
with on-going processes. To get a better picture of direction perpendicular to the traffic lane. Dis-
the normal stress, shear-stresses and displacements placement in other directions were found as irrel-
a Finite Element Modelling (FEM) was done for evant for the purposes of this study.
four cases, by considering two different structures,
with the averages and second worst modulus, and

•- •
two different loads, for the standard truck axle and
a typical mining truck axle. The characteristics of
such structures are shown in the Table 7 and the
3.00m

Table 7. Pavement structure characteristics.

Modulus (MPa)

Layer Thickness (mm) Average Second worst

AC

Sub-base
(granular)
50
Base (granular) 100
600
10000
464
190
10000
56
92
·. .\. ~ .'.(1\j ..
/
Subgrade – 185 88 . . 0 0 .·
Table 8. Axle load characteristics.
lj,r·~~~·rl•vZ
/
··.· ~
~
~---
Axle type Grain 1 Grain 2

Characteristics Standard Exempt Figure 5. Grains rotation under the load.

Number of axles 2 2
Axles separation (m) 3.00 3.00 Vertical plane parallel to X-Z

Tires separation (mm)

Load per axle (kN)
350
100
350
180
" ... ., .
y\:.;;; .,, .. .. .:. ;;::-'. !./
'.. I Normai StressZZ

Load per tire (kN) 25 45 I I I 7

Tire type Regular Exempt (non-off-road)

.. II I !1 ...
-38

. II I I -129

·--·
- 175
. ,......'J I II - -221
Tire pressure (kPa) 552 862 -266
l• ••
E' -_,'"'-1+--+-+-

·-·-••.....
-312
Tire pressure (PSI) 80 125 11 • • •• I -357
~ -400·++-+-+-

-·-·••
Tire radius (mm) 120 129 N ltr• . , fJ .....
-403

-494
-539

....,,.
-585

E
•
SOD -722
.§. 350 __,,
-767

>- -ass
111111111111
1DDD 2DDD 3DDD Avtra,emodulus
500 1DOD 1500 2000 2500 3000 Minina; ti r e
X(mm) X (mm )
-SOD
Figure 6. Normal stresses for structure with average
Figure 4. Tires position (load configuration). modulus and mining tire load.

471
 Vertical plane parallel to X-Z Ve rtical plane parallel to X-Z

-50
Normal Stress ZZ Sht9r Stress VZ
- 100
38.7
- 15:) -24 36.6
-53 3<.5
-zoo -83 32.5
- Z5:l -112 30.0
-U1 28.3
~ -300 -171 26.3

~
-200 24.2
- 3"' -229 22.1
...____. - 400 -259 20.1
-208 18.0
N - 450 -3 17 15.9
-3<7 13.9
-500 -376 11.8

-
-55:) -405 9.7
-435 7.7
- 600 5.6
-493 3.5
-6"' -S23
-S52
1.5
-<J.6
-700
-75:)
500 1000 1500 2000 2~00 3000 Avera, emodulus AvHa,e modulus
Re,ull!lrtire M ini ng tire
X ( mm) X (mm)

Figure 7. Normal stresses for structure with average Figure 10. Shear-stress for average modulus structure
modulus and regular tire load. and mining tire load.

Vertical plan e paralle l to X-Z Vertical plane parallel l o X-Z

\\ '" I IIII. ;~ '· i 'l

- 50 I

- 100 I I II I Normai SiressZZ

Shear Stress YZ

I I II 7

...
20.6
-1 5:) - 36
- 200 I I II -129
19.5
18.4

- 250 II I -115 162

I
.,., '
II -221 15.1
~ - 300 -266

~ - 350
I I L -312 12.9

·-·-...•• -·-·• •
-357
...____. - 400 1• •• ...._.,.
-403
11.8
10.7

N - 450 1••• ~ 9 .6

- 500 Mr• • .g -539

8.5
7 .0

- 550
- 600
..,.
-585
-631
-722
6 .3
5.2
0.1
3.0
-767 1.9
- 650 -813 0.8
- 700
- 75:)
500 1()00
X
1500
(mm)
200D 25 00
•• 3000
Second worst modul us
Miningtire 500 1000
X
1500
(mm)
2000 2 500 3000
-0 .3

Alferage modu lus

Regul ar tire

Figure 8. Normal stresses for structure with second Figure 11. Shear-stress for average modulus structure
worst modulus and mining tire load. and regular tire load.

Vertical plane parallel to X-Z Ve rtical plane parallel to X-Z

<II'' I
I "
.. I I I
I
I
_,.5
Shear StressVZ

• I

.··!!··-••••,. .,...·-·l-·••.
I 60.3
I ..,
56.8

••
•• I U:
I -53

-112
53.3
49.8

_I 06.3

•••.,:
-141

•••
42.8
-171 39.2
'E -3"'* -+.f- -200
-229 'E - 30>()-!-\-'IJ-+++* 35.7
32.2
-S -<oo-H-11-f- -259
-288 -S -40<)-ft-1:+/'+t-!f- 28.7
25.2
-317
2 1.7
-3<7
18.1

-
14.6

••
-405
11.1
-435

••
....
7.6
_.,.
••·: ....
••
0,1

500 1000 1500 2000 2 500

•• 3000
Second worst modulus
Regul artirt
-523
-552

500 1 000 1 500 2000 2500 3000 s~cond worst modul us

0.6

....
-2.9

X ( mm) M inin&tire
X (mm)

Figure 9. Normal stresses for structure with second Figure 12. Shear-stress for second worst modulus struc-
worst modulus and regular tire load. ture and mining tire load.

472
 Vertical plane parallel to X-Z Vertical pl~ne par~llel to X-Z

Vertical plane parallel to X-Z Vertical pl~ne par~llel to X-Z

-50
She11rStressVZ
- 100
-50
32.5
She11rStressVZ 0.0353
--1~
100 30.6 0.0300
32.5
28.7 0.0353
0.0216
-200
- 1~ 30.6
26.8 0.0300
0.0193
28.7
24.9 0.0216
0.0110
-2~
-200-fi+\+1-H-fH- 26.8
23.0 0.0193
0.0086
300 -ttt-ti-J'tttt-
~ - -2~ -fi+\+1-H-fH- 24.9
21.1 0.0110
0.0033

~
1923.0
.2 0.0086
-0.0021
~--'"' -ttt-ti-J'tttt-
300-t\''<:1-i'-H-t+ 21.1
17.2 0.0033
-0.007~

~ -400+'\+H-1-t-
1519.3.2 -0.0021
-0.0127
~
- '"' -t\''<:1-i'-H-t+ 13417.2 -0.007~
-0.018 1
N~-4 ~ -f\-"i-'-+-1'-f--
-400+'\+H-1-t- 15 .3
11.5 N -0.0127
-0.023<'
134
96 -0.018 1
N -4 ~-H--H--t-11-
-500 -f\-"i-'-+-1'-f-- 11.5
7.7 N -0.023<'
- 0.03<'1
96
5.8 -0.0394
-500-11---F--+-++-
-5~ -H--H--t-11- 7.7
3.9 - 0.03<'1
-0.0447
5.8
-5~-t-\--hl--ft-
-0.0394
-600 -11---F--+-++- 2.0
3.9
0.1
-0.05.01
-0.0447
-0.055-4
-6~
-600+---f'-++
-t-\--hl--ft- 2.0
-1.8 -0.05.01
-0.0607
0.1
-3.7 -0.055-4
-0.0661
-7oo+-+--J4
-6~ +---f'-++ -1.8 -0.0607
-3.7 -0.0661
- -7oo+-+--J4
7~ _..,.._-t--4--1'
Second worst modulus
500 1000 , 500 2000 2500 3000 Regular tire 500 1000 1500 2000 2500 3000
Miningtir~
- 7~ _..,.._-t--4--1'
X (mm) Second worst modulus X (mm)
500 1000 , 500 2000 2500 3000 Regular tire 500 1000 1500 2000 2500 3000
Miningtir~
X (mm) X (mm)
Figure 13. Shear-stress for second worst modulus struc- Figure 16. Y displacement for second worst modulus
ture and regular tire load. structure and mining tire load.

Ve rtical plane parallel to X-Z

Vertical plane parallel to X-Z
Ve rtical plane parallel to X-Z
Vertical plane parallel to X-Z
-50
-100
-50
0 .0 153
0.0197
-100
-1~ 0.0 124
0.0167
0 .0 153
0.0096
-200
-1~ 0.0 124
0.0067
0.0197
0.0137
0.0167
0.0107
-250
-200 0.0096
0.0038
0.0067
0.0009
0.0137
0.0078
0.0107
0.0018
~ -250
-300 0.0038
-0.0019
~ -300
~
_,,., 0.0009
-0.00413
..o.oon
-0.0019 E'
0.0078
0.0018
0.0018
-0.0012

~ _,,.,
~ -400 -t\-\r-P+tf+lll+\--'f--+--++HH-++I~'-1-/J, -0.00413
-0.0105
..o.oon
-0.01S4 ~
E'-
0.0018
-0.0042
-0.0012
-0.0071
-0.0042
-0.0101
N
~--400
4~ -lr\-1"'-fltltf-t-t-lt--+--++HHK+\-"1~ ~-
-0.0105
-0.0163
-t\-\r-P+tf+lll+\--'f--+--++HH-++I~'-1-/J, N -0.0071
-0.0131
-0.01S4
-0.0192
-0.0101
N-500
- 4~-1\-~~~+4~1--'-+--H-+~~~~
-lr\-1"'-fltltf-t-t-lt--+--++HHK+\-"1~ -0.0163
-0.0220
-0.0192
-0.0219
N
-0.0161
-0.0131
-0.0191
--500
550 1rL+T7~-t-t-l+-+--l++-rt++~~
-1\-~~~+4~1--'-+--H-+~~~~ -0.0220
-0.0278
-0.0161
-0.0220
-0.0191
-0.0250
-600-ft+ +cH-t-l-++----+- -H-++11-t-'<-1-fi
- 550 1rL+T7~-t-t-l+-+--l++-rt++~~
-0.0219
-0.0306
-0.0278
-0.0335
-0.0220
-0.0280
-0.0250
--600-ft+
6"' ~~~~1-+-~-+---++~~~~~
+cH-t-l-++----+- -H-++11-t-'<-1-fi -0.0306
-0.0361
-0.0310
-0.0280
-O.OS-10
-0.0335
-0.0393
-7oo
- 6"'-I\-+-+--+!+-+-H-+--+++-++H,....-l~
~~~~1-+-~-+---++~~~~~ -0.0361
-0.0310
-0.0369
-0.0393
-O.OS-10
-7~
-7oo~~~~~-+~-+-1---f~_u~~~
-I\-+-+--+!+-+-H-+--+++-++H,....-l~ -0.0369
~00 1000 1 ~00 2000 2500 3!XXJ Averllgemodul us - 7:)() -'.,Lf--4--J'-!'-..J.----'f-- - + - - ¥ - - -+--lf-L--,L+--',J Second worst modulus
-7~ ~~~~~-+~-+-1---f~_u~~~ Mining tire 500 1000 1500 2000 2500 3000 Re~:ular tire
~00 X (1mm)
1000 ~00 2000 2500 3!XXJ Averllgemodul us - 7:)() -'.,Lf--4--J'-!'-..J.----'f-- - + - - ¥ - - -+--lf-L--,L+--',J Second worst modulus
X (mm)
Mining tire 500 1000 1500 2000 2500 3000 Re~:ular tire
X ( mm) X (mm)
Figure 14. Y displacement for average modulus struc- Figure 17. Y displacement for second worst modulus
ture and mining tire load. structure and regular tire load.

Vertical pl.nne par~llel to X-Z

Vertical pl.nne par~llel to X-Z 7 DISCUSSION

- 50
- 100
- 50
--1 100
~
0.0031
0.0068
The previous Figures show that the shear stresses
-200
-1~
0.0031
0.0052
00.0068
.0036 are quite high for the exempt loads on the weaker
--200
250 00.0052
.0020
0 .0036
0.0005 structure. However, the suggestion to increase the
,..--.. - -300
250 0 .0020
-0.0011
0.0005 total truckload does not seem to be challenging to
~ - -3'-0
-0.0027
,..--.. 300 -0.0011
-0.0043
~~ -- 400
3'-0-tt\,.,..Httfll-t+-'1----t--11--f-HIItt\-T"F-ffJ
-0.0027
-0.0059 road deterioration as long as the legal maximum
-0.0043
-0.0075
~--4~
N 400-1-'c--'t''-Hff+l+t--t---+--++HIIf-l+'r'-1<-f-1.
-tt\,.,..Httfll-t+-'1----t--11--f-HIItt\-T"F-ffJ -0.0059
-0.0091
-0.0075
axle load is not changed. As can be seen in the
-0.0 107
N -500
- 4~~~+H~~#-~---#~+f~~~
-1-'c--'t''-Hff+l+t--t---+--++HIIf-l+'r'-1<-f-1. -0.0091
-0.0 122
-0.0138
107
same Figures, axles spaced by three meters do not
-0.0
-550
-500~~T7~++-H--+--~+-rt++~~
~~+H~~#-~---#~+f~~~ -0.0151
-0.0 122
-0.0170
-0.0 138
interact much in the base and sub base material,
-550-ti:-t-+
-600 ~~T7~++-H--+--~+-rt++~~
H ++-t+-+-----cf++-H +'rl--fi
-6"'
-0.0 151
-0.0 186 but the left and right-hand wheels do. Some mate-
-600+-~~~+-+--r+-+--+++-~~~~
-ti:-t-+ H ++-t+-+-----cf++-H +'rl--fi -0.0 170
-0.0202

-7oo-ft++ -i+H-++--+-+-t-++ ft-''<-l-fl

-6"' +-~~~+-+--r+-+--+++-~~~~
-0.0 186
-0.0218
-0.0202
rials, such as hot asphalt and saturated unbounds,
-7oo-ft++
-7~ -i+H-++--+-+-t-++
~~~~~L-~~+-~+~~~~~ ft-''<-l-fl -0.0218
will have a shorter time to recover between loads
1000 1 2000 2500 3000
~00 ~00 Averlll[emodulus
-7~ ~~~~~L-~~+-~+~~~~~
~00 X
1000 (mm)
1 ~00 2000 2500 3000
Rei[Ulllr tire
Averlll[emodulus
though. This will likely increase rutting rates.
X (mm) Rei[Ulllr tire If axle loads increase, it is a very different story
Figure 15. Y displacement for average modulus struc- though. From the field we gathered data from a
ture and regular tire load. site that had been subjected to exempted heavy

473
 Shakedown Material strength specifications may have to
be changed, as the higher gross weights combines
the loads to higher stress levels, and possible more
grinding. We do know this from previous experi-
ence and the mining road example as well. This
may mitigate the sustainability of allowing higher
gross weight, and further lead to shortage of avail-
able materials.
At this point, we do not exactly know what the
stress and respective strain limits should be in real
numbers for any given unbound material. The inter-
shakedown action between layers, the surrounding environ-
ment and the loads comprise as complex situation,
Figure 18. Unbound Layer behaviour (Adapted from which is difficult to model. For performance-based
Werkmeister et alia, 2001). construction in particular, there is a need for sim-
ple guidelines to control the deterioration and keep
loads. In addition this occurred before all asphalt risks manageable. With field tests from the FWD,
layers were in place. All layers, including the sub- one can effortlessly derive vertical strain in any layer.
grade were subjected to much higher loads than in As we can see from the present example unbound
a normal design of a road. base material having a vertical strain less than
We did observe that the stresses were high, but 1000 micro strain did fairly well. Sub base layers
many elastic responses were normal nonetheless. having a higher confining pressure could probably
At the weaker points, one can see the following endure somewhat higher strain. Weaker materials
stages occur: being close to the LA drum test percentage cur-
rently used should not be used. Expect stricter limits
1. Post-compaction of upper unbound layers,
on materials, when higher loads are permitted.
energy is being dissipated.
2. As stiffening occurs, less energy is dissipated;
layers further down are being compacted.
REFERENCES
3. Plastic creep limit is exceeded; the process starts
all over again, as shown on Figure 18. Bonaquist, R.F., Witczak M.W.; A Comprehensive Con-
The explanation to the process by Werkmeister stitutive Model For Granular Materials In Flexible
et al (2001) and the influence of the octahedral shear Pavement Structures, Proceedings Eighth Interna-
stress as presented by Bonaquist (1997) are impor- tional Conference on Asphalt Pavements, Seattle, Wa
1997, p. 783–802.
tant pieces of the puzzle of understanding the dete- James, R., Noel, J., Furr, H., and Bonilla, F. 1986. Pro-
rioration of unbound materials. Add the seasonal posed New Truck Weight Limit Formula. Journal
variations and the dissipation of energy; it is clear of Structural Engineering, 10.1061/(ASCE)0733–
that we are dealing with a very complicated process. 9445(1986)112:7(1589), 1589–1604.
Liimatainen, Heikki and Nykänen, Lasse. 2014. Possible
impacts of increasing maximum truck weight—Case
8 CONCLUSIONS Finland. Transport Research Arena. Paris, France.
McKinnon, Alan C. 2005. The economic and environ-
Allowing higher loads on fewer trucks is a good mental benefits of increasing maximum truck weight:
the British experience. Transportation Research Part
idea, from a sustainability standpoint, as the D: Transport and Environment. Volume 10, Issue 1,
amount of greenhouse gases produced per weight January 2005, Pages 77–95 http://dx.doi.org/10.1016/j.
unit moved will be reduced. The maximum axle trd.2004.09.006.
load is not to be increased. Swedish Road Design Code. TRVK Väg. 2011:072
Using a linear elastic model, the fatigue and TDOK 2011:264. Trafikverket 2011. On-line. https://
rutting will not increase for a constant amount trafikverket.ineko.se/se/trvk-väg.
of goods being hauled. However, as rest periods Werkmeister, S; Dawson, A.R. and Wellner, F. Perma-
between loads become shorter a factor increasing nent Deformation Behavior of Unbound Granular
the number of loads by 1.25 is suggested for pave- Materials and the Shakedown Theory. Journal of
the Transportation Research Board, No. 1757, pp
ment design. For higher roads categories, in effect 75–81.2001.
this factor is about 1.09.

474
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Modeling of dynamic loads in pavement structural response models

J.A. Ramos-García, F. Sánchez-Domínguez & J.M. Sanz-García

Euroconsult Group, Madrid, Spain

ABSTRACT: The simplest and most commonly used pavement structural response models are linear elas-
tic multilayer models, although these entail substantial simplification. In particular, with regard to load mod-
eling, it is generally not possible to simulate dynamic loads, to modify the shape of the tire-pavement contact
surface or the load distribution on this contact surface. There are other more advanced structural response
models that offer a better approach to this issue, while it is the 3D Finite Elements (FE) models that really
allow for a deeper understanding of the actual behavior of pavements. In this paper we analyze different
aspects that affect the load modeling of a moving axle, including: applied load, eventual oscillation during
the circulation, moving speed, inflation pressure, tire-pavement contact surface and pressure distribution on
that surface. In addition to this, some recommendations for the modeling of dynamic loads are offered.

1 INTRODUCTION The practical advantage mainly depends on the

error that appears in each case when using simple
Modeling of dynamic loads in pavement structural linear elastic models. As for the finite elements
response models is a complex task and it requires models in the context of pavement analysis, it was
additional studies where it is frequent to have in- also concluded that the 2D axisymmetrical models
sufficient information available. The main aspects offer little benefits compared to multilayer elastic
affecting the modeling that need to be known are models, while 3D Finite Elements (FE) models
the actual applied load and its eventual oscillations are very useful, particularly in certain cases such
while moving, the moving speed, the inflation pres- as having singular boundary conditions or when
sure, the contact pressure, and the geometry of the analyzing localized discontinuities.
tire-pavement contact surfaces. Among the typically used models, the following
In the mechanistic analysis, it is common to use are worth noting:
simplifications, some of them quite significant.
– Linear elastic multilayer model. It does not
The multilayer elastic models, mostly used, con-
allow for simulations with moving loads, modi-
sider static loads, although the loads applied to the
fications in the shape of the tire-pavement con-
pavement surface are dynamic. In this research,
tact surface or in the load distribution on that
the factors to consider for an adequate modeling
contact surface.
of loads using the most advanced models, such as
– Modified linear elastic multilayer model
3D-finite elements models (3D-FE), are analyzed.
(non-linear, visco-elastic). As an example, the
KENLAYER program (Huang 2004) allows for
2 PAVEMENT RESPONSE MODELS simulating different load configurations which
are more complex than the usual ones (single
Nowadays there are different software programs, tires or single axle with twin tires). In addition,
that can be used to calculate pavement structures, it uses an approximation to convert moving
that allow for simple to complex modeling of the loads (approximately constant in magnitude and
said structures. With the aim of comparing and with varying positions) into loads with varying
analyzing the different programs used in Europe, magnitudes and a constant position. Besides,
the Amadeus project (European Commission the program allows for configuring the load in
2000) took place between 1998 and 1999. One of two different ways: as a stationary load and as
its objectives was to know the different programs a moving load that moves at an approximate
used for the mechanistic analysis of pavements to speed of 17.9 m/s (40 mph).
serve as a basis to action Cost 333: “Development – Linear elastic multilayer dynamic model. The
of new bituminous pavement design method” solution of the linear elastic multilayer model
(European Commission 1999). with stationary harmonic loads was developed
One of the conclusions of this project was that (Thomson 1950, Haskell 1953) by the middle of
visco-elastic models provide a better representa- last century. The hypothesis are similar to those of
tion of the waves generated by the traffic loads. linear elastic multilayer models with static loads,
475
 but there are some differences: the load consid- Table 1. Deflection survey systems (source: Benatov &
ered is stationary, not static; the behavior of each Sánchez 2007).
layer may be elastic or visco-elastic; and dynamic
effects (inertial forces and damping forces) are System Spanish standard Measuring speed
considered. Benkelman Beam UNE_41250_1* Stationary device
– FE model. The most advanced models currently Falling Weight UNE_41250_3* Stationary device
available to deal with most of the engineering Defl.
issues are FE models (different geometries, many LaCroix Defl. UNE_41250_2* 2–4 km/h
load static or dynamic configurations, several Curviameter UNE_41250_4* 18 km/h
constitutive models and various types of analy-
sis). There are some FE programs adapted to the * AENOR (2016a, b, c, d)
pavement structures, such as ILLI-PAVE (Raad
& Figueroa 1980), MICH-PAVE (Harichandran
et al. 1989) and AXIDIN (Antunes 1993) (axisym- systems. Besides, there are other devices, like laser
metrical programs where the greatest disadvan- deflectographs, that are being assessed by different
tage is that they do not allow for simulating more Administrations and National Laboratories to use
than one load) or the three-dimensional program them at network level; the first stages of the assess-
CAPA-3D (Scarpas 1992). There are also specific ment are yielding promising results, as included in
codes for pavements, such as DIANA (Vos et al. the “Pavement Structural Evaluation at the Net-
1996), developed by the Delft University, and work Level: Final Report” of the Federal Highway
CESAR, developed by the Laboratoire Central Admi-nistration (FHWA 2016).
des Ponts et Chaussées (LCPC 1996). In addi-
tion to this, there are some other very advanced
generic programs, such as ABAQUS (Dassault 4 FACTORS TO BE CONSIDERED IN THE
Systèmes 2013), PLAXIS (Plaxis 2001), ANSYS MODELING OF DYNAMIC LOADS
(Ansys Multiphisics 2009) and SYSTUS (Systus
2006), which allow for a great variety of con- In this section, the main factors to consider when
stituent rules and types of analysis, thus offering modeling dynamic loads are analyzed.
full flexibility to elaborate the model. Despite
the complexity in their use, they are considered 4.1 Dynamic consideration of the model
the most accurate and suitable models for any
The importance of considering the dynamic load
detailed study of pavement structures.
respect the pavement response is shown in the
behavior of asphalt mixes. As pointed out by several
3 DEFLECTION SURVEY DEVICES authors, such as Huet (1964), the stiffness of asphalt
mixes decreases when reducing the traffic speed of
The study of pavement structural behavior is usu- vehicles, which is basically due to their visco-elastic
ally performed by means of empirical and analyti- nature. In this same line, various studies, such as
cal methods, or a combination of both. The main those by Romero et al. (1994), Chatti et al. (1996) and
goal of the empirical methods is to study the behav- Mateos & Snyder (2014) among others, show that
ior of pavements using observations and real-scale there is an increase in both the strain of the lower
measurements, whereas the goal of the analytical part of asphalt mixes and the pavement deflection
methods is to model the pavement response in the when the moving speed of the same load is reduced.
same conditions as the real ones. It is advisable to Additionally, the dynamic consideration of
calibrate and verify the theoretical response mod- the load carries with it the possible occurrence of
els (mechanistic analysis) against real-scale obser- dynamic effects on the pavement, such as damping
vations of the pavement behavior. and inertial forces.
Pavement deflection survey is a non-destructive In this regard, Mateos (2003) conducted a
empirical test of the pavement structure response detailed analysis on the importance of consider-
when subject to a certain load. The comparison ing or dismissing damping and inertial forces when
between the measured deflection and the theoreti- modeling a flexible pavement. Mateos verified that,
cal deflection obtained from the pavement struc- for a maximum speed of 35 km/h, both effects were
tural response model requires, among other things, not relevant, and obtained almost identical results
the correct modeling of the applied load (which both when considering and dismissing these forces
depends on the device used). in the model. Likewise, he extended the study to
Table 1 shows the main deflection survey devices higher speeds, and the effect of both damping and
currently used. This table includes systems that inertial forces respect the results was not significant
measure the maximum deflection and the entire until the simulation speeds were very high. This
deflection bowl. There are different international allowed him to address the problem using a quasi-
published studies on the correlation among these static calculation procedure which is included in
476
many FE programs, where neither the damping nor ±20% respect the static weight of the vehicles in
the inertial forces are taken into account. However, it roads that show certain longitudinal unevenness
is considered that the properties of materials depend for traffic speeds close to 80 km/h. On the con-
on the frequency, thus allowing for the considera- trary, Ma & Caprez (1995) registered oscillations
tion of the visco-elastic behavior of asphalt mixes. below 5% of the static load in pavements in good
In the same line, Jones et al. (1967) pointed out condition for speeds of 40 km/h.
that the inertial effects do not apply to pavement Considering the different literature sources con-
engineering as the regular speed of vehicles (from 0 sulted, it should be noted that the oscillations of
to 30 m/s) is quite different from that of the waves moving loads mainly depend on the longitudinal
propagation speed at the pavement surface (from evenness of the pavement and the speed of the
100 to 600 m/s). Likewise, it is worth noting that moving load together with the damping system of
most of the pavement studies conducted with FE the vehicle. The better the longitudinal evenness
that are included in the literature do not consider and the lower the moving speed, the more reduced
either the damping or the inertial forces. the potential oscillations of the load.
Consequently, when modeling the loads applied As for the Benkelman beam, the LaCroix deflec-
by the Benkelman beam, the LaCroix deflecto- tograph and the Curviameter, where the measure-
graph and the Curviameter, with due considera- ment speed is reduced compared to the traffic speed
tion to their measurement speed, the damping and (much lower than those of the mentioned studies),
the inertial forces are considered negligible, and it it could be considered that the moving load oscil-
will be possible to accept as valid the quasi-static lations are negligible, with maximum variations
calculation procedure in order to consider the always lower than 5%, as registered by Ma & Caprez
dynamic behavior of the moving load. As for the (1995). Therefore, the simplification that the load of
laser deflectographs, it would be advisable to con- these devices during the deflection measurement is
duct specific studies in order to assess the effect of a constant magnitude will be considered valid, en-
both the damping and the inertial forces. tailing no significant alteration of the results.
At present, this aspect is being thoroughly ana-
4.2 Magnitude of the applied load lyzed due to the development of new survey devices
able to measure different parameters at high speeds,
The applied load by the deflection survey devices sometimes even reaching standard traffic speeds.
is, generally, an actual rolling load on the pave- As for the laser deflectographs, Rabe (2012) points
ment, of the same nature as the traffic loads. This out that, at high measurement speeds, there are sig-
load is usually applied by a single axle with twin nificant variations regarding the load applied by
tires, moving at a constant speed during the test. In these devices, which might reach point deviation
Spain, the same 13 t load is applied for measures values of up to 50%, so it is advised to continuously
with the Benkelman beam, the LaCroix deflecto- register the vertical load in this kind of measure-
graph and the Curviameter (AENOR 2016a, b, d). ment systems. Jansen (2014), in turn, reports load
On the contrary, the load applied by the Fall- variations close to 15% in a controlled and moni-
ing Weight Deflectometer (FWD) aims to simulate tored measurement section. Thus, for the study and
traffic loads. The nature of this load, which varies interpretation of measures at high speeds obtained
in magnitude and is constant in its position, differs by means of laser deflectographs, it is deemed nec-
greatly from the traffic loads, which are approxi- essary to take into account the load magnitude
mately constant in magnitude and variable in their variation during the deflection measurement.
position. In this regard, Mateos (2003) points out
that the “representativeness” of the test with FWD
should not be narrowly understood, as it depends 4.4 Geometry of the load
on the load pulse, the pavement structure (particu- As indicated, the rolling-load-type devices apply
larly the thickness of the asphalt mixes) and the load onto the pavement using a single axle with twin
structural response variable for which the equiva- tires. The tire footprint of the Benkelman beam, the
lence is defined. LaCroix device and the Curviameter, according to
current regulations (AENOR 2016a, b, d), must
4.3 Variation of the load during the tests comply with the indications included in Figure 1.
For its part, the FWD transmits the load impulse
It is worth noting that the longitudinal uneven- to the pavement surface through a circular load plate
ness of the pavement produces load oscillations measuring 300 mm in diameter (AENOR 2016c).
that could lead to alterations in the magnitude of
such load. The most important thing is to know
whether the effect on the actual load during the 4.5 Moving speed of the load
measurement is significant or not. The actual rolling load applied by the deflection sur-
Huhtala & Jacob (1995) and Barbour & vey devices moves at a constant speed during the test.
Newton (1995) registered oscillations of up to Table 1 includes the measurement speed of the Ben-
477
kelman beam, the LaCroix deflectograph and the
Curviameter. Besides, the new laser deflectographs
can apply different speeds during the measurement.
On the other hand, the FWD applies a load
which is variable in magnitude and constant in
its position in order to simulate traffic loads. As
indicated by Mateos (2003), it is generally accepted
that the FWD simulates the passing of a load at,
approximately, 60 km/h, although some recent
research works have question this, such as that of
Siddharthan et al. (2014), where different orders of Figure 1. Benkelman beam, LaCroix and Curviameter
magnitude are considered. tire footprint (source: AENOR 2016a, b, d).

floor and using a jack to lift and lower the twin tires
4.6 Inflation pressure and contact pressure
of these devices). From this information and that of
In the pavement mechanistic studies, it is usually the magnitude of the applied load it is possible to
used the simplification that the stress distribution calculate the actual average contact pressure; this is
of the tire load on the pavement surface is uni- the value to use when modeling the dynamic loads
form, assuming that the inflation pressure is equal of the pavement response models.
to the average contact pressure. In addition to this, it is possible to verify the
In this respect, some authors have analyzed in hypothesis commonly used according to which
detail the differences between inflation pressure the inflation pressure is equal to the average con-
and average contact pressure, and have determined tact pressure. To this end, the calculated value (the
the relations among them, which yield different actual average contact pressure) is compared to the
results. On one hand, Van Vuuren (1974) set the tire inflation pressure (controlled using a gauge
relation of Equation 1. manometer). In this respect it should be noted that
the empirical tests conducted with the Curviameter
qm 0 61 pi + 145 (1) device did not show great differences between both
pressures, and it was also verified that such differ-
where qm = average contact pressure (kPa); ences had little effect on the results of the pave-
pi = inflation pressure (kPa). ment response models developed.
On the other hand, other studies reveal that It would be possible to proceed in the same man-
the average contact pressure depends both on the ner to calculate the actual average contact pressure
inflation pressure and the applied load, as shown of the laser deflectographs, although for measure-
by Ikeda & Itoh (1985) through Equation 2. ments taken at high speeds it would be necessary
to account for the variation of magnitude dur-
qm 0 42 Fa + 0.29 ⋅ pi + 1.448 (2) ing the deflection measurement, as indicated in
Section 4.3 of this research work.
where qm = average contact pressure (kp/cm2);
Fa = applied load (103 kp); pi = inflation pressure
(kp/cm2). 4.7 Distribution of the contact pressure
Upon examination of both studies some diffe- As already mentioned, in the pavement mechanis-
rences between inflation pressure and average tic studies it is common practice to consider the
contact pressure are identified, contrary to the simplification that the stress distribution of the tire
simplification most widely used (which indicates load on the pavement surface is uniform.
that the inflation pressure is equal to the average In this respect, different research works, as
contact pressure). those conducted by Huhtala et al. (1989), Greer
In order to clarify the difference between the & Palazotto (1998), De Beer (1996), Himeno
regular practice and some theoretical studies in this et al. (1997), Blab (1999), Woodside et al. (1999),
field, it is recommended to determine the average Blackman et al. (2000), Hua & White (2002) and
contact pressure in each case, using the magnitude Soares (2005), concluded that the distribution of
of the applied load and the actual geometry of the the contact pressure over the tire-pavement contact
tire footprint. surface is not uniform, and it could significantly
As indicated, according to Spanish standards vary between the center and the edges of the said
UNE_41250_1, 2, 4 (AENOR 2016a, b, d), the surface. Moreover, some of them were consistent
tire footprint of the Benkelman beam, the LaCroix in pointing out that the inflation pressure substan-
deflectograph and the Curviameter must be adjusted tially controls the contact pressure at the center of
to the indications of Figure 1. The tire footprints of the tire-pavement surface, while the applied load
the twin tires of the axle could be manually meas- controls the pressure at the longitudinal edges of
ured in detail (for example, by placing papers on the the mentioned tire-pavement contact surface.
478
 In order to know the possible effect of the deflectograph and the Curviameter device,
different distribution of the pressure on the tire- included in Spanish standards UNE_41250_1, 2, 4
pavement contact surface, some published research (AENOR 2016a, b, d), the average contact pressure
works based on experimental results have been obtained at the tire center (q60) would be 0.83 MPa
analyzed, such as those by De Beer et al. (1997), while the average contact pressure in the tire edges
Blab (1999) and De Beer (2001). (qe) would be 0.76 MPa, using Equations 3 & 4.
First, it should be mentioned that the study For its part, q60 would be 0.55 MPa and qe would be
published by De Beer (2001) established the influ- 0.54 MPa if the Equations 5 & 6 were used.
ence of the inflation pressure and the applied load It is important to note that q60 results are rather
respect the distribution of the contact pressure on different when applying the first or the second sets
the tire-pavement surface. As for the loads and the of equations; the same applies to qe (which could
inflation pressures of the devices (Benkelman beam, be reasonable, as the studies were based on their
LaCroix deflectograph and Curviameter) included own experimental data). Nevertheless, the differ-
in Spanish standards UNE_41250_1, 2, 4 (AENOR ence between q60 and qe in all studies is very small,
2016a, b, d), Figure 2 shows a virtually uniform dis- and these values are almost the same in Blab (1999)
tribution of the contact pressure. On the contrary, equations.
when using other loads and inflation pressures (for Thus, from the results obtained in these three
instance, with 35 kN loads and 520 kPa inflation studies (De Beer et al. 1997, Blab 1999, De Beer
pressures) this distribution of the contact pressure 2001) upon application of the pressures and loads
on the tire-pavement surface is clearly not uniform. of the Benkelman beam, the LaCroix deflecto-
On the other hand, Equations 3 & 4 illustrate graph and the Curviameter, it is possible to con-
the relations proposed by De Beer et al. (1997), clude that the distribution of the contact pressure
while Equations 5 & 6 correspond to the relations on the tire-pavement contact surface is homogene-
set out by Blab (1999). In both cases, the average ous, which makes it possible to assume the hypoth-
contact pressures are related to the inflation pres- esis that stress distribution of the tire load is
sures and the applied load, both at the center of uniformly distributed over the said tire-pavement
the tire footprint (the central section corresponds contact surface.
to 60% of the tire width) and the longitudinal As for other devices, such as those of rolling-
edges of the tire footprint. load-type for the measurement of the deflection,
the corresponding loads and inflation pressures
q60 0 86 pi + 175 (3)
applied in each case will make it possible to deter-
qe 0 53 Fa + 57.46 ⋅ Fa − 534.05 (4) mine, in the same way as described in this section,
the uniformity of the stress distribution of the tire
q60 0 864 Fa + 0 438 ⋅ pi + 190.23 (5) load over the tire-pavement contact surface.
Regarding the FWD, it is usually assumed that
qe 0 087 Fa2 + 19.189 ⋅ Fa + 17.615 (6) the applied load is uniformly distributed under the
load plate. However, some research works, as the one
where q60 = average contact pressure in the tire conducted by Nazarian & Boddapati (1995) point
center (kPa); qe = average contact pressure in the out that the stiffness of the load plate could lead to
tire edges (kPa); Fa = applied load (kN); pi = infla- a non-uniform distribution of the contact pressure.
tion pressure (kPa).
Regarding the tire inflation pressures and the
applied loads of the Benkelman beam, the LaCroix 5 MODELING OF DYNAMIC LOADS IN
THE 3D-FE RESPONSE MODELS

The most widely used pavement response models,

mainly due to their ease of use, are the linear elas-
tic multilayer models. These models are very limited
regarding the modeling of the applied loads (they
use significant simplifications). Among other things,
these models do not allow for simulations with mov-
ing loads, modifications in the geometry of the tire-
pavement contact surface or the load distribution
over the said contact surface. Under the hypothesis
applied in these models, it is assumed that the static
load produces a vertical pressure which is uniformly
distributed over a circular area of the pavement sur-
Figure 2. Influence of the inflation pressure and the face. Besides, dynamic effects are not considered.
applied load in the contact pressure distribution (source: At the other end are the 3D-FE response models.
De Beer 2001). Despite the complexity in their use, these models
479
are the ones that best represent the behavior of
pavement structures. Regarding the modeling of
loads, these models allow for many load static and
dynamic configurations. Listed below are some
recommendations about the modeling of dynamic (a) Til'f:FOOiprint\
loads in this type of response models (3D-FE).

5.1 Dynamic consideration of the model

Both the linear elastic multilayer dynamic models
and the 3D-FE models allow for a dynamic consid-
eration in pavement analysis, which makes it possible
to approximate the model to the actual situation, (b) Ti~ Prcu wc Distnlmtiou\ Applied to the Mc-.h

including the damping and the inertial forces. How-

ever, as already stated, when modeling the applied Figure 3. Example of the modeling of the applied load
loads of the Benkelman beam, the LaCroix deflecto- (source: Kim et al. 2011).
graph and the Curviameter, it is possible to disregard
the damping and the inertial forces in the pavement that use an approximation to convert moving
response model due to their measurement speed. loads into loads with varying magnitudes and a
constant position. Besides, the linear elastic multi-
5.2 Magnitude, variation, geometry and moving layer dynamic models define stationary harmonic
speed of the load loads. As for the 3D-FE models, these allow simu-
lating the movement of the tire-pavement contact
All software programs offer the option of including
surface by applying moving boundary conditions
the magnitude of the applied load in the response
in the driving direction.
model. The main problem lies in the simulation of
the variation of this magnitude, which occurs with
laser deflectographs during the measurement of the 5.3 Contact pressure
deflection at high speeds. The different configuration As stated above, different research works concluded
options offered by the 3D-FE programs could help that the distribution of the contact pressure on the
to find an approximate solution to this problem. tire-pavement contact surface is not uniform. In this
As for the geometry of the applied load, most same line, Rahman et al. (2011) propose to model
models assume the simplification that the tire- the tire-pavement surface using a rectangle in the
pavement contact surface is circular. By contrast, central part and two half circles on the sides. It is
the models developed with 3D-FE programs make assumed that the rectangular surface, correspond-
it possible to model nearly any geometrical shape, ing to approximately half the contact surface, bears
which allows for the detailed simulation of the two thirds of the total load of the tire, thus modi-
actual shape of every tire-pavement contact sur- fying the concept of homogeneous distribution
face. As an example, Figure 3 shows the modeling of the load on the tire-pavement surface. Hadi &
of the applied load used by Kim et al. (2011). In Bodhinayake (2003), for their part, indicate that most
this research, the tire-pavement contact surface of the pressure is concentrated on a rectangular sur-
was divided into several rectangles that were sub- face located at the central part of the tire-pavement
ject to different pressures. contact surface, with a length of approximately 0.87
In addition to this, many of the existing software times the total length of the tire-pavement surface.
programs only allow to analyze these two configu- Other works, such as those conducted by
rations: single tires or single axle with twin tires. Al-Qadi et al. (2004) and Yoo (2007) among oth-
Although these are the configurations most widely ers, divide the contact surface into parallel rectan-
used in road pavement analysis, sometimes it is nec- gles, and the actual pressure obtained from the
essary to model different configurations (two-axle, measurements of the tires used in the research is
three-axle, six-tire, etc.), particularly when working applied to each of them.
with airport pavements. 3D-FE programs admit all In this regard, the versatility of the 3D-FE mod-
types of load configurations. Similarly, there are some els makes it possible to apply different pressures on
multilayer software programs, such as KENLAYER the various geometrical shapes in order to define
(Huang 2004), that make it possible to simulate more the tire-pavement contact surfaces in each case.
complex axle configuration loads than the usual ones However, in the case of the Benkelman beam, the
(single tires or single axle with twin tires). LaCroix deflectograph and the Curviameter, it is
Regarding the moving speed of the load, the lin- concluded (as seen in Section 4.7 of this paper)
ear elastic multilayer models do not allow for simu- that the distribution of the contact pressure on the
lations with moving loads. There are some modified tire-pavement contact surface is homogeneous, so
linear elastic multilayer models (KENLAYER) in these cases is valid the hypothesis that the stress
480
distribution of the tire load is uniformly distrib- Ansys Multiphisics. 2009. Ansys Release 12.1. Ansys
uted over the said tire-pavement contact surface. Incorp.
Antunes, M.L. 1993. Avaliação da capacidade de carga
de pavimentos utilizando ensaios dinâmicos (Pavement
6 CONCLUSIONS bearing capacity evaluation using dynamic tests). PhD
thesis, Technical University of Lisbon, October [in
The main conclusions of this research are the Portuguese].
following: Asociación Española de Normalización y Certificación—
AENOR. 2016a. UNE 41250-1. Test methods for meas-
– Linear elastic multilayer models (the most widely urement of pavement deflection. Part 1: Benkelman
used pavement response models) are very limited Beam. Madrid: AENOR [in Spanish].
regarding the modeling of the applied loads, and Asociación Española de Normalización y Certificación—
use significant simplifications. For their part, the AENOR. 2016b. UNE 41250-2. Test methods for meas-
urement of pavement deflection. Part 2: Deflectograph.
3D-FE response models, despite the complexity Madrid: AENOR [in Spanish].
in their use, are the ones that best represent the Asociación Española de Normalización y Certificación—
behavior of pavement structures. AENOR. 2016c. UNE 41250-3. Test methods for meas-
– Regarding the dynamic consideration of the urement of pavement deflection. Part 3: Falling Weight
model, the linear elastic multilayer dynamic Deflectometer. Madrid: AENOR [in Spanish].
models and the 3D-FE models allow for the Asociación Española de Normalización y Certificación—
inclusion of damping and inertial forces in pave- AENOR. 2016d. UNE 41250-4. Test methods for meas-
ment analysis. Nevertheless, when modeling urement of pavement deflection. Part 4: Curviameter.
the applied loads of the Benkelman beam, the Madrid: AENOR [in Spanish].
Barbour, I.A. & Newton, W.H. 1995. Multiple-Sensor
LaCroix deflectograph and the Curviameter, it is Weight-1n-Motion. First European Conference on
possible to disregard the damping and the iner- Weight-ln-Motion of Road Vehicles, Zurich.
tial forces in the pavement response model due to Benatov, B.B. & Sánchez Domínguez, F. 2007. Pavement
their measurement speed. surveys using high performance devices in Spain. XIV
– The simplification that the load is a constant Ibero-Latin-American Congress of Asphalt, November,
magnitude will be considered valid when the La Habana, Cuba [in Spanish].
deflection is measured at reduced speed, entailing Blab R. 1999. Introducing Improved Loading Assump-
no significant alteration of the results. However, tions into Analytical Pavement Models Based on Meas-
for the study and interpretation of deflection ured Contact Stress of Tires. International Conference
on Accelerated Pavement Testing, Reno.
measurements at high speed, it is deemed nec- Blackman, D.I., Hadillay, A.R. & Merrill, D.B. 2000.
essary to take into account the load magnitude Effects of tyre type, inflation pressure and load on contact
variation during the tests. The different configu- area and pavement performance. TRL. P. Report PR/
ration options offered by the 3D-FE programs IP/66/00, June.
could help to find an approximate solution. Chatti, K., Kim, H.B., Yun, K.K., Mahoney, J.P. & Moni-
– The 3D-FE models allow for simulating the smith, C.L. 1996. Field Investigation into Effects of
movement of the tire-pavement contact surface Vehicle Speed and Tire Pressure on Asphalt Concrete
by applying moving boundary conditions in the Pavement Strains. Transportation Research Record 1539,
driving direction. TRB, National Research Council, Washington, D.C.
Dassault Systèmes. 2013. Abaqus 6.13. Dassault Systèmes
– The empirical tests conducted with the Cur- Simulia Corp., Providence, RI, USA.
viameter device do not show great differences De Beer, M., Fisher, C. & Jooste, F.J. 1997. Determination
between inflation and contact pressures. These of Pneumatic Type/Pavement Interface Contact Stress
differences have little effect on the results of the under Moving Loads and Some Effects on Pavements
pavement response models developed. with Thin Asphalt Surfacing Layers. Proc. of the 8th
– The distribution of the contact pressure on the International Conference on Asphalt Pavements, Seattle,
tire-pavement contact surface can be consid- Washington.
ered homogeneous for the Benkelman beam, De Beer, M. 1996. Measurement of tyre/pavement inter-
the LaCroix deflectograph and the Curviameter. face stresses under moving wheel load “. Heavy Vehicle
Systems, Special Se., Int. J. Vehicle Design, Vol. 3, No
When using other loads and inflation pressures, 1–4: 97–115.
the distribution of the contact pressure on the De Beer, M. 2001. Stress-ln-Motion Technology. Work-
tire-pavement surface can be non-uniform. shop on accelerated Load Testing, Cologne, Germany.
European Commission. 1999. Cost 333. Development of
new bituminous pavement design method. Final report.
REFERENCES Luxembourg: Office for Official Publ. of the Europ.
Communities.
Al-Qadi, I.L., Elseifi, M.A. & Yoo, P.Y. 2004. In-situ vali- European Commission. 2000. Avanced models for ana-
dation of mechanistic pavement finite element model- lytical design of european pavement structures “AMA-
ling. Proceedings of the 2nd International Conference on DEUS”. Final report. Pr. funded by the European C.
Accelerated Pavement Testing, September, Minneapolis, under Transport RTD Progr. 4th Framework Progr.
USA. RO-97-SC-2137.

481
Federal Highway Admin. – FHWA. 2016. Pavement Struc- Research Record, Nº1896, V.1806, DOI: 10.3141/1806-
tural Evaluation at the Network Level: Final Report. Pub. 03: 107–118.
No: FHWA-HRT-15-074. Georgetown P. McLean, VA, Mateos Moreno, A. 2003. Modelización numérica de la
USA. respuesta estructural de los firmes flexibles a partir de
Greer, J. & Palazotto, A. 1998. Tire contact using two- medidas experimentales en pista de ensayo a escala real.
dimensional finite elements. ASCE, Journal of Engineer- PhD thesis. ETS. ICCP. Polyt. Univ., UPM, Madrid [in
ing Mechanics. Vol. 124, No. 3, April: 32–36. Spanish].
Hadi, M. & Bodhinayake, B.C. 2003. Non-linear finite ele- Nazarian, S. & Boddapati, K.M. 1995. Pavement Falling
ment analysis of flexible pavements. Advances in Engi- Weight Deflectometer interaction using dynamic finite-
neering Software, 34: 657–662. element analysis. Transportation Research Record, 1482,
Harichandran. R.S., Baladi, G.Y. & Yeh, M. 1989. Develop- Transportation Research Board, Washington, D.C.
ment of a Computer Program for Design of Pavements Plaxis BV. 2001. Plaxis V8. Delf University, The
Systems Consisting of Bound and Unbound Materials. Netherlands.
Dep. Civil & Environmental Engineering, Michigan Raad, L. & Figueroa, J.L. 1980. Load Response of Trans-
State Univ. portation Support Systems. Transportation Engineering
Haskell, N.A. 1953. The Dispersion Surface Waves on Journal, ASCE, N = TE1, Vol. 106.
Multilayered Media. Bul. Seismological Soc. Amer., Rabe, R. 2012. Structural pavement monitoring with non-
Vol.43, N.1. destructive measuring devices at traffic speed-current
Himeno, K., Ikeda, T., Kamijima, T. & Abe, T. 1997. Dis- research and experience in Germany. German Fed.
tribution of Tire Contact Pressure of Vehicles and its Highway Research Inst. BASt. 4th European Pavement
Influence on Pavement Distress. Proceedings of the 8th and Asset Management Conf. EPAM4, September,
International Conference on Asphalt Pavements, Seattle, Malmö, Sweden.
Washington. Rahman, M.T., Mahmud, K. & Ahsan, S. 2011. Stress-
Hua, J. & White, T.A. 2002. Study of nonlinear tire pres- strain characteristics of flexible pavement using finite
sure effects on HMA rutting. International Journal of element analysis. Inter. Journal of Civil and Structural
Geomechanics, AN 7073527, Vol. 2, Issue 3, June, 353, Enginerering. Vol. 2, No 1. ISSN 0976-4399. September:
24p. 233–240.
Huang, Y.H. 2004. Pavement analysis and design. 2nd Romero, R., Ruiz, A., Rodil, R., & Lechuga, M.A. 1994.
Edition. Pearson Prentice Hall. Upper Saddle River, Variation of Deflection with Measuring Equipment
NJ47058, USA. and with Load Speed on Test Track. Transportation
Huet, C. 1964. Une méthode d’etude en régime dynamique Research Record 1448, TRB, Nat. Research Council,
du comportement visco-elástique des matériaux Washington, D.C.
enrobés. Bul. de liaison des Lab. Ponts et Chaussées, N° Scarpas, A. 1992. CAPA-3D Finite Elements System—
8 [in French]. User’s Manual. Dept. of Structural Mechanics, Faculty
Huhtala, M. & Jacob, B. 1995. OECD DIVINE Project— of Civil Engineering, TU-Delft, The Netherlands.
Spatial Repeatability of Axle Impact Forces. First Euro- Siddharthan, R., Nasimifar, M., Rada, G., Nazarian, S.,
pean Conference on Weight-ln-Motion of Road Vehicles, Sivaneswaran, N. & Thyagarajan, S. 2014. Investiga-
Zurich. tion of applicability and use of a pavement response
Huhtala, M., Pihlajamaki, J. & Pienimaki, M. 1989. Effects model with High Speed Deflection Devices (HSDDs).
of tires and tire pressures on road pavements. Transpor- National Pavement Evaluation Conference. September,
tation Research Record, Nº 1227. Blacksburg, VA, USA.
Ikeda, T. & Itoh, M. 1985. Considerations on equations for Soares, R.F. 2005. Finite element analysis of the mechanics
estimating tire contact pressure of a large truck. Proc. of viscoelastic asphaltic pavements subjected to varying
60th Annual Conf. of the Japan Soc. of Civil Engineers, tireconfigurations. PhD thesis. University of Nebraska,
Vol. 5. Lincoln, Nebraska, USA.
Jansen, D. 2014. Evaluation of the TSD in Germany. Systus/Sysweld. 2006. User’s manual, ESI Group-Systus
National Pavement Evaluation Conf., September, Int.
Virgnia, USA. Thomson, W.T. 1950. Transmission of Elastic Waves
Jones, R., Thrower, E.N. & Gatfield, E.N. 1967. Surface Through a Stratified Soil Medium. Journal Applied
wave method. Proc. 2nd Int. Conf. of Structural Design Physics, Vol. 21.
of Asphalt Pavements, Univ. of Michigan, Ann Arbor, Van Vuuren, D.J. 1974. Tyre Pressure and its Effect on
USA, 505–519. Pavement Design and Performance. National Institute
Kim, Y.R., Ban, H. & Im, S. 2011. Impact of truck load- for Road Research (NITRR), Council for Scientific and
ing on design and analysis of asphaltic pavement struc- Industrial Research (CSIR), (NITRR Report; RR 192).
tures. Phase II. Report MACT-UNL: 321. Final report. Pretoria.
Mid-America Transportation Center. University of Vos, E., Hendriks, N., Galjaard, P.J. & Van Schelt, W. 1996.
Nebraska, Lincoln. The non-linear elastic model of Boyce: An evaluation of
Laboratoire Central des Ponts et Chaussées—LCPC. 1996. the model in the FEM—Program Diana. En: Flexible pave-
Manuel d’utilisation du logiciel éléments finis. CESAR. ment. Ed. Gomes Correia, Balkema, Rotterdam, 201–207.
LCPC, Paris [in French]. Woodside, A., Woodward, W. & Siegfried. 1999. The Deter-
Ma, S. & M. Caprez. 1995. The Pavement Roughness mination of Dynamic Contact Stress. International Con-
Requirement for WIM. First European Conference on ference on Accelerated Pavement Testing, Reno.
Weight-ln-Motion of Road Vehicles, Zurich. Yoo, P.J. 2007. Flexible pavement dynamic responses analy-
Mateos, A. & Snyder, M. 2014. Validation of Flexible sis and validation for various tire configurations. PhD.
Pavement Structural response Models with Data from Dissertation. University Illinois at Urbana-Champaign,
the Minnesota Road Research Project. Transportation USA.

482
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Impact of longer and heavier vehicles on the performance of asphalt

pavements: A laboratory study

A.W. Ahmed & M.S. Rahman

Pavement Technology, VTI, Linköping, Sweden

S. Erlingsson
Pavement Technology, VTI, Linköping, Sweden
Faculty of Civil and Environmental Engineering, University of Iceland, Reykjavik, Iceland

ABSTRACT: Historically, Longer and Heavier Vehicles (LHVs) have been permitted to operate in
Sweden. Since 1996 and as of the beginning of 2015, the maximum gross vehicle weight of 60 tons and
a length of up to 25.25 m have been permitted. The Swedish Transport Administration has decided to
further increase the maximum gross vehicle weight to 74 tons and studies are undergoing to evaluate the
impact of the proposed LHVs on existing transport infrastructure. To this end, repeated load triaxial tests
and indirect tensile fatigue tests were conducted on selected conventional asphalt mixtures to investigate
and quantify the impact of single, tandem and tridem axle configurations on permanent deformation and
fatigue performances of conventional asphalt pavements. In addition, fatigue tests for selected LHV sce-
narios were conducted. This paper presents the results of the laboratory tests and simulations conducted.
The test results have clearly demonstrated the impact of the different axle configurations on the rutting
and fatigue performances of the mixture. Furthermore, such results can explain the significance of axle
configuration on modelling the rutting and fatigue performances of asphalt pavements.

1 INTRODUCTION cal GVW limits in Sweden and the EU (VTI 2008).

The Swedish Transport Administration has decided
Longer and Heavier Vehicles (LHVs) are believed to to further increase the maximum GVW to 74 tons
benefit the economy and the environment by reduc- and studies are underway to evaluate the impact of
ing the costs of road freight transport (Leduc 2009, the proposed LHVs on traffic safety, economy and
Steer et al., 2013). Historically, LHVs have been the infrastructure, mainly on roads and bridges.
permitted to operate in Sweden. Since 1996 and Recently, neighboring Finland has increased the
as of the beginning of 2015, the maximum Gross allowable maximum GVW from 60 to 76 tons.
Vehicle Weight (GVW) of 60 tons and a length Several studies have been conducted to evaluate
of up to 25.25 m have been permitted. Within the the impact of the LHVs on transport infrastruc-
European Union (EU), however, the maximum tures (BAST 2006, Akerman & Jonsson 2007,
gross weight for an articulated vehicle and truck- Aurell & Wadman 2007, TRL 2008, Leduc 2009,
trailer combination on five or more axles has been Chatti et al., 2009, Weissmann et al., 2013).
harmonized at 40 tons. Figure 1 shows the histori- BAST (2006) conducted laboratory simulations
to study the effect of increasing GVW of tractor-
trailer combinations from 40 to 60 tons on Ger-
Sweden man federal trunk roads. The study concluded
EU that both fully as well as partially loaded tractor-
trailer combinations may lead to a reduction in
road stress and thus to less rutting on the road as
r-----------
40 far as weight is concerned. However, the 60-ton
28
r-- -- 1 tractor-trailer combination might increase the
risk of bottom up fatigue cracking in the asphalt
Year
layers. Moreover, it was anticipated that replacing
30% of the 40 ton vehicles by 60 ton tractor-trailer
Figure 1. Historical limits of gross vehicle weight in combinations could increase the pavement life by
Sweden and the European Union (EU). about 5.25%.

483
 TRL (2008) carried out a numerical analysis of tonnage compared to single axles. Furthermore,
the effects of LHVs on the UK’s transport infra- increasing the number of axles carrying the same
structure. In this study, the relative road wear of load resulted in less damage. Significant differ-
eight LHVs scenarios was estimated using the ences in damage were found between single, tan-
standard assumption that wear is proportional to dem and tridem axles. Similarly, trucks having
the fourth power of the axle weight, as shown in more axles and axle groups produced lower truck
Equation 1. The study indicated that all the pro- factors (average ESALs per vehicle) per tonnage
posed LHVs resulted in lower wear factors per than those with single axles.
100 tons than those vehicles which were permitted The main objective of the work presented here
under regulations: was to study the impact of LHVs on the existing
Swedish road infrastructure. The scope of the
study was limited to conventional asphalt pave-
m ⎛ Lx, i ⎞ 4 ments commonly used in Sweden. Furthermore,
ESALs = ∑ ⎜ ⎟ (1)
the study was limited to laboratory simulations
i = 1⎝ Ls, i ⎠ and testing.

where ESALs = equivalent standard axle loads,

Lx,i = axle load, Ls,i = the corresponding standard 2 MATERIALS AND TESTS
axle load for the axle configuration, m is the total
number of axles per vehicle. Indirect Diametrical Tensile (IDT) fatigue tests
Akerman & Jonsson (2007), Aurell & Wadman and Confined Triaxial Tests (CTT) were conducted
(2007) and Leduc (2009) employed similar pro- in this study. The CTTs were conducted on con-
cedures based on the fourth power law to com- ventional asphalt binder mix with a pen grade of
pare LHVs scenarios for current and prospective 70/100 and maximum aggregate size of 11 mm,
modular vehicle combinations with a total length denoted as ABb11 70/100, whereas a conventional
greater than 25.25 m. Their studies concluded that road base mix with a pen grade of 160/220 and
the road wear from LHVs was typically less than maximum aggregate size of 22 mm (denoted AG22
from vehicle combinations permitted under EU 160/220) was prepared for the IDT fatigue tests.
regulations. It should be noted that the studies The CTT specimens having a diameter of 100 mm
based on the fourth power law assumed that the and height of 150 mm were prepared using gyra-
LHVs scenarios do not necessarily result in higher tory compaction. On the other hand, for the IDT
axle loads. Also, the ESALs per ton are normal- fatigue tests, asphalt specimens were drilled from
ized with the GVW, and thus the results may vary three roller-compacted asphalt slabs having a
if the LHVs are not fully loaded. Furthermore, thickness of 60 mm, width of 500 mm and length
several studies have questioned the validity of the of 700 mm. Table 1 presents the volumetric prop-
fourth power law for conditions different from erties of the mixtures.
the AASHTO test roads used to empirically derive
the fourth power law (Cebon, 1999).
The Finnish Ministry of Transport and Com- 2.1 Indirect Diametrical Tensile (IDT)
munications (FMTC) recently decided to increase fatigue test
the GVW limit to 76 tons (Steer et al., 2013). It was
The stress controlled IDT fatigue test was
found that with regards to road wear, the critical
conducted in accordance with SS-EN 12697-
factors have been found to be the weight per axle
24 Appendix E. The method involves applying
and even load distribution (FMTC, 2002).
a certain number of load pulses along the verti-
Weissmann et al., (2013) studied the impact
cal diametric plane of a cylindrical sample thus
of LHVs on selected Texas corridors based on a
inducing a tensile stress along the horizontal dia-
mechanistic-empirical approach, and it was con-
metric plane perpendicular to the loading plane.
cluded that the LHVs scenario had no impact on
rigid pavements. However, for flexible pavements,
the LHVs scenario improved the expected life of
segments carrying overweight axles, and decreased Table 1. Mixtures volumetric properties.
the life of segments with thick Hot Mix Asphalt
(HMA) layers carrying no overweight axles. Binder
Chatti et al. (2009) conducted a series of labora- content Air void
tory tests and simulations to evaluate the impact of Mix (%wt) (%) Test
multi-axle trucks on pavement distress in the state
AG22 160/220 4.2 4.0 IDT Fatigue
of Michigan. From the tests, it was concluded that
ABb11 70/100 5.1 6.2 CTT
multiple axles were found to be less damaging per

484
The standard fatigue test consists of a single load
pulse of 0.1 sec loading and 0.4 sec rest period,
as shown in Figure 2(a). In addition, tests under
tandem and tridem loaded pulses with 0.15 sec
and 0.2 sec, loading times, respectively, were con-
ducted. A rest period of 0.4 sec was adopted for
all the axle configurations to establish the fatigue
curves representing multiple axle configurations
(Figs. 2b and 2c). The loading time and the rest
periods were estimated based on average speed of
the trucks and axle spacing. All the tests were con-
ducted at a temperature of 10°C, which is consid-
ered a representative temperature for fatigue tests
in the Nordic countries. The interaction level of
the load pulses (the depth of the valley of the load
pulses in Figs. 2b and 2c) were selected to represent
an asphalt layer that is neither very thin nor very
thick. The actual setup of the IDT fatigue test is Figure 3. IDT fatigue test setup.
shown in Figure 3.
Furthermore, IDT fatigue tests for the three
LHVs scenarios shown in Figure 4 were conducted.
According to national Bridge Weigh-In-Motion
(BWIM) measurements, the LHV scenario type
A (Fig. 4) is currently permitted an 8 axle tractor-
A
semitrailer combination with a legal GVW limit of

Slres3

··~1
cro/2

(a)
cr.,..

cr..., u 100 400

Time(ms)
B

Stress c
Figure 4. LHVs scenarios selected for IDT fatigue
tests.

64 tons and length of 22.5 m and those in type B (9

axles) and C (11 axles) in Figure 4 are the proposed
(b) 400 combinations for higher GVW limits (74 tons).
The loading time and rest periods for the fatigue
tests were selected based on the estimated average
Stress speed of 80 km/h and the spacing between axles

cro/~
and axle groups.

cr.,.. ! 2.2 Confined Triaxial Tests (CTT)

cro/2
cr...., CTT were conducted to evaluate the permanent
~----------------------~·
deformation characteristics of the selected asphalt
(c)
I 200 I 4oo Time (ms) mixture subjected to different load configurations
of single, tandem and tridem load pulses. Fig-
Figure 2. (a) Single, (b) tandem and (c) tridem pulses ure 5 shows a schematic and the actual test set-
used for IDT fatigue tests. up of the CTT. Δq and σ3 in Figure 5a denote the

485
 0"1 = 0"3 + t!.q was selected based on the average maximum sum-
mer temperature expected and the limitation of the
testing machine. In addition, observed CTT per-
manent deformation for a temperature lower than
15°C was insignificant. Each specimen was tested
at constant temperature under multi-stage loading
conditions, i.e., four load levels were applied, each
E.zl with 5000 load cycles (N). The vertical deforma-
tion was measured over the middle 100 mm of the
specimen with three extensometers spaced 120°
apart around the cylindrical specimens. The sum-
mary of the CTT test program is shown in Table 2.
a)

3 RESULTS AND DISCUSSION

3.1 IDT Fatigue test

It was observed that the fatigue curves under mul-
tiple axle loading were dependent on the definition
of the initial strain. For the single axle loading, as
the case for standard IDT fatigue test, the initial
strain was obtained from the single peak strain
pulse. However, for tests under multiple axle load-
ing, the magnitude of the peak strain significantly
increased from the first to the last peak due to the
accumulation of viscoelastic strain. Thus, different
fatigue curves were obtained depending on the ini-
tial strain considered, i.e., first peak or last peak.
Figure 6 depicts the peak strains under tridem axle
configuration.
Figure 7 presents the strain-based fatigue curve
Figure 5. Actual setup for Confined Triaxial Test (CTT). determined, based on the last peak strain as an
initial strain. It was demonstrated that a unique
curve representing the fatigue behavior under sin-
Table 2. Load levels and stress paths for Confined Tri- gle, tandem and tridem axle configurations was
axial Tests (CTT). obtained when the last peak strain was used as an
initial strain, i.e., the resulting fatigue curve was
Single Double Tripple independent of the axle configuration.
Tem- On the other hand, if the first peak strain was
perature Stress σ3 Δq Δq Δq N used as an initial strain, a slightly separate fatigue
(°C) path (kPa) (kPa) (kPa) (kPa) (–) curve was obtained for each axle configuration,
15 1 130 382.0 382.0 382.0 5000
2 509.3 509.3 509.3 5000
3 636.6 636.6 636.6 5000 14
4 382.0 382.0 382.0 5000
12
25 1 130 254.6 254.6 254.6 5000 10
2 382.0 382.0 382.0 5000
3 509.3 509.3 509.3 5000
4 254.6 254.6 254.6 5000

deviatoric and the confining stresses, respectively,

and εa1 and εa2 are the axial strains. 0.1 0.2 0.3 0.4 0.5 0.6
The CTT tests under multiple axle configu- Time (sec)
rations were conducted at two temperatures
(15 and 25°C). The maximum test temperature Figure 6. Definition of initial strain under tridem axles.

486
 1000 . - - - - - - - - - - - - - ; - - - - - - - - - - - - - . , 350

300
(a)

w
,::, 250
<=:
·~ 100 '(;;' 200
"
]
~
150
--Fitted curve • Single
100
o Tandem • Tridem
10 +-~~~~~~~~~~-~~~~-~~~~
50
l.E+03 l.E+04 l.E+05 l.E+06 l.E+07

Number of cycles to failure 0

150 200 250 300 350 400
Figure 7. Fatigue curve based on last peak strain.
Strain (f.LE)

1000 . - - - - - - - - ; - - - - - - - - - ; - - - - - - - - - - - - - - - ,

300

250
'(;;'
~ 200
(/)

~ 150
(/)

100
Single - - S i ngle Tandem
---·Tandem Tridem ··········· Tridem 50
10 +-~~~~~-~~~~-~~~~-~~~~
l.E+03 l.E+04 l.E+05 l.E+06 l.E+07
Number of cycles to failure 450 500 550 600 650 700

Figure 8. Fatigue curve based on first peak strain. Strain (f.LE)

350
(c)
as shown in Figure 8. This was because the first 300
peak strain was not affected by the accumulation
of viscoelastic strain due to multiple axles. Thus, 250
for a given load level (for load controlled test), ,-._

equal initial first peak strains were recorded for all ~ 200
axle groups. However, as the number of axles in (/)

the group increased the number of cycles to fail- ~

(/)
150
ure decreased. Chatti et al. (2009) reported similar
results and thus recommended using the strain- 100
based fatigue curve based on the last peak strain
when analyzing multi-axle configuration of trucks. 50
In addition, the fatigue curves based on the
initial dissipated energy were established. Fig-
350 400 450 500 550 600
ure 9 shows typical hysteresis loops for single, tan- Strain (f.LE)
dem and tridem axle configurations under IDT
fatigue tests. The areas enclosed by the hysteresis Figure 9. Typical hysteresis loops for (a) single, (b)
loops were determined to establish the dissipated tandem and (c) tridem axle configurations under IDT
energy density fatigue curves shown in Figure 10. fatigue tests.
The dissipated energy fatigue curves showed
more spread than the strain-based fatigue curves
(Figs. 7 and 8). Also, as shown in Figure 7, a unique peak strain could represent the fatigue behavior
fatigue curve was sufficient to fit the data points under any combination of multiple axle configu-
for single, tandem and tridem axles. Therefore, a ration. Therefore, this approach was used in the
unique strain- based fatigue curve using the last subsequent analysis to evaluate the impact of the

487
 100 . - - - - - - - - - - - - - - - - - - - - . - - - - - - - - - . - - - - - - - - - , 10000 ,...-------------,------------,------------,---------;---------,

(a)
8000

,.oo
0 0 0 0 0 0 _,_,0 0 0 0 0 0 0 0

_,_, o o o Single
2000
o o o o o o o o ----- Tandem

--Tridem
Single ----Single 0 Tandem
2500 5000 7500 10000 12500 15000
··········... ...
- - - · Tandem Tridem ··········· Tridem Load cycles
·· ...
1 +-~~~~mo--~~~~--~~~~--~~~~
l .E+03 l.E+04 l.E+05 l.E+06 l .E+07 10000

Number of cycles to failure (b)

8000
:::!.

Figure 10. Fatigue curve based on initial dissipated .s 0 0 0

0 0 0

energy. ~ 6000
"E

I
p.,
4000

2000
selected LHVs scenarios. Equations 2 and 3 were --Tridem
used to fit the fatigue test results.
2500 5000 7500 10000 12500 15000

Nf K1 ( )n1 (2) Load cycles

Nf K 2 (w )n2 (3) Figure 11. CTT accumulated permanent deformation

measurements at test temperatures of (a) 15°C and (b) 25°C.

where Nf = number of load cycles to failure;

ε = initial strain; w = initial dissipated strain energy; of axles. Typical registrations of the measured
K1, K2, n1 and n2 are regression constants. tensile strain are shown in Figures 12a, b and c.
A smoothing of the raw data was applied, using a
3.2 CTT permanent deformation 10-point moving average method.
The damage for a given axle configuration was
Figures 11a and b present the measured accumu- calculated by (Huang, 2004),
lated permanent strains under single, tandem and
tridem axles at 15 and 25°C test temperatures. The
1 1 (4)
results shown were the average of 3 specimens, D=
each with 3 extensometers. N f GVW
A clear difference in the development of per-
manent deformation was observed between sin- where D = damage per ton of the LHV;
gle and multiple axle configurations (tandem or GVW = gross vehicle weight. The GVW was cal-
tridem axles). However, a slight difference was culated by multiplying the axle load by the total
observed between the tandem and tridem axles. number of axles. The damage per ton normalized
This was probably due to the viscous behavior of by the damage of the reference LHV (type A shows
the material. in Fig. 4) was given by

3.3 Impact assessment of the LHVs scenarios N ffA GVW

WA N ffA noAxlesA
DX , norm = = (5)
The relative impact of the LHVs was evaluated N fx GVW
WX N fX noAxlesX
by means of the measured and calculated dam-
age per ton of GVW. Three or more samples
were tested for each LHVs scenario to measure where DX,norm = damage per ton of the LHV nor-
the initial strain and number of cycles to failure malized by the damage by the reference LHV
(Nf). The tests conducted were load controlled; type A, NfA = number of cycles to failure of ref-
thus, the amplitude of each pulse was kept equal erence LHV type A; NfX = number of cycles to
and the same amplitude was applied for each type failure of LHV (type B or C in Fig. 4); noAxlesA
of LHV. The amplitude of the pulse was assumed = the number of axles of reference LHV (type A)
to simulate the axle load. Therefore, the GVW of and noAxlesX = the number of axles of type B or
the LHVs was proportional to the total number C LHVs.

488
 140
Table 3. Damage per ton of the LHVs scenarios meas-
(a)
120
ured Nf based on the IDT fatigue tests.

100 Measured No. cycles Damage

w LHV initial strain to failure per ton
80
5 type (με) (Nf) of goods
.5
1:l
CZl
60
A 394 6769 1.00
40
B 411 5756 1.05
20 C 408 5028 0.98

0.2 0.4 0.6 0.8 1.2 1.4 1.6

Time (sec)

140 Table 4. Damage per ton of the LHVs scenarios calcu-

(b) lated using the fatigue curves.
120

Measured No. cycles Damage

100
LHV initial strain to failure per ton
w
5 80 type (με) (Nf) of goods
.5
1:l 60
CZl
A 394 13931 1.00
40 B 411 12244 0.88
C 408 12478 0.81
20

0.2 0.4 0.6 0.8 1.2 1.4 1.6 1.8

Time (sec)

140
Additionally, the damage per ton of the LHVs
(c) was estimated from the CTT permanent defor-
120 mation measurements shown in Figures 11a and
100 b. To achieve this, the relative damage of the
single, tandem and tridem axles was obtained
w 80
:::!_
in each case from the CTT results shown in Fig-
.5
1:l 60 ure 11. Here, a plastic strain level of 4000 με
CZl
(accumulated permanent strain) was selected
40
as the damage criterion and the corresponding
20 number of cycles (Nf) to reach this strain level
was obtained. The 4000 με criterion was selected
0.2 0.4 0.6 0.8 1.2 1.4 1.6 1.8 arbitrarily to stay within the measured CTT per-
Time (sec)
manent strain range. The damage per ton of the
LHVs was then calculated by adding the dam-
Figure 12. Typical registrations of tensile strains under age contribution of the component axle groups.
a) type A, b) type B and c) type C LHVs scenarios. Tables 5 and 6 show the damage per ton of the
LHVs derived from the CTT permanent defor-
mation measurements.
Tables 3 and 4 present the fatigue damage per As shown in Tables 5 and 6, the LHVs types B
ton of the LHVs scenarios. As shown, in terms of and C produced less damage per ton due to perma-
the fatigue damage per ton of transported goods, nent deformation. Like the IDT fatigue test results,
the LHVs type B and C were not more damag- the type C LHV contributed less damage per ton
ing than the currently permitted LHV (type A compared to type B.
in Figure 4) which was used as the reference Note that the damage contributions are added
LHV in the tests. Furthermore, the type C LHV linearly and the effect of rest periods between
caused less damage compared to type B LHV. It the different axle groups was not considered. The
is essential to note that these results were purely effect of material hardening was also not consid-
laboratory simulations and only one type of mix ered. Thus, a more detailed and rigorous analysis
was considered. Furthermore, each simulation is needed to verify the test results. Moreover, the
was conducted by applying a single LHV type load pulses selected for the tests are representative
until the end of the fatigue life of the specimens. of an intermediate thickness of asphalt layers so
Thus, more tests should be conducted to verify the the results may not be representative for thin or
results of this study. very thick asphalt pavement structures.

489
Table 5. Damage per ton of the LHVs scenarios at considered (one for the IDT fatigue test and one
15°C derived from the CTT tests. for CTT); thus, more tests and simulations should
be conducted to verify the results of this study.
Plastic No. cycles Damage
LHV strain level to failure per ton at
type (με) (Nf) 15°C
REFERENCES
A 4000 17200 1.00
B 4000 6600 0.89 Akerman, I. & Jonsson, R. 2007. European Modular
C 4000 6500 0.56 System for road freight transport—Experiences and
possibilities, TFK Report, 2007.
Aurell, J. & Wadman, T. 2007. Vehicle combinations
based on the modular concept, Committee 54: Vehi-
Table 6. Damage per ton of the LHVs scenarios at cles and Trans-ports, Report 1/2007.
25°C derived from the CTT tests. BAST, 2006. Effects of new vehicle concepts on the infra-
structure of the federal trunk road network, Final
Plastic No. cycles Damage Report, Federal Highway Research Institute.
LHV strain level to failure per ton at Cebon, D. 1999. Handbook of Vehicle-Road Interaction.
type (με) (Nf) 25°C Swets & Zeilinger Publishers, Lisse, The Netherlands.
Chatti, K., Manik, A., Salama, H., El Mohtar, C. & Lee,
A 4000 6000 1.00 H.S. 2009. Effect of Michigan multi-axle trucks on
B 4000 5025 0.88 pavement distress, Michigan State University, Depart-
C 4000 5200 0.57 ment of Civil and Environmental Engineering Pave-
ment Research Center of Excellence, Final Report
Volume II Project RC-1504.
Finnish Ministry of Transport and Communications
4 CONCLUSIONS (FMTC), 2002. Nordic vs Central European Vehicle
Configuration; Fuel Economy, Emissions, Vehicle Oper-
The study conducted IDT fatigue and CTT tests ating Costs and Road Wear.
Huang, Y.H. 2004. Pavement Analysis and Design. USA:
to evaluate the impact of LHVs on asphalt pave-
Pearson Prentice Hall.
ments. The tests were conducted under single, tan- Leduc, G. 2009. Longer and Heavier Vehicles: An over-
dem and tridem load pulses. Furthermore, IDT view of technical aspects, JRC Scientific and Technical
fatigue tests for selected LHV scenarios were con- Report, European Commission.
ducted to measure the initial strain and the fatigue Steer D. et al. 2013. A Review of Mega-Trucks: Major
life of asphalt mixtures under different LHVs load- Issues and Case Studies, Directorate General for Inter-
ing. The results demonstrated that, if damage due nal Policies, European Parliament.
to permanent deformation is concerned, the type TRL, 2008. Longer and/or Heavier Good Vehicles-A
B and C LHVs proposed in this study produced Study of the Likely Effects If Permitted in the UK.
TRL Report PPR285.
less damage per ton of goods compared to the per-
VTI 2008. The Effects of Long and Heavy Trucks on the
mitted LHVs. On the other hand, the eleven axles Transport System, VTI report 605.
LHV type C produced less fatigue damage per ton Weissmann, A.J., Weissmann, J., Papagiannakis, A. &
compared to the permitted LHV type A. Kunisetty, J.L. 2013. Potential Impacts of Longer
The results reported here are purely labora- and Heavier Vehicles on Texas Pavements, Journal of
tory simulations and only two mix types were Transportation Engineering, Vol. 139, No. 1.

490
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Investigation of influence of heavy traffic loads on asphalt pavement

response by SAFEM

P. Liu, D. Wang & M. Oeser

Institute of Highway Engineering, RWTH Aachen University, Germany

ABSTRACT: A computational model which is able to calculate the response of the pavement fast and
precisely is greatly important to accurately assess the impact of heavy traffic loads on asphalt pavements.
A specific program SAFEM was developed based on a semi-analytical finite element method for this
objective. It is a three-dimensional FE program that requires only a two-dimensional mesh by incorporat-
ing the semi-analytical method using Fourier series in the third dimension. The impact of heavy traffic
loads was analyzed in terms of stress distribution, surface deflection and fatigue life. The results indicate
that the SAFEM-program is an efficient and fast tool that is capable of accurately predicting the struc-
tural response of pavements to traffic loads.

1 INTRODUCTION

In the past decades the Finite Element (FE) method

has been developed rapidly and was increasingly
used in many industrial fields as well as in the rou-
tine pavement design and assessment process. But
several limitations exist in the conventional FE
packages such as ABAQUS, e.g. the complexity
and hence the time-consuming user training proc-
ess often renders it impractical to be used by a road
engineer. The specifically developed FE tools for
pavement analysis usually offer the results quickly
but sometimes the accuracy is relatively low due
Figure 1. Pavement structure geometry and load
to oversimplifications of the modelling, e.g. from mode.
a Three-Dimensional (3D) condition to a Two-
Dimensional (2D) plane strain or axisymmetric one.
Therefore, it is necessary to find means that both In the following sections, the mathematical basis
improve the computational speed without increas- of the SAFEM will be shortly introduced, and the
ing the resource requirement and still offer the com- impact of heavy traffic loads on the asphalt pave-
putational accuracy. The Semi-Analytical Finite ment is analyzed. A brief summary and conclu-
Element Method (SAFEM) is one of the methods sions are provided at the end of this paper.
which can meet the requirements. For a typical
pavement structure problem as seen in Figure 1,
the geometry and material properties usually do 2 DESCRIPTION OF SEMI-ANALYTICAL
not vary in one of the coordinate direction (out- FINITE ELEMENT METHOD
of-plane direction, for this case the z-direction),
but the boundary conditions, e.g., the load terms, The first step in the FE formulation is to express the
exhibit a significant variation in that direction. As element coordinates and element displacements in
a result, the pavement structure problem could not the form of interpolations using the natural coor-
be simplified as a 2D plane strain case. However, by dinate system of the element. By using SAFEM,
assuming that the displacements in the geometrical the general form of the shape functions defining
out-of-plane direction can be represented using a the variation of displacements can be written as a
Fourier series and exploiting its orthogonal proper- Fourier series in which z ranges between zero and a.
ties, the problem of such a class can be numerically The pavement is assumed to be held at z = 0 and
solved by a series of 2D FE-meshes which will be z = a in a manner preventing all displacements in
shown later (Fritz 2002, Hu et al. 2008, Liu et al. the XY plane but permitting unrestricted motion
2013, Liu et al. 2014, Zienkiewicz & Taylor 2005). in the z-direction, as shown in Figure 2 (Fritz 2002,
491
 to obtain the strain and stress of one element.
The strain-displacement matrix Bkl is defined as
follows:

⎡ ∂N k lπ z ⎤
⎢ ∂x sin a 0 0 ⎥
⎢ ⎥
⎢ ∂N k lπ z ⎥
0 si
sin 0
⎢ ∂y a ⎥
⎢ ⎥
⎢ lπ lπ z ⎥
0 0 − N k sini
⎢ a a ⎥
Bk = ⎢
l
⎥
Figure 2. Schematic representation of an SAFEM ⎢ ∂N k sin lπ z ∂N k
sin
lπ z
0 ⎥
situation. ⎢ ∂y a ∂x a ⎥
⎢ ⎥
⎢ lπ lπ z ∂N k lπ z ⎥
0 N k cos cos
⎢ a a ∂y a ⎥
Hu et al. 2008, Liu et al. 2013, Liu et al. 2014, ⎢ ⎥
Zienkiewicz & Taylor 2005): ⎢ lπ N cos lπ z 0
∂N k
c s
co
lπ z ⎥
⎢⎣ a k a ∂x a ⎥⎦
⎧⎪u ⎫⎪ (4)
U = ⎨v ⎬
⎩⎪w ⎭⎪ By using the principle of minimum potential
⎡ lπ z ⎤ energy, a typical sub-matrix of the element stiff-
⎢sin a 0 0 ⎥ ness matrix (Klm)e is (Hu et al. 2008):
⎢ ⎥ ⎧ukl ⎫
L 6
⎪ ⎪
= ∑ ∑ N k ⎢⎢ 0 0 ⎥⎥ ⎨vkl ⎬
l z
(K l )e ∫∫∫vvol (B l )
T
si DB m dxdydz (5)
a
⎥ ⎪⎩wk ⎪⎭
l =1 k =1 l
⎢
⎢ 0 lπ z ⎥
0 cos A typical term for the force vector becomes:
⎢⎣ a ⎥⎦
L
= ∑ N l ⋅U l (F l )e ∫∫∫vvol (N l ) { p}l dxdydz
T
d (6)
l =1
(1)
where ukl , vkl and wkl are the displacements of the From Equations 4 and 5, the stiffness matrix of
node at the term of the Fourier series along x-, y- one element includes (Zienkiewicz & Taylor 2005):
and z-directions, respectively. l identifies the term
a lπ z mπ z
of the Fourier series and L is the number of terms I1 = ∫ sin ⋅ cos ⋅ dz
considered; Nk is the shape functions of the node 0 a a
in the XY plane which are the same as those for a lπ z mπ z
I 2 = ∫ sin ⋅ siin ⋅ dz (7)
displacement approximation used in 2D prob- 0 a a
lems, and their definition in detail can be found in a lπ z mπ z
I 3 = ∫ cos ⋅ cos ⋅ dz
(Zienkiewicz & Taylor 2005). 0 a a
The loading function defining the variation of
load along the z-direction is given a similar form as The integrals exhibit orthogonal properties
the displacements (Hu et al. 2008): which ensure that (Zienkiewicz & Taylor 2005):
L L
l z ⎧1
∑ p ( x y) = ∑ { p}
l
(2) ⎪ a ffor l m
l =1 a l =1 I2 I3 = ⎨ 2 (8)
⎪⎩ 0, for l ≠ m
n
⎛ 2 Pt ⎞ ⎡ lπ lπ ⎤
p ( x, y ) ∑ ⎝ lπ ⎠ ⎢⎣cos a Z t1 − cos Zt 2 ⎥ (3)
t =1 a ⎦ Only when l and m are both odd or even num-
bers, the first integral I1 is zero. Due to the special
where Pt is the tire load pressure; Zt1 is the z coor- structure of the Bl matrix, all terms that include
dinate where the tire load starts; Zt2 is the z coordi- I1 become zero. This means that the matrix (Klm)e
nate where the tire load ends. becomes diagonal. Thus, the stiffness matrix can
After determining the element displacement, be reduced and the final assembled equations have
geometrical and physical equations can be used the following form:

492
⎡ K 11 ⎤ ⎧ U1 ⎫ ⎧ F1 ⎫
⎢ ⎥ ⎪ U 2 ⎪ ⎪F 2 ⎪
⎢ K 22 ⎥ ⎪⎨ ⎪ ⎪ ⎪
⎬+⎨ ⎬= 0 (9)
⎢  ⎥⎪  ⎪ ⎪  ⎪
⎢ ⎥
⎣ K LL ⎦ ⎪⎩U LL ⎪⎭ ⎪⎩F L ⎪⎭

The Equation 9 shows that the large system of

Figure 3. The configuration of single axle with dual
equations splits up into L separate problems, i.e., wheels.
the equilibrium equations are fully decoupled for
each harmonic of Fourier series, which can be well
adapted to the individual processors of a parallel
Table 2. Loading parameters of the asphalt pavement
computer, and thus reduce the computational time
(Yu et al. 2013).
significantly compared to the sequential solving
procedure. Distance
The accuracy and efficiency of SAFEM have Axle Contact Contact Wheel Wheel between
been validated by former studies (Liu et al. 2017a, load pressure area width length wheels
2017b).
P p A W L D
kg MPa mm2 mm mm mm
3 ANALYSIS OF THE IMPACT OF HEAVY 10000 0.70 35256 156 226 320
TRAFFIC LOADS ON THE ASPHALT 12500 0.76 40172 166 242 320
PAVEMENT 15000 0.82 45056 176 256 320
17500 0.85 50220 186 270 320
3.1 The definition of models 20000 0.89 55185 195 283 320

The thicknesses and the material properties of the

asphalt pavement layers are listed in Table 1. The
excess length in both ends of the traffic direction which is proved by former study (Liu et al. 2017a),
was defined as 20 times of the loading radius to therefore only one pair was loaded in the center of
limit the computational time. The width of the the pavement surface. Considering different work-
pavement was defined as 3750 mm (Gohl 2006, ing conditions, five different levels of axle loads
Rabe 2004, 2007, 2014). The upper three asphalt were applied. The corresponding loading param-
layers were totally bound. The two contact layers eters are shown in Table 2.
among the asphalt base course, gravel base layer,
frost protection layer and sub-grade were defined 3.2 The response of the asphalt pavement
as being partially bound, which means same dis- structure under the heavy traffic loads
placement happens in vertical direction but differ-
ent displacements in horizontal directions. With an increase of the axle load, the contact pres-
A standard single axle with dual wheels was uti- sure between the tire and pavement increases. As a
lized and its configuration can be seen in Figure 3. result the vertical compressive stress in the asphalt
Assuming that the influence of one pair of dual pavement structure will increase, which will cause
wheels on the other pair on the other end is small rutting. The variation tendencies of the compres-
enough it can be neglected to simplify the model, sive stress distribution along the depth of the
asphalt pavement subjected to the different axle
loads are similar, seen from Figure 4. The com-
Table 1. Thicknesses and material properties of the pressive stress caused by larger axle load is obvi-
pavement layers. ously larger than that caused by smaller one, which
means an overweight load will lead to more serious
Thickness E destruction of the asphalt pavement.
The values of the compressive stresses on the
Layer Mm μ MPa top of each pavement layer are listed in Table 3.
The values decrease as the depth from the pave-
Surface course 40 0.35 11150
ment surface increases. The upper three asphalt
Binder course 50 0.35 10435
layers bear much larger compressive stress than
Asphalt base course 110 0.35 6893
those lower ones. The compressive stress on the
Gravel base layer 150 0.49 157.8
top of the surface course caused by the axle load
Frost protection layer 570 0.49 125.7
Sub-grade 2000 0.49 98.9
of 20000 kg is 1.11 times that caused by a stand-
ard axle load of 10000 kg. The compressive stress

493
 0.2 0 .4 0 .6 0.8 a significantly increased compressive stress in
'E the asphalt pavement structure. The compressive
.s
Q)
Surface cours·
stress in asphalt binder course is relatively large
u
~
40 I I I I I I l.oo!l
and increases more significantly; the service life of
~
:::J
f/) Binder course 1/ ~ ....- the asphalt pavement will decrease if this phenom-
c
Q) enon is not considered enough.
80
E
Q)
>
ro ~
Asphalt base course
Similarly, the surface deflection of the asphalt
a.
120 pavement will increase as the axle load increases.
~ - Axle load = 10000 kg
The deflection distribution along the traffic direc-
E - Axle load = 12500 kg
.g 160
r- h ~ - Axle load = 15000 kg tion offset from the midpoint between the dual
.c
a.
Q) II'A
"" -
Axle
Axle
load = 17500 kg
load= 20000 kg
wheels centers is shown in Figure 5.
0
200
The maximum surface deflections due to differ-
Stress [MPa] ent axle loads are listed in Table 4. The deflection
(a) The depth from surface course to asphalt base course
increases to 159% when the axle load of 10000 kg
increases to 20000 kg, i.e., the surface deflection
0.03 0.06 0.09 0.12 0.15 is sensitive to the change of the axle load. It indi-
'E 2oo cates that when the asphalt pavement is designed
.s
Q) }/
I "'- 1
'1'..
for a region with many heavy traffic loads, it
Gravel base layer
~ 600 L...L.....L.: should be considered that the thicknesses and stiff-
i)l Frost protection layer ness of the pavement structural layers should be
c
~ 1000 II"". increased adequately in order to support the sur-
~
ro
Sub-grade face deflection.
a. I I I I I
Q) 1400 - Axle load= 10000 kg
The fatigue life of the asphalt pavement sub-
£
E - Axle load= 12500 kg jected to the different axle loads can be calculated
e
~ 1800
- Axle load= 15000 kg according to the Equation 10 (FGSV 2009):
Axle load= 17500 kg
a.
Q) - Axle load = 20000 kg
0
2200
Ninsitu N f ⋅ 2.8283 ⋅ ε −4.194 (10)
Stress [MPa]

(b) The depth from gravel base layer to sub-grade. where Ninsitu is the number of load cycles until the
macro-crack initiates; Nf is the shift factor, for this
Figure 4. Comparison of the compressive stress distri- case Nf = 1540; ε is the computational tensile strain
bution along the depth of the pavement derived from the at the bottom of the asphalt base course.
different axle loads.

0.00
Table 3. The compressive stresses (MPa) on the top of
each pavement structure layer. -0.15

Axle load Surface Asphalt -0.30

kg course Binder course base course 'E
E
'-;-0.45
0 Axle load 10000 kg ~
10000 0.801 0.663 0.410 u
12500 0.845 0.720 0.451 ~-0 . 60 Axle load 12500 kg
~
r-
0 Axle load 15000 kg r-
r-
15000 0.871 0.776 0.490 -0.75 Axle load 17500 kg r-
r-
17500 0.874 0.803 0.511 Axle load 20000 kg
~
20000 0.891 0.840 0.538 -0.90
500 1000 1500 2000
Axle load Gravel Frost Offset from midpoint between centers of the wheels [mm)
kg base layer protection layer Sub-grade
Figure 5. The deflection distribution offset from the
10000 0.088 0.076 0.037 midpoint between the centers of the dual wheels.
12500 0.101 0.088 0.042
15000 0.115 0.099 0.048 Table 4. The maximum surface deflections due to
17500 0.125 0.109 0.053 different axle loads.
20000 0.136 0.119 0.058
Axle load
kg 10000 12500 15000 17500 20000
on the top of the binder course caused by an axle
Maximum
load of 17500 kg is already larger than that on the
deflection
top of the surface course caused by a 10000 kg mm −0.511 −0.594 −0.678 −0.742 −0.814
axle load. As a result, overweight loads will cause

494
Table 5. The tensile strain and the fatigue life of the FOR 2089. The authors are solely responsible for
asphalt pavement subjected to different axle loads. the content.
Axle load
kg 10000 12500 15000 17500 20000 REFERENCES
Tensile
FGSV, (2009). Instructions for determining the stiffness
strain
and fatigue performance of asphalt with the dynamic
μm/m 193 224 255 279 306
indirect tensile test as an input variable in the dimen-
Fatigue life 1.67e19 8.86e18 5.09e18 3.51e18 2.38e18 sioning, AL Sp-09 asphalt, 2009 edition. FGSV Pub-
lisher, Research Society for Road and Transportation,
Cologne. (in German).
Fritz, J.J. (2002). “Flexible pavement response evalua-
The tensile strain and fatigue life of the asphalt tion using the semi-analytical finite element method”.
pavement is shown in Table 5. The theoretical value International Journal of Materials and Pavement
seems large enough and thus the fatigue may be not Design, Vol 3(2), 211–225.
a big deal. However, considering the other influenc- Gohl, S. (2006). “Vergleich der gemessenen mechanischen
Beanspruchungen der Modellstraße der BASt mit den
ing factors such as climate, the fatigue life would be Berechnungsergebnissen ausgewählter Programme”.
much less than the theoretical value. And according TU Dresden, Professur für Straßenbau, Diplomarbeit
to the computation, the fatigue life decreases to 14% (In German).
when the axle load is from 10000 kg to 20000 kg, Hu, S., Hu, X., Zhou, F. (2008). “Using semi-analytical
i.e., the increase of the axle load will significantly finite element method to evaluate stress intensity factors
speed up the destruction of the asphalt pavement. in pavement structure”. Pavement Cracking, 637–646.
Liu, P., Wang, D., Oeser, M. (2013). “Leistungsfähige
semi-analytische Methoden zur Berechnung von
4 SUMMARY AND CONCLUSION Asphaltbefestigungen”, Tangungsband, 3. Dresdner
Asphalttage (in German).
Liu, P., Wang, D., Oeser, M., Chen, X., (2014). “Einsatz
This paper proposes to use the SAFEM for pre- der Semi-Analytischen Finite-Elemente-Methode zur
dicting the asphalt pavement structural responses Beanspruchungszustände von Asphaltbefestigungen”.
under heavy traffic loads to different extents. The Bauingenieur, Vol 89(7/8), 333–339 (in German).
compressive stress distribution, surface deflection Liu, P., Wang, D., Otto, F., Hu, J., Oeser, M. (2017a).
and fatigue life derived from the SAFEM is ana- “Application of Semi-Analytical Finite Element
lysed to investigate the impact of the heavy traffic Method to Evaluate Asphalt Pavement Bearing
load on the asphalt pavement. The results indicate Capacity”, International Journal of Pavement Engi-
that the overweight load will have a serious impact neering. DOI: 10.1080/10298436.2016.1175562. http://
dx.doi.org/10.1080/10298436.2016.1175562.
on the destruction of asphalt pavement structure. Liu, P., Wang, D., Hu, J., Oeser, M. (2017b). SAFEM – Soft-
For further investigation, the SAFEM allows the ware with Graphical User Interface for Fast and Accurate
application of dynamic analysis and various mate- Finite Element Analysis of Asphalt Pavements, Journal
rial properties, such as viscoelasticity for asphalt of Testing and Evaluation. Vol. 45, No. 4, pp. 1–15.
and nonlinear elasticity for the sub-base of the Rabe, R. (2004). “Bau einer instrumentierten Modellstraße
pavement. In order to further reduce the compu- in Asphaltbauweise zur messtechnischen Erfassung der
tational time, infinite elements should be coupled Beanspruchungssituation im Straßenaufbau”, AP 03
with the finite element model. With these improve- 342, interner Bericht, Bundesanstalt für Straßenwesen,
ments, the SAFEM should be more appropriate to Bergisch Gladbach.
Rabe, R. (2007). “Messtechnische Erfassung der
predict the impact of the heavy traffic load on the Beanspruchungen im Straßenaufbau infolge LKW-
asphalt pavement. Überfahrten über eine Modellstraße in Asphaltbau-
weise”, AP 04 342, interner Bericht, Bundesanstalt für
Straßenwesen, Bergisch Gladbach.
ACKNOWLEDGEMENTS Rabe, R. (2014). “Angaben zum Aufbau der Modell-
straße und Angabe von ausgewählten Ergebnissen
This paper is based on parts of the research projects und Materialkennwerten”, Bergisch Gladbach.
carried out at the request of the Federal Ministry Yu, L., Yang, J., Bao, L., Yan, F. (2013). “Analysis on
of Transport and Digital Infrastructure, requested the mechanical responds of asphalt pavement with
subbase used fly-ash-flushed by seawater under heavy
by the Federal Highway Research Institute, under traffic”. Journal of Shenyang Jianzhu University (Nat-
research projects No. 04.0259/2012/NGB and FE ural Science). Vol. 29, No. 1, 110–115 (in Chinese).
88.0137/FE88.0138, as well as parts of the research Zienkiewicz, O.C., Taylor, R.L. (2005). The finite element
project carried out at the request of the German method for solid and structural mechanics. 6th edition,
Research Foundation, under research projects No. Elsevier Butterworth-Heinemann, Oxford, 498–516.

495
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

The increase of pavement cost due to the traffic overloads

J.C. Pais & P. Pereira

University of Minho, Guimarães, Portugal

ABSTRACT: The traffic on road pavements is characterized by a large number of different vehicle
types, and these vehicles are considered in the pavement design by converting their effects through the use
of truck factors which transform the damage they apply to the pavement in the damage that would be
applied by a standard axle. However, there is an important number of vehicles that travel with axle loads
above the maximum legal limit. These axles/vehicles cause significant damage to the pavements which
increases the pavement construction and rehabilitation cost. This paper investigates the impact of over-
loaded vehicles using a weight database by studying the truck factors for different vehicle categories. The
study concluded that the presence of overloaded vehicles increases pavement damage and life cycle costs
compared to the cost of the same vehicles with legal loads.

1 INTRODUCTION fatigue life of flexible pavements based on weigh-

in-motion data and concluded that the increase of
Among the factors responsible by the distresses that percentage of overloaded vehicles from 0% to 20%
appear in the pavement, the traffic is the main cause can reduce the fatigue life of asphalt pavement up
of pavement problems due to the loads applied by to 50%. This result is due to the fact that in Poland
the axles and tires of vehicles. While light traffic the percentage of overloaded vehicles is in the
doesn’t cause structural problems in the pavement, range from 6% (road where continuous control of
heavy traffic causes the most important failures in traffic is performed) to 16.5%. The study concluded
the pavements producing fatigue cracking and rut- that most of overloaded vehicles exceeded the axle
ting that require pavement rehabilitation. load limit, while the gross weight was exceeded less
The main concern in traffic characterization frequently.
for pavement analysis and design is due to the fact Because the knowledge of the impact of over-
that it presents a large number of different types loads are important for pavement analysis and
of vehicles with different carrying loads, different design as recognized by Pais et al. (2013), this work
number of axles and group of axles (e.g., single analyzes the effects of overloads on pavements in
axles when the distance between axles is large, tan- terms of ESAL, required pavement thickness and
dem axles when two axles are very close and far pavement costs by studying a traffic records of
from the other axles, or tridem axles when three 420 manual weighs obtained in a national road,
axles are very close and far from the other axles). divided into 4 classes of vehicles.
The consideration of the traffic for pavement This work analyzes the percentage of vehicles
design is based in the expected number of vehicles in each traffic class, percentage of overloads and
and its loads. The number of vehicles is considered average axle loads for each type of vehicle. The
based on the actual traffic and the traffic growth impact of the overloads in the pavement perform-
while the traffic loads can be obtained by consid- ance is analyzed by studying the ESAL, required
ering the maximum legal load of each vehicle or pavement thickness and pavement cost consider-
using weight measurement from weigh-in-motion ing the following conditions: i) Vehicle with the
systems which record the weight of each axle of maximum legal load in each axle; ii) Vehicle with
the vehicles. While the second takes into account average observed load in each axle; iii) All vehicles
the exact loads of the axles, the first doesn’t but the observed in this work; iv) Legal vehicles (not con-
effect of the overloaded vehicles can be compen- sidering the overloaded vehicles); v) Overloaded
sated by the vehicles that circulate below the maxi- vehicles (vehicles with at least one axle weighing
mum limit. more than the maximum legal load); vi) Vehicles
The effect of the overloads in the pavement limited to the legal limits (vehicles where the load
analysis and design was studied by Rys et al. (2016) of the overloaded axles is considered equal to the
who analyzed the effect of overloaded vehicles on maximum legal limit for that axle).

497
2 TRAFFIC DATA axles (class F1). Classes F2 and H3 have the same
amount of vehicles, around 15%.
The road traffic has weight limitations due to tech- More important than the number or the per-
nical, economic and competitive factors and these centage of heavy vehicles in a road, is to know the
weight limits are function of the number of axles number of overloaded vehicles, meaning that at
and the axle configuration. Tandem and tridem least one axle has a weight greater than its maxi-
axles have different weight limits because the total mum legal limit. These results are presented in
load of the axle is applied in several points whereas Figure 2 where it can be observed that for classes
in the single axles the load is applied only in one. F2, H3 and H5 (vehicles with more than two axles)
In Portugal, heavy vehicles are classified into the percentage of overloaded vehicles reach more
different classes based on vehicle configuration than 50% of the vehicles in that class. Only for
and the number of axles. The classes analyzed in class F1 (vehicles with two axles) the percentage
this study are described in Table 1 where it can be of overloaded vehicles is reduced but even so it
observed the load limit of each axle. Tandem and reaches almost 20%.
tridem axles are considered as a set of axles. The The distribution of the axle loads inside all
limit on total vehicle load is the sum of the maxi- classes allows to realize the amount of overload
mum load for each axle. The load limits indicated axles. This is presented in Figures 3 through 6,
in the table are used to define the overloads for respectively for F1, F2, H3 and H5 class.
the traffic considered in this work. In this work an For class F1 (Figure 3), one can observe that the
overloaded vehicle has at least one axle carrying a loads of the first axle are very low and almost are
weight greater than that allowed for the axle. below the legal limit. Only 14% of the first axles
For the development of this work, 420 manual have overloads. For the second axle, the percentage
weighs were carried out on a two lanes main road, is identical, i.e. 13%.
during a period of one month, not in consecutive For class F2 (Figure 4), it is notorious that
days to avoid that the overloaded vehicles escape more than 50% of the axles have overload. In fact,
for other roads or decrease the typical loads. The 56% of the first axle have overloads while 61% of
measurements were done by weighing the load of the tandem axle (second + third) have overloads.
both sides of each axle. On average, the axle load
was equally divided by both sides of the axles. 50
The classes presented in Table 1 follows the 45
trend observed by Pais et al. (2013) where, for a 40
motorway, and using a weigh-in-motion system, .,
V> 35
u
the observed traffic was concentrated mainly in :;:: 30
~
F1, H3 and H5 classes. 25
~
In this study, the percentage of vehicles are pre- ~
20

sented in Figure 1 where it can be observed that ~ 15

almost half of the traffic is composed by vehicles 10

in the H5 class that has a tractor with two single

axles and a trailer with a tridem. One fourth of f1 f2 H3 H5
the traffic is composed by trucks with two single Vehicle class

Figure 1. Percentage of heavy vehicles observed in the

Table 1. Vehicle classes and maximum legal load. study.

Maximum load per axle (kN) 90

1st 2nd 3rd 4th 5th ., 80

u
:;::
Vehicle classes axle axle axle axle axle ~
70

.,
"tJ
60
F1 75 120 "'0 50
~ 40
'0
F2 75 190
f
30

20

H3 75 120 190 ~ 10

F1 F2 H3 H5
H5 75 120 240 Vehicle class

Figure 2. Percentage of overloaded vehicles.

498
 180 400
0 Axle 1 - limitAxlel 0 Axle 1 - LimitAxle l
160 350 0 Axle 2 - LimitAxle2
0 Axle 2 - LimitAxle2
0 Axle 3+4+5 - limit Axle 3+4+5
140 300

~
120
"'0 100
~ 250

"'
..Q 0
200
\0
80 0
"'
~ ~
150
60

40
50
20

50 100 150 200

20 40 60 80 100 120
Observations
Observations

Figure 3. Distribution of the loads of the vehicles of Figure 6. Distribution of the loads of the vehicles of
class F1. class H5.

350
0 Axle 1 - LimitAxle1
culate with overloads. For the second the load
300
0 Axle 2+3
0 variation is larger and the percentage of over-
250 loads reached 42% but it is in the tridem that
~ 200
the load variation has a huge variation and the
"'0 number of overloaded axles was 32%. Consider-
150
ocJXP ing this type of vehicles, 61% of them circulated
Ji 100
cOO
0
0 with overloads.
50

10 20 30 40 so 60
3 MODELS
Observations

Figure 4. Distribution of the loads of the vehicles of

The impact of the vehicles and mainly the overloads
class F2. on pavement performance was studied by convert-
ing all axle loads and vehicles into a representative
axle, i.e., which in pavement design is referred to as
350 a standard axle resulting in the Equivalent Single
0 Axle 1 - LimitAxle1 0
300 0 Axle 2 - LimitAxle2
Axle Load (ESAL).
250
The ESAL is generally expressed as the relation-
[ ship between the actual axle load (Px) and the load
of the standard axle (Pref), mainly with α = 4, even
200
"'0
"'
..Q
"'
150
though it is recognized that there is not an unique
~ 100 power value and it varies with pavement type,
so distress considered, failure level, contact stresses
(Hong and Prozzi, 2006) (Equation 1).
10 20 30 40 so 60
Observations α
⎛P ⎞
ESAL = ⎜ x ⎟ (1)
Figure 5. Distribution of the loads of the vehicles of ⎝ P80 ⎠
class H3.

To take into account the type of axle, i.e., sin-

In terms of vehicle, 80% of F2 class circulated with gle, tandem or tridem, LCPC (1994) proposed
overloads. Equation 2 for the calculation of ESAL, based
For class H3 (Figure 5), the conditioning axle on the previous and adding the coefficient k that
is the tandem composed by axle 3 and 4 that pre- is a function of the type of axle (single, tandem
sented 66% of overloads while the first and the or tridem) and α is a coefficient that is a func-
second axles circulated only with 17 and 25% tion of the type of pavement (most importantly,
of overloads, respectively. In terms of vehicles, the pavement stiffness). K coefficient is given in
83% of this class H3 have at least one axle with Table 2.
overloads. α
Axle loads distribution of class H5 presented ⎛P ⎞
ESAL = k ⎜ x ⎟ (2)
in Figure 6 shows that the first axle of all vehi-
cles have almost the same load and only 15% cir- ⎝ P80 ⎠

499
Table 2. Values of the k and α coefficients for the
(H ) ( )
3 3
French method (LCPC, 1994). assp Easp + H gra E gra
ET = 3 (4)
k E subg

Single Tandem Tridem For tandem and tridem axles, coefficient k of

α axle axle axle Equation 3 was defined to be used considering all
Flexible Pavement 5 1 0.75 1.1
the set of axles of tandem and tridem axles as only
Rigid and Semi-rigid 12 1 12 113
one axle, on what the load is the sum of the load of
Pavements the individual axles.
For the calculation of ESAL for the individual
axles of tandem and tridem, k coefficient defined
Table 3. Axle Parameter (AP). in Equation 3 must be calculated according Equa-
tion 5 and 6, respectively for tandem and tridem,
Single axle Single axle Tandem axle where kset is the k coefficient defined for the set of
Single wheel Dual wheel Single wheel axles and kindividual is the k coefficient to be applied
for the individual axles.
1.0 2.0 2.7
K individual kset .2α −1 (5)
Tandem axle Tridem axle Tridem axle
Dual wheel Single wheel Dual wheel K individual kset .3α −1 (6)
4.1 3.8 5.2

4 ANALYSIS OF THE EFFECT OF THE

Table 4. Constants for Equation 3. OVERLOADS

α ET (m) a1 a2 a3 The analysis of the effect of the overloads in pave-

ment cost was made calculating the ESAL for two
4 ≤1.2 1.08E+01 −9.41E-01 6.69E-02 different cases: for the axles and for the vehicles
>1.2 5.20E+00 3.33E-02 1.82E-03 and for both applying the model defined in Equa-
α ET (m) a4 a5 a6
tion 1, referred in this work as 4PW model, the
model defined in Equation 2, referred in this work
4 ≤1.2 −2.85E-01 3.04E-01 −1.41E+00 as FR model, and the model defined in Equa-
>1.2 1.15E-01 −1.17E-01 −1.33E+00 tion 3, referred in this model as JP model. These
models have significant differences producing dif-
ferent results as function of the type of axle and in
Because the effect of a load into a pavement the JP model the results are function of the type of
is completely different of the effect into another wheel and pavement.
pavement, coefficient k of Equation 2 must quan- The differences of these models can be observed
tity this effect. Also the effect of a single load or in Figure 7 for the second axle of F2 vehicles (sin-
a dual load have different effects into the pave- gle axle + tandem axle) where 4PW model gives the
ment, coefficient k can also be used to measure highest ESAL followed by the FR model that cor-
this effect. Thus, Pais and Pereira (2016) proposed responds to 75% of the 4PW model while JP model
a model to calculate the coefficient k as function presents the lowest values due to the fact that the
of the pavement constitution and type of axle and axle has a dual wheel. Because the load is divided
wheel. This coefficient is expressed by Equation 3, 16
where Hasp is the thickness of the asphalt layer e Axle 2: Modei4PW

(m), Hgra is the thickness of the granular layer (m),

14
• Axle 2: Model FR •
Easp is the stiffness of the asphalt layer (MPa), 12
e Axl• /: Morl•IIP •
Esubg is the stiffness of the subgrade (MPa), AP is 10 •
the Axle Parameter as defined in Table 2, and a1, ••
.....
a2, a3, a4, a5, and a6 are the constants presented in
-~- ...........·.
Table 4 for α = 4.
Fjjllli!!F~-:-:-......:
k=a1.H
H assp H gra Eassp Esubg a5 × e a6 AP (3) 0 Ill
10 20 30 40 so 60
Vehicles
The coefficient ET (Equivalent Thickness) of
Table 4 is defined in Equation 4. Figure 7. ESAL for the second axle of F2 vehicles.

500
 20 0.40
18 0 Axle 4: Modei4PW

16 O A>de4:ModeiFR 0.3S

14 O Axle 4: ModeiJP
I o.3o
12
~
~ 0.25
~
10
.!,!
~ 0.20
E
~ 0.1S

~ 0.10

80 100 o.os
on O T2 O T3 O T4 O TS O T6
Vehicles
0.00
10 20 30 40 so 60
Figure 8. ESAL for the fourth axle of H5 vehicles. Vehicles

Figure 10. Pavement thickness for all vehicles of class

Table 5. Traffic levels. H3 limited to the maximum axle load.
Traffic level T1 T2 T3 T4 T5 T6
0.14
on O T2 O T3 O T4 O TS O T6
Trucks (1E6) 18 14 8.9 5.0 2.7 1.0 0.12
E"
j 0.10

0.60 -t; 0.08

~
E o.oG
~
~ 0.04

0.02

30 40 so 60
Vehicles

Figure 11. Pavement thickness due to the overloads for

0.00
10 20 30 40 so 60
all vehicles of class H3.
Veh icles

ranges from 0.30 m up to 0.49 m, meaning that the

Figure 9. Pavement thickness for all vehicles of class
range of weight of these vehicles varies consider-
H3.
ably. This graph was made by the truck factors
ordered in increasing order meaning that the last
by two wheels, its effect in the pavement is reduced vehicles correspond to the overloaded vehicles.
and the ESAL given by JP model is low. However, The required pavement thickness was also cal-
for single loads the JP model gives great ESAL culated for all vehicles but limited to the maximum
when compared with the other models as indicated axle legal load. For the legal vehicles the axle loads
in Figure 8 for the fourth axle of the H5 vehicles. were kept constant but for the axles that exceed the
For this case, the ESAL given by FR model is 10% legal load limit its load was reduced to the legal
greater than the one obtained by 4PW model. limit. Thus, all vehicles considered in this analysis
The analysis of the impact of the overloads on can be considered as legal vehicles. The required
pavement performance was established by the cal- pavement thickness for these vehicles limited to the
culation of the asphalt layer thickness required for legal load is represented in Figure 10. These results
a pavement to support a given traffic level defined are equal to the one presented in Figure 9 for all
in Table 5. The pavement was defined with the legal vehicles. For the overloaded vehicles the
following characteristics: thickness was reduced and the direct comparison
between both figures gives the thickness due to the
• Stiffness of the asphalt layer: 5000 MPa
overload which is represented in Figure 11.
• Thickness of the granular layer: 0.20 m
Knowing the pavement thickness due to the
• Stiffness of the granular layer: 120 MPa
overloads, and because the increase of pavement
• Stiffness of the subgrade: 60 MPa
cost is due to that increase of thickness, the increase
The calculation of the required thickness for the in pavement cost can be calculated which is shown
vehicles of class H3, and considering the track fac- in Figure 12. Two conclusions can be drawn from
tor defined by the JP model, is represented in Fig- this increase of prices, namely the traffic level as a
ure 9. It is possible to verify that for example for reduced effect on the increase of pavement cost but
T1 class the thickness required for the pavement the effect is greater for lower traffic levels and the

501
 40
O Tl O T2 O T3 O T4 O TS O T6 0
~ 35 i
t; 30
8
E
E
25
i~~
~ 20 8
'0 15 i

10 20 30 40 50 60

Figure 14. Increase of truck factor due to the overloads.

Figure 12. Increase of pavement cost due to the over-
loads for all vehicles of class H3. 0.36

40

• Vehicles with maximum legal load

35
• Vehicle with average load - Vehicles with maximum legal load
30
• All vehicles __._ Vehicle with ave rage load

25 • Legal vehicles
__..__ Legal vehides
• Overloaded vehicles
20 0.11 ~ OVer l oaded vehicles
• Vehicles limited to the legal limits
- Vehicles limited to the legal limits
15

10
Traffic level

Figure 15. Pavement thickness for all traffic configura-

Fl F2 H3 H5
tions studied.
Vehicle class

Figure 13. Truck factor for all traffic.

increase for this class of vehicles can reach more

than 30%.
In this study an analysis of the truck factor was
carried out for different configurations of the traf-
fic, namely for i) Vehicle with the maximum legal
load in each axle; ii) Vehicle with average observed
load in each axle; iii) All vehicles observed in this l.OOE+07
Traffic level (N80)
work; iv) Legal vehicles; v) Overloaded vehicles; vi)
Vehicles limited to the legal limits. The truck factor
Figure 16. Increase of pavement thickness due to the
for these configurations is indicated in Figure 13 overloads function of the traffic level.
and allows to have an indication about the effect
of the legal and the overloads and with maximum
and average load and support the definition of design. Without overloads, the design traffic can
truck factors to be used in pavement design and be significantly reduced producing pavement less
analysis. The use of truck factors given by the max- expensive than for the case that a huge number of
imum legal load represents well the actual traffic of overloaded vehicles circulate in the road.
classes F2 and H5 while for class F1 it represents Considering the data obtained for Figure 13,
a value greater than the actual and for class H3 it the required thickness for the pavement was cal-
represents a value lesser than the actual. culated for the different traffic configurations. The
The presence of overloads increases the truck results are presented in Figure 15 and it is possi-
factor that can reach 100% as indicated in Fig- ble to observe that the pavement thickness follows
ure 14 for class H3 while for classes F1 and F2 is a logarithmic variation with the traffic level and
almost 60%. Only for class H5 the increase of the the difference among all traffic configurations are
truck factor due to the overloads is relatively low relatively reduced meaning that the influence of
(30%). However, these values emphasize the effect the overloads in the pavement thickness, consider-
of the overloads in the pavement analysis and ing all traffic and not only one class of vehicles, is

502
relatively low. The influence of the overloads of all fast calculation of the pavement thickness for any
traffic in the pavement thickness is given in Fig- traffic level.
ure 16 and it ranges from 3.5% up to 5% depending In terms of pavement thickness required to sup-
of the traffic levels considered that is a relatively port a given traffic level with the vehicles observed
low value. This is explained by the fact that there in this work one can conclude that the pavement
are a lot of legal vehicles that compensate the effect thickness due to de overloads reach 0.10 m for
of the overloaded vehicles. some vehicles and the pavement cost due to the
overloads can reach 30%. However, on average, the
increase of the pavement cost reach 10% for class
5 CONCLUSIONS H3 when considering a low traffic level.
In terms of truck factor, the analysis carried out
This work uses a database of vehicle weighing meas- in this work allowed to conclude that it increases
urements to study the impact of the overloaded considerably due to the overloads and can reach
vehicles on pavement cost by the calculation of the 100% for class H3.
ESAL for each axle and vehicle and the use of that For the analysis of the influence of the over-
values to calculate the required pavement thickness loads in the pavement thickness, it depends on the
with which the pavement cost is calculated and the traffic level but it ranges between 3.5% and 5%.
effect of the overloads is obtained.
The traffic weighs were obtained by manual
measurements in a national road in Portugal and ACKNOWLEDGMENTS
it is composed by four different type of vehicles:
class F1 composed by vehicles with two single The first author would like to express the support
axles; class F2 composed by vehicles with a sin- of Portuguese national funding agency for science,
gle axle and a tandem axle; class H3 composed by research and technology (FCT) through scholar-
vehicles with a tractor with two single axles and a ship SFRH/BSAB/114415/2016.
trailer with a tandem axle; class H5 composed by
vehicles with a tractor with two single axles and a
trailer with a tridem axle.
Excepted to class F1, the percentage of over- REFERENCES
loaded vehicles, meaning that at least one axle has
Hong, F., Prozzi, J.A. 2006. Comparison of equivalent
load above the legal limit for that axle, is greater single axle loads from empirical and mechanistic-
than 60% and for classes F2 and H3 it reaches 80%. empirical approaches, Transportation Research Board
This happens because the average load of the axles Annual Meeting.
is approximately the maximum legal load limit. LCPC. 1994. Laboratoire Central des Ponts et Chaussées,
The analysis of the traffic was carried out by French design method for flexible pavements.
calculating the ESAL for all observed axles using Pais, J.C., Amorim, S.I.R., Minhoto, M.J.C. 2013. Impact
three different models: the traditional four power of traffic overloads on road pavements performance.
model, the French model and the model developed ASCE Journal of Transportation Engineering. Vol.
by Pais and Pereira (2016) that takes into account 139, N. 9, pp. 873–879.
Pais, J.C., Pereira, P.A.A. 2016. Development of a model
the type of axle, wheel and pavement for the cal- for equivalent axle load factors. 8th International
culation of the ESAL. These models produce dif- Conference on Maintenance and Rehabilitation of
ferent results function of the axle and wheel type. Pavements - MAIREPAV8. Singapore.
Despite the differences among the different mod- Rys, D., Judycki, J., Jaskula, P. 2016. Analysis of effect of
els, the trend is almost identical. overloaded vehicles on fatigue life of flexible pavements
To calculate the required pavement thickness for based on weigh in motion (WIM) data, International
different traffic levels, a model was developed that Journal of Pavement Engineering, 17:8, 716–726.
expresses Shell fatigue laws. This model allowed a

503
Structural design methods
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

A comparison of mechanistic-empirical pavement design methods for

Norwegian conditions

I.E. Trangen & H. Mork

Department of Civil and Transport Engineering, Norwegian University of Science and Technology, Trondheim,
Norway

ABSTRACT: Pavement design has traditionally been based on empirical methods, and the current
Norwegian pavement design method appears as purely empirical to the user. However, for years there have
been ideas to include analytical aspects into the design process. Mechanistic-empirical design is used in
different pavement design software in different parts of the world. In this paper three different mechanis-
tic-empirical design software tools, PMS Objekt, MMOPP and ADtoPave, are considered for Norwegian
conditions. The outputs from these software tools are compared against empirical output from the Nor-
wegian design method. The study is based on a literature review and a case study. The case study is a mix
of real and fictitious data. However, the material data used are already in the software. The empirical
analysis was conducted by using Handbook N200 from the Norwegian Public Roads Administration
(NPRA). The pavements analysed in the case study were the one found empirically, an under-designed
one, an overdesigned one and an optimum pavement from each software tool. The following three ques-
tions are discussed: (1) How is the usability, (2) which results are found for the case study, and (3) are
these software suitable for Norwegian conditions? All the software give different results, but all indicate
if the pavements are within their design criteria. The software have different advantages, limitations and
complexity. Considering Norwegian conditions, PMS Objekt seems to give the best results, taking frost
heave into account.

1 INTRODUCTION This paper will look into three such software

tools; PMS Objekt from Sweden, MMOPP from
There has been a normal practice to use empirical Denmark and ADtoPave from Germany. The
design methods for the pavement design in differ- study is looking at the usability of the software
ent countries in the world, where one of these is tools and how well they might work for Norwe-
Norway. The Norwegian method is based on cal- gian conditions.
culations with input parameters taken from tables
and diagrams in handbooks from the Norwegian
Public Roads Administration (NPRA). This is
2 BACKGROUND
a method that could work fine, but can be the
reason for under- and over-designed pavement 2.1 Definitions
structures, and does not facilitate new methods
and materials or support changes of climate and 2.1.1 Empirical design
traffic load. This is critical looking at economical Empirical models are usually based on road tests
reasons and safety. Therefore, this paper will look and experience. These models depend on the traf-
at mechanistic-empirical design methods that fic, and parameters such as material properties and
might reduce possible incorrect design, and at the measures of structural strength, for example the
same time fit Norwegian conditions. Mechanistic- Strength Index (SI) (Caliendo 2012). The Ameri-
empirical design includes an analysis of pavement can Association of State Highway and Transpor-
responses and field calibrated distress models, and tation Officials (AASHTO) developed the first
is therefore just applicable as software tools. Input formula for empirical design of pavement in the
parameters are usually site conditions, climate US, based on the so-called AASHO Road Tests.
data and traffic conditions. By calculating and This test and formula is the origin of many types
analyse these, and checking the output against of empirical design systems in the world. Equa-
empirical criteria, the pavement will be fit for the tion 1 shows the formula developed (AASHTO
project site by layer and/or material adjustment. 1993).

507
log(w18 ) l g((
log( ) − 0.20 protection, and material properties related to frost
R 0
log ( ΔPSI
S heave and bearing capacity. These parameters are
+ )
+ 2.32 log(M R ) − 8.07 (1) also found in tables in the handbook, as well as the
0.4 + ( SN1094
S +1)5.19 final tables on how to design different pavement
structures (Statens vegvesen 2014).
Empirical design is the method currently used
for pavement design in Norway. The method is
2.1.2 Mechanistic-empirical design
based on the index method, and is composed of
Mechanistic-empirical design is composed of a
simple static parameters (Statens vegvesen 2014).
mechanistic part and an empirical part. The empiri-
The principals for the index methods are shown in
cal part of an M-E design is the calibration of labo-
Figure 1. This design method uses a fixed expres-
ratory distress criteria into field damage models,
sion for traffic loads, where the sum of ESALs
based on observations from a full-scale test section
(10 t) over the design period is calculated. The
or in-service roads. The laboratory distress criteria
traffic calculation is given by equation 2 (Statens
used, are usually based on stress or strain values. The
vegvesen 2014).
mechanistic part of the M-E design method is the
N f × AADT H × 365 × stress and strain analysis and the damage accumula-
tion calculation based on these damage models (Stat-
(1 + 0.01 × p ) − 1
n
(2)
×C E ens vegvesen 2014). The design method is divided
0.01 × p into different levels. The first level is the input param-
eters categorised as traffic, climate, material proper-
In principle, this is the same as the N values ties and road geometry. These input data are taken to
given by equation 6 and equation 9. the next level, where the analysis is performed. The
A load distribution coefficient, ai, is used analysis includes calculation of the responses, the
to express the relative load distribution ability performance, and the damage accumulation. The
for materials in the pavement. This parameter final level is looking at the final process of the pave-
expresses the relative stiffness compared to the ment design. It might include the life cycle analysis
reference material, and the values for the different and strategies around how to select the pavement
materialsare given in Handbook N200 (Statens (NCHRP 2004).
vegvesen 2014). The parameters for bituminous
materials are given by equation 3, and the values
are used in the index formula shown in equation 4.
3 ANALYSIS TOOLS
ai = 0.21 × 3 E i (3)
3.1 ADtoPave
SI = ∑(a h) (4) ADtoPave (Analysing and Design Tool for Pave-
ments) is a software tool developed in Germany.
Depending on the traffic level, there are require- ADtoPave is currently used by researchers, but
ments connected to some of the summed index will probably take over for the empirical system
values at specific levels of the pavement (BI and SI RStO in the future. ADtoPave is using a linear
in Figure 1). In addition to the index method, the elastic model to describe the materials. Further-
Norwegian empirical method incorporates frost more, it uses the Multi-Layer Theory (MLT) for
the calculation of load responses (Wellner 2009)
by the calculation tool BISAR from Shell (Zeissler
2016). The main focus of the software tool is the
design of new pavements, and it calculates asphalt
fatigue for the design period (Wellner 2009). The
input parameters for ADtoPave are axle load dis-
tribution, master curves for the asphalt materials,
modulus of elasticity for the granular material,
temperature distribution and temperature gradi-
ent in the pavement layers. In addition, the design
period is chosen for the calculation. These data
are analysed by the software tool using Miner’s
I
hypothesis for calculation of the fatigue proc-

~uh~· ·::_-________I ess. A factor, SF, is included in the calculation of

Miner’s hypothesis to transform the result from
laboratory to real-life damages. In addition, safety
factors are used to ensure that there will be no lack
Figure 1. The index method (Statens vegvesen 2014). of bearing capacity in the pavement. A table with

508
these factors is found in “Guidelines for math- N Æ10 P KF KK KR FSSS
(6)
× ∑ FÆ10
ematical design of foundations of traffic surfaces
with a course asphalt surface RDO-Asphalt 09” L
(Wellner 2009). The results of this software tool ( + )n − 1
are given as maximum number of ESAL’s when P= (7)
α
100% fatigue is reached, and the time until the
fatigue process and the structure has reached its AADT H
L = AADT × 365 × × 0.86 (8)
100% capacity. When the results are obtained, one 100
can go back and optimise the pavement to fit the
design period.

3.3 PMS Objekt

3.2 MMOPP
PMS Objekt is a Swedish software tool used to design
MMOPP is an acronym for “Mathematical Meth- new pavements, and an aid for maintenance and reha-
ods of Pavement Performance” and is a Danish bilitation of existing roads. The software tool includes
software tool (Vejregler 2013b). The software tool a database with a set of “standard pavements” that
is based on the Danish manual “Dimensjonering are dependent on the subgrade. The software tool
af Befæstelser og Forsterkningsbelægninger”, and begins with the “standard pavement”, then the user
thissets the criteria for the software tool (Vejregler has to modify the pavement to meet the specific
2013a). The “Vejregelgruppen” called “Design af demands for the selected project (Busch et al. 2010).
befæstelser” developed MMOPP, and the cur- The software tool includes analytical extrapolated
rent software tool and user manual were first models. The models are based on linear elastic theory
published in 2003. For this paper the revised ver- (Saba et al. 2006). PMS Objekt follows criteria and
sion from 2013 and its manual is used (Vejregler advice set in “TRVK Väg—Trafikverkets tekniska
2013b). krav Vägkonstruktion” (Andersson and Winnerholt
MMOPP is divided into three parts: The first 2011a) and “TRVK Väg—Trafikverkets tekniska
is the catalogue based, the second is the analytical- råd Vägkonstruktion” (Andersson and Winnerholt
empirical method and the third is calculation by 2011b). The software tool is also a pavement manage-
simulation. The first part is entirely empirical and ment system that is used for the public road network
follows the Danish handbooks. The software tool in Sweden (Andersson and Winnerholt 2011a).
MMOPP is used for analytical-empirical design PMS Objekt is composed of different models,
and for design by simulation (Vejregler 2013a). For of which some are calculation models. The mod-
this paper, only the analytical-empirical mode of els are listed beneath. In addition, there are more
MMOPP is used. advanced functions in the software that are more
The analytical-empirical design in MMOPP is a specific for in-depth investigations (Winnerholt
traditional analytical analysis. The design criterion and Viktorsson 2015).
is given by formula 5 (Vejregler 2013a). The input
data are traffic volume obtained from the formu- 3.3.1 Climatic model
las 6, 7 and 8, and material data embedded in the In the climatic model in PMS Objekt, Sweden is
software. These parameters are used for stress and divided into five climate zones and six climatic
strain calculation, by the use of the American soft- periods. The climatic periods are divided into win-
ware tool WESDEF (Baltzer 2016). The stresses ter, thaw winter, thaw, late spring, summer and
and strains are calculated for an equivalent 10 t load autumn, for which the temperature, and thereby
called an Æ10 load. An Æ10 is a 10 ton axle with the stiffness of biuminous materials, vary by a cer-
twin wheels, with 20% dynamic addition. The axle tain factor (Evensen 2007).
load is distributed over two circular areas, at each
end of the axle, with a 0,7 MPa contact pressure. 3.3.2 Traffic model and load model
The center-to-center distance between these two The first calculation model is a traffic model,
areas is 350 mm (Vejregler 2013b). The calculated where ESAL is calculated as shown in formula
stresses and strains should be less than the critical 9 (Winnerholt and Viktorsson 2015). The load
stresses and strains, which are shown in formula 5. model gives the parameters for the traffic calcula-
These critical stresses and strains are used as design tion. The design load is given as a 10 t axle load
criteria (Vejregler 2013a). MMOPP will adjust the with twin wheels with contact pressure 800 kPa and
layers in the pavement until the design criteria are c-c 300 mm. (Andersson and Winnerholt 2011a).
fulfilled.
B Nekv AADT k 3.6 A × B just
⎛ E ⎞ ⎛ N Æ10 ⎞
C
n
k j (9)
p A×⎜ ⎟ ×⎜ Æ
⎝ 106 ⎠ ⎟ (5) × ∑ (1 + )
⎝ Erref ⎠ i =1 100

509
B just B × fa f b × fc (10) The model simulates the freezing process, and how
the frost penetrates through the pavement (Busch
et al. 2010). The outputs of this model are frost
3.3.3 Material model heave [mm] and frost depth [mm] (Winnerholt and
Here the stiffness moduli, E/Mr, are set for differ- Viktorsson 2015).
ent pavement layers in relation to the climatic peri-
ods and the climate zones (Evensen 2007).
3.3.4 Damage model 4 FINDINGS AND DISCUSSION
The second calculation model in PMS Objekt deals
with bearing capacity calculations. Bearing capacity This exploratory case study was performed by
is defined as the highest load that can be accepted using qualitative, quantitative and mixed methods
considering the development of cracks and defor- (Yin 2014). The case study was divided into three
mation (Andersson and Winnerholt 2011a). This different experiments; (1) is a comparison between
model is based on calculation of strains. Horizontal horizontal strains at the bottom of the asphalt lay-
tensile strain at the bottom of the asphalt layers and ers calculated by the software tools PMS Objekt
the vertical compressive strain at the top of the sub- and MMOPP, (2) is a comparison of different
grade are calculated for the design load. In addition, results between all the software tools and the
a vertical strain in the subgrade due to a simple load Norwegian empirical method, and (3) is an evalua-
is calculated (Winnerholt and Viktorsson 2015). tion of the usability of the different software tools
Formula 11 and 12 show the criteria for the allowed and the Norwegian empirical method.
number of standard axles for the different types of
strains (Winnerholt and Viktorsson 2015). 4.1 Comparison between horizontal strains for
PMS Objekt and MMOPP
Ntill , bb Nekv (11)
Table 2 shows the result of the calculations made
Ntill , te 2 Nekv (12) in the software tools. The input parameters in
PMS Objekt and MMOPP were the same. The trial
was performed for six different pavements, shown
3.3.5 Frost calculation in Table 1. As shown in Table 2, the values from
The third calculation model in PMS Objekt deals PMS Objekt are around 4–6% higher than the
with the frost depth and the frost heave calculation. results given by MMOPP, even though the inputs

Table 1. Material information.

Pavement Layer Name Thickness [mm] E [MPa]

1 AB 70/100 ABB Mod. GAB 0 70/100 271 2755

A 2 SG I 290 350
3 Frost protection layer 440 100
4 Frost susceptible subgrade – 40
1 AB 70/100 ABB Mod. GAB II 40/60 236 3809
B 2 SG I 310 350
3 Frost protection layer 440 100
4 Frost susceptible subgrade – 40
1 SMA Mod. ABB Mod. GAB II 40/60 231 4051
C 2 SG I 310 350
3 Frost protection layer 450 100
4 Frost susceptible subgrade – 40
1 AB 70/100 GAB 0 70/100 GAB II 40/60 240 3609
D 2 SG I 310 350
3 Frost protection layer 440 100
4 Frost susceptible subgrade – 40
1 SMA Mod. GAB 0 70/100 GAB II 40/60 272 2736
E 2 SG I 290 350
3 Frost protection layer 430 100
4 Frost susceptible subgrade – 40
1 AB 70/100 ABB Mod. GAB 0 70/100 271 2755
F 2 Crushed asphalt (KAS) 270 250
3 Frost protection layer 460 100
4 Frost susceptible subgrade – 40

510
are supposed to be similar. The load models are set 4.2 Comparison of results and optimum
to be the same by using the same pressure in the pavements
wheels and the same c-c distance as in the software
Three pavements were designed by the Norwe-
PMS Objekt. A simplification that 5 t is the same
gian empirical method; one normally designed,
as 50 kN is also done. In addition, the Poisson’s
one over-designed and one under-designed. The
ratio is not given as input to the calculation. Both
data used for the design were subgrade material
of these might be the reason for different results.
T3 (average frost susceptible), climate and traffic
The software tools do not use the same calculation
data from Gardermoen/Skedsmo/Berger in Nor-
program either. PMS Objekt uses BISAR while
way. The climate and traffic data was collected
MMOPP uses WESDEF. Neither of the programs
from the Norwegian Meteorological Institute and
is calculating anything wrong, but there might be
the NPRA, respectively. The three calculated pave-
a variance in the results due to different math-
ments where analysed with the software tools. The
ematical calculation techniques. However, there is
results in Table 3 are given by symbols, where “+”
a difference that seems constant throughout the
means accepted and “−” means rejected. The sym-
experiment. It is difficult to make a conclusion
bols are according to the criterion for the software.
regarding these values, as there is no right or wrong
Looking at the results in Table 3, it is seen that the
answer to what the correct values are.

Table 2. Comparison between horizontal strains at the bottom of the asphalt layers for PMS Objekt and MMOPP.

Critical strain, εh [μs]

Pavement MMOPP PMS Objekt Difference[μs] Difference[%]

A 128.1 121 7.1 5.5

B 126.2 121 5.2 4.1
C 125.2 120 5.2 4.2
D 127.4 122 5.4 4.2
E 127.9 121 6.9 5.4
F 121.7 116 5.7 4.7

Table 3. Evalulation of criteria.

Description N200 MMOPP* PMS Objekt** ADtoPave***

Wearing course: 45 mm
Binder course: 35 mm
Normal empirical method Base course: 140 mm +-++ +++ +
Subbase: 750 mm
FPL: 1350 mm
Total thickness: 232 cm
Wearing course: 30 mm
Binder course: 20 mm
Under-designed Base course: 100 mm --+- --- -
Subbase: 550 mm
FPL: 100 mm
Total thickness: 80 cm
Wearing course: 60 mm
Binder course: 50 mm
Over-designed Base course: 160 mm ++++ +++ +
Subbase: 850 mm
FPL:1500 mm
Total thickness: 626 cm

*Each symbol indicates the layers combined asphalt, subbase, frost protection and subgrade. “+” means that the
stresses and strains are within the criteria, and “−” that they are outside the criteria.
**Each symbol represents the calculation done in the software; the first two indicates whether the pavement is within
the bearing capacity criteria for the bottom of the asphalt layer and the top of the subgrade, the last symbol indicates
if the pavement is within the frost criterion.
***The symbol indicates if the pavement is within the fatigue criterion.

511
normally designed pavement is within the criteria empirically calculated one, is 108 cm. However,
for both PMS Objekt and ADtoPave. However, this MMOPP has a thicker lower asphalt layer that
is not the case for MMOPP, where the second layer might result in a bearing capacity more similar to
is outside the criterion. For the over- and under- the result for the empirically found pavement. The
designed pavements it is seen that both generally load distribution coefficient, see subsection 2.1.1,
follow the empirical models for all three software. is high for asphalt, and by calculating the strength
The over-designed is within the criteria for all the index for all the pavements this conclusion can be
software tools, while the under-designed is not. This drawn. As ADtoPave has limited input on frost,
might show that with major differences in pavement a simple bearing capacity test, without the frost
structures, the software tools tell the same, but when protection layer (FPL), is carried out for both the
there are more optimum solutions there might be empirically calculated pavement and the software
differences. Furthermore, it seems like MMOPP tools. The empirical pavement was calculated to
might have some stricter criteria than the other two. follow the index method, see subsection 2.1.1, while
For the given climatic and traffic conditions at the material parameters for the analytical tools still
Gardermoen, Norway, Table 4 shows the design are as shown by Table 6. The results of this test are
by the Norwegian empirical method and the final shown in Table 5. The differences between the soft-
design by the software tools. The material param- ware tools and the empirical method are smaller
eters used for the analythical analyses are shown in than when the FPL is included. PMS Objekt and
Table 6. The upper asphalt layers were set, while the ADtoPave have relatively similar results, and a
rest were adjusted to fit the criteria for the design conclusion can be that the bearing capacity cal-
period. Looking at the pavements there is a major culation for these models is rather similar. How-
difference between the empirical pavement and the ever, Sweden is a cold country, similar to Norway,
optimum pavement from ADtoPave. PMS Objekt and is probably using softer asphalt materials than
and MMOPP are closer to the empirical design. Germany. This might be the reason for the differ-
The difference in thickness between the thickest ence in layer thicknesses that has occurred in this
pavement calculated by the software tools and the test. MMOPP still has the thickest asphalt layers,

Table 4. Comparison of optimum pavements.

Description N200 MMOPP* PMS Objekt** ADtoPave***

Total thickness 232 cm 100.6 cm 124 cm 72 cm

Wearing course 45 mm 45 mm 45 mm 45 mm
Binder course 35 mm 40 mm 35 mm 35 mm
Base course 140 mm 161 mm 110 mm 90 mm
Subbase 750 mm 320 mm 400 mm 250 mm
FPL 1350 mm 440 mm 650 mm 300 mm
Subgrade T3 Frost susc. 4a Frost susc. E = 40 MPa
Evaluation – ++++ ∼++ +

*Each symbol indicates the layers combined asphalt, subbase, frost protection and subgrade. “+” means that the
stresses and strains are within the criteria, and “−” that they are outside the criteria.
**Each symbol represents the calculation done in the software; the first two indicates whether the pavement is within
the bearing capacity criteria for the bottom of the asphalt layer and the top of the subgrade, the last symbol indicates
if the pavement is within the frost criterion. The symbol, tilde “∼”, means that the pavement is within 10% of the cri-
terion, but still inside the criterion.
***The symbol indicates if the pavement is within the fatigue criterion.

Table 5. Comparison of optimum pavements by looking at the bearing capacity only (no frost protection layer).

Description N200 MMOPP PMS Objekt ADtoPave

Total thickness 107 cm 87,9 cm 70,5 cm 61,5 cm

Wearing course 45 mm 45 mm 45 mm 45 mm
Binder course 35 mm 40 mm 35 mm 35 mm
Base course 140 mm 174 mm 125 mm 85 mm
Subbase 850 mm 620 mm 500 mm 450 mm
Subgrade T3 Frost susc. 4a-Frost susc. E = 40 MPa

512
Table 6. E moduli [MPa] for each layer in the calculation of optimum pavements.

Description MMOPP PMS Objekt ADtoPave

Wear. course 3000 4000

Binder course 3000 4000 –
Base course Tot. asphalt layers: 2923 2500
Subbase 350 450 150
FPL 100 100 120
Subgrade 40 50 40

Table 7. Evaluation of the usability.

N200 MMOPP PMS Objekt ADtoPave

Positive Easy to use with good Calculates the optimum Frost heave calculations A continious fatigue
explanations in the pavement by itself. Easy included. Easy to damage graph for
handbook N200. to understand the understand the the period. Good
software tool, good software tool, good layout. English trans-
layout. layout. Enlish lated manual.
interface. in the
software
Negative Not specialized for each Does not consider forst Climate zones just for Does not consider
project. No analysis of heave. Not possible to Sweeden. The user frost. The user has
the response, pavement customize material data. has to adjust the to adjust the layers
performance and dam- The climate is not taken layers when when performing
age accumulation. May into consideration. Dif- performing optimization.
lead to design errors. ficult to get the project optimization. Complicated
Few design parameters. to be specialized to one Swedeish interface material data.
Interface in Norwegian. place. Danish interface in manual and German interface for
for the software tool, handbooks. the software tool and
the manual and the the handbooks.
handbooks.

which increases the total bearing capacity. Again, MMOPP, as the software tool has a simple setup
the thickness of the pavement shows that MMOPP and a good manual. However, PMS Objekt was
has stricter criteria and will therefore give a pave- easier to understand after learning the differ-
ment structure closest to the empirical one. ent functions, and this software tool also has
some extra applications. ADtoPave is still in the
start-up face, and therefore the software tool is
4.3 Evaluation of the usability of the different
still mainly in German, consequently, it is more
software and the Norwegian empirical method
complicated to use. The material data used in this
Both the software tools and the Norwegian empiri- software tool are also very complex, and certain
cal method have their limitations, advantages and specific laboratory tests are needed to find the
different level of complexity. First, the different master curves. The overall result from the testing
input and output parameters were investigated. shows that PMS Objekt has the most to offer for
In addition, the extra applications of the software Norwegian conditions, as it includes a frost heave
tools were reviewed. From this an evaluation was calculation in addition to the bearing capacity
performed as shown in Table 7. The positive and test. Among the extra applications there is also a
the negative aspects indicate how these methods function looking at rutting due to studded tires.
works and how the usability is. This function is very likely to be used for Norway.
MMOPP (analytical-empirical method) is a good
software tool, but with the lack of adjusting cli-
5 CONCLUSIONS mate data and no focus on frost, it is not well
suited for Norwegian conditions. ADtoPave has
This paper has considered different parts of a lot of potential, but also has a lack in focus on
the software tools PMS Objekt, MMOPP and frost and winter climate. And the software inter-
ADtoPave. The usability of the software tools face is still in German, which makes it hard to
is generally good. The easiest to use might be implement at the moment.

513
ACKNOWLEDGEMENTS Evensen, R. (2007). Vurdering av EDB-system for bereg-
ning av nedbrytning av veg. Statens vegvesen and Veg-
We want to thank the German, Swedish and direktoratet 2520, 36.
Danish experts helping with the research for this NCHRP (2004). Guide for mechanistic-empirical design
of new and rehabilitated pavement structures—final
project, and who made it possible to learn to use report. Transportation Research Board 1–37A.
the software tools in three months. Saba, R., A. Huvstig, G. Hildebrand, E. Sund, R. Even-
sen, H. Sigursteinsson, & J. Elsander (2006). Perform-
ance prediction models for flexible pavements: A
REFERENCES state-of the-art report. Statens vegvesen 2477, 50.
Statens vegvesen (2014). Håndbok N200 Vegbygging.
AASHTO (1993). AASHTO Guide for Design of Pave- Statens vegvesen and Vegdirektoratet.
ment Structures. AASHTO. Vejregler (2013a). Håndbog: Dimensjonering av befæs-
Andersson, P. & T. Winnerholt (2011a, 6). Trvk vä g. telser og forsterkningslægninger brukervejledning—
Anlä ggningsstyring—krav 2011:072, Trafikverket, anlæg og planlægning. Vejdirektoratet.
781 87 Borlä nge. Vejregler (2013b). Håndbog: MMOPP dimensjonering-
Andersson, P. & T. Winnerholt (2011b, 6). Trvk vä g. sprogram for vejbefæstelser brukervejledning—anlæg
Anlä ggningsstyring—råd 2011:073, Trafikverket, 781 og planlægning. Vejdirektoratet.
87 Borlä nge. Wellner (2009). Guidelines for mathematical dimension-
Baltzer, S. (2016). Personal correspondence. ing of foundations of traffic surfaces with a course
Busch, C., M. Holst, & A. Christiansen (2010). Pavement asphalt surface RDO-Asphalt 09 (Drafted.). Dresden:
performance models (NordFoU-PPM)—identifica- Research Society for Roads and Traffic.
tion and selection of pavement performance models. Winnerholt, T. & P. Viktorsson (2015). Manual fr PMS
Nord- FoU—Nordic Cooperation Program 1, 67. Objekt 5.0.1 (draft ed.). Trafikverket.
Caliendo, C. (2012). Local calibration and implementa- Yin, R. (2014). Case Study Research (5th ed.). Sage Pub-
tion of the mechanistic-empirical pavement design lications, Inc.
guide for flexible pavement design. Journal of Trans- Zeissler, A. (2016). Personal correspondence.
portation Engineering 138(3).

514
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Tool for enhancing the 1993 AASHTO pavement design

method to incorporate the dynamic modulus of asphalt mixture

Y.S. Hamdar & G.R. Chehab

American University of Beirut, Lebanon

ABSTRACT: The objective of this research is to enhance the 1993 AASHTO design methodology for
asphalt pavements by integrating the dynamic modulus of asphalt mixtures in the design process. The
aim is to provide a more accurate estimate of the structural coefficient ‘ai’ of the asphalt layer (i), repre-
sentative of the prevalent climatic and traffic conditions, by establishing a new relationship between the
structural coefficient and the effective dynamic modulus (|E*|eff.) of the asphalt mix. The research resulted
in the development of a multivariate relationship between the structural layer coefficient ‘ai’, the effec-
tive dynamic modulus of the asphalt mix (|E*|eff.), and the resilient modulus of the aggregate base layer
(Ebase). A Microsoft-Excel-based pavement structural design tool was subsequently built to support the
developed ai–|E*|eff.–Ebase relationship, and simplify calculations. Design thicknesses acquired using the
tool are generally close to those acquired using AASHTO’s Mechanistic-Empirical Guide through the
Pavement-ME software.

1 INTRODUCTION and financial resources, leaving many agencies

constrained to empirical design methods such as
1.1 Background that of the 1993 AASHTO guide. As a result of
the stated constraints, not all of these states have
The 1993 AASHTO Design Guide for Pavement
put into effect the implementation of the MEPDG.
Structures is currently the most widely used pave-
ment design manual by highway agencies and
consultants around the world, despite its empiri-
1.2 Objective and scope
cal nature and numerous limitations. In the 1993
Design Guide, the structural coefficient ‘ai’ of the The objective of this research is to enhance the
asphalt layer (i) is typically assumed to be 0.44 for 1993 AASHTO design methodology for asphalt
dense graded asphalt mixes, or is acquired from pavements by integrating the dynamic modulus of
a relationship that ties ‘ai’ to the resilient modu- asphalt mixtures in the design process. The aim is to
lus (MR) (AASHTO 1993). Thus, the assumed ‘ai’ provide a more accurate estimate of the structural
value does not reflect the effect of mix type and coefficient ‘a1’ of the asphalt layer, representative
properties, traffic volume and speed, layer thick- of the prevalent climatic and traffic conditions, by
nesses (thin versus thick pavements), climate, and establishing a new relationship between the struc-
unbound layer properties. Moreover, the advances tural coefficient ‘a1’ of the asphalt layer and the
made in the asphalt material technology are not effective dynamic modulus (|E*|eff.) of the asphalt
accounted for when determining the design thick- mix.
nesses of asphalt layers. Note that the dynamic modulus (|E*|) is the
Many US states and countries outside the US ratio of stress to strain of an asphalt mix obtained
have initiated official or research-driven imple- at a range of temperature and loading frequency
mentation and local calibration studies for AASH- combinations. The effective dynamic modulus
TO’s Mechanistic-Empirical Pavement Design |E*|eff. is the |E*| value of the mix that corresponds
Guide (MEPDG) (Olson 2009; Kim & Muthadi to a specific combination of frequency (feff.) and
2007; Fernando et al. 2007; Schwartz 2007; Anon temperature (Teff.) that is prevalent for the road-
2007; Cochran et al. 2009; Pierce 2007; Williams way pavement under design (El-Basyouny & Jeong
et al. 2007; Sadek et al. 2014; Khattab et al. 2014; 2009; El-Basyouny & Jeong 2010; Jeong 2010).
Galal & Chehab 2005a; Nantung et al. 2005; Galal The research scope includes: (1) developing a
& Chehab 2005b). However, adopting the MEPDG multivariate relationship between the structural
has seen its challenges, which include need for layer coefficient ‘ai’, the effective dynamic modu-
extensive data, a relatively high level of expertise, lus of the asphalt mix (|E*|eff.), and the resilient

515
Table 1. Main findings of studies to improve the in Table 2. Those inputs that are not mentioned
estimate of the structural layer coefficient for asphalt were kept constant at their default values in Pave-
layers. ment-ME. Thirty-three asphalt mixes belonging to
five main mix type categories were considered: (1)
Reference Technology Main findings(s)
conventional Hot Mix Asphalt (HMA), (2) HMA
(Qi et al.
Polymer- Layer coefficients of polyethyl- with Polymer Modified Binder (PMB), (3) Warm
1995) modified ene-modified mixtures were Mix Asphalt (WMA), (4) mixes with Reclaimed
asphalt 75–85% higher than layer Asphalt Pavement (RAP), and (5) Stone Matrix
coefficients of unmodified Asphalt (SMA). Figure 1 shows the dynamic mod-
ones ulus mastercurves of the mixes considered in the
(Hossain Crumb- For CRM asphalt mix overlays, scope. Details about the sources and properties of
et al. Rubber the average surface layer coef- these mixes can be found in (Hamdar 2016; Ham-
1997) Modified ficients were found to vary dar & Chehab, in press).
asphalt between 0.11 and 0.46, with
The methodology for establishing the ai–|E*|eff.
(CRM) most values falling around 0.3
For newly constructed CRM relationship is summarized below:
pavements, the layer coef- 1. Running the scenarios summarized in Table 2
ficients varied from 0.25 to on Pavement-ME for the 33 mixes, and acquir-
0.48, with the average around
0.35
ing the design thickness to meet the failure cri-
(Marquis Foamed The layer coefficient of foamed
terion of 15% fatigue cracking,
et al. asphalt asphalt ranged between 0.22 2. Find the effective frequency for each scenario
2003) and 0.35 based on the acquired design thickness,
(Peters- Hot Mix The average recalibrated layer 3. Finding the effective temperature for each sce-
Davis & Asphalt coefficient of HMA was nario based on the acquired design thickness,
Timm (HMA) found to be 0.54, with a 4. Finding the effective reduced frequency for each
2009) standard deviation of 0.08 scenario from the mix’s shift factor equation,
5. Finding the effective |E*| for each scenario from
the |E*| mastercurve of the relevant mix,
modulus of the aggregate base layer (Ebase), and 6. Finding the Structural Number (SN) from
(2) building a Microsoft-Excel-based pavement AASHTO’s 1993 Design Equation and based on
structural design tool to support the developed the design thickness acquired from Pavement-ME,
a1–|E*|eff.–Ebase relationship. 7. Back-calculating the layer coefficient of the
asphalt layer using SN = a1D1 + a2D2, where
a1 = layer coefficient of the asphalt layer;
2 LITERATURE D1 = thickness of the asphalt layer; a2 = layer

Multiple research initiatives have re-evaluated the Table 2. Design input values considered in study scope.
structural coefficient of asphalt layers for conven-
tional as well as new asphalt technologies and mix Input parameters considered
types. The methodology that is commonly adopted Category in scope
to back calculate the structural layer coefficients
relies on measurements of the Falling Weight Climate Chicago, IL (Cold, MAAT* ∼ 11°C)
Deflectometer (FWD). This methodology is based St. Louis, MO (Moderate-Cold,
on recommendations of the 1993 AASHTO guide MAAT∼13.5°C)
as well as NCHRP Project 10–48 (Kim et al. Dallas, TX (Moderate-Hot,
MAAT ∼ 19°C)
2001). Findings of such studies are summarized Phoenix, AZ (Hot, MAAT ∼ 24°C)
in Table 1. These studies are limited to a narrow Traffic volume Low volume (1,500 AADTT**)
set of asphalt mix types, climatic and traffic condi- High volume (15,000 AADTT)
tions. Additionally, the back-calculation algorithm Traffic speed 96 km/hr
assumes the asphalt layer to be undamaged and Aggregate base 200 MPa
elastic as opposed to the more accurate viscoelas- modulus
tic with possibility of existing damage. Aggregate base 30 cm for low traffic volume
layer thickness 38 cm for high traffic volume
Subgrade layer 172 MPa
3 THE DEVELOPED A-E* RELATIONSHIP modulus
Design life 10 years
3.1 Methodology Failure criterion 15% fatigue cracking

The design input values and variables used to * Mean Annual Air Temperature.
develop the ai-|E*|eff. relationship are summarized **Average Annual Daily Truck Traffic.

516
 criteria for interstate pavements (AASHTO
~ 10000 1 (a) 10000
1993; Hamdar 2016).
.., ~ 1000 ~ ~ 1000 • Reliability is assumed to be 90% for both design
~i - M -HMA-1
c
~ methods. Although the concept of reliability
·e 100
M ·HMA-2
100

! - - - HMA-1
0 HMA-Fib
is different in each of the two design guides,
10 +---~~--~--~
l.E-06 l.E -03 1.E+OO l.E +03 1. E+06
10 +-~~~------~
l.E -06 l .E-Q3 l. E+OO l. E+03 l .E+06
change in reliability is assumed to have a similar
Reduce d Fraquancoy at lOC Reducecl Frequency at 20C effect on design thickness (Carvalho & Schwartz
g 2006).
~
(c) (d)
• Conversion of traffic from AADTT to Equiva-
~
10000 100000 v

-;; _
'g g_ 1000
~~
·e 100
~
X WMA-F
i:: 100
0 HMA-RA-1
X HMA-RA-2
lent Single Axle Loads (ESALs) is based on
calculations performed in Pavement-ME, which
assume a SN of 5 and a terminal serviceability
~ 1!1
0 W MA-C
0 HMA-RA-3
10 10 ' of 2.5.
l.E-06 l. E-03 1.E+OO l.E +03 1.E+06
Redu ce d Frequency at 20C

3.3 Results and analysis

~ 10000 (e)

Q
The structural layer coefficient for each of the
j;! - HMA-'1 scenarios (material-climate-traffic combination)
ag 0
D HMA-RAP-1
defined in the scope was determined based on
8 X HMA·RAP-2 X SM-2
X
0 HMA·RAP-3 0 SM-4 the methodology described above. The structural
1.E-06 l .E-03 1.E+OO l.E +03 1.E+06 asphalt layer coefficient was found to be mainly
Reduced Frequ ency at 20C
dependent on the three variables: mix type, cli-
~ t (g) mate (temperature), and traffic speed. A one-way
10000
~0000 j (h) ANOVA was conducted on the structural layer coef-
j ~~ooo x~>s ~ 1000 D ficient values to statistically validate the effect of mix
::;;~ ' X BM-J ::;; D

l
-~

100 ~.
0 0 BM-4
- - - BM-5
• BM-6
-
100
D
-
IJ
SMA·l
SMA· 2
type on the layer coefficient. The analysis revealed a
significant effect of mix type on the average struc-
1 0 +---~~--~--~ 10 +-~~~------~
1 E-06 l .E-03 1.E+00 l.E+03 1.E+06 l .E-06 l .E-Q3 l.E+OO l.E+03 1.E+06 tural layer coefficient at the 95% confidence level.
Reduced Frequ ency at 20C Reduced Frequency at 20C The research methodology culminated in estab-
lishing a linear relationship (Eq. 1) between the
Figure 1. Dynamic modulus mastercurves of mixes structural layer coefficient of the asphalt layer and
included in scope: (a) polymer modified and fiber modi- the effective dynamic modulus |E*|eff. of the asphalt
fied mixes, (b) WMA with PMB, (c) WMA with neat
binder, (d) mixes with recycled aggregate, (e) mixes with
mix.
RAP (f) Virginia DOT surface mixes, (g) Virginia DOT
base mixes, (h) Virginia DOT SMA mixes. i .+ ε
ai = m | E * |eff . int (1)

where ai = structural coefficient of asphalt layer

coefficient of the aggregate base layer; and (i); m = slope, i.e. unit change in a1 for every unit
D2 = thickness of the aggregate base layer, change in |E*|eff.; int. = intercept, function of mix
8. Correlating a1 to |E*|eff. and establish the type; and ε = error.
a1–|E*|eff. relationship. The regression parameters of Equation 1 are
summarized in Table 3. It is evident, from the R2 val-
More details about the methodology and an in-
ues presented in Table 3 that the regression improves
depth analysis of the results can be found in (Ham-
significantly when the mixes are categorized and
dar 2016; Hamdar & Chehab, in press).
regressed by mix type. This further validates the
effect of mix type on ‘ai’. More information on the
3.2 Assumptions regression can be found in (Hamdar 2016).
By taking a closer look at the regression param-
The study is based on an underlying assumption
eters presented in Table 3, it can be inferred that
that the MEPDG methodology is more reliable
in general, ‘ai’ decreases as temperature increases
than that of the 1993 design guide, and generates
due to the decrease in stiffness of the asphalt mix.
design thicknesses that are closer to optimal. Other
Mixes with polymer-modified binder have, on
assumptions include the following:
average, higher layer coefficients than those with
• 15% fatigue cracking is equivalent to a change neat binder.
in Pavement Serviceability Index (PSI) of 1.2. Design input values, such as the failure crite-
This equivalence is obtained by matching the rion, thickness and modulus of the aggregate base
MEPDG recommended performance criteria layer, and traffic volume, were initially assumed for
for fatigue cracking (Anon 2008) to the rec- the back-calculation of the structural layer coef-
ommended 1993 AASHTO serviceability loss ficient, as specified in Table 2. These inputs may

517
Table 3. Summary of regression parameters of linear for every unit change in log (Ebase); int. = intercept,
models. function of mix type; and ε = error.
Based on the mixes considered in the scope, it
Mix Category* Intercept Slope R2
was found that the effect of the base modulus is
All mixes (universal) 0.434 4.35E-6 0.276 fairly independent of mix type, and is represented
All mixes with unmodified 0.422 3.70E-6 0.298 by an average slope of ‘−0.32’. As such, for every
binder unit increase in log (Ebase), i.e., as Ebase increases by
All polymer-modified 0.468 4.77E-6 0.356 10 times, the layer coefficient decreases by 0.32,
mixes given that |E*|eff. is constant.
Unmodified surface mixes 0.405 3.00E-6 0.791
Unmodified base mixes 0.413 3.00E-6 0.791
Unmodified SMA* mixes 0.438 3.00E-6 0.791
4 FEATURES OF THE DESIGN
RAP surface mixes 0.441 3.00E-6 0.791
RAP-base mixes 0.410 3.00E-6 0.791
SUPPORT TOOL
Polymer modified HMA 0.505 3.00E-6 0.791
and WMA mixes This section presents a primer of a design sup-
Polymer modified mixes 0.440 3.00E-6 0.791 port tool that was built to complement the objec-
with RAP tive of the study. The tool, which was developed
using VBA in Microsoft Excel, employs the
*Stone mastic asphalt. a1–|E*|eff.–Ebase relationship (Eq. 2) and the design
methodology of the 1993 Design Guide to pro-
vide a more accurate estimate of the structural
affect the output of the MEPDG (Pavement-ME) layer coefficient of asphalt and resultant design
runs, and in turn, may affect the back-calculated thickness.
structural layer coefficient. Therefore, a sensitivity At the core of the design tool are the various
analysis was conducted to examine the sensitivity a1–|E*|eff.–Ebase relationships that were developed
of the back-calculated ‘ai’, and subsequently the for the various mix categories in the scope of this
ai–|E*|eff. relationship, to each of the input param- study. The computation methodology used in the
eters. It was found that the relationship is sensi- tool is summarized in the flowchart in Figure 2.
tive to the modulus of the aggregate base layer The interface of the tool is shown in Figure 3.
(Ebase). On average, the structural layer coefficient The tool requires the following inputs:
decreases by 0.008 for every increment in modulus • General: The general inputs include design reli-
of the aggregate base layer of 10 MPa. This is rea- ability (%), and the assumed initial and terminal
sonable since the modulus of the aggregate base serviceability.
has a central effect on the pavement’s structural • Climate: The user has the option to select a
number in the 1993 Design Guide, and the stress- pre-defined climatic location, or input custom
strain response of the asphalt layer. climatic data. This data consists of the mean
Here, it is important to distinguish between annual air temperature (°F), the mean monthly
asphalt layer (structural property) and asphalt standard deviation (°F), wind speed (mph), sun-
material. By definition, the layer coefficient is shine (%), and cumulative annual rainfall (in),
a combined structural and material indicator. as shown in Figure 4. These climatic parameters
It not only indicates the integrity of the mate- are required to calculate the effective tempera-
rial, but also its ability to act as a structural ture for fatigue cracking and asphalt rutting.
component in the given pavement. Therefore, it • Traffic: The required inputs are traffic volume
is expected that the structural coefficient be not (equivalent single axle loads, ESALs) and traffic
only dependent on the asphalt material type and speed (mph).
material properties, but also on the layer’s bound- • Unbound layer moduli: The inputs entail the
ary conditions, represented here by the modulus modulus of the aggregate base material and the
of the base layer. subgrade material (psi).
By including the modulus of the aggregate base • Asphalt materials: The basis of asphalt material
layer in the regression, the a1–|E*|eff.–Ebase relation- inputs is the dynamic modulus mastercurve. The
ship is found to be as shown in Equation 2. user may select one of three input levels:
Level 1: The user inputs the sigmoidal and shift
a1 = m | E * |eff . n log ( Ebas
basee ) i t.
int (2) factor coefficients of the dynamic modulus mas-
tercurve (Fig. 5).
where ai = structural coefficient of asphalt layer Level 2: If mastercurve data is not available,
(i); m = slope, i.e. unit change in a1 for every unit the user may select the mix type and Nominal
change in |E*|eff.; n = slope, i.e. unit change in a1 Maximum Aggregate Size (NMAS) (Fig. 6). In

518
,--------------,
I Level1: : General
: Dynamic Modulus Mastercurve· 1 Oes&gn Reliabilfty
Sigmoidal Coefficients 1 Initial and Terminal Se.rviceabflily
1
r--------------~ Standard Deviate
I Level2: I Traffic IData
I Mix Type and NMAS 1, Tra,flic Volume ( ESALs)
I (MIX catalogue) II
r- --------- ----~11\
I
Oes&gn Speed (mph)

I . Level 3: . I 1 Climatic Data

M1x Volumetncs 1 Default station
1 (Witczak Predictive EquaUon) 11 I
1._ _______________ I \
or
Mean Annuall Air Temperature (F)
\ Mean Monthly standard Deviation (F)
\ Cummulative Annual Rainfall! (in)
I Wind Speed (mph) Iterate until new
I %Su nshine aral <0.01

Asphalt Mater·a! Properties

Unbound Material Properties

Modulus of Aggregate Base (psO
Modulus of Subgrade (psO

Find new a 1 from the a 1-1E' eff.l

relationship

Calculate the effective temperature, Calculate the reliability of the design

frequency, and dynamic modulus for thickness based on the confidence
asphalt rutting intervals o1 the ariPefi.l model

Figure 2. Computation methodology adopted in design support tool.

this case, the software selects a mix from the

tool’s database that best represents the user’s
input. The selection process is optimized to
account for the given traffic volume and mean
temperature.
Level 3: The user may input mix volumetrics
(Fig. 7). The software then uses the data to
calculate the effective E* based on the Witczak
predictive equation (Andrei et al. 2005; Bari &
Witczak 2006). The user enters the PG grade
of the binder, from which G* and delta are cal-
culate based on the Witczak database (Andrei
et al. 2005).
Figure 3. Interface of the design support tool.
The tool then generates the design thicknesses
based on the improved structural layer coefficient,
and displays the reliability of the design thickness
based on the reliability of the regressed a1–|E*|–
Ebase relationship (Fig. 8).
The tool has two additional options:
1. The tool generates the effective dynamic mod-
ulus for fatigue and for rutting which may be
used for testing during mix design or during
Quality Assurance/Quality Control (QA/QC).
2. The tool includes a “suggest design” option,
which is based on an algorithm that suggests an
asphalt mix, and subsequent structural design,
Figure 4. Custom climate inputs in the design support based on the climate and traffic level. This
tool. option may be used for benchmarking.

519
Figure 5. Level 1 material inputs in the design support Figure 7. Level 3 material inputs in the design support
tool. tool.

Figure 6. Level 2 material inputs in the design support

Figure 8. Sample output of the design support tool.
tool.

4.1 Limitations design thicknesses that are significantly close to

those acquired using the MEPDG (Pavement-ME),
The tool has two main limitations: (1) the inte-
where the average difference in design thickness is
grated a1–|E*|–Ebase relationship, as well as the
1% with a standard deviation of 3%; whereas, the
mix database, is limited by the number of mixes
1993 design guide tends to consistently over-predict
and mix types included in the scope of the study.
the design thickness compared to the MEPDG,
Therefore, the results of the tool are as reliable as
where the average difference in design thickness is
the developed model, (2) the tool does not accom-
17% with a standard deviation of 11%. It is thus evi-
modate more than one asphalt layer.
dent that the ai–|E*|eff.–Ebase relationship developed in
this research efficiently accounts for material proper-
ties (asphalt mix type and aggregate base modulus),
5 VALIDATION and climatic and traffic conditions. Further valida-
tion against field data is necessary in the future.
To assess the developed ai–|E*|eff.–Ebase relationship,
32 scenarios of different asphalt mix types, climatic
locations, traffic volumes and speeds, and aggre- 6 CONCLUSIONS
gate base moduli were considered. These scenarios
are detailed in (Hamdar 2016; Hamdar & Chehab, The research presented in this paper resulted in
in press). The design thicknesses acquired by the the development of a relationship between the
ai–|E*|eff.–Ebase relationship were compared to those structural layer coefficient of the asphalt laye (ai),
acquired by the 1993 AASHTO design method the effective dynamic modulus of the asphalt mix
(assuming ai = 0.44) as well as those resulting from (|E*|eff.), and the resilient modulus of the aggregate
analysis using the MEPDG (Pavement-ME). The base layer (Ebase). As a result, the empirical nature
results revealed that the developed relationship yields of the asphalt layer coefficient is significantly

520
reduced, and is now correlated to fundamental Complex Modulus of Asphalt Mixtures (NCHRP
material properties (|E*|eff. and Ebase). 1-37 A).
A new ‘ai’ value greater than 0.44 means that Anon, 2008. MEPDG Guide: A Manual of Practice.
designing the pavement using the recommended American Association of State Highway and Transpor-
tation Officials.
value of 0.44, i.e., neglecting mix type and climate/ Anon, 2007. VDOT Preparation Plan for the Implemen-
traffic, yields a pavement that is over-designed. tation of the Mechanistic-Empirical Guide for Design
On the other hand, a new ‘ai’ value that is less of New and Rehabilitated Pavement Structures.
than 0.44 means that designing the pavement using Bari, J. & Witczak, M.W., 2006. Development of a new
the recommended value of 0.44 yields a pavement revised version of the Witczak E* predictive model for
that is under-designed. Both these scenarios can hot mix asphalt mixtures (with discussion). Journal of
hold serious economic and environmental implica- the Association of Asphalt Paving Technologists, 75.
tions, depending on the magnitude of the differ- Carvalho, R. & Schwartz, C., 2006. Comparisons of flex-
ence between the new ‘ai’ value and 0.44. ible pavement designs: AASHTO empirical versus
NCHRP project 1-37 A mechanistic-empirical. Trans-
The findings of this research present a signifi- portation Research Record: Journal of the Transporta-
cant improvement in the area of empirical pavement tion Research Board, (1947), pp.167–174.
design. Acquiring the structural coefficient from Cochran, G. et al., 2009. Implementation of the MEPDG
the developed relationship is proven to yield design for New and Rehabilitated Pavement Structures for
thicknesses that are generally close to those acquired Design of Concrete and Asphalt Pavements in Minne-
using the MEPDG (Pavement-ME). A Microsoft- sota. Report MN/RC 2009-06. Minnesota Department
Excel-based pavement structural design tool was of Transportation, St. Paul, MN.
also built to support the developed relationship, El-Basyouny, M. & Jeong, M., 2009. Effective tem-
and The findings of this research present a signifi- perature for analysis of permanent deformation and
fatigue distress on asphalt mixtures. Transporta-
cant improvement in the area of empirical pavement tion Research Record: Journal of the Transportation
design. Acquiring the structural coefficient from Research Board, (2127), pp.155–163.
the developed relationship is proven to yield design El-Basyouny, M. & Jeong, M.G., 2010. Probabilistic Per-
thicknesses that are generally close to those acquired formance-Related Specifications Methodology Based
using the MEPDG. A Microsoft-Excel-based pave- on Mechanistic-Empirical Pavement Design Guide.
ment structural design tool was also built to support Transportation Research Record: Journal of the Trans-
the developed relationship, and simplify calcula- portation Research Board, 2151(-1), pp. 93–102.
tions. pavement design. This in turn leads to a more Fernando, F.G., Oh, J. & Ryu, D., 2007. Phase I of
accurate design thickness of the asphalt layer. A new MEPDG Program Implementation in Florida. Report
D004491/PR15281-1. Texas Transportation Institute,
‘ai’ value greater than simplify calculations. College Station, Tx.
Future work includes expanding the scope of Galal, K. & Chehab, G., 2005a. Implementing the mecha-
the research to include a wider array of mix types nistic-empirical design guide procedure for a Hot-Mix
and structural design scenarios, and continuously Asphalt-rehabilitated pavement in Indiana. Transpor-
enhancing the a1–|E*|–Ebase relationship. Future tation Research Record: Journal of the Transportation
work also entails improving the design support Research Board, (1919), pp.121–133.
tool to accommodate multiple asphalt layers, and Galal, K. & Chehab, G.R., 2005b. Considerations for
validating the design thicknesses acquired from the implementing the 2002 ME design procedure using a
tool across measured performance data. HMA rehabilitated pavement section in Indiana. In
Presented and published at the Transportation Research
Board (TRB) 84th Annual Meeting, Washington, DC.
Hamdar, Y., 2016. Effective Incorporation of Asphalt
ACKNOWLEDGMENTS Mixture Properties in the Structural Design of Asphalt
Pavements as a Precursor for Implementing Perform-
The authors acknowledge the University Research ance-Based Design. American University of Beirut.
Board at the American University of Beirut for Hamdar, Y. & Chehab, G.R., Integrating the Dynamic
funding this study. Thanks are due to Mr. Jad Modulus of Asphalt Mixes in the 1993 AASHTO
Khalil for his help with the Pavement-ME runs, Design Method. Transportation Research Record,
and Mr. Hussein Kassem, Miss Dima Hassanieh Journal of the Transportation Research Board, in print.
and Mr. Bryan Smith for data support. Hossain, M., Habib, A. & Latorella, T., 1997. Structural
layer coefficients of crumb rubber-modified asphalt
concrete mixtures. Transportation Research Record:
REFERENCES Journal of the Transportation Research Board, (1583),
pp. 62–70.
AASHTO, 1993. AASHTO Guide for Design of Pave- Jeong, M.G., 2010. Manual of practice HMA quality
ment Structures, 1993, American Association of State assurance spreadsheet program using measured values
Highway and Transportation Officials. of E* and d.
Andrei, D., Witczak, M.W. & Mirza, M.W., 2005. Devel- Khattab, A.M. et al., 2014. Evaluation of Witczak
opment of Revised Predictive Model for the Dynamic E* Predictive Models for the Implementation of

521
 AASHTOWare-Pavement ME Design in the King- Pierce, L., 2007. New Pavement Design Guide. Wash-
dom of Saudi Arabia. Construction and Building ington State Department of Transportation, Olympia,
Materials, 64, pp. 360–369. WA.
Kim, R. & Muthadi, N., 2007. Implementation Plan for Available at: http:/resources/07Fall_NewPavemtDesign-
the New Mechanistic Empirical Pavement Design Guide.pdf.
Guide. NCDOT Project 2005-28. North Carolina State Qi, X., Sebaaly, P.E. & Epps, J.A., 1995. Evaluation of
University, Raleigh, NC. polymer-modified asphalt concrete mixtures. Journal
Kim, Y.R. et al., 2001. Assessing Pavement Layer Condi- of materials in civil engineering, 7(2), pp.117–124.
tion Using Deflection Data, Washington, DC. Sadek, H.A. et al., 2014. Implementation of Mechanis-
Marquis, B. et al., 2003. Determination of structural tic-Empirical Pavement Analysis in the State of Qatar.
layer coefficient for roadway recycling using foamed International Journal of Pavement Engineering, 15(6),
asphalt. Durham, NH: Recycled Materials Resource pp. 495–511.
Center. Schwartz, C., 2007. Implementation of the NCHRP
Nantung, T. et al., 2005. Implementation initiatives of the 1-37 A Design Guide. Report No. SP0077B41. Mary-
mechanistic-empirical pavement design guides in Indi- land State Highway Administration, Lutherville, MD.
ana. Transportation Research Record: Journal of the Williams, C.R. et al., 2007. Testing of Wisconsin Asphalt
Transportation Research Board, (1919), pp.142–151. Mixtures for the Forthcoming AASHTO Mechanis-
Olson, R., 2009. Mechanistic-Empirical Pavement Design tic-Empirical Pavement Design Procedure. Report No.
Guide in Indiana. ACEC Indiana Short List, 5(5). WHRP 07-06. Wisconsin Department of Transporta-
Peters-Davis, K. & Timm, D.H., 2009. Recalibration of tion, Madison, WI.
the asphalt layer coefficient,

522
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Determination of AASHTO 1993 layer coefficients considering

time- and temperature-dependency of the asphalt mixture

M. Lanotte & M.E. Kutay

Department of Civil and Environmental Engineering, Michigan State University, East Lansing, MI, USA

ABSTRACT: Empirical pavement design methods resulting from the AASHO Road Tests are still the
most adopted around the word. Since their development, asphalt properties have greatly improved and
advances in the theories of mechanics enhanced the ability to predict the response of pavement materials.
However, layer coefficients for asphalt mixtures have not been updated to accommodate these advances
and pavement structures are currently designed using inaccurate values. The objective of the study pre-
sented in this paper was to present a four-step procedure for determination of structural layer coefficients
for asphalt mixtures, considering their viscoelasticity and temperature-dependency. Layer coefficients
were calculated for thirty-eight (38) HMAs and most of the values were above the single layer coefficient
suggested in the original design guide. Moreover, the variability of the new values indicates the ability
of the procedure to discern the behavior of different materials. Five HMAs were randomly selected for
a simple design exercise of two structures generally used in Michigan for low and high traffic volume
roads. The original AASHTO 1993 layer coefficients were initially adopted, then the two structures were
re-designed using the new structural coefficients and the AASHTOW are Pavement ME design software
(Level 1 analysis). Results indicate that the thinner pavement structures obtained using the new structural
coefficients are still overdesigned if compared with those obtained with the latest and more accurate Pave-
ment ME design methodology.

Keywords: layer coefficient, pavement design, AASHTO method, flexible pavement, empirical design
methods

1 INTRODUCTION 1-37 A (ARA, Inc. 2004) has been relatively slow, it

is reasonable to think that these empirical methods
Governmental agencies and departments of trans- will be used for many years to come.
portation around the world currently use differ- All versions of the AASHTO 1993 procedure
ent approaches to design flexible pavements. Even are based on the experimental program developed
though advances in the theories of mechanics in Illinois in the late 1950’s and early 1960’s by the
have actually improved the ability to predict the American Association of State Highway Officials
response of pavement materials and structures (AASHO) (AASHTO 1961). Pavement distresses
under different environmental and loading con- and surface roughness were periodically measured
ditions, the empirical methods are still by far the to calculate until the pavement structures reached
most popular. the end of the service life. Data from this experi-
A survey conducted in 2013 and published ment was used to develop the currently known
in 2014 by the National Cooperative Highway AASHTO 1993 guide. The design procedure
Research Program (NCHRP), indicated that most described in the AASHTO guide links the pave-
of the responding agencies are using the AASHTO ment serviceability to traffic volume, structural
1993 Guide for the Design of Pavement Structures capacity and statistical parameters using the fol-
(AASHTO 1993). Other agencies in the US adopt lowing equation:
the AASHTO Guide for Design of Pavement
Structures, with 1998 Supplement (AASHTO
1998) and the AASHTO Interim Guide for Design log10 (W18 ) ZR S0 9.3366 log logg10 ( 1) 0.20
of Pavement Structures (AASHTO 1972). This is ⎛ ΔPSI S ⎞
log10
principally due to the fact that these procedures ⎝ 4.2 − 1.5 ⎠ (1)
have served well for several decades. Since the rate + + 2.32 log10 ( ) − 8.07
1094
of the implementation of the mechanistic-empiri- 0.40 +
cal method developed under the NCHRP Project ( +) 5.19

523
Where W18 is the number of Equivalent Single-Axle using |E*| values form both the Hirsh model and
Load (ESAL), SN is the structural number, MR is the HMA master curve. Two pavement structures
the resilient modulus of the subgrade and ZR and typically used in Michigan for low and high traffic
S0 are the standard normal deviate and the stand- volume have been designed using layer coefficients
ard error in predicting pavement serviceability, suggested in the AASHTO 1993 Guide. Then, using
respectively. From Equation (1) designers obtain obtained traffic volumes, these structures were re-
the required structural capacity (SN) for given designed using the new layer coefficients. For com-
combinations of traffic volume (ESALs), soil sub- parison purposes, structures were also designed with
grade, terminal serviceability index, and road cate- the AASHTOW are Pavement ME design software
gories (e.g., arterial, collector …etc., which defines using the Level 1 analysis procedure.
the ZR and S0). Then, the SN of the final struc-
ture must be equal or greater than the required
SN. The SN is an abstract number that combines 2 PROCEDURE FOR THE
layer thicknesses and the materials’ characteristics DETERMINATION OF ASPHALT
through the so-called layer coefficients (ai), which LAYER COEFFICIENT
were obtained during the original AASHO experi-
mental plan. The AASHTO Guide published in The following four-steps procedure allows the eval-
1993 specifies how they may vary depending on uation of the HMA structural coefficients consid-
material properties (e.g. HMA resilient modulus at ering viscoelasticity and temperature dependency
20°C) and type of layer (e.g. surface or base HMA of the asphalt mixtures. Site-specific data such as
layers). However, only a one type of subgrade and climate and traffic speed are considered for the cal-
one HMA in a single environmental condition culation of the dynamic modulus (|E*|) of the mix,
were considered in the AASHO Road Tests. which is then used for the determination of the
Asphalt properties have advanced since that time layer coefficient. As in the AASHTO 1993 proce-
to support increasing traffic loads and extreme cli- dure, the design process starts with the selection of
matic conditions. More durable materials are pro- layer thicknesses for the asphalt concrete wearing,
duced nowadays using polymer-modified bitumen levelling and base courses, then sub-layers can be
and even the mix design processes changed signifi- created if necessary.
cantly. Moreover, several tests are now available to
characterize the viscoelastic properties of HMAs 2.1 Step 1 – Evaluation of the pavement
and their behavior under loading. Despite this fact, temperature spectrum
asphalt layer coefficients have not been revised to Designers should obtain the maximum and mini-
accommodate these advances and changes and mum monthly average air temperature for the
pavement structures are currently designed using design location for one year. The maximum and
inappropriate layer coefficients. minimum monthly pavement surface temperature
Until now, only few attempts have been made shall be estimated using the following formulations
to calibrate layer coefficients for ‘modern’ asphalt (Huber 1994):
mixtures (Hossain et al. 1997; Peters-Savis &
Timm 2009; Bahia et al. 2000). All these studies Ts ( a ) Ta ((max)) − 0.00618Lat 2 0.2289Lat + 24.4
highlighted a significant difference between the (2)
new and the original coefficients suggested in the
AASHTO 1993 Guide. However, even these new Ts ( ) i ) + 1.7
0 859Ta (min) (3)
coefficients are obtained in a way that make them
applicable only to specific conditions. where: Ts(max) is the maximum pavement surface
The main objective of this paper is to present temperature (°C), Ta(max) the maximum air tem-
a four-step procedure to determine asphalt layer perature (°C), Lat is the latitude of the pavement
coefficients. Starting from a predetermined trial location, Ts(min) the minimum pavement surface
pavement structure, the procedure consists initially temperature (°C) and Ta(max) the minimum air
in the evaluation of the pavement temperature temperature (°C).
spectrum and the calculation of the load frequency The maximum and minimum pavement surface
at different depths from the vehicle speed. Then, temperatures are used to calculate the pavement
using temperature and frequency, the dynamic temperature at the center of each layer for each
modulus (|E*|) of the mixes are selected for each month using the BELLS2 model given in the fol-
layer and lift of the pavement structure. The struc- lowing equation:
tural coefficient is finally determined using the
relationship between modulus and layer coefficient Td 2.78 0.912 IIR { (d ) − }
available in AASHTO 1993 Guide.
In this study, layer coefficients of thirty-eight (38) { i ( hrr18 − 15.5)
−0.428IR + 0.553 ( − day) + 2.63 sin }
HMAs commonly used in Michigan were obtained si ( hrr18 − 13.5)
+00.027 IR sin (4)

524
where Td is the pavement temperature (°C) at the obtained at a reference temperature (Tref) by shift-
depth (d), IR is the surface temperature (°C), d the ing horizontally the |E*| values collected during
depth at which the material temperature should to the test. The amount of shift is different at each
be predicted (mm), 1 – day is the average air tem- temperature and defined by the so-called shift fac-
perature the day before (°C), sin is the sine function tor coefficients a(T). The following second-order
on an 18-hr clock system, with 2π radians equal polynomial equation can be used to develop the
to one 18-hour cycle, hr18 the time of day, in 24-hr relationship between shift factors and the corre-
system, but calculated using an 18-hr Asphalt Con- sponding temperature:
crete (AC) temperature rise- and fall-time cycle.
The sine function varies upon the maximum and
minimum pavement surface temperatures reached
log (aT (T )) = a T 2 − Tr2ef ( ) (
a2 T − Treef ) (7)

during the day. For this reason, IR = Ts(max) and

IR = Ts(min) are input separately in this model which where Tref is the reference temperature (chosen by
is then used twice for each month and each layer the user), T is temperature, a1 and a2 are the poly-
of the pavement structure. The final monthly pave- nomial fit coefficients for the temperature shift
ment temperature at the depth d, is the average of factor.
the two values obtained. During the shifting process, shift factors vary
until a good fit to the |E*| data at all the tempera-
tures is obtained using the following sigmoidal
2.2 Assessment of the load frequency function:
The equivalent loading frequency is calculated
log ( E * ) b1 +
from the average vehicle speed using the following b2
(8)
formulation (Losa & Di Natale 2012): e p(
1 + exp 3 4 log ( fR ))

V −2.65 z + β (T ) where b1, b2, b3, b4 are the sigmoid coefficients and
f = 0.043 e (5)
2a fR is the reduced frequency, which is the product of
frequency (f) and the shift factor coefficient (i.e.,
where: f is the load frequency (Hz), V the vehicle fR = f⋅a(T)).
speed (m/sec), a is the radius of tire pressure (m), The |E*| master curve is by far the best option,
z the depth from surface to the center of the AC however whenever this is not available several alter-
layer (m) and β is a function of the average pave- natives can be considered to estimate |E*| from sim-
ment temperature (°C) defined as: pler material properties and volumetric (Bari 2005;
Christensen, Jr. et al. 2003; Al-Khateeb et al. 2006;
β (T ) 1.255 10 5T 3 − 1.6 ⋅10
1 −3T 2 9.2 10 2 T (6) Andrei et al. 1999). The following Hirsch model is
one of those:

2.3 Step 3 – Determination of dynamic ⎡ ⎛ M ⎞

VMA M VFA ⎞ ⎤
⎛ VMA
E* Pc ⎢ 4, 200, 000 1 − + 3 G*
modulus |E*| m
⎣ ⎝ 100 ⎠ b
⎝ 10, 000 ⎠ ⎥⎦
According to the available data, |E*| values can be +
( )
determined using the |E*| master curve or a predic- ( )+ VMA (9)
tive model. To obtain the |E*| master curve, all test 4, 200, 000 3 G * b ⋅VFA
samples should be prepared in accordance with
the AASHTO PP60 specification and air voids
(20 + 3 G (VFA M ))
V ) (VMA
0.58
of all samples (three replicates per each mixture) *

should be within the range of 7% ± 0.5%. Tests P = b

(10)
650 + (3 G (VFA M ))
V ) (VMA
c 0.58
*
must be conducted at five temperatures (−10, 10, b
21, 37 and 54°C) and six frequencies (25, 10, 5, 1,
0.5 and 0.1 Hz) using the Asphalt Mixture Per- where |E*|m is the dynamic modulus of HMA (psi),
formance Tester (AMPT), in accordance with the |G*|b the dynamic shear modulus of the asphalt
AASHTO T342 specification. It is well known that binder (psi), VMA the voids in mineral aggregate,
the minimum AMPT test temperature is approxi- VFA the voids filled with asphalt. The |G*| fre-
mately 0°C therefore, to get |E*| data at −10°C, it is quency sweep tests should be conducted in accord-
suggested to condition samples overnight at −13°C ance with AASHTO T315 on Rolling Thin Film
using an external environmental conditioning sys- Oven (RTFO) aged residue at the following sug-
tem and then quickly transfer the sample into the gested temperatures: 15, 30, 46, 60 and 76 °C. At
AMPT chamber which was kept at 0°C overnight. each temperature, tests are suggested to be run at
A single dynamic modulus master curve can be 11 frequencies varying between 1 and 100 rad/sec.

525
The second-order polynomial (Equation (7)) and ducted by the Michigan Department of Trans-
the sigmoid function (Equation (8)) described for portation (MDOT). Mixes are designated using
the development of the |E*| master curve may be a two-part nomenclature system: the first number
used for the development of the |G*| master curve corresponds to the layer of the pavement structure
as well. (3 - Base, 4 – Leveling/Surface, and 5 – Surface),
and the second is the designation of the expected
2.4 Step 4 – Assessment of HMA layer design traffic volume in million ESALs. Based on
coefficients this nomenclature, the database has been divided
in two groups: HMAs for low-traffic volume roads
The dynamic moduli (|E*|) computed in the pre- (LTV – 1 to 3 mESALs) and HMAs for high-
vious step can be converted to layer coefficients traffic volume roads (HTV – 10 to 30 mESALs).
using the following formula (AASHTO 1993): General information and properties of the materi-
0.171 ln ( E ) − 1.784 als employed in this study are provided in Table 1
a1 (11) and Table 2. All asphalt binders and mixtures were
tested for the determination of |E*| and |G*| mas-
where |E*| is the HMA dynamic modulus in psi. ter curves as described in the previous paragraph.
A single layer coefficient shall be calculated for These data and the volumetric characteristics of
each month of the year and for each layer/lift of the mixtures have been used as input in the pro-
the pavement structure. Then, the yearly average of cedure for the assessment of the layer coefficients
the coefficients shall be used in the design. and for the subsequent pavement design exercise.
Figure 1 and Table 3 show the trial pavement
structures used for the initial structural analysis
3 MATERIALS AND INPUT FOR
PAVEMENT DESIGN
Table 2. Details and properties of HMAs for HTV.
Asphalt mixtures and binders employed in this
study were sampled from several projects con- MDoT
mix HMA Binder Va Vbe VMA VFA
type ID PG [%] [%] [%] [%]
Table 1. Details and properties of HMAs for HTV.
3E30 2 64-22 7.4 10.4 17.8 58.6
MDoT 3E10 18B 58-22 7.0 10.1 16.0 64.1
mix HMA Binder Va Vbe VMA VFA 200 58-28 7.3 10.2 17.5 58.6
type ID PG [%] [%] [%] [%] 4E30 4 70-28 7.0 10.7 17.7 60.6
203 70-22 6.4 10.6 17.0 62.2
3E3 205 58-28 6.7 9.8 16.6 59.4
4E10 20C 64-28 7.5 10.9 14.8 49.2
26A 58-22 7.1 10.7 17.8 60.3
102 64-22 7.8 10.5 18.3 57.5
62 58-28 6.7 10.8 17.4 61.6
5E30 204 70-22 6.9 11.4 18.0 63.8
4E3 28 64-28 6.7 10.7 17.4 61.8
5E10 20B 64-28 6.4 10.7 17.3 62.7
31A 70-28 7.5 10.9 18.3 59.2
21 64-28 7.4 11.9 19.3 61.7
31B 70-28 7.5 10.3 17.8 57.8
24 A 70-28 6.9 11.7 18.5 62.9
67 64-34 7.2 11.0 18.3 60.5
49 A 70-28 7.0 13.2 20.2 65.3
108 64-22 7.5 10.7 18.2 58.7
103 64-22 7.2 11.7 18.9 61.7
111 70-22 7.2 10.7 17.8 59.8
202 64-22 7.6 11.8 19.3 60.8
4E1 44 58-28 7.0 10.7 17.7 60.6
47 64-28 6.8 10.6 17.4 60.8
80 58-34 7.2 10.8 18.0 60.3
85 64-34 7.5 11.2 18.7 60.1
5E3 29B 64-28 9.1 11.4 20.5 55.6
32A 70-28 6.9 11.4 18.3 62.2
37 58-28 7.3 12.1 19.4 62.4
65 58-34 7.2 11.7 18.9 62.0
68 64-34 7.5 11.4 18.8 60.2
109 64-22 7.6 11.6 19.2 60.5
112 70-22 7.5 11.9 19.4 61.6
5E1 45 58-28 7.5 11.8 19.0 62.2
48 64-28 7.2 11.6 18.8 61.8
86 64-34 6.7 12.5 19.4 64.2
206 64-22 7.8 11.6 19.9 61.1
207 64-22 7.6 11.6 19.2 60.4 Figure 1. Cross-section of LTV and HTV pavement
structures.

526
Table 3. Input for pavement design. 0.64 • Hirsch model IE* I Master curve
0.60
HMA Rel. St. 0.56
Layer ID [%] Dev. ΔPSI 0.52
0.48
0.44 -
LTV Surface 31 A 80 0.45 2.2
0.40
Leveling 26 0.36
HTV Surface 204 97 0.45 1.2 HMA2 HMA HMA HMA4 HMA HMA HMA
Leveling 4 18B 200 203 20C 102
Base 2 BaseE30 Base E10 Intermediate E30 Intermediate E10

0.64
0.60

-~ -1 -~-1~-
0.56
and the main input employed, respectively. Cli-

-~-~~-1 -m
0.52
matic conditions (Lansing, MI), vehicles’ speed 0.48
(35 km/h for the LTV structure and 60 km/h for the 0.44
HTV structure), and mechanical characteristics of 0.40
0.36
the unbound materials and reliability levels were
HMA HMA HMAHMAHMAHMA HMAHMAHMA HMAHMA
kept constant for all the analysis. 4 203 204 20C 102 20B 21 24A 49A 103 202

Surface E30 Surface E10

4 RESULTS AND DISCUSSION

Figure 3. Layer coefficients for HMAs used in HTV
structures.
4.1 Evaluation of layer coefficients
Structural coefficients calculated using the new calculated using the |E*| values from the Hirsch
proposed procedure are presented in Figure 2 and model while the second chart is the one with the
Figure 3. In both figures, the first chart shows the values obtained from the |E*| master curves. The
results obtained when the structural coefficient was dashed line in all graphs indicate, for comparison
purposes, the layer coefficient of 0.44 which is the
sole value used in Michigan for all type of HMAs.
Most of the coefficients are above 0.44 which indi-
cate that the structural capacity of the HMAs is
currently underestimated. Moreover, the variability
of the new values demonstrates the ability of the
procedure to discern the behavior of different mate-
HMAHMAHMAHMAHMAHMAHMAHMAHMAHMAHMAHMA HMA rials under certain loading and climatic conditions.
26 62 205 28 31A 316 67 108 111 44 47 80 85
Furthermore, differences are also noticed
Base E3 Intermediate E3 Intermediate El between coefficients calculated through the Hirsch
model and the |E*| master curve. These variations
0.64
0.60 can be attributable to the limitations of the appli-
0.56 cation of the Hirsch predictive model.
0.52
0.48
0.44 4.2 Pavement design
0.40
0.36 Pavement structures shown in Figure 1 have been
HMAHMAHMAHMAHMAHMAHMAHMAHMAHMAHMAHMAHMA analyzed initially using the AASHTO 1993 method
28 31A 316 67 108 111 296 32A 37 65 68 109 112
with a single layer coefficient of 0.44 for all HMAs
Surface E3
for back-calculation of the admissible (design)
0.64
0.60 traffic volumes. The choice of a unique structural
0.56 coefficient follows the common practice in Michi-
0.52
0.48
gan. The allowable traffic volumes were calculated
0.44 - to be 1.6 and 18 m ESALs for the LTV and HTV
0.40 structures, respectively.
0.36
H~ H~ H~ H~ H~ H~ HMA HMA HMA
The new layer coefficients for all five HMAs
44 47 80 85 45 48 86 206 207 selected for this design exercise are higher than
Surface El 0.44. Therefore, since the design traffic volumes
remain constant, a reduction of the layer thick-
Figure 2. Layer coefficients for HMAs used in LTV nesses was necessary. It was carried out with the
structures. following criteria:

527
(a) 50mm • HTV structure: reduction of the granular unbound
base and HMA base course, then consider the
50mm
removal of the HMA leveling course if the reduc-
120mm tion of the HMA base course exceed 50 mm.
30 mm reduction in thickness
As illustrated in Figure 4 and Figure 5, the
(b)
thickness reduction of the LTV structure is only
50mm in the granular unbound base layer. However,
50mm even a decrease of 30–50 mm of the overall layer
lOOmm thickness results in a significant saving of material.
50 mm reduction in thickness Moreover, this reduction decreases the number of
lifts needed for the construction of the overall layer
(c) 50mm with an obvious economic return. Such effects are
20 mm - 30 mm reduction in thickness even more evident in the case of the thicker pave-
JOOmm ment structure where the application of the layer
50 mm reduction in thickness coefficients calculated through the |E*| master
curves allows to remove the entire HMA leveling
Figure 4. Results of the re-design process of the LTV
course. In both cases, the thinner pavement struc-
structure using new layer coefficients from (a) Hirsch tures obtained using the new structural coefficients
model, (b) |E*| master curve, and (c) Pavement ME are still overdesigned if compared with those
(PavME) design software. obtained with the latest and more accurate Pave-
ment ME design methodology. It is worth noting
(a) 50mm that the AASHTO 1993 procedure produces simi-
No levelin!! course - 50 mm thickness reduction lar thicknesses as compared to the Pavement ME
design software, when the actual (measured) |E*|
values are used to compute the layer coefficients.
150mm

5 CONCLUSIONS
150mm
The empirical AASHTO 1993 pavement design
method is still the most widely used design method
all around the word. However, since its develop-
(b) 50mm ment, HMAs’ layer coefficients have not been
No levelin!! course - 50 mm thickness reduction updated to accommodate advances of the modern
asphalt materials.
140mm
In this paper, a four-steps procedure for the
10 mm thickness reduction
assessment of the AASHTO 1993 asphalt layer
coefficients have been presented. Coefficients are
110mm calculated considering HMAs’ viscoelasticity and
40 mm thickness reduction temperature dependency as well as site-specific
parameters. Values obtained by analyzing 38 HMAs
(c) 50mm are often above the unique layer coefficient adopted
No levelin!! course - 50 mm thickness reduction in Michigan (0.44). Thus, the materials’ structural
capacity is currently underestimate. This finding,
140mm has been proven by performing a simple design exer-
10 mm thickness reduction cise. Pavement structures designed using the new
layer coefficients are thinner than the ones obtained
lOOmm with the common coefficient. Moreover, the design
50 mm thickness reduction carried out with the latest and more accurate Pave-
ment ME design methodology showed that even the
Figure 5. Results of the re-design process of the HTV thinner pavement structures are still overdesigned.
structure using new layer coefficients from (a) Hirsch
model, (b) |E*| master curve, and (c) Pavement ME
(PavME) design software. REFERENCES
• TV structure: initial reduction of the granular AASHTO, 1961. The AASHO Road Test: Report 1, His-
unbound base, then reduction of the HMA lev- tory and Description of the Project. Special Report
eling/base course. 61A., Washington, D.C.

528
AASHTO, 1972. AASHTO Interim Guide for Design of Bari, J., 2005. Development of a New Revised Version
Pavement Structures, American Association of State of the Witczak E* Predictive Models for the Hot Mix
and Highway Transportation Officials. Asphalt Mixtures. Arizona State University.
AASHTO, 1993. AASHTO Guide for Design of Pave- Christensen, Jr., D.W., Pellinen, T.K. & Bonaquist, R.F.,
ment Structures, American Association of State High- 2003. Hirsch Model for Estimating the Modulus
way and Transportation Officials. of Asphalt Concrete. Journal of the Association of
AASHTO, 1998. AASHTO Guide for Design of Pave- Asphalt Paving Technologists, 72, pp. 97–121.
ment Structures with 1998 Supplement 4th ed., Ameri- Hossain, M., Habib, A. & LaTorella, T.M., 1997. Struc-
can Association of State and Highway Transportation tural Layer Coefficients of Crumb Rubber-Modified
Officials. Asphalt. Transportation Research Record, 1583, pp.
Al-Khateeb, G. et al., 2006. A New Simplistic Model 62–70.
for Dynamic Modulus Predictions of Asphalt Paving Huber, G.A., 1994. Weather Database for the Superpave
Mixtures. Journal of the Association of Asphalt Paving Mix Design System, Washington, D.C.
Technologists, 75, pp. 1254–1293. Losa, M. & Di Natale, A., 2012. Evaluation of Repre-
Andrei, D., Witczak, M.W. & Mirza, M.W., 1999. sentative Loading Frequency for Linear Elastic Anal-
Development of a Revised Predictive Model for the ysis of Asphalt Pavements. Transportation Research
Dynamic (Complex) Modulus of Asphalt Mixtures, Record: Journal of the Transportation Research Board,
College Park, MD. pp. 150–161.
ARA, Inc., E.C.D., 2004. Guide for Mechanistic–Empir- Peters-Savis, K. & Timm, D.H., 2009. NCAT Report
ical Design of New and Rehabilitated Pavement 09-03 - Recalibration of Asphalt Layer Coefficient.w
Structures. Final report, NCHRP Project 1-37A,
Washington, D.C.
Bahia, H. et al., 2000. Layer coefficients for new and
reprocessed mixes - Final Report, Madison, WI.

529
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

A mechanistic empirical design concept for low volume flexible

pavement using unbound granular materials with application of
concentration factor in a layered system

P.P. Biswas, M.K. Sahis, G.C. Mandal & D. Majumder

Department of Construction Engineering, Jadavpur University, Kolkata, India

ABSTRACT: In this paper, low volume road pavement has been considered as pavement with only
unbound granular base and sub base layer resting on subgrade soil. Design of pavement has been done
with stress based design approach. The allowable stress on top of the subgrade to limit rutting has been
considered from mechanistic-empirical design approach. Present paper deals with dispersion of stress at
interface from a stiffer to softer layer in a flexible pavement so that the thickness of pavement may reduce
the vertical stress on subgrade to its allowable level. A factor defined as Concentration factor has been
used in this paper to find out vertical stress reduction at interface of granular base and subgrade using
Boussinesq’s approach. Concentration factor for different modulus ratio have been determined analyti-
cally and expressed as a function of modulus ratio, i.e the ratio of elastic modulus of stiffer and softer
layer in a two layered system.

1 INTRODUCTION any role to withstand stress strain or deflection

due to wheel load. In view of these, present paper
India’s road network being one the largest road attempts to find out a design methodology for low
network in the world has a total road length of 4.7 volume rural road sections with unbound materi-
million kilometers with a road density of 0.66 kil- als on virgin subgrade using mechanistic-empirical
ometer per square kilometer. Though in terms of design approach.
gross road length India, ranks second in the glo-
bal order but the riding quality on such road raises
2.1 Design basics
serious question about the durability of road pave-
ment in India. Especially 46% of total road length Pavement with bound or unbound granular materi-
in India is unsurfaced road which requires primary als is characterized as a system of layered materi-
attention for repair and up gradation to make als, stiffness of which reduces from top to bottom.
those functionally adequate and safe. The thickness of pavement primarily depends on the
strength of subgrade as well as unbound granular
base and sub-base placed on the top of subgrade
2 PROBLEM STATEMENT soil. The increase in strength of respective layer
reduces the overall thickness of the pavement and
In this backdrop appropriate design methodology vice versa. Therefore, design of pavement essentially
need to be developed to make in service and new prescribes to limit the stress, strain or deflection in
road pavements more durable. The large volume each layer of the pavement system which occurs
of unsurfaced road in India is primarily the rural due to wheel load application. In solution of two
roads with comparatively low volume traffic. Such layered system, the materials in respective layers are
pavement generally can be characterized as two assumed to be homogeneous isotropic and elastic.
layer system with unbound granular base or sub The top layer is assumed to be infinite in lateral
base course as top layer resting on subgrade soil. direction but with a finite depth, whereas the under-
If the stresses in the subgrade, the half space, due lying subgrade layer is infinite in both horizontal
to the wheel load are too high, a stiff top is needed and vertical direction. The boundary and continuity
to reduce these stresses. Such a system with a stiffer condition ensure the continuous contact of two lay-
layer on top of a softer half space, is characterized ers and the surface layer is free of shearing and nor-
a two layer system. Thin bituminous surfacing are mal stress outside the loaded area. Stress at interface
often provided on the top such road pavement to of subgrade and pavement layer was obtained by
ensure a better riding quality but without having Burmister and Huang (Burmister, D.M., 1958 and

531
Huang, Y.H., 1969b) and found to be dependent on and semi-infinite in nature. The equation (1) may
modulus ratio, the ration of elastic modulus of top be used to calculate vertical stress at z depth in a
and bottom layer in a two layer system. If the verti- system if the contact area of surface load is flex-
cal compressive stress at the top of the bottom layer ible and circular.
is too high, excessive deformation will develop in the
subgrade which may be considered as a major cause
of failure of pavement system. ⎡ z3 ⎤
σ z = q ⎢1 − ⎥ (1)
⎢⎣ (a + z )
Flexible pavement is designed against two 3/ 2
⎥⎦
major failure mode, known as rutting and crack-
ing (IRC:37-2012). Rutting is caused due to exces-
sive vertical compressive stress or strain acting on q = surface stress intensity
the top of subgrade where as cracking occurs due a = equivalent radius of contact of surface load.
to excessive radial tensile strain acting at the bot- z = depth at which stress to be determined σz = ver-
tom of bituminous layer. In present work, major tical stress
mode failure of low volume road pavement with However, all equations, which are derived from
unbound base has been considered as rutting Boussinesq’s approach, are based on such assump-
because no bituminous mix has been considered as tions, which are all ideal and not real. Specially
structural layer in pavement system. the pavement, which is a layered structure, there-
Therefore, design of low volume flexible pave- fore cannot be considered as homogeneous and
ment with unbound granular base can be done therefore needs a modification of Boussinesq’s
with stress based design approach. Vertical com- approach in order to make it applicable in layered
pressive stress on top subgrade has been accepted system. In the present work, attempts have been
as failure criteria in flexible pavement design. The made to modify Boussinesq’s analysis to take in to
allowable stress on top of the subgrade to limit rut- account the elasticity of pavement considering the
ting has been considered as design input from rel- system with two layers. In this paper, a concept has
evant mechanistic-empirical design approach. The been used to apply a new parameter (n), termed as
fatigue criterion in pavement design in present case concentration factor (Yoder, E.J., and M.W. Wite-
has not been considered due to absence of bitumi- zak, 1975) in the following manner for determina-
nous layer in the pavement system. tion of stress under a uniform circular load.

2.2 Theory of stress distribution in layered system ⎡ ⎛ ⎞ ⎤

n
z
of with different elastic modulus σ z = q ⎢1 − ⎜ 2 ⎟ ⎥ (2)
⎢⎣ ⎝ a + z ⎠ ⎥⎦
2
Pavement is essentially a layered system of mass
lying on soil subgrade. The stiffness of such lay-
ered system of mass reduces with respect to the The value of concentration factor (n) varies with
depth in order to withstand high stress produced at the modulus ratio (MR) in a two layered system
pavement surface due to application of wheel load. and has been determined by an analytical approach
Therefore, design of flexible pavement involves based on the research work of Huang (Huang,
determination of thickness and characterization Y.H., 1969b) related to the variation of vertical
of constituent paving layers to meet up the func- stress applied by a circular footing on a two lay-
tional requirement of pavement during its design ered system. Such variations of interface stress with
period. The major variables involved in such design respect to modulus ratio and depth of the pavement
are subgrade strength, intensity and frequency of are shown in Figure 1. It can be observed that verti-
wheel load application, environmental and climatic cal stress decreases significantly with the increase in
factor. modulus ratio. It was proved by Burmister that ver-
In pavement design vertical stress on the top of tical stress at pavement subgrade interface is about
the subgrade is an important factor. The function 68% of surface contact pressure if MR is equal to
of base or sub base in such case is to reduce the 1.0 and it reduces to about 8% of contact pressure
vertical stress on subgrade to withstand every pos- if MR becomes 100. Appreciable reduction of inter-
sible failure. However, the allowable vertical stress face stress due to change in modulus ratio proves
on a subgrade is found to be related to the strength the significance of quality of paving layer.
or the modulus of subgrade. (Huang, Y.H., C. Lin, The resilient modulus of subgrade and granu-
X. Deng and J.G. Rose, 1984b) lar base can be obtained using following empirical
In dealing with stress in ideal masses, the relationship to find out the ratio of such modulus
basic equation was formulated by Boussinesq to determine concentration factor.
(Boussinesq, J., 1885) in which the materials was
assumed perfectly elastic, homogeneous, isotropic MR (Mpa) = 10*CBR for CBR ≤ 5 (3)

532
 0.9
0.8

0.7 +0
I '>?
I~'>'· 0
'b-.......:«-,.,Ll
......
1.
L
~
!-""
- .,.-::

I-"'
.......
- ~

"""'
<f! \ './
0.6 «- 'IL <:. , ~ .......
-; "-L /
0.5 v ~ ..... ~

--
cr v 1/ ~f;:)/ I-'
~ 0.4
L L IL' k': I--" ~
t:)
;/ / -?
0;3 li' / L <;,0
/ / ~
...... ~oo f.-"'

- - -"""
0.2 v lL I" k-" ~;..;-
/•/ v / """' ..... !"'"

0.1
~ ~
,.
I / './
....... f.-
0 I
0 0.4 0.8 1.2 1.6 2.0 2.4
a/h,

Figure 1. Vertical interface stress for two-layered system.

Table 1. Variation of concentration factor with modu-

MR (Mpa) = 17.6 (CBR)0.64 for CBR ≥ 5 (4) lus ratio.

where MR = Resilient modulus of subgrade, Modulus Ratio (MR) Concentration factor (n)
CBR = Ca California Bearing Ratio.
1.0 3.02
It is evident from Figure 1 that variation of ver- 2.5 2.11
tical stress with respect to depth is almost linear 5.0 1.47
from a depth (h1/a) = 1 to (h1/a) = α for different 10.0 1.029
modulus ratio. It is important to note that non- 25.0 0.572
linearity of variation of vertical stress with respect 50.0 0.469
to depth of pavement is distinct in lower values of 100.0 0.240
modulus ratio and becomes almost linear where
the modulus ratio is high.
Therefore, in order to determine the value of Table 2. Pavement thickness for 1 Msa.
concentration factor (n), the vertical stress at depth
(h1/a) = 1 was obtained from Figure 1 for a particu- Pavement thickness ( z/a) for 1Msa load
lar modulus ratio and has been put in equation (2)
CBR Present work IRC: SP-72-2007
considering h1 = z. The concentrating factor n thus
obtained for different modulus ratio are presented 2% 2.41 2.79
in Table 1. 3–4% 2.17 2.46
The correlation equation between concentration 5–6% 1.90 2.14
factor and modulus ratio has been developed using
the data of Table 2 and is as following:
2.3 Theoretical formulation of pavement design
n = 3.3394 (MR)-0.5384 (5) model based on vertical compressive stress
It is relevant to mention that the lowest layer in a
R2 = 0.9871 = Regression Co-efficient.
pavement structure is the softest one, in terms of
It is relevant to mention that equation (2) elastic modulus which needs to be protected from
becomes identical to Boussinesq’s equation for higher stress due to wheel load repetitions on pave-
point load if n is equal to 3.02 is considered for ment surface. Therefore, the vertical stress, which is
homogeneous system with a modulus ratio of 1.0. produced at the top of the subgrade due to wheel
The significance of equation (5) involves determi- load repetitions, must be within the allowable
nation of concentration factor (n) with varying stress in subgrade soil.
modulus ratio with different strengths of subgrade The thickness and modulus of paving layers are
and pavement materials and its application in to be designed in a manner to save the subgrade
stress distribution equations. layer from such failure. In the present work the

533
load applied from tire to pavement is considered In the present work ‘q’, the contact stress
similar to a flexible plate with a radius (a) under an between tyre and pavement = 5.6 kg/cm2.
uniform pressure (q). The vertical stress produced Standard axle load = 8200 kg, with dual tyre at
at top of the subgrade along the centerline of such rear axle.
plate has been considered in this work as a param- The radius of contact between tyre and pavement
eter for examination of pavement failure. The rea- 4100
son of such consideration may be substantiated a π *5.6 152 mm
by the work of Foster et al (Foster, C.R. and R.G. and Modulus of unbound granular base have been
Ahlvin 1954) where it was shown that the vertical considered as 100% CBR.
compressive stress along the center line of flexible
plate is the highest one with respect to radial or So, E1 = 17.6*(100)0.64 = 335.36 Mpa
tangential stress at any given depth of pavement.
Now by solving eqns. (6) and (7), the thickness
The allowable vertical stress in subgrade depends
of pavement (z/a) has been determined for different
on the number of load repetitions and subgrade
subgrade strength (E2) and wheel load repetitions (NS).
property. Based on the shell design-criteria (Claus-
The thickness of pavement in this work has been
sen, A.I. et al. 1977 and Shell, 1985) and ASSHTO
expressed in the form of a non dimensional param-
(AASHTO, 1986, 1993) equation, Huang et al
eter (z/a).
developed the following relationship.

NS = 7.199 × 10−5σz−3.734E23.583 (6)

3 RESULTS AND DISCUSSION
E2 = Elastic modulus of subgrade in (kg/cm ), 2

σz = Vertical compressive stress on the surface of Design of low volume pavement guidelines in
the subgrade in (kg/cm2). India both for gravel and flexible pavements are
performance based, drawing on extensive expe-
The eqn. (2) already stated earlier may also be rience in USA on low volume road design, as
expressed in the following form for ease of analysis. brought out in the AASTHO guide for design of
pavement structure. In this backdrop, design load
⎧ ⎫ range of 1–20 Msa and the range of subgrade
⎪ ⎪ strength from 2–10% CBR have been considered
⎪ 1 ⎪ for determination of pavement thickness in this
σ z = q ⎨1 − ⎬ (7)
⎡ 2 0.5 n
⎤ model for comparative analysis.
⎪ ⎛ a⎞ ⎪
+
⎪ ⎢ ⎝ z⎠ ⎥ ⎪
1 The modulus ratio has been determined as the ratio
⎩ ⎢⎣ ⎥⎦ ⎭ of modulus of elasticity of pavement and subgrade
for different subgrade CBR. Values of modulus ratio

3.5

3
,-..,

~
2.5
_._ 2%CBR -.r 3%CBR

2
-6. 4%CBR -e- 5%CBR

1.5
0 5 10 15 20 25
Ns(Msa)

Figure 2. Thickness vs Design load for different CBR value.

534
thus obtained have been used to find out the values of ysis that pavement thickness obtained from present
concentration factor for different subgrade strengths. study is comparable and marginally less than the
It has been found from present study that the pave- thickness recommended in IRC: SP 72-2007.
ment thickness increases with decrease in subgrade The thickness of pavement obtained in this paper
strength and increase in axle load repetitions, which is has also been compared with Kentucky design curves
most logical in flexible pavement design. for flexible pavements and has been presented in
The thickness of pavement expressed in this Table 3. In order to convert the wheel load of 5000
paper in a non-dimensional form (z/a) obtained by lb to 9000 lb, an equivalent wheel load factor of 16
solving equation (6) and (7) for different axle load was used in the present analysis. Such conversion is
repetitions (NS) with different subgrade strength important as the dual wheel load of 9000 lb is con-
and have been presented in Figure.1 and Figure 2. sidered to generate the effect of 8200 kg axle load,
The thickness of flexible pavement thus obtained which is used as a design load in the present model.
from present analysis has been compared with the Beside this, comparison of thickness of pave-
recommended thickness of pavement as laid down ment obtained in this work has also been made
in IRC: SP 72-2007. Pavement thickness for sub- with the results of Corps of Engineers (HRB, 1956
grade CBR ranging from 2–6% for 1 msa load has and U.S. Army Corps of Engineers, 1961) corre-
been considered for comparative analysis and has sponding to 9000 lb wheel load and has been pre-
been presented in Table 2. It is evident from the anal- sented in Table 3 and Figure 2.

Table 3. Comparison of pavement thickness from present model with Corps of Engineers and Wyoming Design
Chart.

(z/a) values for different CBR

Method NS 2% 3% 4% 5% 6% 7% 8% 9% 10%

Kentucky 20 Msa 3.76 3.20 2.91 2.73 2.57 2.47 2.35 2.28 2.22
Present work 3.71 3.38 3.16 3.00 2.93 2.86 2.81 2.75 2.70
Kentucky 10 Msa 3.55 3.03 2.73 2.56 2.44 2.33 2.23 2.16 2.08
Present work 3.38 3.06 2.88 2.71 2.64 2.58 2.53 2.48 2.43
Kentucky 2 Msa 3.09 2.64 2.38 2.23 2.10 2.01 1.95 1.85 1.78
Present work 2.67 2.42 2.25 2.12 2.06 2.01 1.97 1.92 1.88
Kentucky 1 Msa 2.84 2.44 2.21 2.02 1.90 1.91 1.75 1.67 1.62
Present work 2.41 2.17 2.02 1.90 1.84 1.79 1.75 1.71 1.67

3.1

2.9

2.7

2.5

'""'
~ 2.3
<::!.-

2.1
- -6%CBR --- 7%CBR
1.9
- . -so/oCBR - - 9%CBR
l.7
---*- I O%CBR
1.5
0 10 15 20 25
Ns(Msa)

Figure 3. Thickness vs Design load for different CBR value.

535
Table 4. Comparison of pavement thickness between Kentucky’s design and present analysis.

Agency 3% 4% 5% 6% 7% 8% 9% 10%

Corps of Engineer 2.37 2.05 1.82 1.64 1.50 1.41 1.31 1.27
Present work with a load of 2.17 2.02 1.90 1.84 1.79 1.75 1.71 1.67
1 Msa
Wyoming Design Chart 2.84 2.44 2.17 2.0 1.85 1.74 1.64 1.56

It is evident from comparative study that, with AASHTO, 1993. Guide for Design of Pavement Struc-
higher axle load repetition and with a lower sub- tures. American Association of State Highway and
grade strength, Kentucky result matches closely to Transport Officials, Washington, D.C.
the findings of present work. Similarly, with lesser Boussinesq, J., 1885. “Application des Potentiels a Petude
de l’equilibre et du Mouvement des Solids Elastiques,
load repetitions but with higher subgrade strength, Gauthier−Villars, Paris.
Kentucky results are also comparable with present Burmister, D.M., 1958. Evaluation of Pavement Sys-
findings. Such nature of match indicates that, tems of the WASHO Road Test by Layered Systems
linear elastic failure is a major phenomenon of Method. Bulletin 177, Highway Research Board,
unbound paving materials, where failure in either pp. 26−54.
way is controlled by subgrade strength or the mag- Claussen, A.I. et al. 1977. “Asphalt Pavement Design −
nitude of load repetitions. The Shell Method,” Proceedings, 4th International
Findings of empirical results are found to be Conference on the Structural Design of Asphalt Pave-
close with mechanistic- empirical design, where the ments, Vol. 1, pp. 39−74.
Foster, C.R. & R.G. Ahlvin, 1954, “Stresses and Deflec-
linear elastic behavior of paving materials is consid- tiond induced by a Uniform Circular Load”, Proceed-
ered. Moreover, the comparison of results of Corps ings, Highway Research Board, Vol. 33, pp. 467–470.
of Engineers and Wyoming design chart with the HRB, 1956. Flexible Pavement Design Correlation Study.
present work show, that stress based model appears Highway Research Board Bulletin, 133.
to be quite reasonable for design of low volume Huang, Y.H. et al. 1984a. Asphalt Pavement Design:
roads, with comparatively lesser number axle load Highway versus Railroad. Journal of Transportation
repetitions. The thickness of pavement correspond- Engineering, ASCE, Vol. 110, pp. 276−282.
ing to CBR 4% to 5% shows a reasonable good cor- Huang, Y.H. et al. 1984b. Kentrack: A Computer Pro-
relations obtained by different approaches. gram for Hot−Mix Asphalt and Conventional Bal-
last Railway Trackbeds, Report No. RR-84-1, The
Asphalt Institute, College Park, Maryland.
Huang, Y.H., 1969b. “Influence Charts for Two−Layer
4 CONCLUSION Elastic Foundations,” Journal of the Soil Mechanics
and Foundation Division, ASCE, Vol. 95, No. SM2,
It can be concluded from present study that stress March, pp. 709−713.
based design approach may be applied for determi- IRC: 37-2001. Guidelines for the Design of Flexible Pave-
nation of pavement thickness for low volume flexible ments. Indian Road Congress, New Delhi, 2001.
road pavements considering the pavement as a two IRC: SP: 72-2007 Guidelines for the Design of Flexible
layered system. Inclusion of allowable stress criteria pavements for low volume rural roads, India Road
Congress, New Delhi. 2007.
from mechanistic-empirical approach in the present Kentucky, 1976, “A Flexible Pavement Design and Man-
analysis makes the solution more realistic. Applica- agement System, Bureau of Highways, Department
tion of concentration factor for determination of of Transportation”, CommonWealth of Kentucky,
vertical stress is an useful tool which can be applied 4th International Conference on Structural Design
for solution of a multilayered system too. Pavement of Asphalt Pavement.
thickness obtained using present analytical approach Shell, 1985. Shell Pavement Design Manual—Asphalt
is close to the thickness obtained by other research- Pavements and Overlay for Road Traffic, Shell Inter-
ers in the relevant field. The thickness of pavement national Petroleum, London.
corresponding to CBR 4% to 5% shows a reasonable U.S. Army Corps of Engineers, 1961. Revised Method of
Thickness Design of Flexible Highway Pavements at
good correlations obtained by different approaches. Military Installations, Technical Report No. 3−582,
Waterways Experiment Station.
Wyoming, 1948, Proceedings of 27 th annual meetring of
REFERENCES Highway Research Board, Vol-27, pp. 84–91.
Yoder, E.J., & M.W. Witezak, 1975. Principle of Pavement
AASHTO, 1986. Guide for Design of Pavement Struc- Design, Wiley, New York.
tures. American Association of State Highway and
Transport Officials, Washington, D.C.

536
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Perpetual flexible pavement design life

G. Kollaros, A. Athanasopoulou & A. Kokkalis

Department of Civil Engineering, Democritus University of Thrace, Xanthi, Greece

ABSTRACT: Perpetual asphalt pavements, long-life, deep-strength, and extended-life pavements refer
to structures designed for lives long enough to surpass the need for costly maintenance. The design of long
lasting pavements is intended to confine pavement distresses to the surface layer which can be replaced
or overlaid with a new one relatively easily. The pavement achieves an extended service life by regularly
renewing its surface layer so that the overall pavement structure can remain in place perpetually. Layered
pavements under specific loading conditions and the effect of the variability of these conditions on pave-
ment’s longevity are studied in order to understand the technical behaviour of structures with a high
strength surface layer. Different choices in software input have been tried. Reduction of damage per mil-
lion equivalent single axle loads and increase in expected service-life has been achieved with increase of
moduli values and thicknesses of different layers in the pavement structure.

1 INTRODUCTION For the implementation of long life or perpetual

pavement concept various prerequisites must be
The main reasons for which a pavement needs accomplished. Besides the financing to be war-
maintenance are: the reduced support of the soil ranted the local conditions must be well known;
and aggregate layers leading to loss of load carry- these include the available materials, the subgrade
ing capacity, the inadequate drainage of the base, types crossed by the road and the volumes esti-
subbase and subgrade courses, the improper selec- mated to be carried by the structure. The main goal
tion of materials for the different layers, the lim- is to assure low costs in the life cycle (low delay
ited compaction and the unsuccessful estimation and environmental costs and the maintenance kept
of the loads to be carried by the road structure. as little as possible). All these can be achieved by
Obviously, the abovementioned reasons could be posing conservative design criteria for a life over
counteracted with reasonable measures yielding a 40 years, specifying desired performance of fatigue
long lasting construction. resistant base course, rut resistant layers and wear
A first priority for an engineer is to remove resistant top layer, and by the provision of resur-
water accumulated in the substructure. Since many facings in defined periods.
failures are originating from the bottom, qual- The concept of a long life or perpetual pavement
ity materials should be placed starting from the is to have a rut resistant surface (hot mix asphalt,
subgrade up. Subgrades can be improved through HMA, or Open Graded Fatigue Course, OGFC)
compaction, chemical stabilization or the use of a fatigue-resistant asphalt rich lower layer, and
geosynthetics. On the other hand, high quality sufficient total thickness to eliminate the develop-
aggregates should be specified in order for flexible ment of fatigue cracking (Figure 1). It is possible
or rigid pavements avoid wear due to environmen- for today’s HMA pavements to be designed to last
tal cycles or the traffic loading. practically in perpetuity. Fifteen years ago, Asphalt
The road must be capable of handling the loads Pavement Alliance (APA 2002) defined Perpetual
moving on it; thus it is necessary to consider all Pavements as asphalt pavements designed and
types of vehicles anticipated in the traffic mix. The built to last more than 50 years without requiring
hypothesis usually made is that 18,000 pounds major structural rehabilitation or reconstruction.
yield one unit of damage in each pass. The only need is for their periodic renewal of the
The first road that ever received the Perpetual surface layer responding to distresses confined to
Pavement designation is the New Jersey Turnpike the top of the pavement.
built in 1950 with a length of 240 kilometers and An empirical concept for HMA mixtures is that
a very high average daily traffic (almost 700,000 a level of flexural strain there exists below of which
vehicles). The characterization has been given in fatigue damage does not occur for any number of
September 2001 (Newcomb et al. 2010). load repetitions (practically infinite). This is called

537
 enough to provide the structural integrity needed
to prevent structural rutting, fatigue cracking, and
permanent deformation, while having the durabil-
ity to resist damage from the environment (Willis
& Timm, 2009). Expanded design of perpetual
pavements includes high-modulus pavements, low-
volume and high-volume pavements, and the reha-
bilitation of flexible and rigid pavements (Harvey
2011).
High-modulus pavements offer a method which
makes use of less material reducing the cost of
perpetual pavements. In this design approach, a
Figure 1. Long life flexible pavement design concept. very stiff asphalt mixture is used in base and inter-
mediate layers. High-modulus asphalt mixes are
endurance limit and its calculation is usually based in use in European countries both in heavy duty
on the relationship of strain to load repetitions and structural rehabilitation projects where it is
to failure. Flexible pavements could be more effi- desirable to minimize the impact of grade change,
ciently designed if an endurance limit is defined. yet still ensure pavement longevity. In these pave-
Factors affecting the value of the Fatigue Endur- ments, the base course mix is made with a stiff
ance Limit (FEL) in hot mix asphalt mixtures are binder combined with a relatively high binder con-
the binder content, the aggregate type and grada- tent and low void content. This allows for a thick-
tion, the asphalt binder grade, as well as modifi- ness reduction between 25 and 30 percent in the
ers incorporated in the asphalt binder (Anderson pavement structure (European Asphalt Pavement
et al. 2010). FEL is closely connected to the healing Association 2009).
potential of the binder. Thus, at higher tempera- Based on life cycle costs analysis, the compari-
tures, healing occurs more rapidly and damage is son between perpetual and traditional asphalt
recovered more quickly; consequently, there is an pavement will benefit the decision for investment
increase in strain levels that can be tolerated with in highway construction. The early stage con-
no damage accumulation in the structure (Fallon struction costs of perpetual asphalt pavements
et al. 2016). are far beyond those of traditional asphalt pave-
The strategy followed towards adequately ment structures; however, the maintenance costs
modelled long lasting pavements is different than are lower. Frequent rehabilitation interferes with
that applied to previously established procedures free-stream flow traffic and markedly increases
for pavement design. The Mechanistic-Empirical the user’s costs, even double that of perpetual
(M-E) method uses a rational engineering analy- asphalt structure during a 50 years analysis period
sis of the reaction of the pavement in terms of (Jackson et al. 2012).
stresses, strains, and displacements in the context The present work aims to relate different input
of the pavement’s expected life (Montuschi 2012). values with the anticipated pavement life in a lay-
This iterative approach uses the pavement response ered structure handled by known software. Prob-
in terms of stresses, strains, or deflections in order able damage and life estimates have been derived.
to estimate the allowable number of loads to fail-
ure (Nf) for given loading and material properties.
The actual number of anticipated traffic loads (n) 2 PERROAD SOFTWARE
is divided by Nf to define the degree of damage
(D). The point at which the damage equals one is The design of perpetual pavements for low-volume
considered failure (Ferne 2006). Engineers typi- roads is facilitated by PerRoadXpress. This com-
cally consider pavement failure to occur either at puter program has been derived based on a large
20% fatigue cracking in the wheel-path or when 0.5 number of runs in low to medium volume pave-
inches of rutting appear (Behiry 2012). A new M-E ment design cases in the PerRoad described in
Pavement Design Guide (Timm & Priest 2006) is the following paragraphs. The input consists of:
nowadays implemented; thus, more attention is (1) functional classification of the road (urban
paid on proper material and pavement response or rural collector), (2) two-way Annual Average
characterization. Daily Traffic, (3) the anticipated traffic growth,
All layers of the pavement should acquire good (4) the soil classification and/or soil modulus, (5)
constructability characteristics and be imperme- the aggregate base thickness, and (6) the asphalt
able to avoid stripping damage along with other mixture modulus. PerRoad uses the M-E design
moisture problems (Newcomb & Hansen 2006). philosophy. To estimate stresses and strains within
Furthermore, the pavement structure must be thick a pavement, layered elastic analysis is coupled with

538
Monte Carlo simulation (Montoya Rodriguez the limiting pavement responses, an estimate of the
2015). In order to predict strains detrimental for damage incurred per single axle load, and a pro-
fatigue cracking or structural rutting, the following jected time to when the accumulated damage is equal
are required: to 0.1 (D = 1.0 is considered “failure”). On high vol-
ume pavements, the critical parameter is the percent-
• Seasonal resilient moduli of unbound materi-
age of load repetitions below the limiting strains. It is
als and annual Coefficient of Variation (COV).
generally recommended that the designer strive for a
Important modulus considerations for unbound
value of 90% or more on high volume roads.
materials are the moisture content and the state
A probabilistic analysis is executed. To develop
of moisture (i.e. frozen or unfrozen)
a range of output, values within the moduli and
• Seasonal pavement moduli and COV. Moduli of
thickness variability are randomly selected (Lane
HMA layers can be influenced by factors as the
et al. 2006). This type of analysis presents a risk
temperature, the grade of the base asphalt, the
assessment of the probability that any given
air void content, and the use of polymer modi-
threshold value will not be exceeded, as well as an
fied binders. Polymers, fibres and fibrous materi-
indication of the rate of damage from loads caus-
als are used to increase performance properties,
ing the criteria to be exceeded. It is probable that
decrease maintenance works to minimum level,
the thresholds should not be exceeded more than
and for longer service life cycles of roads.
about 90 to 95% of the time, for a high-volume
• Thickness of bound materials and COV
pavement.
• Thickness of unbound materials.
A check on the quality of a perpetual pavement
• Load spectrum for traffic loads according to the
design for a road carrying a high traffic volume
distribution of axle weights.
is to consider the damage caused by a Million
• Location for pavement response analysis (top,
Equivalent Single Axle Loads (MESAL). When
middle, bottom of the layer).
the damage equals 1.0 in conventional mechanistic
• Magnitude of limiting pavement responses
design, the pavement is considered having reached
• Transfer functions for pavement responses
its design distress level. For a long life pavement,
exceeding the user-specified level for accumulat-
D/MESAL must be held to a very low level (10−2
ing damage.
to 10−3).
For each environmental condition to be taken Fatigue ratios are provided through the division
into account, the length of the season must be of cumulative strains by the fatigue endurance
specified (in weeks). The number of pavement lay- limit of the asphalt base layer. Fatigue ratios better
ers ranging between 2 and 5 is specified by the user. predict bottom-up fatigue cracking compared to
The output of PerRoad consists of an evalua- field measured strains or to field fatigue cracking
tion of the percentage of load repetitions lower than performance (Molenaar 2007, Wang et al. 2016).

Table 1. Ranges of maximum AC thicknesses.

Maximum Asphalt Thicknesses (in.)

Subgrade Base 6-in. Aggregate Base 8-in. Aggregate Base 10-in. Aggregate Base
modulus modulus, Mr
Mr (ksi) (ksi) Average Range Average Range Average Range

5 30 14.0 12.5–15.5 13.8 12.5–15.0 13.3 12.0–143

5 50 13.7 12.0–15.0 13.2 11.5–143 12.3 11.0–133
5 100 13.2 12.0–14.0 12.2 11.0–13.0 11.2 10.0–12.0
5 250 10.2 8.5–12.0 9.0 7.5–10.5 7.7 6.5–9.0
5 500 9.3 8.0–11.0 83 7.0–10.0 7.2 6.0–8.5
10 30 12.2 10.5–14.0 11.7 10.5–13.0 11.0 10.0–12.0
10 50 11.8 10.5–13.0 11.2 10.0–12.0 10.0 9.0–11.0
10 100 11.0 10.0–12.0 10.2 9.0–11.0 9.0 8.0–10.0
10 250 9.7 8.5–11.0 8.7 7.5–10.0 7.3 63–8.5
10 500 8.8 7.5–10.5 7.8 6.5–9.5 6.7 53–8.0
20 30 10.5 9.0–12.5 10.7 9.0–12.5 10.0 8.5–11.0
20 50 10.2 8.5–12.5 10.2 8.5–12.0 9.0 73–10.0
20 100 9.7 8.0–12.0 9.0 7.5–10.5 8.0 63–9.0
20 250 8.8 73–10.5 8.0 7.0–9.5 6.8 6.0–8.0
20 500 8.3 12.5–15.5 7A 6.5–9.5 6.5 5.0–8.0

539
 In Table 1, approximate ranges of maximum affects the performance of a roadway pavement,
design thicknesses for asphalt pavements are a pavement system with four layers has been
shown, as proposed by the National Center of elected as input in PerRoad software version 3.5.
Asphalt Technology (NCAT 2016) for the design The modulus has obtained various values in a
of pavements to meet two basic criteria: (a) the log-normal distribution type and for the thick-
cumulative tensile strain at the bottom of the ness variability a normal distribution type has
asphalt and (b) the vertical compressive strain dis- been assumed.
tribution at the top of the subgrade. The bottom soil layer is practically infinite in
The second criterion calls for microstrain values depth. Poisson’s Ratio values were assigned to the
below 200 (at least in the 50% of the measurements), top ΗΜΑ, intermediate, granular base layers and
while the limiting cumulative strain distribution the subgrade as 0.35, 0.35, 0.40 and 0.45 respec-
and fatigue ratios have to be in a range of 29 to 326 tively. PerRoad poses by default a load spectrum
and between 0.68 and 2.19, respectively. In such a depending on the road type. In the calculations
way, the rather conservative limiting strain value of presented here the hypothesis made is for a rural
70 microstrain or the FEL for the base course as highway.
defined through laboratory testing is replaced by The truck growth factor has been taken as 2%
more refined material characteristics. The predic- per year, instead of 4% proposed by the software,
tion of fatigue cracking based on laboratory deter- due to cost reasons mainly (extended economic cri-
minations is rather cumbersome due to the effects sis period).
of traffic load lateral wander, healing, differences In the trials, performance criteria found to
in geometry and test conditions. be valid in the literature have been used. These
included: (1) on top of the HMA layer vertical
deflection with a threshold value of 20 milli-inch,
3 ESTIMATING RANGES OF MAXIMUM (2) at the bottom of the 2nd layer a horizontal
PAVEMENT THICKNESSES strain (−70 microstrain, tension), and (3) on top
of the subgrade vertical strain (200 microstrain).
Mechanistic-empirical design incorporates the Cracked asphalt concrete has been elected as mate-
properties of the pavement layer materials directly rial of the second layer.
as input (Martínez-Diaz & Pérez 2015). Meth- The damage per million ESAL of the 2nd
ods to determine the properties and a means to and 4th layer versus the surface layer modulus is
understand how they varied with environmental depicted in Figure 2.
conditions are required (Karakas & Ortes 2017). The results are plotted in graphs of modulus
The way the properties vary is crucial to Perpet- against life estimate and modulus versus layer
ual Pavement design because most of the damage thickness versus damage per million ESAL.
will occur when the pavement structure is weak- In Figure 3, life estimates in years are presented
est and the loads are the highest. So, the goal is for the second and fourth layer, when the values of
to minimize this damage. Referring to perpetual surface layer modulus range from 2 × 105 to 6 × 105
pavements, asphalt thicknesses have in most cases pounds per square inch (psi) and those of the 2nd
ranged from 20 to 40 cm depending on the avail-
able materials, the traffic volumes, and site condi-
tions (Sargand et al. 2015, Tran et al. 2016). The 0.14
optimal pavement structure could be obtained by
employing a multilayered system where thickness
and mix design are analogous to the type of dis-
0.1 2 l\
tress intended to be diminished by the layer (Tare-
fder & Bateman 2012). :;!
r:a
0.1 ~ r-------,~ ~

~~ ~
Bottom-up fatigue cracking may propagate to ~
~
the surface of an asphalt pavement, affecting all ~ 0.08
the layers, while the water is allowed to alter the " ~
behaviour of the unbound material layers. This ~
0.06
action of the water leads to accelerated surface
- - - - - - . - 2..tl.a.y u
deterioration, pumping, and rutting. Fatigue -e--- 4ilo.la}"!r
cracking begins due to high repeated strains at 0.04
I
the bottom of an asphalt layer from heavy loads 4
M odulus (10' psi)
(Athanasopoulou & Kollaros 2011).
In an effort to determine the way the thick- Figure 2. Damage/MESAL of the 2nd and 4th layer
ness and the modulus of each individual layer versus 1st layer modulus.

540
 45
years, while the corresponding values of the sec-
40
/ ' ond layer have remained almost intact.

3l
~: / / '
4 CONCLUSIONS
;l~l/ ,----- /
,--/

IIv ---- Pavement performance is more than a function

of design. Traffic, climate, pavement materials,
subgrade parameters, construction, maintenance

20
I levels all contribute to how a pavement will per-
form during its life. A new approach is needed to

ll 1/ recognize that there is a point beyond which addi-

tional thicknesses of HMA offer very little return
"
Surface layer modulus 1
Second layer modulus 2
----...-.....-- 2ndJayer
on investment.
10
- - e - & - 4th Ja~ r Perpetual pavements must have enough struc-
tural integrity and thickness to preclude distresses
Modulus(tOJ psi)
such as fatigue cracking, permanent deforma-
tion, and structural rutting. They must be durable
Figure 3. Life estimate of the 2nd and 4th layer versus
enough to resist damage from traffic, such as abra-
surface layer modulus, and versus second layer modulus.
sion, and the environment. While proper design
for a perpetual pavement is of great importance, it
must also be recognized that design life is a func-
layer are in the range of 50,000 psi to 3 × 105 psi, tion of the design requirements, material charac-
respectively. teristics, construction practices, layer thicknesses,
To properly calibrate the transfer functions, maintenance activities, and the failure criterion.
material properties, detailed traffic information, Future research must be targeted on the vali-
and sufficiently detailed pavement management dation and calibration of functions to be used in
data from a number of sites are required. For the Mechanistic-Empirical design procedures, on the
intermediate layer the transfer function coeffi- formation of suggestions for material characteriza-
cients were selected as k1 = 2.83e-006 and k2 = 3.15; tion based on mechanistic approaches aiming the
for the subgrade k1 = 6.03e-008 and k2 = 3.87, accurate predictions of pavement response, on the
respectively. characterization of pavement performance in reha-
An increase of the 1st layer modulus leads to bilitated structures, as well as on the determination
a reduction of Damage/MESAL for the 2nd and of thresholds for fatigue response in the field to
4th layer, while it causes the expected life of these provide long life pavements. Research has also to
layers to be longer. It must be noted that the reduc- be performed in order to investigate the existence
tion of Damage in the intermediate layer is by far of a fatigue endurance limit for hot mix asphalt
greater than that in the subgrade. Though this ten- mixtures, the effect of HMA mixture characteris-
dency is true for the increased life estimates, the tics on the endurance limit, and the potential for
differences are no as great. the limit’s incorporation in structural design meth-
The modulus of the 2nd layer is the most criti- ods for flexible pavements.
cal for longevity of this layer, as well as for the The PerRoad software is a useful tool for ana-
subgrade. When the thickness of the 2nd layer is lyzing pavement structures under the conditions
kept equal to 8 in and the thickness of the 3rd layer they will likely experience during the service life.
has a value of 6 in, then an increase in the asphalt PerRoad should be used in conjunction with appli-
surface layer thickness leads to a reduced Damage/ cable design standards.
MESAL and to a significant rise of the design life Since long-term support of traffic loads and
of both the second and forth layer. minimization of volume change are crucial to per-
The reduction of the quantity Damage/MESAL formance, guidelines are needed for the assessment
is directly affected by the increase of the modulus of stiffness at the time of construction, required
of the intermediate layer. The estimated life in stiffness for long-term performance, and provi-
both layers examined in this paradigm is increased sions to minimize volume change due to expansive
by appointing a higher value to the modulus of the behaviour or frost heave.
second layer. A thicker second layer led to longer Another approach ensuring the fatigue life
life. When a larger thickness has been attributed would be to design a thickness for a stiff struc-
to the third aggregate layer, the expected life of ture such that the tensile strain at the bottom of
the subgrade has been significantly rose to over 50 the asphalt layers is minimized to the extent that

541
cumulative damage would not occur. This would Montuschi, A. 2012. Flexible pavement design using
allow for a single mix design to be used in the base Mechanistic-Empirical methods: the Californian
and intermediate layers, precluding the need to approach. MSc thesis. University of Bologna, Depart-
switch mix types in the lower pavement structure. ment of Civil, Chemical, Environmental and Material
Engineering.
Montoya Rodriguez, C. 2015. Predicting pavement
performance under traffic loading using genetic
REFERENCES algorithms and artificial neural networks to obtain
resilient modulus values. PhD Dissertation, Ohio State
Asphalt Pavement Alliance (APA). 2002. Perpetual pave- University.
ments: A synthesis. APA 101, Lanham, Maryland. Newcomb, D.E. & Hansen, K.R. 2006. Mix type selection
Athanasopoulou, A. & Kollaros, G. 2011. Pavements can for perpetual pavements. Proceedings International
last longer. 5th International Conference Bituminous Conference on Perpetual Pavements. Ohio University,
Mixtures and Pavements. Thessaloniki, Greece, 1–3 Columbus.
June 2011: 1334–1343. Newcomb, D., Willis, R. & Timm, D. 2010. Designing
Behiry, A.E.A.E.M. 2012. Fatigue and rutting lives in perpetual pavements. Hot Mix Asphalt Technology:
flexible pavement. Ain Shams Engineering Journal. 26–35.
3(4): 367–374. Sargand, S., Khoury, I., Jordan, B., Scheer, M., &
European Asphalt Pavement Association (EAPA). 2009. Cichocki, P. 2015. Implementation and thickness
High modulus asphalt. Publication (99)2-098-05001, optimization of perpetual pavements in Ohio. Report
Brussels. FHWA/OH-2015/17 for Ohio Dept. of Transporta-
Ferne, B. 2006. Long-life pavements—A European study tion, Athens, OH: Ohio Research Institute for Trans-
by ELLPAG, International Journal of Pavement Engi- portation and the Environment, Ohio University.
neering, 7(2): 91–100. Tarefder, R.A. & Bateman, D. 2012. Design of optimal
Harvey, J. 2011. Long-Life Asphalt Pavements. Technol- perpetual pavement structure. Journal of Transporta-
ogy Transfer Program. University of California Pave- tion Engineering. 138(2): 157–175.
ment Research Center. 3(2): 1–12. Timm, D.H. & Priest, A.L. 2006 Material properties of
Jackson, N., Mahoney J. & Puccinelli, J. 2012. Using the 2003 NCAT test track structural study, Report
existing pavement in place and achieving long life. No 06-01, National Center for Asphalt Technology,
SHRP 2 R23, Final Report, TRB, Washington, D.C. Auburn University, Alabama.
Karakas, A.S. & Ortes, F. 2017. Comparative assessment Tran, N., Robbins, M.M., Timm, D.H., Willis J.R. &
of the mechanical properties of asphalt layers under Rodezno, C. 2016. Refined limiting strain criteria
the traffic and environmental conditions. Construction and approximate ranges of maximum thicknesses for
and Building Materials. 31: 278–290. designing long-life asphalt pavements. NCAT Report
Lane, B., Brown, A.W. & Tighe, S. 2006. Perpetual pave- 15-05R. National Asphalt Pavement Association,
ments: the Ontario experiment. International Confer- Auburn Al.
ence on Perpetual Pavements. Columbus Ohio. Wang, Y., Norouzi, A. & Kim, Y.R. 2016. Compari-
Martínez-Diaz, M. & Pérez, I. 2015. Mechanistic-empiri- son of fatigue cracking performance predictions in
cal pavement design guide: features and distinctive ele- asphalt pavements using Pavement ME and LVECD
ments. Journal of Construction, 14(1): 32–40. programs. Transportation Research Record, Journal of
Molenaar, A.A.A. 2007. Prediction of fatigue cracking the Transportation Research Board: 2590: 44–55.
in asphalt pavements. Transportation Research Record, Willis, J.R. & Timm, D. 2009. Field-based strain thresh-
Journal of the Transportation Research Board: 2001: olds for flexible perpetual pavement designs, NCAT
155–162. Report 09-09, Auburn University, Alabama.

542
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Mechanistic-empirical pavement design for unpaved roads:

Development of a damage function for structural rutting

J.-P. Bilodeau, G. Doré & M. Le Vern

Department of Civil Engineering, Laval University, Québec, Canada

P.M. Thiam
FPInnovations, Montréal, Canada

ABSTRACT: The Canadian economy relies heavily on the extraction of natural resources. In that con-
text, low-volume (mostly unpaved) roads play a major role in the socioeconomic development of the
country. Up to now, empirical design approaches have been used for these roads but the current trend in
pavement engineering is to adopt mechanistic-empirical design procedures. Therefore, a research project
was undertaken to develop a mechanistic-empirical design procedure for unpaved roads. A small-scale
heavy vehicle simulator was used on pavement structures built over four different sub grade soils in a
laboratory size test pit. The experimental pavement structures were instrumented to measure stress and
water content throughout the pavement layers, surface rutting, as well as permanent and resilient strain at
the top of the sub grade soils. A multistage load sequence was applied to the pavement structures to deter-
mine the relationship between permanent deformation, the number of load cycles and the characteristic
resilient strain for various stress levels. The results were used to develop a Wöhler curve relating resilient
strain and the required number of loads to reach a rutting failure criterion at the top of the sub grade.
Finally, the proposed empirical transfer function, validated with field measurements, has been coupled
with a multilayer elastic analysis to complete the procedure for unpaved roads. The results from this study
were compared with other available methods to validate its reliability.

1 INTRODUCTION From a structural behavior perspective, unpaved

roads response to wheel loading is classified as
Unpaved roads are of great importance for the flexible (TAC 1997), the stress being gradually
socioeconomic development of vast countries, distributed and attenuated by the stiffer granu-
such as Canada, where the extraction of natural lar layers throughout their thickness down to an
resource and access to remote communities heav- allowable level for the typically softer subgrade soil
ily rely on transportation networks. In Canada, (Doré and Zubeck 2009). Pavement deterioration
the unpaved road network is managed in large part and damage of unpaved roads differ from paved
by the private sector, but a significant part of the roads near the surface as distress like fatigue and
network is public. Unpaved roads are usually char- transversal cracking will not occur on a granular
acterized by a very low traffic flow of fewer than surface. However, potholes, rutting, loose material,
400 vehicles per day (AASHTO 2001), but heavy erosion and wash boarding are among the typi-
traffic may also be significant (up to 2000 vehicles cal distress observed on unpaved roads (Skorseth
per day) for some mining and forest industry net- and Selim 2000). One main advantage of unpaved
works. In some areas, the traffic consist primarily roads is the relatively easy surface maintenance
of heavy vehicles, sometimes heavily loaded as the operations that can allow resetting the pavement
load restrictions may differ on private roads, and surface condition. However, when it comes to rut-
non-standard loads are commonly encountered. ting, just like paved roads, a significant proportion
More over, for economic and practical reasons, the is typically associated with rutting of the softest
use of a paved surface may not be justified. The layer in the pavement system, which is typically
typical layout of an unpaved road vary signifi- the sub grade (Asphalt Institute 1991). This creates
cantly from one agency to another, but typically ruts deep in the pavement structure which are more
consists of a gravel wearing course, granular base expensive to rehabilitate, and the associated deep
and granular sub base (optional) placed above the drainage deficiency increases the deterioration rate
native soil (AASHTO 1993). of the pavement. The rutting mechanism in the

543
subgrade soils is typically associated with the verti- The design of unpaved roads is however, more
cal resilient strain (ε r1 ) occurring at the top of the typically based on experience or empirical methods,
sub grade soil (Figure 1), which is influenced by the such as the one proposed by the American Associ-
resilient modulus (Er), the thickness (h) and Pois- ation of State Highways and Transportation Offi-
son ratio (μ) of the layers in the pavement system. cials (AASHTO 1993), USDA Forest Service (True
The strain amplitude is related to the amount of be and Evans 1994), CBR method (Department of
load repetition prior reaching a failure condition, the army 1990), Tannant and Regensburg (2001),
a higher strain amplitude (low structural capacity the critical strain method (Douglas and Lawrence
of the pavement, higher loads…) being associated 2014), etc. By definition, these methods provide
with a faster failure (lower load repetitions). The adequate designs in the context where they were
failure is typically defined as a specific rut depth, developed, but loose reliability when used in other
with criteria of 25 to 75 mm being proposed in the contexts. As such, Mechanistic-Empirical (ME)
literature. Relationships between elastic strain and methods, which allow combining physics with
the number of load repetitions are referred to as empirical observations, are now the recommended
empirical transfer function, or Wohler curves, and practice. Therefore, this project aims at developing
are one of the key elements in a pavement mech- a ME design method, adapted for unpaved roads
anistic-empirical design procedure. With a trial in Canada. In order to achieve good control of the
pavement structure and known or assumed mate- experimental conditions, the research was under-
rial mechanistic properties, the design engineer taken in the laboratory using a small-scale heavy
can compute the strains occurring in the structure vehicle simulator. Also, results from other studies
when submitted to a specific load configuration. are used to validate the proposed damage function.
Using an empirical transfer function, the designer
can therefore obtain an associated number of
loads prior failure for a specific damage mecha- 2 MATERIALS AND METHODS
nism. Some of these functions were defined in the
past for unpaved roads (Gupta et al. 1994; Shell Accelerated pavement testing was performed on
1978; The yse et al. 2004; TRRL 1987, reported by four pavement structures using a Laboratory-Scale
Paterson 1986; Austroads 2004; Sahoo 2009). Vehicle Simulator (LSVS) at Laval University
(Canada) (Figure 2). As presented in Figure 3, four
sub grade soils were used for testing, one for each
experimental pavement: a Silty sand (SM), a low
' I ,

~- ~···· ··· ···· ····· ·· ····· 1 .,8 .3 m,...... .. ...... ... .... ..~

···*
: Acoeleraoon Constant speed zone Ac:cel'eration '

I·~~:~~.t~~~-+ ······~. ~-~-'?';.~ ······ ··.1 --~~~.:~.~ ···i

: O'o •
~
Wearing course

Granular base

IStage 1 : a:0 =442 kPa, 5x11J4 cycles. water level= •

Subgrade 'Stage 2 : o0 :oS.11 kPa, 5x1Q4 cycles,, water level == •
'Stage 3 : o 0 =56·1 kPa , 5x 10Acycles, water level= •
'Stage 4 : o0 =S61 kPa, Sx10• cycles, water level= .,

Figure 1. Resilient vertical strain at the top of the sub- Figure 2. Schematic illustration of the experimental
grade soil in the unpaved road pavement system. setup and applied test conditions.

544
 Four load stages, each applied for 5 × 104 cycles
for a given surface contact stress σ0, and water level
conditions were applied to the experimental pave-
ments to obtain various levels of resilient strain
in combination to permanent deformation evo-
lution. Each laboratory pavement was built with

--
';;1.
200 mm granular base and a 25 mm surface treat-
ment. The purpose of the surface treatment is to
avoid material displacement during testing due to
wheel wandering (+/− 150 mm). Measurements of
strains were only made when the wheel was above
the strain measurement system. The displacement
of the first 150 mm in the sub grade layer was
measured using a multi depth deflectometer and a
JD OOOl
displacement transducer that captured the relative
100 10 1 0 .1 0.01 0 .0011 movement between the two plates in the sub grade,
Diameter (mm) allowing capturing the permanent deformation
and resilient strain.

3 RESULTS AND DISCUSSION

100 10 0.1 0.01 0.001
Figure 4 and Figure 5 present the typical results
Figure 3. Grain-size distribution of the soils and mate- collected with the experimental setup and the
rials used for the experimental pavements. average characteristic resilient strain for each
experimental conditions. The results obtained
Table 1. Material properties after construction.
show that the permanent deformation response
accumulates with increasing number of load
w DPI CBR** Er cycles for each stress stage. The permanent defor-
mation rate increases with the increase of contact
(%) % MDD (mm/bl.) (%) (MPa) stress, but the increase of the water level appears
to be, by far, the most significant factor influenc-
Base* 5 98.3 3.90 68 167 ing this damage parameter. The change of experi-
ML 10 90 16.08 4 95 mental conditions (contact stress and water table
SM 8 99 3.33 52 172 levels) changes the characteristic resilient strain
CL 26 82 59.75 2.5 53 ε1r occurring under wheel loading (Figure 5),
SP 8.5 99 7.95 28.2 46 with the finer soils (ML and CL) being much
*Average from the four experimental test pavements.
4 4 5 5
**Laboratory measurement. 2.5x10 7.5x10 1.3x10 1.8x10

~ 0.0275
plasticity silt (ML), a low plasticity Clay (CL) and ~ 0.025
a poorly graded Sand (SP). The pavement sam- 0 0.0225
;:
ples were compacted inside a stainless steel tank,
with dimensions of 563 mm × 610 mm × 1829 mm e o.o2
1.. 0.0175
(depth × width × length). The pavement system
~ 0.015
was optimized in order to obtain strains in the sub
grade soil similar to those that would be expected
for the typical cases of unpaved roads. Table 1 sum-
marizes the main material properties, such as the
-
"C

;
r::::
C'l:l
0 .01
0.0125

0.0075
water content by weight (w), the compaction per- eo.oo5
Cl)
centage (%MDD) with respect to the Maximum D.. 0.0025
Dry Density (MDD), the Dynamic Penetration
4 5 5 5
Index (DPI), as well as the estimated resilient mod- 5x1 0 10 2x1 0 2x1 0
ulus (Er) using the equations proposed by Boutet Load cycles
et al. (2010). For reference, California Bearing
ratio (CBR) test results on compacted samples are Figure 4. Multi-stage permanent deformation accumu-
also provided in Table 1. lation—SP subgrade soil.

545
E
0.007 r::========------,
cr 0=442 kPa 0 .0065 5
E 0.006 cr 0=5 11 kPa
4.5
E
.§. 0.005
cr 0 =561 kPa
cr 0 =56 1 kPa(sat.)
0 .0055

0.004 5
e4 ~~-1--........---l--~.....1.-......:.....::~::.......;.....~...
,. w 0.004 ~-~-----1 ·1- - - .§. 3.5
s:::

--
0 .0035
'§ 0.003 g'3
en
s::: 0.002
~
'iii 0.001
0 .0025

0 .001 5
! 2L~~V-~---~·~-----~:------+·------~:-_·-~-~-----~t-----l
CJ . '
2.5

C1)
0::: 0 .00 05 ~ ·- --------------:---- --··t·· ----r-· 1.5
0 ,~ 1 • ---'---1--
ML SM CL SP

Figure 5. Characteristic resilient strain for each load

w ~---- ___ (sR=0.11595fV0·334 0 0.5
stage and subgrade soil. a~~~~~~--L-._~~~
4 4 4 4 4
0 10 2x1 0 3x1 0 4x1 0 5x1 0
Load cycles
more affected by the increase of the water table.
The poorly graded sand presents a resilient strain Figure 6. Example of power model fitting for the deter-
response that is less influenced by the surface con- mination of long-term damage—SP sub grade.
tact stress and water table.
During the tests, the surface rutting was also
measured frequently, with a higher measurement of distress (rut that may develop in the base layer).
frequency in the early phase of each loading stage. The relationship between characteristic resilient
Three transversal rut profiles were measured in the strain and the allowable number of load repeti-
constant speed zone at the experimental pavement tions prior failure corresponds to a damage func-
surface (Figure 1), i.e. one in the centre, and two tion referred to as Wöhler curve, or to an empirical
150 mm on each side of the transversal centre line. transfer function. In pavement engineering, when
Figure 6 presents an example of the average maxi- it comes to the structural rutting failure criteria,
mum rut depth, measured with the three transver- the empirical transfer function can be expressed as
sal rut profiles with respect to the number of load
cycles. For each experimental condition (surface ε1r = aNr
N b (2)
contact stress and water table level), the Surface
Rut (SR) was modelled with respect to the number in which Nr is the allowable number of load rep-
of load cycles using a power model (Sweere 1990) etitions prior reaching a failure criteria, or a given
expressed as rutting value within the subgrade, ε1r is the char-
acteristic vertical resilient strain, and a and b are
SR = AN B (1) fitting coefficients.
Figure 7 presents the model obtained with
in which A and B are fitting coefficients. the experimental data collected throughout this
The power models developed for each of the project. Data introduced by Little (1992) using
four experimental pavements were used to extrapo- field measurements were used to improve the
late the number of load cycles prior to structural structural rutting damage function. The model
rutting failure. This number was estimated for developed for a rut of 25 mm, which has a coef-
an accumulated 25 mm structural rutting with in ficient of determination R2 of 0.742, is defined by
the sub grade. The 25 mm rutting is in the range
of values previously reported in the literature for ε1r = 0.04792 Nr 0.29475
(3)
unpaved roads (Little 1992). This is attributed to
the fact that structural rutting is mostly associated in which Nr is the number of load repetitions, ε1r
with the weakest layer in the unpaved road sys- a = 0.04792 and b = −0.029475. It can be observed
tem, which is typically the subgrade. Therefore, a that a satisfying agreement exists between the data
25 mm rut, developed in the sub grade, may lead collected in this study and the data collected by Little
to significant performance challenges. Moreover, (1992). It should be noted that Little (1992) studied
the maintenance of an unpaved road implies rou- pavements with and without geosynthetics; the data
tine surface grading and resurfacing, which in part points from this study used for the comparison are
contributes to the maintenance and rehabilitation from experimental pavements without geosynthetics.

546
 3 5 7 1a2 1a3 1a4 1a5 1a6 1a7 1as
-1.5 a.1
-1 .75 0.05
-2
-2.25 eE a.a1
0.02
~ -2.5
0
...J -3
-2.75 e.§. a.aa1
0.005
0.002
-3.25 0.0005
-3.5 w'-
0.0002
-3 .75 a.aaa1 2
4 6 8 1a 1a3 1a4 1a5 1a6 1a7 1as
N
Log N
/_ Gupta et al. (2014)-····· TRRL (1987)
o ML + CL o Little (1992) - Model Shell (1978) Theyse et al. (1996)
() SM A SP •••• +!- RMSE - Sahoo (2009) This study
1
······ Austroads (2004)

Figure 7. Proposed models using the data collected in

this study and the ones from Little (1992). Figure 8. Comparison of the proposed model with
some existing empirical transfer functions for structural
rutting.
The proposed model also includes the effect of vari-
ous soil types on the Wöhler curve. One interesting it has become more relevant to adapt these meth-
observation is that, for three of the four tested sub ods to the design of unpaved roads. This is par-
grade soils (ML, CL and SM), the measured data ticularly important considering the importance of
point for the case where the water table was raised the unpaved roads networks for the socioeconomic
are slightly off a linear trend between the three data development of vast countries, such as Canada,
points collected for the low water table conditions. where the extraction of natural resources and access
As a Wöhler curve assumes elastic behaviour of to isolated communities heavily rely on transporta-
materials, the data points collected for high water tion network.
table may be less representative of these conditions For the implementation of an adapted unpaved
(as observed in Figure 5). This is in agreement with roads mechanistic-empirical design method, the
the fact that some experimental pavements with high development of an empirical transfer function
water table conditions reached failure prior to com- for structural rutting is necessary. A small-scale
pleting 50000 load sequence. It is also interesting to laboratory vehicle simulator was used to develop a
note that the sub grade soil that can be considered transfer function (Wöhler curve) for structural rut-
the less sensitive to high saturation levels (SP) does ting. Four experimental test pavements were built
not show this particular trend on the Wöhler curve. in a test pit using four different sub grade soils,
Figure 8 presents a comparison between the and they were submitted to a multistage loading
proposed empirical transfer function for structural procedure with varying experimental conditions.
rutting and other similar equations available in the The data collected allowed proposing a relation-
literature. One can find an agreement between the ship that links the characteristic resilient strain
proposed empirical function and most of the avail- at the top of the sub grade soil and the number
able models. The Wöhler curve developed in this of load repetitions required to reach a 25 mm rut
study is particularly close to the ones proposed associated to the sub grade soils response. This
by Shell (1978) and Sahoo (2009). The proposed relationship was validated using data available in
model is also in the average range of the models the literature. The proposed model adequately rep-
considered for high number of load repetitions resents a general empirical transfer function for all
(>50000). For weaker pavement structures, which is the soil types and experimental conditions consid-
associated with high characteristic resilient strain, ered. A comparison with existing Wöhler curves
the model developed in this study will give the high- for structural rutting allows concluding that the
est allowable number of load repetitions among the model compares well to several other models, espe-
models considered for the comparative analysis. cially for load repetitions higher than 50 000.

4 CONCLUSION REFERENCES
The design of unpaved roads in many countries American Association for State Highway and Transpor-
is mostly experience-based or based on empirical tation Officials 1993. Guide for design of pavement
methods. With the gradual shift of pavement design structures. American Association of State Highway
methods towards mechanistic-empirical approaches, Officials, Washington D.C.

547
American Association for State Highway and Transpor- Paterson, W.D.O., 1987. Road deterioration and mainte-
tation Officials 2001. Guidelines for geometric design nance effects. Washington, DC: The World Bank.
of low volume local roads. American Association of Sahoo, U.C., 2009. Performance evaluation of low vol-
State Highway Officials, Washington D.C. ume roads. Unpublished thesis (PhD). Department
AustRoads, 2004. Technical basis of Austroads pavement of Civil Engineering, Indian Institute of Technology,
design guide. AP-T-33. Sydney: Austroads. Kharagpur.
Boutet, M., Doré, G., Bilodeau, J.-P. and Pierre, P. 2010. Shell, 1978. Shell pavement design manual. London: Shell
Development of Models for the Interpretation of the International Petroleum Company Ltd.
Dynamic Cone Penetrometer Data. International Skorseth, K. and Selim, A. (2000). Gravel Roads Main-
Journal of Pavement Engineering, Vol. 12(3), 201–214. tenance and Design Manual. U.S. Department of
Department of the Army, 1990. Design of Aggregate Transportation.
Surfaced Roads and Airfields. Washington D.C. Sweere, G. 1990. Unbound granular bases for roads. Ph.D.
Doré, G. and Zubeck, H. 2009. Cold regions pavement dissertation, Delf University, Delf, The Netherlands.
engineering. McGraw-Hill, NY. TAC 1997. Pavement design and management guide. Dr.
Douglas, R.A. and Lawrence, K. 2014. Optimizing haul Ralph Haas Ed., Waterloo, ON.
road design—a challenge for resource development in Tannant, D. and Regensburg, B. 2001. Guidelines for
Northern Canada. In the Proceedings of Transporta- Mine Haul Road Design. Research report, 115 p.
tion Association of Canada Montreal, Quebec. Theyse, H.L., De Beer, M., and Rust, F.C., 1996. Over-
Gupta, A., P. Kumar, and R. Rastogi, 2014. Mechanis- view of the South African mechanistic pavement
tic–Empirical Approach for Design of Low Volume design methods, Transportation Research Record No.
Pavements. International Journal of Pavement Engi- 1539, Washington, DC: TRB, 6–17.
neering, DOI: 10.1080/10298436.2014.960999. Truebe, M. and Evans, G. 1994. Lowell Surfacing
Little, P. 1992. The Design of Un surfaced Roads Using Thickness Design Test Road. U.S. Department of
Geosynthetics. PhD dissertation, University of Not- Transportation.
tingham, United Kingdom.

548
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Reliability demystified, at last

A.M. Ioannides
Department of Civil and Architectural Engineering and Construction Management,
University of Cincinnati, Ohio, USA

D.D. Rodriguez
Airfields and Pavements Branch, U.S. Army Engineer Research and Development Center (ERDC),
Vicksburg, Mississippi, USA

ABSTRACT: The apparent complexity and sophistication of reliability approaches in pavement design
has rendered them the exclusive domain of statistical experts, while the rest of the profession remains
vociferously polarized and painfully divided. This paper lifts the artificial obscurity of theoretical manip-
ulations by demonstrating that reliability is simply our modern equivalent to good old safety factors,
which every engineer can comprehend. This is accomplished by appealing to common-sense engineering
and employing simplified terminology. It is discovered that the 1986/1993 AASHTO Design Guide sends
the engineer on an impossible mission, with very little guidance as to how to select the necessary param-
eters. The scarcity of information in the Guide itself, as well as in textbooks and journals, is attributable to
the fact that some parameters are simply impossible to determine. Rather than selecting a single value in
an admissible range, the engineer is well advised to turn to Monte Carlo simulation, which will elucidate
the repercussions of each parameter chosen.

1 INTRODUCTION heretofore shrouded only in artificial obscurity,

and to illustrate options offered by common-sense
Few issues vex pavement engineers more than engineering in its application to pavement design.
reliability. Many practitioners have despaired of
understanding it, and leave decisions about it to
more statistically oriented colleagues, or rely on 2 RELIABILTY IN AASHTO GUIDES
cook-book answers from available textbooks or
on prior experiences. When reliability is discussed In the context of the AASHTO series of pavement
at all, the conversation often degenerates into an design guides, reliability has appeared belatedly as a
argument, as opinions are helplessly polarized, second thought, of sorts. Pioneering methodologies
and no amount of statistically generated evidence had included a safeguard against uncertainties and
suffices to bring about a rapprochement. There is variabilities in the form of a safety factor, SF, most
a widespread feeling that, at least in the context of often defined as the ratio of a calculated working
the series of design guides released by the Ameri- stress to a selected ultimate stress (e.g., Sheets 1933):
can Association of State Highway and Transpor-
tation Officials (AASHTO), including the 2008 SF = σw/σu (1)
Mechanistic-Empirical Pavement Design Guide
(MEPDG), reliability is a topic fraught with com- In contrast, the exclusively statistical/empiri-
plexity, intricacies and sophistication, which render cal design procedures developed in the aftermath
it simply beyond the reach of most practicing engi- of the AASHO Road Test (1958–1960) could not
neers. The purpose of this paper is to contribute to adopt such a definition of a SF, since there was
the conversation about reliability by demystifying no mention or calculation of a working stress, or
the theory behind it, simplifying the presentation of any other mechanistic response parameter. This
offered in existing guides and suggesting ways in omission was probably recognized as a weakness
which decisions about it may be formulated by to be changed “in the future” (Liddle, 1962), even
engineers, through application of their general as AASHTO insisted that “if you want to know
design background, knowledge and expertise. It is how the pavement is going to hold up under traf-
anticipated that the result of this effort will be to fic, you have to go one step further” than mechan-
demonstrate that the issue of reliability has been ics of stress calculation permits (Carey, 1962).

549
A self-evident option would have been to define In this mathematical description of reliability, N
SF as the ratio of the calculated working number can be considered analogous to ultimate strength,
of repetitions (or Equivalent Single-Axle Loads, while n as analogous to working stress, which for a
ESALs) divided by a specified ultimate number of safe structure, should be smaller than the strength.
repetitions. This solution was not pursued, how-
ever, because the profession had at the time already
been moving away from the concept of the SF. The 3 AASHTO RELIABILITY MADE SIMPLE
latter had been criticized by more mathematically
inclined investigators as “theoretically inadequate What is intrinsically a simple idea, serving the
for the rational solution of any but the most sim- need to guard against unforeseen failure in a man-
ple structural problems” (Freudenthal 1947). To ner directly analogous to the function of the con-
replace SF, statistical analysis had been proposed ventional and familiar safety factor, emerged as a
for the estimation of “probabilities of [structural] nebulous and highly technical concept in the pres-
unserviceability and of failure” (Freudenthal 1956) entations published at the time of the release of the
as well as of the “reliability of fatigue-sensitive 1986 AASHTO Design Guide (AASHTO 1986).
structures” (Freudenthal 1961). Over the years, These presentations appeared scattered in several
therefore, engineers have tended to gravitate toward individual reports, lacking continuity and uniform-
reliability considerations. The shift in favor of reli- ity, and were characterized by the use of multiple
ability was anticipated by Freudenthal (1962), who inconsistent symbols, subscripts and definitions,
had declared that “no engineering significance thereby creating a formidable morass of informa-
can be attached to any specific single value of [the tion, and creating more questions for every answer
safety] factor” and that “the probability or risk provided. The story related in these presentations,
of failure or of survival is a unique parameter of however, is a very familiar one to any design engi-
the distribution function of the safety factor, and neer, and it would be fruitful to repeat it here.
should replace the conventional concept of safety. Every pavement recommendation is essentially a
Its generalization in the form of a ‘reliability func- statement by its designers than it is capable of car-
tion’ provides the basis for an integrated safety rying the traffic that will be applied to it during its
analysis of structures”. life. We may postulate that personnel in the Office
Drawing on a concept proposed by investigators of Engineering use material characterization and
in the nuclear industry, Lemer and Moavenzadeh prevailing structural theories to estimate pavement
(1968) defined reliability as: “The probability that capacity, N; on the other hand, experts in the Bureau
the system will perform in a satisfactory manner, for of Statistics rely on data from weigh stations and
at least a given length of time, under a prescribed information from sociology, political affairs, eco-
set of conditions”. To estimate it, they calculated nomics, etc., in order to estimate the expected traf-
the probability of failure of various pavement sec- fic count, n. The purpose of design is to ensure that
tions and then established reliability as: the n never exceeds N. Naturally, neither estimate is
infallible, since both pertain to future predictions.
R = 1 – pf (2) In other words, it may be taken for granted that esti-
mates from both the engineering and the statistics
This formulation was subsequently adopted by teams incorporate some unknown degree of error;
Darter and Hudson (1973), who applied “proba- in fact, if the pavement ultimately fails it would be
bilistic design concepts so that the various uncer- impossible (and entirely unproductive) to pinpoint
tainties and variations of the design system might which of the two teams was responsible for the fail-
be quantified”. Such uncertainties included: (a) ure. Table 1 presents this information visually.
variation of design parameters within a design Since N and n are not known a priori, their
project length; (b) variation between the design approximate estimates at the time of design are
and actual values of the design parameters; and, denoted in this Table as W and w respectively.
(c) variation due to the lack of fit of the design The Table also defines theoretically four ratios
methodologies. Darter and Hudson (1973) for- that may be extracted on the basis of these four
mulated reliability on the basis of the number of parameters, only one of which actually lends itself
ESALs the pavement could carry (i.e., the capacity to calculation (FR = W/w). The error inherent in
of the pavement), N, and the number of ESALs the statistical traffic estimate, g = w/n, is a variable:
that will actually be applied to the pavement (i.e., considering a large number of designs prepared,
the demand on the pavement), n. Mathematically, it may be expected that the average error will be
they described reliability “as the probability that N close to 1.0 (indicating that statisticians are com-
will exceed n”, or: petent, albeit fallible); its variance is denoted by
SS2. Similarly, the corresponding error in the engi-
R = P[N > n] (3) neering estimate of capacity, g* = W/N, may also

550
Table 1. Simplified parameters.

OFFICE OF BUREAU OF
ENGINEERING STATISTICS

PAVEMENT CAPACITY TRAFFIC COUNT RATIO

(Sustainable traffic before (applied over design CAPACITY/
failure e.g., PSI = 2.5) life e.g., 20 years) COUNT

ESTIMATED W w FR
(can only be predicted;
may vary with each staff member)
ACTUAL N n F R*
(precise, but only known in Error in overall
retrospect) pavement design
Variance, SO2
RATIO g* g
ESTIMATED/ACTUAL Error in Section Error in Design
Performance Period Traffic
Variance, SE2 Variance, SS2
Capacity Prediction Count Prediction
Uncertainty Uncertainty

be expected to have a mean close to 1.0 (indicat- pavement design was adopted by AASHTO and
ing that engineers are competent, albeit fallible); was officially first introduced in 1986 (AASHTO,
its variance is denoted by SE2. Finally, the error in 1986). The 1993 revised version (AASHTO 1993)
overall pavement design, FR* = N/n ≈ FR = W/w, left the reliability analysis untouched. There-
may be expected to have a mean value of about fore, these two guides will be termed together as
1.0 and a variance of SO2; statistical theory (which AASHTO 86/93 in this study.
need not distract us at this time) requires that the
value of SO2 may be obtained from SS2 and SE2, as
follows: 4 AASHTO GUIDELINES FOR
RELIABILITY IMPLEMENTATION
SO2 = SS2 + SE2 (4)
AASHTO 86/93 refers to zRSO as a “positive spac-
Darter and Hudson (1973) assumed that N and ing factor” between estimated capacity, W, and
n (as well as their approximate estimates W and w) estimated demand on the pavement section, w, call-
follow lognormal distributions. N and W were con- ing it the “reliability design factor,” FR:
sidered to have lognormal distributions because
“number of applications to failure in a fatigue FR = exp10 [–zR SO] (6)
test” and “error in predicting the N to failure”
were found to be lognormally distributed; n and It is not difficult to see that the reliability design
w were postulated to be lognormally distributed factor is analogous to the traditional SF, since both
because they depended on design [average daily strive to make the pavement design safe by intro-
traffic] ADT, percent trucks, axle factor and axle ducing conservatism into the pavement design.
load distributions and which have normal distribu- Unlike traditional safety factors, which are usually
tion. The introduction of logarithms complicates multiplicative, the reliability design factor is addi-
somewhat the mathematics at hand, but does not tive; yet, like the safety factor, it is mandatory for
alter the underlying fundamental principles out- pavement design because it is impossible to predict
lined above. Therefore, Darter and Hudson (1973) the pavement’s future precisely. It can also be seen
postulated that a reliable pavement design would that the reliability design factor consists of two
result by setting: terms, i.e., the standard normal deviate, zR, cor-
responding to a specified reliability level, R, and
LogW = logw + zR√(SS2 + SE2) = logw + zR SO (5) the overall standard deviation, SO. zR is obtained
from the standard normal distribution table and is
in which zR = standard normal deviate of the nor- always negative. zR seeks to increase the confidence
mal distribution curve with mean zero and vari- that a proposed design will perform as intended;
ance 1.0, for a given level of reliability, R. This mathematically, it is the area under the standard
methodology of introducing conservatism into normal distribution curve, to the left of the mean

551
value of 0. Since SO is positive, zRSO is negative and Therefore, in either case estimation of SO is
decreases the capacity of the pavement design, impossible. Consequently, the reliability design
thereby introducing conservatism into the design. factor (or safety factor) remains incomputable,
In applying reliability, AASHTO divided rural and this makes it difficult for designers to under-
and urban pavements into four different functional stand the repercussions of their reliability choices
categories and recommends a reliability range for on the safety margin inherent in their design.
each one, beginning from a low value of R = 50% When the recommended SO-values are introduced
for “local” roads to an upper value of R = 99.9% in Eq. (6), multiplicative SF values are obtained,
for “interstates and other freeways”. AASHTO ranging between 0 and 17 for reliability levels of
also provided the following values of SO: (a) For 50% to 99.9%. Most engineers are accustomed to a
flexible pavements: 0.4–0.5; and, (b) For rigid pave- typical range of SF between 1 and 5. These short-
ments: 0.3–0.4. According to AASHTO, the lower comings have plagued the development of reliabil-
value in each range “corresponds roughly to the ity concepts in the aftermath of the publication of
estimated variances associated with the AASHO the AASHTO 86/93 guide.
Road Test and the original pavement performance
models presented in the previous (1972 and 1981)
Design Guides”. Moreover, it was suggested that 6 RELIABILITY IN NCHRP 1-26
in Eq. (4), “SS2 is around 20–25% of the overall (1988–1998)
variance, SO” (AASHTO, 1985), from which it may
be inferred that SE2 is around 75–80% of SO2. Upon the release of AASHTO 86/93, “it was
decided that research should be initiated immedi-
ately with the objective of developing mechanistic
5 SHORTCOMINGS OF AASHTO pavement analysis and design procedures suitable
RELIABILITY FORMULATION for use in future versions of the AASHTO guide”.
The first such effort was titled “Calibrated Mecha-
As stated in conjunction with Table 1, above, nistic Structural Analysis Procedures for Pave-
determination of SO requires four terms: N, n, W, ments”, and was funded under NCHRP Project
and w. We shall attempt to evaluate these for the 1-26 (Barenberg and Thompson, 1992). It sought
following cases: (1) Survival of a pavement section; initially to use principles of engineering mechanics
and, (2) Failure of a pavement section. in computing deflections, stresses and strains in the
pavement structure, subjected to combinations of
Case 1: Survival of a pavement section means that traffic and environmental loadings. These mecha-
this has not failed at the end of its design nistic responses were subsequently to be related to
life and it continues to serve traffic. In this pavement damage through statistical “transfer”
case, W and w are, of course, known (since functions, established using actual distress obser-
design time). Moreover, the actual traffic vations. Given that SO could not be evaluated,
that has been applied on the pavement sec- NCHRP 1-26 opted for the use of a “traffic mul-
tion during its design life, n, can also be tiplier”, MFR, as “the most realistic approach for
ascertained, if adequate records have been considering design reliability.” The new parameter
kept. The actual capacity, N, however, was nominally considered to be equivalent to the
cannot be computed since the pavement “reliability design factor,” FR. Nonetheless, focusing
section has not failed, yet. As a result, SS2 on the engineering estimates of pavement capacity,
can be calculated, but SE2 cannot be deter- investigators admitted that “since all factors which
mined. Since, SE2 cannot be estimated, SO2 affect pavement performance cannot be identified
cannot be calculated either. and quantified at this time, it is difficult to accurately
Case 2: Failure of pavement section means that quantify the pavement performance error term,” SE2.
this went out of commission before the A similar argument could have been made for statis-
end of its design life and it is no longer tical traffic count estimates, as well, since calculation
serving traffic. Again, the values of W and of SS2 poses analogous challenges. Consequently,
w have been available since design time. for rigid pavements, the following traffic multipliers
Given that the pavement has failed before were recommended “based on the level of informa-
the end of its design life, its actual capac- tion currently available”: (a) For high reliability (90
ity, N, is known (again, assuming adequate to 95%): 2.5 to 3.5 or 4.0; and, (b) For medium reli-
records have been kept), but its actual traf- ability (75 to 85%): 2.0 to 2.5. In contrast, for flex-
fic count, n, cannot be ascertained, since ible pavements the recommendation was to calculate
the pavement section no longer serves traf- MFR using the variability of a measured or calcu-
fic. Therefore, it is infeasible to calculate lated mechanistic response, e.g., surface deflection,
SS2, and, consequently, SO2. Δ, in the following regression formula:

552
MFR = [1 + (zR COV)]4.6 (7) loadings, as well as of material inputs. Each accu-
mulated distress type is called a performance indi-
where COV is the coefficient of variation of Δ, cator, whereas the corresponding specified critical
assumed to be 0.20 for “full depth” and 0.25 for distress level not be exceeded is termed a design
“conventional” asphalt pavements. Flexible pave- criterion or threshold limit. The latter is selected
ment traffic multipliers between 3 and 4 were either from a set of MEPDG recommendations or
quoted. The choice of surface deflection in this for- at the discretion of the designer of record. Deter-
mula was justified by the statement that “surface mination of performance indicators starts with the
deflection is perhaps the most well documented calculation of mechanistic responses using struc-
flexible pavement transfer function currently avail- tural analysis techniques on the basis of average
able.” Moreover, it was advocated that the use of input values. These responses are then correlated
measured pavement responses “offers a direct and to distress levels using statistical regression algo-
accurate variability estimate as contrasted with rithms, previously established using a database of
the [AASHTO] procedures previously presented.” field measurements. Invariably, the latter exhibit
It was anticipated that “adoption of mechanistic a lack of fit, which is expressed as the “standard
based procedures with reliability analysis based on error of the distress prediction”, i.e., the stand-
the analysis of variance should gradually reduce ard deviation of the difference between predicted
[the] error term in the design process” and rem- and observed distress levels, STDmeas; records of
edy the “lack of fit for the AASHO Design equa- these statistics are available for each of the statis-
tion”, which was deemed to be “the largest single tical regression algorithms (or “models”) used in
factor affecting its ability to predict performance” the 2008 MEPDG. In order to accommodate the
(Barenberg and Thompson, 1992). desired level of reliability, P, say, the performance
indicator thus calculated using mean input values,
Distress_mean, is incremented as follows before
7 RELIABILITY IN NCHRP 1-37A comparing it to the corresponding threshold limit:
(1998–2008)
Accumulated distress over design period
Both the concerns and the arguments of the = Distress_mean + STDmeas × zP (9)
NCHRP 1-26 researchers were echoed in the
reliability considerations adopted in the second In other words, the reliability examined is a
attempt to compile a set of AASHTO mechanistic- measure of the mechanistic-empirical distress pre-
empirical design procedures. The new effort com- diction methodology employed, rather than that of
menced in 1998 under program NCHRP 1-37A, the pavement section itself. It was anticipated that
whose transfer functions were extracted from a since the distress model formulation and the design
much larger database, assembled as a result of the calculation constitute a closed loop, this approach
Long Term Pavement Performance (LTPP) stud- would result in the elimination of the “lack of fit
ies and other field experiments. The reliability problem” as well as in more reliable pavements.
assessment in the resulting MEPDG (AASHTO Experience to date has provided some insights
2008) employs multiple mechanistic responses to with regard to the first of these two expectations.
produce individual reliability estimates for each Unfortunately, the lack of fit persists as statistical/
distress type guarded against, each of which pro- empirical models are applied in a variety of geo-
vides a measure of the error between predicted and graphic and climatic regions, evincing the inabil-
observed (or expected) values. This approach ful- ity of such approaches to capture the underlying
fills the expectation of NCHRP 1-26 for a remedy mechanistic phenomena leading to the distresses.
to the “lack of fit problem”, but in the process reli- A “local recalibration” is found to be necessary in a
ability is redefined as the probability that each of large proportion of the cases reported; this has cre-
the key distress types, accumulated over the entire ated significant concerns about the overall reliabil-
design period, will be less than a specified critical ity of pavement sections designed using the 2008
level, with a desired level of reliability. Expressed MEPDG, as well.
mathematically, the 2008 MEPDG reliability defi-
nition is:
8 MONTE CARLO SIMULATION
R = P[Accumulated distress over design
period < Specified critical distress level] (8) It is apparent that the practical impossibility
of computing SO has led investigators involved
All distresses are assumed to be normally dis- in both the NCHRP 1-26 and 1-37A studies to
tributed variables that are functions of variations abandon the reliability methodology introduced
and uncertainties in estimates of traffic and other in the AASHTO 86/93 guide, despite its obvious

553
attractiveness. Not only does the 1986 reliability set of variables, e.g., success or failure of the pave-
approach provide a convenient bridge to the con- ment system; and (e) Assessment of the probabi-
ventional and familiar factor of safety concept, listic nature of the problem upon the completion
but it is also consistent with contemporary text- of a large number of solutions, per (d), above, e.g.,
books on reliability in other branches of engineer- calculation of the reliability of the pavement sys-
ing (e.g., Haldar and Mahadevan 2000). These tem. The validity of the result obtained from (e) is
textbooks also provide a number of choices for verified by assessing the accuracy and efficiency of
overcoming difficulties such as those confronting the simulation, e.g., by repeating the entire exercise
pavement engineers using numerical procedures several times, with a larger or a smaller number of
that exploit the power of modern computational simulation iterations.
tools. A very popular such approach goes by the
vernacular name of Monte Carlo Simulation
(MCS), and deserves further exploration. 9 APPLICATION OF MCS TO AASHTO
MCS is one of the oldest computational appli- 86/93
cations to the solutions of probabilistic problems
deemed too complex to have an analytical solu- Even a cursory literature review suffices to illus-
tion. It was developed in the 1940s primarily out trate that the treatment of reliability in MEPDG
of the US effort to build the first atomic bomb. (AASHTO, 2008) is significantly different from
The names associated with these early efforts are that in the NCHRP 1-26 Reports (Barenberg and
those of Los Alamos scientists John von Neumann Thompson, 1992). Moreover, both are unlike the
(1903–1957) and Stanislaw Ulam (1909–1984), but preceding decade-old methodology, to be found
the term itself was coined by Nicholas Metropo- in the statistical/empirical AASHTO Guides
lis (1915–1999) during the top secret Manhattan (AASHTO, 1986; 1993), whose own roots can
Project. The method involves random number be traced to the Texas Flexible Pavement System
generation, which explains the reference to Monte (FPS) of the early 1970s (Darter and Hudson,
Carlo, Monaco, where the main attractions are 1973). Perhaps the only common thread con-
casinos of games of chance. As one of several necting all these AASHTO reliability efforts is an
simulation techniques, MCS is intended to repli- unwillingness (or inability) to implement the one
cate numerically a real complex system by succes- methodology that could have provided answers to
sive generation of the random variables entering many vexing concerns, namely Monte Carlo Simu-
it and solution of the equations governing it for lation. Yet, since at least the early 1970s, the use of
every set of such values. Harnessing the power of MCS had been advocated in implementing reliabil-
computers for this purpose, the problem is solved ity analysis, since “simulation methods become the
hundreds or even thousands of times, permit- only feasible approach” for systems posing difficul-
ting investigators to extract useful conclusions ties “to attain the initial estimates of failure proba-
that would have otherwise required a prohibitive bilities [and, consequently, of reliability] because of
amount of experimentation. Using MCS in 1948, the complexity of the physical processes involved”
Metropolis obtained estimates for the eigenvalues (Lemer and Moavenzadeh, 1971). Over the years,
of the Schrödinger equation, but since then the such recommendations have been dismissed
method has been applied in other branches of as being complex and unfit to use (Darter and
physics, medicine, and engineering, as well as in Hudson, 1973; Barenberg and Thompson, 1992).
business, economics, investing, communications, Most recently, while admitting that the reliability-
management, and risk analysis, to mention but a based design approach implemented in MEPDG
few areas. “is not as sophisticated as a Monte Carlo–based
Application of MCS involves: (a) The definition one,” Darter et al. (2005) opined: “The MC-based
of the problem, usually though a set of governing approach was recognized as preferable, but it may
equations, e.g., the AASHTO 86/93 design equa- be too computationally demanding to be combined
tions for flexible and rigid pavements; (b) Identifi- with the incremental damage design.”
cation and quantification of the random variables To rectify this situation, it is proposed that
that enter these equations in terms of their entire MCS may be applied to address the debilitat-
probability distributions, e.g., the mean and stand- ing weaknesses in the reliability methodology of
ard deviation of a normal distribution; (c) The AASHTO 86/93, which led to its abandonment
repeated generation of random variables satisfying in the NCHRP 1-26 and 2008 MEPDG efforts. It
the specified probability distributions using avail- was shown above that in any given pavement case,
able software, e.g., a spreadsheet; (d) The solution the pavement either fails or it does not, and this
of the problem for each random variable set gener- allows the determination of only three of the four
ated in (c), above, and the extraction of a conclusion required inputs for So. It is suggested that this prob-
regarding the system observed for that particular lem may be overcome through the use of MCS.

554
Monte Carlo Simulation can provide a means to introduced in AASHTO 86/93 is difficult to grasp
resolve the problem identified in establishing So, and impossible to calculate. Consequently, an
by relying only on three of the required pieces of engineer is left without the benefit of an intuitive
information, plus an assumed conventional factor appreciation of the selected level of reliability, as
of safety or FR value. Accordingly, reliability may might be feasible if R were related to the more con-
be established as the number of successful designs ventional factor of safety, SF.
divided by the total number of simulations con- To address these limitations, MCS is proposed
ducted. The overall standard deviation may then as the basis of an approach that combines com-
be obtained from Eq. (4). For this purpose, all that mon sense and well understood measures of mate-
is required is a computer spreadsheet, which can rial and traffic variability, viz., means, standard
perform the simulations using typical input param- deviations and coefficients of variation. A compu-
eters and their respective coefficients of variation, ter spreadsheet is envisaged for this purpose, simu-
available in the literature (e.g., Huang, 2004). lating several thousand pavement sections designed
The same spreadsheet implementation of MCS using the AASHTO 86/93 pavement design equa-
may also lead to the resolution of the second obsta- tion and assumed values of SF. As a result, reli-
cle identified above, namely the significance of the ability estimates will be obtained by counting the
recommended AASHTO reliability levels. This number of successful or failed sections. Moreover,
may be accomplished by correlating the assumed the MCS methodology will allow the determina-
conventional factor of safety or FR values with the tion of SO in a manner that dispenses with the
reliability values resulting from the simulations. guidelines presented in AASHTO 86/93.
Consequently, guidelines may be formulated for
selecting the design reliability level, in a manner
that is meaningful to practicing engineers accus- ACKNOWLEDGMENT
tomed to the conventional factor of safety.
In addition, such an MCS approach would pro- This paper was developed under Contract Number
vide answers to questions such as the following: W912HZ-16-P-0123: Framework for Mechanistic-
What is the relationship between Reliability and Empirical Concrete Pavement Design Procedure,
Factor of Safety? How is the relationship between from The Engineering Research and Development
Reliability and Factor of Safety influenced by vari- Center (ERDC) of the United States Army Corps
abilities in material properties and in traffic? How of Engineers (USACE) to the University of Cincin-
is Overall Standard Deviation, SO, related to Reli- nati (UC). Dr. Jeb S. Tingle was the Contracting
ability (or Safety Factor)? Can we obtain estimates Officer’s Representative (COR). Dr. Anastasios M.
of standard error of traffic prediction, SS and Ioannides was the Principal Investigator (PI).
standard error of capacity prediction, SE? What is
the effect of COV values for individual inputs? To
what extent are the assertions and recommenda- REFERENCES
tions of AASHTO 86/93 verified by MCS?
AASHTO (1985), “AASHTO Guide for Design of
Pavement Structures, Appendix EE-Reliability Ele-
10 CONCLUSIONS ments and Variance Estimates,” American Associa-
tion of State Highway and Transportation Officials,
Reliability considerations were originally intro- Washington, D.C.
duced into the series of AASHTO pavement design AASHTO (1986), “AASHTO Guide for Design of Pave-
guides for the purpose of imparting a degree of ment Structures,” American Association of State High-
way and Transportation Officials, Washington, D.C.
conservatism to the design as a defense against AASHTO (1993), “AASHTO Guide for Design of Pave-
various uncertainties. Inadvertently, however, this ment Structures,” American Association of State High-
also resulted in making the pavement design proc- way and Transportation Officials, Washington, D.C.
ess more cumbersome and nebulous. Although AASHTO (2008), “Mechanistic Empirical Pavement
the mathematics entailed is fairly elementary, reli- Design Guide,” NCHRP 1-37A, American Associa-
ability analysis is frequently misunderstood and tion of State Highway and Transportation Officials,
misapplied, resulting in potentially overdesigned Washington, D.C.
or underdesigned pavements. This paper reviews Barenberg, E.J. and Thompson, M.R. (1992), “Cali-
the fundamental engineering principles implicit brated Mechanistic Structural Analysis Procedures for
Pavements,” NCHRP Project 1-26, Phase 2, NCHRP,
in reliability analysis and restates the AASHTO National Research Council, Washington, D.C.
methodology in simple, easy-to-understand terms. Carey, Jr., W.N. (1962), “Report on Session I - AASHO
It is shown that a major contributor to the unfor- Road Test and Performance Criteria,” Proceedings,
tunate situation prevailing today is the fact that (First) International Conference on the Structural
the parameter Overall Standard Deviation, SO, Design of Asphalt Pavements, p. 15, Ann Arbor, MI.

555
Darter, M.I. and Hudson, W.R. (1973), “Probabilis- Freudenthal, A.M. (1962), “Safety, Reliability and
tic Design Concepts Applied to Flexible Pavement Structural Design,” Transactions, Proceedings of the
System Design,” Report No. CFHR 1-8-69-123-18, American Society of Civil Engineers, Vol. 127, Part II,
Center for Transportation Research, University of ASCE, Reston, VA, pp. 304–323.
Texas, Austin, TX. Haldar, A. and Mahadevan, S. (2000), “Probability,
Darter, M.I., Khazanovich, L., Yu, T., and Mallela, J. Reliability, and Statistical Methods in Engineering
(2005), “Reliability Analysis of Cracking and Faulting Design,” John Wiley and Sons, New York, NY.
Prediction in the New Mechanistic-Empirical Design Huang, H.Y. (2004), “Pavement Analysis and Design,” 2nd
Procedure,” Transportation Research Record 1936, Edition, Pearson Prentice-Hall, Upper Saddle River, NJ.
TRB, National Research Council, Washington, D.C., Lemer, A.C. and Moavenzadeh, F. (1968), “An Integrated
pp. 150–160. Approach to Analysis and Design of Pavement Struc-
Freudenthal, A.M. (1947), “The Safety of Structures,” ture,” Highway Research Record 291, pp. 173–185.
Transactions of the American Society of Civil Engi- Lemer, A.C. and Moavenzadeh, F. (1971), “Reliability of
neers, Vol. 112, ASCE, Reston, VA, pp. 125–180. Highway Pavements,” Highway Research Record 362,
Freudenthal, A.M. (1956), “Safety and the Probability pp. 1–8.
of Structural Failure,” Transactions of the American Liddle, W.J. (1962), “Report on Session I - AASHO Road
Society of Civil Engineers, Vol. 121, ASCE, Reston, Test and Performance Criteria,” Proceedings, (First)
VA, pp. 1337–1397. International Conference on the Structural Design of
Freudenthal, A.M. (1961), “Reliability Analysis Asphalt Pavements, p. 14, Ann Arbor, MI.
of Fatigue-Sensitive Structure,” Acta Technica, Sheets, F.T. (1933), “Concrete Road Design Simplified
Akademia Kiado, Hungarian Academy of Sciences, and Correlated with Traffic,” Portland Cement Asso-
Budapest, Hungary, pp. 55–81. ciation, Chicago, IL.

556
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Effect of treated subgrade layer on mechanistic-empirical

pavement design

S. Islam, A. Sufian & M. Hossain

Kansas State University, Manhattan, Kansas, USA

ABSTRACT: The AASHTOWare Pavement ME Design software represents a major development in

pavement design and rehabilitation over previous empirical design procedures. This software requires com-
prehensive input data including traffic load distribution, bound and unbound pavement layer properties
to do design analysis of pavements. This study focused on evaluating the effect of chemically-treated sub-
grade layers on flexible pavement performance as predicted by the AASHTOWare Pavement ME Design
software. In this study, Lime-Treated Subgrade (LTSG) as well as mechanically-stabilized subgrade types
was considered as subgrade soil treatment. Ten new flexible pavements with treated subgrade layer were
analyzed to evaluate the effect of subgrade treatment. In addition, the effect of treated subgrade layer on
new Jointed Plain Concrete Pavement (JPCP) design was also evaluated. The AASHTOWare software
predicted higher design thickness for flexible pavements designed with mechanically-stabilized subgrade
compared to the LTSG layer. Higher total permanent deformation was predicted for flexible pavements
without stabilized subgrade layers. Pavements designed with mechanically-stabilized subgrade was pre-
dicted to have lower rut depth compared to those with LTSG. Top-down fatigue cracking was also found
to be higher for the hot-mix asphalt pavements designed LTSG compared to the pavements designed with
mechanically-stabilized subgrade. The rigid pavement thickness design was found to be insensitive to the
subgrade soil treatment.

1 INTRODUCTION This study focused on evaluating the effect of

chemically-treated and mechanically-stabilized
The Kansas Department of Transportation subgrade layers on flexible pavement performance
(KDOT) is in the process of implementing the new as predicted by the AASHTOWare software. To
AASHTO Mechanistic-Empirical Design Guide account for high plasticity in the subgrade soil,
(MEPDG) for pavement design. AASHTOWare KDOT typically performs chemical or mechanical
Pavement ME Design is the software version of stabilization of the top six inches of the subgrade
MEPDG. The software allows designers to input layer. In case of chemical stabilization, lime is often
layer material properties and traffic input param- used alone or in conjunction with other stabilizing
eters in three levels: Level 1 is project-specific agents such as Class C fly ash to treat the subgrade
with extensive laboratory test results and precisely layer. Mechanical stabilization includes changing
counted traffic volume, distribution, and axle load the gradation of the treated soil layer by mixing
spectra; Level 2 contains regional input parameters two or more types of soil materials with one being
with minimal test results; and Level 3 is based on granular. Such treatment usually has a beneficial
global default values in the software. Based on effect on the properties of the soil—it reduces
input parameters, the software predicts selected plasticity and swell potential, improves workabil-
pavement distresses and smoothness, referred to as ity, and enhances durability (2). Soil stabilization
the International Roughness Index (IRI). A satis- of subgrade not only provides a strong platform
factory design is the one that limits these distresses to the pavement structure but also improves pave-
to some preset levels. MEPDG design models were ment performance and increases service life (3).
developed and calibrated using the Long Term Ten flexible pavements in Kansas were analyzed
Pavement Performance data (1). with the AASHTOWare software and KDOT-sug-
The AASHTOWare software requires soil gra- gested resilient modulus (MR) values and soil gra-
dation properties, Poisson’s ratio, and resilient dation properties were used as input parameters
modulus (MR) values in order to characterize sub- to investigate the long term performance of these
grade soil. pavements.

557
2 PROBLEM STATEMENT 4 AASHTOWare PAVEMENT
ME DESIGN INPUTS
The use of chemical and mechanical stabilization
techniques to improve roadbed soil properties Mechanistic-empirical design is an iterative proc-
has increased recently to modify subgrade and ess. Based on input parameters, the software pre-
base materials due to lack of high quality aggre- dicts selected pavement distresses and smoothness
gate in many states of the United States. Previous (International Roughness Index, IRI). For a suc-
researchers have studied the effect of treated sub- cessful design, a pavement section needs to satisfy
grade layer on the long term performance of the the distress and smoothness criteria at a specified
pavement. Chou et al. (2004) conducted a study to reliability level. These failure criteria and reliabil-
investigate the structural benefits of subgrade soil ity values are decided upon by a specific agency.
treatment and concluded that chemically-treated In this study, for flexible pavement analysis, the
soil exhibits higher strength at early age in terms selected failure criteria were 10% of total lane area
of resilient modulus and unconfined compressive for bottom-up fatigue cracking, 0.5 inch for total
strength compared to untreated soil (4). Sargand rutting, and 164 inch/mi for IRI as suggested by
et al. (2014) incorporated a finite element model KDOT. The design reliability chosen was 90%.
to investigate structural benefits of the chemically
stabilized layers and conclude that that the stiff-
ness of treated subgrade layer increases signifi- 4.1 Material properties
cantly over time. They also found that the treated In case of flexible pavement design, AASHTOWare
subgrade layer provides structural stability to flex- pavement ME design software requires dynamic
ible pavements (5). In this study, AASHTOWare modulus (E*), creep compliance, and indirect ten-
Pavement ME software was used to investigate the sile strength of the asphalt mix as Level 1 inputs.
benefits of treated subgrade layer in terms of long Dynamic shear modulus (G*) and phase angle (δ)
term pavement performance. Ten flexible pave- values of the asphalt binder are also required to
ments were analyzed incorporating lime-treated, generate dynamic modulus master curves for the
mechanically-stabilized and untreated subgrade asphalt mixes. Since these data were not available
layers. Furthermore, paired t-tests were conducted for the selected projects, Level 3 input values (aggre-
to investigate if there are significant differences in gate gradation, binder grade and mix volumetric
predicted design thicknesses due to different sub- properties) were extracted from the mixture designs
grade treatment. Ten new JPCP pavements were for the surface, intermediate, and base layers.
also analyzed with the AASHTOWare software to Table 2 lists the volumetric properties of the Hot-
investigate the effect of treated layer. Mix Asphalt (HMA) mixtures for these projects.

3 TEST SECTIONS 4.2 Subgrade properties

The flexible pavement sections chosen in this study The gradation and Atterberg limits of the subgrade
have full-depth HMA on top of six-inch treated soil are necessary inputs for the AASHTOWare
subgrade. The test sections identified with project Pavement ME Design software. In this study, sub-
ID for the purpose of analysis of this study Table 1 grade soil gradation values were obtained from the
lists the general features of the test sections. county-specific soil survey reports of the Soil Con-
servation Service. Additionally, AASHTOWare soft-
ware requires resilient modulus (MR) values in order
Table 1. General features of the test sections.
to characterize subgrade soil. KDOT provided a list
Subgrade Initial of average county-wide soil resilient modulus values
Project no. Route County soil type AADT that were used in this study. Table 3 lists the sub-
grade soil properties used in this study.
008U0005400-EB US-54 Butler A-6 235 The AASHTOWare software requires resilient
008U0007700-NB-3 US-77 Butler A-6 315 modulus, Poisson’s ratio and gradation proper-
019K0000700-NB-2 K-7 Crawford A-7-6 275 ties to characterize the treated-subgrade layer.
028U0005000-EB US-50 Finney A-6 1,170 KDOT suggested the resilient modulus value to be
031K0001800-WB K-18 Geary A-6 600 31,000 psi for the mechanically-stabilized subgrade
007U0007500-NB US-75 Brown A-6 1,000 layer. For the lime treated subgrade layer, the Mr
011U0006900-NB US-69 Cherokee A-7-6 540 was obtained using Equation 1.
033U0028300-NB US-283 Graham A-6 340
065K0002700-NB K-27 Morton A-7-6 275
095U0005600-EB US-56 Stevens A-6 500
LTSG MR (psi)
=2.03 * untreated subgrade MR (psi) + 225 (1)

558
 Table 2. HMA mixture volumetric properties of the flexible pavements.

PG binder grade Binder content (% by volume)

Project name Surface layer Binder layer Base layer Surface layer Binder layer Base layer

008U0005400-EB 64–28 64–28 64–22 12.3 10.7 10.1

008U0007700-NB-3 64–28 64–28 64–22 11.6 10.2 9.2
019K0000700-NB-2 64–28 64–28 64–22 10.3 10.9 10.9
028U0005000 EB 70–28 64–22 64–22 11.8 11.0 11.0
031K0001800-WB 70–28 70–28 64–22 11.8 9.6 9.6
007U0007500-NB 70–28 64–22 70–22 11.6 11 10.1
011U0006900-NB 64–28 64–28 64–22 12.0 10.9 11.1
033U0028300-NB 64–28 64–22 64–22 11.7 10.3 10.3
065K0002700-NB 58–28 58–28 58–28 11.9 9.0 9.0
095U0005600-EB 70–28 64–22 70–28 13.1 10.5 9.9

Table 3. Site-specific subgrade soil properties for new flexible pavements.

Gradation (% Passing) Atterberg limits Resilient

modulus,
Project No. #4 Sieve #10 Sieve #200 Sieve Liquid limit Plasticity index Mr (psi)

008U0005400-EB 100 100 90–100 52 29 2600

008U0007700-NB-3 100 100 95–100 47 27 2600
019K0000700-NB-2 100 100 91 57 32 2600
028U0005000-EB 100 100 90–100 30–55 10–35 3200
031K0001800-WB 100 100 90–100 30–45 10–20 2800
007U0007500-NB 100 100 95–100 30–50 15–25 2600
011U0006900-NB 100 100 90–100 47 27 3900
033U0028300-NB 100 100 95–100 25–40 9–17 3900
065K0002700-NB 92 64 9 16 2 4400
095U0005600-EB 100 100 90–100 30–55 10–35 4500

30 30

25 25

" 20 "' 20

~ 15 £ 15
~ 10 ; 10
-:
0 u
s§> s§' ~ ~ -"<:> -"<:> -"<:> -"<:> -"" -"<:> -"<:> ~-..."' -""
"'" ""' ,<:s fl ,,'J"' -v'S"' ,.,'J" ,-::>- ~'J" ..,~ 'v'"'"' "''"'~ -v""'
Axle Load (lb) Axle Load (lb)
Cl AASHTOWare ME • Kansas Rllral r:lAAS4TOWi1rP. MF • Krmso:~s R1 1rr:~l

Figure 1. Single (left) and tandem (right) axle-load spectra for class 9 trucks in Kansas.

4.3 Traffic
The Poisson’s ratio of the treated subgrade layer
was considered to be 0.2 as per KDOT suggestion. Traffic data is a primary input for the structural
Overall, the top 18 inches of the subgrade gets design of pavements using the mechanistic-empir-
compacted to 95% of the maximum dry density ical pavement design methodology. Annual Aver-
obtained in the standard Proctor test (AASHTO age Daily Truck Traffic (AADTT) for the year of
T99). construction, percentage of trucks in the design

559
direction and on the design lane, and traffic growth c25o
rate are required inputs. =
-;- 200
KDOT typically collects Weigh-In Motion
(WIM) data for short (48 hr.) durations at 73 port-
j 150
able WIM sites and five permanent sites. This study ~ 100
developed statewide distribution of axle loads for ; 50
CQ
rural principal arterials using average axle loads
across five WIM sites. AASHTOWare software also
requires axle load spectra to represent loads due
to mixed traffic. These spectra were derived from
traffic data collected at WIM stations and repre-
sent the percentages of total axle load applications
within designated load intervals for various axle lliiLTSG D Submod Iii Untreated
configurations. Figure 1 compares distributions of
single and tandem axle loads for truck class 9 on Figure 2. HMA base thicknesses as predicted by
AASHTOWare Pavement ME.
rural principal arterials in Kansas with the default
AASHTOWare software values. Vehicle class dis-
tributions and monthly adjustment factors were
conducted to investigate if the AASHTOWare-pre-
also developed for Kansas by analyzing Automated
dicted design base thicknesses for both subgrade soil
Vehicle Classification (AVC) data of eight stations
treatment strategies are significantly different. The
located on rural principal arterials in Kansas.
p-value (probability of occurrence of a given event)
obtained for this analysis was less than 0.05 (0.0051),
which suggests that the required design thickness is sig-
5 ANALYSIS AND RESULTS
nificantly less than for the pavement sections designed
with mechanically-stabilized subgrade layer than
In this study, the thicknesses of the surface and
that for sections with—the LTSG layers. Paired t-test
binder (intermediate) layers of the flexible pave-
conducted between the design thicknesses obtained
ments were kept constant for each case and the
for the LTSG and untreated layers did not show any
required AC base thickness was determined.
significant differences. However, AASHTOWare-pre-
Design analysis by the AASHTOWare software
dicted HMA base thickness was significantly less than
was performed based on a strategy that, if a section
when designed with mechanically stabilized layer than
passed all criteria for smoothness (IRI) and other
for the LTSG layer. In all design cases, total pavement
distresses, the thickness was reduced by 0.5 inch,
deformation and top-down fatigue cracking were the
and the analysis was redone. The section with the
contributing failure factors.
resulting bituminous base thickness (thickness of
previous trial just before failure) was then accepted
5.2 Rut depth
as the design thickness. For the purpose of analysis,
mechanically-stabilized subgrade soil was referred To investigate the rut depth predicted by the
to “Submod” in this study. AASHTOWare Pavement ME Design, the bitu-
minous base for all the flexible pavement sections
was fixed at 2.5 inches. Figure 3 shows the total
5.1 Flexible pavement design
pavement permanent deformation for the pave-
The AASHTOWare software was locally cali- ment sections designed with LTSG, Submod
brated for Kansas prior to this study (6). However, and untreated subgrade layer as predicted by the
local calibration of the bottom-up fatigue crack- AASHTOWare software. Lowest total pavement rut-
ing model was not done since KDOT recognizes all ting was obtained for the sections with the Submod
load-related cracking as top-down cracking. Cali- layer. On the other hand, sections with untreated
bration of the top-down cracking model yielded subgrade layer had the highest total rutting.
changes in the fatigue damage model, thereby
inducing significant bottom-up fatigue cracking.
5.3 Longitudinal cracking
Therefore, this study used the nationally-calibrated
fatigue model to predict distresses and the locally- The total top-down longitudinal cracking with
calibrated model to predict pavement rutting and 2.5-in. bituminous base for all design strategies
IRI. The thicknesses of the flexible test sections is shown in Figure 4. In all cases, pavement sec-
obtained from the AASHTOWare ME software tions designed Submod showed lower longitudinal
are illustrated in Figure 2. cracking than the LTSG and untreated subgrade
Figure 2 shows that test sections designed with layers. Lower AASHTOWare software-predicted
Submod layer required lower base thickness than longitudinal cracking was obtained with LTSG
the sections with the LTSG layer. A paired t-test was compared to the untreated subgrade layer.

560
 • The HMA design thickness predicted by the
AASHTOWare software for flexible pavement
sections is significantly higher when designed
with a LTSG layer compared to the mechani-
cally-stabilized subgrade layer. AASHTOW are
software predicted higher HMA thickness for
pavements sections without chemically treated
subgrade. r.
• AASHTOWare software predicted higher rut
depth for flexible pavements without a treated
subgrade layer. The predicted rut depth is lower
for pavements designed with mechanically-sta-
bilized subgrade layer compared to pavements
Figure 3. Total rutting for LTSG, submod and designed with a LTSG layer.
untreated soil subgrade. • AASHTOWare software predicted higher longi-
tudinal cracking for flexible pavements without
treated subgrade. The longitudinal cracking is
lower for pavements designed with mechani-
cally-stabilized subgrade compared to the pave-
ments with the LTSG layer.
• The required slab thickness for jointed plain
concrete pavements did not show any sensitivity
toward subgrade soil treatment.

ACKNOWLEDGEMENTS

This study was funded by the Kansas Department

IUTSG OSUbmod 8 Untreated of Transportation (KDOT).
Figure 4. Top-Down fatigue cracking for LTSG, sub-
mod and untreated soil subgrade.
REFERENCES

5.4 Jointed Plain Concrete Pavement (JPCP) Al-Jhayyish, Anwer K. “Incorporating Chemical Stabi-
design lization of the Subgrade in Pavement Design and Con-
struction Practices.” Phddissertation, Ohio University,
Ten JPCP sections were designed with the Athens, 2014.
AASHTOWare Pavement ME Design software Carpenter, S. H., Crovetti, M.R., Smith, K.L., Rmeili,
incorporating untreated, LTSG and Submod E., and Wilson, T. Soil and Base Stabilization and
layers. The rigid pavement sections include a Associated Drainage Considerations. Volume I: Pave-
Portland Cement Concrete slab over on Port- ment Design and Construction Considerations. Final
land Cement-Treated Base (PCTB) on top of a Report No. FHWA-SA-93-004. 1992.
Chou, E., Fournier, L., Luo, Z., and Wielinski, J. Struc-
six-inch treated subgrade. In case of rigid pave- tural Support of Lime or Cement Stabilized Sub-
ments, the failure criterion for IRI was selected grade Used with Flexible Pavements. No. FHWA/
to be 164 in/mile. Failure joint faulting and trans- OH-2004/017, 2004.
verse slab cracking were chosen as 0.125 inch NCHRP. Guide for Mechanistic-Empirical Design of New
and 5 percent, respectively. The AASHTOW are and Rehabilitated Pavement Structures. Final Report.
software predicted the same required JPCP slab NCHRP Project 1–37A. Transportation Research
thickness for each section irrespective of the sub- Board, National Research Council, Washington,
grade layer type. D.C., 2004.
Sargand, S., Khoury, I., Gray, J. and Al-Jhayyish, A.
Incorporating Chemical Stabilization of the Subgrade
in Pavement Design and Construction Practices. No.
6 CONCLUSIONS FHWA/OH-2014/12, 2014.
Sufian, A. A. Local Calibration of Mechanistic. Empiri-
The following conclusions were drawn based on cal Pavement Design Guide for Kansas. M.S. Thesis,
this limited study: Kansas State University, Manhattan, 2016.

561
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Comparison of practice for aggregate use in road construction—results

from an international survey

M. Fladvad
Department of Geoscience and Petroleum, NTNU—Norwegian University of Science and Technology,
Trondheim, Norway
Norwegian Public Roads Administration, Trondheim, Norway

J. Aurstad
Norwegian Public Roads Administration, Trondheim, Norway

B.J. Wigum
Department of Geoscience and Petroleum, NTNU—Norwegian University of Science and Technology,
Trondheim, Norway

ABSTRACT: Conventional flexible pavement structures are constructed using aggregates, in a com-
bination of unbound and bitumen bound materials. Crushed rock aggregates are fundamental for road
construction in many countries. This paper presents the results of a survey examining differences in the
use of aggregates for road construction in 18 countries. The purpose of the study is to gain knowledge
about the varying practices for aggregate use and pavement design, to enable better interpretation of
international research on this topic. The results confirm that crushed rock is the standard material used in
road construction. The aggregate market is international; the aggregate industry in e.g. Norway produces
aggregates for both export and domestic use. Pavement materials are regulated through international
standards, provided by e.g. CEN and ASTM, in combination with national standards and requirements.
The international standards do not set direct requirements; they provide a common system for classifi-
cation of road materials. The comparison of typical flexible pavement structures in different countries
shows that even though the materials overall meet equivalent standards; the practices for material use are
divergent. There are considerable differences in both pavement thickness and aggregate size used in the
studied countries. In the current situation, traditional physical quality tests dominate aggregate choice.
Functional aggregate testing could increase aggregate utilization.

1 INTRODUCTION ture, in unbound and bound construction materi-

als. The nature of the materials and the amount of
1.1 Aggregates—raw material for construction materials used will vary, depending on local condi-
The purpose of this study is to gain knowledge tions such as traffic, subgrade, and climate.
regarding the different practices for aggregate use Road pavements are generally divided into two
and pavement design in various countries, to ena- types: flexible and rigid pavements. The main differ-
ble better interpretation of international research ence is the use of bituminous vs. concrete materials
on this topic. in the upper parts of the pavement structure. Com-
A definition for aggregates is given by CEN mon for both pavement types is the use of aggre-
(2009): Granular material of natural, manufactured gates, both unbound and bound. Hot mix asphalt
or recycled origin used in construction. A simpler consists of approximately 95% aggregates, whereas
definition is provided by the European Aggre- concrete has an aggregate content of about 70%.
gates Association (2016): Aggregates are a granular National and international standards regu-
material used in construction. late the use of aggregate materials through e.g.
Crushed rock aggregate is a non-renewable the standards EN 13242 (CEN 2009) and ASTM
resource of great interest in road construction and D2940 (ASTM International 2015).
other branches of the construction industry. Uti- Arnold et al. (2003) analysed a selection of
lization of aggregates should be considered care- unbound granular materials of marginal quality
fully to prevent resource scarcity. and found that it is possible to assess suitability and
When a road is built or rehabilitated, aggregates performance of non complying aggregates for use in
are used as raw materials for the pavement struc- different layers in the pavement structure. The pos-

563
sible use of alternative test methods showing the Table 1. Countries represented in survey results.
functional properties of granular materials failing in
traditional mechanical tests, including recycled (sec- Continent Country Abbreviation
ondary) materials, has also been discussed by Aurs-
Asia Japan JP
tad and Hoff (2002) and Aurstad et al. (2005). South Korea KR
Europe Austria AT
Denmark DK
2 SURVEY
Finland FI
Germany DE
This study is based on a questionnaire sent to
Iceland IS
selected experts from the World Road Organization
Netherlands NL
(PIARC), European Committee for Standardiza-
Norway NO
tion (CEN) and Nordic Road Association (NVF). Slovakia SK
The survey consisted of three main subjects; Slovenia SI
Pavement structures, Quality requirements, and Spain ES
Import and export of aggregates. Sweden SE
The participants were asked to outline a normal Switzerland CH
pavement structure in their region for a two lane United Kingdom* UK
road with Annual Average Daily Traffic (AADT) North America Canada—Ontario CA-ON
of 10 000 vehicles. The information should include Canada—Quebec CA-QC
layer thicknesses for all individual layers (bound USA US
and unbound materials).
Furthermore, the participants were asked to give *The survey reply represents England and Wales.
information about the aggregate sizes used in the dif-
ferent layers of a pavement structure in their region. The two Canadian provinces Ontario and Que-
The participants were also asked to refer the bec are considered as separate countries in this
specific quality requirements regulating the choice paper, as they have individual pavement design
of aggregates in their region, including both stand- guidelines and requirements, and have provided
ardized and non-standardized national or regional separate answers to the survey. The participants
requirements. in the study represent national or regional road
Information about the use of natural and authorities, research institutes, or contractors.
crushed aggregates was also enquired, including The level of detail of the result presentation
any regulations regarding the use of crushed vs. is reliant on the level of detail in the information
natural (sand and gravel) aggregates. received from respondents. The information pre-
A question regarding the utilization of local sented should not be regarded exhaustive.
aggregates and efforts to ensure such use was also
included in the questionnaire. The term local aggre-
3.1 Pavement structures
gates can be interpreted as (lower-quality) locally
available aggregates, but also in-place reuse of road 3.1.1 Pavement thickness
materials in rehabilitation projects. Example pavement structures were gathered from
Finally, information about import of aggregates 16 countries, all designed for a traffic level of
from other countries to each country/state was AADT ≈ 10 000.
enquired, including the use of such imported aggre- All 16 countries take sub grade conditions into
gates. Corresponding information was also requested account when deciding the full pavement thick-
about export of aggregates from the same region. ness. 12 countries mention the importance of frost
in the pavement design procedures.
The pavement thicknesses for the 16 different
3 RESULTS pavement designs are illustrated in Figure 1.
Most countries provide a pavement thickness
The survey was answered by 17 participants, rep- span with a minimum and a maximum value.
resenting 16 different countries in Europe, North The minimum total pavement thickness for each
America and Asia. 13 out of 17 answers represented country ranges from 220 mm to 1220 mm, with an
European countries, all of which involved in the average of 576 mm. The maximum total pavement
European Committee for Standardization, CEN. thickness for each country ranges from 450 mm to
Including answers representing Norway pro- 2400 mm, with an average of 1116 mm.
vided by the authors, the results presented in this The difference between minimum and maxi-
paper represent 18 individual geographic regions. mum pavement thickness for a country displays
An overview of the participants is given in Table 1. how the pavement design can be adapted to dif-
The abbreviations in Table 1 are used when results ferent climatic, sub grade and drainage conditions.
are presented in figures. The difference between minimum and maximum
564
 s·
~
•

0..
~
~-

to
r.n.
;·~
D

D
3

::il
E;
(")

(")

(")
(1)

(1)

(1)
~

s::
0

0
.j>.
Vl
8 8 8 8 8
N

0
Pavement thickness [mm]

Upper sieve size (mm]

N0 ~~~~:;::;=;:;::;::;=;:;:~:;:;:1

Fi t~~~~~
N
0

(;J
SE ~~=i==i==

0
,_.
Vl

N
0
JP ~
.....,..
0

0
Vl
0

UK
SK

SK
CA-ON
DK

DK
UK

NL

NL

JP
SE
SI

SI
CA-QC

DE
CA-ON

FI
CH

CH
AT

AT

NO
DE

ES

ES
CA-QC

us
IS

IS
Figure 1. Minimum and maximum total pavement Figure 3. Upper sieve size limits for materials in surface
thicknesses for AADT ≈ 10 000. and binder layers.

s·
•
~

"
D
a

I>)
•

::::::
5.
t:d

~
D

3
::il

"0
(1)

(1)
~
s::
0

.,J::;a.
0
0000000
00000000
N~O'\ooON~

Upper sieve size [mm]

II

!
Pavement thickness [mm]

SE m m

I
w
0

I
I

I
'I
CA-QC m r l

".ll_l ll

I
N
0
I
I

SK ~
__________L__,
I

,_.

~Jl1
0

CH r - -
DK 1--

r----
[------
0

UK

NL
SE
CA-ON

FI

DE
AT

NO

ES
us
UK
SK
CA-ON

DK

NO

NL
SI

FI
CH

DE
JP

AT
ES
IS

Figure 4. Upper sieve size limits for bound base layer

Figure 2. Bound layer thickness and full pavement materials.
thicknesses for AADT ≈ 10 000.

Maximum allowed upper sieve size for HMA in

pavement thickness for each region varies from surface and binder courses can be seen in Figure 3.
zero (Austria and the Netherlands) up to 1890 mm These numbers form the upper aggregate size
(Norway). boundary for a range of material fractions used
All 16 structures outlined for the example in HMA materials in the uppermost layers of the
are flexible pavement structures using Hot Mix pavement structure.
Asphalt (HMA) as the topmost material. Asphalt For the surface HMA, the maximum size speci-
Concrete (AC) and stone Mastic Asphalt (SMA) fied is 22 mm (Japan and Finland), while 8 of 16
are the commonly used HMA types. countries specifies 16 mm as their maximum size
The division between bound (asphalt) and for surface materials. Five countries specify an
unbound materials for each structure is displayed allowed upper sieve size for surface HMA in the
in Figure 2. The minimum full pavement thickness range 9.5–11 mm.
from Figure 1 is used for reference to full pavement Eight countries do not distinguish between aggre-
thickness. The bound layer thickness varies from 95 gate sizes used in surface and binder layers, while
to 250 mm, with an average of 178 mm. The bound Austria presents the biggest difference in maximum
layer thickness compared to full pavement thickness sizes for surface and binder layer material, 21 mm.
ranges from 14% (Finland) to 100% (Germany). For HMA used in base layers, data was pro-
vided from 13 countries, as presented in Figure 4.
3.1.2 Aggregate sizes 10 of 13 countries report a maximum grain size of
The results regarding aggregate sizes are focused 32 (or 31.5) mm.
on upper limits for aggregate sizes in the different Nine countries also supplied requirements for
levels o f the pavement structure. 17 respond- the lowest allowed upper sieve size for a bound base
ents provided information about aggregate sizes in material. These figures span from 11 to 22 mm.
bound and unbound materials.
3.1.2.2 Unbound materials
3.1.2.1 Bound materials 11 countries gave information about aggregate
For bound materials, the size fractions vary from sizes used in unbound base layer materials, see
an open and fine graded 2/6 mm for porous asphalt Figure 5. The maximum allowed aggregate sizes
surface material to the coarse 0/32 mm fraction for span from 19 to 63 mm. For the six countries that
asphalt concrete base layer ma terials. additionally specified a minimum upper sieve size,
565
 DMin • Max Table 2. Most common quality parameters for aggre-
gates in bound materials.

Number of
Quality parameter countries

Grain shape (e.g. flakiness index, shape 15

index)
Resistance to fragmentation (e.g. 14
Los Angeles)
Percentage of crushed and broken 13
surfaces
Figure 5. Upper sieve size limits for unbound base Polished Stone Value 8
materials. Resistance to wear (MicroDeval, Nordic 8
Ball mill)
DMin • Max Resistance to freezing and thawing 7

60
50
;.<
~ 40
.s
5 30
=
~ 20 r-
~

Figure 6. Upper sieve size limits for subbase materials.

10

0
[i ~ - f- - f- - f- f-

IS SE FI DK SI ES NL DE CH NO
those values are in the range 19 to 32 mm. The
Figure 7. Upper Flakiness Index (FI) limits for aggre-
most common upper limit for aggregate size in the gates used in bitumen bound materials.
base layer is 32 mm, although Norway and Sweden
set 32 mm as their lower limit.
Information about material sizes in unbound sub- higher FI indicates a larger amount of elongated
base materials was provided from 17 countries. Fig- particles.
ure 6 displays the upper sieve sizes for aggregates used For most countries, the requirements in Fig-
in subbase materials. The maximum aggregate sizes ure 7 are valid for surface, binder and base mate-
for subbase span from 38 to 600 mm. Five regions rials. The strictest requirements apply for surface
allow aggregates larger than 100 mm in the subbase. layers, while the highest limit usually applies for
Norway and Finland use the coarsest materials, with base layer materials.
upper limits of 300 mm and 600 mm, respectively. In Denmark, FI values are only required for
Seven countries refer specifications where maxi- open-graded surface materials, while the values for
mum grain size is proportionately linked to layer Iceland are only valid for surface and binder lay-
thickness for the unbound materials. ers, as HMA base layers are not used in Iceland.
The strictest requirements are FI ≤ 15, specified by
three countries. The most common lower limit is
3.2 Quality requirements for aggregates FI ≤ 20, used in five countries.
3.2.1 Bound materials
The survey gave information on quality require- 3.2.1.2 Resistance to fragmentation
ments for bitumen bound aggregates for 15 coun- The Los Angeles abrasion test (LA) is another
tries. The most common quality parameters are important quality parameter for aggregates.
listed in Table 2. The LA test is described in EN 1097-2 (CEN
From these results, the three most important 2010), ASTM C131/C131M-14 (ASTM Inter-
parameters are grain shape, resistance to frag- national 2006), and other equivalent standards.
mentation, and percentage of crushed and broken LA requirements for bitumen bound materi-
surfaces. als from 14 countries are given in Figure 8. The
required limits are in a wide range, from LA ≤ 15
3.2.1.1 Grain shape to LA ≤ 50.
Of the 15 countries that use requirements for grain 11 of 14 countries report a range of LA require-
shape, 10 provide specific limits for the Flakiness ments. Countries differentiate requirements
Index (FI), as shown in Figure 7. The FI values are according to several factors, e.g. traffic level, mate-
tested according to EN 933-3 (CEN 2012), where a rial types, and layer (surface, binder, base).
566
 0000000
-NW.,J:::.VlO\

-tvW.,J:::.VlO\
0
Los Angeles abrasion loss

Los Angeles abrasion loss

0
0

[%]
[%]
0

I
I

I
,------'-------
0

I
I

I
0
0

SK
DK
UK
NL

JP
SE

SI
DE
CH

CA-QC
NO
AT

ES
IS

UK
SE
SI

CH
CA-QC
NL

AT
NO
IS
Figure 8. Upper limits for Los Angeles abrasion loss Figure 10. Upper limits for Los Angeles abrasion loss
for aggregates used in bitumen bound materials. for unbound materials.
00000000000
-NW.,J::::..VlO\-...l00\.00

I
I
I
Amount of crushed or

For unbound materials, the quality parameter

broken particles [%]

best suited for comparison is the resistance to

fragmentation measured by the Los Angeles abra-
I

sion test (LA), as nine countries provided require-

IIIII I

ments for this value. LA requirements are given in

Figure 10. Some countries specify a range of LA
NO r---"-'-1
I

requirements; the range is partly due to different

requirements for fine and coarse aggregates, partly
SI
DK

FI

CA-ON

NL
SE
CH

DE
CA-QC
IS

ES

differentiation for traffic volume.

For countries that do not use unbound materials
in base layers, the LA requirements in Figure 10
Figure 9. Requirements for minimum amount of represent subbase materials only.
crushed or broken particles in bitumen bound materials. Only two countries specifically mention that
geological parameters should be considered when
3.2.1.3 Crushed and broken surfaces deciding requirements for aggregate quality; Iceland
12 countries specify requirements for the amount considers petrography when setting limits for LA in
of crushed and broken surfaces for particles in unbound base layers, while Germany considers min-
bitumen bound materials, as shown in Figure 9. eralogy when setting limits for LA in bound base
A crushed or broken particle is defined in EN layers.
933-5 as a particle with more than 50% of its surface
crushed or broken (CEN 1998). 3.2.3 Natural vs. crushed aggregates
For this parameter, most countries provide mul- 17 countries describe requirements for a minimum of
tiple requirements, many with a wide range. These crushed faces for natural aggregate particles. The use
ranges account for variations in HMA types, with of uncrushed natural aggregates is unusual, except in
stricter requirements in surface layers than lower frost protection materials and embankments.
layers in the pavement structure. The numbers Several countries report on a decreasing use
presented in Figure 9 represents the minimum of natural aggregates over the later years, partly
required amount of crushed or broken particles. caused by limited availability and environmen-
What Figure 9 illustrates is that almost all tal concerns. Denmark however, describes a
countries have some materials where they require near exclusive use of natural aggregates, supple-
95–100% of crushed or broken particles, while most mented by requirements for amount of crushed
countries also use mass types where more rounded particles.
materials can be utilized. 8 of 13 countries specify
a lower limit for crushed and broken particles at 3.2.4 Utilization of local aggregates
50%. The requirements from The Netherlands are None of the respondents mentions any efforts to
the strictest, requiring 95–100% of crushed or bro- utilize lower quality locally available aggregates.
ken particles in bitumen bound materials. Two countries point out that using local aggregates
has hauling/logistical benefits that also can pro-
3.2.2 Unbound materials vide an economic advantage, and that this is the
14 countries provide information on quality primary motivation to utilize local aggregates.
requirements for unbound aggregates, of which 11 Recycling and reuse are mentioned by five coun-
describe specific requirements for physical proper- tries as a response to the question about utilization
ties. Three countries do not set requirements for of local aggregates. The answers include several
any physical properties of unbound aggregates. kinds of reuse or recycling, e.g.:
567
– Reuse of existing road materials in rehabilitation 4 DISCUSSION
projects
– Recycling of Hot Mix Asphalt (HMA) in new 4.1 Pavement design
HMA production The values for minimum pavement thickness in
– Reuse of crushed concrete as unbound material Figure 1 represent optimal conditions, in which no
or as aggregate in production of new concrete extra efforts are necessary to manage frost, drain-
For the rehabilitation projects, especially remix- age, and sub grade challenges. The wide variation
ing of HMA and base materials into base or sub- between the countries, from 220 to 1220 mm, can be
base materials are mentioned as a common practice. explained as differences in pavement design systems
but also differences in the quality and availability of
road construction materials among the countries.
3.3 Import and export of aggregates The varying need for frost protection and other pro-
All 18 countries taking part in the survey pro- tective measures connected to poor sub grade con-
vide information about import of aggregates; the ditions account for the differences between specified
results are summarized in Table 3. minimum and maximum pavement thickness.
Ten countries import aggregates. Some countries Four countries stand out having significantly
have an overall shortage of aggregate resources, higher maximum pavement thickness than the rest
while others have regional differences in aggre- (Fig. 1). These are all regions heavily affected by
gate availability within their country, where import frost action. Thus, the maximum pavement thick-
gives lower hauling costs than non-border crossing ness can be assumed caused by a combined effect
transport. of pavement design for poor sub grade conditions
In The Netherlands, imported aggregates are and large frost volumes.
used in all layers in the pavement structure, while From basic pavement design theory, it could be
Denmark, Iceland, and Slovenia import aggregates assumed that a pavement structure with thicker
to fulfil quality requirements for surface layers. bound layers will have lower total pavement thick-
Eight countries report no shortage of aggregates, ness due to the better load distribution capacity
and thereby no import. of bound materials. However, from the data in
Table 3 also shows the eight countries that provide Figure 2, no clear correlation can be found between
information regarding export of aggregates. The the full pavement thickness and the percentage of
replies from Iceland and Finland specifies that the the thickness being constructed by bound materials.
exported aggregates are not used for road construc- As no preconditions were given regarding sub-
tion purposes. Four respondents distinctly stated that grade conditions in the questionnaire, it must be
there was no export of aggregates from their country. assumed that some replies are valid for a specific
(normal) subgrade condition in their region, while
others have provided a variation range for the thick-
Table 3. Overview over import and export status. ness depending on subgrade conditions. This will
have affected the specified pavement thicknesses.
No No Traffic volume is another main factor influencing
Region Import import Export export the pavement design, besides subgrade conditions.
Additionally, the data does not provide informa-
Austria X
tion on pavement design life or pavement design
Canada—Ontario X X
methods, which must be assumed to vary among
Canada—Quebec X X
the 16 countries.
Denmark X X
Finland X X
Germany X X 4.2 Aggregate size
Iceland X X The variation in aggregate sizes is most evident in
Japan X the subbase materials, where the largest maximum
Netherlands X X upper sieve size (Finland) is 15 times bigger than
Norway X X the smallest maximum upper sieve size (US, Japan,
Slovakia X X Netherlands, Fig. 6).
Slovenia X X For surface and binder layers, the correspond-
South Korea X ing size relation is 2.3 and 3.4, respectively (Fig. 3).
Spain X X In base materials, the same relation is 1.7 for bitu-
Sweden X
men bound materials (Fig. 4) and 3.3 for unbound
Switzerland X X
materials (Fig. 5).
United Kingdom X X
For most countries, the subbase layer is the
USA X
largest constituent of the full pavement thickness.

568
Comparing full pavement thickness (Fig. 1) and 4.4 Aggregate accessibility and
maximum aggregate size allowed in subbase international trade
(Fig. 6), we see that the four regions allowing the
The use of natural uncrushed aggregates compared
thickest pavements are also among the five coun-
to crushed aggregates are clearly related to the
tries that allow the coarsest unbound materials.
availability of the different kinds of aggregates in
This corresponds well with the information given
the regions. Countries reporting a need for import
on requirements linking maximum grain size pro-
of aggregates are also the countries that are accept-
portionately to layer thickness.
ing a wider use of natural aggregates.
There is not a clear tendency that the countries
4.3 Quality requirements with the most difficult aggregate accessibility use
less aggregates in their pavement structures. As
A more differentiated quality requirement system
an example, both Denmark and the Netherlands
will enable better utilization of aggregates, as it
describe shortage of good quality aggregates, while
allows for choosing different qualities for different
they are not among the countries using the thin-
purposes, permitting lower quality for less demand-
nest pavement structures. This shows how import
ing purposes. This means using the appropriate
of aggregates is crucial for the road construction
quality for each purpose, not using the best avail-
business in many countries.
able aggregate at all times. As aggregates are a non-
Although the practice for aggregate use is diver-
renewable natural resource, this mindset should be
gent among the 18 studied countries, the quality
used to achieve better utilization of aggregates.
requirements meet equivalent standards issued
The results show that the common practice is
mainly by CEN and ASTM International. The
to set a firm limit for the measured material prop-
standards describe test methods and product cat-
erties, without differentiating the requirements
egories, rather than setting specific requirements
concerning geological properties. The ASTM
for the test results. This way, the countries can
Standard for the LA test does however point out
decide their quality requirements according to the
that test results from different sources should not
national practice, areas of application, and availa-
be compared (ASTM International 2006). This
bility of materials and resources. At the same time,
reservation indicates that a mechanical test like LA
the use of standards allow comparison between
should be complemented by information about the
countries and facilitates international cooperation.
composition of the aggregates before it can be used
Data from the annual report of the European
to decide aggregate choice for a project.
Aggregates Association (2016) and the results in
Considering sustainability through aggregate
Chapter 3.3 show that the aggregate market is inter-
utilization, functional tests could be used for the
national. The information gathered clearly point
purpose of successfully utilizing what today is con-
out that international aggregate trade is a prerequi-
sidered borderline or rejected aggregate resources
site for the ability of several countries to construct
for certain applications. Current research is work-
roads of satisfactory quality. The use of standards
ing towards functional requirements for road con-
enables an efficient international aggregate market.
struction materials, rather than the traditional
In light of the results of this study, the varying
physical and geometrical tests. An approach to
quality demands for aggregates can be viewed as an
functional testing of unbound granular material is
opportunity for aggregate utilization, both nation-
given in the work by Arnold et al. (2003), where
ally and in a regional international perspective.
a repeated tri axial test is applied in pavement
design using unbound aggregates that fail material
specifications.
5 CONCLUSIONS
Developments in this direction, where laboratory
testing and stress calculations are used to obtain
The results presented have shown that there are
a pavement design where the focus is to design a
large variations in practice for aggregate use inter-
pavement that can use locally available aggregates
nationally. The selection of countries is sufficient
could increase overall aggregate utilization.
to illustrate a variety of pavement designs, aggre-
The results of this survey do however show that
gate sizes, and quality requirements, in addition to
the use of traditional quality assessment methods
enlightening use of natural vs. crushed aggregates
is still dominating. None of the countries repre-
and international trade of aggregates.
sented in this survey mention functional quality
The following points summarize the conclu-
requirements specifically, and all countries that pro-
sions obtained from the results and discussion:
vided specific information on quality requirements
still use traditional tests like grain shape, amount – No clear relationship was found between bound
of crushed and broken surfaces, and resistance to layer thickness and full pavement thickness for
fragmentation. 16 compared pavement structures.

569
– Countries affected by frost use the largest pave- road construction. The study is part of the industrial
ment thicknesses. innovation research project Use of local materials,
– Pavement design regulations differentiated for funded by the Norwegian Research Council. This
traffic and other parameters can allow for better project is dealing with utilization of rock materials
utilization of aggregate resources. from tunnels and other construction sites.
– For subbase aggregates, sizes differ by a factor of The authors are very grateful for the contribu-
15 among the 17 countries. Countries in North- tion from each expert who replied to the survey.
ern Europe and Canada allow and use signifi- The main author is responsible for any misconcep-
cantly coarser materials in subbase materials than tions of the received information.
other countries. There is a connection between
total pavement thickness and allowed aggregate
size in subbase materials; countries that design REFERENCES
thick pavements use coarse subbase materials.
– Functional aggregate testing, e.g. tri axial tests Arnold, G., Hughes, D., Dawson, A. & Robinson, D.
could increase aggregate utilization. The par- (2003) Design of Granular Pavements. Transporta-
ticipating countries have not yet implemented tion Research Record: Journal of the Transportation
functional requirements for pavement perform- Research Board, 1819, 194–200.
ASTM International (2006) ASTM C131/C131M-14
ance; traditional quality tests are still dominat- Standard Test Method for Resistance to Degradation of
ing aggregate choice. Small-Size Coarse Aggregate by Abrasion and Impact
– Quality requirements for aggregates do rarely in the Los Angeles Machine, West Conshohocken, PA,
take the geological composition of aggregates ASTM International.
into account, although the literature point to ASTM International (2015) ASTM D2940/D2940M - 15
geological composition as an important factor Standard Specification for Graded Aggregate Material
in the interpretation of test results. For Bases or Subbases for Highways or Airports, West
– Crushed aggregates are the standard materials Conshohocken, PA, ASTM International.
used in road construction. Use of natural aggre- Aurstad, J. & Hoff, I. (2002) Crushed asphalt and con-
crete as unbound road materials—Comparisons of
gate is diminishing due to reduced availability field measurements and laboratory investigations. 6th
and environmental concerns. International Conference on the Bearing Capacity of
– International aggregate trade is widespread, and Roads, Railways and Airfields. Lisboa.
for certain countries necessary. Although aggre- Aurstad, J., Aksnes, J., Dahlhaug, J. E., Berntsen, G. &
gate trade often is associated with long transport Uthus, N. (2005) Unbound crushed concrete in high
distances, several countries use foreign aggre- volume roads—A field and laboratory study. 7th
gates to reduce transportation. International Conference on the Bearing Capacity of
– Despite divergent practice for aggregate use, qual- Roads, Railways and Airfields. Trondheim.
ity requirements in the studied countries meet CEN (1998) EN 933-5:1998 Tests for geometrical prop-
erties of aggregates—Part 5: Determination of per-
equivalent standards. The use of international centage of crushed and broken surfaces in coarse
standards for product specifications and quality aggregate particles, Brussels, European Committee for
tests aids the international aggregate market. Standardization.
– The results of this study show that the varying CEN (2009) EN 13242:2002+A1:2007 Aggregates for
quality demands for aggregates can be viewed unbound and hydraulically bound materials for use in
as an opportunity for aggregate utilization, civil engineering work and road construction.
both nationally and in a regional international CEN (2010) EN 1097-2:2010 Tests for mechanical and
perspective. physical properties of aggregates—Part 2: Methods for
the determination of resistance to fragmentation.
CEN (2012) EN 933-3:2012 Tests for geometrical proper-
ties of aggregates—Part 3: Determination of particle
shape—Flakiness index, Brussels, European Commit-
ACKNOWLEDGEMENTS tee for Standardization.
European Aggregates Association (2016) Annual Review
This paper is part of a PhD-study at NTNU, named 2015–2016. Brussels, European Aggregates Associa-
Optimal utilization of unbound crushed aggregates in tion (UEPG).

570
In-situ measurements techniques and monitoring
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Influences of measurement conditions on structural indicators obtained

from FWD data

C. Van Geem
Belgian Road Research Centre, Brussels, Belgium

ABSTRACT: In a previous publication new structural indicators (KPI1, KPI2 and KPI3) for the evalu-
ation of a primary road network were developed and applied, based on deflection measurements. These
measurements were all executed under very similar conditions and on road structures with similar thick-
nesses for the layers above the base course. Here we present observations made on the variation of the
indicator values when varying the drop height of the FWD load and in case of different temperature
conditions. For the Tragfähigkeitszahl (Tz) we observed a linear variation with respect to load charges.
Temperature does influence Tz but the categorisation with KPI1 based on Tz is not much affected. Indi-
cator KPI2evaluating bonding in the upper layers showed higher values in summer. For indicator KPI3
expressing the coherence of the road structure, we propose an alternative definition when monitoring
road structures with smaller thicknesses as on a secondary road network.

1 INTRODUCTION on an entire road network. Instead, in (Van Geem

et al. 2015) a set of structural indicators referred to
Worldwide, major efforts have been made for mak- as KPI1, KPI2 and KPI3 were defined and applied
ing deflection measurements obtained with FWD’s to the primary road network of Wallonia, the south-
repeatable and reproducible, including compara- ern region of Belgium. Values for these indicators
tive tests with a fleet of FWD under identical prob- can be computed directly from measurements with
ing conditions on real roads and the determination a Falling Weight Deflectometer (FWD). In the defi-
of correlation factors between the participating nition of KPI1 we used the Tragfähigkeitszahl (Tz),
devices. A description of such a procedure was defined in (Jendia 1995).
given in the final report of European COST action On the primary road network of Wallonia the
336 (Van Gurp et al., 2005), regularly used in The FWD was used on rigid and semi-rigid road struc-
Netherlands as reported in (Koster 2015). The tures, most of the time on concrete roads, and the
method was also used in the comparative trial in applied load charge level was 100 kN. The meas-
the SPENS project (Stryk & Erjavec 2008), (Spiel- urements were concentrated in autumn and spring,
hofer 2009). A correlation trial is performed annu- sometimes at night, so that the temperature at the
ally in the UK: see (Brittain 2010). This makes the road surface always stayed within the same, restricted
FWD a reliable measurement device dedicated to interval.
the evaluation of bearing capacity and the struc- In the summer of 2015, at higher temperatures,
tural health of existing road structures. we applied the same indicators on a secondary road
For a network analysis of the structural perform- network where the thicknesses of the different lay-
ance of the roads in the network, deflection measure- ers in the road structure were lesser than on the pri-
ments can be transformed into values for structural mary road network of Wallonia. Therefore we also
indicators. In the final report (Litzka 2008) of COST reduced the applied load level to 50 kN. Rather than
action 354 a structural indicator based on bearing blindly applying the same indicators, we investigated
capacity, possibly combined with an evaluation of what factors could have an influence on the values
rutting, longitudinal evenness and cracking, was sug- of these indicators when applied to a secondary road
gested. In (Crispino et al., 2005) and (Gaborit et al., network under different conditions during the FWD
2012) for instance, deflection measurements are used measurements. The sensitivity of the different indi-
at network level. However, for estimating the bearing cators in function of the applied load was verifie-
capacity by use of back-calculation detailed infor- don a test track in the premises of the Belgian Road
mation is needed on the materials and thicknesses of Research Centre (BRRC). There we also investigated
the different layers in the road structure. Estimating the influence of layer thicknesses on KPI3, in par-
residual life expectance asks for further information ticular for roads with a thinner layer structure.
on traffic: the traffic of the past and an estimate The application of these indicators on another
of future traffic. All these data are rarely available type of road network also implied a review of the
573
thresholds that could possibly be used for classifica- over all measured deflections. Initial values in the
tion of the road sections of a road network in func- process are computed from D(0), D(600), D(900)
tion of their bearing capacity and structural health. and D(1500): these must be available when the rec-
In Section 2 we recall the definitions of Tz, KPI1, ommendations in (Van Gurp et al., 2005) are fol-
KPI2 and KPI3 given in (Van Geem et al., 2015). lowed during FWD measurements. The computed
In Section 3 we present the test track in the BRRC radius is accepted (or said “of good quality”) if
premises, used for the sensitivity study of the indi- the difference between measured and computed
cators in function of layer thicknesses and applied value for D(300) is smaller than 3,5% of the meas-
force by the FWD. We discuss the effect of layer ured value. Since on the primary road network of
thickness on KPI3 and propose an alternative defi- Wallonia we were confronted with stiff road struc-
nition of KPI3 for secondary roads in Section 4. tures, we had to apply a force of 100 kN in order
In Section 5 we discuss the effect of the applied to get deflections larger than 20 μm on the outer
load force on the indicators. We determined a lin- geophones, a minimum deflection recommended
ear relationship between Tz and the applied load. in (Van Gurp et al., 2005). We used Tz obtained
This immediately had consequences for the thresh- with a force of 100 kN in the definition of KPI1.
olds on the indicators. In Section 6 we discuss the For KPI1 we first use a criterion based upon the
importance of the choice of thresholds in function product R0.D(0). For each homogeneous road seg-
of the type of road network when classifying the ment containing at least 5R0 values of good qual-
road segments in the network according to their ity, we compute RA:
structural health as evaluated with the indicators.
In Section 7 we present our first observations on σ (R 0.D ( ))
RA = (4)
the influence of the temperature during the FWD ave (R 0.D ( ))
measurements on the values of the indicators.
with ave(R0.D(0)) the average of product R0.D(0)
on the segment and σ(R0.D(0)) its standard devia-
2 STRUCTURAL INDICATORS
tion. For the other segments we set RA to 0. When
For completeness, we repeat the definitions of Tz, ( .25) the structural performance is consid-
KPI1, KPI2 and KPI3. In the following, the deflec- ered “bad” since an almost constant value for prod-
tion measured by the geophone at x millimeters uct R0.D(0) is expected in case of a healthy road
from the center of impact will be denoted by D(x). structure. In this case KPI1 is defined as:
Deflections are expressed in μm. We write gx to
denote the geophone at x millimeters from the center KPI1 i (5 1 75 5 RA) . (5)
of impact (e.g. g600 for the geophone at 600 mm).
In (Jendia 1995) Tz was defined as:
When (RA ≤ ) and if ( ( .D (0)) <
LIM), where LIM = 150,000, KPI1 is defined as:
Tz = R 0 / D ( ) (1) KPI1 i (5; 3 LIM / ave( 0.D (0)). (6)
where D(0) is the maximal deflection measured The choice of the value of LIM was done in
by the FWD and R0 is the radius of curvature function of the observations on rigid and semi-
(expressed in meters) determined from the deflec- rigid road sections in the primary road network of
tion curve at D(0). In (Jendia 1995) Tz is proposed Wallonia.
as indicator for the structural evaluation of bitumi- With this definition a low value of KPI1 should
nous roads. The applied force with the FWD was indicate good bearing capacity while values higher
set to 50 kN and critical values for Tz were deter- than 4 should indicate road structures at their
mined which could be used as thresholds for pave- end-of-life.
ment management. A computation of the radius From all FWD data stations in a homogeneous
R0 is proposed in (Jendia 1995) but, as mentioned road section we compute the value L defined by:
in (Van Geem et al., 2015), we use another, itera-
tive approach with the theoretical model f(x) of a L (
median ( D ( )− D( )) D( )) . (7)
deflection bowl expressed by formula (2) for which
the radius than can be computed with equation (3). We define KPI2 as:
a0 + a1x 2
if ( L ) : KPI
K 2 = max ( + L), (8)
f (x) = (2)
1 + a2 x 2
otherwise : KPI
K 2 m i (5; 2 1.3.L
min
mi L), (9)
1
R0 = (3)
2. (a1 a0 a2 ) Low values for KPI2 may indicate poor bonding
in the upper part of the road structure.
We set a0 = D(0) and we look for a1 and a2 such The response of each geophone gx and of the
that i ( f i) D(i ))2 is minimal when summed force sensor of the FWD is stored over a period

574
of 60 ms, which gives the load-displacement dia- is larger than 2 times the number of points NP in
gram for each geophone. Surface E(gx) inscribed the segment, then
in the load-displacement curve represents energy
(expressed in J = N.m). Also slope S(gx) of the line KPI 3 2 2 ( NTS
N 4) / NP, (12)
through the origin and the point with the maxi- (a value between 3 and 4), otherwise, if the number
mal deflection recorded by that geophone is used of jumps in the “lower part” NINF (energy and
in the definition of KPI3. The load-displacement slope counted together) in the same homogeneous
diagram and the line defining slope S(gx) are illus- segment is at least 1, then
trated in Figure 1. We define KPI3 as follows. When KPI 3 2 2 ( NIN
N NF 4) / NP N , (13)
all energies E(gx) have an absolute value less than 1:
(a value between 2 and 3), and otherwise, denote
KPI 3 (
ave E ( g0
g ) , ) (10) the number of jumps in the “upper part” (energy
and slope counted together) in the same homoge-
averaging over all points in the homogeneous seg- neous segment by NSUP and
ment, |X| denoting the absolute value of X. When
all energies E(gx) are larger than 15: KPI 3 1 2 ( NSU
N UP 4) / NP
N , (14)

I3 (
i 5 3 ave ( E ( g0
g )) 15 , ) (11) (a number between 1 and 2).

averaging over all points in the homogeneous seg- 3 TEST TRACK

ment. In the other cases, “jumps” in energies and
slopes are computed. For this, only the geophones In order to investigate the effect of layer thicknesses
at 0, 900, and 2100 mm distance from the centre on KPI3 and the applied force by the FWD load on
of force impact are considered. When the ratios the indicators, we made use of the test track that was
between consecutive energies and slopes are great, constructed in the frame of a research project on
this may indicate a lack of cohesion in the “upper GPR: see (Van der Wielen et al., 2015). The test site
part” or in the “lower part” of the road structure: is a stretch of 23 m long and 2.7 m wide, inside one
If ||E(g0) – E(g900)| – |E(g0)/2|| > 0.2.|E(g0)|, of the buildings of BRRC, representing 4 different
then count 1 jump for the energy in the “upper road structures that are commonly used in Belgium:
part”,
If ||S(g0) – S(g900)| – |S(g0) / 2|| > 0.25. – asphalt covering cobblestones (section 1),
| S(g0)|, then count 1 jump for the slope in the – asphalt layer on concrete plates (section 2),
“upper part”, – asphalt layer on lean concrete (section 3), and
If ||E(g900) – E(g2100)| – |E(g900) / 2|| > 0.2. – concrete layer on lean concrete separated by a
| E(g900)|, then count 1 jump for the energy in the thin asphalt layer (section 4).
“lower part”, The asphalt layer in section 1 has a thickness of
If ||S(g900) – S(g2100)| – |S(g900) / 2|| > 0.25. 4 cm; the cobblestones laid on a sand bed are about
| S(g900)|, then count 1 jump for the slope in the 12 to 14 cm high and rest on 18 cm of aggregates.
“lower part”. The asphalt layers in section 2 have a total thick-
If the total number of jumps NTS (energy and slope ness of 10 cm. They are laid on top of a concrete
counted together) in the same homogeneous segment slab with a thickness of 20 cm. Underneath the
concrete slab there is a layer of 20 cm of aggre-
gates. In section 3 the same asphalt layers with

:!
roo·
~
/'
-
? ~ ~- ~

/
,,"' r} total thickness of 10 cm are laid on top of a layer
of 20 cm of lean concrete and the same layer of
20 cm of aggregates. Section 4 is made of a 20 cm
'! - O(O)o'"rra"'

i I layer of concrete on top of a sandwich layer in

-v
- · - $1opelheO(O)

• asphalt with a thickness of 6 cm, a layer of 20 cm

/

~
of lean concrete and a layer of 20 cm of aggregates.
k»d !kNl

f#t·tIf*=---·
4 INFLUENCE OF LAYER THICKNESSES
- - slooo liov 0(900) When we applied the indicators developed for
~
the primary road network to secondary roads, we
k>od (kN) observed that the values for KPI3 were all very high.
This would have meant that all roads had a struc-
Figure 1. Load-displacement diagrams and associated tural weakness or failure in the base layer. Then we
line defining slope S(gx) for geophones g0 (top) and g900 noticed that the layer thicknesses on the secondary
(below). road network were significantly inferior to the usual

575
Table 1. Effect of layer thicknesses on the different defini- On 15 different stations, distributed over all 4
tions of KPI3. road structures, FWD measurements were per-
formed at different forces, by steadily increasing the
KPI3 KPI3
Section (structural road network) (secondary roads) drop height. The stiffer the road structure, the more
and the higher forces we applied. The measurements
1 4.578 4.578 on different stations but on the same road structure
2 2.333 3.500 were considered as being in the same homogeneous
3 2.167 3.500 segment. On each segment one value was computed
4 1.750 3.750 for each indicator and for Tz. The different average
values for Tz, KPI1, KPI2 and KPI3 are presented
in Tables 2 to 5 for the different sections. For sec-
tions 1 and 2, Tables 2 and 3 present KPI3 following
thicknesses on the primary road network. The idea the definition for secondary roads (using the geo-
behind the definition of KPI3 is that the deflections phones at 600 and 1800 mm away from the centre
registered by the geophone at 900 mm from the place of the load impact). For sections 3 and 4 the cor-
of impact are representative for the base course and responding Tables 4 and 5 present KPI3 as defined
that those of the geophone at 2100 mm are represent- for a primary road (using geophones at 900 and
ative for the subbase. This is a reasonable assumption 2100 mm distance). Each of these tables presents
for road structures of the primary road network but the indicator values obtained at different loads.
for a thinner road structure the selected geophones For Tz we observed a linear behaviour in func-
are too far away from the place of impact. Hence we tion of the applied force F. Hence, with equation
redefine KPI3 when probing a secondary road net- (3) we can transform Tz obtained with applied
work, by replacing the geophone at 900 mm from the
place of impact by the geophone at 600 mm and by Table 2. Indicator values for different loads on section 1.
replacing the geophone at 2100 mm by the geophone
at 1800 mm. In (Van Gurp et al., 2005) the positions F (kN) Tz KPI1 KPI2 KPI3
at 600 and 1800 mm are among those recommended
for FWD measurements on roads that present a cen- 30 0.188 5.000 2.482 2.750
tre deflection between 500 and 1000 μm at a load of 40 0.182 5.000 2.462 2.750
50 kN, so it will always be possible to compute KPI3 50 0.178 5.000 2.455 4.578
with this new definition for secondary roads. Note
also that nowadays FWD’s are equipped with 15 or Table 3. Indicator values for different loads on section 2.
more geophones. Hence there is no need for a physi-
cal repositioning of any of the geophones: it suffices F (kN) Tz KPI1 KPI2 KPI3
to position geophones at the distances needed in the
definitions of KPI3 for primary and secondary road 30 3.267 3.272 1.506 1.667
networks; positions where they are recommended to 40 3.404 3.206 1.443 2.167
be in (Van Gurp et al., 2005) anyway. 50 3.284 3.264 1.517 2.167
We used the test track to evaluate the effect of
the layer thicknesses on the different definitions for Table 4. Indicator values for different loads on section 3.
KPI3. The layer thicknesses are increasingly impor-
tant from section 1 toward section 4. The definition F (kN) Tz KPI1 KPI2 KPI3
of KPI3 for secondary roads should not be applied
to the road structure of sections 3 and 4. The defini- 30 3.423 3.197 2.034 2.292
tion of KPI3 for the primary road network should 40 2.958 3.423 2.072 2.208
50 2.851 3.474 2.079 2.208
not be applied to sections 1 and 2. In Table 1 we see
65 2.644 3.575 2.073 2.292
that there can be a significant difference between
the values for both definitions of KPI3.
Table 5. Indicator values for different loads on section 4.

F (kN) Tz KPI1 KPI2 KPI3

5 FORCE INFLUENCE
30 4.707 2.573 1.333 1.234
The higher the force applied to the road sur- 40 5.194 2.336 1.448 2.125
face by the falling weight, the higher the maxi- 50 5.137 2.364 1.338 1.875
mal deflection will be. We used the test track 65 4.889 2.484 1.503 1.875
80 4.941 2.459 1.539 1.875
mentioned in Section 3 for an evaluation of the
85 4.920 2.469 1.517 1.875
effect of the load on the different indicators. The
95 4.977 2.441 1.437 1.875
applied load is measured with the force sensor of 100 5.054 2.404 1.467 1.875
the FWD.

576
force F to a value for Tz at reference force Fref. In Table 6. Classification of secondary network using thresh-
(Jendia 1995) Fref is set to 50 kN, in our application olds for primary roads based on KPI1, class 1 = “end of life”.
in (Van Geem et al., 2015) Fref is set to 100 kN.
Percentage of kilometres
( )
Tz Freef T ( F ) F / Fref
Tz (3)
Class Range KPI1 of road segments

1 4.5 – 5.0 24
Since the sections of the test track are very 2 4.0 – 4.5 42
short, in this situation KPI1 is fully determined by 3 3.5 – 4.0 21
Tz, and so the value for KPI1 does not significantly 4 3.0 – 3.5 9
change with varying charges when Tz is normal- 5 2.0 – 3.0 3
ized to a reference force using equation (1). 6 0.0 – 2.0 1
On the test track we also did not find a large
variation of the values for KPI2 with varying
charges. The values for KPI3 however did vary. Table 7. Classification of secondary network using thresh-
olds inspired by (Jendia 1995), class 1 = “end of life”.
This can be explained by the nature of the defini-
tion of KPI3: since we only count the number of Percentage of kilometres
“jumps” in energy or slope determined from the Class Range Tz(100 kN) of road segments
load-displacement diagrams of three geophones, a
small difference in the number of “jumps” makes a 1 0.000 – 0.375 20
rather big difference in the value for KPI3. 2 0.375 – 0.500 4
3 0.500 – 1.750 48
4 1.750 – 3.500 22
6 THRESHOLDS FOR INDICATORS 5 more than 3.500 6

In (Jendia 1995) thresholds were determined for

Tz based on a series of FWD measurements with mated to 20 or even 24% (classes 1 and 2). This can be
reference load equal to 50 kN on main roads (other explained by the original design of the roads: prob-
than motorways) with bituminous road surface. ably insufficient for the large number of trucks it is
It is stated in (Jendia 1995) that Tz(50 kN) < 0.75 carrying now in an industrial area. However, another
almost always corresponds to a road with lack of parameter of influence on Tz is the temperature.
bearing capacity (for the traffic that makes use of
that road) and also that values between 0.75 and
1 sometimes indicate a lack of bearing capacity. 7 TEMPERATURE INFLUENCE
These thresholds are applicable to Tz(50 kN) but
they for Tz(100 kN) they must be divided by 2. It is well-known that under different temperature
In contrast, the interpretation for the primary conditions the obtained deflections will not be
road network in Wallonia, described in (Van Geem the same since the stiffness modulus of an asphalt
et al., 2015), worked with transformation functions layer is significantly affected by its temperature.
that implicitly introduced other thresholds inspired Some temperature correction formulas are devel-
by evaluations of residual life using back-calcula- oped and reported in the literature. In particular,
tions on a limited number of road sections. It was in (Jendia 1995) a transformation on the deflection
observed that for this primary road network (includ- measurements to a standard temperature of 20°C
ing motorways) values Tz(100 kN) larger than 3.5 is proposed before computing Tz.
(which corresponds to KPI1 < 3) were structurally Usually the transformation formulas are very
healthy. When Tz(100 kN)< 1.75 (which corresponds much dependent on the composition of the asphalt
to KPI1 > 4), the bearing capacity of the road based and on the layer thickness. We therefore chose not
on back-calculations seemed insufficient. to apply a correction formula to the deflections
When we apply the different thresholds to the sec- measured with the FWD. Instead, we did start an
ondary road network of bituminous roads on which investigation on the influence of temperature on
FWD measurements were performed in the summer Tz and the indicators KPI1, KPI2 and KPI3 in
of 2015, we get quite a different classification of order to estimate the sensitivity of these indicators
the road sections. A classification using KPI1 and to temperature conditions.
thresholds similar to those used on the primary road On three secondary road sections FWD meas-
network in Wallonia is given in Table 6. Another urements were repeated, first in July 2015 (with
classification using Tz and thresholds inspired by surface temperature at 22°C) and then in Decem-
those given in (Jendia 1995) is given in Table 7. ber 2015 (with surface temperature at 9°C). We
Surely the classification in Table 7 is the more observed a higher value for Tz in December, little
appropriate one for the secondary road network, variation on KPI1, somewhat lower values for KPI2
even if the percentage of kilometres of road seg- in December and almost no variation on the value
ments with insufficient bearing capacity is still esti- of KPI3 when defined with the above proposed var-

577
Table 8. Indicator values in summer. contribution we presented the results of our inves-
tigations on the influence on the indicator values
Section Tz KPI1 KPI2 KPI3 (str) KPI3 (sec) by the temperature during the measurements and
A 0.55 4.20 2.18 4.36 4.36 by the applied FWD load. Moreover, we observed
B 0.42 5.00 2.45 2.30 2.40 that the definition of one of the indicators (KPI3)
C 0.61 4.60 3.10 2.50 2.60 should be adapted when applied to secondary roads
with less important layer thicknesses. We also dis-
cussed the thresholds on the indicators, used for
Table 9. Indicator values in winter. classification of the road sections in a road network.
Section Tz KPI1 KPI2 KPI3 (str) KPI3 (sec)
When we applied these indicators on a secondary
road network, we noticed the need for a prudent
A 0.70 4.60 1.15 4.41 4.41 interpretation of the results when making the final
B 0.97 4.44 2.14 2.70 2.60 classification.
C 1.41 4.19 2.11 2.50 2.50

REFERENCES
iant for secondary roads. Table 8 presents the values
obtained in summer on three sections (A, B and C) Brittain, S., Highways Agency 2009 National Falling
for the average Tz (normalized to 100 kN), KPI1, Weight Deflectometer Correlation trials, TRL Report
KPI2, KPI3 as defined for a primary road network PPR437, ISBN 978-1-84608-935-0, January 2010.
and KPI3 as defined for a secondary road network. Crispino, M., Olivari, G., Poggiolo, M., Scazziga, I., Includ-
ing Bearing Capacity into a Pavement Management
Table 9 presents the values for the same indicators
System, International Conference on Bearing Capacity of
and the same sections based on measurements in Road Pavements, Trondheim, Norway, 2005.
winter. Gaborit, Ph., Di Benedetto, H., Sauzéat, C., Pouget,
This exercise on only three road sections is too S., Olard, F., Quivet, S., Analyse d’une structure de
limited for deriving any general conclusions but we chausséeautoroutière par auscultation in situ etes-
are comforted in the following hypotheses: saisenlaboratoire, Actes des 30ième Rencontres Uni-
versitaires de Génie Civil (AUGC et IBPSA 2012),
– temperature effects to Tz, KPI1 and KPI2 exist Chambéry, France, 6-8 June 2012 (in French).
but only to a limited extend; Jendia.S., Bewertung der Tragfähigkeit von bituminösen-
– Tz seems to be lower at higher temperature; Straßenbefestigungen. Veröffentlichungen des Institutes
– temperature does not have a significant effect on fürStraßen- und Eisenbahnwesen der Universität Karlsruhe
KPI3, especially for values in the higher range. (TH), Heft 45, 1995 (ISSN 0344-970-X, in German).
Koster, I.W., CROW Falling Weight Deflectometer Correla-
Also, the exercise shows that it is recommend- tion Trial 2015, CROWReport D15-03, November 2015.
able to monitor a network under similar tempera- Litzka, J., et al. The way forward for pavement perform-
ture conditions in order to be able to compare the ance indicators across Europe, Final report of COST
values of the indicators between the different road action 354, Performance Indicators for Road Pavements,
sections in the network. ISBN 978-3-200-01238-7, 2008.
Spielhofer, R. et al., Guidelines of a complex methodology
for non-destructive pavement measuring techniques,
8 CONCLUSIONS Deliverable D11 of the Sustainable Pavements for European
New Member States (SPENS) project, February 2009.
Whereas structural indicators based on back-calcu- Stryk, J., Erjavec S., Analysis of FWD Harmonisation
test. SPENS WP 2 report, CDV, 2008.
lation, elasticity moduli for layers in the road struc-
Van der Wielen, A., C. Van Geem, C. Grégoire, GPR on
ture and residual life expectance can be of great Roads: Layer Thickness Estimation, in Proceedings
interest on project level, they ask for detailed knowl- of the Third Action’s General Meeting, London, UK,
edge of the road structure and the local traffic. For March 4-6, 2015(COST Action TU1208 “Civil Engi-
the structural evaluation of a whole road network neering Applications of Ground Penetrating Radar”),
these details may simply not be available. Therefore, editors: L. Pajewski, A. Benedetto, A. Alani, S. Lambot;
it is our conviction that other indicators directly Publishing House: Aracne; Rome, Italy, ISBN 978-88-
based on deflection measurements obtained with an 548-8486-1, May 2015.
FWD could be used for network evaluation. Such Van Geem, C., Nigro, P, and Berlémont, B., The Use of
Deflection Measurements in Pavement Management of
indicators were defined in (Van Geem et al., 2015)
the Primary Road Network of Wallonia, Belgium, 9th
and used on the primary road network (including International Conference on Managing Pavement Assets
motorways) in Wallonia, Belgium. In that case, (ICMPA9), Alexandria, Virginia, USA, May 18–21, 2015.
the variation of temperature conditions during the Van Gurp, C.A.P.M. et al., COST 336, Use of Falling Weight
FWD measurements was deliberately limited and Deflectometers in Pavement Evaluation, Final Report of
the applied FWD load was always the same. In this the Action, Main Report, 2nd Edition, April 2005.

578
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

A comparison of TSD and FWD deflections at Norwegian roads with

an interpretation of bearing capacity from TSD measurements

V. Antonsen
The Norwegian Public Roads Administration, Lillehammer, Norway

H. Mork
NTNU, Trondheim, Norway

ABSTRACT: Norway has a long history of deflection measurements on the road network, including
Benkelman Beam (BB), Dynaflect (DF), LaCroix Deflectograph and single- and two-mass Falling Weight
Deflectometer (FWD). Except FWD, all these are merely discontinued. For all these equipment types,
measurements are performed point by point, except for LaCroix normally at a 50 m interval, and the
center deflection only (BB and LaCroix) or the center deflection with some measure of surface curvature
(DF, FWD) has been used with AADTH to calculate a bearing capacity value. Though intended for net-
work level measurements, the capacity of these types of equipment is limited, and the most relevant use
is for measurements at project level. Despite network measuring devices like the Traffic Speed Deflectom-
eter, TSD, has been around for several years, no measurements have been performed in Norway with this
type of equipment until fall 2015. Then, measurements were performed with a TSD from Poland on some
600 km at low and high volume roads in the south-eastern part of the country. At some of these roads,
FWD measurements were also carried out. For these concurrent sections, TSD and FWD measurements
are compared, and the BB and FWD methodology is used to calculate bearing capacity for the TSD
measurements, with astonishing similarity. The most promising methodology is then used to calculate
TSD bearing capacity for the sections without FWD measurements, and the values are compared to other
indicators of pavement performance like rutting and unevenness.

1 INTRODUCTION However, the capacity also of FWD is limited

when it comes to network level measurements, and
1.1 Background in the south-eastern part of Norway, systematic net-
work level deflection measurements have not been
Norway’s history of deflection measurements goes
carried out since the early 90’ies. The investigation
back to 1965, when Benkelman Beam (BB) was used
was then taken as point-by-point measurements by
on the entire national road network, based on which
DF at 50 meter intervals. Instead, the functional
a nationwide “Bearing Capacity (BC) Register” was
performance of the road network has been moni-
established. Due to changes in traffic and legal axle
tored on an annual basis since 1987, first based on
load limits, rehabilitation and the construction of
ultra-sound, since 2008 by laser-based equipment.
new roads, this register pretty soon was outdated,
The need for heavy maintenance is again rising
and a new round of nationwide deflection meas-
in order to keep the bearing capacity, accessibility
urements was carried out 1976. By then the first
and road safety at a reasonable level, especially for
Dynaflects (DF) were introduced, and from then
Norwegian county roads. There is therefore a need
on, annual measurements were carried out with
for an efficient tool for analyzing bearing capacity,
either DF, of which there eventually was at least one
prioritize and maximize the effect of maintenance
in each county except one, which operated a LaC-
efforts. The Traffic Speed Deflectometer (TSD)
roix Deflectograph. Around 1980 the first two Fall-
could be such a device.
ing Weight Deflectometers (FWD) were introduced,
but none of the counties’ Public Roads Administra-
tion operated a FWD until 1988. Eventually, there
1.2 Measurements with traffic speed
were four FWDs operated by NPRA. Two of these
deflectometer
machines were replaced by new ones 2006, after
NPRA was reorganized into five regions 2003. By For the first time in Norway, measurements with
then, all the previous equipment were discontinued. Traffic Speed Deflectometer were performed the

579
autumn 2015. This was part of a Research and Table 1. Applied force for FWD and TSD
Development project at the Norwegian Public measurements.
Roads Administration. The main goal of the R&D Fv 24 Hp 02 Force applied Force applied
project was to test the TSD for Norwegian condi- [m] FWD [kN] TSD [kN]
tions, and to check whether it is possible to calculate 3200 49,3 50,8
the Norwegian Bearing Capacity term from TSD 3250 49,7 46,2
3300 49,7 50,0
measurements.
3350
The TSD measurements were carried out on -- - - 49,3 - 55,2
3400 49,8 52,0
some 600 km of road in the county of Hedmark, 3450 48,8 53,7
and included both high volume national roads and 3500 48,6 53,8
low volume county roads. In addition, a few test 3550 49,3 47,5
sections were selected among these where FWD 3600 49,8 49,5
3650 49,6 49,4
measurements were taken. The goal of these test
3700 49,5 51,2
sections was primarily to compare FWD and TSD 3750 48,8 49,1
measurements, and to use the BC calculated from 3800 49,6 49,1
FWD measurements as a reference for the analyses 3850 49,2 52,7
of TSD measurements. 3900 49,6 49,4
The working principle for the TSD in described 3950 48,7 52,7
in Flintsch et al. (2013). 4000 49,1 52,2
4050 49,5 53,1
4100 48,8 51,2

2 COMPARISON OF FWD AND TSD Fv 202 Hp 02 Force applied Force applied

[m] FWD [kN] TSD [kN]
MEASUREMENTS 2000 48,9 51,6
2049 49,3 51,3
2.1 Sections 2099 49,3 49,6
2150 48,8 50,2
Six two-lane test sections were selected for compar- 2200 48,3 48,8
ison on the county roads 24, 202 and 222, where 2250 47,9 51,9
both TSD and FWD measurements were carried 2299 48,6 51,6
out. Each test section was 1 km long and a few 2349 48,8 51' 1
kilometres apart. The TSD measurements were 2400 48,8 51,8
averaged over subsections 10 m long, and FWD 2450 49,1 52,5
2500 49,1 52,9
measurements were taken in the outer wheel path 2549 49,1 49,0
at one side at 50 m intervals. The averaged TSD
measurement at the closest location was used to - - 2599 - - 49,5 - 48,3
2650 49,1 52,2
compare with the FWD measurement. 2700 49,3 49,4
It should be noted that the measurements with 2750 49,6 48,4
the two different equipment were taken a month 2799 49,5 50,5
2849 49,8 51,0
and a half apart during fall, but weather and tem-
2900 48,7 50,6
perature conditions were rather similar. 2950 47,2 49,7
3000 49,1 51' 1
2.2 Method
taken at the same time or corrections had been
For each section, plots of both the maximum made. Corrections for applied force is made when
deflection (d0) and the Surface Curvature Index the bearing capacity is calculated in section 3.
300 (SCI300) were made. The latter is defined
as the maximum deflection minus the deflection 2.3 Results
300 mm from the load center (d0 – d300).
A subjective judgement of the correlation The results from the statistical tests did not give a
between the values for the two equipment was conclusive answer. The regression analyses did not
made, as well as a regression analysis and a T test give an R2 value above 0,8 for any of the sections, nei-
of the relationship. ther individually nor combined. Figure 1 shows the
However, no corrections of the values related analysis results for d0 for the six sections combined.
to air or ground temperature or applied force were
made, as these factors were considered to be rather 2.4 Deflection comparison of FWD and TSD
similar for both measurements, see Table 1. It is Figures 2–5 show the measured d0 and SCI300 val-
possible that the correlation between the measure- ues for FWD and TSD for two 1000-meter long
ments would be different if the measurements were sections.

580
 y = 0,5945x+ 211,94 Fv 202 Hp 02 m 2000-3000 d0
Maximum deflections for FWD and TSD R' =0,5196
1000

E
900
BOO
700
600
.......,..... .. ..
,~.._._,-,-.;.· ···:
• •
········ •

""'c~
500
400
... .,~.

300
•• •
200
100
••
200 400 600 BOO 1000
FWD[IJ.m] - -dO-TSD[pm[ - - dO-rWO]I'f'l]

Figure 1. Regression analysis for all sections Figure 4. Deflections d0 for TSD and FWD at county
combined. road 202.

Fv 202 Hp 02 m 2000- 3000 SCI300

Fv 24 Hp 02 m 3200- 4100 d 0

!~nsnEti~n ~ H§n~~~~~~~n~~
Meter

- -SCI300-TS0[1>11'1) - -SCIJOO-FWD[I.tm]
~dO-TSOfpm] ~dO-fWD[IJtll]

Figure 5. SCI300 for TSD and FWD at county road 202.

Figure 2. Deflections d0 for TSD and FWD at county
road 24.
these sections for the two equipment types, as
well as a pretty similar variation pattern. Despite
Fv 24 Hp 02 m 3200 • 4100SCI300 a rather low R2 value from the statistical analyses,
it was therefore decided to continue the evaluation
of the measurements according to the Norwegian
bearing capacity concept.

3 BEARING CAPACITY

3.1 The Norwegian method

- -S<I300-T50]1Jt11] --S<I300-FWD~m[
The Norwegian method for the calculation of
Bearing Capacity (BC) from deflection measure-
ments was initially based on just average daily
Figure 3. SCI300 for TSD and FWD at county road 24.
traffic for heavy vehicles (AADTH) and the meas-
ured deflection by the particular equipment. When
The statistical analyses for the relationship FWD’s were later introduced, the applied force
between the deflections measured by the equip- was also included. Equation 1 below gives the gen-
ments for these two sections gave R2 values varying eral expression.
between 0.30 and 0.79, and it was not conclusive
whether the value was better for d0 or SCI300. BC = 64 × AADT
TH 0 072
DIM −0,6 (1)
Nevertheless, except for SCI300 at Fv 24, the two-
tailed T test did not indicate that the values were Here DIM is an expression including the meas-
significantly different at 5% level. ured deflection(s). (Mork, 2014)
The main impression from these figures is that The BC is designated in metric tons, and indi-
both the maximum deflection and the SCI300 cates the maximum axle load the road can take
value apparently have comparable values along without excessive degradation over the design life.

581
3.2 Estimating DIM for the TSD Fv 24 Hp 02 m 3200 - 4100
Bearing capacity
The basis for estimating a DIM value for the TSD 30,0
was the general DIM expressions for FWD and
25,0
BB, given by equations 2 and 3 below, respectively.
~ 20,0
~
( d 0 × (d ))
0 ,5
× p −1
~ 15,0
DIM
I , d (2) 1!
~ 10,0

DIM
I ,5 d 0 (3) 5,0

0,0

In the FWD equation, p is contact pressure

[MPa], and the expression assumes a contact plate Meter

radius of 150 mm. - FWD - TSD(BB-method) - TSD(FWD-method)

The TSD is equipped with load cells that contin-

uously measure vertical strain for each wheel group Figure 6. Calculated bearing capacity at county road 24.
of the measuring axle. Based on the constant grav-
ity this is transformed to wheel group load [N].
Furthermore, based on the TSD inflation pressure
of 0.82 MPa, the twin wheels’ equivalent radius is Fv 202 Hp 02 m 2000- 3000
Bearing capacity
calculated as 138 mm. This gives the following equa- 30,0
tions 4 and 5 for the TSD DIM value, labelled the
25,0
“FWD method” and the “BB method”, respectively.
~20,0
~
(9850893, d 0 × (d ))
0 ,5 G ts,o
IM d
(4) ~
× p 1 × a −1
10,0

5,0

DIM
I , (d 0 × (Pbbb Ptsd )) (5) 0,0

~ £~ ~
Meter

In these equations, the load radius a and the - FWD - TSO(BB-method) - TSO(FWD-method)

dx values are [mm]. In equation 5, Pbb is the force

applied by the BB which is constant at 39, 24kN, Figure 7. Calculated bearing capacity at county road 202.
while Ptsd is the force applied by the TSD and varies.

3.3 Results Bearingcapadtyforas.electionofseetioi'IS

.. l.-..~-
Figures 6–7 show examples of calculated BC for
FWD and TSD according to the two methods.
Note that the TSD measurements were carried out hlUHIIOI,.~'IOD - 6900

1, 5 month before the FWD measurements. ••lW .... Oino~ - 0900

hl~ .........
llol.,.,._-__ -
>d

As can be observed, both the values and the ..,._ ..._

hXIIHIIOI"'lOOD- JOOO
.

variation patterns are closely related for the two .. u•.. -vlllt<

types of equipment. However, the TSD bearing

capacity according to the FWD method is closest
to the calculated bearing capacity for the FWD.
Based on this observation, the “FWD method” is
chosen for further analyses. Results from regression
analyses confirm that the fit is better for the “FWD Figure 8. Mean and sectional bearing capacity for a
method”. selection of roads.

3.4 Bearing capacity for road network sections

The mean value is the average of the calculated
For some of the roads at which deflections were bearing capacity values for each 10 m interval
measured by the TSD only, the bearing capac- along the section.
ity is calculated according to the FWD method The sectional bearing capacity according to the
described above. Both an average value and a sec- Norwegian methodology is the lowest remaining
tional value is given. The results of these analyses value when the 10% poorest values of the section
are given by Figure 8. are disregarded, ie. the 10% percentile value.

582
 Table 2. Rutting, IRI and calculated BC for Rv 3

"'a.
:;·
"'
"'<w

"'~
"""' :J
<
Svingen—Rustad.
1> 23

Critical Critical Mean

Bearing capacity [t]

Section Last Rutting year year BC

V1
0

[meters] repaved [mm] rutting IRI unevenness [t]

.r:.
0
u.o

f{
0

8500–9500 2014 7,9 2034 1,5 2076 17,4

l
,.,_,
0

9500–10500 2009 18,1 2020 2,4 2044 12,6

b

10500–11500 2009 19 2020 2,8 2029 13,4

0

11500–12500 2009 12,4 2027 2,2 2062 21,6

11883
12203
12523

13483

16363
16683
17003
17323
13803
11563

15083
15403
15723
9003
9323
9643
9963
10283
10603
10923

141 23

16043
14763
~ ~~~~
112 43

14443
12500–13500 2009 12,7 2028 2,4 2058 25,1
13500–14500 1998 16,4 2025 2,5 2056 24,7
~

14500–15500 1998 15 2026 1,9 2068 21,1

15500–16500 1998 14,6 2027 2 2066 23,2
Figure 9. Bearing capacity for Rv 3 section 16500–17500 1998 17,4 2024 2,7 2045 23,2
Svingen—Rustad.

If there is a big gap between the mean bearing

w
:;.
~

~
a.
<>

~
:;;

2
c
capacity and the sectional bearing capacity, it usually

0
0
~ ~ ~ ~
means that the bearing capacity along the section
varies substantially, this can also be seen by a rela-
tively high standard deviation. As can be seen from
[~m)

Figure 8, the main arteries E6, E16, Rv 2 and Rv 3

8
SCBOO

~ ~ ~

generally have the highest bearing capacity values,

but some of these roads also show the greatest dis-
crepancy between the mean and the sectional value.
~

Especially for Rv 3 Svingen-Rustad, this deviation

0

:f3274

5587
5818
2118

2580

6050
4198

4660
....
119'1
1425
1655

3043

~ 3505
3735
3966

5124
5355
269

732
963

2349

2812

4429

6512
"
501

4891

6281
is severe. This is a sign of some parts with either
a rather poor or parts with high bearing capacity
along this section, as can be seen by Figure 9.
Figure 10. E6 Uthus—Basterud SCI300.
3.5 Bearing capacity compared to other
indicators of pavement performance from 13500 to 17500 have a surface layer which is
Some attempts to relate calculated bearing capacity more than double as old, and have nearly the same
for 1000 m long sections to rutting and unevenness rut depth. But as can be seen from the extrapolated
measurements for the same sections are made. This is critical year for rut depth, the rut development is less
done to check whether there is a connection between rapid. The registered traffic volume is the same for
poor bearing capacity and an increase in rut devel- all sections. This is a clear indication that a shorter
opment or unevenness or rutting. For the section of pavement life and a poor bearing capacity goes hand
Rv 3 Svingen—Rustad the bearing capacity is clearly in hand.
more poor between station numbers 9000 and 11800. For the 6.5 km long E6 section between Uthus
The following data has been extracted from the Nor- and Basterud opened to traffic 1972, Figure 10
wegian Pavement Management System. shows how the SCI300 varies along the section.
The mean BC values are from the TSD meas- The TSD performed measurements three times
urements. The IRI and rutting values are the 90% for this section to check reproducibility, and the
percentile values for unevenness and rutting in the figure shows the average of the three. Obviously,
outer wheel path along the subsections, respectively. there are five spots with higher SCI300 and there-
The critical years for rehabilitation measures are for lower bearing capacity than the rest. From the
calculated from the recent IRI and rut develop- measurements it can be concluded that at least the
ment numbers, as the year the 90% percentile for spots 2288, 3303 and 5134 have consistently higher
IRI is prognosticated to reach the triggering level of d0 and thereby SCI300 values than the surrounding
4.5 mm/m and the year the 90% percentile for rutting parts, giving a relatively low BC, while the other two
reaches the triggering value of 25 mm, respectively. might be due to erroneous measurements. Two of
It is observed that the sections from station 9500 these three spots coincide almost perfectly with parts
to station 11500 are relatively recently repaved, has of the road which lacked a bottom stone layer above
the deepest ruts, and will be the first to have a need the subgrade to drain the pavement, documented
for further repaving. These are also the sections by excavation efforts 1976, see Figure 11. This is no
with the lowest calculated BC value. The sections evidence, but could be regarded as an explanation

583
 E6 Uthus- Basterud layers
measurements, especially when both maximum
deflection and some measure of deflection bowl
shape is included. However, this comparison is
based on a limited amount of data, and further
18
comparative measurements should be carried out
" to verify this observation, and/or to come up with
a possible correction factor.
Nevertheless, there are strong indications that
the calculated TSD bearing capacity can be closely
HPffi HPill HPill HPffi HPffi HPffi HPffi HPffi ~ffi HPffi HPffi HPffi related to functional performance indicators used
M1000 MlSOO M2000 M2500 M30Cl0 M3500 M4000 M4 500 MSOOO MSSOO MGOOO M6500

• Stonelayer • Subb.ISO! •Penetratedcrushedston<? • Asph<~ltconcrete

for timing rehabilitation measures. This relation-
ship needs further investigation, though.
Figure 11. E6 Uthus—Basterud layers.
ACKNOWLEDGEMENT
for the lower BC at these points. However, it has not
been possible to relate this to pavement life indica- This research is financed by several projects within
tors, probably because the bearing capacity in gen- the Norwegian Public Roads Administration and
eral is high, and as the IRI and rutting values in the the Directorate of Public Roads, and their financial
Pavement Management System are connected to support is highly appreciated. We also would like
longer subsections than those giving the local low to thank the Polish research institute Instytut Bad-
BC values. awczy Dróg i Mostów (IBDiM) which performed
the TSD measurements over a limited period of
time, and Greenwood Engineering for analyzing
4 CONCLUSIONS the raw data from the TSD measurements and pre-
paring the results for further analyses.
The TSD measurements performed in Hedmark
county 2015 clearly show that this type of equip-
ment is well suited for network level measurements REFERENCES
for Norwegian conditions, and that high measuring
capacity is achieved even for relatively narrow, wind- Antonsen, V. (2016). Måling med Traffic Speed Deflec-
ing low volume roads. FWD and TSD measure- tometer på riks - og fylkesveger I Hedmark høsten
ments carried out on the same roads under similar 2015.
Flintsch, G., Katicha, S., Bryce, J., Ferne, B., Nell, S. &
conditions show a high degree of coincident values
Diefenderfer, B. (2013). Assessment of Contunuous
both regarding deflections and variation patterns. Pavement Deflection Measuring Technologies. Trans-
An attempt to calculate bearing capacity from portation Research Board.
the TSD measurements show great degree of Mork, H. (2014). Bereevne for eksisterande veg. Notat
coincidence with values calculated from FWD 991 Institutt for veg - og jernbanebygging. NTNU.

584
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Assessment of pavement structures at traffic speed

A. Zofka, J. Sudyka & D. Sybilski

Road and Bridge Research Institute (IBDiM), Warsaw, Poland

ABSTRACT: Traffic Speed Deflectometer (TSD) is a modern device intended to assess pavement
bearing capacity at traffic speeds, i.e. up to approximately 90 km/h. The main concept behind the TSD
operations are the ultra-high speed sensors based on Doppler phenomenon. Several of those sensors are
installed on the special beam inside the semi-trailer while its rear axle is loaded with a static load, typically
100 kN. In the first part, this paper provides a general information on the TSD, including its mechanical
features, concept of operation, and data interpretation approaches. Since it is still relatively new device,
there is no globally accepted approach in analyzing the TSD measurements. TSD has been so far used
mainly on the network level however it can also provide detailed information on the project level. In order
to demonstrate the capabilities of the TSD measurements, a second part of this paper demonstrates
example data and associated analysis for the network and project levels. Presented results show high
effectiveness of the TSD measurements conducted at traffic speeds at both levels. It is believed that the
TSD-type devices will enter soon routine pavement management activities and even more comprehensive
systems will be developed.

1 INTRODUCTION & Nell & Diefenderfer 2012). Currently, there are

eight working TSD devices in the following coun-
In many countries pavement structures are increas- tries: UK, Denmark, Italy, Poland, USA, South
ingly exposed to higher than design traffic load- Africa, Australia and China. The following section
ings which cause accelerated damage growth and presents a brief development of deflection devices
premature failures amplified further by the severe and the next section explains the TSD operational
climate changes. This leads to the growing number concept and data interpretation procedures.
of pavements in the deteriorated condition that
not necessary show the damage through the sur-
face distresses. For that reason there is an urgent 3 PAVEMENT BEARING CAPACITY
need to implement quick and reliable method(s) ASSESSMENT
able to acquire information on pavement bearing
capacity especially on a network level. Typically, Deflections have been traditionally measured using
bearing capacity is assessed using pavement deflec- static devices, such as Falling Weight Deflectometer
tions due either to static, dynamic or moving load. (FWD) or Benkelman beam. The main disadvan-
Deflections or their derivative parameters can be tage of these method is that they are relatively slow
then implemented into Pavement Management and thus ineffective to implement at the network
Systems (PMS) in order to directly capture the level. Further, they create unsafe working condi-
structural aspect of the pavement condition and tions for their operators as they require lane closure
more effectively manage maintenance and rehabili- and traffic redirection. On the other hand, recent
tation activities under common budget constraints technological developments have allowed to com-
(Bryce & Flintsch & Katicha & Diefenderfer 2013; mercialize new devices that are able to record pave-
Stubstad & Carvalho & Briggs & Selezneva 2012; ment deflections under the moving load (Flintsch
Choummanivong & Martin 2010; Kim 2001). & Katicha & Bryce & Ferne & Nell & Diefenderfer
2012; Rada & Nazarian 2011). An example of such
a device is the Traffic Speed Deflectometer (TSD)
2 TSD BACKGROUND that continuously measures pavement response as
it is moving at the highway speed. Several highway
The first prototype of the Traffic Speed Deflec- agencies around the world are currently assessing
tometer (TSD) was developed in Denmark in the and/or implementing the TSD, and generally the
late 1990’s (Flintsch & Katicha & Bryce & Ferne TSD is considered as productive, safe and effective

585
tool for the network level evaluation (Hildebrand a) Left side wheel b) Right side wheel
0.7, - - - - - - - - - - - ,
0.7 ......--------.--~-,
& Rasmussen 2002; Rasmussen & Krarup & Hilde-
brand 2002). However, there is still a significant 0.6 0.6

effort to be made in order to fully understand all ~0.5 0.5

phenomena present during the TSD operations ~

~0.4 1
04.9 m/~l. '.._ 0.4

and to determine how they affect the actual meas-

urements as well as the post-processing algorithms
(Muller & Roberts 2012; Graczyk & Rafa & Zofka
& Sudyka 2014; Zofka & Graczyk & Rafa 2015;
t: 0.1
114 9 " ' , ,
0.3

0.2

0.1

Zofka & Sudyka & Maliszewski & Harasim & 80 110

Sybilski 2014).
Deflection devices can be categorized by load-
Figure 1. Simulated wind velocity effect at 90° angle
ing characteristics into following groups (example and 22 m/s TSD speed.
devices listed in parentheses):
1. Static (Plate Load Test) 0. 1 ,..,---~--,-------.------,--~----,--,

2. Impulse (Falling Weight Deflectometer, FWD) 0.09

0.08
3. Steady-state vibratory (Road Rater, Dynaflect)
: ; 0.07
4. In-motion (moving): -~ 0.06
a. Deflectographs (Benkelman Beam, LaCroix, ~
g oos Ju mm/ml
Curviameter) ~ 0.04 "\.
b. Deflectometers (Road Deflection Tester, e o.o3
il.
RDT; Rolling Wheel Deflectometer, RWD; 0.02

TSD) 0.01

90 100 11 0
It should be noted that deflectographs use % of static load
deflection acquisition systems that are in-contact
with a pavement. On the other hand, deflectome- Figure 2. Simulated pavement roughness effect at
ters employ loading and deflection acquisition sys- 22 m/s TSD speed.
tems that are moving as one system and deflections
are measured via non-contact techniques. (SN), radius of curvature and effective pavement
modulus. Alternatively, indices related to the struc-
tural condition such as Structural Adequacy Index
4 TSD CONCEPT OF OPERATION (SAI), Structural Condition Index (SCI), Remain-
ing Service Life (RSL) and/or Surface Curvature
The TSD device is a semi-trailer truck with a Index (SCI) (at various distances) can be also
standard 2-axle tractor that host the driver and determined which alleviates the adaptation process
the operator in the passenger seat. The single axle of the TSD into agency’s operations and enhances
semi-trailer host the Data Acquisition System its PMS algorithms.
(DAQ) together with the measurement system Accuracy and precision of TSD measurements
and the fixed (dead) load attached to the bottom can be influenced by a number of internal and exter-
of the main body towards the rear axle. The main nal aspects, e.g. calibration and QA procedures,
part of the measurement system is the horizontal wind and temperature during the measurement,
beam hold in that position by the array of high- pavement roughness and tire-pavement interaction
speed actuators and equipped with a number of (Zofka & Graczyk & Rafa 2015). Figure 1 and 2
Doppler sensors. The pavement deflection veloci- present respectively example effects of wind veloc-
ties measured by these sensors are divided by the ity and pavement roughness on dynamic TSD load
instantaneous vehicle speed which gives the deflec- (Zofka & Graczyk & Rafa 2015). One can notice
tion slopes. Slopes can be used directly in the anal- that presented effects can significantly influence
ysis or the absolute deflections can be obtained the TSD dynamic load (up to 30% in case of
by integrating the deflection slopes numerically roughness) and appropriate measures should be
or by fitting the slopes to the solution of a suit- employed during the data interpretation.
able pavement mechanical model containing typi-
cally viscoelastic elements. Once the slopes and
corresponding deflection basins are determined, 5 TSD DATA INTERPRETATION
the TSD data allows to calculate well-established PROCEDURES
structural parameters (similar to the FWD) such
as normalized maximum deflection d0, strain at Typical TSD interpretation procedure is focused on
the bottom of asphalt layer, Structural Number analysing moving TSD truck which is well-known

586
as Lagrangian description (Zofka & Sudyka & -200 1 . , • , I

: : 1I~~~~1'f\'· ;/~~ ~rall\~~!t.:~3i(\~.,~.,•~f•.~

Maliszewski & Harasim & Sybilski 2014; Graczyk
& Rafa & Zofka 2016). In Lagrangian description
the focus is on an object (particle) moving through 6 b:: ~& ~~ · ~6d ::! ~~ ::~
the space and specific equations are solved to to -500 " ,. 6 i '
determine the location and status of individual 0
I
~ · '
objects solely as a function of time t. However
:s
~

recent study (Zofka & Sudyka & Maliszewski & -700

0
Harasim & Sybilski 2014) demonstrated an alter- "0

native procedure based on the Eulerian descrip-

tion. In this approach, the focus is on a specific
location in space (defined by its position x) and
all processes are observed that occur with time t
at that specific location. By considering a set of 1200
points the processes can be expressed in terms
of both position x and time t. Examples of
both approaches are presented in the fo llowing
sections.

6 TSD RESULT DEMONSTRATION

TSD collects high-rate data that can be post-

processed using two general approaches depend-
ing on the analysis scope (project vs. network
level) and objective as well as other factors, such
as number of Doppler sensors and available ana-
lytical tools. The following two sections present
several example approaches in interpreting TSD - Section #2
data. ~~--~2~
oo~--~4~oo~--~6~o~o----M
~o~==1~o~oo~==~
Mileage [m]

6.1 Network level Figure 3. Typical results from two sections in terms of,
Figure 3 presents TSD post-processing results a) d0, b) SCI300 parameter.
based on the Lagrangian approach. The data were
collected at 70 km/h and are presented in terms
Sensor location [mm]
of d0 and SCI300 parameters along 1220 m sec-
00 110 210310 41 0 610 910 1510
tions. It can be observed that Section #1 is fairly ~~·o--------- ---------------a
uniform in terms of both parameters but Sec- -100
__,.,,n-······.a····
tion #2 could be split into two sub-sections. One -200 -
can clearly observe that sub-section after approx. ~ _,rl'
~ -300 p'
600 m is weaker and produces more variability. 0 /
This represents a great advantage of the TSD to
quickly “scan” structural condition at the network :s
s:::
-400 '

0
level and identify “weaker” or doubtful sections for :g
L________J .•

more detailed investigation. Q)

'\iS
0

6.2 Project level ~ I

Figure 4 shows typical deflection basins collected -1ooo l__J____j_--L_____c__ __ L_ __ _ _ _L __ _ _ __ __ __ ________:I

at 70 km/h from two different pavement sections.

Data was obtained from a standard (i.e. Lagrang- Figure 4. Typical results (deflection basins) from two
ian) processing algorithm. It can be noted that sections at particular TSD location.
both pavement sections differ significantly in
their bearing capacities. This type of analysis is
suitable more for a project level and it is similar corresponds to different pavement cross-section
to the FWD analysis. Finally, Figure 5 demon- and such deflection isochrones maps can be useful
strates the Eulerian approach. Each vertical profile for both network and project level.

587
 Level—Interim Models. Austroads Publication No.
" ' AP-T159/10, ARRB 2010.

" ,Illii/II ~ IpJ I~I dIIHf~ .:

Flintsch G., Katicha S., Bryce J., Ferne B., Nell S.,
Diefenderfer B., Evaluation of Traffic-Speed Deflec-
tometers. Transportation Research Record: Journal of
the Transportation Research Board, No. 2304, Trans-
portation Research Board of the National Academies,
' '· Washington, D.C., 2012, pp. 37–46.
Graczyk M., Rafa J., Zofka A., Sudyka J., Analytical
Eulerian points (0 .1m) (-------> TSD direction)
Solution of Pavement Deflections and its Application to
the TSD Measurements, presented at 26th ARRB Con-
Figure 5. Example of deflection isochrone map pre- ference 2014, October 19–22, 2014, Sydney, Australia.
pared from the actual TSD measurements at 70 km/h Graczyk M., Rafa J., Zofka A., The selected problems of
(color-coded deflections are in mm). multi-layer pavements—influence of composite impacts
vehicles and climatic factors on the behavior of roads
pavements, Transport Research Procedia, Vol. 14, pp:
7 SUMMARY 2487–2496, 2016.
Hildebrand G., Rasmussen S., Development of a High
Presented results demonstrate a high effectiveness of Speed Deflectograph. Danish Road Institute, Report
117, 2002.
the TSD measurements conducted at traffic speeds. Kim Y.R., Assessing Pavement Layer Condition Using
It is believed that the TSD-type devices will enter Deflection Data. NCHRP digest, NCHRP Project
quickly routine pavement management activities 10-48, 2001.
and even more comprehensive systems will be devel- Krarup J., Rasmussen S., Aagaard L., Hjorth P., Output
oped soon enhanced with, for example, Ground From The Greenwood Traffic Speed Deflectometer. Pre-
Penetrating Radar (GPR) or infrared (IR) cameras. sented at the 22nd ARRB Group Conference, 2006.
Such additions will allow for more complex assess- Muller W., Roberts J., Revised approach to assessing traf-
ment of pavement condition and will eliminate fic speed deflectometer data and field validation of
duplicate measurement efforts with other vehicles. deflection bowl predictions. International Journal of
Pavement Engineering, DOI:10.1080/10298436.2012.
It should be also mentioned that although this 715646, 2012.
paper presents one specific equipment (TSD), Rada G.R. Nazarian S., The State-of-the-Technology of
observations and procedures shown here are Moving Pavement Deflection Testing. Final Report,
equally relevant to other deflectometers operating FHWA-DTFH61-08-D-00025, U.S. Department of
with the same or similar concept. As of now (year Transportation, Washington, D.C., 2011.
of 2016) there is no other TSD-like commercially Rasmussen S., Krarup J., Hildebrand G., Non-contact
available equipment but this situation may change Deflection Measurement at High Speed, Presented at
in the near future due to the growing demand and the BCRRA, 2002.
revolutionary developments in the digital and elec- Stubstad R., Carvalho R., Briggs R., Selezneva O., Sim-
plified Techniques for Evaluation and Interpretation
tronic industries. of Pavement Deflections for Network-Level Analysis:
Guide for Assessment of Pavement Structure Perform-
ance for PMS, Report FHWA-HRT-12-025, 2012.
REFERENCES Zofka A., Graczyk M., Rafa J., Qualitative evaluation of
stochastic factors affecting the Traffic Speed Deflec-
Bryce J., Flintsch G., Katicha S., Diefenderfer B., Devel- tometer results, presented at Transportation Research
oping a Network-Level Structural Capacity Index for Board 94th Annual Meeting, January 11–15, 2015,
Structural Evaluation of Pavements. Final Report Washington DC, USA.
VCTIR 13-R9, Virginia Department of Transporta- Zofka A., Sudyka J., Maliszewski M., Harasim P., Sybil-
tion, 2013. ski D., Alternative Approach for Interpreting Traffic
Choummanivong L., Martin T., Predicting Struc- Speed Deflectometer Results, Transportation Research
tural Deterioration of Pavements at a Network Record, No. 2457, pp: 12–18, 2014.

588
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Integration of traffic speed deflectometer and ground penetrating

radar for network-level roadway structure evaluation

K.R. Maser & A. Carmichael

Infrasense, Inc., Woburn, MA, USA

P. Schmalzer
Nichols Consulting Engineers, Chtd., Reno, NV, USA

B. Shaw
Idaho Transportation Department, Rigby, ID, USA

ABSTRACT: A pilot project was conducted on 683-mile network of roadways in eastern Idaho with the
goal of generating network-level roadway structure data to support future project planning and design
efforts and advance the management of assets into a more efficient set of priorities. The roadways were
surveyed continuously with a Traffic Speed Deflectometer (TSD) and with Ground Penetrating Radar
(GPR) and 2-mile segments were selected for Falling Weight Deflectometer (FWD) testing to confirm the
TSD data and calibrate in the TSD analysis. TSD and GPR data were combined to calculate subgrade
modulus, effective structural number, and remaining life throughout the network. Using these results,
the network was divided into “homogeneous” sub-segments of typical project length for the purpose of
identifying and programming rehabilitation projects. The level of condition detail available at the sub-
segment level allows for the development of preliminary rehabilitation design without the need for addi-
tional data.

1 INTRODUCTION

The reported work represents a pilot project to

incorporate roadway structural capacity into Ida-
ho’s state highway pavement model to inform cor-
ridor planning and project design. The pilot project
combines the 515-mile East Idaho Loop Corridor
(EILC) with 84 miles of Interstate 15 in each direc-
tion. The makeup of the pilot network is described
in Table 1 and shown spatially in Figure 1.

Table 1. Corridor segments for east Idaho loop.

Segment Begin End Centerline

name milepost milepost miles

US 93 82.6 350.819 183.508*

US 20 244.555 306.056 61.501
SH33 (west) 0.00 78.236 78.236
SH 22 24.67 68.606 43.936
SH 28 15.15 135.645 120.495 Figure 1. East IdahoPilot network.
SH 75 217.122 244.325 27.203
I-15 NB 111.859 196.00 84.141
I-15 SB 196.00 111.859 84.141 Typical pavement management systems depend
Total 683,161 entirely on visible distresses and functional charac-
teristics such as roughness to characterize the con-
*due to equation in section. dition of pavements. These characteristics relate to

589
the functionality of the pavement at the time it was personnel to visualize the results and to facilitate
measured, but relate less well to the condition of network and project level decisions. This type of
the pavement at a future point in time or the type data visualization is an outgrowth of an ongoing
and cost of repair necessary to effectively improve effort being carried out at the Idaho Transporta-
the condition of the pavement. This is problem- tion Department (6).
atic for the many of the key goals of a pavement The scope of work described in this paper com-
management system, including determining main- bines Traffic Speed Deflectometer (TSD) testing
tenance needs and the impact of various funding to evaluate the strength of the pavement struc-
scenarios on future network condition. Structural ture; Falling Weight Deflection (FWD) testing
information, such as pavement thickness and to calibrate and confirm the TSD data; Ground
stiffness, is required to determine the degree of Penetrating Radar (GPR) to determine the thick-
structural deficiency and enable robust predictive ness of asphalt and unbound base layers; and core
models for future condition. Obtaining this infor- and boring data for calibration and confirmation
mation requires a quality and quantity of data that of the GPR data. This combined data was applied
up until very recently has been deemed cost-pro- to an initial 44-mile segment of SH-28 to calcu-
hibitive to collect by most agencies. late pavement and subgrade moduli and resulting
Falling Weight Deflectometer (FWD) testing is effective structural number on a point by point
the typical practice in the United States for deter- basis, using the methodology in the AASHTO
mining in-situ pavement structural properties, and 1993 Guide for the Design of Pavement Struc-
as of 2008 over 90% of state highway agencies rou- tures (7). The effective structural number was
tinely perform FWD testing (1). The FWD, how- used to calculate remaining life in ESALs, and,
ever, is a stationary test that requires lane closures, along with traffic projections, remaining life was
and is not well suited to network-level evaluation. represented in years. The pattern of remaining life
Typical production rates for FWD testing ranges values was used to divide the sample segment into
from 200 to 400 test points per day, so testing any sub-segments of relatively uniform remaining life,
reasonably sized network requires tremendous and these sub segments will serve as the basis for
effort even with very widely spaced test points. As future project-level rehabilitation design. All the
a result of this limitation, only 16% of state high- data generated during this process has been incor-
way agencies regularly collect structural property porated into a geodatabase within ArcGIS.
data at the network level (2). The cost and disrup-
tion to traffic of data collection activities often
cited as the reason for not doing so. 2 DATA COLLECTION
To address these limitations, technologies for
deflection testing at highway speeds have been Continuous collection of GPR and TSD data was
introduced, including the Traffic Speed Deflec- carried out throughout the pilot network. FWD
tometer (TSD). The TSD technology, which was data was collected at 100-foot intervals along select
developed in Denmark and has been deployed in 2-mile long segments based on the TSD data review.
Europe and Australia over the past 5 years, has Cores and borings were extracted at 99 locations
recently been imported to and evaluated in the determined from the GPR data review to best rep-
United States. The FHWA has recently completed resent the pavement structure conditions through-
a pooled fund study (3), in which the TSD was out the project section. GPR data was collected
brought to the USA and tested in a number of continuously in each wheel path in both directions
participating states. The TSD measures the deflec- of each roadway. The following sections provide
tion slope under the load of a loaded tractor trailer details related to the GPR, FWD, and physical
several positional offsets from the tire load while sampling testing protocols and equipment.
driving at normal driving speed. Various studies
have been carried out to calculate various indices
from these deflection slopes to correlate TSD with 3 GPR DATA COLLECTION
FWD results (4) and to correlate these indices to
strains at the bottom of the pavement (5). GPR pavement data was collected by Infrasense
Since the TSD data allows for the computation from August 12–15, 2015. The pavement survey
of a deflection bowl, much like what is provided by utilized a dual antenna system. This system con-
the FWD, it can provide the basis for a compara- sisted of a GSSI SIR-20 radar control and data
ble evaluation of the strength of the subgrade and acquisition unit, a pair of 1 GHz horn antennas,
pavement layers. This information, coupled with and a high resolution electronic Distance Measur-
the pavement layer thicknesses, is key in determin- ing Instrument (DMI). The data collection rate
ing the pavement structural capacity and predicting was one scan per foot of linear travel. The sur-
remaining life. The final product has been incor- vey data was continuously digitized and stored in
porated into an ArcGIS database to allow agency the SIR-20 along with the encoded distance and

590
 well established, the TSD technology is relatively
new to the US, and has not previously been used
for the type of analysis planned for this project.
In order to determine appropriate areas for FWD
testing, the TSD deflection data was reviewed and
13 representative sections were identified for FWD
testing. These sections were selected to represent
a range of pavement conditions, ranging from
“strong”, “weak”, and “variable”. Each FWD sec-
tion was 10,000 feet long, and FWD data was col-
lected at 100 foot intervals with in each of these
sections, yielding 100 test points per section.
Figure 2. Greenwood TSD equipment. The FWD data was collected during the week
of October 26–30, 2015, using a Dynatest truck-
mounted deflectometer, model number 8002. At
differentially corrected Global Positioning System each test point, five drops were performed includ-
(GPS) coordinates. Markers were placed manually ing a seating drop at a nominal load level of
in the data at specified locations (including mile- 12,000 pounds, two recorded drops at a nominal
posts and bridges) for ground reference. 9,000 pounds and two recorded drops at a nomi-
nal 12,000 pounds. Each drop produces a vertical
deformation, or deflection of the pavement which
4 TSD DATA COLLECTION is measured by a series of 7 deflection sensors
spaced at offsets 0, 8, 12, 18, 24, 36, and 60 inches
The TSD survey was carried out on the EILC by from the load. The deflection values, coupled with
Green wood between September 15 and 19, 2015, the pavement layer thicknesses obtained with the
and the I-15 segments were collected by ARRB on GPR, are used to calculate the pavement layer
August 2, 2016. Data on the EILC was collected moduli. In addition to the load and deflection
primarily in one direction of each road in the sur- data, air temperature, pavement surface tempera-
vey loop, while I-15 data was collected in the travel ture, and differential GPS data were collected at
lane in each direction. A photo of the Greenwood each location.
TSD equipment is shown in Figure 2.
The TSD is an articulated truck with a nomi-
nal rear axle load of 20 kips (US service), which 6 CORE SAMPLING AND BORINGS
utilizes seven Doppler lasers mounted on a servo-
hydraulic beam to record the deflection velocity of Pavement cores and borings were carried out at
a loaded pavement. Six Doppler lasers are posi- 157 locations with each location specified based
tioned such that they measure deflection velocity on a preliminary screening of the GPR data. The
at a range of distances in front of the rear axle, GPR data was initially reviewed to identify loca-
at 100, 200, 300, 600, 900, and 1500 mm (3.9, 7.9, tions where pavement material sampling would be
11.8, 23.6, and 59 inches). The seventh sensor is most beneficial. This review highlighted variations
positioned 3500 mm (138 inches) in front of the in the layer thickness and sought to ensure that the
rear axle largely outside the deflection bowl, act- material sampling will be fully representative of
ing as a reference laser. The beam on which the the observed pavement structures. Recommended
lasers are mounted moves up and down in oppo- locations for pavement sampling were provided in
sition to the movement of the trailer in order to terms of GPS coordinates and local milepost refer-
keep the lasers at constant height from the pave- ences. In addition to extracting cores and boring
ment surface. To prevent thermal distortion of the samples, sample testing was carried out to deter-
steel measurement beam, a climate control system mine moisture content, R-value, and percent pass-
maintains the trailer temperature at a constant ing a 200 sieve.
20°C (68°F). Data is recorded at a survey speed of
up to 50 mph and is reported at 10 meter intervals
on the EILC and 0.01 mile intervals on I-15 7 GPR DATA ANALYSIS

The GPR pavement data was initially reviewed to

5 SUPPLEMENTAL FWD TESTING select boring locations for calibration and to con-
firm composition of the observed layer structure.
FWD testing was carried out to confirm the TSD Following this review, a more detailed analysis of
results and to establish a methodology for analyz- the data was carried out to calculate the thickness
ing the TSD data. While the FWD methodology is of the bound and unbound material layer thick-

591
nesses. The pavement layer thicknesses were cal- 9 TSD AND FWD DATA COMPARISON
culated and reported at each reported TSD test
locations and reported as the average of data ± 5 Upon review of the TSD data, 13 comparison seg-
feet on either side of the reported location. The ments were identified in which FWD testing would
GPR layer thickness results were used to calculate be carried out. The segments were selecteded to
pavement resilient modulus and pavement modu- represent TSD indications of, strong, weak, and
lus from the TSD data. highly variable pavement conditions. Each com-
parison segment was approximately 2 miles long,
and FWD data was collected at 100-foot spacing
8 COMPARISON OF GPR VS. CORE DATA
within each segment. A methodology for calculat-
ing deflections from the TSD deflection was devel-
Prior to carrying out the full GPR pavement thick-
oped based on a combination of Boussinesque half
ness analysis, the GPR data was analyzed at each of
space theory and piecewise integration (8,9). Resil-
the 99 boring locations, and the resulting layer depth
ient modulus (Mr) was then calculated from the
results were compared to data obtained from the
TSD and FWD data using the method described
cores taken at the boring locations. This core infor-
in the AASHTO 93 Design Guide (7). A compari-
mation was used by the GPR analyst to facilitate the
son of centerline deflections (D0) and Mr values
layer interpretation of the pavement structure. The
calculated from both TSD and FWD data sources
comparison of the GPR vs. core total asphalt thick-
showed similar spatial patterns, with TSD-based
ness shown in Figure 3 was completed once the layer
Mr values generally higher and having a higher
interpretation was carried out. Note that the in-hole
standard deviation than the values calculated from
measurement of the asphalt thickness is used in this
the FWD. This Mr difference was more signficant
comparison rather than the extracted core thickness
in segments that had a Cement Recycled Asphalt
measurement. This is because, in a number of cases
Base System (CRABS). Average structural number
material at the bottom of the core was not fully
for each segment calculated from the D0, Mr and
extracted and the in-hole measurement was greater
the GPR layer thickness data for both FWD and
than the core measurement. At 8 of the 99 locations
TSD data showed reasonably good agreement,
the in-hole measurement deviated from the GPR
with an average absolute difference of 8.2% (8, 9).
measurement by greater than 2 inches, regardless
Therefore, it was concluded that the analyzed TSD
of the layer interpretation. These locations, all of
data reasonably replicates the results that would
which have cores that broke into two or more pieces,
have been obtained using an FWD.
were treated as outliers and were not included in the
correlation presented in Figure 3.
Figure 3 shows a very good correlation between the
10 DETAILED ANALYSIS OF A SELECTED
GPR-based asphalt thickness values and the in-situ
EILC SEGMENT
values measured in the core holes. No further calibra-
tion is suggested by this data, and the layer interpre-
Using the analysis methods described above and
tations supporting this analysis served as a basis for
in (7), pavement modulus Ep, effective structural
interpreting the GPR data on roadway sections in
number, SNeff, were calculated at each TSD test
the regions surrounding each core. For the 8 outlier
location on a selected ∼45-mile section of SH-28
core values, the presence of a broken core suggests the
from Leadore to Salmon. Remaining life in ESALs
possibility that the GPR reported one of the internal
was then calculated and converted to remaining
boundaries rather than the bottom of the core.
life using projected traffic volume data from ITD.
Figure 4 shows a small sample of results reveal-
16 ~-----------------------, ing the spatial level of detail resulting from this
14
analysis. Based on this analysis, the data was fur-
y =1.0207x ther divided into subsegments, 1 to 3 miles long,
Rz = 0.8812
of relatively homogenous condition, which repre-
sent potential maintenance projects. The results of
the remaining life calculation and the division into
homogeneous sections is depicted in Figure 5.
.. :-
.. .'·.,...'· .! . 11 GEOSPATIAL DATABASE

10 12 14
The geodatabase developed under the initial US
AC thickness from GPR {in.) 26 project with ITD was reviewed and updated to
reflect the larger scale of EILC, and the addition
Figure 3. GPR vs. core thickness. of the TSD data. Since the previous project was

592
 Figure 6. Core and boring locations and sample core
popup.

Figure 4. Google Earth kml plots showing the sub-

grade resilient modulus (A), effective pavement structural
number (B), and remaining life of pavement (C).

35 Remainng Life .. .
- -Avg remaining life ...
0 30 ,I
b"' 25 '
<2
:J 20
OJ)
c:: 15
:soe
s
<1.)
10 Figure 7. Remaining life for network segments.
0:::
5
Figure 4. Figure 7 shows an example of a corre-
0
sponding ArcGIS map of all of the 99 boring loca-
90.5 100.5 110.5 120.5 130.5 tions for the east Idaho loop. The identify tool can
Milepost be used at each location to view each material layer
of any particular boring, along with the properties
Figure 5. Sample segment—calculated remaining life of that layer.
and division into homogeneous sub-segments.

12 DISCUSSION
based on FWD data alone, one primary modifi-
cation was the expansion of the FWD analysis The above results show how continuous network-
table of results to a more generic deflection test- level pavement structure data can be generated
ing analysis table. This new table combines data using a combination of Traffic Speed Deflectom-
generated both by the FWD and the TSD. The eter (TSD) and Ground Penetrating Radar (GPR)
reported results, including subgrade modulus, data. At a network level, the pavement structure
pavement modulus, effective structural number, data generated represents a considerable advance-
remaining ESAL’s, and remaining life in years ment over surface distress alone for the prediction
remain consistent between the two sources of of remaining life and the projection of pavement
data. rehabilitation needs. At the project level, the data
The geodatabase was initially set up using Arc- provides considerable detail for estimating overlay
GIS, and then published in Idaho’s IPLAN, which requirements and for evaluating other rehabilita-
is ITD’s web-based portal to geospatially refer- tion alternatives. The incorporation of all the data
enced transportation data sources. Figure 6 shows into the geospatial database provides agency per-
an IPLAN map of the TSD analysis segment on sonnel with access to all data at various level of
SH28. The map shows the point-by-point remain- detail, along with the ability to evaluate spatial pat-
ing life calculation based on the data shown in terns for planning and programming purposes.

593
 As an example of the power of the geodata- realistic pavement performance and scenarios and
base, Figure 8 shows detailed subsegment values of associated resource requirements and allocations,
remaining life, Mr and SNeff for a segment of SH and how this data can be used to propose more
28 between Salmon to just south of Leodore. The appropriate rehabilitation designs. While it is rea-
remaining life values in Figure 8(a) shows two defi- sonable to assume that better information leads to
cient areas (remaining life < 3 years), one just south lower costs, the ultimate economic benefit of using
of Tendoy and the other just south of Lemhi (both this roadway structure data vs. the cost of generating
circled). The SNeff data for these two areas, depicted this data has yet to be studied. ITD District 6 will
in Figure 8(b), shows a weak pavement structure, be completing its pilot program in 2017, and such
while to Mr data for these two areas, depicted in a study is planned as part of that completion effort.
Figure 8(c), shows a strong subgrade south of Ten-
doy but a relatively weak subgrade south of Lemhi.
ACKNOWLEDGEMENT
These differences in subgrade condition suggest
differences in rehabilitation design to deal with the
The authors wish to acknowledge the Idaho Trans-
differences in the subgrade condition.
portation Department (ITD) for supporting this
The above example shows how the roadway
work, the FHWA for making the TSD available to
structure data can be used to project generate more
the project through ITD’s participation in the pooled
fund project, TPF 5(282), and ARRB for providing
TSD services for the I-15 portion of this project

REFERENCES

[1] Alavi, S., LeCates, J. & Tavares, M. 2008. Falling Weight

Deflectometer Usage, NCHRP Synthesis 381, Transpor-
tation Research Board, Washington, D.C., 2008.
[2] Flintsch, G. & McGee, K. 2009. Quality Management of
Pavement Condition Data Collection, NCHRP Synthe-
sis 401, Transportation Research Board, Washington,
D.C.
[3] FHWA, “Pavement Structural Evaluation at the Net-
work Level”, DTFH61-12-R-00040, Statement of Work.
[4] Rada, G.R. & Nazarian, S. 2011. The State-of-the-Tech-
nology of Moving Pavement Deflection Testing. Final
Report, FHWA-DTFH61-08-D-00025, U.S. Depart-
ment of Transportation, Washington, D.C..
[5] Rada, G.R., Nazarian, S., Visintine, B.A., Sid-
dharthan, R., & Thyagarajan, S. 2015. Pavement
Structural Evaluation at the Network Level, Report
No. FHWAHRT-15-074, FHWA, U.S. Department of
Transportation, Washington, D.C.
[6] Shaw, W., Loudon, W, Schulte, R, & Liu, J. 2009. Visu-
alizing Long-Range Health of Rural Recreational and
Agricultural Corridors. InTransportation Research
Record: Journal of the Transportation Research Board,
No. 2119, Transportation Research Board of the
National Academies, Washington, D.C.
[7] AASHTO (1993). AASHTO Guide for Design of Pave-
ment Structures, American Association of State High-
way and Transportation Officials, Washington, D.C.
[8] Maser, K. & Schmalzer, P. 2016. East Idaho Loop Cor-
ridor, District 6 Subsurface Pavement Evaluation and
Substructure Database Development. Interim Report
submitted to the Idaho Transportation Department,
District 6, June.
[9] Schmalzer, P. & Weitzel, N. 2017. Methodologies for
Computing Pavement Deflection from Pavement Deflec-
tion Velocity, and Their Impact on Pavement Rehabilita-
tion Design. Paper submitted for the Proceedings of the
Figure 8. Evaluation of SH28 between Leodore and Tenth International Conference on Bearing Capacity of
Salmon. Roads, Railways, and Airfields, Athens, June.

594
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Methods for calculating deflections from traffic speed deflectometer

data and impacts on pavement design

P. Schmalzer & N. Weitzel

NCE, Reno, Nevada, USA

ABSTRACT: The Traffic Speed Deflectometer (TSD) is a relatively new device that is used to measure
a stiffness-related response of an in-place pavement to a moving wheel load. The TSD mounts a number
of doppler lasers that measure a ratio of the downward velocity of the pavement surface as it deflects due
to the vehicle load to the forward velocity the vehicle. This ratio is taken to be the slope of the deflection
basin at the measurement location with respect to distance from the point of load application. While
deflection can be determined through integration of the deflection slope curve, a boundary condition
must be determined. Existing methods for setting that boundary condition relate to the behavior of the
pavement system at points well away from the load, which is also the area where the measurement error
is highest. Therefore some methodologies also include modification of the raw deflection slope data to
better match a theoretical model.
This paper discusses and compares four methodologies for determining deflection from deflection slope
to data collected by a Falling Weight Deflectometer (FWD), including a method developed by Greenwood
Engineering (unpublished and unreproducible by the authors), two methods developed by Australian Roads
Research Board (one previously published and reproduced by the authors, the other not previously pub-
lished but reverse-engineered by the authors) and one developed by the authors (not published previously).
Methodologies are compared in terms of raw deflection, as well as subgrade modulus, effective pavement
structural number, and required overlay structural number computed according to the AASHTO 1993 pave-
ment design method. While maximum deflection and effective structural number are similar amongst the
methods, the outer deflections and hence subgrade modulus and required overlay vary more significantly.

1 INTRODUCTION does introduce the problem of determining deflec-

tion from deflection slope. While in principal it
Attempts to develop a device that measures pave- would appear to be easy to simply integrate the
ment deflection due to an actual moving vehicle deflection slope curve, as with any integration a
have resulted in the Rolling Wheel Deflectometer boundary condition must be provided.
(RWD) and Traffic Speed Deflectometer (TSD). Methodologies for computing deflection from
Both vehicles are superficially similar, and consist deflection slope have been developed by Green-
of a road tractor and an instrumented single-axle wood Engineering (the manufacturers of the TSD),
semi-trailer. The RWD directly measures pave- the Australian Roads Research Board (ARRB, the
ment deflection using triangulation lasers, similar owner of the only TSD currently in the United
in principal to the technique used by high-speed States), and the authors. In this paper these meth-
inertial profilers. Difficulties with this approach ods are discussed and the results are compared to
center on the very small magnitude of pavement FWD measurements, both in terms of raw deflec-
deflections relative to the pavement surface tex- tion data and in terms of analysis results when used
ture, roughness, movement in the laser mounting in the AASHTO 1993 pavement design process.
system, and vehicle vibration and movement.
The TSD instead measures pavement deflec-
tion velocity using doppler lasers. These lasers are 2 DATA SOURCES
mounted at slight angle from vertical, and there-
fore measure a vector sum of the vertical deflection This paper utilizes TSD data collected in eastern
velocity of the pavement surface and the forward Idaho by Greenwood Engineering between Sep-
velocity of the vehicle. The ratio of these veloci- tember 15 and 19, 2015 as part of the TPF 5(282)
ties is taken to be the slope of the deflection basin pooled fund project. Comparison FWD data was
at the measurement location. While this eliminates collected by Nichols Consulting Engineers, Chtd.
many of the difficulties with the RWD approach, it (NCE) between October 26 and 30 2105. Ground-

595
Penetrating Radar (GPR) was collected by Infra- Table 1. Summary of FWD comparison segments.
sense on August 12 through 15, with coring and
boring performed by American Geotechnics. Average thickness,
FWD, GPR and coring boring was performed on Postmile mm
FWD
behalf of the Idaho Transportation Department Segment Route Begin End AC CRABS AB
(ITD) as part of the Eastern Idaho Loop Corridor
(EILC) project. All data was reported along with 1 SR 33 59.5 61.4 94 185 178
GPS coordinates, which was used for alignment. 2 SR 33 28.7 30.4 188 216 –
The Greenwood TSD had six Doppler lasers 3 SR 33 0.3 2.2 97 208 –
to measure the pavement surface responses, which 4 SR 22 25.0 27.0 104 – 145
were offset 0.1, 0.2, 0.3, 0.6, 0.9, and 1.5 meters from 5 SR 22 44.1 46.1 104 – 135
the center of the wheel-load. The data was measured 6 SR 28 105.5 107.4 163 – 135
continuously and an average value was reported 7 SR 28 132.1 134.0 193 – 221
every 10 meters of travel by the TSD. In total, TSD 8 US 93 262.51 264.4 109 – 89
data was collected for approximately 830 kilometers 9 US 93 245.05 246.9 124 – 114
of pavement in Idaho. The pavements analyzed con- 10 US 93 130.17 132.1 208 – 74
sisted primarily of State Highways 22, 28, and 33, 11 US 93 96.9 98.8 71 318 –
and US Highways 20 and 93 within ITD’s District 6. 12 US 20 253.9 255.8 53 224 140
FWD testing was performed on 13 selected seg- 13 US 20 294.8 296.7 99 259 –
ments of the route tested by the Greenwood TSD,
as shown in Figure 1. Each segment was nominally
3050 meters long, and FWD tests were performed
3 FWD COMPARISON SECTIONS
at a 30 meter interval. The FWD used seven dif-
ferent sensors to measure the surface deflections,
The location of the segments with FWD compari-
which were offset 0, 0.20, 0.30, 0.41, 0.61, 0.91, and
son data is shown in Table 1. Table 1 also includes
1.52 meers from the center of the load. At each
the average pavement structure as determined by
location, two drops were performed and the aver-
GPR. Average values are shown for reference pur-
age value was reported.
poses—all analysis results presented in this paper
GPR testing was performed using a 1-GHz
were performed using point-by-point GPR and
horn antenna. GPR analysis results were reported
deflection data.
at a 10 meter interval, corresponding to the TSD
Six out of thirteen FWD comparison seg-
reporting intervals.
ments contain a layer known locally as Cement
This paper also utilizes TSD data collected in
Recycled Asphalt Base Stabilization or CRABS.
eastern Idaho by ARRB on June 17, 2016 as part of
This is a pulverized mixture of existing flexible
a different project. For the purposes of this paper,
pavement, mixed in-place with portland cement
this data was used solely for developing and check-
and re-compacted. This was a common reha-
ing the authors’ implementation of ARRB’s algo-
bilitation procedure in ITD District 6 up until
rithm. The ARRB TSD had seven Doppler lasers
approximately 5 years ago, although it has since
mounted at offsets of 0.110, 0.210, 0.308, 0.610,
fallen out of favor due to uneven performance.
0.910, 1.510 and 3.500. Data was collected con-
The authors have previously analyzed FWD data
tinuously, and the average reported every 16 meters
collected on CRABS material, and found it to
(0.01 miles). In total, data was collected on approxi-
be highly challenging due to high variability in
mately 140 km of I-15 within ITD’s District 6.
thickness and condition.

4 TSD DEFLECTION CALCULATION

METHODS

4.1 Greenwood Engineering

Greenwood Engineering (GE) supplied both raw
deflection slope values and computed deflection
values. Based on personal communication with
Jack Jorgensen of Greenwood, the methodol-
ogy used to compute those deflection values is
unpublished, but generally based on a visco-elas-
tic model developed by Louis Pedersen (Pedersen
Figure 1. Map of FWD comparison segments. 2013). This model is asymmetric, and accounts

596
for a maximum deflection that is located behind the 1.5 meter sensor), this rule typically has the
the tire load. Beyond that, the authors’ knowl- effect of forcing the slope of the outer portion
edge of the method is limited, however it appears of the deflection basin to be equal to a constant
to be based on fitting the measured deflection times the inverse of the measurement offset. For
slopes to a pre-computed database of typical the purposes of this paper, this will be referred to
deflection curves. as the ARRB2 algorithm.
Once the tail-taming algorithm was adjusted,
the authors were able to match the deflections
4.2 ARRB
supplied by ARRB for the I-15 project based on
ARRB employees have published a series of the raw deflection slopes provided by ARRB. The
papers on their development of an alternate ARRB1 and ARRB2 algorithms were then applied
method for calculation of deflections based on to the raw deflection slopes provided by Green-
TSD measurements (Wix, Murnane, and Mof- wood for the EILC project.
fatt 2016, Muller 2015, Roberts, Al, Toole, and
Martin 2014, Muller and Roberts 2013). The
4.3 NCE
basic concept described has been consistent, and
involves numerical integration of the deflection The authors developed an alternate approach for
slope curve, with the boundary condition con- computing deflections due to dissatisfaction with
sisting of zero deflection at a point 3.5 meters the comparison of the Greenwood and ARRB1
in front of the tire. Integration is performed by data to the FWD data, especially for the outer sen-
fitting the curve with cubic hermite splines, and sors necessary to compute subgrade modulus (the
then integrating those splines. ARRB2 algorithm was not known to the authors
Variations in the methodology over time appear at the time).
in the process that ARRB terms “tail-taming”. As The NCE approach is similar to the ARRB1
described in the most recent publication on the approach, except that the boundary condition is
subject, tail-taming consists of: set based on Boussinesq theory. A convinent form
of the Boussinesq equation for deflections in a
Limiting the magnitude of the 900 mm deflection homogeneous half-space is (SHRP 1993):
slope value to no more than two-thirds of the value
at the 600 mm laser eliminates the possibility of a P( − μ 2 )
E= (1)
flat or bulged curve (from 900 to 3500 mm) and def
e R
produces a more tapered typical deflection curve.
If the difference is greater, the tail is cut off and
where E is the modulus of the layer in Pascals, P is
excluded from the analysis.
the load in Newtons, μ is the Poisson’s ratio (typi-
cally assumed to be 0.5 for fine-grained materials),
The authors interpreted this as meaning that
def is the deflection in meters, and R is the distance
if the deflection slope at an offset of 900 mm is
from the center of the load plate in meters.
greater than two-thirds of the value at 600 mm,
While a layered pavement system is certainly
then the deflection slope at 900 mm will be set to
not a homogeneous half-space, if the subgrade is
two-thirds of the value at 600 mm. For the pur-
homogenous, then for locations further from the
poses of this paper, this will be referred to as the
point of load application than the thickness of the
ARRB1 algorithm.
pavement the surface deflections can be approxi-
However, in reviewing the results provided by
mated by a homogenous half-space, where the
ARRB the I-15 project, it was determined that
modulus of the half-space is equal to the subgrade
a very different process was in use, as the ratios
resilient modulus.
of the adjusted 0.910 meter and 1.510 meter
An obvious consequence of the Bousinesq equa-
deflection slopes to the 0.610 meter deflection
tion is that, where it is valid, deflection is inversely
slopes were typically equal to 0.670 and 0.404,
related to distance from the load application,
respectively. While in some cases they exceeded
and therefore deflection and deflection slope are
these ratios, the never were less than these ratios.
directly related. Taking the derivative of deflection
Therefore it is apparent that the current ARRB
with regards to offset in Equation 1 yields:
tail-taming algorithm is essentially the opposite
of the published one: the minimum value of
P( − μ 2 )
any slope measurement at an offset greater than E= (2)
0.6 meters is equal to the ratio of 0.6 to that off- − def
e R
set. As this minimum value was hit very often
in the data analyzed in this study (84,971 out of where def’ is the deflection slope in meters per
93,030 basins qualify for either the 0.9 meter or meter.

597
 Note that in the NCE method, the slope of the not consistent with static or quasi-static pave-
outer portion of the deflection basin is assumed to ment modeling.
vary with the inverse of the offset squared, whereas Differences between the ARRB1, ARRB2 and
in the ARRB method the slope is assumed to vary NCE methodologies are depicted graphically in
with the inverse of the offset. Figure 2.
The NCE approach starts with computing
the effective homogenous half-space modulus
for each reported deflection slope value using 5 TEMPERATURE CORRECTION
the derivative of the Bousinesq equation (Equa-
tion 2). The minimum of these values is taken As the FWD and TSD data were collected at dif-
to be the subgrade modulus. Then, the deflec- ferent times, all of the computed center deflec-
tion at 0.9 meters is computed using the sub- tions were corrected to 20°C (68°F), using the
grade modulus and Equation 1. Finally, the thickness-dependent temperature adjustment
remainder of the deflection base is established factors in Figures 5.6 (for granular base) and
using piece-wise integration. The specific inte- 5.7 (for CRABS base) of the AASHTO 1993
gration method used involves solving equations Guide for the Design of Pavement Structures
3 through 6 for each interval, starting from the (AASHTO 1993). Mid-depth asphalt tempera-
outermost sensor as ’i’, and solving the set of ture at the time of testing was estimated using
equations in turn for each sensor. The maximum air temperature, pavement surface tempearture
deflection is taken by solving equation 3, with and the time of testing, in accordance with
the parameters determined for the final interval, the BELLS3 method (Lukanen, Stubstad, and
and R = −0.100 meters. Briggs 2000).
Ri3 + bRi2 + cR
aR Ri + d ddeffi (3)
3 i
2
2bRi + c
2bR dde f ′ i (4) 6 PAVEMENT DESIGN CALCULATIONS
3 2
2bRi −1 + c
2bR dde f ′ i −1 (5)
i 1 In order to evaluate the practical effect of the
3 2
i 2 2bRi − 2 + c
2bR dde f ′ i − 2 (6) variations in computed deflections analysis was
performed in accordance with the AASHTO
Selection of −100 mm as the location of the 1993 Guide, including computation of subgrade
maximum deflection is based on a recommen- modulus, effective pavement modulus, effective
dation from Greenwood. The 0.9 meter sensor structural number, required structural number
was selected as the point to begin reconstruction for a 20 year design life, and the required over-
of the deflection curve, as the data at 1.5 meters lay structural number. All calculations were
appeared to be highly variable, and also included performed on a point-by-point basis (10 meter
a significant number of negative values (i.e. intervals for TSD data, 30 meter intervals for
deflections increasing with increasing distance FWD data).
from the load). It may be that measurement error
was excessive at the 1.5 meter sensor, or that
6.1 Subgrade modulus
influence from the tractor wheels became sig-
nificant at that location. It is also possible that The subgrade modulus (MR) values were calculated
these measurements are actually valid but just using the deflection values for each TSD analysis
method, and for the FWD deflections. The sub-
grade modulus values were calculated as shown in
Equation 7. Equation 7 was performed for each
sensor, and the minimum value was taken to be the
subgrade modulus.

P( − μ 2 )
M R ,i = (7)
def
efi Ri

SensorOffset(m) 6.2 Effective pavement modulus and

structural number
Figure 2. Depiction of TSD deflection calculation Effective pavement modulus was computed
methods. using Equation 8. This equation can not be

598
re-written in terms of effective pavement mod- standard deviation were determined on recom-
ulus, so an iterative solver was written for the mendations from the pavement design chapter in
equation. The subgrade resilient modulus used the ITD Materials Manual, and are based on the
in equation 8 was corrected using a 0.33 correc- functional classification of the various roadways
tion factor as recommended by the AASHTO (ITD 2015). The input parameters areshown in
guide. Table 2. As with the calculation of effective pave-
ment modulus, the subgrade modulus was cor-
Dmax 1 5 pa rected using a factor of 0.33.
⎧ 1− 1 ⎫
⎪ ⎪ log10 (W18 ) =
2
1 1+ t
⎨ + a
⎬ (8)
Ep ZR S0 + 9 36 log S R 1) − 0.20
l g10 (SN (10)
⎪ M R 1 + ( at 3 Ep
) 2 ⎪
⎩ MR
⎭ log ( ΔPSI S
)
+ 10 44..2 −1.5
+ 2.32 × log10 (M R ) − 17.0 (11)
where p is the load plate pressure in pascals, a is the 0.40 + 1094
NR +1)5.19
( SN
load plate radius in meters, t is the total pavement
thickness in meters, Ep is the effective pavement where W18 is the cumulative lifetime 8.2 tonne (18
modulus in pascals and MR is the sugrade modulus kip) equivalent single axle loads, ZR is the reliabil-
in pascals. ity, S0 is the standard deviation, SNR is the required
Effective structural number was computed from structural number, ΔPSI is the design serviceability
the effective pavement modulus and pavement loss and MR is the subgrade resilient modulus in
thickness as shown in Equation 9. Pavement thick- pascals.
ness was determined from the GPR data for each The pavement design inputs used for each analy-
TSD deflection basin. sis segment are shown in Table 2.

Nefff = 0 0093 × t
SN EP (9)
6.4 Required overlay
where t is the total pavement thickness in meters The required overlay structural number was cal-
and EP is the effective pavement modulus in pascals. culated by simply subtracting the current effective
pavement structural number from the required
structural number to carry the projected future
6.3 Required structural number 20-year traffic, using the following equation:
The required structural number calculation was
performed using the flexible pavement equation NO = SNR − SN
Nefff (12)
from the AASHTO Guide, as shown in Equa-
tion 12. The ESALs are for a 20-year design life where SNO is the structural number of the required
and were calculated from traffic data provided by overlay, SNR is the structural number required for
ITD. The change in pavement PSI, reliability, and 20 years of additional traffic, and SNeff is the effec-
tive structural number of the existing pavement.
The thickness of the required overlay in millime-
ters can be determined by multiplying SN0 by 57.7
Table 2. Summary of pavement design inputs.
(which is equivalent to a layer coefficient of 0.44 in
FWD segment W18 ZR S0 ΔPSI US customary units),

1 238,000 90% 0.45 2.5

2 238,000 90% 0.45 2.5 7 RESULTS
3 238,000 90% 0.45 2.5
4 2,247,000 90% 0.45 2.5 Comparisons of the four TSD analysis method-
5 2,247,000 90% 0.45 2.5 ologies with FWD data are shown in Tables 3
6 1,279,000 90% 0.45 2.5 through 7, for maximum deflection, deflection at
7 1,279,000 90% 0.45 2.5
0.9 meters, subgrade resilient modulus, effective
8 1,451,000 90% 0.45 2
structural number and required overlay structural
9 1,451,000 90% 0.45 2
number, respectively. Deflection values shown
10 1,292,000 90% 0.45 2
in Tables 3 and 4 for both TSD and FWD were
11 1,292,000 90% 0.45 2
load-normalized to 40 kN to allow direct com-
12 2,975,000 90% 0.45 2
13 2,975,000 90% 0.45 2
parison, as testing was performed at slightly dif-
ferent loads.

599
Table 3. Comparison of maximum deflections.

Average maximun deflection, μm Absolute error with FWD,%

FWD segment GE ARRB1 ARRB2 NCE FWD GE ARRB1 ARRB2 NCE

1 149 136 178 180 222 33 39 20 19

2 155 150 146 151 203 24 26 28 26
3 361 346 436 451 433 17 20 1 4
4 321 318 399 424 446 28 29 11 5
5 405 377 519 508 558 27 32 7 9
6 491 461 553 557 375 31 23 48 49
7 191 177 224 224 212 10 17 6 6
8 354 341 424 429 393 10 13 8 9
9 203 218 243 256 225 10 3 8 14
10 165 179 206 223 171 4 4 20 31
11 307 284 361 381 393 22 28 8 3
12 140 152 180 180 192 27 21 6 6
13 209 201 247 257 306 32 35 19 16
Average Error: 20 21 14 15

Table 4. Comparison of deflections at 0.9 meters.

Average 0.9 meter deflection, μm Absolute error with FWD,%

FWD segment GE ARRB1 ARRB2 NCE FWD GE ARRB1 ARRB2 NCE

1 38 38 70 65 79 52 52 11 19
2 53 64 54 53 82 35 21 34 36
3 22 37 107 80 71 70 48 50 12
4 13 29 94 67 59 77 50 60 14
5 41 36 157 106 121 66 70 30 13
6 112 96 174 145 129 13 25 35 12
7 49 47 85 78 82 40 43 3 4
8 46 55 122 89 91 50 39 34 2
9 21 46 65 51 52 59 11 26 1
10 19 48 60 48 48 61 0 26 0
11 45 38 106 89 103 56 63 3 13
12 21 34 60 48 57 63 41 7 15
13 30 40 74 61 90 66 56 18 32
Average Error: 55 39 27 13

Table 5. Comparison of computed subgrade moduli.

Average subgrade modulus, mPa Absolute error with FWD,%

FWD segment GE ARRB1 ARRB2 NCE FWD GE ARRB1 ARRB2 NCE

1 389 596 177 197 136 185 337 30 44

2 430 266 251 276 129 234 106 95 114
3 361 694 113 148 120 202 480 5 23
4 498 820 127 169 143 249 474 11 18
5 197 498 70 105 78 152 538 10 34
6 103 150 65 81 83 25 82 21 2
7 303 471 139 157 133 128 255 5 19
8 252 383 104 136 110 129 247 6 24
9 570 453 190 229 185 209 145 3 24
10 724 509 262 263 218 232 134 20 21
11 277 483 124 146 104 167 365 19 41
12 613 632 190 262 188 226 236 1 39
13 437 592 170 204 115 280 415 48 78
Average Error: 186 290 20 36

600
Table 6. Comparison of effective structural number.

Average effective structural number Absolute error with FWD,%

FWD segment GE ARRB1 ARRB2 NCE FWD GE ARRB1 ARRB2 NCE

1 6.02 5.99 8.41 6.23 5.39 12 11 56 16

2 4.98 5.22 5.23 5.37 4.55 9 15 15 18
3 2.42 2.46 2.78 2.58 2.61 7 6 7 1
4 1.7 1.68 2.2 1.86 1.85 8 9 19 1
5 1.79 1.62 2.27 1.91 2.01 11 19 13 5
6 2.36 2.34 2.53 2.36 2.72 13 14 7 13
7 4.16 4.18 4.37 4.4 4.08 2 2 7 8
8 1.84 1.84 2.33 2.08 2.1 12 12 11 1
9 2.6 2.7 3.06 2.74 2.77 6 3 10 1
10 2.82 3.14 3.3 2.9 3.22 12 2 2 10
11 3.19 3.11 3.42 3.23 3.18 0 2 8 2
12 4.29 4.17 4.59 4.4 4.23 1 1 9 4
13 3.97 3.98 4.21 4.03 3.78 5 5 11 7
Average Error: 7 8 10 6

Table 7. Comparison of design overlay thickness.

Average required overlay, mm of AC Absolute error with FWD,%

FWD
segment GE ARRB1 ARRB2 NCE FWD GE ARRB1 ARRB2 NCE

1 0 0 0 0 1
2 6 5 6 7 11 42 53 47 37
3 40 36 85 77 87 54 59 3 11
4 53 54 103 100 112 53 52 8 11
5 96 77 142 131 148 35 48 4 11
6 93 83 115 109 87 8 4 33 26
7 2 1 3 2 3 50 67 17 33
8 84 78 114 107 115 27 33 1 7
9 18 26 31 32 39 54 33 19 18
10 7 9 23 23 19 64 55 18 21
11 20 11 46 43 58 66 81 21 26
12 1 2 6 3 9 88 81 31 69
13 6 6 18 14 44 87 87 58 67
Average Error: 56 58 22 28

8 DISCUSSION The ARRB1 and GE methods do significantly

less well in matching the raw deflection basins,
The authors developed the NCE method due to however the effective SN calculation is very good
dissatisfaction with the GE and ARRB1 meth- and on par with the NCE and ARRB2 methods.
ods, particularly with their matching of the outer This is because these methods typically over-esti-
portion of the deflection basin and resultant sub- mate the subgrade modulus, and under-estimate
grade moduli. During the writing of this paper, center deflection. As can be seen from Equation 10,
the authors became aware of the ARRB2 method, these errors will tend to compensate. Another way
which, although developed independently and of looking at this is that effective SN is primarily
apparently without consideration of same theory a function of the curvature of the inner portion
as the NCE method, is effectively quite similar. of the deflection basin, and not very sensitive to
Depending on the parameter of interest, either the boundary conditions that shift the entire curve up
NCE or ARRB2 method is marginally superior or down. However these methods do significantly
when applied to this data set. under-estimate the future required SN as well as
601
the required overlay SN, as they significantly over- Australian Roads Research Board and Infrasense
estimate the subgrade modulus. for sharing their knowledge.
Previous studies of TSD deflections have pri-
marily focused on parameters such as maximum
deflection or SCI300, which are largely functions REFERENCES
of the pavement stiffness itself. Investigation of the
outer portion of the deflection curve, which is pri- AASHTO (1993). AASHTO Guide for the Design of Pave-
marily a function of the subgrade stiffness, have ment Structures. Washington, D.C.: American Associa-
been neglected by comparison. This is despite the tion of State Highway and Transportation Officials.
fact that both pavement modulus and subgrade ITD (2015). Materials Manual. Boise, Idaho: Idaho
Transportation Department.
modulus are required to evaluate the adequacy of Lukanen, E., R. Stubstad, and R. Briggs (2000). Temper-
an existing pavement and to design a rehabilitation ature predictions and adjustment factors of asphalt
treatment. pavement. Technical report, FHWA, U.S. Department
Further investigation of the deflection response of Transportation, Washington, D.C.
of the portions of the pavement further than Muller, W. (2015). A comparison of tsd, fwd and gpr
0.9 meters ahead of the tire is warranted. Given field measurements. International Symposium on
limitations of legal axle loads and doppler laser Non-Destructive Testing in Civil Engineering, Berlin,
technology, it may be that deflections in those Germany.
locations arebelow the noise threshold for typi- Muller, W. and J. Roberts (2013). Revised approach to
assessing traffic speed deflectometer data and field
cal US highway pavements. It may also be that validation of deflection bowl predictions. International
deflections in those locations are influenced by the Journal of Pavement Engineering 14, No. 4, 388–402.
tractor loading. However it may also be that the Pedersen, L. (2013). Viscoelastic Modeling of Road
assumption of quasi-static behavior that underlies Deflections for use with the Traffic Speed Deflectom-
the NCE and ARRB2 methodologies is violated in eter. Ph. D. thesis, Technical University of Denmark,
that region. Lyngby, Denmark.
Roberts, J., U. Al, T. Toole, and T. Martin (2014). Traffic
speed deflectometer: Data review and lessons learnt.
ACKNOWLEDGEMENTS Austroads APT279–14.
SHRP (1993). Analysis of section homogeneity, nonrep-
resentative test pit and section data, and structural
The authors wish to acknowledge the Idaho Trans- capacity. Technical report, National Research Coun-
portation Department (ITD) for supporting this cil, Washington, D.C.
work, and the FHWA for making the TSD avail- Wix, R., C. Murnane, and M. Moffatt (2016). Experience
able to the project through ITD’s participation in gained investigation, acquiring and operating the first
the pooled fund project, TPF 5(282). The authors traffic speed deflectometer in australia. Transportation
also wish to acknowledge Greenwood Engineering, Research Procedia 14, 3060–3069.

602
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Quality control tool for asphalt emulsion-based chip seal curing times

M.A. Montoya & J.E. Haddock

Lyles School of Civil Engineering, Purdue University, West Lafayette, IN, USA

W.J. Weiss
School of Civil and Construction Engineering, Oregon State University, Corvallis, OR, USA

ABSTRACT: Due to its ease of use, favorable economics and effectiveness, asphalt emulsion-based
chip seals are one of the most widely used preventive maintenance treatments for pavements. The main
advantage of applying emulsified asphalt is the low-temperature and low-volatile emission application.
However, one inherent concern of asphalt emulsion-based chip seals is curing time. This paper discusses
the potential of using electrical resistance measurements as a quality control tool to manage or reduce the
uncertainties related to asphalt emulsion curing, and also as a means to ensure quality of construction.
Furthermore, the implementation of such an approach has great potential as an early-life performance
tool for diverse asphalt emulsion applications.

1 INTRODUCTION however to a great extent the actual curing time

required by the asphalt emulsion to gain sufficient
Since the 1920 s chip seals have been used to pro- mechanical strength is uncertain. The actual cur-
vide cost-effective wearing surfaces. The early uses ing time depends on the formulation of the asphalt
were predominantly as wearing courses in the con- emulsion and, several uncertain factors, such as the
struction of low-volume gravel roads. Over the past climatic conditions at the construction site.
75 years, chip seals have evolved into maintenance A key aspect for the success of any construc-
treatments that can be successful for both low- tion project is to manage or reduce uncertainty.
volume and high-volume pavements (Gransberg & Therefore, to improve chip seal performance, field
James, 2005). However, the last two decades have inspectors should measure in-situ chip seal curing
seen growing popularity among highway transpor- times. Previous work by the present authors has
tation agencies to use chip seals as a pavement pres- demonstrated that electrical resistance measure-
ervation technique. The popularity of chip seals is a ments can be effectively used to quantify chip seal
direct outcome of their low initial costs in compari- curing times, and thus determine the optimal time
son with those of thin asphalt overlays and other to sweep and open the newly placed chip seal to
preservation options (Gransberg & James, 2005). unrestricted traffic (Montoya et al. 2017). How-
Consequently, chip seals have become increas- ever, this paper discusses the implementation of
ingly important in USA pavement preservation electrical resistance measurements as a quality
programs, leading to seemingly constant revision control tool of chip seal curing times to ensure
by many state and local highway agencies of both the project success. Additionally, this paper sug-
maintenance policies and construction specifica- gests how a similar measurement technique can be
tions to improve chip seal performance (Mahoney employed as an early-life performance assessment
et al. 2014). Transportation officials have estab- tool for diverse asphalt emulsion applications.
lished on-going chip seal research programs with
the purpose of delivering successful and longer-
lasting surface treatments (Cole & Wood, 2014). 2 IMPORTANCE OF CHIP SEAL
The goal is to develop new methodologies that CURING TIMES
contribute to lengthening the service life of chip
seal treatments and thus maximize the available In order to deliver reliable and durable chip seal
funding for pavement preservation. projects, transportation officials must rely on chip
A potential area for chip seal performance seal best practices. By implementing sound con-
improvement is to measure or reduce the actual struction techniques, chip seal average life expect-
curing time of fresh chip seals at the construction ancy, which is typically between 6 to 8 years, can
site. Curing times range from two to four hours, be prolonged to 12 or even 15 years (Cole & Wood,

603
 fied) and CRS-2P (cationic, rapid-set, polymer-
Chip Seal Success, Performance and Life modified) emulsions. Table 1 shows the asphalt
t<xp•ctoncy emulsion and cover aggregate used at each chip
seal project location.
“SC” is the Indiana Department of Transpor-
Hest. l-lrnc:tic:es tation’s (INDOT) designation for aggregate that
has been produced primarily for use in chip seal
applications (INDOT, 2015). The gradation corre-
<:onstmction 1-'h:Jse sponding to each designation is shown in Table 2.
Figure 2 shows the 7.6 cm two-point probe elec-
trical resistance measurement configuration used to
quantify chip seal curing times. Two probes (or steel

Table 1. Asphalt emulsion and cover aggregate used at

each pavement section.

Pavement section Asphalt emulsion Cover aggregate

Figure 1. Importance of asphalt emulsion curing time SR 352 Oxford AE-90S SC 16 Limestone
on chip seal project success. SR 38 Kirklin AE-90S SC 16 Gravel
US 52 Brookville CRS-2P SC 11 Dolomite
US 52 Metamora CRS-2P SC 11 Dolomite
2014). Chip seal best practices have been identified SR 827 Angola AE-90S SC 16 Dolomite
as a set of maintenance policies, design specifica-
tions, and construction techniques that determine
the service life of a chip seal project. Gransberg & Table 2. Aggregate gradation.
James (2005) grouped chip seal best practices into
four main categories: (1) contract administration, Percent passing %
warranties, and performance measures, (2) pave-
ment selection, design, and material selection, (3) Sieve size mm SC 11 SC 16
construction and, (4) chip seal equipment and,
quality assurance and quality control. 12.5 100 100
Any of these four best practices can positively or 9.50 75–95 94–100
negatively impact the success of a chip seal project. 4.75 10–30 15–45
2.36 0–10
However, the quality and performance of chip seals
1.18 0–4
during their service life are driven mainly by the
0.60
construction phase (Gransberg & James, 2005).
Decant 0–1.5 0–1.5
Furthermore, if the material and equipment best
practices are satisfied and the climatic conditions
are favorable, the success of the chip seal comes
down to the timing of the construction process
(NCDOT, 2015). As a result, the chip seal curing
time is a critical input element to deliver longer-
lasting surface treatments, as presented in Figure 1.

3 FIELD IMPLEMENTATION

In 2016, electrical resistance measurements were

implemented at five different chip seal projects in
the state of Indiana, USA. The methodology was
implemented on SR 352 (approximately 2 miles
west of Oxford), SR 38 (approximately 1 mile east
of Kirklin), at two locations on US 52 (one near
Brookville and one near Metamora), and on SR
827 (approximately 3 miles north of Angola). The
measurements were conducted at chip seal projects Figure 2. Two-point probe electrical resistance meas-
using AE-90S (anionic, rapid-set, polymer-modi- urement configuration.

604
 100
e SR 352 Oxford
c Mass Loss Data fl. SR 38 Kirldin • •

80 - -Regression Model •
0
US 52 Brookville
US 52 Metamora
~~
•
X SR 827 Angola
- - - - NRI = 10
:%
:
:.
••••••• MCR 15% - 25% X· ~.ill

-----------~~~~--
,.-.,
60
::R
0

~
u
~ 40

20
• JSK.
~oe
<s><41
..
'4: ~
:
0 I ~--~~. .r--,-----,----~~--~
0 2 3 4 5 6 100 80 60 40 20 0
Curing Time (b) 1\IICR(%)

Figure 6. Plate sample mass as a function of curing Figure 7. Normalized resistance Indices as a function
time for US 52 Brookville (Specimen 1). of moisture content ratio.

the curing times are within or closed to the criti-

mt mf (2) cal moisture content range suggested by Shuler
C =
MCR ×100%
mi mf (2011), between 15 and 25 percent. This range
of remaining moisture corresponds to sufficient
where MCR = moisture content ratio (%); binder adhesive strength to allow sweeping and
mt = plate sample mass at any given curing time traffic on the newly placed surface (Shuler, 2011).
(g); mi = initial plate sample mass (g); and mf = final These findings validate the effectiveness of using
plate sample mass (g). electrical resistance measurements for the quality
control of chip seal curing times.

6 CHIP SEAL CURING TIMES

7 ELECTRICAL RESISTANCE
Table 3 presents the curing times that were deter- MEASUREMENTS FOR QUALITY
mined for each pavement section using electrical CONTROL
resistance measurements.
Implementation of the methodology at full- From an asphalt emulsion product performance per-
scale chip seal systems shows that curing times for spective, the measured curing times presented are in
the projects ranges from three and a half to four substantial contradiction with references in the liter-
hours. This is in good agreement with chip seal ature that suggest chip seal curing times are reduced
construction guidelines in the USA that suggest to one and one-half hours for polymer-modified
that asphalt emulsions in chip seals gain adequate asphalt emulsions, such as AE-90S and CRS-2P
mechanical strength within two to four hours after (Testa & Hossain, 2014). This disagreement high-
sealing (Caltrans, 2014). As shown in Figure 7, lights the importance of measuring the actual curing
time of each chip seal project, as several changeable
variables can delayed the curing process.
Table 3. Actual chip seal curing time at each pavement Considering that an aggressive quality control
section. testing program, combined with close inspection,
can contribute to a chip seal project’s success (Grans-
Curing Moisture content % berg & James, 2005), the use of electrical resistance
time measurements can serve as a quality control tool
Pavement section h Specimen 1 Specimen 2 for manufacturing and performance acceptance of
asphalt emulsion products. Also, this measurement
SR 352 Oxford 4.0 19.0 22.0
technique can help to select asphalt emulsion-cover
SR 38 Kirklin 3.8 21.0 19.0
aggregate combinations that ease the curing proc-
US 52 Brookville 4.0 16.0 15.0
ess, and thus improve chip seal performance.
US 52 Metamora 3.5 20.0 22.0
Implementing this methodology as a quality
SR 827 Angola 4.0 26.0 28.0
control tool could potentially ensure quality of

606
materials used on the project, prevent minimal mulation of the asphalt emulsion. However, sev-
windshield claims and chip seal repair work, pre- eral changeable factors can significantly impact
vent unnecessary construction delays, provide the curing time required for the emulsion to gain
safety for the public and construction workers and ample mechanical strength. The use of electrical
ensure a successful chip seal project (Davis, 2005). resistance measurements can provide a quality
control tool to manage or reduce the uncertainties
related to asphalt emulsion curing, and thus ensure
8 QUALITY CONTROL TOOL FOR the chip seal project success. Additionally, the
DIVERSE ASPHALT EMULSION implementation of this methodology can provide
APPLICATIONS quantitative data to formulate faster curing asphalt
emulsions. By reducing asphalt emulsion curing
Asphalts in their emulsified form are widely used times, chip seal performance and life expectancy
for road construction and maintenance, but there could be improved. Additional field trials carried
is still need for testing that would tie together the out by highway agencies from around the world
asphalt emulsion products with their final applica- would be interesting to validate the findings of this
tion for performance (Kadrmas, 2006). study using different asphalt emulsions and cover
Uncertainties about asphalt emulsion work- aggregates under various climatic conditions. Pilot
ability and the development of early mechanical projects should be performed and analyzed using
strength are still an inherent concern for any asphalt engineering-based and qualitative performance
emulsion application, thus limiting the wider use indicators in order to optimize the construction
of this paving material. Asphalt emulsions require technique. Furthermore, the use electrical resist-
minimum viscosity to prevent run-off, but the ance measurements as a quality control or early-
workability of the emulsified bitumen also must life performance assessment tool can be extended
permit spraying, placement, and compaction using to various other asphalt emulsion applications.
conventional construction equipment. Asphalt
emulsion properties such as viscosity, homogene-
ity, stability and final residue directly affect work- ACKNOWLEDGEMENTS
ability, and ultimately binder performance. These
characteristics of asphalt emulsion need to be This research work was supported by the Indiana
properly identified in the field to ensure the appro- Department of Transportation (INDOT) and
priateness of the emulsion for a specific use. Joint Transportation Research Program (JTRP).
As an alternative, electrical resistance measure- The contents of this paper reflect the views of the
ments show great potential to serve as a quality authors, who are responsible for the facts and the
control tool for asphalt emulsion products. Future accuracy of the data presented herein. The con-
work should aim to establish whether handheld tents do not necessarily reflect the official views
electrical devices could be employed to determine or policies of the sponsors. These contents do not
asphalt emulsion properties during construction. constitute a standard, specification, or regulation.
The use of electrical resistance measurements as a
quality control and early-life performance assess-
ment tool for asphalt emulsion paving applications REFERENCES
would be extremely beneficial. For example, it can
assist in determining the curing time required for Caltrans Division of Maintenance (Caltrans). 2014. Cal-
emulsified asphalt tack coats and thus maximize trans Maintenance Technical Advisory Guide (MTAG),
the cohesion between pavement layers. Another Chapter 7: Chip seals. Sacramento: Caltrans.
Cole, M.K. & Wood, T.J. 2014. Effective Use of Chip
potential use of electrical resistance measurements Seals in Minnesota. TR News 290: 51–52.
is to evaluate the readiness for traffic and surfacing Davis, L. 2005. Quality control and quality assurance on
of cold recycled pavements. A field measurement chip seal projects. First national conference on road-
technique that characterizes the actual material way pavement preservation; 31 October-1 November
being used for paving will guarantee the quality 2005. Kansas City: Transportation Research Board,
and performance of asphalt emulsion products. National Research Council.
Gransberg, D. & James, D.M. 2005. Synthesis 342: Chip
Seal Best Practices, A Synthesis of Highway Practice.
9 CONCLUSION Washington, D.C.: Transportation Research Board,
National Research Council.
Kadrmas, A. 2006. Emulsion Test Methods, Do We Need
This paper has discussed the use of the electrical Them? Transportation Research Circular E-C102:
resistance measurement as a quality control tool Asphalt Emulsion Technology. Washington, D.C.:
for chip seal curing times. The actual curing time Transportation Research Board, National Research
of fresh chip seals depends primarily on the for- Council.

607
Mahoney, J.P., Slater, M., Keifenheim, C., Uhlmeyer, J., Shuler, S. 2011.When to Broom or Remove Traffic Con-
Moomaw, T. & Willoughby, K. 2014. WSDOT Chip trol Safely on Fresh Emulsified Asphalt Chip Seals.
Seals—optimal timing, design and construction consid- Journal of the Transportation Research Board 2235:
erations. Olympia: Washington State Department of 82–87.
Transportation. Testa, D.M. & Hossain, M. 2014. Kansas Department of
Montoya, M.A., Weiss, W.J., & Haddock, J.E. 2017. Transportation 2014 Chip Seal Manual. Topeka: Kan-
Using Electrical Resistance to Evaluate the Chip Seal sas State University Transportation Center, Kansas
Curing Process. Journal of the Association of Asphalt Department of Transportation.
Paving Technologists, Vol 86.
North Carolina Department of Transportation
(NCDOT). 2015. Chip Seal Best Practices Manual.
Raleigh: NCDOT.

608
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

F/HWD international round robin tests on the STAC’s test facility

M. Broutin & S. Belon

French civil Aviation Technical center, Bonneuil-sur-Marne, France

ABSTRACT: The French civil Aviation center (STAC) organized in October 2013 an international
Falling and Heavy Weight Deflectometer (F/HWD) round robin survey on its test facility (Bonneuil-
sur-Marne, France) in order to assess interchange ability of these pavement testing devices. This survey
involved seven (7) devices from four (4) countries. Each device was first calibrated using the STAC’s preci-
sion dynamic weighing system. Crossed tests between the different devices were then performed on several
tests points positioned over the flexible pavement of the test facility. All devices were tested in turn repeat-
edly on each point. For each measurement cycle, each device, on each test point, several test sequences
were performed at several load levels. A thorough analysis was performed using the large collected data
base. A repeatability study using the different drops from a same fall height was first carried out for all
devices and a representative drop was selected. A linearity study was then processed using results from the
different load levels which allowed standardizing results relative to applied load. Next step consisted in
studying the effect of temperature by comparing for a given device on a given test point the results from
the different measurement cycles, allowing standardizing the results with respect to temperature. Then,
the reproducibility was analyzed by comparing the standardized results from the different devices on a
given test point and a given measurement cycle. Finally, this parameter was compared with the discrepan-
cies between results from a given device when considering the different test points.

1 OBJECTIVE AND SCOPE

The first European F/HWD round robin tests were

organized by the STAC in October 2013 on its
test facility, Bonneuil-sur-Marne, France (STAC,
2012). Seven (7) devices from four (4) countries
were involved.
The aim was to evaluate repeatability, repro-
ducibility of these pavement testing devices, and
to assess their interchange ability by studying the
scattering of the results from several test points
performed on a homogeneous pavement.
Each device was first calibrated using the STAC’s
precision dynamic weighing system. Crossed tests Figure 1. Exploded view of the STAC’s precision
between the different devices were then performed weighing system.
on several tests points positioned over the flexible
pavement of the test facility.
elements, and is intended to withstand 600 kN
(60 metric tons) impact loads. The measurement
is made by three precision force sensors, which
2 LOAD MEASUREMENTS
ensure an isostatic system. This was validated by
RELIABILITY ASSESSMENT
an experiment whose aim was to study the influ-
ence of the centering of the load plate over the
2.1 Comparison with a precision dynamic
weighing scales. It is embedded in a reinforced con-
weighing system
crete basement.
The STAC’s precision weighing system shown in All F/HWD measured force signals were veri-
Figure 1 was used for F/HWD force measurements fied by performing three series of tests on the
assessment. This device which was worked out and weighing system. A 125 kN target load was cho-
designed by the STAC is composed of special steel sen for the FWD whereas multi-height tests (three

609
 &
QJ
~ MeanvalueF=250
+----+--+----~------- kN
"G
-.; ~ • Meanvaluef=125
r~ 4% +---'~--++---------c :-------- kN

I
~~ 2%
-
1 70

r~ Dispertion

App2 Ap 3 App 4 AppS App6 App 7 App 8

Figure 4. Comparison between force signals peak

value delivered by the F/HWD and the STAC’s weighing
Figure 2. F/HWD tests on the STAC’s precision system.
dynamic weighing system.

Table 1. Force signal peak values dispersion.

300
254,21
250

200

~ 150
~ o Scale
..... 100
• HWD Table 2. Calculated force correction coefficients.
50

58 36 58,37 58,38 58,39 58,4 58,41 58,42

-50
t[s]

Figure 3. Example of comparison between force 1,07 - , - - - - - - - - - - - - - - - - - - - - - - - - - -

signal histories measured by HWD and the STAC’s
weighing system.

:~
~ 1,05
.....

~
1,06

1,04
+ - - - - - - - - - - - - - - - - - -.......--=D

---
- - -
-
-
:....:e:.. v....:

- -
.:; 2o_

____ - - _ _.... ...

_ - --· ..
- Dev o

Dev3
~ 1,03 1 ~ :...:. - · .--- .. • ·A
250 kN drops plus three 125 kN drops) were per- 8 1,o2 "".,_
::.___ _ _ _ _ _ _ _ _ _ _ _ _• -_ _____,.----;;;;o_..,
~ '~'<-8ev-8-
formed for HWD.
Figure 3 presents an example of results for one of the ·~ 1 ~ 01 t=====~;::;::~~~;:::..:_:=--;;;;;;;;;;;;=;;;;..,..~Sr:;Tr;;A--rC:--
~ 1,00 "1'~
'---·_- _ ·_-_ _• _ _ _ _ _ _ _ _ _ _ _ _ _ _ __
three 250 kN drops, for the STAC’s HWD. 0
u 0,99 -+----------------------
A synthesis of the results relative to all devices
when considering peak values is presented in Figure 4. 0,98 +----------------------
0,97 +---------------------,------
125 250

2.2 Dispersion of the results HW D load [kN]

The plus or minus twice the standard error scatter- Figure 5. Comparison between force signals peak val-
ing bars are depicted in Figure 4. Table 1 recaps the ues delivered by the F/HWD and the STAC’s weighing
corresponding values. system.
It appears that all values are within the STAC’s
pavement testing guidance (STAC, 2014) require-
ments (uncertainty < 2%) or close to them, except Let α be the correction coefficient relative to
for one device (Dev. 3) which will not be consid- HWD values:
ered in the following analyses.
1
α (i , k ) = (1)
2.3 Deviation to true value and correction 1 − ΔFmax (i, k )
The results given by the high precision (uncertainty
< 0.01%) STAC’s weighing system are assumed to with i and k respectively the considered F/HWD
represent the true values. device and load level.

610
 True peak force value corresponds actually to: results from a given device when considering the
different test points.
1 These analyses were carried out on the basis of
l =
Fmax scales Fmax (ii k Fmax HWD
1 − ΔFmax (i, k ) the corrected force measurements (see 2).

(2) 3.2 Repeatability study

Table 2 recaps α(i, k) values. For each fall height, from each test, three drops
A linear interpolation is applied for HWD were performed. The repeatability of the devices is
loads between both extreme levels as presented in studied by comparing the normalized deflections
Figure 5. obtained, i.e. the ratio deflection/applied load, for
Corrected force data will be considered in the the three drops.
next analyses. This analysis is performed for all tests. A syn-
thesis is presented in Figure 7 which depicts the
variation coefficient mean values obtained for
all F/HWD device, for each deflection, d1 cor-
3 ROUND ROBIN TESTS
responding to central geophone G1 to d9 corre-
sponding to outer geophone G9 positioned 2.10 m
3.1 Scope and research plan
from plate center.
The main goal was to perform crossed tests between A good repeatability, within the STAC’s guid-
the different devices over the flexible pavement of ance tolerances (uncertainty < 2%), is observed,
the test facility, assumed to be perfectly homo- which makes the results reliable. For each test a
geneous in terms of pavement layer thicknesses representative drop is chosen which corresponds to
and materials properties, and to assess the inter- the closest to the average deflection basin in terms
changeability of F/HWD devices by comparing of root mean square minimization.
the reproducibility with the discrepancies between
the results from different test points. 3.3 .Linearity
Three test points were defined. All devices were
tested in turn repeatedly (three passes) on each The deflection linearity against applied load is then
point. For each measurement cycle, each device, verified. This verification is a key step since a good
on each test point, several test sequences were per-
formed at several load levels. The test sequence
was for the HWD 3 × 250 kN + 3 × 200 kN + 3
× 150 kN + 3 × 125 kN + 3 × 100 kN + 3 × 75 kN
+ 3 × 250 kN, and for the FWD 3 × 125 kN + 3 × 10
0 kN + 3 × 75 kN + 3 × 125 kN, a 45 cm 4-split plate
was used and the distances to center were respec-
tively 0, 0.30, 0.40, 0.60, 0.90, 1.20, 1.50, 1.80 et
2.10 m for geophones 1 to 9.
The full research plan is available in (STAC, 2015).
A thorough analysis was performed using the
significant collected data base.
First step consisted in preliminary studies
needed before the reproducibility and interchange-
ability studies to be achieved. A repeatability study
using the different drops from a same fall height Figure 6. The F/HWD on the STAC’s test facility.
was first carried out for all devices and a repre-
sentative drop was selected. A linearity study was 2,5%

then processed using results from the different load 2,0%

levels which allowed standardizing results relative
to applied load. Next step consisted in studying ~
1,5% -
• 250 kN
the effect of temperature by comparing for a given I~
1,0% - -
device on a given test point the results from the dif- • 125 kN

ferent measurement cycles, allowing standardizing

the results with respect to temperature.
Finally, the reproducibility was analyzed by
comparing the standardized results from the dif-
0,5%

0,0% ~ ITI I
d1 d2 d3 d4
I It
d5 d6 d7
-

d8
-

d9

ferent devices on a given test point. This param- Figure 7. Variation coefficient mean value for normal-
eter was compared with the discrepancies between ized deflections.

611
linearity allows normalizing results with regards ggg jgg 0 QQR 0 ggg n QQR
1,000
to the applied load. All attendees tried actually to
0,995
perform their tests with an applied load as close
as possible to the target load, within the tolerance i 0,990
0,991

~ 0,985
defined in the research plan, but the only way to
8 0,980
study comparable data is to normalize them.
Each test comprises six load levels for HWD ·~ 0,975
0,97 1
and three ones for FWD. ·~ 0,970

~ 0,965
An example of results is presented in Figure 8
where a very good linearity is observed. 0,960

The same work is done for all tests. The deter- 0,955
STAC Dev 2 Dev 3 Dev 4 Dev 5 Dev 6 Dev 7 Dev 8
mination coefficients (R2) mean values obtained
for geophone 1 to 9 when considering all STAC’s Figure 10. Linearity study—Average determination
HWD tests are provided in Figure 9 and the aver- coefficients, all F/HWD.
age values for all geophones in Figure 10.
A good linearity is demonstrated (except for
Dev. 3). The linearity criterion derived from the
STAC’s guidance [Ref] gives R 2 ≥ 0.973 for FWD
(three loads levels) and R 2 ≥ 0.987 for HWD
(6 load levels).
As a consequence, the normalization of the
deflections against applied load is possible and will
be used for the following further analyses.

3.4 Elastic behavior

The pavement elastic behavior is studied by com-
paring results from the first and the last sequence Figure 11. Linearity study—example of linear
regression.

350

300
y - 1,2 638x + 0,1283
,. representative drops, normalized against load. An
example of relative gap is shown in the Figure 11.
E'
~ 250
R2 = 0,9997 / Obtained values are lower than the repeatability
~ 200
/ uncertainty (see 3.2).
~ 150
/ This conclusion is confirmed for all test
/ points, so that the pavement elastic behavior is
~ 100
demonstrated.
~
50 This implies that the tests sequence has no influ-
ence on results.
50 100 150 200 250 300
Fmax [kN]

3.5 Reproducibility and temperature correction

Figure 8. Linearity study—example of linear
regression. This part is dedicated to the reproducibility consid-
ered here as the scattering between results provided
by the different devices on a same test point.
1,ooo .--o- ,9- 9_9 _ 1,.--,,o=-=o-=-
o - o-,9-9_9 _ _ _ _ _ _ _ _ _ _ _ _ __ It has been demonstrated in (Broutin, 2010)
0,999 -t--i__--il__--il___ o_.,9•9-9 _ o-'--
,9_99_ _,o""9"'"'
99" --=o-=
,9-::-:
99: : - - - - -
o,'--
99_ 9
that HWD test reproducibility (including device
0,998 0,998
repositioning) is very good in the case of a single
0,998 -t--i__--il__--il________________________________ l--
device on flexible pavement within a short time
range. In the frame of our round robin tests, the
0,997 t---l-----l-----l-------l------ll------ll-----ll------l--1- major parameter to be considered is the effect of
0,996 t---l-----l-----l-------l------ll------ll-----ll------l--1- temperature in the bituminous materials whose
variation during the survey was not negligible.
0,995 +--"--.......- -.......--....--....--....----,~---,-----.-----.-.___,
Figure 12 presents the temperatures in the bitu-
minous materials during the survey, and Figure 13
the deflections obtained for the three STAC’s
HWD tests, performed with respectively 11, 15.5
Figure 9. Linearity study—Determination coefficients and 17°C mean temperatures in the bituminous
vs considered deflection, STAC’s HWD. material layer.

612
 20% .-----------------------
Linear regressions are performed for each device
using results relative to the 3 passes. Figure 13
presents an example of results for the STAC’s
16%
18% --~~~~--------------
+-------~~~r---------

14% +---------------~~----
HWD. 12% +-----------------~~~

A standardization to 15°C of the deflections is ~ 10% +----------------------- ~ Scattering among test

1"<::1 points
then performed, using these regressions, in order ~ 8% +-------------==~~~--
.,.._Scattering among
to make results relative to all devices comparable. 6% +---~~~~----------~
devices

3.6 Scattering among test points vs among

devices 30 60 90 120 150 180 210
Distance to load center r[cm]
Assessing the devices interchange ability consists
in comparing the scattering among devices for a Figure 14. Scattering among test points vs scattering
given test point and the scattering among test among devices.
points for a given device.
In the first case, data standardized to 15°C are
considered (see 3.5). are well calibrated. Note that higher discrepancies
Results are provided in Figure 14. would be expected for in situ pavement.
The scattering among devices is lower than the
scattering among test points, which means that the
devices are interchangeable, under condition they 4 CONCLUSIONS AND PERSPECTIVES

The international round robin tests organized in

Oct. 30, 2013 2013 by the STAC on its test facility have been pre-
sented in this paper. The STAC thanks again all
attendees.
This event was a valuable experience for this
+ p = Ocm

- p = 3cm
kind of surveys and the obtained significant data
.t.

x
p=Scm

p=7cm
base allowed enriching the scientific community
:.. p = 13cm knowledge about F/HWD testing and particularly
- p = 19cm
• p=25cm
test repeatability and reproducibility, and devices
- Average interchange ability. The latter was demonstrated,
under the condition the devices are well calibrated.
13:00 14:00 15:00 16:00 17:00 18:00 It is planned to organize again such kind of
Day t ime [hh :mm)
international survey in 2017, on the rigid test facil-
ity this time.
Figure 12. Temperatures in the bituminous materials
during the survey.

REFERENCES

Broutin, M. 2010. Assessment of flexible airfield pave-

ments using Heavy Weight Deflectometers; Develop-
ment of a FEM dynamical time-domain analysis for
the back calculation of structural properties, PhD
thesis.
300
d3
Broutin, M. et al. 2013. A full-scale instrumented test-
facility for airport pavement modeling improvements,
~ d4 d4; y = 3,3905x + 231,94
Airfield and Highway Pavement Conference proc.
8 250
·~
d5; y = 2,596x + 202 Miami, June 2013.
d5
';
Q d6; y = 1,9608x + 178,21
Broutin, M. & Belon. S.. 2015. International HWD
200 round robin tests, report, http://www.stac.aviation-
d7; y = 1,403x + 158,24 civile.gouv.fr/.
dB; y = 1,2022x + 137,02 STAC, 2012. STAC’s test facility technical manual,
150 report, http://www.stac.aviation-civile.gouv.fr/.
d9
d9; y = 0,5906x + 125,59
STAC, 2014. Airfield pavement structural assessment
100
using Heavy Weight Deflectometer, technical guid-
10 11 12 13 14 15 16 17 ance, http://www.stac.aviation-civile.gouv.fr/.
Mean temperature in the bituminous material [•c]

Figure 13. Deflections vs mean temperature in the bitu-

minous materials.

613
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Reproducibility of decisions for rehabilitation of existing roads based

on deflection measurements with curviameter or FWD

C. Van Geem
Belgian Road Research Centre, Brussels, Belgium

ABSTRACT Managers of ageing road networks are confronted with the pragmatic question where to inter-
vene on the bearing capacity of the lower layers of the road structure in the forthcoming years. The answer
can be sought on network level based on structural indicators, and on project level based on deeper analysis
methods such as back-calculation of elasticity moduli. The answer should be independent of the device used
for deflection measurements. In this contribution the reproducibility of the answer to this pragmatic question
is discussed, illustrated by three case studies. Raw data and structural indicators for network analysis defined
in a previous publication are compared. The usefulness of the determination of homogeneous segments and
of reporting structural indicator values per homogeneous segment or by blocks of constant length for analysis
on project level and for network analysis with a Pavement Management System (PMS) is discussed.

1 INTRODUCTION a Falling Weight Deflectometer (FWD) or the

Curviameter.
In the case of ageing road networks, road man- Curviameter data can be used both at net-
agers are confronted with the pragmatic question work and project level, as illustrated by (Ramos,
where to intervene on the bearing capacity of the Sanchez, et al., 2013) where several indicators as
lower layers of the road structure when program- well as back-calculation are used for the evaluation
ming road works in the forthcoming years. The of bearing capacity, and where the influence of
answer to this question can be sought on network temperature on the measurement data is taken into
level based on structural indicators, and on project account. The FWD allows a more controlled way
level based on deeper analysis methods such as of deflection measurements and is very well suited
back-calculation of elasticity moduli (E-moduli). for project level.
The answer should be independent of the device In this contribution the reproducibility of the
used for deflection measurements. answer to this pragmatic question is discussed and
For a network analysis of the structural per- illustrated by three case studies. In two cases the
formance of the roads in the network, deflection interpretation of Curviameter data and FWD data
measurements can be transformed into values for are compared. In one of these cases FWD data are
structural indicators. In the final report (Litzka available at 5 m intervals. In the third case data were
2008) of COST action 354 a structural indicator only collected with aCurviameter, but collected
based on bearing capacity, possibly combined twice (in 2012 and again in 2015).
with an evaluation of rutting, longitudinal even- Raw data and structural indicators (different
ness and cracking, was suggested. In (Crispino for FWD and Curviameter) for network analy-
et al., 2005) and (Gaborit et al., 2012) for instance, sis defined in (Van Geem, Nigro et al., 2015) are
deflection measurements are used at network compared. The usefulness of the determination of
level. However, for estimating the bearing capacity homogeneous sections, and of reporting structural
by use of back-calculation detailed information is indicator values per homogeneous segment or by
needed on the materials and thicknesses of the blocks of constant length for analysis on project
different layers in the road structure. Estimating level and for network analysis with a Pavement
residual life expectance asks for further informa- Management System (PMS) is discussed. Indica-
tion on past and estimated future traffic. All these tor values and back-calculation results are com-
data are rarely available on an entire road net- pared for project level analysis.
work. Instead, in (Van Geem, Nigro et al., 2015)
a set of structural indicators referred to as KPI1,
KPI2 and KPI3 were defined and applied to the 2 FWD, CURVIAMETER, GPR
primary road network of Wallonia, the southern
region of Belgium. Values for these indicators can The FWD is a device that measures at standstill the
be computed directly from measurements with deflections of the road structure at several points

615
while a falling weight is transferring a load to the for the quality of the bonding in the upper layers
road. The FWD of the Belgian Road Research Cen- of the road structure, using D(0) and D(300) from
tre (BRRC) is equipped with 9 geophones and a force the FWD data and the “road noise indicator” IBR
sensor. The maximal deflection measured with the from the Curviameter data; and KPI3 is conceived
geophone at x millimetres away from the centre of for the evaluation of the cohesion of the entire
impact of the falling weight will be denoted by D(x). road structure (that is, “are all layers playing their
For the three case studies addressed in this paper, role in bearing the load?”), exploiting the variation
measurements were performed as recommended in in the form of the load-displacement diagram for
(Van Gurp et al., 2005). In literature some other indi- three of the geophones of the FWD and the non-
cators are derived from the deflection curve through symmetrical form of the deflection basin measured
points (x,D(x)), such as the radius of curvature R with the Curviameter.
in point (0,D(0)). In (Van Geem, Nigro et al., 2015) The reader is referred to (Van Geem, Nigro
a method to determine Ris proposed different from et al., 2015) for the rigorous definitions of these
the method given in (Jendia 1995). indicators.
The Curviameter is a lorry that allows measur-
ing the deflection bowl at sample points within
a 5 m interval. It uses a geophone for the meas- 4 ROAD SECTIONS
urement of the deflection bowl and operates at
a speed of 18 km/h. The Curviameter is the only Three road sections have been subject to deflection
device that measures the deflection basin includ- measurements. All of them are regional roads car-
ing the maximal deflection Dmax and the radius rying important traffic loads by lorries. All have a
of curvature Rc of the basin at the position on bituminous road surface and can be considered as
the basin where the maximal deflection occurs. A semi-rigid or flexible pavements.
detailed description of this device is given in (Van The first road section is about 14 km long. The
Geem 2010). For the three case studies addressed Curviameter was used over the whole length of the
in this paper, measurements were performed in section. On the same day in 2015, FWD measure-
accordance to the French standard NF P 98-200-7. ments were executed at 126 stations, with an inter-
The “road noise indicator” IBR defined in (Van mediate distance of about 100 m or more between
Geem, Nigro et al., 2015) is computed on the raw two consecutive stations.
signal coming from the geophone and expresses The second road section is only 600 m long but
the “cleanliness” of the signal. Experience shows FWD measurements are available at stations only
that the IBR may indicate bad bonding in the 5 m apart from each other. Also Curviameter data
upper layers of the road structure. and GPR images were collected. Some cores were
The Ground Penetrating Radar (GPR) is a taken as well. All data were collected within a time
tool for non-destructive geophysical investigation frame of less than 2 months in the spring of 2016.
based on the propagation and reflection of electro- The third road section is about 15 km long and
magnetic waves. It reacts to electromagnetic varia- was measured in both directions with the Curvi-
tions in a medium (permittivity, conductivity and ameter in 2012 and in 2015.
magnetic sensitivity). Measurements are generally
carried out by moving the antenna over the surface
(either in contact with the ground or in air) and at 5 HOMOGENEOUS SEGMENTS
regular intervals, which allows to rapidly generate
Road managers prefer planning and applying the
an image of the subsurface.
same maintenance intervention over the whole
length of a road segment with similar structural
3 STRUCTURAL INDICATORS properties and behaviours. Standard procedures
exist for the determination of “homogeneous seg-
Three structural indicators KPI1, KPI2 and KPI3 ments” derived from maximum deflections.
were defined in (Van Geem, Nigro et al., 2015) Homogeneous segments from FWD data are
and applied to the primary road network of Wal- obtained using dynamic segmentation by the cumu-
lonia, Belgium. They take values in the range from lative sum methodas recommended in (Van Gurp
0 (perfect condition) to 5 (poor condition). Each et al., 2005). Homogeneous segments from Curviam-
of these indicators are defined somewhat differ- eter data are obtained using the techniqueby statisti-
ently when they are computed from FWD data cal analysis of the maximal deflectionsas described
or from Curviameter data. Nevertheless, KPI1 is in (du Mesnil-Adelée, Peybernard 1984) and (Leb-
meant to express the bearing capacity of the road eau 1992). Both computation methods are designed
structure and is using the Tragfähigkeitszahl (Tz) for the characterisation of structural performance—
defined in (Jendia 1995) as well as the variation of expressed by D(0) or Dmax – of the road structure.
product Rc.Dmax or R.D(0) within a homogene- Changes in the road structure, such as changes in
ous segment as a first filter; KPI2 is an indicator layer thicknesses or in the materials of which the lay-
616
ers are made, will give rise to a significant change comparisons between D(0) and Dmax have been
in maximal deflection and will mark the separation made and are made again and again. When done
between adjacent homogeneous segments. for the two road sections addressed in this contribu-
Rather local variations in the maximal deflec- tion on which data from both devices are available,
tions may also give rise to the creation of a homo- very similar tendencies in their maximal deflection
geneous segment but this will depend on the step values can be seen. This is illustrated with the graphs
size between consecutive stations where measure- in Figures 1 and 2. In both cases the load applied by
ments were performed. For FWD data, the distance the FWD was 65 kN. In the case of the first road
between two consecutive stations is rarely shorter section the measurements were done on the same
than 20 m and often even 100 m or more. Hence, day, under very similar weather and temperature
local variations can be missed. When a road section conditions (about 22°C). In the case of the second
is probed with GPR first, local discrepancies in the section, the measurements were performed on dif-
road structure can be identified and the FWD sta- ferent days and different road surface temperatures
tions can be chosen in function of the interpreta- (8 to 10°C for Curviameter, 18 to 22°C for FWD).
tion of the GPR radargram. The Curviameter has a When homogeneous segments are determined
fixed interval of 5 m between consecutive measure- from similar raw data, the resulting segments will
ment points where a deflection basin is obtained— also be very similar. Since on a same road section
that is, when the quality of the data is high. Often the number of measurement points from a Cur-
some part of the data is of poorer quality—even if viameter will be much higher than the number of
in that case the IBR may still be usable. Hence, the stations monitored with an FWD, the number of
Curviameter will pick up more local variations than homogeneous segments identified from Curviam-
the FWD but may still miss some local discrepan- eter data tends to be higher than the number of
cies between two consecutive measurement points. homogeneous segments identified from FWD data.
Another way of looking at the raw data may be Less common is a further comparison between
based on the computation of average values in inter- the deflections measured with FWD and Curvi-
vals of fixed length. This is a common approach for ameter. In fact, the same tendencies are still visible
indicators of surface characteristics, e.g. the IRI for but it becomes more obvious that the measurement
roughness. It has the advantage that the intervals results are different due to the different character of
can be set on beforehand, based on geo referenced the measurement devices. Figure 3 shows the com-
positions on the road. This easily allows compar- parison between radius Rc measured by the Cur-
ing the different values obtained at measurement viameter and radius R estimated from the FWD
campaigns in consecutive years for a same indicator deflections with the method described in (Van
on the same road interval. However, unlike both the Geem, Nigro et al., 2015). The latter also gives an
cumulative sum method for FWD and the statistical evaluation of the quality of the estimated value for
analysis method on Dmax of the Curviameter, inter- R. It turns out that the well-estimated values for R
vals of fixed length complicate the identification of correspond well to the measured Rc values.
homogeneous segments and local discrepancies.
Since usually the distance between two consecutive
stations where FWD measurements are available is
about 100 m, it then becomes impossible to report
on segments with constant length less than 100 m.
Hence, the choice of reporting on deflection meas-
urements is to be made in function of the available
data and the collection of the data must be in agree-
ment with the objectives of the road manager.

6 RAW DATA COMPARISONS Figure 1. Dmax and D(0) on the first road section, with
FWD data every 100 m or less.
One way of obtaining the same decision on the
approach for rehabilitation of a road section from
data obtained with different deflection measure-
ment devices would be the repeatability of the
measurements themselves. This however is a vain
expectation for FWD and Curviameter data since
these devices apply a different measurement tech-
nique. Both do evaluate the temporary deformation
of the road structure under a load but the applied
loads are different in nature and the ways in which Figure 2. Dmax and D(0) on the second road section,
the deflections are measured are different too. Still, with FWD data every 5 m.

617
 - - - - - - - Oistance(km)

u
Figure 3. Radii of curvature on the second road sec- lPiil.fWD ..,
Kf!l~

)~ illiltttm ~ i d
tion: measured Rc (Curviameter), estimated R (FWD).

Figure 4. Dmax measured in 2012 (below) and 2015 (on Figure 6. Comparison of Tz, KPI1, KPI2 and KPI3
top) on the same road section (same scales). computed from FWD and Curviameter measurements
and presented on the homogeneous segments determined
from FWD data on the first road section.

from the FWD data and in 40 homogeneous seg-

ments from the Curviameter data. For each of the
homogeneous segments values for Tz, KPI1, KPI2
and KPI3 were computed. Then, for each of the
Figure 5. IBR in 2012 (below) and 2015 (on top) on the homogeneous segments obtained from the FWD
same road section (same scales). data, the homogeneous segments obtained from
In the case of the third section, Curviameter data the Curviametercovering the FWD-segment were
from 2012 and 2015 are at hand. The repeatability grouped together.
of the raw data not only depends on the device and For each group of Curviameter segments, the
the external conditions like surface temperature but averages of the values for Tz, KPI1, KPI2 and KPI3
also on the evolution of the bearing capacity of the in these segments were compared to the values
road structure. With respect to the repeatability and obtained for the corresponding FWD segment. This
reproducibility of Curviameter data, the reader is comparison is presented graphically in Figure 6.
referred to (Van Geem, Ramos Garcia et al., 2015) The definitions of KPI1, KPI2 and KPI3 are
for a recent comparison between two Curviameter- not the same for Curviameter and FWD data. This
sand to (Castillo Diez 2009) for a comparative test explains the difference in values for KPI2, although
executed in Spain. The differences observable in the the tendency—or the relative variation between
raw data of the third road section may rather be consecutive segments—is very similar. For most of
due to further degradation of the bearing capacity the segments, the road manager will come to very
of the road structure. A graphical representation similar first conclusions from these indicators for
of the comparison of Dmax measured in 2012 and network analysis. Note that the variation of the
2015 on the third road section is given in Figure 4. indicator values obtained from Curviameter data
Of course the comparison of Rc and IBR obtained within the same “homogeneous” FWD segment
in 2012 and 2015 could be presented too, both giving can be great. This is illustrated in Figure 7, where
similarly comparable data. However, as illustrated in the FWD values for Tz, KPI1, KPI2 and KPI3 are
Figure 5, the large IBR values of 2012 have become put against the Curviameter values per homogene-
even larger in 2015. This could be an indication of ous segment determined from the maximal deflec-
locally increased delamination in the upper layers. tions measured with the Curviameter. Obviously,
the sparse FWD data do not show such detail.
The FWD data on the second road section are
7 INDICATOR COMPARISONS as densely collected as the Curviameter data (only
5 m between two stations). Since the raw deflection
On the first road section more data from the Cur- data correspond well, so do the frontiers of the 5
viameter (almost every 5 m over 14 km of road) is homogeneous segments independently identified
available than from the FWD (only 126 stations). from FWD and Curviameter data. The values for
Homogeneous segments were independently deter- indicators Tz, KPI1, KPI2 and KPI3 computed
mined from the FWD and from the Curviameter from the FWD data are presented in Table 1 and
data. This resulted in 16 homogeneous segments from the Curviameter data in Table 2.

618
 segments 1, 2 and 4 the values for KPI3 obtained
from the Curviameter data are much higher than the
values for KPI3 obtained from the FWD data. The
values of KPI3 obtained from the FWD data do not
allow making a distinction between the 5 segments
but KPI3 > 2 indicates a weakness in the lower part
of the road structure and KPI3 > 3 adds a weak-
ness in the upper part to it. The by KPI3 indicated
weakness in the deeper part of the road structure is
confirmed for segments 1, 2 and 4 by Tz, KPI1 and
high maximal deflections Dmax, all indicating poor
bearing capacity. So in this case, despite somewhat
different values for the individual indicators, reason-
ing on the combination of KPI1, KPI2 and KPI3
obtained with the FWD results in quite similar con-
clusions on the structural health of the road section
as reasoning on the combination of KPI1, KPI2 and
KPI3 obtained with the Curviameter.
Figure 7. Comparison of Tz, KPI1, KPI2 and KPI3 On the third road section the evolution of the
computed from FWD and Curviameter measurements indicators computed from Curviameter data can
and presented on the homogeneous segments determined
be studied over a time laps of 3 years. The vertical
from Curviameter data on the first road section.
lines on the graphs in Figures 4 and 5 are the bor-
ders of the homogeneous segments determined from
Table 1. Indicators from FWD data on 5 homogeneous
segments of the second road section. Dmax measured in 2012. When homogeneous seg-
ments would be determined from Dmax measured
Segment Tz KPI1 KPI2 KPI3 in 2015, their borders would not coincide with those
of 2012. In general, additional homogeneous seg-
HS 1 1.19 4.32 2.14 2.30 ments could appear, or a sequence of homogeneous
HS 2 0.55 3.79 3.37 3.09 segments could be replaced by a different sequence.
HS 3 1.56 4.11 3.33 3.03 This makes it difficult to compare indicators com-
HS 4 0.84 3.31 3.99 2.91 puted on each individual homogeneous segment. In
HS 5 1.39 4.20 2.07 3.14 order to compare the indicators obtained in 2012
and 2015, the homogeneous segments of 2012 were
Table 2. Indicators from Curviameter data on 5 homo-
applied to the raw data of 2015. The four graphs in
geneous segments of the second road section. Figure 8 give the values obtained for indicators Tz,
KPI1, KPI2 and KPI3 in 2012 and 2015 for each of
Segment Tz KPI1 KPI2 KPI3 the homogeneous segments. A very good correla-
tion is obtained between the values of 2012 and 2015
HS 1 1.7 3.1 3.3 4.8 for Tz and KPI1. Some bigger differences occur for
HS 2 0.9 3.3 3.6 4.4 KPI2 and KPI3. This is due to the definition of these
HS 3 2.6 3.3 3.0 2.0 indicators, where first a filter based on a threshold is
HS 4 0.9 3.3 3.7 4.6 used. This filter then determines which formula to
HS 5 2.1 3.8 3.0 2.3 apply in order to determine the value of the indica-
tor. For some homogeneous sections the threshold
Note that the definitions of the KPI’s are different when
of the filter was not reached in 2012 but was reached
based on raw FWD or Curviameter data.
in 2015 (or vice versa). Especially for KPI2 this
Data were collected under different road surface results in a very significant difference for its value (as
temperature conditions (FWD at 18 to 22°C, Cur- for segments 11, 12, 27 and 43). The filter for KPI2 is
viameter at 8 to 10°C). As reported in (Van Geem, based on the ratio RR between the second decile of
2017), with a higher surface temperature Tz tends the Rc values and the average value of the Rc in the
to decrease and KPI1 tends to increase. Taking this homogeneous segment. When ratio RR is smaller
effect into account, the values for Tz are compara- than 0.75, the variation of Rc is considered too
ble and the values of KPI1 are quite consistent. It important, indicating a great risk of lack of bonding
can be seen that the tendency for KPI2 seems to be in the top layers. In that case, KPI2 is set to a value
respected but that the values for KPI2 from the FWD greater than 3. Otherwise, KPI2 takes a value smaller
data make a more significant distinction between than 3 computed with the value for IBR. It turns out
the homogeneous segments than the values for KPI2 that for some segments the “noise” is not very impor-
from the Curviameter data. Both data sets seem to tant but the variation of Rc is. This may rather be an
indicate a lack of bonding in segments 2 and 4. For indication that the homogeneous segment is not as

619
 Table 3. Layer structures in three segments (thicknesses
expressed in m).

Thickness Thickness Thickness

Layer segment 1 segment 2 segment 3

Asphalt 0.206 0.200 0.230

Unbound aggregates 0.250 0.250 0.250
Gravel 0.300 0.300 0.300
Ground 4.500 4.500 4.500
Stratum ∞ ∞ ∞

between the asphalt layer and the unbound aggre-

gate layer is not good at all in segments 1 and 2, and
not optimal in segment 3. The ratio between FWD
deflections D(300) and D(0) confirm this. The val-
ues for KPI2 are rather low, which should indicate
Figure 8. Tz, KPI1, KPI2 and KPI3 in 2012 and 2015 poor bonding. For the back-calculations both good
for all homogeneous segments on the third road section. and bad bonding between the layers was modelled
in order to investigate the effect on the E-moduli.
homogeneous after all. Note that the segmentation Versionv1.5 of the software “Alizé-LCPC” was
from Dmax measured in 2012 is applied to the data used for back-calculation of Curviameter and
in 2015: this allows investigating the repeatability of FWD data. Both loads by the double-wheel of the
the conclusions from the indicators but this may not Curviameter and by the circular plate of the FWD
be the optimal segmentation for the data of 2015. is modelled in Alizé-LCPC.
Small changes in the segmentation may result in a For the Curviameter data, 15 points on the deflec-
bigger RR, so that a more similar value for KPI2 is tion basin were introduced as input to the compu-
obtained. In general, when (after the application of tations. These points were selected from a basin as
a preliminary filter) the same formula can be applied follows: the first point is the place on the deflection
on the data from 2012 and 2015, very similar values basin where the maximal deflection Dmax occurs
are obtained for all indicators. and this place is denoted as the point at reference dis-
tance “zero”; then the other points on the basin are
selected at a distance of 0.04, 0.08, 0.12, 0.16, 0.20,
8 PROJECT LEVEL COMPARISONS 0.24, 0.28, 0.40, 0.50, 0.60, 0.75, 0.80 and 0.90 m
away from the point at reference distance “zero”.
Sufficiently detailed information about materials The deflection basin measured by the Curviameter
and layer thicknesses as needed for back-calcula- is stored as a sequence of 100 points, with interme-
tion of E-moduli for the layers in the road struc- diate distance of 0.04 m between two consecutive
ture was only available for the second road section. points on the basin. Most of the selected points are
Over the whole length of the section GPR data registeredpoints. The few selected points that do not
were collected, 15 cores were taken, and the road man- coincide with a position on the basin are the result
ager provided additional information about the base of linear interpolation from the two neighbouring
course. From the Curviameter data 5 homogeneous points that are registered. Back-calculations using
segments were identified. The comparison between points further away from Dmax did not result in any
the results of back-calculation with Curviameter and better results. Hence, a somewhat larger number of
FWD data are presented on only 3 of these segments points close to D max were selected.
because they represent all significant differences that For the FWD data, the deflections measured by
could be observed on the road section. The materi- the 9 geophones of the FWD were introduced as
als in the road structure are the same everywhere but well as the load pressure obtained from the load cell
the layer thickness and the age of the asphalt layer measurement. The geophones were positioned at 0,
vary. Table 3 presents the road structures in the three 300, 600, 900, 1200, 1500, 1800, 2100 and 2400 mm
homogeneous segments. The layer thicknesses of the away from the centre of impact of the load force.
asphalt layers were determined from the cores: 1 core The E-moduli were computed for all stations
in segment 1, 3 cores in segment 2 and 2 cores in seg- measured with FWD or Curviameter in each seg-
ment 3.A Poisson coefficient of 0.35 was used for the ment and then the stations that allowed best conver-
asphalt layers, of 0.50 for all other layers of the road gence between measured and computed deflections
structure, and of 0.25 for the stratum. were selected. The results that gave unrealistic values
From the radius of curvature Rc measured by (from an engineering point of view) for some of the
the Curviameter, one would say that the bonding E-moduli were ignored. For segment 1 the selected

620
stations from FWD and Curviameter data are pre- (between stations from km 14.871 to 14.896). In the
sented in Table 4 when perfect bonding was assumed rest of the stations one may hesitate on the quality
and in Table 5 when total lack of bonding between of bonding although the E-moduli seem somewhat
the asphalt and the unbound aggregates layers was less realistic when perfect bonding is modelled. The
considered. There is little difference between the back-calculation from Curviameter data conver-
results based on FWD or Curviameter data. Differ- gences better when modelling perfect bonding but
ences in back-calculation results for the same device then the resulting E-moduli for the unbound aggre-
between different stations are explained by differ- gate layer are often unrealistic. So, from the Curvi-
ences in measured deflection values. Large varia- ameter data one would rather conclude that there is
tions in stratum moduli obtained from Curviameter a lack of bonding. Convergence with FWD data is
data could be explained by the lack of data beyond better than with Curviameter data. A few represent-
0.90 m away from the point at reference distance ative results for segment 3 are presented in Table 7.
“zero”. Some results in Table 4 are unrealistic so one Still, the conclusions from back-calculations are
would rather conclude that bonding is fine but that not different: the asphalt layer is in good condition
the second and third layers are weak. This stands in and the unbound aggregate layer is not as weak as
contrast to KPI2 and the small values for Rc (rather in the other two segments.
indicating bad bonding) but in accordance to high Previously it was reported in (Van Geem, Perez,
values for KPI3 (lower layers failing). 2015) that back-calculations from Curviameter
For segment 2 different bonding models do not data were more difficult to perform than from
make back-calculation converge better. The conver- FWD data. But in those cases only 4 points were
gence of the back-calculation is better for the FWD selected on the deflection basins measured by the
data than for the Curviameter data. Also, in this Curviameter: only the points at 0, 0.30, 0.60 and
case back-calculation from Curviameter data gave 0.90 ma way from the point at reference distance
quite often unrealistic values for the E-moduli of the “zero” were used. Using a higher number of points
second and the third layer. Still, both from FWD improves the performance of the back-calcula-
and Curviameter data, and in all stations where real- tions but it still is more difficult to obtain realistic
istic E-moduli are found, the E-moduli for both lay- E-moduli from Curviameter data than from FWD
ers turn out to be quite low. A further discussion on data. Since Alizé-LCPC is using a linear elastic
the quality of the bonding between these two layers model for the theoretical multi-layer road struc-
is not useful since the road manager would have to ture, it can be expected that taking into account
decide to replace both layers. Table 6 gives a few rep- the viscosity of the materials may further improve
resentative back-calculation results for segment 2. back-calculationfrom Curviameter data.
In segment 3 the results of back-calculation on
FWD data seem to indicate that there is a lack of
bonding, especially in the first part of the segment Table 6. Representative E-moduli for segment 2.

FWD Curviameter
Table 4. E-moduli for segment 1, perfect bonding. E-modulus
(layer) s.14490 s.14545 s.14456 s.14531
FWD Curviameter
E1 (asphalt) 6614 7343 6270 3317
E-modulus (layer) s.14340 s.14331 s.14341 Bonding no perfect no perfect
E2 (unb. aggr.) 555 132 491 213
E1 (asphalt) 9901 8285 4604
E3 (gravel) 109 167 22 34
E2 (unbound aggregates) 248 859 734
E4 (ground) 179 210 114 105
E3 (gravel) 235 116 127
E5 (stratum) 202 208 162 89
E4 (ground) 242 89 91
E5 (stratum) 271 1611 41209

Table 7. Representative E-moduli for segment 3.

Table 5. E-moduli for segment 1, no bonding.
FWD Curviameter
FWD Curviameter E-modulus
E-modulus (layer) s.14896 s.14920 s.14822 s.14927
(layer) s.14327 s.14347 s.14331 s.14341
E1 (asphalt) 8464 13415 9286 6150
E1 (asphalt) 8662 4147 14863 8605 Bonding no no no no
E2 (unb.aggr.) 1965 37 991 593 E2 (unb. aggr.) 1867 3101 772 1126
E3 (gravel) 492 1774 34 173 E3 (gravel) 200 60 139 212
E4 (ground) 223 175 115 103 E4 (ground) 238 257 157 183
E5 (stratum) 285 370 570 70055 E5 (stratum) 165 151 201 207

621
9 CONCLUSIONS Crispino, M., Olivari, G., Poggiolo, M., Scazziga, I.,
Including Bearing Capacity into a Pavement Manage-
In this paper the pragmatic question was addressed ment System, 7th International Conference on Bearing
where on an ageing road network to intervene on Capacity of Road Pavements, Trondheim, Norway, 2005.
the bearing capacity of the lower layers of the road duMesnil-Adelée, M., Peybernard, J., Traitementautoma-
structure, based on three case studies and on FWD tique des résultats de mesureencontinu, in Bulletin de
liaison des Laboratoires des Ponts et Chaussées N° 130,
and Curviameter data. The answer to this question pp. 76–81, March-April 1984.
should be independent of the deflection measure- Gaborit, Ph., Di Benedetto, H., Sauzéat, C., Pouget,
ment device used. On network level, the road man- S., Olard, F., Quivet, S., Analyse d’une structure de
ager can consider indicators that do not need deep chausséeautoroutière par auscultation in situ etessaisen-
knowledge on the road materials and the past and laboratoire, Actes des 30ième Rencontres Universitaires
future traffic load. Some indicators computed from de Génie Civil (AUGC et IBPSA 2012), Chambéry,
the raw data of FWD and Curviameter were com- France, 6–8 June 2012.
pared and it is shown that they give rather consist- Jendia.S., Bewertung der Tragfähigkeit von bituminösen-
ent results. On project level, the road manager may Straßenbefestigungen. Veröffentlichungen des Institutes
fürStraßen- und Eisenbahnwesen der Universität Karl-
dispose of detailed information allowing back- sruhe (TH), Heft 45, 1995 (ISSN 0344-970-X).
calculation of elasticity moduli and the estimation Lebeau, J., Traitementautomatique de mesures pour undé-
of residual service life. It is shown on one of the cases coupage de routes en zones homogènes, rapport de stage
that the conclusions from FWD and Curviameter de DESS,étudiant de l’Université Paris VI sous la direc-
data on the necessity for rehabilitation of the bearing tion de Y.Lerolle du CEBTP, Octobre 1992.
capacity of an existing road are very similar. The pos- Litzka, J., et al., The way forward for pavement perform-
sibility to perform back-calculation with 15 points ance indicators across Europe, Final report of COST
of the deflection basin measured with the Curviam- action 354, Performance Indicators for Road Pave-
eter (rather than with the traditional way using only ments, ISBN 978-3-200-01238-7, 2008.
Ramos-Garcia, J.A., Sanchez-Dominguez, F., Alvarez-
4 points) has a positive influence on the success of Loranca, R., Structural Road Surveys Using a Curvi-
the computations. The difficulty of determining the ameter Device. Analysis of the Temperature Effect on
same homogeneous segments from FWD or Curvi- Asphalt Mixes Properties, 9th International Conference
ameter data (due to the smaller number of stations on Bearing Capacity of Road Pavements, Trondheim,
where FWD data are collected) or from Curviameter Norway, 2013.
data collected on the same road section but in differ- Van Geem, C., Influences of measurement conditions on
ent years was evoked. It is recommended to deter- structural indicators obtained from FWD data, 10th
mine homogeneous segments from the data rather International Conference on the Bearing Capacity of
than to use segments with fixed length because this is Roads, Railways and Airfields, Athens, June 2017.
Van Geem, C., Nigro, P, and Berlémont, B., The Use of
the most relevant way for the attribution of indicator Deflection Measurements in Pavement Management of
values to a road segment. Moreover, the changes in the Primary Road Network of Wallonia, Belgium, 9th
the decomposition of the road section in homoge- International Conference on Managing Pavement Assets
neous segments obtained from data collected in two (ICMPA9), Alexandria, Virginia, USA, May 18–21,
different years can help evaluating the evolution of 2015.
the structural performance of the road section. Van Geem, C., Overview of interpretation techniques
based on measurement of deflections and curvature
radius obtained with the Curviameter, 6th European
ACKNOWLEDGEMENTS FWD User Group Meeting, Sterrebeek, Belgium, 10–11
June 2010.
The case study with the FWD measurements at Van Geem, C., Perez, S., Structural Evaluation of a Dual
5 m intervals is part of the pre-normative project Two-Lane Road by Curviameter and FWD Data,
on the use of ground penetrating radar in the Back-Calculation and Dimensioning for Rehabilitation,
road sector, supported by the “Belgian Bureau for Proceedings of The International Symposium on Non-
Destructive Testing in Civil Engineering (NDT-CE),
Normalisation” (NBN) and the Belgian “Federal Berlin, (in “Open Access e-journal of Non-destructive
Public Service Economy, S.M.E.s, Self-employed Testing—ISSN 1435–4934”: http://www.ndt.net/), Sep-
and Energy”, and realized in collaboration with tember 15–17, 2015.
“l’Institut Scientifique de Service Public” (ISSeP). Van Geem, C., Ramos Garcia, J.A., Sanchez Dominguez,
The author would also like to thank the techni- F., and Van Buylaere, A., Comparative test between
cians of division SRM of BRRC for the realisation Curviameters of BRRC and Euroconsult and Quality
of the measurements and for their technical support. checking of the Curviameter at BRRC, Presentation at
the Curviameter Workshop in Wavre, Belgium, (see web
page http://www.ocw.be/en/curviametre, last accessed:
REFERENCES 24 October 2016), January 2015.
Van Gurp, C.A.P.M. et al., COST 336, Use of Falling
Castillo Diéz, E., Nota técnicasobre la armonización de la Weight Deflectometers in Pavement Evaluation, Final
medida de la deflexion con equiposdeltipocurviámetro, Report of the Action, Main Report, 2nd Edition, April
Ministerio de fomento, Madrid, Spain, 23 March 2009. 2005.

622
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Investigation of sampling strategy on estimating strains in flexible

pavements through Response Surface Method (RSM)

R. Aswathy & A.K. Swamy

Department of Civil Engineering, Indian Institute of Technology, New Delhi, India

ABSTRACT: Various numerical techniques used for analysis of flexible pavement use constitutive rela-
tionships to compute stress-strain fields in pavement under given traffic loading, material properties and
environmental conditions. Even with recent improvements in computational algorithms, more time and
resources are required to achieve a viable numerical solution. Under such circumstances, Response Sur-
face Method (RSM) is a convenient method to use. Published literature in other areas of engineering
has indicated that accuracy of RSM model is highly influenced by sampling method. This work presents
details about various sampling strategies used while constructing response surfaces for predicting strain
in flexible pavement. A three layered pavement structure was considered for this study. Numerical values
of all significant input variables obtained from different sampling methods were used to predict strain
at critical locations. Further, response surfaces of different orders were developed and evaluated using
statistical techniques for their accuracy.

Keywords: Response Surface Methodology, Sampling, Surrogate model, Design of Experiments

1 INTRODUCTION surface. The mathematical functions are developed

using training data set. This training data set com-
Various numerical methods have been used in the prises of input values at selected data points and
area of pavement engineering to compute stress- corresponding output values. DoE approach uses
strain field in pavement layers. Such an exercise optimized sampling strategies while choosing the
of computing stress-strain field involves exten- number and location of input variables. Such an
sive utilization of constitutive models and matrix optimized sampling strategy guarantees extraction
manipulation. Even with recent improvements in of maximum information with least effort. Some
numerical algorithms and computational power, of the popular sampling methods used in RSM are
more time and resources are required to achieve Box Behnken (BBM) method, Full Factorial Design
at a viable solution. This issue of increased time (FFD) and Central Composite Design (CCD). The
requirement and resources might be a bottleneck current work compares various sampling strategies
when several possible design alternatives have to be used in DoE while generating response surface for
evaluated quickly on routine basis. Under such cir- predicting strain at critical locations in three lay-
cumstances, surrogate modelling techniques come ered pavement structure.
handy. Surrogate models are compact, predict
response with reasonable accuracy, cheap to evalu-
2 BACKGROUND
ate, and have proven very useful for various areas
mentioned previously. These surrogate models imi-
2.1 Response surface methodology
tates the behaviour of the computationally expen-
sive simulation model, while being economical to RSM includes a set of mathematical and statis-
evaluate. In general, surrogate models relate vari- tical techniques, in which the interrelationships
ous influencing factors (i.e. independent variables) between independent parameters and response
and the parameters of interest through mathemati- are found (Myers & Montgomery, 1995; Draper
cal functions (Dirk Gorissen, 2010). Some exam- & Box 1951; Baş & Boyacı, 2007). RSM helps in
ples for surrogate modelling are Response Surface finding the most suitable approximation using
Methodology (RSM), space mapping technique, polynomials. A typical first order and second order
polynomial chaos expansion etc. RSM uses Design response surfaces used are presented in Equation 1
of Experiments (DoE) approach to select independ- and 2, respectively. In general, scatter is found
ent variables and then to finally develop a response with experimental data and numerical simulation.

623
In experimental data, this scatter can be attributed
to measurement errors. On the other hand in numer-
ical simulation, scatter is due to numerical approxi-
mations in the model adopted (Toropov & Polynkin,
1993). In this context, RSM accounts for possible
errors through fitting averaged function through
sampling space. Its application is mainly aimed at
reducing the computational cost of the expensive
analysis and the numerical noise associated with it. a) FFD b) BBM
k
y β0 ∑β X
j =1
j j (1)
k k
y β0∑ X + ∑β j =1
j j
j =1
jjj X 2j +

∑∑ β X X iij i j (2)
i j

where, y = response; β0, βj, βjj, βij represents the c) ceo

regression coefficients for intercept, linear, quad-
ratic and interaction terms respectively; Xj and Figure 1. Schematic representation of DoE techniques.
xj = independent variables.
Development of typical response surface con-
sists of mainly six steps i.e. (i) choice of independ- of these levels across all such factors. If cube points
ent variables, (ii) identification of typical range of (lower limit, upper limit) and centre point of all
values, (iii) selection of input parameters at critical design variables are considered, the total number of
locations, (iv) development of database of response combinations in CCD is given in Equation 3 (Lopo
at these critical locations, (v) determination of de Araújo et al., 2012). The major advantage of using
regression coefficients, and (vi) validation of pre- FFD is that the information regarding the factors
dicted response. (Myers & Montgomery, 1995; Baş is obtained to the maximum possible extent as the
& Boyacı, 2007; Mohammed et.al, 2012; Penny & number of degrees of freedom is high compared to
Lindfield, 1999; Steenackers & Guillaume, 2009). other techniques. (Antony, 2003). Since the number
More details regarding DoE is presented in the next of experimental runs increase significantly, FFD is
section. preferred when less than five independent variables
have to be evaluated (Tijana Rakić et al., 2014).
2.2 Sampling techniques in DoE N = 3K (3)
RSM mainly involves DoE through which combina-
tions of the input parameters at critical locations are where K = number of independent variables
identified (Baş & Boyacı, 2007; Penny & Lindfield, In case of CCD, sampling of independent
1999; Steenackers & Guillaume, 2009). DoE enables parameters is obtained at cube points, centre points
a planned testing when the objectives or the response and axial points. The total number of combinations
function are outlined clearly. Hence RSM along with in CCD is given in Equation 4 (Lopo de Araújo
DoE offers several advantages over other prediction et al., 2012). Inclusion of axial points in CCD
methodologies. RSM has been used in various fields allows for predicting curvature in response surface.
including research, quality control, product devel- However inclusion of axial points beyond identified
opment, manufacturing, market research, purchas- range of values limit the construction of accurate
ing, engineering etc. (Antony, 2003). response surface model. In comparison with FFD,
The efficiency of DoE depends on the selection number of critical points are less in CCD since
of the sampling technique adopted in identifying some of the degrees of freedom are not considered
critical locations. Some of the popular sampling which would be helpful for better reliability models
techniques are BBM, FFD and CCD. The sche- particularly when experimental errors are expected
matic diagram of each sampling technique is pre- in the test (Tijana Rakić et al., 2014).
sented in Figure 1.
In case of FFD, individual parameters are speci- 2K 2 + Cp (4)
fied, each at discrete possible locations and whose
experimental units take all possible combinations where CP = number of center points

624
 In case of BBM, design points in the centre are
considered while eliminating axial points and centre Material properties I I Geometric properties
points. This method will be particularly useful if the
boundaries are known beforehand. The number of
+ l
combinations given by BB method is given in Equa- Identify limits
tion 5 (Lopo de Araújo et al., 2012). BBM can give
best combinations even when large number of vari-
ables is considered. However, BBM generates less Develop combinations of input parameters
number of combinations resulting in lesser degrees usingRSM
of freedom (Tijana Rakić et al., 2014). Full Factorial Design
Box Behnken Method
N K (K
2K ) CP (5) Central Composite Design

Linear Elastic Analysis using llT Pave

2.3 Use of RSM in pavement engineering
Recent studies in pavement engineering have made
use of RSM in their research incuding evaluation Strain at critical locations
of stripping potential (Khodai et al. 2012, 2013),
modeling of asphalt concrete properties (Khodai
et al., 2014; Kavussi et al., 2014), reliability based
design of flexible pavements (Deepthi & Shivku- I Develop Response surface model
mar 2013), evaluation of rutting (Manik et al.,
2007), and finite element analysis of pavements
(Bassam Saad 2014). Generate strain at critical locations from the
I model

3 CURRENT METHODOLOGY I Statistical analysis and validation

According to Indian guidelines of flexible pave-

ment design (IRC 37-2012), the pavement is Figure 2. Methodology adopted for the study.
modelled as a linear elastic multilayer structure.
3.1 Identification of significant parameters
Further, tensile strain at bottom of asphalt con-
and their limits
crete and vertical compressive strain on subgrade
induced by standard vehicle are computed. These In this work, first step consisted of selecting critical
strain values are computed using multilayer elas- parameters that affect stress-strain fields in flexible
tic analysis approach. The thickness of individual pavement. Based on published literature, variables
layers is adjusted until allowable number of rep- that affect stress-strain significantly i.e. modulus
etitions matches with expected number of repeti- of Hot Mix Asphalt layer (HMA) (E1), modulus
tions. However, use of multilayer elastic analysis of subbase (E2), modulus of subgrade (E3), thick-
software for routine work increases computa- ness of HMA layer (h1) and thickness of subbase
tional time and resources. This necessitated the (h2) were selected. Other parameters like stand-
development of surrogate model that can reason- ard individual wheel load (P = 20 kN), tyre pres-
ably predict strain at critical locations with little sure (0.56 kPa), wheel spacing (s = 155 mm) were
effort. assumed as per relevant Indian guidelines. In India,
The current research work involved four major VG 30 grade binder is widely used in pavement
steps. Initially, independent variables that affect construction. The recommended resilient modulus
pavement response significantly and their practical values for HMA mixture with VG 30 binder in IRC
range (upper and lower limits) were identified. In guidelines were used. The resilient modulus values
the second stage, using different sampling strate- of subgrade corresponding to 20°C, 25°C, 30°C,
gies training datasets were generated. In the third 35°C, 40°C were 3500 MPa, 3000 MPa, 2500 MPa,
stage, response surfaces were developed using 1700 MPa, 1250 MPa respectively.
these training datasets. Finally, statistical analysis
was conducted to measure quality of predictions
3.2 Data set generation
made using these response surfaces. Schematic dia-
gram of overall methodology used is presented in DoE approach was used to obtain the different
Figure 2. combinations of the variables selected for the study.

625
 p p 4 RESULTS AND COMPARISON
s
In order to estimate the strain in pavement lay-

LJ LJ HMA/ayer
ers at critical locations, five independent param-
eters with their range were specified using DoE
techniques (FFD, CCD and BBM). The output
obtained from IIT Pave was used as input for devel-
opment of response surface. The response surfaces
1 2 were developed for strain at four critical locations
( t,, , ε t ,d , v,, ,d ). The fitted responses were
Subbase layer statically analysed using ANOVA and the R2 val-
4 ues, MAPE and SSR. With each RSM technique,
3
four critical conditions were analysed. This resulted
Subgrade in 12 response surfaces. The strain values obtained
from RSM approach was cross plotted against val-
Figure 3. Three layered pavement structure. ues obtained from IITPave. The worst fit and best
fit plots are presented in Figures 4 and 5, respec-
tively. Figure 4a represents the crossplot by CCD
Based on practical range of values within which and gave a worst fit. Inorder to justify whether this
independent variables are expected to lie was used was due to the presence of outliers, Figure 4b was
to generate training dataset. Based on sampling plotted by eliminating the outliers. However it was
approach, various combinations of design varia- observed that the predictive accuracy was very less.
bles were generated through DoE. FFD, CCD, and
BBM sampling strategies were used in this study. 0.35
With five independent variables, the number of
combinations generated by FFD, BBM, and CCD 0.3
were 243, 41, 53 respectively. "'>
c.."' 0.25
Strain at four critical points was computed t:
using IITPAVE, linear-elastic layered media analy- E 0.2
sis software widely used in India. The strain at four 0
J:
critical locations was considered. These locations "0 0.15
.B"'
are indicated in Figure 1. Tensile strain in HMA c
0.1
under one of the tyres (1), tensile strain at middle .<:>
0
R2 =0.5517
of the dual tyres in HMA (2), vertical strain under '@
c 0.05
one of the tyres in subgrade (3), and vertical strain b
U)
at middle of the dual tyres in subgrade (4) are rep- 0
resented by ε t ,s , ε t ,d , εv ,s d εv ,d , respectively. The -0.05 0 0.05 0.1 0.15 0.2 0.25
three layered pavement structure considered for Predicted strain from CCD
the study is shown in Figure 3. Thus a training
dataset consisting of strain at critical locations and 4(a)
corresponding input values was developed.

"> 0.0006
3.3 Generation of response surface c.."'
0.0005
The training dataset generated in previous step was t=
E 0.0004
used as input for developing response surface. For 0
<./:::
R2 =0.0075
developing the response surface, XL Stat software
was used. First and second order polynomial response
"0

·s""
0.0003
,- T , .
. .•• I•
0.0002
surfaces were developed in this work. The relevant .£)
0
expressions are presented in Equations 1 and 2. c
·;;; 0.0001
):;
C/J 0
-0.05 0 0.05 0.1 0.15
3.4 Statistical analysis and validation
Predicted strain from CCD
The response surfaces obtained were evaluated for
quality of fit through statistical analysis. Various 4(b)
parameters like Coefficient of Correlation (R2),
MAPE and SSR were used for evaluating the Figure 4. Crossplot of tangential strain at middle of
goodness of fit. dual tyres in HMA by CCD.

626
 1.2 ,------------~

"'>
~ 4.0E-04
t:
E
~ 3.0E-04
"0

"'c
I9
.g Z.OE-04
c
·~

tl l.OE-04
l.OE-04 Z.OE-04 3.0E-04 4.0E-04
E(t,s) E(t,d) E(v,s) E(v,d)
Predicted strain from BBM
"'FFD ~BBM EC CCD

Figure 5. Cross plot of tangential strain at middle of

one of tyres in HMA by BBM. Figure 6. Co-efficient of determination for various
sampling strategies using first order response surface.

The values of R2 obtained while fitting first order

and second order response surface is presented in 1.1
c
Figures 6 and 7, respectively. Comparison of R2 0
-~ 1
values indicated that for horizontal strain in HMA c
(both under single wheel and dual tyres assem- '§ 0.9
bly) BBM gave the best results while CCD gave ~
the worst. This was because of the in acceptable "'
:: 0.8
combinations of the input parameters generated 0

by CCD which are far beyond the specified limits. ~ 0.7

·u
Similarly, for strain in subgrade (both under single <C
'a; 0.6
wheel and dual tyres assembly), BBM gave the best 0
u
R2 values compared to the other two methods. The
best and least R2 values were obtained from BBM E(t,s) E(t,d) E(v,s) E(v,d)
and CCD respectively for all the four critical condi-
tions. This was true for both first order and second "'FFD ISBBM E CCD
order response surfaces. In general, second order
response surface fitted well when compared to first
Figure 7. Co-efficient of determination obtained for
order response surface. various sampling strategies using second order response
For further estimating the prediction accuracy surface.
of these sampling methods, MAPE was calculated
for the second order response surfaces generated. It
100 ,------------~
was observed that BBM showed least mean abso-
lute error compared to other sampling methods.
While comparing the MAPE (Figure 8) for strain
at middle of one of the tyres in HMA, least error

--1~1~----l---1
was observed with BBM and maximum error was
obtained with CCD compared to other methods.
This was the same for other critical conditions too.

I ..i~~,~~:.:1':.•:.~•!_' -__=~ -_=1 ~~

For comparing the deviations predicted from
actual empirical values of data, SSR was calcu- 1
]
~
lated. The same is presented Figure 9. It is clear :.·:_ :_.
:!:_:
:._: ::'.·:!:_:_::. !::
'.·:

from Figure 9 that CCD gave higher values since 1

•_::_::_: :_ :_:1, 1::
1 ..L....O-~~..,.........==~;r:~o.:-...·~
i:>;_
: .....--"~'-~~L
' --"~l-..l
the deviations from the actual empirical values
were beyond the limits. Further, SSR was least in E(t,s) E(t,d) E(v,s) E(v,d)
the case of BBM and hence it indicates a tight fit
of the model to the data. For comparing the SSR " FFD ;.:- ssM ::ceo
for the individual strain combinations such as for
middle of one of the tyres, SSR was least for BBM Figure 8. MAPE for various sampling strategies using
and highest for CCD. For strain at middle of dual second order response surface.

627
 [3] Baş, D. & Boyacı, İ. H. 2007. Modeling and optimi-
l.OE-10
zation I: Usability of response surface methodology.
Journal of Food Engineering 78(3):836–845.
l.OE-08 [4] Saad, B. 2014. Factorial Numerical Analysis of
Flexible Pavement Foundations with Emphasis on
i
~I=
l.OE-06 Groundwater Table Effect. Geo-Congress 2014.
[5] Bezerra, M. & Santelli, R. E. 2008. Response Sur-
"
C/')
C/')
~ face Methodology (RSM) as a tool for optimization
l.OE-04

l.OE-02 -; l"!l
in analytical chemistry. Talanta 76(5):965–77.
[6] Box, P. & Draper, N. (2nd ed.) 2007. Response Sur-
faces, Mixtures and Ridge Analyses. New Jersey:
John Wiley and Sons.
[7] Breitkopf, P., Naceur, H., Rassineux, A. & Villon, P.
l.OE+OO 2005. Moving least squares response surface approx-
£(t,s) £(t,d) £(v,s) £(v,d) imation: Formulation and metal forming applica-
tions. Computers and Structures 83:1411–1428.
"FFD I:II BBM ili! CCD
[8] Chavez-Valencia, L., Manzano-Ramírez, A., Luna-
Barcenas, G. & Alonso-Guzmán, E. 2004. Model-
Figure 9. SSR for various sampling strategies using sec- ling of the performance of asphalt pavement using
ond order response surface. response surface methodology. Building and Envi-
ronment 40(8):1140–1149.
[9] Gorissen, D. 2010. A Surrogate Modeling and
tyres in HMA, SSR was highest for CCD while Adaptive Sampling Toolbox for Computer Based
BBM was the least. Design. Journal of Machine Learning Research
The results obtained from the three sampling 11:2051–2055.
techniques indicate that the response surface [10] Gourab, S & Biligiri, P. 2016. Cracking performance
model developed by BBM have the highest R2 analysis of asphalt mixtures using response surface
values and the least MAPE and SSR. Higher R2 methodology: experimental investigations and sta-
values of BBM indicate that the variability of the tistical optimization. Materials and Structures (in
response data around its mean is well predicted by press).
[11] Groenwold, A. & Etman, L. 2007. Incomplete series
this model. Hence MAPE (prediction accuracy) expansion for function approximation. Structural
and SSR (deviations predicted from the data) will Multidisciplinary Optimisation 34: 21–40.
be very minimum. [12] Hanrahan, G., Garza, C., Garcia, E. & Miller, K.
2007. Experimental design and response surface
modeling: A method development application for
5 CONCLUSION the determination of reduced inorganic species in
environmental samples. Journal of Environmental
This article presented a comparison of various Informatics 9(2):71–79.
sampling strategies that are used while constructing [13] IRC 37–2012, Design of flexible pavements.
[14] Kavussi, A., Qorbani, M., Khodai, A. & Haghshe-
response surfaces. For this purpose strain in flex- nas, H.F. 2013. Quantification of Parameters Affect-
ible pavement at critical locations was selected as ing Moisture Resistance of Warm Mix Asphalt
an example. These response surfaces and sampling using Response Surface Methodology. International
strategies can work efficiently in terms of compu- Journal of Pavements Conference (IJPC)183:1–12.
tational time and resource requirements. Among [15] Khodaii, A., Haghshenas, H.F. & Kazemi, H. 2012.
various sampling strategies used in present work, Effect of Grading and Lime Content on HMA
best fit and worst fits were obtained using BBM Stripping Using Statistical Methodology. Journal of
and CCD. For a particular strain criterion, better Construction and Building Materials 34:131–135.
fit was obtained with second order response surface [16] Khodaii, A., Haghshenas, H.F. & Kazemi, H.
2013. Application of Response Surface Method-
when compared to first order response surface. ology to Evaluate Stone Matrix Asphalt Strip-
ping Potential. KSCE Journal of Civil Engineering
17(1):117–121.
REFERENCES [17] Khodaii, A., Haghshenas, H.F. & Tehrani, H. 2014.
Frequency and Temperature Interactive Effects on
[1] Anshu, M. & Kasthurirangan, G. 2007. Statistical Hot Mix Permanent Deformation using Response
Modeling of Accelerated Pavement Failure Using Surface Methodology. Journal of Materials in Civil
Response Surface Methodology. International Jour- Engineering 49(6):2425–2435.
nal of Civil, Environmental, Structural, Construction [18] Khodaii, A., Khedmati, M. & Haghshenas, H.F.
and Architectural Engineering 1(2):185–195. 2014. Statistical Evaluation of Hot Mix Asphalt
[2] Jiju, A. (2nd ed.). 2003. Design of experi- Resilient Modulus Using a Central Composite
ments for engineers and scientists. Burlinton: Design. ISSN 1997–1400 Int. J. Pavement Research.
Butterworth-Heinemann. Technology 7(6):445–450.

628
[19] Lopo, A. & Costa, F. 2012. Factorial and Doehlert Journal of Computers and Industrial engineering
designs in the optimization of a separation procedure 59:847–855.
using polyurethane foam as a solid phase extractant [24] Tekindal, M, A., Bayrak, H., Ozkaya, B. & Genc,
for platinum determination in ferrous matrices using Y. 2012. Box—behnken experimental design in fac-
inductively coupled plasma optical emission spec- torial experiments: The importance of bread for
troscopy. Analytical Methods 4(2):508–512. nutrition and health. Turkish Journal of Field Crops
[20] Mohammed, B., Ong, G., Khandaker, M. & 17(2):115–123.
Lachemi, M. 2012. Mix proportioning of concrete [25] Tijana, R., Vujanovic, I. & Ivanovic, D. 2014.
containing paper mill residuals using response sur- Comparison of Full Factorial Design, Central
face methodology. Construction and Building Mate- Composite Design, and Box-Behnken Design in
rials 35:63–68. Chromatographic Method Development for the
[21] Myers, R. H. & Montgomery, D. C. 1995. Response Determination of Fluconazole and Its Impurities.
surface methodology: Process and product optimi- Analytical Letters 47(8):1334–1347.
zation using designed experiments. New York: John [26] Toropov, H. & Polynkin, I. 1993. Multiparameter
Wiley & Sons, Inc. structural optimization using FEM and multipoint
[22] Penny, D. & Lindfield, G. (3rd ed.). 1999. Numerical explicit approximations. Structural Optimization
methods using MATLAB. New Jersey: Prentice-Hall 6:7–14.
Inc. [27] Yuanping, G. & Kewen, L. 2011. New models
[23] Steenackers, G. & Guillaume, P. 2009. Develop- for calculating the viscosity of mixed oil. Fuel
ment of an adaptive response surface method for 95:431–437.
optimization of computation-intensive models.

629
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Effectiveness of Spectral Analysis of Surface Waves (SASW) method

for pavement evaluation

Andreas Loizos, Christina Plati, Brad Cliatt & Konstantinos Gkyrtis

Laboratory of Pavement Engineering, National Technical University of Athens, Athens, Greece

ABSTRACT: The Spectral Analysis of Surface Waves technique is based on the dispersive nature of
surface waves generated and transmitted through a medium and recorded via geophones. This nonde-
structive technique can be potentially used for in-situ quality control during pavement construction.
SASW is investigated in this study to ascertain the technique’s effectiveness to estimate mechanical char-
acteristics in the base layer and subgrade of a pavement under construction. SASW measurements were
recorded at multiple locations and for investigative purposes Falling Weight Deflectometer measurements
were undertaken along the same test locations. Upon comparison of the calculated material characteris-
tics, deviations were observed between the SASW and FWD back-calculated moduli that are potentially
attributed to the varying strain ranges achieved during each technique. Although preliminarily found to
be promising, SASW requires additional research to more thoroughly investigate its applicability and
effectiveness for pavement quality control purposes, as well as to overcome potential weaknesses detected
during the current research.

1 INTRODUCTION layers (base-subbase layers). Araya et al. (2011)

claim that it is important to properly consider the
1.1 Problem statement behavior of unbound aggregate layers and sub-
grade soils in order to accurately predict pavement
For quality control purposes on under construc-
responses. Thus, it is vital to obtain accurate data
tion pavements, it is vital to determine the in-situ
to effectively evaluate the structural characteristics
achieved bearing capacity of the unbound layers
of the pavement during the construction phase of
early on during the construction process in order
the project.
to accurately compare in-situ achieved bearing
Currently, structural quality control is carried
capacity with the design values assumed for the
out by collecting data from the mechanical char-
unbound layers. Beyond this, the successful imple-
acteristics of the individual materials and layers
mentation of any mechanistic pavement design
using a variety of equipment. This process is
process requires the development of essential tools
directly followed by the processing and analysis
that can measure the modulus and thickness of
of the collected data; thus allowing for the deter-
each layer in the field (Nazarian et al. 2005).
mination of the bearing capacity. Data collection
Current laboratory methods available to test
can be performed either through destructive or
unbound layers do not fully reflect actual in-situ
non-destructive testing techniques, which are
conditions, while traditional in-situ testing such as
often preferred as they leave the structure and
the Dynamic Cone Penetrometer (DCP) can give
integrity of the newly constructed layers intact.
only point source information and it is also time
More traditional and time-consuming non-
consuming to gather large sets of data. Thus, a
destructive methods that are currently used for
need exists for a practical and portable method to
quality control purposes on the unbound and sub-
test the material in-situ, quickly over larger areas
grade layers include among others, the static Plate
during construction phases in order to effectively
Load Test (PLT), the Dynamic Cone Penetrometer
control compaction during construction. In-situ
(DCP) and the Light Falling Weight Deflectometer
non-destructive testing allows for the materials
(LFWD). While the Falling Weight Deflectometer
utilized during pavement construction to be tested
(FWD) can also be utilized for non-destructive
more extensively under realistic stress conditions
quality control, its function is most suitable when
with actual environmental and climatic conditions
the asphalt layers construction is completed (Chen
taken into consideration.
et al. 2001).
The overall structural and functional behav-
In addition to the above mentioned techniques,
ior of a pavement is strongly influenced by the
the Spectral Analysis of Surface Waves (SASW)
mechanical strength of the unbound aggregates

631
seismic method, better known in geotechnical 2 THEORETICAL BACKGROUND
engineering, has attracted increased interest from
road engineers, due to its theoretical validation for The Spectral Analysis of Surface Waves (SASW)
usage in pavement evaluation. According to Lu relies upon the theory of stress waves propagat-
et al. (2015), the SASW method is known to be ing in elastic media (Ellis, 2008). Particularly, the
very popular for its capability of estimating the method considers the dispersive nature of the sur-
thickness and elastic modulus of subsurface layers. face waves that propagate horizontally in the profile
when subjected to a transient impact load (Nazar-
ian & Stokoe, 1983; Nazarian et al. 2005; Lu et al.
1.2 Objective
2015). The dispersion of surface waves allows for
The objective of the present investigation was to the estimation of the elastic modulus and thickness
determine the effectiveness and/or limitations of of pavement layers. Nazarian et al. (1999) sum-
the SASW method for usage during quality con- marized the main advantages and potential weak-
trol measurements of pavements under construc- nesses of SASW for pavement evaluation. While
tion. As it is well known, funds and resources Widodo & Rosyidi (2009) and Rosyidi et al. (2007)
allocated for the construction, maintenance and have attempted to correlate SASW measurements
rehabilitation of infrastructure projects are lim- to other more conventional tools including the
ited. Taking this into consideration, a need cur- DCP index and CBR index.
rently exists for a quick, reliable and portable According to the elastic wave theory, the rate
in-situ method for quality control purposes during at which seismic waves move through a material
pavement construction. Widodo & Rosyidi (2009) (expressed by seismic wave velocities) is directly
refer to the SASW method as a cost-effective and related to the elastic material properties (stiffness)
non-destructive method, which is able to estimate of the material (Dennis & Bennett, 2005). The
accurately the mechanical strength of unbound propagation of surface waves (known as Rayleigh
and soil materials. With this in mind, the investi- waves) through an elastic half-space or a layered
gation of the effectiveness of the SASW technique medium (i.e. pavement system) is caused by a tran-
on unbound materials used in road construction is sient impact load imposed to the pavement surface.
an important objective of the current research. If The required equipment for SASW implementa-
the SASW method can be proven to be a reliable tion includes a source for generating surface waves,
and resource effective quality control method, a set of receivers (geophones), whose function is to
relevant stakeholders participating in pavement recognize the arrival of surface waves and an elec-
construction projects might consider replacing tronically operated system that performs spectrum
conventional methods utilized for quality control analysis and controls the overall procedure (Fig. 1).
purposes. When performing the SASW method, surface
For this purpose, a SASW system (GDS instru- waves of various frequencies are necessary in order
ments—primarily designed for geotechnical pur- to sample different depths. Various frequencies
poses) was initially tested on a pavement under are achieved by changing receivers spacing and
construction. To investigate and evaluate the sys- using different impact sources (Is mail et al. 2012).
tem’s effectiveness to accurately determine in-situ This is often accomplished by dropping multiple
achieved base and subgrade moduli, an in-situ hammers of different weight onto a metallic plate
experiment was conducted at multiple locations (source). In inversely dispersive pavement systems
along the pavement during the unbound layer con- (Rix et al. 1991), where materials of high stiffness
struction phase. The method’s applicability was
tested by performing SASW tests on the surface of Dynamic Spectrum
Analyser
the base layer, which was comprised of unbound
compacted gravel aggregate. For correlation pur-
poses, the SASW calculated moduli were com-
pared with the back-calculated FWD moduli for
the same test locations.
Due to the individual characteristics of the two
methods (engineering knowledge required, type
and value of imposed loading, different strain val-
ues) deviations are expected between the SASW
and FWD methods for their predicted resilient
moduli. For this reason, a correlation between the
methods was deemed necessary to be investigated,
so that the SASW method can be calibrated and
rated more robustly. Figure 1. Test set up for SASW technique.

632
exist at the upper layers, blows of high frequencies 3 CASE STUDY
must be aimed at shallow depths. At these depths,
the corresponding wavelengths are relatively small 3.1 In-situ procedure
and this is the rationale for strategically position-
In order to investigate the SASW method effective-
ing the source and the receivers at small distances.
ness and capabilities for pavement evaluation, an
As the depth of interest increases, the source and
in-situ testing regime was conducted on a pavement
the receivers are moved apart, so that waves of
under construction in southern Greece. SASW test-
reduced frequencies and increased wavelengths can
ing was performed along the surface of the base layer
be identified.
at multiple locations, hereafter, referred to as K1, K2,
As previously implied, SASW is based on the
and K3. The main material under investigation is the
dispersive nature of Rayleigh waves in order to
compacted unbound gravel of the base layer, while
obtain the shear wave velocity profile and deter-
the subgrade layer was also examined. The SASW
mine material properties (Is mail et al. 2012). To
system used in this study is from GDS instruments,
accomplish this in pavement systems, three basic
which was designed mainly for geotechnical usage.
steps need to be followed. Initially, wave genera-
For sensitivity analysis purposes, four source-
tion is caused by a transient impact load and wave
receiver arrays were investigated. Six geophones
motion is monitored by the receivers. Thereaf-
were utilized and were positioned at equal distances
ter, the processing of raw data (characteristics of
of 15, 30, 45 and 50 cm between the individual geo-
wave propagation) leads to the construction of the
phones. Additionally, for the 15 cm array a light
experimental dispersion curve, which is a plot of
hammer was utilized to potentially achieve higher
Rayleigh wave velocity VR versus wavelength λ.
frequencies. For the remaining array positions, an
Finally, an inversion of the experimental disper-
8-kg hammer was utilized. For each array, two rep-
sion curve occurs in order to estimate the real shear
etitions of five blows were carried out in order to
wave velocity profile, which is a plot of shear wave
enhance the quality of the observed signals due to
velocity VS versus depth d. A conceptual illustra-
wave motion and to reduce scatter.
tion of the inversion process can be found in Rix
The geotechnical nature of the system that sup-
et al. (1991).
ports the SASW’s philosophy should be taken into
When the shear wave velocity profile is deter-
consideration, since for the present experiment the
mined, the shear modulus G and the Young
system’s applicability for pavement sites is prelimi-
modulus E for each layer are calculated using the
narily investigated. Thus, for comparison purposes,
following equations:
the SASW estimated moduli are compared to the
(1) back-calculated moduli from the FWD measure-
G ρ VS2
ments at the same locations. SASW and FWD meas-
Ε = 2 ⋅ G ⋅ (1 + ) (2) urements were conducted simultaneously to ensure
identical environmental and climatic conditions.
where ρ is the mass density, Vs is the shear wave
velocity, ν is the Poisson’s ratio and G, E are the
shear and Young moduli respectively. If not 3.2 Data processing
known, typical values for the Poisson ratio ν and After each completed set of five blows, the col-
the mass density ρ are assumed without inducing lected data include for each wave: its frequency f,
significant error. wavelength λ and the value of the coherence func-
It should be noted that during field seismic tion. The receivers record signals in the time domain
tests, the observed strains are less than 0.001% which are subsequently transformed into the fre-
(Ismail et al. 2012). At this strain range, most soil quency domain (Is mail et al. 2012). During spec-
and unbound pavement materials behaviour is trum analysis, the coherence between the signals
linear elastic. Therefore, the modulus estimated and the phase information of the transfer function
by SASW (ESASW) corresponds to the maximum needs to be as high as possible. Accepted values for
value of Young modulus which is equal to the the coherence function were above the limit of 0.90.
maximum value of the resilient modulus. Many Given the f (Hz) and λ (m), the velocity of
articles refer to the seismic modulus ESASW as the Rayleigh waves VR (m/s) is calculated through the
low-strain modulus (Yuan & Nazarian, 2003; equation:
Correia & Loizos, 2004; Meshkani et al. 2004).
While multiple researchers (Nazarian et al. 1996;
V f ⋅λ (3)
Meshkani et al. 2004; Celaya et al. 2010) have R
highlighted the necessity to adjust the seismic
modulus to the strain range that occurs under Thereafter, VR is multiplied by a factor (1/K)
actual loads. that depends on the Poisson ratio v, to determine

633
shear wave velocity VS (m/s). The value of K is the in-situ seismic method. For the base layer,
calculated according to Dennis & Bennett (2005). an average modulus for the unbound compacted
By assuming a Poisson’s ratio of 0.35, the factor gravel of approximately 2650 MPa was calculated
1/K takes the value 1.0697. The shear wave velocity and for the subgrade, the moduli ranged from
VS is calculated through the equation: 373-690 MPa. Varying thicknesses and different
material properties may account for the observed
1 0697 (4) deviations. Regarding the bedrock, the modulus
S R
ranged from 983–1311 MPa.
Regarding the missing values of Table 1, base
Each wavelength λ (m) is converted to a sam-
modulus at K1 was not estimated due to absence
pling depth d (m), by dividing it with a factor that
of detected waves with proper frequency. At K2
ranges from 2 to 3. In the present study, the depth
the subgrade modulus was estimated to be approx-
is calculated by assuming d = λ/3.
imately 950 MPa, an unreasonably high value, and
From laboratory results of the modified Proctor
hence it was rejected. Also, no bedrock was con-
test, the in-situ density ρ was assumed to be equal
sidered at K3 because the subgrade was assumed
to 2.05 t/m3 (92% of the laboratory maximum dry
to extend up to 1.10 m.
density). Hence, the shear modulus G (MPa) is
Note that moving from locations K1 to K2, the
found from the equation:
estimated subgrade thickness increases (from 30 to
40 cm) and the bedrock modulus is smaller at K2.
G ρ VS2 ⋅10 −3 (5) When the underlying layers appear looser, the geo-
technical infrastructure needs improvement that
Finally, ESASW (MPa) is defined as: reflects a subgrade with higher bearing capacity.
In Figure 2, variations of modulus versus depth
E 2 G ⋅ (1 + ) = 2.7 ⋅ G (6) and variation of frequency versus depth are pre-
sented for the case of K3, where both base and
subgrade moduli were estimated. Constant moduli
The construction of the plot ESASW versus d
were considered for each layer, although a limited
becomes feasible. Thus, the dispersion of VR is con-
number of wavelengths was observed.
verted to dispersion of ESASW enabling the modulus
At the base-subbase, the observed frequencies
estimation for each layer. As it will be shown, the
were in the range of 790–1450 Hz. The deviation is
SASW technique provided results at larger depths
acceptable, because a progressive reduction in fre-
and this is the rationale for estimating the subgrade
quency is combined with an opposite progressive
modulus below base-subbase layer, which is of
increase in depth (and in wavelength as well), so
40 cm thickness.
that rational values are estimated for the base mod-
Upon comparison of SASW and FWD analysis,
ulus. This phenomenon is also obvious at depths
a subgrade with constant modulus that vertically
0.50–0.60 m (subgrade). The different order of
extends to infinity is assumed during FWD back-
calculation. As the SASW technique has the ability
to estimate different values of modulus at various Table 1. Averaged SASW moduli—ESASW (MPa).
depths, a subgrade with finite and varying thick-
ness was assumed in the present case study. Test locations Base-subbase Subgrade Bedrock
Subgrade thickness determination was based on
the estimated bearing capacity expressed by CBR K1 – 373 1311
K2 2445 – 983
(%) index. CBR index was assumed to be the 1/10
K3 2830 690 –
of back-calculated FWD subgrade modulus.
Below the subgrade, the existence of bedrock
was assumed, in order to account for different
stiffness estimation at larger depths. It is noted EsASw (MPa ) Frequency f (Hz)
0 1000 2000 3 000 0 1000 2000
that during the analysis, depths up to 1.10 m were 0 0

examined. 0, 1 0,1

E
; ; 0,3
o,2 E o,2
; ; 0 ,3 •
•
""'a 0.4 ""'~ 0,4
4 RESULTS "'
c 0,5 I c 0 ,5

0,6 I 0 ,6
I
The SASW moduli for the three test locations are 0,7 I 0 ,7

presented in Table 1 for each layer.

As expected, the estimated values are in the Figure 2. Modulus and frequency versus depth at loca-
high range due to the low strains achieved from tion K3.

634
magnitude in the frequencies probably exists due ferent perhaps due to the different mechanical
to different depths and different expected materi- strength of the materials used or due to other rea-
als’ stiffness. sons that require further investigation.
The limited number of the available waves with
proper frequency at small depths is maybe justi-
fied by system’s intended geotechnical usage, while 5 CONCLUSIONS
even the use of light hammer did not yield suffi-
cient and reasonable information for the unbound The use of seismic methods on pavement evalua-
gravel. The low system resolution is obvious at tion and potential quality control during construc-
many frequencies, especially at higher frequencies tion is promising. In the present study a GDS
that are necessary for the base-subbase layers. As system for SASW was utilized. Four different
previously implied, waves with coherence func- source-geophones arrays were implemented and
tion below the limit of 0.90 were rejected, because two hammers of different weight were used to gen-
a high coherence is desirable for enhancing signal erate wave propagation. During data analysis, a
quality and reducing background noise. Utilizing limited number of acceptable waves with sufficient
the available waves’ information (frequency, wave- frequency was observed and no source-geophone
length) led to the modulus estimation for the base array was deemed to be optimal, thus frequency
and the subgrade. control may be considered as a more critical aspect.
Below, a comparative assessment between Observed deviations between SASW and FWD
SASW and FWD results is presented. Compari- calculated moduli potentially arise due to the vary-
son results are tabulated in Table 2 and illustrated ing strain ranges that are achieved during the dif-
at Figure 3. Base moduli through SASW method ferent methods. For the subgrade, the ratio ESASW/
are estimated approximately 2.6–3.0 times higher EFWD was found to be larger than the ratio in base/
when compared to the FWD back-calculated mod- subbase, and this fact was attributed to the low
uli. This was to be partially expected due to the low geo-materials resistance to imposed strains in com-
strain range that occurs during SASW testing. The parison to unbound materials.
ratios appear to be slightly higher for the subgrade Similar trends for the estimated SASW moduli
layer where SASW moduli are estimated 4.0–4.6 have been observed in previous research as well.
times higher than FWD back-calculated moduli. However, the exact quantitative differentiations
The observed higher moduli (SASW vs. FWD) may depend on the different materials investigated
are in partial agreement with other international or are potentially due to other causes that require
in-situ studies. The exact moduli values are dif- additional investigation. Moreover, a correlation
needs to be derived between SASW and methods
Table 2. Comparison of SASW and FWD moduli. including the DCP and LFWD, which are most
commonly used on pavement construction evalu-
Test locations K1 K2 K3
ation of unbound and soil materials, in order to
Base-subbase conclude whether SASW is able to replace these
ESASW (MPa) – 2445 2830 more time-consuming conventional methods.
EFWD (MPa) 907 936 932 Regarding the SASW system used, additional
ESASW/EFWD – 2.6 3.0 modifications to the procedure followed need to
Subgrade be investigated to overcome the limitations of low
ESASW (MPa) 373 – 690 resolution at high frequencies. The SASW’s applica-
EFWD (MPa) 82 102 171 bility needs to be further tested on a wide range of
ESASW/EFWD 4.6 – 4.0 pavement materials and environmental conditions
to obtain a more complete view of the potential
applicability and effectiveness of the method for
potential utilization during pavement construction.
Ba.-e-Subbaselayer Sub grade layer

·~g
3000
2500 800
";;" 2000 ~
REFERENCES
600
;§, !500 .5A5W
e
"" 400
.5A5W
r.l 1000 • FWD r.l • FWD Araya, A.A., Huurman, M., Houben, L.J.M. 2011. Char-
500
0 ~} ~ 200
0
acterizing mechanical behavior of unbound granular
materials for pavements. TRB 2011 Annual Meeting.
Kl K2 K3 Kl K2 K3
Celaya, M., Nazarian, S., Yuan, D. 2010. Implementa-
tion of quality management of base materials with
Figure 3. Modulus from SASW—FWD at the base- seismic methods: case study in Texas, TRB 2010
subbase and subgrade layers. Annual Meeting.

635
Correira, A.G. & Loizos, Α. 2004. Geotechnics in pave- Nazarian, S., Pezo, R., Picornell, M. 1996. An approach
ment and railway design and construction. Millpress, to relate laboratory and field moduli of base materials.
Rotterdam. The University of Texas at El Paso.
Dennis, N. & Bennett, K. 2005. Development of testing Nazarian, S., Abdallah, I., Yuan, D., Ke, L. 1999. Design
protocol and correlations for resilient modulus of sub- modulus values using seismic data collection. The Uni-
grade soils, Arkansas State Highway and Transporta- versity of Texas at El Paso.
tion Department, Little Rock, Arkansas. Nazarian, S., Yuan, D., Tandon, V., Arellano, M. 2005.
Ellis, T.B. 2008. A comparison of nondestructive test- Quality management of flexible pavement layers
ing backcalculation techniques for rigid and flexible with seismic methods. Center for Transportation
pavements. An undergraduate thesis. The University of Infrastructure Systems. The University of Texas at El
Arkansas. Paso.
Ismail, M.A., Samsudin, A.R., Rafek, A.G., Nayan, Rix, G., Stokoe II, K., Roesset, J. 1991. Experimental
A.M. 2012. Road pavement stiffness determination study of factors affecting the SASW method. The
using SASW method. UNIMAS E-Journal of Civil University of Texas at Austin.
Engineering 3: 9–16. Rosyidi, S.A., Taha, M.R., Nayan, K.A.M., Chik, Z., Ismail,
Lu, Y., Cao, Y., McDaniel, J.G., Wang, M.L. 2015. Fast A., Siegfried, M. 2007. Development of a VS-CBR-DCP
inversion of air-coupled Spectral Analysis of Surface empirical model for determining dynamic stiffness of
Waves (SASW) method using in-situ particle displace- pavement base layer using SASW. The University of
ment. ISPRS International Journal of Geo-Informa- Yogyakarta.
tion 4: 2619–2637. Widodo, W. & Rosyidi, S.A. 2009. Experimental inves-
Meshkani, A., Abdallah, I., Nazarian, S. 2004. Determi- tigation of seismic parameters and bearing capacity
nation of nonlinear parameters of flexible pavement of pavement subgrade using Surface Wave Method.
layers from nondestructive testing. The University of Jurnal Ilmiah Semesta Teknika 12(1): 67–77.
Texas at El Paso. Yuan, D. & Nazarian, S. 2003. Variation in moduli of
Nazarian, S. & Stokoe II, K. 1983. Evaluation of modu- base and subgrade with moisture. TRB, Annual meet-
lus and thickness of pavements systems by SASW ing 2003.
method. Center for Transportation Research, the
University of Texas at Austin.

636
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Continuous health monitoring of asphalt concrete pavements using

surface-mounted battery-free wireless sensors

H. Hasni, A.H. Alavi, K. Chatti & N. Lajnef

Department of Civil and Environmental Engineering, Michigan State University, East Lansing, USA

ABSTRACT: This paper presents a new surface sensing approach for health monitoring of Asphalt
Concrete (AC) pavements utilizing a new class of self-powered wireless sensors. The proposed method
was based on the interpretation of the data stored in the memory gates of the sensor. A three-dimensional
finite element analysis was performed to obtain the dynamic strain at the surface of the pavement for
different damage scenarios. Damage states were defined using the element weakening method. The sen-
sor output data was generated from the time-history of the surface strains. Thereafter, the sensor data
was fitted to a Gaussian Mixture Model (GMM) in order to define an initial damage indicator features.
Finally, probabilistic neural network classification scheme was used to classify the damage states. The
results indicate that the proposed method is effective in detecting and classifying bottom-up cracks in AC
pavements using a surface-mounted network of sensors.

1 INTRODUCTION pavement deterioration (Dong et al., 2012; Xue,

2013; Lajnef et al., 2013; Yang et al., 2014).
Pavement health monitoring is an emerging field A major drawback of these methods is that
in pavement engineering for continuous damage they need precise individual placement, wiring
assessment and safety evaluation of the pavement systems and external power supply. To cope with
infrastructure system. Distresses concentrated in these limitations, recent development in the field
the asphalt concrete layer can lead to the failure of pavement health monitoring has revealed the
of the pavement structure over time. Bottom-up capabilities of the Wireless Sensors Networks
fatigue cracking is one of the main failure modes (WSN) in detecting damage in pavement structures
in Asphalt Concrete (AC) pavements. The fatigue (Chatti et al., 2016). A new class of Self-powered
life of pavements is mainly related to the nature Wireless Sensors (SWS) is developed by authors
and the amplitude of the applied loading (Miller at Michigan State University (MSU) (Alavi et al.,
and Belliger, 2003). A realistic modeling of the 2016a-b; Alavi et al., 2017). The advantage of this
dynamic loading as well as the pavements layers new sensing technology is that the sensor is har-
material properties are essential to provide accu- vesting the micro-strain energy of the host struc-
rate predictions of the mechanical response of the ture under mechanical excitation via a piezoelectric
structure. Previous studies were conducted to ana- transducer. This type of materials can be used to
lyze the effect of the analysis type on the pavement transform a mechanical loading into an electrical
response. A dynamic analysis may increase the charge using the direct piezoelectric effect process.
fatigue damage and rutting damage by 4 times and The generated electricity is used to empower the
at least 40% respectively(Cebon, 1986). Further- sensor and asses the health state of the structure.
more, other researchers showed that the dynamic A prototype of the sensor is shown in Figure 1.
response of the pavement is usually higher than The sensor has seven memory gates that cumula-
the response given by a quasi-static analysis. Static tively record the voltage/strain droppage over time
analysis may results in about 39% difference on (Figure 1(a)). Figure 1(b) is the interface board of
the tensile strain compared to a transient dynamic the sensor and Figure 1(c) is the sensor electronics
analysis (Yoo and Al-Qadi, 2007). Furthermore, part. The communication between the sensor and
Hot-Mix Asphalt (HMA) behaves as a viscoelas- a service vehicle could be effectuated using a Radio
tic materials and highly depends on both loading Frequency Identification (RFID) scanner to read
frequency and temperature. From a sensing per- the data stored on board the memory cells of the
spective, strain gages and traditional wired sensors sensor. Previous studies on the self-powered wire-
for pavement health monitoring are widely used to less sensor showed that cracks in pavement could
detect variations of the strain amplitude due to the be detected based on the interpretation of the data

637
 Sensor
V o ltage
/ Strain

(a) Time(s)
(b) (a)
(c) Sensor clcclronic. GMMfit

Figure 1. Prototype of the SWS.

Sensor
output
of a constant injection rate class of SWS (Chatti
et al., 2016). In their study, the SWS was embed-
ded inside the asphalt layer. However, the device
could be damaged and their replacement might be
expensive. Therefore, placing the sensors network
near the top surface of the pavement seems to be Gate Number

an attractive solution. In addition, for the case of a (b)

constant injection rate SWS, the data can be fitted
Figure 2. Procedure of obtaining the sensor output
to a Cumulative Density Function (CDF). How-
histogram.
ever, in this paper, each gate of the sensor has a
specific injection rate, which makes the interpreta-
tion of the data more complicated.
This study proposes a new method for pave- Sj S0 − I gj ∑T
j =1:7
i
j
(1)
ment health monitoring based on a surface sensing
approach. The proposed detection mechanism is
based on integrating the Finite Element Method where Sj is the sensor strain at gate j, Ti j is the
(FEM) and Probabilistic Neural Networks (PNN). duration of time intersection number i at the prese-
Intensive Finite Element (FE) analysis of a mov- lected threshold j (see Figure 2), and Igj is the injec-
ing load on a pavement section was performed to tion rate of gate j. The injection rates are property
obtain a realistic response. The proposed method of the sensors and they control the strain/voltage
uses features extracted from the sensors output droppage rates over time.
distributions to define initial damage indicators. As seen in Figure 2, there is a considerable
Thereafter, the extracted features from the WSN loss of the sensed information because the data is
were fused to increase the classification accuracy. compressed as a function of the cumulative time.
Therefore, a statistical method was proposed to
extract valuable features from the sensor distribu-
2 SMART SENSOR AND PROPOSED tion. In this paper, the output histogram is fitted to
DAMAGE DETECTION SYSTEM a Gaussian mixture model. GMMs are powerful
tools to describe many types of data.
The smart sensor is capable of continuously moni- The Probability Density Function (PDF) of a
toring the strain events within the host structure. Gaussian Mixture (GM) distribution is given by
As mentioned before, the memory cells records the the following expression:
cumulative drop of voltage/strain at a preselected
threshold level. M
ck ⎡ 1 ⎛ x − μ ⎞2⎤
A schematic representation of the working p (x) = ∑ exp ⎢ − ⎜ k
⎟ ⎥ (2)
k =1 2πσ k 2 ⎢⎣ 2 ⎝ σ k ⎠ ⎥⎦
principle of the sensor is presented in Figure 2. Fig-
ure 2(a) represents the input signal and Figure 2(b)
displays the output of the sensor. where: ( μk , σ k , k ,...M ) are mixture component
The recorded strain drop page is a function of parameters (mean and standard deviation) and ck
the cumulative time intersections and the gates ae the mixture weights. The mixture weights of the
injection rates as follow: PDF should satisfy the following condition:

638
Table 1. The preselected strain levels and the gate injec- Weakening Method (EWM) was used to introduce
tion rates considered for the analysis. the damage to the pavement. A total of 13 damage
states were introduced to the pavement including
Strain threshold Injection different combinations of the reduced AC modu-
Gate number level (με) Rates (με/s)
lus and the damage zone height. The damage zone
1 80 0.001000 was assumed to have a rectangular parallelepi-
2 100 0.005710 ped shape of 120 × 120 × h mm. The parameter
3 120 0.023162 h denotes the height of the damage zone which is
4 140 0.027822 variable in the analysis. The damaged AC volume
5 160 0.006562 was located at the center bottom of the AC layer.
6 180 0.005989 The modulus of the damaged zone was reduced to
7 200 0.032792 30%, 50%, 70%, and 90% from the instantaneous
modulus of the AC layer. The damage heights (h)
were 20 mm, 40 mm, and 60 mm.
M
The damage states are displayed in Table 2.Fig-
∑c
k =1
k =1 (3) ure 3 shows the damage location, cross section of
the pavement, and the measurement locations.

In this paper, the strain droppage histogram was

fitted by a bimodal (M = 2) GMM as follows: Table 2. Damage scenarios.

⎛ 7 ⎞ 2 αk ⎡ 1 ⎛ g − μ ⎞2⎤ Damage height Reduction in

Δε ( ) = ∑Δ ⎟ ∑ exp ⎢− ⎜ k
⎟ ⎥
Damage class (mm) modulus (%)
⎝ j =1 ⎠ k =1 2πσ
σ k2 ⎢⎣ 2 ⎝ σ k ⎠ ⎥⎦
C0 0 0
(4)
C1 20 30
where g is the gate number, ( μk σ k ) are the mixture
C2 20 50
C3 20 70
component parameters, α k is a parameter that rep- C4 20 90
resents the mixture weights and Δε j is the cumula- C5 40 30
tive droppage of strain at gate number j (j = 1,...7). C6 40 50
The GMM parameters ( μk σ k ) can be C7 40 70
regarded as damage indicators as they represent C8 40 90
the only viable tools to characterize the SWS data. C9 60 30
Furthermore, the strain response of the pavement C10 60 50
changes with damage progression in the structure. C11 60 70
Therefore, based on the relative variation of the C12 60 90
strain, the GMM parameters will change accord-
ing to the damage growth. Consequently, the dam-
age is logically considered a function of ( μk σ k ).
This will be shown below. Sensing location: Top
In this paper, the damage detection procedure
is divided into three major phases: (1) Structural
simulations of the targeted structure, (2) extraction
layer I
Surface of the AC

of preliminary damage indicator features, and (3)

damage detection and classification based on fusing AC
Damage z: ) .......
the sensors network data. The last step is achieved
through an Artificial Intelligence (AI) approach.
Base
For the analysis, the initial strain value of each
memory gate was set to S0 = 500 με. The gate injec-
tion rates as well as the strain threshold levels are
displayed in Table 1.
Subgrade

3 FINITE ELEMENT MODEL

3.1 FE model description

The FE simulations were performed using Figure 3. A schematic description of the damage and
ABAQUS (ABAQUS, 2010). The Element the measurement locations.

639
 The load contact area was assumed to be rec-
tangular (19.43 × 13.38 cm). The tire pressure was
862 kPa and the vehicle speed was 67 mph (30 m/s).
An implicit dynamic analysis was selected
for the analysis. In order to simulate the moving
load at the desired speed, a FORTRAN code was
developed to define the DLOAD subroutine in
ABAQUS.
The time increment was set to 0.001 s, and
therefore, the loading progresses by 30 mm (30
m/s × 0.001) during one time step.
The sensors were placed at the surface of the
AC layer. Up to 32 sensing nodes were used to
measure the strain changes due to damage pro-
gression. The network of sensors is displayed in
Figure 4. Figure 5. Loading path and FE model mesh.
The pavements layers were modeled using dif-
ferent material properties. The AC layer had vis-
coelastic properties while the base and subgrade
were assumed to behave as a linear elastic material.
A 5% damping ratio was defined for the base and
subgrade layers.
The pavement model was meshed using two
different types of elements: eight-node linear
brick elements with reduced integration (C3D8R)
and eight node linear infinite elements (CIN3D8).
The infinite elements were used to minimize the
reflection of the shear and dilatational waves
back into the FE mesh. The loaded strip as well
as the location of the infinite elements are shown
in Figure 5.

3.2 FE results
Figure 6 shows the time history of the first prin-
cipal strain (in absolute value) at the location of
sensor 1 for the C0 (intact), C4, C8 and C12 dam-
age classes. Figure 7 displays the results of the FE Figure 6. Strain history of sensor 1 for the intact, C4,
C8 and C12 damage classes.

analysis for different damage heights and modulus

reductions at sensors locations 1 and 25.
As it seen in Figure 6, the amplitude of the
maximum principal strain increases with dam-
age progression. Furthermore, the strain ampli-
tude continuously increases between damage
classes at sensor 1 location (Figure 7(a)). How-
ever, the behavior is inverted for sensor 25. This
sensor is located at an offset of 180 mm from
the longitudinal direction (parallel to the traffic
direction).
It can be concluded that when the sensor is
located along the wheel path, the strain increases
with damage progression, however, an offset with
respect to the longitudinal direction could create a
Figure 4. Sensors locations. different behavior of the strain amplitudes trends

640
 340 Sensor 1
- h=20mm
250
--e-- h = 40 mm
~ co
320 l

-
-;:;- 300
=L
---e-- h= 60 mm
200 =---· C12
.~-;
"" ·-.b 280
'0 .- 150
w
E~ :::!.
=~ '-'
.§ ·~ 260 w
.,"=
._ <] 100
::;<>':;

50
20 40 60 80 100
Percentage of Modulus Reduction ('Yo)
2 4 6 8 10
(a) Gate Number
Figure 8. GMM fit to sensor 1 data for the intact and
C12 damage classes.
75.5
(decreases) and the second mean μ2 shifts to the
right (increases). In addition, the standard devia-
tions σ1 and σ2 increase with damage progression
as the distribution expands. Therefore, the GMM
components seems to be a good indicators of
damage progression. Thus, the damage could be
defined as follow:
- h=20mm
--a-- h = 40 mm Damage = function (μ1 μ2 ,σ 1,σ 2 ) (5)
--e-- h = 60mm

20 40 60 80 100 However, if a sensor is located far from the

Percentage of Modulus Reduction(%) damage zone, the amplitude of the strain does not
change remarkably between damage classes, and
(b) therefore, the sensor histogram outputs become
Figure 7. Maximum principal strain vs percentage of
very similar. As a consequence, the GMM param-
modulus reduction for different damage heights at sensor eters (initial damage predictors) also become very
location: (a) 1 and (b) 25. close and the damage classes cannot be separated.
On the other hand, when the sensor is located at
certain offset from the traffic direction, the strain
as a function of the damage evolution in the struc- might decrease with damage progression. In this
ture as shown in Figure 7(b). case, the GMM parameters can detect the damage
event but it cannot estimate its severity (the dam-
age class). In order to overcome these issues, a data
4 DAMAGE DETECTION RESULTS fusion model was proposed based on the ‘sensor
group effect’.
4.1 Individual sensing results
After obtaining the FE results, the sensor output 4.2 Data fusion model
was generated as discussed in section 2. Thereaf- In this section, the initial damage indicator features
ter, the histograms of all the sensors were fitted to of all the sensors were fused to define a new set of
a GMM distribution in order to obtain an initial damage predictors based on the group effect of a
damage indictor features ( μk , σ k ), k = 1, 2. Fig- network of sensors concept. If a sensor is located
ure 8 displays the GMM fit to sensor 1 data for the far from the damage zone, the group effect of the
intact and the C12 damage states. sensors network will help to increase its sensitiv-
As it can be seen in the figure, the mean (μ1) of ity to the damage growth. The new features were
the first components of the GMM shifts to the left proposed as follow:

641
 ⎧ x1 x1ave where:
⎪ y1 = x
⎪ 1STD ⎧ x1 = μ1
⎪ y = x2 x2 ave ⎪x = σ 2
⎪ 2 x2STD ⎪ 2 1
⎨ (7)
⎪ x3 x3ave x
⎪ 3 = μ 2
⎪ y3 = ⎪⎩x4 = σ 22
⎪ x3STD
⎪ x4 x4 ave
⎪ y4 = x
a
and:
⎪ 4 STD

⎪ y = x1 x1STD − xi: The ith feature of the initial feature vector,

⎪ 5 x1ave − xiave: The average of xi for all patterns corre-
⎪ sponding to a specific damage state,
y=⎨ x2 x2STD
⎪ y6 = x − xiSTD: The standard deviation of xi for all pat-
(6)
⎪ 2ave terns corresponding to a specific damage state.
⎪ x3 x3STD
⎪ y7 = x As indicated by Equation 6, the new input
⎪ 3ave feature vector has 10 dimensions, and each com-
⎪ x4 − x4STD ponents is a function of the average/standard
⎪ y8 = x deviation of the initial feature vector components.
⎪ 4ave
Different feature selection algorithms were used
⎪ ( x1 + 3 ) − ( x2ave + x4ave )
⎪ y9 =
to select the best set of damage predictors. The
x1ave + x3ave Sequential Forward Selection (SFS), Sequential
⎪
⎪y = 2 ( x + x 4 ) − ( x1ave + x3ave ) Backward Selection (SBS), and the exhaustive
⎪ 10 x2ave + x4ave search methods were performed on the new feature
⎩ vectors. According to the results, the exhaustive

Confusion Matrix Confusion Matrix

8 0 0 0 100% 4 0 1 0 BOO%
5.8% 0.0% 0.0% 0.0% no% 13 .8'% 0 .0% 3.4% 0.0% 20.'"'

0 41 0 0 100% 0 0 0
0 .0"4 2 9.9% 0.0% O.Oo/e 0.0% 0.0% 0.0% 0.0% "'"'
0.0%

~ I
t----
~
" 3
100%
" 3
0 8 0 l OO'o/o

l i
0.0% 0.0% 27.6 % 0.0% 0.0%

- t---
0 0 0 100% 0 0 0 100%
0.0% 0.0% no% 33.6% 0.0% 0.0% 0 .0"4 0.0% 2 >1. 1%

100%
0.0%
100%
0.0%
100%
0.0%
100%
0.0%
100%
00%
100%
O.O'l'o .....
88.6 %

Target Class Target Class

(a) Training (b) Validation

Confusion Matrix

3 1 0 0 75.0%
10 .3% 3.4% 0.0% 0 0% 25.0%

0
0.0%
•
:'1 1.0"/..
0
0.0%
0
0 0%
100%
0 .0%

~
J!!

"s 3
90.0"/..
10 .0%

g
0 0 0 6 100%
0.0% O.O'l'o 0.0% 20.7% 0.0%

100%
0.0%
9CLOo/o
10.0%
100%
0.0%
85. ~
14.3% .....
03.1%

T a rget C lass

(c) Testing

Figure 9. Confusion matrixes for: (a) training, (b) validation and (c) testing.

642
search algorithm gives the optimal set of features are very dependent on the location of the sensor
that provides the best accuracy. The optimal set with respect to the damage location.
was found to be: 4. The data fusion model gives satisfactory dam-
age detection accuracy (93.1% on testing).
Soptimal {y y y } (8)

REFERENCES
For the classification, 4 damage classes were
defined based on the damage height as follow: ABAQUS, (2010). ABAQUS/CAE User’s Manual Ver-
Class 1: Intact structure sion 6.10, Dassault Systèmes.
Class 2: h = 20 mm Alavi, A. H., Hasni, H., Jiao, P., Borchani, W., & Lajnef,
N. (2017). Fatigue cracking detection in steel bridge
Class 3: h = 40 mm girders through a self-powered sensing concept. Jour-
Class 4: h = 60 mm nal of Constructional Steel Research, 128: 19–38.
The detection accuracy for the training, valida- Alavi, A.H., Hasni, H., Lajnef, N., Chatti, K., & Fari-
dazar, F. (2016a). An Intelligent Structural Damage
tion and testing data are equal to 100%, 96.55% Detection Approach Based on Self-Powered Wireless
and 93.10%, respectively. Figure 9 displays the con- Sensor Data. Automation in Construction 62: 24–44.
fusion matrixes. A confusion matrix is a table that Alavi, A.H., Hasni, H., Lajnef N., Chatti K., & Farida-
describes the performance the classifier. The col- zar F. (2016b). Damage Detection Using Self-Powered
umns of the each matrix represents the predicted Wireless Sensor Data: An Evolutionary Approach.
class, while each row represents the actual class. Measurement 82: 254–283.
Cebon, D. (1986). Road damaging effects of dynamic
axle loads. In: Proceedings of the international sym-
5 CONCLUSION posium on heavy vehicle weights and dimensions, Kel-
owna, BC, Canada: 37–53.
Chatti, K., Alavi, A. H., Hasni, H., Lajnef, N., & Farida-
This paper presents a new approach for detecting zar, F. (2016). Damage Detection in Pavement Struc-
bottom-up cracking in AC pavements based on tures Using Self-powered Sensors. In 8th RILEM
a surface sensing approach. Different numerical International Conference on Mechanisms of Cracking
simulations were performed to obtain the response and Debonding in Pavements (pp. 665–671). Springer
of the pavements under various damage scenarios. Netherlands.
The damage was created at the bottom of the AC Dong, Z. J., Li, S. L., Wen, J. Y., & Chen, H. C. (2012).
layer based on the reduction of its relaxation modu- Asphalt Pavement Structural Health Monitoring Uti-
lus and the damaged volume height. The pavement lizing FBG Sensors. Advanced Engineering Forum,
Vol. 5, pp. 339–344.
was subjected to a moving load at a highway speed. Lajnef N., Chatti K., Chakrabartty S., Rhimi M., &
The load was created using a DLOAD subroutine Sarkar P. (2013). Smart Pavement Monitoring Sys-
developed in FORTRAN programing language. In tem. Report: FHWA-HRT-12-072, Federal Highway
addition, the sensor output was generated using Administration (FHWA), Ishington, DC.
the dynamic strain extracted from the pavement Miller, J. S. & Bellinger, W. Y. (2003). Distress identi-
surface. Initial damage predictors were extracted fication manual for long-term pavement perform-
from the sensor output distribution by fitting the ance program (fourth revised edition). Report
sensor data to a GMM. Thereafter, a PNN classi- FHWA-RD-03-031, Federal Highway Administration
fication scheme was applied to a data fusion model (FHWA), Washington, DC.
Xue, W. (2013). Integrated Transportation Monitoring
derived from the initial damage indicators in order System for Both Pavement and Traffic. PhD Thesis,
to classify the damage states. Virginia Polytechnic Institute and State University,
The following conclusions can be drawn from Blacksburg, VA.
the results presented in this paper: Yang, S., Ceylan H., Gopalakrishnan K., & Kim S.
(2014). Smart airport pavement instrumentation and
1. The amplitude of the surface strains change health monitoring, Construction and Environmental
with damage progression. Engineering Conference Presentations and Proceed-
2. The location of the sensor with respect to the ings. Paper 8.
damage controls the trends representing the var- Yoo, P.J. & Al-Qadi, I.L. (2007). Effect of transient
iation of the strain amplitude with respect to the dynamic loading on flexible pavements. Transporta-
damage state (size and reduced AC modulus). tion Research Record, 1990: 129–140.
3. The bi-modal GM components are good sur-
face indicators of damage occurrence but they

643
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Instrumentation to evaluate the field performance of composite overlays

using accelerated pavement testing

Hashim Rizvi, Ayman Ali, Yusuf Mehta, Andrae Francoise, Caitlin Purdy & Aaron Nolan
Center for Research and Education in Advanced Transportation Engineering Systems (CREATEs),
Department of Civil and Environmental Engineering,Rowan University, Glassboro, New Jersey, USA

ABSTRACT: The objective of this paper is to explain the design and the process of installation of sen-
sors to compare the performance of thin composite overlays loaded using the heavy vehicle simulator.
The sensors were designed and installed to capture the critical mechanical responses in the surface layer,
load transfer efficiency in the slab, and stresses in the unbound subbase layer. The purpose of the instru-
mentation is to develop a prediction model. Four types of sensors were installed in this study: a) pressure
cells in unbound subbase; b) soil compression gauges in unbound base, and c) asphalt strain gauges in
surface layer of HMA and thermocouples in lower subbase as well as in surface HMA layer. Difficulties
and challenges were encountered and possible solution and recommendation are also explained in this
research paper.

1 INTRODUCTION models [2]. It can be noticed that first objective

depends on successful completion of the second
In the state of New Jersey (NJ) approximately 50% objective. Collection of reliable and continuous
of pavements are composite pavements. These data from the test sections is the key to evaluate
pavements are typically composed of a Portland or compare and to develop models. Instrumenta-
Cement Concrete (PCC) layer overlaid with a thin tion is the key to fulfill the second objective. Types
asphalt mixture layer. Thin asphalt layers are typi- and quantity of instrumentation depends on the
cally used as a quick and inexpensive rehabilitation information required. For example, if temperature
and preservation strategy for aged and deterio- gradient needs to be recorded, thermocouples will
rated PCC pavements [1]. Several types of asphalt be used to measure the temperature in the surface.
mixes have been used successfully to construct thin Quantity of thermocouples will also depend upon
asphalt overlays such as: Ultrathin Bonded Wear- the point of interests such as: number of layers
ing Course (UTBWC), High Performance Thin i.e. Sub-base, Base, and PCC and wearing course,
Overlays (HPTO), Stone Matrix Asphalt (SMA), Thicknesses of the layers, number of joints, etc.
Open Graded Friction Course (OGFC), Rubber- Likewise, sensors are installed in the test section at
ized Gap Graded (RGG), 9.5 Nominal Maximum various places of interests and depths to measure
Size, Binder Rich Intermediate Course (BRIC) certain characteristics such as: elastic and perma-
etc. [1]. The most effective, fast, and reliable way nent vertical deflections, pressure in compacted
of evaluating the performance of the thin overlays soils, vertical deformations, longitudinal strains
treatments over PCC is the Accelerated Pavement etc. These characteristics are measured using spe-
Testing (APT) [2]. APT is generally defined as con- cific sensors.
trolled application of a realistic wheel loading to a In this research project the New Jersey Depart-
pavement system simulating long-term, in-service ment of Transportation (NJDOT) required per-
loading conditions [3]. APT has predominantly formance evaluation of four types of thin overlays
used in recent years to evaluate the performance over the PCC slabs i.e. 9.5 ME, 12.5 SMA, HPTO
of flexible overlays [3]. As the name suggests, APT and BRIC. Rowan University Accelerated Pave-
is used as a means of providing rapid comparison ment Testing Facility (RUAPTF) named; Center
and evaluation results that can be used to compare for Research and Education in Advanced Trans-
the performance [2]. Generally there are two objec- portation Engineering Systems (CREATEs) was
tives to perform APT: 1) comparison of treatments awarded the project to build six test sections, apply
with control or within different treatment types; instrumentation and evaluate the performance of
and 2) collect performance data under controlled thin overlays on PCC slabs. This research paper
climate and loading conditions for calibration of covers the instrumentation part of the project.

645
2 OBJECTIVE 5 ft. 5 ft. 5 ft. 5 ft. 5 ft. 5 ft.

The objectives of this research study can be sum-

marized as:
− Design the sensor layout to capture the critical
mechanical responses.
− Observe and report limitation and challenges
related to installation of sensors.

3 METHODOLOGY AND
INSTRUMENTATION

As mentioned earlier that type, number and selec-

tion of instrumentation depends on the require-
ments of the client; therefore, NJDOT was
consulted for the required data and characteristics
to be determined during the accelerated testing. It
was found that joint movement of slabs and reflec-
tive cracking in thin overlays due to the joint move-
ments was required to be observed. The first step
was to determine what types of movements need
to be captured during field testing. It was decided J ft. 5 ft. :> ft. 5 ft. 5 ft. 5 ft.
that vertical, longitudinal and lateral movements
of PCC slabs along with temperature (in lower c:::::J Longitudinal Asphalt Strain Gauges (Total: 2)
and upper surfaces) and pressure in lower surfaces
need to be measured. These characteristics would
0 Pressure Cell (Total: 2, 2 600KPa and 1
250KPa)
provide detailed information to determine the
impact of joint movements on thin overlay. c:::::::J Macrosensors LVDT (Total: 2)
Sensors were selected to capture all the afore-
mentioned movements. List of the sensors is pro- D CTL Soil Compression Gauge (Total: 2)
vided below:
_,Ac_ HMA Temperature Sensors (Total: 3 T-type
− Pressure Cells (PC) "'¥"' Thermocouples)
− Soil Compression Gauges (SCGs)
− Asphalt Strain Gauges (ASG) IIWI Type T thennocouples will be used for
− Thermocouples II temperature measurements.
First four of the mentioned instrumentation Type T thennocouples will be used for
were installed in six test sections. The location of temperature measurements.
the sensors to be installed was a critical decision as
this depends on the nature of experimentation and
required data. This research study focusses on joint Figure 1. Sensor layout to capture the joint movements
movement and reflective cracking in thin over lays, and reflective cracking in the thin overlays.
therefore, all the sensors were installed to capture
all types of motions under or around the joints. Slab one and slab two represent PCC slabs i.e.
Application of sensors will be discussed in installa- below the thin overlay of hot mix asphalt. PCs are
tion part of this paper. applied right underneath the slab joints, SCGs are
Layout of the sensor locations also depends installed across the joints under both slabs, LVDTs
upon the required data and purpose of experimen- are installed on the sides of the joints, ASGs are
tation. The layout of the sensors was developed for installed in HMA on top of slab joints, and thermo-
this research study and shown in Figure 1. Cross- couples are installed in I-3 and asphalt layers. The
section of the test section can also be observed in total length of the PCC slab is 15 feet and width
Figure 1. It shows that 3 inch of HMA is on top is 12 feet. The dimensions of one test section is 30
of 8 inch PCC (the HMA can be any of the four by 12 feet. The wheel of HVS used for AP testing
mixtures). Fifteen inches of I-3 is below PCC layer was 24 inches wide. The combination of two tires,
(divided in two layers of 7.5 inches). Compacted nine inches each with a gap of 4 inches between
soil of five inches is on top of the regular soil bed. both tires, was used for this testing. HVS axle

646
will be rolled on top of HMA in the center of the the two plates together causing a corresponding
test section to replicate the field truck traffic. The increase of fluid pressure inside the cell. A vibrat-
traffic type selected for this research project was: ing wire pressure transducer converts this pressure
Unidirectional-Forward position. The HVS axle into an electrical signal which is transmitted to the
will apply load in only one direction and that direc- readout location i.e. Data Acquisition System.
tion will be forward. This setting was selected to
simulate one-way traffic. The placement of ASGs
4.1.2 Application procedure
also depends on the size and location of HVS axle.
Six Pressure cells, in total, were applied under
It can be observed that ASGs are installed at two
NJDOT sections i.e. each test section was applied
different locations. One ASG will be right under
with one pressure cell. The application procedure
right tire of the axle and the second one will be on
for PCs is provided as follows:
the side of left tire of the axle. The fatigue strains,
responsible of fatigue cracking will be captured by 1. Initial laboratory readings were taken to verify
right ASG and compressive strains will be captured if the PCs were working properly before instal-
by left ASG. All the test sections were in the open lation in the soil bed as the process required a
to simulate the field conditions. The weathering lot of compaction activity and that would have
effect (freeze-thaw) is one of the major factors in been very cumbersome to find that PC was not
cracking. Therefore, it was decided to perform test- working after installation. Figure 2 shows three
ing in the open to avoid any artificial environment pressure cells prepared in laboratory before
for the test section i.e. indoor test section. installation in the test section. It can be observed
that each PC is labeled with a number. This
number represents the corresponding test section
4 INSTALLATION OF SENSORS i.e. the PC labeled 1 is applied in test section 1, so
on and so forth. The numbers were also used as
As mentioned in previous section, PC, SCGs, reference for readings during field testing.
ASGs, LVTS and thermocouples were selected to 2. Reference lines or threads were established to
be installed in the test sections. The installation identify the exact location and depth of the
process and constraints will be discuss independ- pressure cells. It can be observed in plain view
ently in the following sub sections: that the position of the PC was right under-
neath the two joints of PCC slabs, in the lower
I-3 layer of the pavement structure (48.2 cm
4.1 Pressure Cell (PC) installation
(19 inches) below the surface) and under the
4.1.1 Introduction right tire of the axle. Once the references are
Pressure Cells (PCs) are used to measure the total established the location was identified to pre-
pressure in earth fills. All cells consist of two cir- pare a soil bed (about 5 cm (2 inches) deep
cular stainless steel plates (22.86 cm diameter) and 30 cm (12 inches) diameter) to place the
welded together around their periphery and spaced PC in the lower subbase (Figure 3). A 7.6 cm
apart by a narrow cavity filled with de-aired oil (3 inches) wide and 5 cm (2 inches) deep trench
(Figure 2). Changing earth pressure squeezes from location of PC to the manhole was dug for
sensor receiver part of the PC and cord to be
placed in from PC to the manhole (Figure 3).
3. PCs were placed in trench beds and levelled in
order to achieve balanced data from the top
(Figure 3). The plate of PC was placed in the
30 cm diameter trench bed and the wire was
placed in the 7.6 cm (3 inches) wide trench
heading towards the manhole (Figure 3).
PCs were covered with finely graded soil
(< 1.2 mm or US 16) in order to avoid bigger
stones hitting the plate or altering the pressure due
to point loading effect. Installed readings were
taken at this point to explore if the sensor still
received excitation upon supply of voltage and
also verified that it sensed the load upon standing
on top of the buried PC plate. Figure 4 shows that
PCs were covered with finely graded soil.
Figure 2. Pressure cells ready after taking the initial 4. The whole area was compacted using regular
readings at laboratory. I-3 material and the position of the sensor was

647
 extremely important to understand its basic struc-
ture, functionality, ranges as to better determine
the right type and material of thermocouple for an
application.

4.2.2 Application procedure

There were forty-eight thermocouples applied
in six NJDOT sections i.e. eight thermocouples
for each section. Thirty out of forty-eight were
installed in the subbase (I-3). Thermocouples were
installed at a distance of 2 inches from top to bot-
tom on a dowel in lower subbase (as shown in
Figure 5). Application procedure for thermocou-
ples is provided as follows:
1. Thermocouple wires were measured and cut
Figure 3. Placement and levelling of the PC on ground in laboratory. The required length of wire was
bed. about 1.5 meter (5 feet) however in order to be
on safe side approximately 2.4 meter (8 feet) of
wire was cut for each thermocouple.
2. Pairs of wire strands were separated in a con-
sistent manner i.e. pairing wires with the same
number (1-1, 2-2, etc.). This was critically impor-
tant step as if these wires were not paired in
proper manner the result could have been erratic.
3. Thermocouple wires were tested using a digital
thermometer. It was ensured that all the ther-
mocouples read between 23.8 to 26.6°C (75
to 80°F) because the instrument testing was
conducted at room temperature conditions. In
order to connect both wires on one end, quick
tip connectors were used. Shrink tubing was
placed over the quick tips (Figure 5) to avoid
damage in the soil. Heat gun was used in order
to let the tubes shrink and hold the quick tip
Figure 4. Marking the location of PCs in order to get connectors. Tip of the shrink tubing was cut
GPS locations. after sealing the shrink tubing.

marked using spray paint in order to take GPS

locations for future references.

4.2 Thermocouple installation

4.2.1 Introduction
Thermocouples are simple, robust and cost-effec-
tive temperature sensors used in a wide range of
temperature measurement processes. They consist
of two dissimilar metal wires, joined at one end.
When properly configured, thermocouples can
provide measurements over a wide range of tem-
peratures. Thermocouples are manufactured in
a variety of styles, such as thermocouple probes,
thermocouple probes with connectors, transi-
tion joint thermocouple probes, infrared ther-
mocouples, bare wire thermocouple or even just
thermocouple wire. They are commonly used in
a wide range of applications. Due to their wide Figure 5. Layout of thermocouple wire wrapped
range of models and technical specifications, it is around dowel.

648
4. A 25.4 cm (10 inch) dowel was cut and graded at 9. Once location was identified the asphalt spray
5 cm (2 inch) intervals (Figure 5). Tips of sealed was made to place the first thermocouple on
thermocouple wires were placed on graded top of the concrete, as shown in Figure 7.
intervals and were tied with dowel using small 10. Wires of the thermocouples were joined using
nylon cable ties (Figure 5). Thermocouple were the quick tip and shrink-tubing technique.
designated numbers to ensure that the tempera- However, thermocouples were directly installed
ture reading was acquired from the right place on the surface instead of mounting them on
in the subbase. dowel bar at designated distances. A sequence
5. The locations of thermocouple were identified of thermocouple with 19 mm thick layer of
i.e. 1.22 meter (4 feet) away from the center line loose HMA was adopted until the third ther-
of the test section and 1.52 meter (5 feet) away mocouple was successfully installed followed
from the PCC slab joint under the right slab by 38 mm layer of loose HMA to achieve the
of the test section in lower subbase. A vertical total thickness of the surface layer i.e. 76.2 mm.
hole of about 28 cm (11 inches) deep was dug The heights of layers were determined using
in subbase using pointed sharp tool to place the measuring tape and compaction was performed
graded dowel having thermocouples tied up at using hand compactor (Figure 7). The desired
desired heights (Figure 6). A 7.5 cm (3 inch) distances were achieved through compaction.
wide and 7.5 cm deep trench was also dug from
the thermocouple location to manhole to place
4.3 Soil Compression Gauges (SCG) installation
the hard rubber pipe that carries the cord. Posi-
tion of numbered thermocouples were noted 4.3.1 Introduction
and verified from top to bottom. The thermo- SCGs are used to measure compression of the soil at
couples were placed in descending order from a particular depth of the pavement structure under
top to bottom i.e. 5, 4, 3, 2, 1, number 5 being loading conditions. Figure 8 shows the SCG before
on top (near surface) and number 1 being at the application in soil. Displacement of two plates of
bottom of hole. SCG is measured by a sensor i.e. placed between the
6. Vertical hole and trench was filled using finely two plates. The sensor translates the displacement
sieved soil material (< 1.2 mm or US 16) and it into voltage that is sent back to the data acquisition
was ensured that the dowel is vertically straight system. The resulting data is measured in percent
during compaction (Figure 6). The soil temper- strain. This instrument does not have a spring to
ature was measured using temperature gun and return it to normal length; therefore, this gauge can
then thermocouple reading were measured. The only measure plastic, or permanent, compression in
reading was found to be between 23°C to 30°C. the soil. The initial distance (fully extended position)
7. Trenches were completely filled with I-3 mate- between the two plates is 152 mm (6 in.).
rial and locations of thermocouples were
marked using spray paints. 4.3.2 Application procedure
8. In order to apply the thermocouples in asphalt Twelve SCGs were installed in six NJDOT sec-
layer, location was identified using center joint tions i.e. two SCGs for each section. The SCGs
of PCC slabs and manhole.

Figure 7. Installing thermocouples in asphalt layer.

Figure 6. Filling the hole manually using finely sieved Measuring the heights of the layer to install the second
soil material while keeping the dowel vertically straight. thermocouple.

649
Figure 8. Soil compression gauges.
Figure 9. Compression gauge stuffed with fine graded
sieved soil.
were applied in upper layer of subbase (Figure 1).
Application procedure for thermocouples is pro-
vided as follows: condition, was used to stuff the SCGs in pipe
assembly. Hands and small wooden piece were
1. SCGs were tested in laboratory prior to instal- used for compaction. The SCGs fully stuffed and
lation to verify if they were providing correct ready to go in trench are shown in Figure 10.
readings. The manufacturer provided standard 3. Third step was to establish the reference lines
reading at fully-compressed position. It was and location of SCGs. Wires or threads and
determined that sensors 7 to 11 did not match measuring tapes were used to establish loca-
the manufacturer readings at fully compressed tions. The exact locations of SCGs in each test
position and at 1 volt excitation. Therefore, fac- section is provided in Figure 1. It was desired to
tors for calculations, provided by manufacturer put both PCs and SCGs under the center line
for SCGs measurements, were also not identical of the test section. However, it can be observed
in sensors 7 to 11. Therefore, sequence of read- that only SCGs were placed under the center
ings were taken from fully compressed position line (Figure 1). Initially, all PCs were placed
to the fully extended position after an interval under the center line of the test section. Later, it
of 7/8th of an inch. New factors were calcu- was found that the SCG trench needed to be at
lated by measuring slope of the curve of new least 28 cm (11 inch) deep in order to place the
values and distances. Moreover, it was desired 24.2 cm (9.5 inch) of the SCG pipe sleeve in the
not to install the SCGs in field having more upper layer of I-3. This could damage the PCs
than 20% compression distance. Therefore, the underneath as PCs were only 20.3 cm (8 inch)
SCGs readings were recorded, in laboratory, at down. Therefore, SCGs were moved about
10 Volts excitation and compression distance 38 cm (15 inch) away from the center line. How-
of 20%. The reading were again acquired after ever; contractor mentioned that PCC slabs need
installing the SCGs in test sections. It was deter- at least 30.5 cm to 38 cm (12 to 15 inch) space
mined that except first two sensors, all other to put the concrete molds in place for the slab
SCGs were either fully extended (0%) or well construction. This resulted in moving whole test
below 20%. The reason of excessive compres- section. Figure 1 now shows that the SCGs are
sion in first two sensors was lack of experience. under the center line whereas the PCs are 38 cm
These were the first two sensors installed by the (15 inch) away from the center line. The SCGs
team. The compaction was a bit too much that were supposed to measure the compression
showed excessive compression in SCGs. across the joint of PCC slabs; therefore, they are
2. In order to place SCGs fully extended in the placed 14 cm (5.5 inch) across the joint at each
soil, it needed to be stuffed with finely graded side.
soil (< 1.2 mm or US 16). Research team used a 4. A 16 inches wide and 11 inches deep trench was
7.5 cm (3 inch) diameter PVC pipe by cutting it prepared to place both SCGs (Figure 9). The
from one side to place the SCGs in the pipe and soil bed of trench was leveled using finely sieved
stuff it with damp soil (Figure 10). This technique soil to prepare the bedding for the compression
was effective to keep the SCGs fully extended gauge. A 7.5 cm (3 inch) wide trench, from sen-
while being compacted in the trench. Approxi- sor location to the manhole, was also prepared
mately 1500 grams of finely graded soil in SSD to pass the wires of SCGs.

650
Figure 10. Mark the location of SCGs using the refer- Figure 11. Asphalt Strain Gauge (H-Gauge).
ence lines.

construction. Signal conditioning complexity are

5. Pipe molding, carrying SCG, was vertically reduced by fully incorporated Wheatstone bridge
placed (wire at the bottom) in the trench and circuit. There are four active electrical resistance
assembly was carefully and manually installed strain gauges incorporated in each sensor: two
using finely graded sieved soil material aligned in the longitudinal direction and two in
(< 1.2 mm or US 16). As soon as the trench is the transverse direction. Twelve (12) Tokyo Sokki
filled about quarter of the pipe height, the pipe KM-100HAS gauges shown in figure were used in
was gently pulled about 5 cm (2 inch), leaving this research project i.e. two for each test section.
down the stuffed SCG. Filling and compaction The Figure 1 shows that the ASGs were installed
started again using dowel, especially around on top of the concrete at the interface of flexible
the pipe assembly. This process was repeated and rigid pavements. Working temperature range
until the whole pipe is pulled out leaving the of ASGs was significantly higher than the com-
fully extended-stuffed SCG in the trench. This paction temperatures of flexible pavements i.e.
process was performed with great care, as the (185°C vs 142 to 162°C in our case).
compression gauge could be compressed during
pipe pulling process. Readings were taken along
4.4.2 Application procedure
with compaction to ensure that no extra com-
paction (> 20%) has been experienced by SCGs. 1. ASGs were tested prior to installation in the
Figure 19 shows the SCGs fully compacted in test sections. ASGs were tested prior to instal-
trench. The trenches were filled using regular lation in the test sections. The initial readings
I-3 material and location of SCGs were care- were taken to verify if the sensors were working
fully marked as shown in Figure 10. properly i.e. responding to the excitation and
signal is stable over a period of few seconds.
2. Locations were carefully marked with reference
4.4 Asphalt Strain Gauges (ASG) installation
to sides and center of the slabs. Figure 12 shows
4.4.1 Introduction that one of the ASGs in each test section is
Asphalt Strain Gauges (or H-gauges) are used to under the right tire of the wheel; whereas, the
simultaneously measure longitudinal and trans- other one is on the side of the left tire of the
verse strains. A typical H-gauge consists of an wheel.
electrical resistance strain gauge embedded within 3. Concrete was cut using hand-driven saw to make
a strip of glass-fiber reinforced epoxy with trans- trenches for ASG cords. The cut was applied
verse stainless steel anchors at each end of the strip two time to widen the cut enough to hold two
to form an H-shape (Figure 11). It is important cords of ASGs. It is important to note that the
that the strip stiffness is approximately the same as cut was only 19 mm (0.75 inch) deep in order
the HMA stiffness [4]. It is recommended that the to accommodate both wires one over another.
length of the sensor should be three to five times 4. The asphalt binder was sprayed at identified
the maximum aggregate size because the H-gauge location of both ASGs and a bed of loose
measures the average strain between the anchors HMA was prepared to place the ASG. The
[5]. ASG are designed especially to withstand bed was 12.75 mm thick (Figure 12). The hand
the high temperature and loads during pavement compactor was used to prepare the loose HMA

651
 of sensors were installed in three phases. The fol-
lowing observations were made:
1. Each sensor installation phase was interlinked
with the following phase. For example soil com-
pression gauges needed to be moved from the
designated locations, initially planned, to avoid
collision with pressure cells.
2. Attention to detail is needed to ensure robust
measurement, such as, over compaction of first
two soil compression gauges resulted in more
than 20 percent compression.
3. Selection of appropriate time for installation is
critical. For example, soil compression gauges
took more than expected time to be installed
resulting in insufficient light and tired staff that
can cause serious damage to the whole purpose
of careful installation.
Figure 12. Cutting concrete using diamond saw to 4. Provide proper guidelines to construction staff
place ASG cords. is also one of the important things to be con-
sidered during construction. For example one
bed. The ASGs were placed on the desired of the soil compression gauges was damaged
location and were embedded in the bed using due to application of heavy nail to install mold-
the weight of the body. ing for concrete slab construction. That was
5. The ASGs were covered using loose HMA and replaced in time.
were manually compacted to a height less than 5. Post construction evaluation of sensors: Check-
the thickness of the pavement layer. ing of sensors after every activity is also crucial
6. The cords of the ASGs were placed in the cut to verify if the latest activity damaged the sen-
trenches and sealed to avoid any lose cords or sor by any means.
seepage. The sealant was applied in abundance 6. Protecting the sensor cables from water and
to ensure complete filling of the cut. other construction activities is also one of the
7. The ASG location were identified using GPS most important activities to avoid signal dis-
location tool. The paint marks were applied on ruption during field testing. If sensor cords are
top of the paved surface before compaction to not properly protected and damaged due to any
ensure careful compaction of the sensor zone. activity then the whole purpose of careful appli-
The reason for this care is the shallow depth of the cation of sensor is jeopardized.
surface layer i.e. 3 inches. Had the thickness been
six or more inches then the compaction impact
would have not been significantly dangerous. REFERENCES

[1] A. Ali, Y. Mehta and H.R.R. Rizvi, “HVS Evalua-

tion of Thin Asphalt Overlays on Composite Pave-
5 CONCLUSIONS ments,” New Jersey Department of Transportation,
Trenton, NJ, 2016.
The installation of sensors is extremely sensitive [2] J.T. Harvey, “Introduction,” Use of Accelerated
and critical stage of any field testing. The field Pavement Testing to Evaluate Maintenance and Pave-
testing depends on the results acquired through the ment Preservation Treatments, pp. 1–10, September
instrumentation installed in the test sections. Once 2009.
installed, it is not practical to excavate and fix the [3] FDOT, “Instrumentation of Florida’s Accelerated
sensor by any means as this will create an artificial Pavement Testing Facility,” FDOT- State Materials
fracture in the pavements and the results will never Office, 2011.
[4] G. Hilderbrand, “Verification of Flexible Pavement
be conclusive. Therefore, experienced professionals, Response for a Field Test,” Danish Road Institute,
detailed planning (i.e. weather details, equipment Roskilde, Denmark, 2002.
and tools availability, materials and supporting [5] N. Tabatabaee, I.L. Al-Qadi and P.B. Sebaaly, “Field
staff etc.), identification of effective spots and Evaluation of Pavement Instrumentation Methods,”
careful installation are the key factors affecting the Journal of Testing and Evaluation, vol. 20, no. 2,
whole project. In this research study different types pp. 144–151, 1992.

652
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Assessment of the bearing capacity of pavements

using fiber optic sensors

V. Papavasiliou & A. Loizos

Laboratory of Pavement Engineering, National Technical University of Athens (NTUA), Athens, Greece

ABSTRACT: The present paper aims to investigate the contribution of the Fiber Optic Sensors (FOS)
system towards the assessment of the bearing capacity of asphalt pavements that have been Cold In-depth
Recycled (CIR) utilizing foamed asphalt as a stabilization treatment for base works. The investigation is
related with the need of understanding the behavior of asphalt CIR pavements under heavy traffic loads
that can result in the prediction of pavement performance. On this purpose, a testing experiment was con-
ducted to monitor and document the in-situ developed strains due to real traffic loading. The developed
strains were measured using a FOS system. The collected data were processed and analyzed taking into
account the local conditions.

1 INTRODUCTION measurements of strain gauges are very sensitive to

a lot of parameters (adhesion, horizontality, pres-
Pavement performance prediction is connected ence of aggregates etc.) (Bueche et al. 2009).
with the locations in the pavement structure, where Apart from the comparison of the calculated and
the critical pavement responses (stress or strain) measured strains, an important issue is to investigate
attain their most extreme values. Stress and strain the impact of the heavy traffic loading. Towards
can be measured or calculated from mechanistic this goal a simple idea is to measure strains under
analysis. The calculation of stresses and strains real traffic loading and to compare with the related
from mechanistic analysis requires knowledge of ones induced from test truck and FWD loading.
the material’s mechanical characteristics, such Taking into account the above considerations,
as elastic modulus and Poisson’s ratio of each the present research investigates the contribution
layer. Material moduli can be obtained by labora- of the Fiber Optic Sensors (FOS) system towards
tory testing, or in-situ by Nondestructive Testing the assessment of the bearing capacity of asphalt
(NDT) (Rens et al. 1997). pavements that have been Cold In-depth Recycled
Nondestructive testing uses deflection basin (CIR) utilizing Foamed Asphalt (FA) as a stabili-
data generated from a loading device to quantify zation treatment for base works. The investigation
the response of a pavement structure to known load is also related with the need of understanding the
drops. The Falling Weight Deflectometer (FWD) behavior of CIR pavements under heavy traffic
is the most common system for measuring non- loads that can result in the prediction of pavement
destructive surface deflections. Generally, surface performance. On this purpose, a testing experiment
deflections obtained from FWD testing are used to was conducted by the Laboratory of Pavement
back-calculate in situ material properties (Kim & Engineering of the National Technical University
Park 2002). The back-calculated moduli are con- of Athens (NTUA) to monitor and document the
sidered representative of the pavement response in-situ developed strains due to real traffic load-
to load and can be used to calculate stresses and ing. The developed strains were measured using a
strains in the pavement structure. However, the FOS system that is perfectly suited to measure any
accuracy of the results (stresses and strains at criti- change in dimension due to mechanical pavement
cal locations in the pavement) depends upon the response. The collected data were processed and
assumptions used for the analysis (Chen & Scul- analyzed taking into account the local conditions.
lion 2008).
On the other hand, a common way of measuring
stresses and strains is to instrument the pavement 2 PREVIOUS RESEARCH
with gauges (Losa & Di Natale 2014). However, it
is important to keep in mind that strain gauges pro- The outcome and the major conclusions of previ-
vide an indicative value of deformation for a specific ous researches using FOS for strain measurements
loading with an order of magnitude. Deformation at the aforementioned test site are the following.

653
 In Loizos et al. (2012), the results of strain pavements (Papavasiliou et al. 2010). Building
response analysis (calculated tensile strains at the upon this knowledge, it would be beneficial to
bottom of the FA layer), as well as the measured investigate the in-situ strains induced in the body
values from the FOS instrumentation system indi- of such pavements.
cate no in-situ stress-dependent behaviour of the Prior to the implementation of the CIR,
FA recycled material. foamed asphalt mix designs had been undertaken
In Loizos et al. (2013), the measured strains on several different blends of material recovered
with FOS were verified through FWD loading. A from the test pits. These blends were treated with
Finite Element (FE) linear analysis was performed foamed bitumen using the appropriate laboratory
for the calculation of strains based on the in-situ unit and several briquettes were manufactured for
back-calculated layer moduli originating from the testing purposes to determine the Indirect Ten-
FWD testing. The recorded and post-processed sile Strength (ITS), the Unconfined Compressive
fiber-optic strain measurements under FWD load- Strength (UCS), the cohesion (c) and the angle of
ing were lower than the related calculated strains. internal friction (Φ), as well as the determination
However, these deviations can be quantified and of the Indirect Tensile Stiffness Modulus (ITSM).
be considered as constant. In general, the aim of the mix design is to establish
In Loizos et al. (2014), the developed strains the application rates for foamed asphalt and active
under both FWD loading and test-truck loading filler (cement), to achieve optimal strengths and to
were measured using FOS and compared with the determine the strength characteristics of the mix.
calculated ones from FE linear elastic analysis. The According to the mix design, 3% foamed asphalt
maximum measured tensile strains resulted from (from 80/100 Pen grade bitumen) and 1% ordinary
FWD loading were similar to the values gener- Portland cement was used for the composition of
ated by traffic (test-truck) loading at the bottom the FA mix. The decision to introduce 1% cement
of the FA recycled layer. Consequently, at least was based on the improvement in the achieved
for this depth in the pavement, the FWD can be soaked strengths. Since milling machines produce
used for tensile strains evaluation of FA stabilized few fines, it was decided to introduce 30% (by vol-
pavement materials without a significant lack of ume) natural fine sandy material to blend with the
accuracy. Furthermore, the measured strains were recovered material. It is noted, that the remaining
lower than the relative calculated strains, indicat- layer after the CIR implementation where the FA
ing that the FE analysis overestimates the tensile layer is based, defined as “soil support” of the
strains, especially at the bottom of the FA layer. recycled layer, was initially consisting of Cement
Bound Material (CBM).

3 EXPERIMENTATION 3.2 Fiber optic system principals and installation

The instrumentation used in the present study for
3.1 Test site and material properties
the measurements indicating the strains within the
A test site was investigated that has been Cold pavement structure is a Fiber Optic Sensors (FOS)
In-depth Recycled (CIR) utilizing Foamed Asphalt system based on the Fabry-Pérot (FP) sensor type
(FA) as a stabilization treatment for base works. (Doré et al. 2007, Deng et al. 2007). The FP sensor
Foamed asphalt can be used as a stabilizing and consists of two semi-reflective mirrors facing each
recycling agent with a variety of materials ranging other. The mirrors are placed on the tips of multi-
from good quality crushed stone, marginal gravels mode optical fibers, which are spot-fused into a cap-
with relatively high plasticity to reclaimed asphalt illary. The air gap between the mirrors defines the
mixture, milled from the existing road. It has been FP cavity (see Fig. 1); the distance separating the
recognized that the FA stabilization treatment in fuse spots is the gauge length. Any strain variation
full—and partial-depth of pavements during cold will change the length of the FP cavity, therefore
in-place recycling is an innovative, economical and inducing optical signals. Given its small size and
sustainable method for rehabilitation of heavily- lightweight, a FOS can be easily surface mounted or
damaged pavements (Jones et al. 2009). embedded into the matrix material of a composite
CIR using foamed asphalt has multiple advan- element or structure, including pavements. Discrete
tages that have been thoroughly reviewed (Lee sensors can be placed at critical locations inside the
et al. 2002). The majority of research related to the pavement structure. To measure strains, a white-
specific technique, until recently, has concentrated light is sent into one end of the fiber optic cable and
on material characterization performed in the the reflected signal is received by a readout unit.
laboratory (Chen et al. 2006, Jenkins et al. 2007). The gauge was retrofitted into the pavement’s
More recent research has focused on assessing the bound layers via a small core hole in order to
field performance of foamed asphalt stabilized minimize disturbances to the pavement layers

654
 Transducer gauge lenglh L• The segments of the proof body carrying the
strains sensors must have small size sections in
order to minimize the perturbation of the mechan-
ical response of the bound material (Doré et al.
2007). The proof body and consequently the sen-
sor were instrumented in the laboratory at the
critical fatigue locations of the cores previously
drilled from the test sites. The gauge is installed
in the pavement by placing it and gluing it at the
Sensing interferometer pa111 desired location in the core hole. The diameter of
length difference ll, = 1 L,.,. Cavity length L ,.,.of the Fabry-
Perot sensing interferometer
the cores was less than 10 cm, in order to mini-
mize any disturbances to the pavement layers being
Figure 1. Schematic of the Fabry-Perot based fiber instrumented.
optic strain sensor. After drilling a core hole and preparation of the
pavement for the installation of the FOS system,
i.e. cleaning and drying the hole, the gauges, i.e.
the cores with the sensors previously instrumented
in the laboratory (see Fig. 3) were installed in the
pavement by correctly positioning them and uti-
lizing epoxy (Fig. 4) to secure them at the desired
location within the core hole.

strain sensors

Figure 2. Schematic illustration of the proof body. Figure 3. Cores with FOS.

being instrumented. Fiber optic strain sensors

were imbedded in a polymeric proof body. The
proof body was comprised of a polymeric plastic
selected to have mechanical and thermal proper-
ties similar to the surrounding material in order to
minimize disturbances to the stress fields around
the gauge when the pavement is subjected to load-
ing by heavy vehicles. The proof body consisted of
a material, which is homogeneous and isotropic.
This material was sufficiently robust to protect the
sensor and has an elastic modulus close but lower
than the bound material of the pavement layer.
The shape of the proof body is such that it allows
direct anchorage on the walls of the borehole. It is
also designed to allow for the installation of two
orthogonal strain sensors allowing for the meas-
urement of longitudinal and transversal strains at Figure 4. Installation of the FOS system utilizing
the bottom of the bound layer (see Fig. 2). epoxy glue.

655
3.3 Field data collection
The FOS system was used for horizontal (tensile)
strain measurements at the top and the bottom of
the FA layer (Fig. 5). These locations were selected,
taking into account the results of the strain
response analysis based on back-calculated mod-
uli at characteristic locations within the body of
the recycled layer in a similar pavement structure,
which is further detailed in (Loizos & Papavasiliou
2007). Tensile strains at these locations are consid-
ered to be critical in terms of possible fatigue fail-
ure and directly related to the performance of the
pavement structure.
With respect of Figure 5, the codes of the four
FOS locations are: TB (Transverse at FA-Bottom),
LB (Longitudinal at FA-Bottom), TT (Transverse
at FA-Top) and LT (Longitudinal at FA-Top).
The FOS system was installed at two test points, Figure 6. Test points with FOS.
on the right wheel path of the heavy trafficked (right)
lane of the pavement (see Fig. 6). The codes for the
test sites are “E” (or East) and “W” (or West).
The test points were near each other and posi-
tioned at different distances from the pavement
edge, in order to record as many as possible high ~e (.)
T
strain values, due to the expected lateral wander of I measured strain
the wheels. roc
After successfully installing the FOS system, an ·u;
C)

FWD loading was applied at five different load (/)

0
levels, ranging from 40 to 75 kN. The center of the LL
FWD plate (R = 0.15 m) was located at the z-axis
of the FOS system (Fig. 5). After that, FOS signals
(strains) were recorded, induced from real heavy Time
traffic loading for a time period of 30 min. The
sampling rate was 500 Hz (measure period: 0.002 Figure 7. Example of a FOS signal for strain
sec), i.e. 500 measurements per sec. The FOS signal measurement.

TB LB
10 10
R,P Traffic flow

=
Asphalt Concrete (AC) overlay
§
{h1, E1=EAc, vi} I
g
(FA-Top: Longitudinal- Transverse) ·an
C/J
0
""
FA recycled layer -5 -5
{hz, !:,=ErA, v,j
Time [msec] Time [msec]

(FA-Bottom: Longitudinal - Transverse) Figure 8. Example of FOS signal (2-axle HGV,

FA-bottom).
Remaining layer
{h,, E,, v, j
recordings were properly corrected, in order to
remove “noise”, as shown in Figure 7.
Characteristic examples of the (corrected) sig-
Figure 5. FOS location and pavement cross-section. nals from the FOS system in the test location under

656
real traffic loading (HGV: heavy goods vehicle) are 15–17oC. In-depth temperature measurements
illustrated in Figures 8–11. were conducted with thermometer into drilling
The temperature in the body of the AC over- holes and using a single drop of glycerol. It is
lay was 16–18oC, while for the recycled base it was worthwhile to mention that during the period of
field data collection, the test site has been opened
to traffic for seven years approximately.
TT LT
10 10
4 ANALYSIS RESULTS
';;;'
4.1 Measured tensile strains

·n
§"
], Figure 12 illustrates the measured tensile strains at
-;;;
-5 -5
the bottom of the FA layer (maximum FOS signal
"
00
·;;; from TB and LB sensors) versus the FWD load, as
f/J
0 -10 -10 well as the relation between the measured tensile
~
strains and the FWD load.
-15 -15
0 0 0
The values of the correlation coefficient (R2) are
0
=:
0
~
0
g =: ~ g high, indicating a strong correlation. This indicates
Time [msec] Time [msec] that there is a nearly linear relationship between the
measured tensile strains and the applied FWD load.
Figure 9. Example of FOS signal (2-axle HGV, FA-top). Considering that the maximum measured tensile
strains resulted from FWD loading were similar to
the values generated by traffic (test-truck) load-
ing at the bottom of the FA recycled layer (Loizos
TB LB
10 10
et al. 2014), the aforementioned results were used
for estimation of the strain induced under a 40 kN
';;;'
wheel load (ESAL) and a 65 KN wheel load (maxi-
"2 mum allowable HGV load in Greece).
.>! The maximum recorded tensile strain values are
s-;;; shown in Table 1. It can be seen, that the maximum
"
bl)
·;;;
values were recorded from sensors TB and LB,
VI indicating that the critical against fatigue cracking
0
~ location of the recycled pavement is located at the
-5 -5
bottom of the recycled layer.
0 0
0
0
0
0
0
'<>
0
~
0
0
0
0
0
0
0
~
The measured tensile strains at the bottom of
"' " "' " 'C
the FA layer (i.e. maximum FOS signal values of
Time [msec] Time [msec]

Figure 10. Example of FOS signal (5-axle HGV,

FA-bottom). t:. West o East
80
p'
TT LT y= 5.9657x _o·
10 10 60 R2 = 0.9909 _o
';;;'
~
-o
.6
y = 6.6266x
"u2 0:: 40 6
0 ..2 R2 = 0.9921
], Ci
-;;;
-5 -5 ~ 20
"
bl)
·;;; ~
f/J
0-10 -10
~

0
-1 5 -15
0 0
~
0
0
0
0
0
0
0 0 0 0 0 5 10 15
~ ~
0 0
'<>
" "" " '<>

Time [msec] Time [msec] FOS signal [microns]

Figure 11. Example of FOS signal (5-axle HGV, Figure 12. Measured tensile strains (FA-bottom) vs.
FA-top). FWD load.

657
Table 1. Maximum measured tensile strains. ---fr- W (LT) ---<>-- W (IT)
10
FOS signal [microns]

Test point TB LB LT TT c
-~
West 12.1 12.1 5.1 3.3
East 10.4 12.5 10.2
~ 0
-~
Vl
0 -5
u..

---fr- West ---- 40kN strain - ·- · 65 kN strain -10

14

-1 5
12
Time

-T
10 -·-·-·-·-·-
'"§
c
Figure 15. Measured tensile strains (West, FA-top).
I-;;; ------
c
·~ ---fr- E (LT) ---<>-- ECIT)
V)
0 4 10

t~ ~r
u..

=
1
0 §
Time I 0
],
·o; ~~
Figure 13. Measured tensile strains (West, FA-bottom). </}
0
u..
-5 ,:: f> & ~ ~
?

J\/~
1> ~~ '%'

14
--0 -- East ---- 40 kN strain - · - · 65 kN strain
-10

-1 5
ii 0
1> 6

mr~ ~~
6
1
6
i
6

12
_______ L J Time

10 "' o'% ':' 6. ~ 0

f!P,! ¢<:!!!, :\ : i: Gb Figure 16. Measured tensile strains (East, FA-top).
§" ~tidf ' "' ~i 0
i!?: 'i' ; f i! <:{)
I
], ---r-t---:-t-~---:r-~------~-~-1"------- It can be seen that at the top of the FA layer,
·o;
V)
0 4 , 6 ~ i6 ~ only the sensor with code “E (LT)”, i.e. longitudi-
u.. <lD <'& nal at East test point, recorded strain values with
2 magnitude similar to the ones measured at the bot-
tom of the FA layer. The other three sensors at the
0
top of the FA layer recorded low tensile strains and
Time
in most cases negative values (compression).
Figure 14. Measured tensile strains (East, FA-bottom).
4.2 Bearing capacity of the recycled pavement
The measured tensile strain values at the bottom
TB and LB sensors), as well as the level of the of the FA layer using FOS were very low, indicat-
strains induced under a 40 kN and 65 kN HGV ing adequate bearing capacity of the recycled pave-
load, are presented graphically in Figures 13, 14 ment. However, these values may be questionable,
for the test points “West” and “East” respectively. considering that according to Loizos et al. (2013),
It can be seen, that in some cases the measured the recorded fiber-optic strain measurements
strains were higher than the estimated under a under FWD loading were lower than the related
65 kN load, indicating overloaded HGVs. calculated strains. Moreover, the measurements
The measured tensile strains at the top of the FA were conducted at mild FA layer temperatures
layer are presented graphically in Figures 15, 16 for (15–17°C). At higher temperatures, larger strain
the test points “West” and “East” respectively. values are expected.

658
 Following the above considerations, the out- experimental work is needed. The latest include
come of the research of Loizos et al. (2013) con- strain measurements under several FWD loads
ducted at the same test point was utilized, aiming and comparison with the calculated ones, as well as
to estimate the calculated tensile strains under the strain measurements under test-truck loading and
HGV traffic loading. The measured tensile strains comparison with the related strains under FWD
at the bottom of the FA layer (i.e. FOS signal val- loading. The above research should be conducted
ues at mild temperatures) were converted to the repeatedly at different pavement temperatures.
calculated ones at mild temperature using multipli-
ers ranged from 2.76 to 4.58.
In order to estimate the tensile strains at medium REFERENCES
temperature (22–23°C), the calculated tensile
strains under the HGV traffic loading (at mild Bueche, N., Rychen, P., Dumont, A.G. 2009. Optical
temperatures) were multiplied with values ranged fiber feasibility study in accelerated pavement testing
from 1.48 to 2.43. Subsequently, the resulted values facility. Proc. 6th Intern. Conf. on Maintenance and
(i.e. estimated measured tensile strains at medium Rehabilitation of Pavement and Technological Control
temperatures) were converted to the calculated (MAIREPAV). Torino.
ones at medium temperature using multipliers Chen, D. & Scullion, T. 2008. Forensic investigations of
roadway pavement failures. J. Perform. Constr. Facil.
ranged from 2.04 to 2.35. 22(1): 35–44.
The maximum estimated (calculated) ten- Chen, D.H., Bilyeu, J., Scullion, T., Nazarian, S. & Chiu,
sile strains under the HGV traffic loading were C.T. 2006. Failure investigation of a foamed-asphalt
57 microns and 64 microns at mild and medium highway project, J. Infrastruct. Syst. 12(1): 33–40.
temperatures respectively. These values are low, Deng, L. & Cai, C.S. 2007. Applications of fiber optic
considering that the calculated theoretical tensile sensors in civil engineering. Structural Engineering
strain using a 65 kN load and the material proper- and Mechanics. Vol. 25, No. 5: 577–596.
ties of the pavement layers according to the pave- Doré, G., Pierre, P. & Duplain, G. 2007. Monitoring
ment design was 120 microns. Consequently, it mechanical response of in service pavements using
retrofitted fibre optic sensors, Proc. Intern. Conf. on
appears that the FOS measurements prove that the Advanced Characterization of Pavement and Soil Engi-
recycled pavement has better performance, than neering Materials. 20–22 June 2007: 883–891. Athens.
could be expected based on the theoretical strains. Jenkins, K.J., Long, F.M. & Ebels, L.J. 2007. Foamed
bitumen mixes = shear performance? International
Journal of Pavement Engineering. 8(2): 85–98.
5 CONCLUSIONS Jones, D., Fu, P. & Harvey, J.T. 2009. Full-depth pave-
ment reclamation with foamed asphalt in California:
The present research aims to investigate the con- Guidelines for project selection, design, and construc-
tribution of the Fiber Optic Sensors (FOS) system tion. UCPRC-GL-2008-01. University of California,
Pavement Research Center, Davis and Berkeley. CA.
towards the assessment of the bearing capacity of Kim, Y.R. & Park, H. 2002. Use of falling weight deflec-
asphalt pavements that have been Cold In-depth tometer multi-load data for pavement strength estima-
Recycled (CIR) utilizing Foamed Asphalt (FA) as tion. Final Report, Research Project No HWY-00-4,
a stabilization treatment for base works. On this Department of Civil Engineering, North Carolina
purpose, a testing experiment was conducted to State University Raleigh, NC.
record the in-situ developed strains under real traf- Lee, K.W., Brayton, T. & Huston, M. 2002. Development
fic loading and FWD load as well. of performance based mix design for Cold In-Place
Based on the analysis results, it can be concluded Recycling (CIR) of bituminous pavements based on
that critical against fatigue cracking location of the fundamental properties. FHWA-IF-05-014, Washing-
ton D.C.
recycled pavement is located at the bottom of the Loizos, A., Papavasiliou, V. & Plati, C. 2012. Investigat-
(FA) layer. ing In-situ Stress-dependent Behaviour of Foamed
It can also be concluded, that the recycled Asphalt Treated Pavement Materials. Road Materi-
pavement has better performance, than could be als and Pavements Design. Vol. 13, Issue 4, December
expected, based on the theoretical (calculated) 2012: 678–690.
strains using the material data taken into account Loizos, A., Plati, C. & Papavasiliou, V. 2013. Fiber Optic
during the pavement design. Sensors for Assessing Strains in Cold in-place Recy-
Although the installation of an FOS system is cled Pavements. International Journal of Pavement
a time consuming and delicate procedure, it has Engineering. Vol. 14, Issue 2, February 2013: 125–133,
Taylor & Francis. UK.
proved to be a useful and promising tool for in-situ Loizos, A., Papavasiliou, V. & Plati, C. 2014. Effective-
strain measurements under real traffic loading. ness of FWD to Simulate Traffic Loading in Recycled
However, due to the deviations between calculated Pavements. Journal of Performance of Constructed
strains (pavement design, FWD data analysis) Facilities, 10.1061/(ASCE)CF.1943-5509.0000708,
and FOS signal records (measured strains) more 04014193.

659
Loizos, A. & Papavasiliou, V. 2007. In situ characteri- Papavasiliou, V., Loizos, A. & Plati, C. 2010. Monitor-
zation of pavement materials stabilized with foamed ing the Performance of Cold In-depth Recycled Pave-
asphalt and cement. Proc. Intern. Conf. on Advanced ments. Proc. 11th Intern. Conf. on Asphalt Pavements.
Characterization of Pavement and Soil Engineering ISAP. 1–6 August 2010. Nagoya.
Materials. 20–22 June 2007: 875–882. Athens. Rens, K., Wipf, T. & Klaiber, F. 1997. Review of nonde-
Losa, M. & Di Natale, A. 2014. Accuracy in predicting structive evaluation techniques of civil infrastructure.
viscoelastic response of instrumented asphalt pave- J. Perform. Constr. Facil. 11(4): 152–160.
ments. Proc. 3rd Intern. Conf. on Transportation Infra-
structures (ICTI): 339–346. Pisa.

660
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Subgrade cracking monitoring using distributed optical fiber

sensing technique

W. Liu, B. Wang, X. Chen & L. Li

Department of Transportation and Logistics, Dalian University of Technology, Dalian, Liaoning, P.R. China

ABSTRACT: In this paper, a new type of distributed optical fiber based sensor with flexible encapsula-
tion developed by our research group has been embedded in a real highway structure with the total meas-
uring range of nearly 300 meters to monitor the subgrade performance. The embedded sensors work well
after the normal construction of upper pavement layers. The sensor data have been collected twice with
the interval longer than a month in winter. Analysis results show that the proposed sensor can successfully
monitor temperature induced shrinkage, cracking and frost heaving of the subgrade. The research works
presented in this paper have proved the potential of the proposed sensor for further large scale application
in different layers to study the complex effects of the environmental factors on pavement structure.

1 INTRODUCTION niques have been successfully used for large scale

structure health monitoring recently (Gu et al.
Subgrade is the native material underneath a con- 2009; Rodriguez et al. 2015).
structed pavement and it is a very important part In this paper, the proposed sensor has been
of a flexible pavement structure (Solanki & Zaman embedded in a real highway structure with the
2014; Ji et al. 2015). It supports the upper base total measuring range nearly 300 meters. Detailed
and surface layers and serious pavement structural sensing system design and embedding procedure
damage is often caused by the bad condition of the will be introduced in the following sections. The
subgrade layer. The frost heaving and thaw settle- structural monitoring data collected at different
ment of the subgrade is common disease in sea- time are discussed.
sonally frozen region. Cracking could happen after
the frozen of the subgrade. It may reflect up to the
base and surface layers of the pavement structure 2 SENSING SYSTEM DESIGN
and may aggravate the influence of water damage.
It is often difficult to monitor the subgrade crack- General information of the supporting construction
ing of an in service highway structure due to its As part of the national road network plan (NDRCC
depth and large size. Embedding sensors can provide 2013), the construction of He-da highway in Jilin
real time structural performance data with relatively province of China was started in Apr. 2014 and will
high accuracy (Sargand 2002). The harsh working be completed at the end of 2016. Jilin province is
environment and unique characteristics of pavement located in the northeast part of China. The lowest
structure, especially asphalt pavement structure, temperature can be −40°C in winter and low tem-
require the sensors to endure high temperature (up perature cracking is one of the major premature
to 160°C), moisture, high compaction force, repeated damages of highway pavement structures. The con-
heavy loading, etc. and have large coverage. Most of struction of He-da highway has been selected as a
the traditional sensors for other civil structures can- technology demonstration project, which is also the
not be used for pavement structure directly. first one in seasonal frozen area in China.
The distributed Brillouin Optical Time Domain The proposed distributed optical fiber based
Analysis/Reflectometry (BOTDA/R) technique sensor has been selected to be embedded in the sec-
can realize the simultaneous measurement of hun- tion of He-da highway near the city of Dunhua. A
dreds of thousand points, other than the common highway administrative office is close to the sen-
advantages of optical fiber sensors (Shimizu et al. sor embedding area, which will help for protecting
1993; Dong et al. 2012). Through many years of sensing system and collecting the monitoring data
development, the measuring range and spatial res- after the opening of the highway. The subgrade at
olution of BOTDA/R technique have been notably that location is in the fill section with the filling
improved. Distributed optical fiber sensing tech- depth of nearly 4 m.

661
3 SENSING SYSTEM DESIGN AND The proposed sensors were embedded on top of
EMBEDDING the subgrade layer. The highway at the monitor-
ing location has two lanes in each direction. The
A distributed optical fiber based sensor (Fig. 1) width of the subgrade layer is nearly 30 m. The
with flexible encapsulation for pavement behavior upper pavement structure for each direction has
monitoring developed by our research group (Liu been constructed separately. The construction of
et al. 2015. Optical fiber based sensors with flexible the median strip structure has been left to the last
encapsulation for pavement behavior monitoring.) and its width is about 1.5 m. 10 sensors have been
has been used for the subgrade cracking monitor- embedded with 4 of them embedded on the cross-
ing in He-da highway. sectional direction and 6 of them embedded on the
For the application in He-da highway, the gauge longitudinal direction (Fig. 2). The longitudinal
length of the sensor is selected as 1.2 m and the sensors were only embedded on one side of the
total length of each sensor is about 30 m. The spa- highway at the location near the wheel path.
tial resolution of the analyzer used for the data Before the sensor embedding, grooves with the
collection in this paper is 0.2 m. The gauge length depth about 3 cm have been digged on top of the
of the sensor is determined by the distance of the subgrade layer (Fig. 3). The sensors were laid in
Fiber Reinforce Plastic (FRP) annuluses setting the grooves and then covered with fine sand. The
on the optical fiber. The gauge length of 1.2 m
can ensure that at least 3 data points within each
gauge length will not be influenced by the strain
concentration caused by the movement of the FRP
annuluses on the sensor. Besides, the maximum
strain that a normal optical fiber can bear is about
10,000 με and the gauge length of 1.2 m makes the
sensor can bear the maximum cracking width up
to 1.2 cm within a gauge length.
Short rubber tube

FRP

Figure 1. Structure of distributed optical fiber based

sensor with flexible encapsulation. Figure 3. Sensor embedding.

Figure 2. Sensing system design for He-da highway substructure.

662
construction procedure of the cement stabilized 4.2.1 Sensor C1
macadam layer on top has not been altered. Dur- For all the sensors, the 0 m position on the hori-
ing the construction of the cement stabilized mac- zontal axis is near the first annulus of the sensor,
adam layer, sensors C4 and L3 were cut by the which is also the start of the measuring section of
workers unintentionally. Other sensors are all work the optical fiber. Negative value on the horizontal
well after twice data collections. axis means the patch cord area. The first annulus
for the cross sectional sensors is located at bottom
in Fig. 2, and that for the longitudinal sensors is
4 DATA COLLECTION AND ANALYSIS located on the right side in Fig. 2 except for the
sensor of L6.
4.1 Data collection Since the strain resolution of the analyzer used
is 50 με, data variance under 50 με on the chart
The construction of the ATB-30 layer at the moni-
can be treated as stable. The strain along the sensor
toring location has been finished in Oct. 2015, and
C1 for the first data collection doesn’t have large
upper layers will be constructed in Apr. 2016. Data
fluctuation before the length of 18.07 m in Fig. 5.
of the embedded sensors have been collected twice
The variation after the length of 18.07 m may be
in Nov. 7∼8 and Dec. 30∼31 of 2015, respectively.
caused by the discriminating length limitation of
The lowest temperatures during the twice data col-
the first analyzer. Between the length of 12.11 m
lections were around −10°C and −30°C, respectively.
and 16.01 m, the points with relatively larger strain
have appeared. The strain variation is still below
4.2 Data analysis 250 με. Around the same location, a strain peak
has appeared in the second data collection, with
The environmental temperature was different dur-
the maximum strain variation of more than 800 με.
ing each data collection and dozens of degrees of
The environmental temperature during the sec-
temperature variation will result in several hun-
ond data collection was much lower than that dur-
dreds to more than a thousand micro strain differ-
ing the first data collection. This larger strain peak
ences. Besides, different analyzers were used for the
at the location between 12.11 m and 16.01 m should
twice data collections of this paper and the start
be caused by the compaction of the subgrade dur-
parameter settings were also different. For the
ing the drop of temperature induced. The maxi-
comparison of the sensor strain change between
mum strain appeared at the location about 13.96 m,
the twice data collections, relative strain is used.
which is within the 12th gauge length of the sen-
That is, the first collected data have been shifted
sor. The area of the strain peak between 12.11 m
certain strain to fit with the second collected data.
and 16.01 m on the chart (integration of the strain
The data comparison for each sensor has been pre-
increment at each point multiply by the spatial
sented in the following subsections.
resolution) can be calculated as about 2 mm, which

15000
14800
14600
14400
(..) 14200
::i.
E 14ooo - 2
....~ 13800
(f)
- 1
13600
13400
13200
13000
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
Location of t he measuring point/m

Figure 4. Data comparison for sensor C1.

663
means that within the strain peak location, the total broken at the length of 24 m during the second
movement of the subgrade is about 2 mm, or most data collection. The length of 24 m is just the
probably there is a crack with the width of nearly position of an annulus for the 20th gauge length.
2 mm within the 12th gauge length. The location of From Fig. 7 it can be seen that after the length of
the crack is near the median strip area. 18 m, the second collected data fluctuate sharply.
If only cracking happens in this area, finite strain
4.2.2 Sensor C2 peaks as that of sensor C1 should be found from
As shown in Fig. 5, the data variation is the larg- the data instead of sharp oscillations. The oscil-
est among all the sensors that have data. It is lations may be caused by the frost heaving of the
hard to find a part along the sensor that has a subgrade layer and the movement of aggregates
stable strain field from the twice data collections. at the surface of the subgrade has bent the sen-
Between the length of 1 m and 11 m, the strain sor seriously. From Fig. 6 it can be seen that the
has dropped with about 2000 με in maximum for width of frost heaving area is at least 6 m (from
the second data collection. This phenomenon can 18 m to 24 m). For the twice data collections, the
be explained as large area shrinkage among the trends of the whole curves are similar. Between
length of 1 m to 11 m. Between the length of 14 m the length of 2 m to 5 m, which is near the edge
and 17 m, the strain has increased with about of the pavement structure, tension strain with
300 με in maximum. It is the area near the median the maximum of 600 με increase can be found.
strip. The subgrade at the median strip was still
exposed during the second data collection (not 4.2.4 Sensor L1
covered by other pavement layers). Cracking is A strain protuberance with the maximum strain
easier to appear on the subgrade surface at the change of 700 με can be found from the first data
median strip due to the larger temperature sen- collection of Fig. 7 between the length of 15 m
sitivity and smaller resistance force given by the and 17 m. For the second data collection, the peak
upper pavement layers. The subgrade shrinkage of the strain protuberance has increased to about
from both sides of the median strip may cause 3000 με. It indicates that a crack may appear at
the tension strain near the median strip area. The that location and the width of the crack is increas-
data seems to have a trend to be consistent after ing with the drop of the environmental tempera-
the length of 18 m. ture. The peak area for the first data collection
can be calculated as 0.4 mm between the length of
4.2.3 Sensor C3 15.4 m and 16.8 m. The peak area for the second
The sensor of C3 worked well during the first data collection has increased to 3.9 mm between
data collection. However, it has been found to be the length of 13.8 m and 16.4 m. The cracking

15200
15000
14800
14600
14400
~ 14200
~ 14000
--2
....~
en 13800
--1
13600
13400
13200
13000
12800
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Locatio n of the m easuring point/m

Figure 5. Data comparison for sensor C2.

664
 14800
14600
14400 [
14200
14000
1 13800
E' 13600
--2
~ 13400
(/) +
--1
13200
13000
12800
12600
12400

Location of t he m easuring point/m

Figure 6. Data comparison for sensor C3.

17200
16800
16400
16000
15600
(,)

~ 15200
c
·~ 14800 --2
....
(/)
--1

14000
13600
13200
12800 +--,-,,-,--,--,--,-,--,--,--,--,-,--,--,--,-,--.--.-
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Locat ion of t he measuring point/m

Figure 7. Data comparison for sensor L1.

influence length has increased from about one cracking or other subgrade movements have been
gauge length to two gauge lengths. found.

4.2.5 Sensor L2 4.2.6 Sensor L4

For the sensor of L2, the measuring area has For sensor L4 (as shown in Fig. 9), before the length
similar data trends for the twice data collections. of 12 m, the trends of the curves for the twice data
The small value variances from the twice data collections are similar. After the length of 12 m, the
collections were caused by temperature variation second collected data has apparently smaller and
and the analyzer resolution limitation. Based on stable strain. This may be caused by the subgrade
Fig. 8, it can be concluded that no apparent shrinkage induced stress relaxation at that area.

665
 14400

14200

14000

.2:- 13800
.!:
- 2
........"'
Ul 13600 - 1

13400

13200

13000 +-~-.-.--.-.-~-.-.--.-.-~-.-.--.-.-~-.-.--,-,
-3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Location of the measuring point/m

Figure 8. Data comparison for sensor L2.

14650

14450

14250

1 14050
......
.!:
- 2
~ 13850
- 1

13650

13450

13250 +-.-.-.-.-~~~-.-.-.-.-,-.-.-.-.-..-.-.-.-.-.-.
-3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Location of the measuring point/m

Figure 9. Data comparison for sensor L4.

4.2.7 Sensor L5
For sensor L5 (as shown in Fig. 10), the twice The strain increment (about 300 με) for the twice
data collections also have similar trends. Near the data collection at the peak area is much smaller
length of 7 m, a strain peak can be found and the comparing to other detected cracking for the sen-
peak is increased for the second data collection. sors such as C1 and L1.

666
 15000

14800

14600

"'::i. 14400
......
" - 2
~ 14200
<n
- 1
14000

13800

13600 +-,--,-,-,--,-,-,--,-,-,--,-,--,-,-,--,-,-,--,-,
-2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Location of the measuring point/m

Figure 10. Data comparison for sensor L5.

be applied to the pavement layers other than the

5 CONCLUSIONS AND DISCUSSIONS subgrade layer and can be used to monitor the syn-
ergistic effects of different layers which cannot be
This paper introduces the application of a distrib- realized by other sensors on the market.
uted optical fiber based sensor with flexible encap-
sulation on a real highway structure for the first
time. The total measured length is designed over REFERENCES
300 m. The conclusions from the works presented
in this paper can be summarized as below. Dong Y., Zhang, H., Chen, L., and Bao, X. (2012). 2-cm-
• The proposed sensor has high survival rate dur- spatial-resolution and 2-km-rangeBrillouin optical
ing real highway construction. The two broken fiber sensor using a transient differential pulse pair.
Appl. Opt., 51, 1229–1235.
sensors were all damaged by the careless of the Gu, G.P., Revie, W., Zou, L., and Sezerman, O. (2009).
workers using metal tools, which were not the Pipeline monitoring by Brillouin-scattering-based
normal construction procedures. The damages fibre optic distributed strain sensors: pipeline wall
can be avoided by educating the workers before thickness detection. Proc. of SPIE, 7503, 75036O:1–4.
construction or better construction management. Ji, R., Nantung, T., Siddiki, N., Liao, T., and Kim, D.
• The proposed sensor can monitor temperature (2015). Field and Laboratory Determination of Sub-
induced shrinkage effect of the subgrade layer as grade Resilient Modulus and its Application in Pave-
the data shown for the sensor C2. The proposed ment Design. Journal of Testing and Evaluation,
sensor can monitor the cracking level of the 43(5), 1109–1119.
Liu, W.Q., Wang, H.P., Zhou, Z., Xing, X.Y., Cao, D.D.,
subgrade layer as the data shown for the sensors and Jiang, Z. (2015). Optical fiber based sensors with
C1 and L1. The proposed sensor may be able to flexible encapsulation for pavement behavior monitor-
detect the frost heaving of the subgrade layer as ing. Structural Control and Health Monitoring, 22,
the data shown from the sensor C3. 301–313.
• Based on the data from the embedded sensors, National Development and Reform Commission of
it can be seen that the temperature induced sub- China (NDRCC). (2013). Plan for national road net-
grade movements are more obvious in the cross work (2013–2030).
sectional direction than in the longitudinal direc- Rodriguez, G., Cacas, J.R., and Villaba, S. (2015). Crack-
tion. The second collected data of the cross sec- ing assessment in concrete structures by distributed
optical fiber. Smart Mater. Struct., 24, 035005.
tional direction sensors C1, C2 and C3 all show Sargand, S. (2002). Continued Monitoring of Pavement
clear changes from the first collected data. The in Ohio, FHWA/OH-2002/035.
longitudinal sensor L2 almost show no change Shimizu, K., Horiguchi, T., Koyamada, Y., and
and the changes for sensor L5 are also small com- Kurashima, T. (1993). Coherent self-heterodyne detec-
paring to the cross sectional direction sensors. tion of spontaneously Brillouin-scattered light waves
in a single-mode fiber. Opt. Lett., 18, 185–187.
The works presented in this paper provide good Solanki, P., and Zaman, M. (2014). Behavior of Stabi-
examples in applying the distributed optical fiber lized Subgrade Soils under Indirect Tension and Flex-
based sensors for pavement structural large scale ure. Journal of Materials in Civil Engineering, 26(5),
performance monitoring. The proposed sensor can 833–844.

667
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Ensuring reliable robust FWD measurements on the English

road network

S. Brittain, A. Wright & D. Gershkoff

TRL, Crowthorne, Berkshire, UK

R. Fairclough
Highways England, Manchester, UK

ABSTRACT: The UK procedure for assessing the structural condition of a road pavement includes the
collection and analysis of deflection response to an impact load using a Falling Weight Deflectometer
(FWD). There is a need to ensure that these devices are accurate and consistent. Initially in 1998, on behalf
of Highways England, TRL introduced correlation trials to compare the levels measured by the devices
on a range of test pavements under closely controlled conditions. More recently TRL have developed,
with the involvement of the UK and Ireland FWD Group, a rigorous procedure which includes annual
accreditation trials together with checking routines for both the contractor and the Auditor(appointed
by Highways England) to apply. Within the standard for designing and maintaining the English strategic
road network it is a requirement that survey data is collected using equipment that adheres to this proce-
dure. This paper describes this procedure.

1 INTRODUCTION 1.2 Falling Weight Deflectometers and

applicable devices
1.1 Background
FWD equipment can either be hand held machines,
An important element in the UK procedure for devices built into existing vehicles, or devices
assessing the structural condition of a road pave- mounted on a trailer towed behind a vehicle. Each
ment is the collection and analysis of deflection device incorporates a circular loading plate that
response to an impact load using a Falling Weight is lowered onto the pavement surface when test-
Deflectometer (FWD). ing. The device lifts a known load to a predeter-
The first FWD was operated on the UK network mined height and drops it onto the plate, creating
in 1983. Since then its usage and application has an impact load. A set of buffers or a system of
grown and there are now over twenty machines in springs, located between the plate and the falling
operation in the UK. Many of these are used on the weight, spread the impact and produce a particu-
Highways England strategic road network to help lar load pulse shape from the impact. The pulse
design maintenance requirements. As the usage grew shape is designed to broadly represent the pulse
it was soon realised that there was an urgent need to produced by a loaded rolling wheel at traffic speed.
ensure that the measurements from these devices and FWD are generally classified by the load that
their interpretation were accurate and consistent. can be applied to the test plate. Procedures have
Initially TRL, on behalf of Highways England, been developed to test devices which can apply
introduced correlation trials to compare the lev- loads of between 30kN and 120kN. This includes
els measured by the different devices on a range heavy weight and super heavy weight devices which
of test pavements under closely controlled condi- can also apply higher loads.
tions. This has now been developed into a rigorous
quality assurance procedure, which includes annual 1.3 Overall accreditation and QA process
accreditation, contractor’s quality assurance checks
and checks by an independent Auditor. To ensure reliable and robust FWD measure-
It is now a requirement within the standard for ments, two processes are undertaken. The first is
designing and maintaining the English Strategic an annual Accreditation/Re-accreditation trial
Road Network (SRN) that contractors commis- where the devices are assessed against reference
sioned to supply survey data use equipment that data for each of the relevant parameters. This
has been accredited to undertake these surveys. process determines whether the devices measure
The data provided must also be checked using the deflections and other survey parameters within
approved Quality Assurance process. an acceptable range of accuracy and repeatability.
669
However, the test only checks the performance of
the devices on the date of the trial.
Therefore a further ongoing Quality Assurance
programme checks that the device performs con-
sistently between Accreditation trials. This is split
into two parts, the checks undertaken by the survey
contractor (i.e. the company using the device) and
checks undertaken by an independent Auditor.
These Accreditation, Re-accreditation and
Quality Assurance processes are described in the
following sections.

2 ACCREDITATION AND
RE-ACCREDITATION
Figure 1. A test site set up for a FWD accreditation
2.1 Accreditation trials trial.
The purpose of the accreditation process is to test
the devices to check they are producing accurate In August 2011 CROW issued an updated ver-
results. This is achieved by assessing the devices sion of their recommendations. Which introduced
against reference data for each survey parameter. an additional test examining the repeatability of the
However, due to changes in deflection of pave- devices, as a separate test from the SDDR assess-
ments with environmental conditions (e.g. tem- ment. This test has also been included in the UK
perature), it is not a trivial problem to determine trials.
the “true” deflection of a pavement which can
be used as a reference. In the absence of a “true”
deflection value, one would ideally like to compare 2.2 UK implementation of accreditation trials
against a stable reference device but, as yet, such TRL have undertaken the UK accreditation trials
a device has not been developed that can be eco- on an annual basis, on behalf of Highways Eng-
nomically applied over a suitable range of test sec- land, since 1998, during which time they have been
tions. Therefore devices are collected together in an subject to continuous improvement. The trials have
Accreditation trial and the average of the deflection developed into a two day programme. On the first
measurements of the fleet is used as the reference. day the machines are inspected to check that they
The accreditation trials have been developed are properly configured and operating correctly.
from the criteria published by the CROW standards The machines then undertake a preliminary test
organisation in the Netherlands. In turn these were lap. This data is not used in the assessment of the
developed from work within a European COST devices but is instead used to provide early indica-
Action, COST336, as discussed by van Gurp (2005). tions of issues that might be resolved before the
The basis of the trial is the calculation of the Field main testing on the second day. The second day of
Calibration Factor (FCF) and Standard Deviation the trial is used to conduct the test laps which are
of the Deviation Ratio (SDDR) for each device and assessed using the FCF and SDDR methodology.
geophone. The FCF is a measure of the bias of the Following completion of the trial, certificates
device from the rest of the fleet. The SDDR is a showing the performance of the devices against the
measure of the repeatability of the device over the trial criteria are issued to the contractors. These
full range of the test sites. Each of these parameters certificates expire within 13 months of the trial
must lie with specified criteria or the device is consid- date. Therefore, to continue surveying, contractors
ered an outlier and is removed from the assessment. must take part in the next Accreditation trial.
The main difference between the UK and
CROW procedure is that the allowable tolerance
for FCF are tighter in the UK. In addition, in the 2.3 Other survey parameters
CROW procedure the derived FCF are used to In addition to checking the deflection measure-
correct the measured deflections. This approach ments it is important to assess other relevant sur-
is not used in the UK. A related difference is the vey parameters. A number of additional tests have
choice of test sites, which encompass a much wider therefore been included in the trial.
range of structures in the CROW test, particularly
in terms of the range of subgrades, since this is 2.3.1 Temperature measurements
considered of particular importance in their appli- As noted above, pavement deflection varies with
cation in the Netherlands temperature. Therefore during the second day of

670
testing the operators are asked to make tempera- and/or the market identifies that it requires the
ture measurements from pre-drilled holes. These data and the systems become fitted to all devices.
are compared against measurements collected by
the Auditor using automatic loggers. This process
2.5 Experience from implementing the
was initially implemented using sets of holes close
accreditation process
together (one for the logger and one for the contrac-
tor measurement). However, although the operator 2.5.1 Choice of test sections
measurements were fairly consistent between opera- As noted above, the CROW implementation covers
tors, they were often different from the logger meas- a wider range of pavement construction, particu-
urements. Therefore, the temperature measurements larly in terms of the range of subgrades in com-
are now taken using the same holes (widened so that parison to the UK implementation. When the UK
the logger and the contractor’s probes can both fit). FWD accreditation trials were first set up it was
This has resulted in much more consistent results. found that covering such a wide range would be
In recent years temperature sensors have been impractical, and a subset of the typical UK con-
added to some of the devices to automatically collect structions was selected.
air and surface temperatures during testing. These Therefore the UK FWD accreditation trials have
sensors have been fitted to accommodate an antici- been developed on sites with 12 test points spread
pated change to the test method which will allow for over three test sections with differing deflection
the predicted temperature at depth (using air and levels (4 test points per section). It has been found
surface temperatures) to be used instead of requir- that using this approach provides confidence that
ing the direct measurement of temperature at depth. the machines are performing acceptably in the likely
It has been necessary to expand the assessment to ranges of deflections on the Highways England net-
include a check on these sensors. This is implemented work. If this approach is to be undertaken in other
in a similar way to the temperature at depth assess- countries then sites should be selected to provide a
ment, by comparing the device measurements to cor- range of values seen on their respective networks.
responding measurements logged by the Auditor. Between 1998 and April 2010 the UK FWD
Accreditation trials were held on the TRL test
2.3.2 Location referencing track. In November 2010 the trials were moved
Although it is important to the end user of the to the HORIBA-MIRA proving ground. It has
survey data that the deflection measurements are been found that in general the SDDR values for
accurately collected and processed, this informa- the devices appear to be slightly lower on the
tion is of no use if there is uncertainty in where HORIBA-MIRA proving ground. Our experience
the data was collected. Therefore the devices must suggests that it is necessary to review the FCF and
be checked to show that they can accurately record SDDR limits when they are applied on different
the location of the measurements. To do this the sites. Or dynamic criteria should be developed,
survey crews are asked to measure the length of a based on the pavement type and deflection levels.
test site. The length measured is assessed against
the actual length to assess the performance of the 2.5.2 Criteria and use of FCFs
distance measurement of the survey device. A further difference between the CROW and UK
Some devices have GPS systems fitted to provide implementations is the range and use of FCF
3-dimensional spatial coordinates of the test loca- values. A larger range FCF values are acceptable
tions. The horizontal error between these measure- under the CROW implementation than in the UK.
ments and the reference data for the test points (i.e. However, these FCF values are then used during
the straight line distance between the two points) is the survey season to correct them back to the fleet
used to assess this data. mean. In the UK FCF values are not used during
the survey season. This approach has been under-
taken because it is believed that a large part of the
2.4 Mandatory and additional criteria
spread of the FCF values is due to random noise
The majority of the assessments undertaken at the in the data, rather than systematic differences in
trial are “mandatory” assessments i.e. the result is the machines.
either pass or fail, and the device would need to Examination of the FCF values from the UK
meet the criteria to be considered acceptable for use. accreditation trials has found that there does not
However some of the assessments are “addi- appear to be consistency in the FCF values from
tional tests” for which a performance level is given year to year (i.e. a machine which was below the
e.g. High, Medium or Low. These typically include fleet mean in one year may be above the fleet mean
the assessment of systems not fitted to all devices in the next year). Initially this would suggest that
and/or tests which are not as mature as the manda- the variation in FCF values is mostly due to ran-
tory assessments. Over time these tend to migrate dom noise in the data, rather than systematic dif-
to being mandatory criteria as the test matures ferences in the machines. However, it is possible

671
that the tighter tolerances imposed in the UK has ensure the on going performance of the devices
removed the systematic differences between the between the accreditation trials. It is therefore
devices (as any machine which is systematically important to implement a Quality Assurance
different is excluded). Therefore applying the UK (QA) programme to monitor the devices during
approach should in theory result in a more con- the course of the year. We have developed a 2 part
sistent fleet of devices. However, it does mean that QA programme. The first and main part is the QA
devices which might provide consistent results after undertaken by the contractor i.e. the company
the application of a correction factor are excluded. operating the device. The second part is the checks
undertaken by an independent Auditor.
2.5.3 SDDR and repeatability
Although the SDDR and the repeatability are both
3.1 Contractor’s quality assurance
measures of the variability of the device it is use-
ful to undertake both assessments. Over the years The contractor’s QA monitors the performance of
there have been instances where a device fails the the device between Accreditation trials to identify
SDDR criteria but not the repeatability criteria anomalies in the performance of the device. The
and vice versa. Therefore, they assess different contractor provides the results and conclusions to
aspects of the variability. the Auditor. The Auditor will then review the data
In recent years it has been seen that, while remain- and provide feedback on the implementation of
ing within the SDDR criteria, some devices appear to the process and add advice where necessary.
be consistently producing higher SDDR values rela- If it is identified that the device is behaving unex-
tive to other devices in the fleet. It is therefore likely pectedly, then the contractor would investigate the
that tighter SDDR criteria could be implemented at cause and fix the issue if possible (seeking advice
a future date following investigations into why some from the Auditor if required). The Auditor would
devices produce higher SDDR values (and eliminat- be notified of the action taken and would advise
ing the causes). It is possible that similar gains may the contractor (and if necessary the end user of the
be made in the repeatability assessment. However data) if any additional work is required to confirm
long term trends in this data are not yet visible. that the device is performing suitably, and/or if any
data previously collected should be disregarded.
2.5.4 Devices not meeting the mandatory criteria
and differences in the reference fleet 3.1.1 Stack/tower consistency check
It has been found that devices may fail to meet the A stack/tower consistency check (or manufacture’s
criteria on the inspection day, or during the early equivalent procedure) aims to check that all deflec-
laps of the main trial day. Following investigation tion sensors are correctly calibrated relative to each
this is usually a result of minor faults (e.g. applica- other. It is not an absolute calibration check but
tion of correction factors, issues with the buffers should ensure consistency across all deflection sen-
etc.), which can be remedied during the trial. sors. The QA regime recommends that this check is
Occasionally it is not possible to resolve the carried out every 6 months.
issue during the trial and a separate assessment On some devices this is commonly achieved by
must be undertaken. When a single device (or a removing all deflection sensors from the device and
small proportion of the fleet) requires assessment placing them in a collinear holder. The holder is
of its deflection performance it is not practical to placed on the pavement close to the loading plate
assemble the entire fleet to provide reference data. and a test drop is performed. All of the deflection
Therefore the process allows for a subset of the sensors should measure the same value (within pre-
fleet to provide the reference, in a “mini trial”. scribed tolerances).
When selecting the reference devices for these If the tolerance levels are exceeded then the con-
mini trials particular care must be taken in selecting tractor may fine tune individual sensors to bring
devices which will produce a reference representa- them into correct relative calibration. However, if
tive of the whole fleet. If the selected reference large adjustments are required faulty sensors may
devices are not fully representative of the fleet (e.g. need to be replaced or repaired. In either case the
if overall they produce slightly lower deflections), test then would be repeated to show that the per-
then the assessed device may, unfairly, be deemed formance is now suitable.
unsuitable if its FCF value is on the high side (i.e. 3.1.2 Contractor’s calibration check
if it is producing low values relative to the selected The QA process requires the contractor to set up a
reference devices). calibration check site. This site provides a reference
site for the long term monitoring of the perform-
3 QUALITY ASSURANCE ance of the Equipment. Typically the site is a single
site close to the contractor’s base where measure-
Although the accreditation approach results in a ments can be taken safely and without unreason-
consistent fleet on the day of each trial, it does not able disruption to other users of the site.

672
 After achieving accreditation/re-accreditation 3.1.4 Daily check
the contractor is required to undertake a reference The final part of the contractor’s QA process is
survey of their calibration check site. The deflec- the daily check. Its purpose is to provide a regular
tion data is used as a reference dataset from which check on the general condition, configuration and
the performance of the device can be assessed. deflection measurement of the device.
The contractor’s calibration check is performed As a minimum the contractor must carry out a
during the survey season on approximately a visual inspection and check of the device to ensure
monthly basis. There are three parts to the calibra- that the device is operational and correctly con-
tion check, temperature measurement check, loca- figured for the survey to be undertaken. It is rec-
tion referencing check and deflection measurement ommended that the contractor also undertakes a
check. “before and after” check. This involves selecting and
The temperature measurement check involves testing a site prior to the survey (typically close to the
the measurement of an ice bucket and by compar- depot used for the overnight storage of the device).
ing against other suitable references. If the perform- After the survey is complete the site is retested and
ance of the temperature sensor is not suitable then checked against the data collected before the survey.
it should be recalibrated and the check repeated. If there is an unexpected deviation in the results
To check the location referencing systems, the then the contractor should investigate, and if nec-
distance and the OSGR positions (if a GPS system essary return to the contractor’s calibration site or
is fitted) measured during the calibration check are a previous (recent) two-weekly, or daily check site.
assessed against reference data for the site. If the
system(s) is not performing suitably then it should
3.2 Quality checks by the auditor
be recalibrated and the check repeated.
To assess the deflection measurement system The role of the independent Auditor is to provide
the data collected is compared against the refer- advice to the contractor and point out any anomalies
ence data set and the other (valid) data previously that may have been missed. However, the Auditor’s
collected for the site. It is expected that the deflec- role is primarily to monitor, advise and arbitrate.
tion measurements collected on test sites will not The basis of the QA is the contractor takes respon-
necessarily be the same at every visit (for example, sibility for planning, undertaking and delivering QA
they will change with pavement temperature and checks. It is expected that this approach will mini-
subgrade condition). With no absolute reference mise the time from when an issue occurs to when it
measurement of deflection the assessment of the is identified and resolved. However, if the contrac-
results is not straightforward. Therefore the con- tor carries out the checks but does not review the
tractor is expected to have a basic understanding of data (or delays reviewing the data) this can result
both the behaviour of the pavement and the device, in the issue being identified only after several addi-
to determine that the deflection measurements are tional surveys have been undertaken. Therefore the
sensible when compared to the reference and other contractor is required to also provide their QA data
relevant data from that site. If the performance is (with their conclusions)to the Auditor.
not as expected then the contractor will investi- Even though the responsibility for the QA checks
gate the cause and fix the issue if possible (seeking is held with the contractor, the Auditor should
advice from the Auditor if required). be proactive in seeking the data for review. The
Auditor should also keep the end user of the data
3.1.3 Two weekly check updated on the level of QA received from the con-
The two weekly check provides a check on the tractor (where possible). The Auditor will examine
medium term deflection data trending of the device the conclusions and data and provide feedback to
and on the calibration of the location referencing the contractor. If any issues are identified in the
systems. The check is carried out approximately data (not already reported by the contractor) then
every two weeks during the survey season, but they the Auditor may ask the contractor for more infor-
can be carried out up to 20 days apart. mation including data and conclusions from the
The process for this check is similar to the con- other QA checks undertaken. If there are doubts
tractor’s calibration check but it can be carried out in the performance of the device or the QA applied
on different sites referred to as primary check sites. following the checks on the QA, or from other
These sites are set-up by the contractor and can be reports, then the Auditor may undertake additional
spread around the survey network. The sites are spot checks. Spot checks may also be conducted on
not as rigorously defined/set-up as thecalibration the request of the end user of the data. These spot
site and hence if there is any issues during the sur- checks may take the form of visits to the premises
vey (including an anomaly in the distance measure- of the contractor or by conducting repeat surveys.
ment) the contractor typically needs to return to the Typically it has been found that issues can be
contractor’s calibration site to verify and resolve the resolved before the need for spot checks or addi-
issue. tional investigations.

673
4 SUMMARY AND CONCLUSIONS 5 ADDITIONAL INFORMATION

As part of the process for assessing the structural Further details of the accreditation and QA proc-
condition of the English Strategic Road Network esses applied (including the criteria applied) can be
a number of FWD devices operate providing found in “Accreditation and Quality Assurance of
measurements of pavement deflection. To ensure Dynamic Plate Test Survey Devices”. This docu-
that the data provided is reliable and robust, an ment was developed by TRL and Highways Eng-
Accreditation and QA programme has been land in conjunction with survey contractors and
implemented. can be found on the UK Roads Liaison Group
The Accreditation process has been derived website (UKRLG):
from accreditation trials published by the CROW http://www.ukroadsliaisongroup.org/en/asset-
standards organisation, which in turn was devel- condition/road-condition-information/data-col-
oped from work within COST336. This proc- lection/dynamic-plate-test-devices-dpt.cfm.
ess provides confidence that the devices are Similar accreditation and QA processes have been
performing consistently on the day of the trial. developed by TRL and Highways England (again
The accreditation trials have been held in the UK with input from the survey contractors) for Deflec-
on an annual basis since 1998. In this time the tri- tograph and sideway-force skid resistance survey
als have been updated to include checks on tem- devices, also available on the UKRLG website.
perature, distance and OSGR measurements and
more rigorous checks on the repeatability of the
devices. ACKNOWLEDGEMENTS
The accreditation is supported by ongoing QA
carried out during the course of the year. Although This paper has been produced by TRL Limited as
survey contractors have been required to conduct part of a contract placed by Highways England.
QA for many years, formalisation of the approach Any views expressed in it are not necessarily those
is relatively recent, with the first version of the QA of Highways England.
specification being introduced in 2013. This for-
malisation was a useful exercise as it made it clear
to all parties what was required and provided guid- REFERENCES
ance on how to conduct each step.
The accreditation and QA process helps reduce CROW 2011. Publication D11-07, Falling Weight Deflec-
the risk of unsuitable data being supplied to the tometer calibration guide. Ede: Crow
end user of the data. It is also beneficial to the RCMG 2016. Accreditation and Quality Assurance of
survey contractor as it reduces the risk of having Deflectograph Survey Devices, http://www.ukroad-
sliaisongroup.org/en/asset-condition/road-condition-
to undertake repeat surveys due to invalid data. It information/data-collection/deflectograph.cfm
also provides a mechanism for tracking long term RCMG 2016. Accreditation and Quality Assurance
trends in the performance of the device. of Dynamic Plate Test Survey Devices, http://www.
The Accreditation and QA programme should ukroadsliaisongroup.org/en/asset-condition/road-
not remain static and instead should develop as the condition-information/data-collection/dynamic-
devices and requirements develop. Consideration plate-test-devices-dpt.cfm
should be given to both changes in requirements RCMG 2016. Accreditation and Quality Assurance of
and feedback resulting from implementation. This Sideway-force Skid Resistance Survey Devices, http://
may lead to the introduction of additional assess- www.ukroadsliaisongroup.org/en/asset-condition/
road-condition-information/data-collection/skid-
ments or changes to the assessments to reflect resistance.cfm
the needs of the contractor or the end user of the Van Gurp, C. 2005. European Approach to Falling
survey data. For example, a review of the SDDR Weight Deflectometer Calibration. Paper P05-0244,
criteria and migration of the additional tests to 84th Annual Meeting of the Transportation Research
mandatory tests is currently under consideration Board, January 2005. Washington D.C.: Transporta-
for the UK process. tion Research Board.

674
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Inferring pavement layer properties from a moving measurement platform

S. Andersen & E. Levenberg

Department of Civil Engineering, Technical University of Denmark, Denmark

M.B. Andersen
Dynatest Denmark A/S, Denmark

ABSTRACT: This work offered a new method for accessing the mechanical properties of pavement
layers based on data obtained by a moving measurement platform. The method has the following
features: (i) calculations are based on tracking distances to a point on the pavement surface, (ii) meas-
urements are not required outside the zone of load influence, and (iii) possibility to access absolute
pavement deflections. First, the envisioned platform-design and required readings were presented; next,
the interpretation method was outlined and then applied to synthetic (computer-generated) data. Perfect
convergence was achieved when the correct pavement model and accurate input data were employed.
Also, when artificial errors were introduced for assessing sensitivities, only small deviations were gener-
ated. Based on these promising outcomes the method will be further studied and developed to support the
construction of an operational measurement platform.

1 INTRODUCTION deflection bowl, the absolute surface displacement

of a point under a laser sensor is estimated by sub-
Measuring the condition of pavement structures tracting the farthest laser reading. The main tech-
nondestructively has always been of interest in the nological challenge in this scheme is for the lasers
engineering community (STP1026 1989, STP1198 to ‘hit’ the same exact surface point. This is crucial
1994, and STP1375 2000). The current industry- because the pavement surface texture has a depth
standard for this purpose is the Falling Weight that is one or two orders of magnitude larger than
Deflectometer (FWD). While stationary, the the sought deflections.
device applies an impact force to the pavement sur- The TSD is a custom-built truck equipped with
face, and collects the resulting deflections at sev- several Doppler lasers mounted on a stiff beam
eral offset distances from the load center (ASTM (Simonin et al. 2005). The Doppler sensors measure
D4694, ASTM D4695). Provided layer thicknesses instantaneous surface velocities in the direction of
are known, information generated by a FWD test the laser rays. These readings are divided by driv-
is considered accurate enough for back calculating ing speed to yield deflection slopes. Assuming that
individual layer properties, see e.g., Varma et al. one of the sensors resides outside the deflection
(2013). bowl, the slopes may be integrated to yield absolute
With an aim for increasing both safety and deflections. Technological challenge in this scheme
efficiency of FWD testing, as well as minimizing include accurate measurement of driving speed,
traffic disruption, recent efforts have focused on accurate assessment of the differences in incidence
developing moving measuring platforms (MMPs). angles between the lasers, and performing instan-
Recent studies evaluating current technologies taneous control of the beam’s height above the sur-
(Flintsch et al. 2013, Rada et al. 2015) have iden- face to ensure the lasers remain focused.
tified two potentially viable devices, namely the At this time, data generated by either device
Rolling Wheel Deflectometer (RWD) and the Traf- (RWD or TSD) are not sufficiently accurate for
fic Speed Deflectometer (TSD). back calculation. This is due to the aforementioned
The RWD is a custom-built semi-trailer technological challenges which are further aug-
equipped with a long beam (on one side) that car- mented by vehicle dynamics effects. It is for these
ries several triangulation laser sensors, each meas- reasons that the current platforms are essentially
uring the distance between the beam and the road used for network level condition screening (Muller&
surface. The rearmost laser is positioned between Roberts 2013, Zofka et al. 2014, Rada et al. 2016)
the trailer’s dual-tire tires while the other lasers are wherein simplified surface curvature indexes are
positioned further away along the beam (Wilke employed (e.g., Harr & Elton 1983), often spatially
2015). Assuming that the beam extends beyond the averaged to eliminate unreliable data.

675
 This work offers a new interpretation method for of sensors installed in-between the tires. A planar
accessing the mechanical properties of pavement view of the device is shown in Figure 1(a) and a side
layers by use of data from a MMP. The method view is shown in Figure 1(b). The MMP is assumed
is not directly applicable to neither the RWD nor to be traveling along a straight line with constant
the TSD because it requires experimental data that speed V. The two circles in Figure 1(a) indicate
are not currently offered by either device. First, the tire-pavement contact areas each with radius a; the
envisioned platform design and required readings distribution of exerted vertical tire-pavement con-
are presented. Next, the formulation is introduced, tact stress, q(t), is displayed in Figure 1(b) with the
followed by application to synthetic cases for dem- argument t suggesting that the loading intensity
onstrating aptitude and potential. varies with time. A right-handed Cartesian coor-
dinate system is located in-between the tires with
its origin at the un deformed surface; this system
2 THE MEASUREMENT PLATFORM is also moving with constant speed V. The x axis
points to the direction of travel, the y axis points
The envisioned MMP consists of a vehicle transversely, and the z axis points downward into
equipped with a long stiff beam with length LB the pavement medium.
installed close to the tire (or tires) atone side of the The expected MMP readings are discussed here-
rear axle. The beam is nominally oriented paral- after for a random Point A which resides on the pave-
lel to the pavement surface in the travel direction, ment surface along the centerline (y = 0.) Under a
and accommodates a linear array of synchronized fictitious situation for which the MMP is completely
sensors. This array is assumed to be mechanically weightless, i.e., q(t) = 0, the coordinates of Point
constrained such that all nodes move together like A at time t = 0 (arbitrarily chosen) are ( 0 , 0,, )
a single rigid body. wherein x0 is a positive distance that is larger than
As the MMP traverses a pavement section, each LB. This un deformed state is indicated by the
sensor continuously measures and records self-el- dashed line in Figure 1(b). At a later time ( ),
evation, i.e., the distances to all pavement surface the coordinates of this point are ( 0 , 0,, ); i.e.,
points that pass through its field of view. In addi- relative to the coordinate system, Point A appears to
tion, the MMP measures and records the timing be moving purely in the (negative) x axis direction,
of all distance readings. With this arrangement, it and only due to the speed V of the MMP.
becomes possible to identify and track a single (yet Under a realistic situation for which the
random) point on the pavement surface. Moreo- dual-wheel assembly is loaded (i.e. , q(t ) )
ver, the MMP is equipped with an inertial unit the coordinates of Point A at any time t are
for estimating changes in height and angle across ( 0 , 0, ) wherein Δx and Δz denote
short time intervals corresponding to the duration load induced displacements in the x and z direc-
it takes for the same surface point to be detected by tions, respectively. Both Δx and Δz are (also)a func-
two neighboring sensors. tion of x0 and time, i.e., Δx Δx(t ) Δx( x0 V Vt )
Consider the MMP in Figure 1 as an example; and Δz Δz (t ) Δz ( x0 V Vt ). This deformed state
it consists of a dual-wheel assembly and a beam is indicated by the solid line in Figure 1(b). Essen-
tially, both Δx and Δz depend on the mechanical
model chosen to represent the pavement system,
the mechanical properties involved, and the x-coor-
dinate of Point A prior to deformation. Finally,
these displacements depend upon the instantane-
ous magnitude of the loading q(t) in the case of
a time-independent medium, or upon the entire
loading history for a time-dependent pavement
model. Only the former case is addressed herein.
The MMP sensors are identified as Li i = 1...I
(b) (I = 4 in the example). Figure 1(b) shows the beam
Undeformed to be ‘floating’ over the undeformed surface with
...C ........... . height h(t) and angle θ(t) (again, the argument t
Deformed
indicates time wise variability). As the MMP trav-
els forward, each sensor detects Point A, measures
Points of measurement
its distance, and also records the detection time.
The timing for which sensor Li detects point A is
Figure 1. Sketch of a MMP consisting of a dual-wheel denoted by ti and the corresponding distance read-
assembly and equipped with a linear array of four sen- ing is denoted by di With k ...I , time incre-
sors. (a) top view; (b) side view. ments between consecutive readings are defined as

676
 Δtk → k −1 tk − tk , beam angle changes are defined optimization procedure to manipulate the assumed
as Δθ k → k −1 θ (tk ) θ (tk ), and beam height mechanical properties pn and the assumed values
changes are defined as Δhhk → k −1 h(t h tk h(tk ). θ ( I ) and h(tI ) until minimum is attained. Cast in
Consequently, the most elaborate measurement mathematical form:
dataset that realistically can be expected from the
MMP includes: (i) travel speed V; (ii) distance read- min ψ (ψ 1 ( i
FS
) ψ 2(
, d i ), FS
k →k 1 , tk →kk 1 ))
,Δ (1)
pn , ( tI ), h(( I )
ings di ’s; (iii) loading intensities qi ’s ( i (t
(ti ));
(iv) time increments Δtk → k −1 ’s; (v) angle changes where in ψ1 represents the mismatch between meas-
Δθ k → k −1 ’s; and (vi) height changes Δhhk → k −1 ’s. Par- ured and forward simulated distance readings, ψ2
tial datasets can also be envisioned, where some represents the mismatch in the time increments,
of the abovementioned experimental information and the error function ψ1 represents a combined
is not collected. The minimum dataset for which (overall or global) discrepancy. The particular opti-
the analysis can be operated (with assumptions) mization procedure is not specified at this time.
consists of: (i) distance readings di’s; and (ii) some In total there are 2 1 measured MMP data
average time-independent loading intensity q0. values for Point A (I distance readings di ’s and
I − 1 time increments Δtk → k −1 ’s). The number
3 THE INTERPRETATION METHOD of unknowns is N + 2, composed of N medium
parameters pn’s, θ ( I ), and h(tI ). However, the
Based on data collected by the MMP an interpreta- distance readings and the time increments are
tion method was developed for inferring mechani- dependent—constrained by the mechanical
cal properties pn ( n = 1...N ) of a medium, with N medium response and the travel speed. This means
being the total number of sought parameters. The that in an ideal situation for which the chosen pave-
method consists of three elements: (i) an assumed ment model is exact and no errors are present in the
mechanical model for representing the pavement’s MMP data, ψ1 and ψ2 contain identical guidance
load response; (ii) forward simulation of the MMP information for the minimization—suggesting
data collection procedure; and (iii) an optimization that one is redundant. In reality though, errors are
algorithm for inference of properties. contained within in the MMP data and the model
The mechanical model describes the deforma- is only approximate. In such a situation, joint mini-
tion response of any surface point of interest due mization of ψ1 and ψ2 as shown in Equation (1)
to the MMP loading conditions, i.e. tire configura- may be advantageous for reducing both random
tion (e.g., center-to-center tire spacing lc and con- and systematic errors. Alternatively, performing
tact radii), stress intensity qi, and the travel speed V. separate minimization for ψ1 and ψ2 can uncover
These responses are only governed by the medium information about the adequacy of the modeling
properties pn Essentially, this model computes the choice. These aspects will serve as topics for future
displacements Δx and Δz with a complexity level investigations.
that depends on the analyst. The choice should be
guided by the intended usage and to some extent 4 SYNTHETIC CASES
by the quality of the MMP data—which should be
matched with the model sophistication level. In the following the interpretation method is dem-
The second element consists of an efficient onstrated for a number of synthetic examples
numerical ‘engine’ that solves the abovemen- wherein MMP data are computer-generated and
tioned pavement model and forward simulates the used as input. The advantage of so doing is that
MMP data collection procedure for Point A. This the data are both accurate and precise, and the
is performed with assumed values for the param- correct results are known. First presented are the
eters pn plus two unknowns: θ ( I ) and h(tI ) rep- generated data; this is followed by application of
resenting the beam’s angle and the beam’s height the method to evaluate the medium properties and
at time tI, respectively. Also included in the simu- to perform preliminary sensitivity analyses.
lation as known inputs are measured values for
Δθ k , k − ,Δh
Δhk → k 1 qi and V. Overall, the output of
Δh
4.1 Computer-generated MMP data
the computations are forward simulated distance
readings d iFS and forward simulated time incre- The simulated MMP data were generated from
ments ΔtkFS→ k −1. a beam with seven sensors (I = 7), spaced L =
The third element has the task of guiding the 0.4 m apart (i.e., LB = 2.4 m) moving with con-
aforementioned simulation such that the out- stant speed V0 = 10 m/s. The loading intensity was
puts d iFS and ΔtkFS→ k −1 best match the measured q0 = 1 MPa applied over two identical circular
MMP data. This is done by first quantifying an areas with a = 140 mm and center-to-center spac-
error function ψ and then applying a nonlinear ing lc = 0.35 m. Furthermore, the beam height and

677
angle(i.e., when Sensor I detected Point A) were with its origin positioned at the (un deformed)
set to h(tI ) = 176.978 mm and θ ( I ) = 0.001 rad. half-space boundary, the following expressions
The pavement medium was modeled as a linear represent the displacements of any surface point
elastic Boussinesq-type half-space with ν 0 = 0.30 due to a vertical load uniformly distributed over a
and E0 = 500 MPa. It should be stressed, that circular area (Van Cauwelaert 2003):
this model is based on elastostatics and therefore
∞
does not account for time dependencies such as qa( ))(( ) J 0 ( mr )J1 ( ma )
viscoelasticity and dynamic effects. Nonethe- uz (r ) =
E ∫
m=0
m
d
dm (2)
less, it is instructive to employ such a rudimen-
tary model for method development. First and ∞
qa( )(
)( ) J1 ( mr )J1 ( ma )
foremost, for convergence demonstration, and ur (r ) =
E ∫
m=0
m
d
dm (3)
secondly for performing basic sensitivity investiga-
tion. Future work will involve more realistic condi-
tions by incorporating models with higher levels of where uz and ur, are, respectively, the vertical and
sophistication. horizontal displacements of a point originally
Table 1 lists the synthetic MMP readings con- located at z = 0 and at aradial distance r from the
sidered; it includes two index columns (i and k) center of the load; J 0 ( ) and J1 ( ) are Bessel
and four data columns representing: distance read- functions of the first kind of order zero and one,
ings, time increments, angle changes, and height respectively; and m is an integration parameter
changes. The distance readings range between with units reciprocal to length.
172.3166 to 183.0409 mm, with a resolution of
0.1 μm. The time increments range within 39.9273 4.2.2 Forward MMP simulation
to 40.1072 ms, with a resolution of 0.1 μs. The As means of forward simulating the MMP, Equa-
angle changes range within 0.01 rad (= 0.57°) tions 2 and 3 were evaluated with assumed val-
with a resolution of 0.001 rad. Finally, the height ues for E and v. Numerically, this was done by
changes range within 5.1 mm with a resolution of means of Gauss integration carried out between
1 μm. The simulated MMP data areexpected to be the roots of the Bessel functions involved. Sub-
in the realistic range, but the resolutions are supe- sequently, the displacements Δx and Δz along the
rior by two orders of magnitude compared to cur- centerline between the two loads (the x axis, see
rent technology capabilities. Figure 1), were generated via simple superposi-
tion. Identifying the location of Point A at t = 0
as x0 = 3 m, values for Δx and Δz were calculated
4.2 Application ofinterpretationmethod
at 1500 equidistant intervals covering the range
The data in Table 1 were taken as input to the pro- x ∈[ x0 / 30 ;x
; x0 ] corresponding to a discrete spac-
posed method with the ultimate aim of converging ing of approximately 2 mm. Intermediate values
to the correct Young’s modulus of the half-space for Δx and Δz where estimated with linear interpo-
(i.e. 500 MPa). The three elements of the inter- lation. Table 2 presents a summary of the param-
pretation method are sequentially described in the eters used as input in the forward simulation.
following. As an example, the displacements Δx and Δz as a
function of the x-coordinate, are shown in Figure 2
4.2.1 Mechanical model for v = 0.30 and an assumed E = 200 MPa. As can
Similar to the generated synthetic case a linear elas- be seen in this figure, the load introduces posi-
tic Boussines q-type half-space model was also con- tive (downward) Δz displacements along the con-
sidered. In terms of a cylindrical coordinate system sidered interval with a maximum value attained

Table 2. Forward simulation input parameters.

Table 1. Synthetic data for interpretation.
Parameter Magnitude Unit
I K di Δtk→k−1 Δθk→k−1 Δhk→k−1
[−] [−] [mm] [ms] [rad] [mm] I 7 –
L 0.4 m
1 2 177.6336 39.9689 0.002 2.000 LB 2.4 m
2 3 179.7447 39.9445 0.003 0.200 V0 10 m/s
3 4 177.8674 40.1072 −0.006 −1.200 q0 1 MPa
4 5 183.0409 39.9273 0.004 −2.500 A 0.14 m
5 6 175.7224 39.9472 0.003 2.600 lc 0.35 m
6 7 172.3166 40.0703 −0.004 −1.300 v0 0.30 –
7 – 179.4087 – – – x0 3 m

678
 -0 .1 i:·-·-·---·~·-·-·-·-·-·-·-·-·-·-·-·-·-·-·-·-·-·-·-·-·-·-·-
Az -------
Point A

c - Defun nat iu n Basin - Deform ation Basi n

~
I

0.5 1=0
,,
I 1=1,

"
]. I

i5 Figure 3. Identification scheme of point A by sen-

0.8 sor LI in the forward simulation. (a) start time for
I which θ AI ( ) θ (tI ); (b) detection time for which
I
I I
I
θ AI ( I ) θ (tI ).
1.1 ,I
x-coordinate[m]

0 0.6 1.2 1.8 2.4 3

were found from the discrete set of displacements
Δx and Δz. This was done by detecting when the
Figure 2. Centerline surface deformations Δz ( ) and
angle θ Ai (t ), defined between the z axis and a line
Δx ( ) based on Equations 2 and 3 for a medium with
E = 200 MPa and ν = 0.30 loaded overtwo circular areas connecting sensor i and Point A, was equal to the
with radii a = 0.14 m, center-to-center spacing lc = 0.35 beam angle, θ ( i ). The principle is illustrated in
andstress intensity q0 = 1 MPa. Figure 3 for Sensor I.
Figure 3(a)refers to the very beginning of the
simulation when t = 0, where Point A is situated at
the very edge of the discrete displacement field; at
between the two loads where x = 0. The value for this initial stage θ AI ( ) θ (tI ). As the simula-
Δz decreases monotonically as the absolute value tion continues ( ), and the MMP moves in the
for x increases. The horizontal displacements Δx positive x-direction with speed V, the angle θ AI (t )
are approximately one order of magnitude smaller continuously decreases. When the point of equal-
than the transverse deformations Δz, and the sign ity is reached, i.e., θ AI ( ) θ (tI ), then tIFS t and
of Δx indicates that points are displaced towards x d IFS is simply evaluated as the distance between
= 0. Therefore Δx = 0 at x = 0, Δx < 0 for x > 0, Sensor I and Point A.
and Δx > 0 for x < 0. The Δx curve is not monot- Formally, denoting the coordinates of sensor i
onous w.r.t. x (in an absolute sense), with values as ( i (t ), 0, i (t )) the distance is described by the
increasing for a certain range and then decreasing expression:
towards zero as x increases. As a result of choos-
ing a time independent model, Δ ( x ) Δz ( x )
(x )
2
and Δx( x ) Δx( x ). V tI x tI xIs tI
d IFS = (6)
With the availability of the displacements +( z tI zIs tI )2
Δx and Δz, as well as knowledge of speed V, the
MMP forward simulation involved tracking Point
A in both space and time. At time t = 0 the coor- Furthermore, the angle between Point A and
dinates of Point A were ( 0 (0 ), 0, (0 )); as sensor i is described as:
the MMP (and the coordinate system) moved for-
ward, the coordinates of Point A were described as ⎛ ( x - V t Δx(t )) - xis (t ) ⎞
θ Ai (t ) = arctan ⎜ 0 0 ⎟⎠ (7)
( 0 t )), 0, Δz (t)).
(t),0 ⎝ Δz (t ) - zis (t )
At this stage, the height h(tI ) and the beam
angle θ ( I ) when Sensor I detected Point A were By repeating the above described detection
assumed. Subsequently, the height and angle of scheme for all sensors, the forward MMP simula-
any sensor i(i < I ) when Point A was detected is tion delivers a set of I distance readings diFS and
given by: I − 1 time increments ΔtkFS→ k −1 corresponding to
i +1
the values in Table 1. This set essentially depends
h(ti ) = h tI ) + ∑ Δhk → k −1 (4) on the initial assumption of E and on the pre-
k I sumed h(tI ) and θ ( I ).
i +1
θ ( i ) θ ( I ) + ∑ Δθ k → k −1 (5) 4.2.3 Optimization and property inference
k I The optimization has the task of manipulating the
unknowns; the Young’s modulus of the half-space
Under these assumptions the distance read- E and both the assumed beam height h(tI ) and
ings, diFS , and the corresponding timings, tiFS , beam angle θ ( I ). The intention is to best match

679
the forward simulated MMP distance readings 80 .----~--~--~-------, 0.8
and time increments with their corresponding
synthetic MMP readings (given in Table 1). The
matching process was carried out with a general 0.6
purpose multidimensional unconstrained minimi-
zation algorithm as implemented by the Mat lab 8
function fminsearch (Lagarias 1998). This optimi- 2;40 0.41
zation method operates only with objective func- ~
tion values without any derivative information. In
this connection, the following objective functions 20 0.2
wereseparately used:
0 '----~--~"'------~------' 0
I
1 0.5 0.75 1.25 1.5
∑( )
2
ψ1 = − (8)
I i =1

I
1 Sensitivity of the objective functions ψ 1 ( )
∑( ) Figure 4.
2
ψ2 = − (9)
I −1 k =2
1 1
and ψ 2 ( ) to the Young’s modulus ratio E FS / E0 .

where ψ1 represents the average matching error in

distances (units of length) and ψ2 represents the 4.3.2 Sensitivity to errors in distance readings
average matching error in time increments (units The second variant considers the sensitivity of the
of time). inferred Young’s modulus EFS to errors in the syn-
After application of the minimization algorithm thetic MMP distance readings. Figure 5 presents
using either Equation 8 or Equation 9, the forward this sensitivity; it plots the ratio FS
/ E0 when
simulated MMP data converged to the synthetic 10 μm were either added or subtracted from the
MMP data in Table 1 with zero error producing synthetic sensor readings (one sensor at a time).
the correct half-space modulus of 500 MPa. Such The range of deviation is depicted by the solid
perfect outcome provides basic confidence in the horizontal lines and the circles in the figure, rep-
proposed method as well as foundation for further resenting the results when using either ψ1 or ψ2 to
increasing the modeling sophistication in future guide the minimization. The plus/minus markers at
work. the line edges indicate whether the 10 μm distance
error was added or subtracted.
When ψ2 was employed to guide the minimiza-
4.3 Sensitivity analysis
tion, the method converged to the correct result
In the following, three variants of the previously regardless of any errors in the distance readings.
analyzed synthetic case are presented. These vari- This outcome is denoted by circular markers in the
ants were generated to investigate built-insensitivi- figure superposed over the solid lines. This is an
ties in the interpretation method. expected output whenever employing a time-inde-
pendent model to represent the pavement medium.
4.3.1 Sensitivity to half-space modulus When ψ1 was employed to guide the minimiza-
This first variant considers the sensitivity of the tion, the method produced E FS / E0 ratios that
objective functions ψ1 and ψ2 to the Young’s mod- deviated from unity. The deviation intervals are
ulus value. This sensitivity is graphically presented seen to be of non-equal size and non-symmetric
in Figure 4 which depicts these two functions vs. around the correct outcome (where E FS / E0 ).
the ratio between the forward simulated modu- This means that the sensitivity is sensor dependent
lus E FS and the exact modulus E0 = 500 MPa. and also dependent upon the error sign.
No minimization procedure was needed for gen- The narrowest intervals appear when an error
erating the curves; they were simply prepared by is introduced into either Sensor 5 or Sensor 6,
varying EFS (with the values in Table 1 as input with deviations ranging from approximately 0.994
parameters). to 1.003. The widest interval appears when add-
From Figure 4 it is seen that both ψ1 and ψ2 ing the distance error to Sensor 1, with deviation
reach their minimum (perfect zero) value for ranging from approximately 0.968 to 1.031. Add-
E FS / E0 . By increasing or decreasing the ratio ing the distance error to Sensors 2, 4 or 7 gen-
EFS /E0, ψ1 and ψ2 are seen to increase monotoni- erates deviations in the range of approximately
cally. Thus, the value FS / E0 = 1 signifies a dis- 0.984 to 1.013. Adding the distance error to Sen-
tinctly clear minimum, independent of the choice sor 3 resulted in a deviation range of approxi-
of an objective function. mately 0.980 to 1.017. Overall, it seems that the

680
 - 'I/J1 As can be seen in the figure, when ψ1 was guid-
7 0 'I/J2 + ing the convergence, a vertical line at E FS / E0 = 1
is generated—indicating insensitivity to errors in

...0
'It:

00
s::
Q)
4
6
5
--
+ ----e- -
+
+
travel speed. This outcome is to be expected for a
medium that is based on time-independent proper-
ties. Conversely, when ψ2 was guiding the conver-
gence, the sensitivity to errors in speed is extreme.
Vl 3 + By decreasing or increasing the speed ratio from
2 + the benchmark point, the quotient E FS / E0 is seen
to experience an exponential change in magnitude.
+ An infinitesimal decrease in the speed increases
the modulus ratio by three orders of magnitude;
0.96 0.98 1.02 1.04 an infinitesimal increase in the speed ratio gener-
EFs I Eo ates a decrease of about one order of magnitude in
E FS / E0 . This indicates that, in reality, minimiza-
Figure 5. Sensitivity of method output in terms of tion based solely on ψ2 is expected to yield ques-
E FS / E0 to an error of ±10 μmin a single distance read- tionable results. On the other hand, if the medium
ing; interpretation was based on either ψ 1 ( ) or ψ 2 ( ). properties are inferred via ψ1 then the MMP speed
can be inferred to a very high degree of accuracy.
Such information can be beneficial when time-
1.001 dependent modeling is considered for the pave-
ment medium.

5 CONCLUSIONS

This work offered a new method for accessing the

mechanical properties of pavement layers by use of
data from a MMP consisting of a beam equipped
\,/lf/2 with several distance sensors. At its core, the
,,
·...... method is based on sequential detection and track-
ing of random pointson the pavement surface, as
0.999 L-----L---~--~--~ well as on a presumed mechanical pavement model
10-1 to constrain and guide the interpretation. Fur-
thermore, measurements outside the zone of load
influence are not required.
The method was evaluated by application to
Figure 6. Sensitivity of method output in terms of rudimentary synthetic data as a first and essential
E FS / E0 to errors in MMP speed; interpretation was development step. The outcomes of the investiga-
based on either ψ 1 ( ) or ψ 2 ( ).
tion are deemed promising given that perfect con-
vergence was achieved when the correct pavement
model and accurate input data were employed.
sensitivity of the inferred Young’s modulus EFS Also, when artificial errors were introduced to
to isolated errors in the synthetic MMP distance the analysis for assessing sensitivities, only small
readings is small. deviations were generated. This suggests potential
accessing layer properties and absolute pavement
4.3.3 Sensitivity to error in speed deflections
This third variant considered the sensitivity of Further work is planned in this connection as
the inferred modulus EFS to errors in the speed V means of supporting the construction of an opera-
employed in the forward MMP simulation. This tional MMP prototype. This will include testing
sensitivity is graphically shown in Figure 6 with the for sensitivities to errors in reported beam angle
quotient E FS / E0 plotted along the abscissa and changes ( k →k 1 ) and beam height changes
the speed ratio V /V0 plotted along the ordinate. ( k →k 1 ) as well as sensitivities when choosing
The point where V /V0 = 1 and E FS / E0 = 1 indi- an incorrect mechanical model for analysis. The
cate zero error in speed and serves as the bench- next development step would deal with enhanc-
mark point for this plot. Two curves are shown ing modeling sophistication (e.g., multilayered
passing through this point, indicating whether ψ1 half-space, viscoelasticity) and dealing with field
or ψ2 were employed in the minimization. measurements.

681
ACKNOWLEDGEMENT Conference on Managing Pavement Assets, Alexandria,
VA.
The authors would like to acknowledge the Inno- Rada, G.R., Nazarian, S., Visintine, B.A., Siddharthan,
vation Fund Denmarkand Dynatest for the project R. & Thyagarajan, S. 2016. Pavement structural evalu-
ation at the network level. Report FHWA-HRT-15-074,
funding. This work is part of a larger effort aimed Federal Highway Administration, Washington, D.C.
at the development of a new moving measure- Simonin, J.M., Lievre, D., Hildebrand, G. & Rasmussen,
ment platform for non-destructive pavement S. 2005. Assessment of the Danish high speed deflec-
evaluation—provisionally called “Raptor.” tograph in France. Paper Presented at the 7th Interna-
tional Conference on the Bearing Capacity of Roads,
Railways, and Airfields, Trondheim, Norway.
STP1026. 1989. Nondestructive testing ofpavements and
REFERENCES backcalculation of moduli. G.Y. Baladi, & A.J. Bush
(eds). Special Technical Publication, ASTM Interna-
ASTM D4694-09. 2015. Standard test method for deflec- tional, West Conshohocken, PA.
tions with a falling-weight-type impulse load device. STP1198. 1994. Nondestructive testing of pavements
ASTM International, West Conshohocken, PA. and backcalculation of moduli: second volume. A.J.
ASTM D4695-03. 2015. Standard guide for general pave- Bush, H.L. Von Quintus & G.Y. Baladi (eds). Special
ment deflection measurements. ASTM International, Technical Publication, ASTM International, West
West Conshohocken, PA. Conshohocken, PA.
Flintsch, G., Katicha, S., Bryce, J., Ferne, B., Nell, S. STP1375. 2000. Nondestructive testing of pavements and
& Diefenderfer, B. 2013. Assessment of continuous backcalculation of moduli: third volume. S.D. Tayabji
pavement deflection measuringtechnologies. Report & E.O. Lukanen (eds). Special Technical Publication,
S2-R06F-RW-1, Transportation Research Board, ASTMInternational, West Conshohocken, PA.
Washington, D.C. Van Cauwelaert, F. 2003. Pavement design and
Harr, M.E. & Elton, D.J. 1983. Non-contact, non- evaluation—the required mathematics and its
destructive airport pavement profile, texture and applications.M. Stet (ed). Federation of the Belgian
deflection measurements. Report DOT/FAA/ Cement Industry, Brussels, Belgium.
PM-83/14, U.S. Department of Transportation, Fed- Varma, S., Kutay, M.E.& Levenberg, E. 2013. A viscoe-
eral Aviation Administration. lastic genetic algorithm for inverse analysis of asphalt
Lagarias, J.C., Reeds, J.A., Wright, M.H.& Wright, P.E. layer properties from falling weight deflections. Trans-
1998. Convergence properties of the Nelder-Mead portation Research Record: Journal of the Transporta-
simplex method in low dimensions. SIAM Journal of tion Research Board 2369:38–46.
Optimization 9(1): 112–147. Wilke, P.W. 2015. Network level structural evaluation
Muller, W.B. & Roberts J. 2013. Revised approach to with rolling wheel deflectometer. Paper presented at
assessing traffic speed deflectometer data and field the 9th International Conference on Managing Pave-
validation of deflection bowl predictions. International ment Assets, Alexandria, VA.
Journal of Pavement Engineering 14(4): 388–402. Zofka, A., Sudyka, J., Maliszewski, M., Harasim, P.&
Rada, G.R., Nazarian, S., Visintine, B.A., Siddhar-than, Sybilski, D. 2014. Alternative approach for interpret-
R.V. & Sivaneswaran, N. 2015. Use of high-speed ing traffic speed deflectometer results. Transportation
deflection devices in network-level PMS applications: Research Record: Journal of the Transportation
are we ready? Paper presented at the 9th International Research Board 2457:12–18.

682
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Effect of embedment scheme on damage detection performance

of self-powered pavement monitoring sensor

A.H. Alavi, H. Hasni, K. Chatti & N. Lajnef

Department of Civil and Environmental Engineering, Michigan State University, East Lansing, MI, USA

ABSTRACT: An experimental study was conducted to determine the effects of geosynthetic reinforce-
ment on performance of piezoelectric self-powered sensors for pavement health monitoring. A new mini-
aturized disc shape packaging system was designed for the protection of the sensing system. The size of
the entire packaging system is of the same order of a coarse aggregate particle. However, there were two
concerns about the application of such embedded sensing packaging system: (1) the area around the sen-
sor might be prone to crack initiation due to the disruption of the continuity of the pavement, and (2) the
protective system may be damaged due to excessive tensile strain at the bottom of the asphalt layer. To
deal with these issues, the area around the packaging was reinforced using Alkali Resistant (AR) glass-
fiber mesh. A series of tests were performed under a three-point bending configuration on asphalt slabs
to analyze the response of embedded piezoelectric strain transducers with and without reinforcement due
to the progression of bottom-up cracks. The cyclic tests were done at 2 and 5 Hz loading frequency for
different loading amplitudes. Based on the results, glass-fiber reinforcement reduces the stress concentra-
tion around the packaging system. This is while it does not affect the damage detection performance of
the sensors.

1 INTRODUCTION sensor might be prone to crack initiation. This is

because the sensor packaging might disrupt the
Wireless sensor networks are now widely used for continuity of the pavement. Reinforcement of the
Structural Health Monitoring (SHM). A major area around the packaging system can be regarded
concern for the application of wireless sensors is as an effective solution to deal with this issue.
to find a sustainable and cost-effective power sup- Accordingly, an experimental study was conducted
ply. This issue becomes more challenging for the to determine the effects of geosynthetic reinforce-
long-term monitoring of pavement structures. The ment on the performance of the sensors for pave-
reason is that periodic replacement of batteries of ment health monitoring. A new miniaturized disc
embedded sensors or use of solar power technol- shape packaging system was designed for the pro-
ogy is usually cost-prohibitive and in some cases tection of the sensing system. The area around the
impractical (Alavi et al. 2016a). A viable solution packaging was reinforced using Alkali Resistant
for this problem is to harvest the mechanical load- (AR) glass-fiber mesh. A series of tests were per-
ing experienced by the pavement and convert it into formed under a three-point bending configuration
electrical energy. To cope with this issue, the authors on asphalt slabs to analyze the response of embed-
at Michigan State University (MSU) have devel- ded piezoelectric strain transducers with and with-
oped a new class of Self-powered Wireless Sensors out reinforcement to the progression of bottom-up
(SWS) (Lajnef et al. 2013, Alavi et al. 2016a). The cracks.
designed SWS is a small size battery-free sensor.
This unique sensor is based on the integration of
the piezoelectric transducers with an array of ultra- 2 EXPERIMENTAL STUDY
low power floating gate computational circuits. The
performance of this sensing systems for the detec- For the experimental study, an asphalt concrete
tion of damage progression in civil infrastructure sample with a span length of 381 mm (15 inches),
has been verified during previous studies (Alavi thickness of 165.1 mm (6.5 inches) and a width of
et al. 2016a-c). 152.4 mm (6 inches) was built. The slab was con-
However, a probable problem with the embed- struct using the HMA, 4E1 mixture type. The weight
ment of the MSU’s self-powered sensor packaging of the HMA was 25 kg and the length of the slab
(spherical or disc shape) is that the area around the was equal to 450 mm (17.75 inches). The tests were

683
done at 2 and 5 Hz loading frequency for 0.1, 0.15 age was read on NI 9220 data acquisition system
and 0.2 mm amplitudes for 30 cycles. The test was s(with 1 GΩ impedance) in parallel with a resistor
repeated with 2 identical samples to assure the accu- with impedance similar to the self-powered wire-
racy of measurements. 2 PZTs with disc shape pack- less sensors. Damage was introduced by making a
aging were embedded within the asphalt layer at a notch at the bottom of the asphalt layer. The dam-
distance of approximately 50.8 mm (2 inches) from age states were defined by increasing the notch size
the bottom of the layer. The discs were located at a (a) as follows:
distance of 76.2 mm (3 inches) from the mid span. • Intact: Intact plate (a = 0 mm)
Before starting the test, a preload equal to 0.5 kN • Damage 1: a = 12.7 mm (0.5'')
was applied to the sample to ensure it is seated on • Damage 2: a = 25.4 mm (1'')
the fixture. Thereafter, the cyclic displacements were
applied. Figure 1 shows the disc shape packaging After introducing the second damage phase,
manufacturing process sensor packaging and three- a crack propagation phenomenon was observed
point bending configuration. The piezo output volt- (Figure 2). During the loading cycles, the crack

(a) Manufacturing of the disc shape packaging (c) PZT locations

Figure 1. Sensor packaging and test setup.

Figure 2. Crack propagation phase during the test

684
propagated 6.35 mm (0.25 inch). This new damage the averages obtained from these tests. The dam-
phase was considered as Damage 3 as follows: age detection results are provided for a typical
amplitude of 0.2 mm and loading frequencies of
• Damage 3 (crack propagation phase): a = 31.75
2 and 5 Hz. Figures 3(a) and (b) show the voltage
mm (1 1/2'')
outputted by PZTs during different damage states
In order to reinforce the area around the pack-
for the 2 and 5 Hz loading frequencies, respectively.
aging, a glass fiber mesh was inserted across the
As observed from these figures, the voltage gener-
diameter of the packaging (Figure 1). Alkali Resist-
ated by both of the PZTs continuously increases
ant (AR) glass fiber mesh with 10 mm (0.4 inch)
due to damage progression for the given frequen-
square opening was obtained from Nippon Elec-
cies. The important observation is that in all cases,
tric Glass America, Inc. The used fibers are heat
PZT 2 which is reinforced by glass fibers generates
resistant with a softening point around 1037°C
less voltage. A reasonable explanation for this phe-
(1900 F) and a melting point of 1315°C (2400 F).
nomenon is that mesh reinforcement reduces the
The glass fiber specifications are as follows:
stress concentration around the packaging system
• Code Name: TD 10X10 and therefore less strain is applied to the PZT. Fig-
• Type of Glass: Alkali Resistant Glass ure 4 shows the percentage of voltage drop due to
• Weight (g/m): 80 ± 10 reinforcement for the 2 and 5 Hz loading frequen-
• Tensile strength (N/25 mm): More than 200 cies. As seen in Figure 4, the voltage drops between
20 to 30% due to reinforcement. The averages of
voltage drop page due to reinforcement for the
3 EFFECT OF REINFORCEMENT ON 2 and 5 Hz loading frequencies are close to each
SENSOR OUTPUT other and are equal to 26% and 22%, respectively.
In order to analyze the effect of reinforcement
The three-point bending tests were repeated with on the damage detection performance, the changes
two identical samples. The result given here are of voltage increments due to damage progression

2.5
- PZT 1 (Without R einfor ce ment)
•••• ,.. ••• PZT 2 (With Reinforcement)
2.0
Z'
=
~ 1.5
0.1
QJ) ·········"'
.::: ·········•············
=
>
1.0
&·····················&···
·········
0.5

0.0
Intact Damage 1 Damage 2 Damage 3

(a) 2 Hz loading frequency

2.5 - r _--...,P
:::Z::-:T:-:-
1 c::
(\::-:
v ,'·th tR
"': -o-u--,:::e-:i-
n~=-
o-I-c-
em
- en-t):-1

• •• •&e•• • PZT 2 (\Vith Reinforce ment)

2.0

Z'
=
~
0.1
QJ)
1.5
~t·····················"'
...... .... .
.::: ·········
=
>
1.0
&····
0.5

0.0
Intact Damage 1 Damage 2 Damage 3

(b) 5Hz loading frequency

Figure 3. A comparison of maximum delivered voltage by PZT discs for 0.2 mm displacement.

685
 40% DFrequency: 2Hz
0
..... • Frequency: 5Hz
"'= :;? 29%
Q ~ 30%
"'g:.l) E
c.s
c..,
~ ~ 20%
Q~
"' =
Oil ·-
~ "'
.::a: 10%
0
>
Oo/o
Intact Damage 1 Damage 2 Damage3

Figure 4. Voltage drop due to the reinforcement for the 2 and 5 Hz loading frequencies.

"'
OJ)
~
50% D PZT 1 (Without Reinforcement)
E • PZT 2 (With Reinforcement)
~
Q 40%
,.-._
E ~
~ 30% 31%
"'= = 30%
29% 28%
Q 0
·:;:
"'"' ..."'
~
"' 20%
"' 0
b
OJ) 16%

=0.....
"'
OJ)
_g
0
> 0% +--..L..--
Intact to Damage l Damage 1 to Damage 2 Damage 2 to Damage 3

(a) 2 Hz loading frequency

50%
"'
OJ)
~
D PZT 1 (Witho ut Reinforcement)
• PZT 2 (With Reinfo•·cement)
E
~
Q 40%
,.-._
E 32% 33%
~
~
"'= =
Q
30% 27% 26%
25%
0 24%
·:;:
"'"' ..."'
"'<J... ..."'Oll
~
20%
0
= 0..
10%
"'
OJ)
~
.....
0
> 0°/o
Intact to Damage 1 Damage 1 to Da mage 2 Damage 2 to Damage 3

(b) 5 Hz loading frequency

Figure 5. Percentage of Voltage increase due to damage progression.

686
were calculated and presented in Figure 5. As the packaging system and therefore less strain is
observed from Figures 5(a) and (b), the increasing applied to the PZT. It is notable that generating
rates of the voltage outputted by PZT 1 (without less voltage due to reinforcement is not always
reinforcement) and PZT 2 (with reinforcement) are desirable because a specified level of voltage
very similar. As an example, for the 5 Hz loading should be delivered to the sensors to activate
frequency, the voltage increases 27% and 26% by their memory cells (floating gates).
transitioning from the Intact to Damage 1 mode, 3. The increasing rates of the voltage of PZTs
respectively for PZTs 1 and 2 (Figure 5(b)). That is with and without reinforcement are very simi-
to say, the detection performance of the sensors is lar. That is to say, reinforcement does not have
not affected by the reinforcement. any effect on the detection performance while it
results in reduction of the generated voltage.

4 CONCLUSION
REFERENCES
The effect of embedded self-powered sensor pack-
aging on the continuity of the pavement was evalu- Alavi A.H., Hasni H., Lajnef N., Chatti K., Faridazar F.,
ated through a series of three-point bending tests 2016a. Continuous Health Monitoring of Pavement
on 2 asphalt slabs. The area around the packag- Systems Using Smart Sensing Technology. Construc-
ing system was reinforced using the AR glass-fiber tion and Building Materials, 114, 719–736.
Alavi A.H., Hasni H., Lajnef N., Chatti K., Faridazar F.,
mesh. The response of embedded piezoelectric 2016c. Damage Detection Using Self-Powered Wire-
strain transducers with and without reinforcement less Sensor Data: An Evolutionary Approach. Meas-
were measured due to bottom-up cracking. The urement 82, 254–283.
results indicate that: Alavi A.H., Hasni H., Lajnef N., Chatti K., Faridazar
F. 2016b. An Intelligent Structural Damage Detection
1. The delivered voltage by PZT discs with and Approach Based on Self-Powered Wireless Sensor
without reinforcement increases due to damage Data. Automation in Construction, 62, 24–44.
progression. Lajnef N., Chatti K., Chakrabartty S., Rhimi M., Sarkar
2. The important observation is that PZT 2 which P., 2013. Smart Pavement Monitoring System. Report:
is reinforced by glass fibers generates less volt- FHWA-HRT-12-072, Federal Highway Administra-
age than PZT 1. This is because mesh reinforce- tion (FHWA), Ishington, DC.
ment reduces the stress concentration around

687
Backcalculation analyses of deflection measurements
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Analysing FWD data—from the deflection bowl to material parameters

M. Čičković
Technische Universität Darmstadt, Darmstadt, Germany

ABSTRACT: Regarding the bearing capacity of roads, there are many ways of gathering and analysing
data. The Falling Weight Deflectometer (FWD) is a testing device, which has established itself over the
past decades. But the FWD itself cannot give us an interpretation of the data gathered—naturally, there
are many ways of analysing and interpreting the deflection bowl. Therefore, the purpose of this paper is to
introduce German methods of analysing FWD data, which give a more indirect assessment of the struc-
tural condition of the pavement, and compare these methods with international methods (backcalculation
of moduli). The added value of this paper regarding the current state of the art is the critical analysis of
simplified two-layer-models and standard modulus back calculations.
Three pavement samples (asphalt), which moderate the comparison, are being utilised, distinguish-
able by the amount of traffic endured. These pavement samples are then analysed according to the three
methods introduced.
The main finding is that a two-layer model and its individual assessment methods are simplifying the
pavement structure. Consequently, the findings are rather simple. A three-layer model and the backcal-
culation of Young’s moduli gives further insight into the structure, but is prone to errors that derive from
the uncertainties of the construction design data (e.g. thicknesses). This leads to the conclusion that a
two-layer model may be too simple at first sight, but is to be seen as a concession to data uncertainties.

1 INTRODUCTION

The first Falling Weight Deflectometer (FWD) was

constructed and developed in 1964 in Denmark,
based on principal thoughts and preliminary stud-
ies made in France (Bohn 1989, Grätz 2009). The
measuring principle is quite simple: a falling weight
drops from a defined height onto a spring damping
system, which transmits the impulse to a steel plate
that is touching the ground (D = 30 cm). The steel
plate itself is equipped with a rubber mat, in order
to ensure an all-over contact with the examined
pavement (see Figure 1).
The vertical deflection that is being provoked by
the falling weight is recorded by geophones. Each
geophone consists of a spring-suspended perma- Figure 1. Operating principle of a Falling Weight
Deflectometer (Kaakkurivaara et al. 2015).
nent magnet and coils. If the falling weight evokes
a vertical deflection of the pavement, the mass
inertia of the geophone causes a relative move- 600 mm–900 mm–1.200 mm–1.500 mm–1.800 mm–
ment of permanent magnet and coil. This relative 2.100 mm (possible arrangement in Germany even
movement produces an inductive voltage, which is for flexible pavements). Not only the maximum
proportional to the vertical velocity of the pave- deflection of each geophone is being recorded,
ment and can therefore be translated into a vertical the deflection time history is being recorded as
movement of the pavement. well and can be used for further examinations. The
The maximum deflection of each geophone time frame of each measurement is set to 60 ms.
is determined by extreme value determination Besides the deflection data, surface temperature,
of the deflection time history. The geophones ambient temperature and stress curve are being
are arranged in the following manner (from the recorded. These values are important to stand-
load centre outwards): 0 mm–210 mm–300 mm– ardise the gathered deflection data to reference

691
conditions. In Germany the reference conditions [N], M0 = layer modulus of the halfspace [MPa],
are defined by the asphalt temperature (Tref = 20°C) Hw = auxiliary value [–].
and the impact force of the FWD (Qref = 50 kN)
(FGSV 2014). ⎛r ⎞ ⎛r ⎞
J0 ⋅t J 0 ⋅t
⎛ r⎞ 2⋅l ∞ ⎝ l ⎠ 1⎝ l ⎠
⎝ l ⎠ π ⋅ r0 ∫0
Hw ⎜ ⎟ = ⋅ d
dt (2)
t ( t )
2 MECHANICAL FUNDAMENTALS
where l = elastic length [mm], r = distance to load
The following chapter shall give a brief introduc- centre [mm], r0 = radius of the load plate [mm],
tion into the three compared evaluation methods J0 = Bessel function of the first type and zero order,
of FWD data. First, the German methods are J1 = Bessel function of the first type and first order,
being described, as their mechanical principle is t = auxiliary variable for the integration of the
based on a two-layered system. After this short Bessel functions.
introduction into these German evaluation meth- Due to the fact that this specific integral equa-
ods, the multi-layer theory is described. tion does not have a closed solution, parameter
studies and several assumptions were needed to
2.1 Darmstadt evaluation method stipulate a regression equation (Grätz 1999):
The basic principle of Grätz’s evaluation method w ( r ) = A (e B a r + a ) (3)
and its advancement is the concept of a slab on an
elastic-isotropic halfspace (see Figure 2). This con- where w = deflection [mm], A = regression param-
cept is not suited for bodies with finite dimensions. eter [mm], B = regression parameter [mm−1],
Therefore, peripheral areas or joints cannot be r = distance to load centre [mm], a0 = regression
described with this concept (FGSV 2011a). Plate coefficient (= 0.392948), a1 = regression coeffi-
theory entains also the following simplifications cient (= −0.398483), a2 = regression coefficient
(Gerlach 1968): (= 0.0137024).
This regression approach is highly satisfying (R2
• Plate is stiff in comparison to the halfspace
mostly higher than 0.99 – see Figure 3). Another
• Plate thickness t is insignificant in comparison
with the remaining dimensions of the plate
For the special case of a circular load and the
resultant vertical deflection of the surface, numeri-
cal equations are needed in order to calculate the
100 r--------r
I

I
I

I
I I

deflection (Zhu 1985, Grätz 1999): I

I
I
I
200 1---------t---
1 I
I I

⎛ r⎞ Q ⎛ r⎞ I I

w⎜ ⎟ = ⋅ Hw ⎜ ⎟ (1) I I I
L--------L--------L--- ----L--------.1. ,
I I

⎝ l ⎠ M0 l ⎝l⎠ 300 1
I
I
I
I
I
I
I
I
I
'E
:::l.
I
I
I
I
I I
I
I
I

where w = deflection [μm], r = distance to load cen-

'C' 400
I I
~---------~------

1 I
I
-----~--------~ ·

I
I

I
0
tre [mm], l = elastic length [mm], Q = impact force
I I I I
"8
Q)
I I I I
I I I I
~ 500 r--------r-----
1 I
------r--------r ·
I I
0
I I I I
600 1---------~-------- ------~--------~ ·
1 I I
I I
I I
I I I I
700 ~--------~------- -------~--------} ·
1 I I I
I I I I
load plate I I I I I

800 ~--------~-------- --------~--------}

I I

Young's modulus E,
plate Poisson's ratio }11 h -2 000 -1 000 0 1 000 2 000
layer thickness h Distance from the Load Centre [mm]

Young's modulus .&

Measured Deflection Bowl
halfspace
Poisson's ratio Jlo Approximated Deflection Bowl
Deflection Bowl Resulting from Plate Theory
Figure 2. Model ‘slab on elastic-isotropic halfspace’
without layer bonding between slab and halfspace (Grätz Figure 3. Comparison of measured, approximated and
1999). theoretical deflection bowl.

692
positive aspect of this regression approach is its linked with the stiffness of the plate and implies
link to mechanical properties of the pavement. information about the layer modulus of the bound
Both regression parameters (A and B) are con- layers and the total thickness of the bound layer—
nected either to the layer modulus of the halfspace the second dominates the characteristic stiffness
M0 or to the elastic length l: of the load-distributing layer M1h3, as it is linked
with the third power.
1
l= (4)
B
2.2 Jendia’s geometry based evaluation method
where l = elastic length [mm], B = regression This approach, developed by Jendia (1995) and
parameter [mm−1]. refined by Roos et al. (2008) analyses the bearing
Q capacity of roads and airfields by taking a look at
M0 = (5) the geometrical properties of the deflection bowls.
A l
The bearing capacity of the bound layers is being
where M0 = layer modulus of the halfspace [MPa], described by the radius of curvature at the load cen-
Q = impact force [N], A = regression parameter tre R0:
[mm], l = elastic length [mm].
1
The layer modulus is a mechanical parameter w′′′′ ( = )= (10)
of the regarded layer. It is derived by the Young’s R0
modulus, adjusted by its Poisson’s ratio. This
adjustment is necessary, because of the voluminous where w’’(x) = curvature of the deflection bowl at x
form of the pavement or other layers and therefore [mm−1], R0 = radius of curvature at the load centre
these induced effects cannot be neglected: [mm].
Ei Because of the fact that the deflection bowl can-
Mi = (6) not be described functionally by using the discrete
1 − υi2 deflection values gathered by the FWD, the radius
of the curvature R0 is determined regressively by
where Mi = layer modulus of the ith layer [MPa],
using the deflection at x = 0 mm and x = 210 mm.
Ei = Young’s modulus of the ith layer [MPa],
It becomes clear that with an increasing bearing
νi = Poisson’s ratio of the ith layer [–].
capacity of the bound layers, the radius of curvature
Taking a look at the mechanical meaning of the
R0 is comparably high and vice versa, i.e. the bearing
elastic length l, it becomes evident that the elastic
capacity of the bound layers can be quantified with
length l is only a moderating factor between the
the radius of curvature at the load centre R0.
stiffness of the superstructure and the stiffness of
the substructure (Bald et al. 2011, Hake et al. 2007):
24 494 (w0 − w210 )
−0 ,899
R0 (11)
2⋅D
l= 3 (7) where R0 = radius of curvature at the load centre
M0 [m], w0 = deflection at x = 0 mm [mm], w210 = deflec-
M1h3 tion at x = 210 mm [mm].
D= (8) The deflection right underneath the load plate is
12
highly depending on the bearing capacity of the sub-
where l = elastic length [mm], D = stiffness of the soil (Roos et al. 2008), as the stiffness of the subsoil is
plate [MNm], M0 = layer modulus of the halfspace very low, compared to the stiffness of the bound lay-
[MPa], M1 = layer modulus of the load-distributing ers above. Therefore the deformations derive from
layer [MPa], h = thickness of the load-distributing
layer [mm].

M1h3
⇒ l= 3
6 ⋅ M0 (9)
⇒ M1h3 = 6 ⋅ M 0 ⋅ l 3

where M1h3 = characteristic stiffness of the load-

distributing layer [MNm].
The characteristic stiffness of the load-distrib-
uting layer M1h3 is strictly corresponding to the Figure 4. Graphical representation of the crucial
mechanical properties of the superstructure. It is parameters introduced by Jendia.

693
 of a n-layered body. It is important to emphasize
“numerically”, as the presence of a plurality of
layers leads to the necessity of solving differential
II I equations of a higher order, which is only possible
with the help of computers.
The multi-layer theory demands the following
conditions and is subject to the following simplifi-
cations (FGSV 2011b):
III N • Road structure consists of horizontal layers,
whose thicknesses remain constant
• Materials are linear-elastic, isotrop and their
properties are homogeneous
Figure 5. General assessment diagram.
• Bottom layer is of infinite thickness
• Layers are infinite in the horizontal direction
Table 1. Definition of the four classes of bearing
capacity. • Layers have no mass
• Loads are being induced as a “pot load”
Sector Definition • Loads act statically

I Bound and unbound layers are rather stiff—

The following values are necessary to perform the
bearing capacity is comparatively high. backcalculation in order to calculate the Young’s
II Bearing capacity of the unbound layers is moduli: thickness t, Poisson’s ratio μ, temperature
rather low and the stiffness of the bound T, impact force Q. The last two parameters can be
layers is relatively high. obtained during the measurement itself and are
III Both, the bound and the unbound layers are being used during the post-processing. The thick-
relatively soft. The structure as a whole has ness t of the respective layers have to be obtained
a low bearing capacity. via core drills or GPR, whereas the Poisson’s ratio
IV Bearing capacity of the unbound layers is μ for each layer can be assumed.
rather high, but the bound layers are The process of backcalculation of the Young’s
relatively soft. moduli follows the principle of a forward calcula-
tion: a combination of Young’s moduli are being
the unbound layers. If one forms the ratio between picked by the program (this can be interpreted as an
the radius of the curvature R0 and the deflection at educated guess—boundaries are therefore needed
the load centre w0, this ratio provides an informa- in order to prevent unrealistic values). With this
tion about the bearing capacity of the structure as combination of initial values the program calcu-
a whole. Jendia (1995) defined this parameter and lates a deflection bowl and compares this calculated
called it load classification number Tz. deflection bowl with the actual one. If the deviation
is too high, then the program iterates the Young’s
R0 moduli and calculates a new virtual deflection bowl.
Tz = (12)
Once the desired deviation has been reached (has to
w0

where Tz = load classification number [-],

R0 = radius of the curvature at the load centre
[mm], w0 = deflection at x = 0 mm [mm].
In order to quantify the bearing condition of
the subsoil, the geometrical approach refers to the
layer modulus of the halfspace M0.
The assessment of the examined pavement is "1
being done with the help of an assessment dia-
gram (see Figure 5). This diagram has four sectors,
which characterise four different states of quality
of the bearing capacity. The boundaries depend
on the load class of the examined road. The four
classes can be defined as follows:

2.3 Multi-layer theory "'

With the help of the multi-layer theory it is pos- Figure 6. General mechanic model of the multi-layer
sible to calculate numerically stresses and strains theory (Riedl et al. 2009).

694
defined at the beginning of this iterative process),
= j ··~~
the calculation is successfully concluded.
--
3 METHODS AND DATA

The following part explains the overall approach

_j
~mm
Frt101. Bionk.<
of the on-site-measurements and the suitability of ~-

the data gathered. Therefore, this part is divided

twice into two parts: description of the approach
and the description of the relevant data.

Figure 8. Construction design of the examined pave-

3.1 Methods ments (a. BAB 2, b. K 86, c. unnamed rural road).
3.1.1 Experimental setup
The experimental setup is fairly simple: for each Table 2. Traffic volumes of the observed roads.
of the three measuring sites, FWD measurements
were performed with a measuring grid width of Road Traffic volume [vehicles per day]
300 m (for BAB 2) and 50 m (K 86, rural road).
Each measurement was made twice, in order to BAB 2 85,000 (20% commercial vehicles)
identify outliers directly. K 86 12,000 (14% commercial vehicles)
The arrangement of the geophones was the Rural road 60 (≈ 50% commercial vehicles)
following: 0 mm–210 mm–300 mm–600 mm–900 (Apelbeck
mm–1.200 mm–1.500 mm–1.800 mm–2.100 mm. to Moisburg)
The target value of the impact force was 50 kN and
the reference temperature was 20°C.
Marienborn), the medium stressed road is the K 86,
3.1.2 Measuring sites located close to Harburg and the less stressed road
The measuring sites chosen are located in Lower is a rural road between Apelbeck and Moisburg.
Saxony in Germany. The highly stressed road is The following images shall give further informa-
the BAB 2 (investigated route is: from Hannover to tions about the location of the measuring sites, as
well as the construction design of the pavements.
It is important to note that the construction
design data has not been gathered with the help of
field surveys. Drilling cores was not possible and
GPR is not widespread in Germany. Therefore the
construction design information is based on infor-
mations provided by the local building authorities.

3.2 Data
The deflections bowls measured by the FWD are
being used to calculate the various parameters of
the German two-layer evaluation methods—layer
modulus of the unbound layer M0, elastic length
l, characteristic stiffness of the load distributing
layer M1h3, load classification number Tz.
Concerning the multi-layer theory, it is not
recommendable to assume a two-layer system,
because of the differing material properties of
asphalt layer, frost protection layer and half-
space. The backcalculation was performed with
EVERCALC and its outputs are the Young’s
moduli of the three layers. In order to perform this
backcalculation it is important to have informa-
tion about the individual layer thicknesses and the
Figure 7. Location of the measuring sites (Wikimedia individual Poisson’s ratios, which can be assumed
Commons 2008). with empirical values:

695
• μ1 = 0,35 (asphalt) 12
d ' '
• μ2 = 0,35 (unbound material) 11 '
--------,--------
'
'
T-------
'
,'- - - - - - - -
'
• μ3 = 0,50 (halfspace) 10 ' ' '
--0 -----:--------~-------~------ --
, ' '
9 ----- ------- -·--------
' ' '
.,_-------
~-------

In order to ensure comparability of the data, it is '

'
'
'
''
' ' '
important to adjust the raw data to reference values
(temperature T, impact force Q). The adjustment of ~~ ~ ~~ ~ ~ ~~ ~ ~ ~ ~ ~ ~ ~~I~~~~~~ ~ J~ ~ ~ ~ ~ ~ ~ ~
0
0
the deflection bowl to the reference impact force Q ----- .0.
is standardised in Germany (FGSV 2014), whereas
the adjustment to a reference temperature T is not
standardised. For this purpose, the approach of
Schulte (1984) is being used, as the adjustment of
I I I I
I I I I
---- ~------- ~------- -~-------"------ .

the deflection bowl is not only normalised to the -- ---- ?_:_- -- ---- ~----- --~- -- ----- ~--- --- -~ ------ -
reference temperature T, but also depending on I
I
I
I
I
I
I
I
I
I

the thickness of the asphalt layer. After these two 200 400 600 800 1 000 1 200

stages of normalisation of the deflection bowl, the Layer Modulu s MO [MPa]

backcalculation can be performed.

For these three case studies, a three-layer-system Figure 10. Assessment diagram for BAB 2 (from west
to east) – Jendia’s geometry based method.
is assumed, which means that three different Young’s
moduli are being backcalculated from the deflection
data: E1 (asphalt layer), E2 (subbase), E3 (subgrade).
Initial moduli and moduli range are determined
after WSDOT (2005). As the backcalculation is an
iterative process, there is always a difference between
measured and backcalculated deflection bowl. The
RMS is defined as a tolerance criterion for the esti-
mation of quality. For the sake of backcalculation,
the RMS tolerance is fixed to 1%. : '-;
i~i --,~ 600

--: ~ ~ ~:-
4 RESULTS AND DISCUSSION :t 400

In this part, the results of the on-site-measurements Kilometrage [lorn]

are being presented and critically analysed, sum-

marising the most important conclusions. With Figure 11. Backcalculated Young’s moduli (BAB 2 –
these conclusions formulated, the focus shall turn from west to east; E1: blue, E2: red, E3: green).
to interactions, links, pros and cons of the meth-
ods of evaluation, in order to stipulate the lessons 4.1 Results – BAB 2
learnt from these methods.
The main “problem”, concerning the examination
of the BAB 2 is that the area examined is very vast.
'1
Therefore, the construction design is rather inho-
mogeneous, especially if one bears in mind that
6 000 ----- - --------·--------
there can be many patches that modify the mechan-
5 000 - - -- - -
ical response to the impulse load of the FWD.
Another factor that has to be kept in mind is that
~
4 000 ----- --- --- --.- ------ - there are many installations or partially hydrauli-
a: cally bonded layers underneath the asphalt layers
3 000 --- _Q -
________<!) ______ _ that cannot be seen. That may also lead to a differ-
Q ent mechanical response of the pavement structure.
Oo
--------,-------- ------, This is why there are a few measuring points, where
oo
0 M1h3, M0, Tz and the Young’s moduli (E1, E2 and
1 000 - - (3 - - - - ~---- - - - - E3) are (unrealistically) high. This is also the rea-
p 0
son, why the backcalculated Young’s moduli often
o +=~~~~~~~~=+==~;=~=;~= reach the upper moduli border—either the pave-
0 200 400 600 800 1 000 1 200
layer Modulus MO (M Pa]
ment structure has a rather stiff layer within (e.g.
hydraulically bonded layer or an old concrete layer)
Figure 9. Assessment diagram for BAB 2 (from west to or the assumed/measured thicknesses differ at these
east) – Darmstadt evaluation method. measuring points or a combination of both.

696
 The mechanical reaction of the pavement structure 4.2 Results – K 86
itself is, of course, heterogeneous. There are areas,
Regarding the evaluation of the K 86, the results
where the pavement structure as a whole is rather
seem plausible. The moduli boundaries are not
stiff, as well as there are areas, where the pavement
reached and the German assessment methods do
structure is weak (in relation to the load endured).
not show any excessive values.

~ 160 ----r ----r----

~
:i
~
140

0
-~--i----i---- :::::::;::::::::::::::::::: :::~:: :::::::::: : :c:::: :
j

~
120

100
~~--~---+~
-~
~-~------~P_-_----~~~------~--~~--~--_._
____ ,. ____ ,. ____
Go
----~----~----~----!----!----!----!----~ ?.
! o e! ~ ! ! ! ! !
:
:·::-;:::::::: r ::;:::r:: r::: _
£5 <
50 --- - ---~--- - ---:---- - ---:------- i--_____ 1______ _1____ --- --
:: ::~ :::: ~: ::: ::: ~f ~~ : f;:~: f:::~~: ::~ f::: :f: ::: f: :::f::
0
-------:--------:--------:--------:-------1-------1------- - ·
: : 0: : : :
---- •---- •---- ---- ~ ~- o- ~---- ~---- ~---- ~---- ~---- j---- j--
3o ------ -~-------!- ------ i------- ~- -o ---- ~------- ~------- --

:::::::t:::::L::~j~:~t~o~~:~:::: -,
o, ' 0 ' ' ' ' ' ' '

----•----•---- ~- - -t--- - t----t----t----t - - - -t--- -t----1--

: :o o~o~G' o9 :o 0
: 0
:o
l ayerModu lusMO{MPa)
LayerModulusMO[MPa)

Figure 12. Assessment diagram for K 86 – Darmstadt

Figure 15. Assessment diagram for the unnamed rural
evaluation method.
road – Darmstadt evaluation method.

3 ---- ----

____ c ____ c ____

---- ---- ---- ---- ----

----r----r----r----t----t----t----j----je-
---- ---- ---- ----- -

,- II -::-- :-_--:-_--: 1
~ f--- -1o--- ~---- ~ ----~ ----j----j-
: : : : : : 0 :
2 -- -- r---- r ---- - -- - - -- - <

' ' 0 00 : ~@ : &: : o: : : :

. :::::::,::::::I::::::;~ :::::r: : : L: : :, :::::::.:
:::r:::l:::>j;:TI T::;: ::;:::r::;: ------ -~------ -:--------:------- ~- -o ---- ~--oo.o - ~- .n ---- ~ -~
: : o~ ' o
__ ___ _- '~ __ ____ -:-' -oce -,.:; 0
~
:omio-____
0 '
0

_S~ - - _____6!;_______ +-: _

o(:}
Q§) ' 0 ' '

" ; oco ~ ~o :
0 0 0
: : : :

250
Layer Modulus MO [MPa) Layer Modulus MO (MPa)

Figure 13. Assessment diagram for K 86 – Jendia’s Figure 16. Assessment diagram for the unnamed rural
geometry based method. road – Jendia’s geometry based method.

T r-----,-----------, ~

---- ---·-- --- -- ,_ ---

-------r-------

-------1--------

-------'--------

-------r-------

~~~~~~~~-+~~~~~~-r~+-~ 0
1400 1 600 1800 2000 2200
Kilometrage[m]

Figure 14. Backcalculated Young’s moduli (K 86 – from Figure 17. Backcalculated Young’s moduli (unnamed
west to east; E1: blue, E2: red, E3: green). rural road – from west to east; E1: blue, E2: red, E3: green).

697
 The pavement structure itself is rather weak, deflection bowl (deformation of the pavement struc-
especially the asphalt layers react very strongly to ture) and assumes that the curvature at the load cen-
the impulse load of the FWD. The backcalcula- tre is mostly influenced by the quality of the asphalt
tion of the Young’s modulus of the asphalt layer layer. The Young’s modulus of the asphalt layer E1
supports this finding—M1h3, Tz and the Young’s is also being (back) calculated with the help of the
moduli are rather low (third quadrant of the geometric form of the deflection bowl (i.e. the defor-
German assessment method). mation). Therefore, one can conclude that there is
also a causality. But still, the radius of the curvature
at the load centre R0 is a very volatile parameter
4.3 Results—unnamed rural road (Apelbeck to
that does not enable a reliable statement about the
Moisburg)
mechanical quality of the asphalt layer. Therefore,
The rural road is a scarcely used road. The asphalt this correlation and causality is not expedient.
layer functions as a sealing of the unbound base Concerning the quality of the underlying struc-
layers underneath. Therefore, the thickness of tures (subbase and subgrade), a correlation and
the asphalt layer should be rather small. Unfor- causality between the two-layer-model and the
tunately, the local road construction office had three-layer-model and their evaluation methods
no records about the actual construction design cannot be found. The main problem is the distri-
of the rural road, so that the construction design bution of the shares of the bearing capacity within
is an assumption, according to the traffic load the three-layer-model. The backcalculation fits the
and the German design guidelines concerning Young’s moduli of these two layers according to
rural roads. Looking especially at the backcal- the deflection bowl (in consideration of the thick-
culated Young’s moduli, it becomes clear that ness of the subbase), whereas the layer modulus M0
the thickness assumptions cannot be correct, as is based solely on the geometric properties of the
the moduli very often reach the upper boundary outer deflection bowl. These assumptions make a
of the individual modulus range. A local thick- causality between these two methods impossible
ness surplus can explain this fact. The German and also not expedient.
assessment diagrams support this theory, as the
characteristic stiffness of the load distributing-
layer M1h3 is relatively high in relation to the load 5 CONCLUSION
class of the road. Drill cores would bring further
enlightenment into this problem. The purpose of this paper is to give an insight into
different structure models and their individual evalu-
ation methods, based on FWD measurements. What
4.4 Comparison of the methods
becomes evident is that the two-layer-based model is
If one excludes the abovementioned restrictions, less dependent on the correctness of structural data
one can see that the evaluation methods show dif- (thicknesses, specific layer characterisations). The
ferent results for the quality of the pavement struc- output is therefore only limited to a general state-
ture. The two German evaluation methods show ment about the structural quality of the pavement.
quite similar results, only the limits concerning the The backcalculation method on the other hand is
quality of the superstructure differ in such a way going into more detail, but is more prone to errors
that Jendia’s evaluation method is stricter and leads that derive from the volatility of the given infor-
to higher failure rates. A linear relation between mation about the construction design. Of course,
these two evaluation methods (regarding the evalu- GPR data can decrease the amount of uncertainty
ation of the asphalt layer) delivers usually a coef- about the actual construction design, but still, even
ficient of determination of R2 > 0.8, which means GPR data is prone to uncertainty and errors and
that there is a connection. This is not a surprise, can lead to a false sense of accuracy about the given
because both methods rely on the observation of pavement. In general, the same applies for the back-
the actual deflection bowl and a two-layer-model. calculation method and the differentiation of the
A linear relation between the backcalculated pavement into three or more layers. Of course, the
Young’s moduli and the parameters of the German results seem to deliver a precise insight into the struc-
evaluation methods remains inconsistent. Only the tural quality of the pavement, but the statements are
radius of the curvature at the load centre R0 and very much depending on the quality of the construc-
the Young’s modulus of the asphalt layer E1 seem to tion data. As pointed out with the case studies (espe-
have a connection (R2 ≈ 0.7), the question is, whether cially: unnamed rural road), if the given data of the
there is a causality as well. To answer this question, construction design is wrong or unknown, a back-
one hast to look at the concepts of these two param- calculation of Young’s moduli and an assessment of
eters—the radius of the curvature at the load centre the pavement structure is merely a poking around in
R0 derives from a geometry-motivated view on the the dark.

698
 So, the overall statement is: of course, a multi- stheoretischer Berechnungen. Mitteilungen aus dem
layered model is, scientifically speaking, a progress Institut für Materialprüfung und Forschung des Bauwe-
concerning the certainty and precision of the eval- sens der Technischen Hochschule Hannover (Heft 6).
uation of pavement structures, but the amount of Hannover.
Grätz, B. (1999). Einfluß der Temperatur, der Belastungs-
uncertainty about the actual construction design frequenz und der Impulskraft beim Falling-Weight-
impedes such a precise statement. A two-layer- Deflectometer (FWD) auf die Größe der effektiven
based evaluation method is therefore not a step Schicht-E-Moduli (FE-Nr. 04.174). Schlußbericht,
back, but a concession to uncertainty. Technische Universität Darmstadt, Fachgebiet
Straßenwesen, Darmstadt.
Grätz, B. (2009). Bewertung der Tragfähigkeit und der
REFERENCES Standfestigkeit bestehender Asphaltstraßen zur Wahl
zweckmäßiger Erhaltungsmaßnahmearten. Habilita-
Bald, J.S., Böhm, S., Nguyen, A.-D., & Schwebel, N. tionsschrift, Technische Universität Damrstadt, Fach-
(2012). Optimierung von Maßnahmen zur Erhaltung gebiet Straßenwesen, Darmstadt.
einer funktionstüchtigen Straßeninfrastruktur—Ermit- Hake, E., & Meskouris, K. (2007). Statik der Flächentrag-
tlung des Substanzwertes. Schlussbericht zu FE-Nr. werke. Einführung mit vielen durchgerechneten Beispie-
88.0100/2009, Darmstadt. len (2., korrigierte Auflage Ausg.). Berlin: Springer
Bohn, A.O. (1989). The History of the Falling Weight Verlag.
Deflectometer (FWD). Abgerufen am 27. Juni 2016 Jendia, S. (1995). Bewertung der Tragfähigkeit von bitu-
von SWECO Danmark A/S: http://www.pavement- minösen Straßenbefestigungen. Veröffentlichungen des
consultants.com/media/6042/HistoryOfFWD_ Institutes für Straßen- und Eisenbahnwesen der Univer-
AxelOBohn.pdf sität Karlsruhe (Heft 45). Karlsruhe.
Deutsche Vereinigung für Wasserwirtschaft, Abwasser Kaakkurivaara, T., Vuorimies, N., Kolisoja, P., & Uusi-
und Abfall e.V. (2005). Arbeitsblatt DWA-A 904. Rich- talo, J. (2015). Applicability of portable tools in
tlinien für den ländlichen Wegebau. Hennef. assessing the bearing capacity of forest roads. Silva
Forschungsgesellschaft für Straßen- und Verkehrswesen. Fennica (49).
(2008). Arbeitspapier Tragfähigkeit von Straßenbe- Riedl, S., Grätz, B., Hothan, J., & Pauli, S. (2009). Über
festigungen. Teil B 2.1: Falling Weight Deflectometer die Mehrdeutigkeit bei der Rückrechnung von Schich-
(FWD): Gerätebeschreibung, Messdurchführung— tsteifigkeiten aus Einsenkungsmessungen. Straße und
Asphaltbauweisen. Köln: FGSV Verlag. Autobahn (8), S. 499–505.
Forschungsgesellschaft für Straßen- und Verkehr- Roos, R., Freund, H.-J., & Thiele, T. (2008). Erarbeitung
swesen. (2011a). Arbeitspapiere zur Dimensionierung eines Bewertungshintergrundes für Tragfähigkeits-
von Verkehrsflächenbefestigungen. Reihe B: Berech- messungen auf Basis von Zustandsindikatoren nach
nungsmodelle. Teil B 1: Methode der Mehrschichten- JENDIA. Forschung Straßenbau und Straßenverke-
theorie und Halbraumtheorie bei der Bemessung. Köln: hrstechnik (Heft 989). Bonn.
FGSV Verlag. Washington State Department of Transportation. (2005).
Forschungsgesellschaft für Straßen- und Verkehr- EVERSERIES User’s Guide. Pavement Analysis
swesen. (2011b). Arbeitspapiere zur Dimensionierung Computer Software and Case Studies. Olympia, WA.
von Verkehrsflächenbefestigungen. Reihe B: Berech- Wikimedia Commons. (18th. March 2008). Karte Deut-
nungsmodelle. Teil B 2: Methode der Plattentheorie bei schland. (D. Liuzzo, Produzent) Abgerufen am 29th.
der Dimensionierung. Köln: FGSV Verlag. September 2016 von Wikimedia Commons: https://
Forschungsgesellschaft für Straßen- und Verkehrswesen. commons.wikimedia.org/wiki/File:Karte_Deutsch-
(2014). Arbeitspapier Tragfähigkeit von Verkehrs- land.svg
flächenbefestigungen. Teil C 2.1: Falling Weight Zhu, Z., Wang, B., & Guo, D. (1985).
Deflectometer (FWD): Auswertung und Bewertung— —Mechanical Design of Pavements (in Chinese).
Asphaltbauweise (AP Trag Teil C 2.1). Peking: China Communications Press.
Gerlach, A. (1968). Zur Beurteilung der Standfestigkeit
von Fahrbahnaufbauten mit Hilfe statisch-elastizität-

699
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

A layered pavement structural and remaining capacity model by analysis

of FWD and TSD data

J.D. Roberts
ARRB Group (Australian Road Research Board), Melbourne, Victoria, Australia

ABSTRACT: Although pavements are built on structural principles, maintenance is usually triggered
by surface condition. Readily available structural indicators would allow selection of more appropriate
cost effective treatments. The challenge was to deliver a low cost reliable structural indicator to satisfy the
above need. The objective was to achieve this reliability by FWD data, and provide consistent outcomes
regardless of date of survey. Low cost, and linear intensity, was to be achieved by porting of the model
to TSD data. The Paper describes an innovative back-analysis of FWD data, delivering a layered model
by pavement type, seasonally adjusted. Remaining capacity estimation follows, also by pavement type. An
example is shown of the similarity between FWD and TSD based outcomes.

Keywords: TSD, FWD, SNP, Pavement remaining life

1 INTRODUCTION The research was first undertaken in Greece,

Roberts et al (2003), using 150 cored sites, then
The use of deflection (FWD) data to assess the expanded to over 1,500 sites throughout NSW. These
structural status of a pavement, for asset manage- covered the full matrix of pavement type, condition,
ment purposes, is reasonably well established. Con- traffic level, subgrade and climate, Roberts et al
ventional approaches are based on the value of the (2010). Each had FWD, was cored to 1.1 m at exactly
central deflection, d0, from which estimates are made d0, inspected by NSW State officers for structural
for remaining capacity, COLTO (1997), Austroads life prognoses, tested for surface condition and traf-
(2011). Other methods use an estimate of SNP as fic surveyed. All sites were analysed by FEM based
input to predicted road Roughness, ISOHDM (2004). models for material strength and pavement capacity.
Deflection is also used as part of detailed design
for strengthening through mechanistic back-analy-
ses. These methods need skilled staff, are relatively 2 STRUCTURAL DESCRIPTORS
slow and need prior coring to determine layers and
material types. To accelerate the work, GPR has 2.1 Use of structural number
been used, but this also needs coring for validation.
Returning to an asset management context, fur- The aim of the work described in this Paper is the
ther work has been done for the more careful inter- development of a multi-layered pavement model,
pretation of full bowl data, Horak (2008). It had been for which the primary input is FWD data.
possible to establish a link between the values of the For description of the model the Structural
central, mid-range and tail of the bowl, and an indi- Number system was used, as shown in Equation 1:
cation of weakness (or strength) in respectively, the
i n −1
base, mid-depth layers, and the subgrade. This infor- SNP
N ∑ i =1
(ai hi ) SNs
S g (1)
mation is helpful, may focus further investigation.
The work described in this Paper builds on the
where SNP = Structural Number of Pavement for
knowledge base to deliver a new deflection based
n layers, including the subgrade term SNsg, hi is the
model that satisfies assets needs from planning
layers thickness (ins), and ai the strength param-
through to pre-design. Model per test site includes:
eter, material coefficient. Further, the product aihi
– 4 layers; surface, base, subbase and subgrade can be represented by a term SNi, which is the
– Material type per layer & overall pavement type structural contribution by layer i. The subgrade
– Thickness of each layer contribution, SNsg, is discussed later. ai is used to
– Estimated material strength of each layer describe the strengths of all material types, from
– Estimated pavement remaining capacity; ESAs. soft subgrades to strong bound materials.

701
 Although all primary modelling will be under- In Figure 1 the new relationship has been
taken using ai, data inputs and reporting will be extended below E values < 1000 MPa, based on
as CBR for softer unbound granular materials and relationships by the same authors for unbound
Modulus (MPa) for bound materials. materials, as highlighted in the hatched rectangle.
The parameter, ai, has the special property of a As such the relationship offers a single solution for
linear relationship with thickness. This means that a range of materials, bound and unbound.
within each SNi term, the components, ai and hi, can For softer unbound materials, a separate rela-
be cross substituted so long as SNi is unchanged. tionship between a and CBR was developed.
This simplifies what would otherwise be a complex Numerous sources were studied for guidance, in
non-linear problem. However, the non-linearity of particular the large amount of information con-
the underlying mechanisms is still retained through solidated for HDM-4, ISOHDM (2004). A new
the transformation relationships between ai and the integrated relationship was developed. The form
parameters CBRi and Ei, as described below. was chosen to encompass the trend of the bulk
The validity of these properties, and of SNi as of individual material specific and ‘range limited’
a measure of pavement structure, is supported by published relationships. Further, the form was
comparison with detailed mechanistic analyses. mathematical pure to allow for perfect inversion.
These show that strain on first loading is similarly The primary relationship is shown in Figure 2, and
related to SNi, regardless of values of ai or hi, an defined by Equation 4, and the inverted relation-
interesting confirmation of the original empirical ship is defined by Equation 5.
SN concept by contemporary analyses.
SNP, and its components, are thus suitable CBR ( ea ) − 4… (4)
parameters for asset management purposes.

a=
{LN ⎡⎣(CBR + ) ⎤⎦} (5)
2.2 Material strength transformations 18.86
Considering material strength transformation rela-
tionships between a and E, two well established The applied zone for the relationship in Figure 2
relationships for asphaltic materials were consid- is for a range up to an extreme maximum of a = 0.2.
ered Ulliditz (1987) and Ioannides et al, (1986). To put this into perspective relative to bound mate-
These, although both developed for a as a function rials and show continuity between them, this same
of E, have been re-plotted in Figure 1 as the inverse zone is as boxed out in Figure 1.
to suit project needs. A new integrated relationship The relationships between a & E and a & CBR
based on these lines, is also shown, and is defined by respectively, were developed independently. How-
Equation 2, and Equation 3, its perfect mathemati- ever, to close the loop, and using a as the common
cal inverse (both required for inputs and outputs). link, the resulting relationship between E & CBR
was examined, and is shown plotted in Figure 3.
E ( ea ) − 270… (2) Referring to Figure 3, the plotted line is a good
cross validation, in agreement with common expe-

a=
{LN ⎡⎣(E + ) ⎤⎦} (3)
rience, and sits well mid-way between a number of
other reported relationships. The reasonableness of
5.564

16,000

y
15,000 200
14,000
180
13,000
12,000 160
-= 11 ,000
140
~ 10,000
e. 9,000 120
..2 8,000 ~ 100
= 7,000
'C 1:1:
~ 6,000
CQ
u
80

"" 5,000
4,000
3,000
2,000
1,000 I - - - - - - - - - - -
Oc: __________ _
,,

0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75
Material Coefficient (a)
60

40

20

0
0.00 O.o2
=/ J
0.04 0.06 0.08 0.10 0.12 0.14

Material Coefficient (a)

0.16 0.18 0.20

Figure 1. Modulus v Material coefficient bound Figure 2. CBR v Material coefficient unbound
relationship. relationship.

702
 600,-------------~~~~~~~~~------, 15 or~-----------------------------,
550 +- -+ ... + 14
500 13 • Full Depth Asphalt
x Spray Seal Stabilised
450 12
+ Spray Seal Unbound
400
' [ 350
i: r + ~ 11
10
• Thin Asphalt Stabilised
• Thin Asphalt Unbound
! 300
~~ 9
+
~ 8
;
~
250
200
z 7 "'
~ 00 6 y = 167x-0·57
::; ISO
100
5 R 2 = 0. 77
50
4
3
10 20 30 40 50 60 70 80 90 100 11 0 120 130 140 ISO 160 170 180 190 200 2
CBR (%) 1
0 +----r--~---.----r---,----r--~~
0 400 800 1200 1600 2000 2400 2800
Figure 3. Modulus v CBR unbound relationship.
FWD dO (microns at 700KPa)

this relationship supports the validity of the rela-

tionship in Figure 1 for values of a below 0.2. Thus, Figure 4. RTA Roberts Loizos method for SNP from
FWD d0.
these material transformation relationships form a
consistent and integrated system, applicable for most
flexible materials. This is essential for this work, but successfully to rural asphaltic pavements in Greece,
also valuable for a more general application. Loizos et al (2002). The relationship as determined
in this Paper is termed the RTA Roberts-Loizos
method. The estimated values of SNP are compat-
2.3 Derivation of SNP from FWD deflection ible with HDM-4 interim approaches, ISOHDM
The next stage is the determination of a relationship (2004), and the Salt method, Salt (1999). There are
for the estimation of SNP, based only on FWD data. benefits, beyond the current work, of a supported
The NSW data set of some 1,500 test sites described simple universal relationship for all flexible pave-
above was examined. For each test point, the match- ment types.
ing drilled core and FWD deflection point results
were collated. The data was back analysed, using an 2.4 Consideration of pavement stiffness
FEM based procedure, and a value for E assigned to
each cored layer down to a depth of 1.1 m. The stiffness of a pavement, as distinct from its
The layer thicknesses expressed in inches and strength, has been understood by pavement analysts
material strengths transformed from E to a, and as an indicator of the performance of the upper lay-
the SNP derived using Equation 1. SNsg term ers of the structure. Usually this relates to structural
in Equation 1 was determined simply by treating crack initiation in asphaltic layers. The indicator often
this layer just like all the others with no special used is the quantity Curvature, which is the arithmetic
subgrade formula. The derived SNP values were difference between the d0 deflection the value offset
plotted versus the d0 FWD central deflection, as 200 mm, d200, Austroads (2011). In this Paper it is pro-
shown in Figure 4. posed that the concept of stiffness be extended as an
The relationship has an r2 of 0.77. It was decided indicator for any flexible pavement structure, includ-
that the accuracy was sufficient for the purposes of ing cemented and even unbound spray seal pave-
the models. The form is defined by Equation 6. ments, with a strong crushed rock base.
It is also proposed, Loizos et al (2002), that the
167 quantity Curvature be expressed as a new param-
N =
SNP (6) eter termed Radius of Curvature, or ROC. ROC is
d0 0.57
the radius of the curve fitted to pass through the d0
and d200 deflection points, and can be thought of
where, d0 is the FWD central deflection normalised
as the radius of the seat of the deflection bowl. A
to 700 kpa. The derived relationship is applicable
larger ROC, equivalent to a smaller quantity Cur-
for a wide range of pavement types, ranging from
vature, indicates a stiffer pavement. The parameter
heavy duty thick asphaltic to light duty rural spray
is defined through simple geometry by Equation 7.
sealed unbound granular. The data was collected
on the network of the RTA (NSW Roads and Traf-
200
fic Authority: currently known as Roads and Mar- Cm =
ROC (7)
itime Services). The form was previously applied 2 × (d d )

703
where ROCm has units of metres and the It is suggested that any point on this line rep-
FWD data at d0 and d200, in microns, normalised resent successive pavements where a balanced and
to 700 kpa. parallel increase in strength is occurring in both
A pavement can of course be stiff, with a high the upper and lower portions of the structure, as
value of ROCm, just because it is strong. The abso- if ‘balanced’.
lute stiffness of a pavement, as measured by ROCm, If for any value of SNP, the ROCm test point
will increase either if the pavement is stronger plots above the ROCe line, then that pavement
(larger SNP), or if the pavement structure, with the structure has a makeup which includes a propor-
same SNP, is configured such that it is stiffer in the tionately greater amount of pavement strength
upper layers, or a combination of both effects. This within the upper layers compared to the bal-
concept is illustrated in Figure 5, where a number anced case. Similarly, if the ROCm test point plots
of actual test point ROCm results are plotted versus below the line, then there is proportionately less
their SNP values. A few of these have their ROCm strengthen within the upper pavement layers com-
marked in for illustration. Of these, two test points, pared to the pavements on the ROCe line.
with very different SNP values, have the same Thus, for any value of SNP, taking ROCm as a
ROCm. Conversely, two other points have the same measure of total stiffness, (due to strength plus that
SNP, but very different ROCm values. due to pavement makeup), then ROCe is a measure
An approach is proposed which permits all data of stiffness due to strength alone.
test points to be understood and used in analyses, Figure 7 shows two real example deflection
however extreme. The procedure aims to describe bowls, plotted, for two different pavements, both
pavement stiffness in a way which allowed any with the same d0 and SNP, but quite different
stiffness effects due to configuration to be isolated ROCm values. The ROCe, a function of SNP is the
from the more general strength effects. same for both cases, and is also plotted.
Figure 6 shows the same data as plotted as It is proposed that a new parameter is cre-
Figure 5, but this time a simple power regression ated to define pavement stiffness independent of
line through the data has been included. The power strength, by describing and quantifying total stiff-
function used is defined by Equation 8: ness, ROCm, relative to strength stiffness, ROCe, as
Equation 9:
ROCe = 8.41 × SNP
N 1.87 (8)
ROC
Cm
Note that the values 8.41 and 1.87 are just appli- PSR = (9)
ROC
Ce
cable for the data in this example to illustrate the
concept.
where PSR = Pavement Stiffness Ratio, with a value
of 1.0 for a balanced pavement (ROCm = ROCe),

1500
1400 1500 -r---..,.-----.----,,....-...,..--,----,.-.--.---.---r-;---.---,
1300 1400 +--+--+---1f-+--+-+-t--+-_L___i._~+--l
1200 LV !---"' 1300 +--+--+---1f--+--+-+-t--+--"-
Y,=_8,.4_1_x,l._s7+--J
1100 _/ 1200 +--+--+-f---+--+ ...7-'.···+·o.:.::
···~·· ·t--+--+---'f~t-+--1
""' 1\ 1100 +--+--t-f--f:.-'··:....
···t--+--+--+-t--+--+..-' ....'i.:--1

t 800
•
~
•
1ooo +--+--+-h'-..+·· -t--+--+--+-t--+---+-v--+_
900 -l--+--+~
......
; ~.-+---1f-+--+-+-t--
,~-+.--+~
.
··...-'-;:1
••
/,

":" II i\ • •
~ 700
0 600
I"" • 1.
j
--.... •
!~:::=~==::~==:-= ~·=·=··=· :-~=·-=·-:...:::::~::==:=:=~
e 6oo +--+--+~~~~+··~~·~-1~·-r-~~·-._~-~-~·+--+--+~
v .-.... 1".4! v ....

g~ L -:~ ;t. •• "" i\

~ 500 -1-4-~+-~'-h·~-+-~
~ 400
4!,+~-'~-r.--t-.t--+-~
K..-······~~• •-~-- _.Y ~ •
30
I
1± ,:l ~~~· ~ ..G vI \ ,:. • ~
,: ; ·.. • •#'It..
)(.:_.:::~-..~-H~--t--1---1
~.

200 ll l4t_ L ~.: rY\1 .L .1 300 +--+--tf.........·ti:·.~.J.t.< ... :

... !_ . . -"..! •. .: ...
9 •\J v v
100
0
14" ~ ~ ~ 1:2: f>< I--
8 9
/

1..----'
10 11
v
12 13 14 15
: : t4~;;;~ ,- ·: .. / .•
2 4 5 7 8 9 10 I I 12 13 14 IS
SNP SNP

Figure 5. Plots of ROCm versus SNP, showing some Figure 6. Derived plot of ROCe versus SNP, showing
extremes. gradual increase in ROC as SNP gets larger.

704
 and PSR are known (from the deflection data),
then SNupper can be calculated. In turn, SNlower,
the lower part of pavement strength is given by
Equation 12:
Nlower = SNP
SN N SNupper (12)
//_.-··················.,,\ROCe
The validity of this proposed definition of
°5:et::on~c(~~n) 200 300 r:ets: ~;G
0
::l)SOO 900 1000 llOO 12001300 1400 1500
SNupper (and by extension also SNlower), based just
600
on deflection data, is confirmed in Section 3.5
against data.

3.2 Definition of four layer model

//_........................,\(OCe
It is proposed that the upper and lower portions of
Offset from 40 {mm) Offset from d 0 (mm) the pavement strength, SNupper and SNlower, be fur-
400 3~ 200 100 200 300 }.too SOO 600 700 800 9001000 IIOO 12001300 1400 1500

ROC ther subdivided by strength as follows:

Nupper = SN sf + SNbs
SN (13)
Nlower = SN sb + SN sg
SN (14)
Figure 7. Plots of two different deflection bowls, same
SNP and ROCe, but quite different ROCm. where SNsf, SNbs, SNsb and SNsg, are the layer
strengths in SN units of the surfacing, base, sub-
base and subgrade respectively, as a conventional
and typically ranging from 0.4 to 2.0. ROCm is as 4 layer structure. As the components of each SN
measured and ROCe = Fn (SNP). This means that part, the overall relationship is Equation 15 (like
it is now possible to describe pavements at deflec- Equation 1):
tion test points by both SNP, and relative stiffness,
PSR. SNP
N asf hsf abs hbs asb hsb asg hsg (15)

where a and h for each SN part are the Material

3 FORMATION OF FOUR LAYER MODEL Coefficient and thickness (ins) respectively.
Along with the deflection and associated data
3.1 Division into upper and lower zones (such as temperature and time of testing), the sys-
A PSR value of 1.0 applies to all pavements, what- tem requires readily available key data about surfac-
ever their strength, when ROCm is equal to ROCe. ing type (asphaltic or spray seal) and subgrade type
Further, for such pavements the upper and lower (plastic/clay, draining/sandy, or a mix). Lastly, for
portions of pavement strength are, as discussed asphaltic surfaced pavements only, knowledge of
above, both equal and of value SNP/2. This situ- any possible mineral cemented layer in the base. This
ation is expressed in Equation 10: to assist model to distinguish between an asphalt
surfaced stabilised base, or full depth asphalt.
Cm 0.5 × SNP
ROC N To start the main process for developing the
PSR = 1 = = (10) detailed 4 layer pavement model, an initial estimate
Ce 0.5 × SNP
ROC N
of generic pavement type as a 4 character Pavement
It is proposed that in the same way that, as per Code is required. Each character in sequence rep-
Equation 10, ROCe is proportional to 0.5 × SNP, resents the surfacing, base, subbase and subgrade.
then ROCm is considered proportional to a quantity The available codes per layer are set out in Table 1.
termed SNupper, where SNupper, represents the value of The initial code is simple as just SRGM, or
the strength of the upper portion of the pavement ARGM, for all seal and normal asphaltic surfaced
in units of SN. This is expressed by Equation 11: pavements, and ACGM, if a possible cemented
base.
ROC
Cm SN
Nupper The model will rationally flesh out the details.
PSR = = (11)
Ce 0.5 × SNP
ROC N
3.3 Definition of surfacing
For any particular value of SNP, ROCe is con- For a surfacing code S, the system allocates a h of
stant and proportional to 0.5 × SNP, but SNupper 1 inch (25 mm, double seal), and an a value = 0.2.
will change as ROCm changes. In this way, if SNP This 0.2 value is automatically reduced if cracking

705
Table 1. Character codes per generic pavement layer. The values in Table 2 (user selectable) are typi-
cal for the long-term average rainfall levels in the
Layer Index Description region of interest, but are automatically adjusted
in the pavement layer development model accord-
Surfacing A Asphalt
ing to the rainfall levels in the month of testing,
S Seal (bituminous)
Base G Gravel (graded)
drainage quality, surface condition and level in the
R Rock (crushed) pavement.
L Lime stabilised The system has been designed to cycle progres-
C Cement stabilised sively through a rational set of options, of mate-
A Asphaltic concrete rial code, thickness and material strength, until a
Sub-base N Natural material solution is found, whatever the initial Pavement
G Gravel (graded) Code.
R Rock (crushed)
L Lime stabilised 3.5 Definition of base layer (asphaltic)
C Cement stabilised
Subgrade P Plastic (clay) If the surfacing is asphalt, the system automati-
M Medium (loam) cally engages the asphalt model. First, the pattern
D Draining (sand) of actual measurable values for SNupper are used to
calibrate a model for estimation of SNupper values
from FWD.
For this a number of asphalt pavements in the
Table 2. Material Coefficient (a) ranges per material
type. network of interest are FWD tested, drilled and
back-analysed to measure the asphalt thickness
Index Lower Nominal Upper and E of the material. 20 drilled sites for every
2,000 FWD test points is found to be suitable.
N 0.02 0.08 0.11 For each site the SN value of the asphalt as one
G 0.11 0.13 0.14 layer (surfacing+base) is calculated using trans-
R 0.13 0.15 0.17 formation Equations 2 & 3, and expressed as the
L 0.13 0.15 0.17 actual value of SNupper. Then for each site Equa-
C1* 0.15 0.17 0.2 tion 11 is expressed as Equation 16, and solved
C2* 0.17 0.2 0.25 for ROCe. In other words, the question asked is,
C3* 0.2 0.25 0.4 “assuming Equation 11 is valid, what ROCe value
C4* 0.25 0.4 0.5 is required to deliver the ‘actual’ SNupper”.
C5* 0.4 0.5 0.7

* Respectively Cement Modified, Cement Stabilised,

⎧ 0.5 × SNP
N ⎫
Ce = ⎨
ROC ⎬ × ROC
Cm (16)
Cement Strengthened, Lean Concrete & Concrete
Pavement.
⎩ SN Nupper ⎭

where SNupper is the actual value determined above.

is reported in the condition survey (Section 3.8). The ROCe values per test point are plotted ver-
Thus for code S, SNsf typically = 0.2 × 1.0 = 0.2. sus SNP, and a trend line is determined as a func-
For a surfacing code A, the system allocates a tion of SNP, as Equation 17:
h of 40 mm (user adjustable), and the a value is
determined by the asphalt model (Section 3.5). =C S
SNP n (17)

This relationship for ROCe as a function of

3.4 Definition of base layer (unbound or SNP, is applicable to the FWD sites in the tested
cemented) network.
Equation 17 is now changed and solved for
The value of SNbs is determined as SNupper – SNsf .
SN upper, but using the deflection estimated
The system works on a rational allocation of thick-
ROCe (itself a function of SNP), as shown in
ness and material strength according to a reasonable
Equation 18:
set of thickness limits, and material strength limits.
The thickness limits are automatically allo- ⎧ ROC
Cm ⎫
cated (user adjustable) for the total thickness (not Nupper = ⎨
SN ⎬ × 0.5 × SNP
N (18)
including subgrade) with a minimum for the base ⎩ ROC
Ce ⎭
(typically 150 mm) and subbase (100 mm).
The material strength limits are set by material The resulting deflection based estimates of
code, regardless of layer, in a master table, Table 2: SNupper versus the actual values are plotted, as in

706
Figure 8. The r2 of an equality plot for this exam- where d900 is the FWD deflection at 700 kpa, and
ple was 0.75. Ksg is a factor allowing for calibration against
The fact that the SNupper values could be detected field investigation data. For a network it is usual
and measured, and that the modelled SNupper values to calibrate separately for each subgrade type
matched them closely over a wide range of con- (P, M or D).
ditions, leads to 2 key conclusions applicable for It now remains to express the subgrade CBR as
flexible pavements: a value for SNsg for inclusion within Equation 15.
Conventional approaches for the estimation of
i. SNupper is a ‘real’ quantity measurable in the
SNsg are based on work by TRL as applied to
field.
HDM-4, ISOHDM (2004). The TRL relationship
ii. SNupper can be accurately modelled from FWD.
plateaus at a maximum SNsg value of just over
The calibrated model will now provide reason- 2.0. However, from work in Greece and Sydney
ably accurate values for SNupper for the thousands on rocky subgrades, Roberts et al (2010), a model
of routine FWD test points across the network. was developed which accounted for such cases of
It remains to analyse the modelled SNupper values strong subgrade.
for asphalt and to derive the thickness and material The model treats subgrade as just another layer
strength a for each test location. Space precludes the (Equation 15), and values are required for asg and
description of the detailed procedure in this Paper, hsg. The estimate of CBRsg from Equation 19 was
but examples of modelled h and E values are shown transformed into a value for asg using Equation 5.
in Figures 9 and 10 for the data as used in Figure 8. A single universal value for hsg was carefully chosen,
Simultaneous estimation of h and E is innovative. ensuring that the resulting value of SNsg was com-
patible with the existing TRL method for the lower,
3.6 Definition of subgrade more usual values of CBRsg. A hsg value of 22 inches
(560 mm) suited, as in Equation 20 and Figure 11.
First the subgrade type, (P, M or D) is selected
based on local observation. For strength, SNsg is
a significant part of SNlower, and can be estimated ::~~ ~~;--;:-~----;~--;;:=:::;:=:==:::;:;:;::::=~-;-~
12000
separately using the d900. deflection data in the 10000
8000
tail of the bowl, Roberts et al (2010), proposed 6000
4000
a simple relationship, which after local regional 2000
0 ·~==--==----=~---------~
Ut98S)·dney IIIO Syd•ry 111 7Sydney IIJOS)·dnry IJI2Wn tffn 131 9 Wn ttm U29 Wtsttrn
calibration versus detailed drilled and tested cores, Trial Site Reference by Region

delivers reasonable outcomes. The relationship is

shown as Equation 19: Figure 10. Plots of actual E Modulus values and mod-
elled values, both plotted against the same core test point
K sg ×1700 locations.
Rsg =
CBR (19)
d900

6.00.------------------------.,
4.00
3.80
, ..
1, ,
5.00
3.60
L I ) .-'
,,,,''
I~ ~..-'
4.00
3.40 ....
......... -
3.00
3.20 --
, ,,
2.00
1.00
3.00
o.oo .L.__ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ___,

l t98Sydney IIIO S)·dnty 111 7S)·dn ty IIJO S)·dn ty IJ12Wtt~trn IJ19Wftttrn IJ29Wnttrn
2.80
0:0 2.60
,
,"" r
Trial Sile Reference by Region
z
~
2.40 ,
I
2.20

;(_
§
Figure 8. Plots of actual SNupper values and modelled § 2.00 ~
deflection based values, both plotted at core test point Cl.
l.80 r--
locations. ~ l.60
2:
... 1.40
00
l.20
I

400 .--------------------~
350 !:I
1.00
I
-~
I
I
- SNsg(TRL)
--·SNsg (Proposed)
r-- r--
300 00 0.80 r-- r--
J~
250
200 0.60
ISO 0.40
ID L-'- --L~ ---f-' L
100
50 . 0.20
oL--------------------~ 0.00
1098Sydney 1110 Sydney 111 7Sydnty 11 38S)·dnC')' 1305 Wntrrn IJ ilWntrrn 13 19 WHttrn 1) 29Wntn-n
Trial Site Reference by Region 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 ISO 160
Subgrade CBR (%)

Figure 9. Plots of actual asphalt thickness values

and modelled values, both plotted at core test point Figure 11. New method for SNsg from CBRsg compared
locations. to TRL.

707
 N sg = asg × 22
SN (20) pavement strength is the input to the models pave-
ment remaining capacity.
The old and new estimates of SNsg are compatible
up to CBRsg = 15%, thereafter the new value contin- 3.9 Final 4 layer pavement model
ues to climb. It is noted that for CBRsg = 80%, under A typical seasonally adjusted pavement model
the new method, SNsg = 3.3, potentially suitable as result is shown in Figure 12, for FWD tests at 50 m
a pavement in its own right, an integrated outcome. spacing. This case covered a zone at the junction
between two pavement types. h (mm), E (MPa) and
3.7 Definition of subbase (unbound or cemented) CBR (%) are seen for each layer. The cement treated
subbase and thick asphalt were correctly identified.
The value of SNs is determined as SNlower – SNsg.
The system works in a similar way as the base layer,
a cascading decision tree operating within thick- 4 PAVEMENT REMAINING CAPACITY
ness limits, available material types and strengths.
For spray seal flexible pavement types the over- 4.1 Default model for new works
all system is capable of identifying the material
types, estimating the thicknesses and material A pavement structural capacity model was devel-
strengths, of the base and sub-base layers, includ- oped for application to in-service roads, but
ing cemented. starts as a model for new roads. The default
model is defined as both primary and inverse in
Equations 21 and 22:
3.8 Adjustment for seasonal effects
1
The determination of material type, thickness and
⎧⎪ ⎡SNP
N (Ki × )⎤⎦ ⎫⎪ K ×0.175
CAP = ⎨ ⎣
c
material strength, as described for each of the four (21)
generic pavement layers above, applies to condi- ⎬
⎡
⎣ g 2.135 ⎤⎦
K ×
tions at the Time of Testing (ToT). Adjustments ⎩⎪ ⎭⎪
need to be made, without change of layer thickness. SNP
N (K i
× ) + (K g
× ) × [CAP Kc
] (22)
The system adjusts the asphaltic material
strength from ToT (at whatever hour), to a repre- where Ki, Kg & Kc are calibration factors. The
sentative monthly, annual and long-term value. For default system emulates representative outcomes
moisture effects, these are assumed to be monthly for three world best practice design systems for
at ToT, then adjusted to annual and long-term. new works on rural sealed or asphalt surfaced
Cemented materials are insensitive. The mod- unbound pavements, COLTO (1985), TRL (1993)
elling system is sensitive to surface cracking, and & QDTMR (2005). These agree over a wide traffic
how this allows moisture to access the base layer. range.
Drainage effectiveness is modelled to affect mate-
rial strength of subbase and subgrade level.
Seasonal modelling allows data captured at 4.2 Calibration as model for in-service roads
different times on a network, especially for net- The 500 test sections in NSW, each with three
work wide surveys using a Traffic Speed Deflec- test sites (1,500 total sites) were each assessed as
tometer (TSD), to all be expressed as a single a blind test by experienced NSW State officers for
long-term condition. This long-term statement of expected time until the next structural treatment
in years. The research analyst, knowing the actual
0 traffic loadings and projected growth rate, was
- 100 able to use this new data to estimate the remaining
-200 capacity of each road section (MESAs). This field
e_§_ -300 data was used alongside parallel FEM based anal-
-400
"E yses for remaining capacity at each site. The out-
c:r -500
~ comes were collated, assessed by pavement type
......
.....
-600
- 700
and Region, and plotted versus SNP. A dominant
<e
....:l
-800 pattern emerged which informed the calibration
-900 of Equations 21 and 22 for in-service remain-
-1000 ing capacity. The calibration factors are listed in
Table 3, and equations plotted in Figure 13.
The spray sealed unbound pavement curve
Figure 12. Valetta Rd, City of Charles Sturt, South is similar to South African practice, COLTO
Australia. (1997).

708
Table 3. K factors to modify default capacity The same points were then tested by TSD, travel-
relationship. ling through the site at up to 80 km/h. The analysis
took a few seconds, with results in Figure 15. The
Pavement type group Ki Kg Ki similarity of the FWD and TSD results is useful.
Spray Sealed Unbound 1.600 1.150 1.450
Spray Sealed Stabilised 1.800 1.450 3.000
Thin Asphalt Unbound 1.650 0.925 2.100 6 CONCLUSIONS
Thin Asphalt Stabilised 1.700 0.700 2.800
Full Depth Asphalt 2.000 0.800 1.400 The described system has successfully analysed FWD
data to model and quantify layered flexible pave-
ment structures, and estimates of their remaining
lo.o rr===========;----,-.,:11 structural capacities. Its extension to TSD applica-
=Spray Sealed Unbound
9.0 tion opens up the potential for continuous real-time
8.0 -Thin Asphalt Unbound
7.0
pavement analysis at highway speed. The system
- ·Thin Asphalt Stabilised
6.0
has found widespread use in local government in
= Spray Sealed Stabilised
~
;z 5.0 Australia, and was used by NSW Roads and Mari-
"' 4.0 I L-----....,.----' time Services in 2010 & 2015 with TSD data to
3.0 develop road strategies and in the rebuilding of sev-
2.0
1.0 +-...--.,..,....,+-....,....,....,....,-.+--.....-......+--.-..,....,..,..,..,..,f---,,.....,..,_,.,.,,f-.,......,...,...,.,..,-l
eral key state roads. The system provides new infor-
0.000 0.001 0.0 I 0 0.100 1.000 10.000 100.000 mation to inform planning and pavement evaluation.
Remainin Ca aci (MESAs)

Figure 13. Pavement in-service remaining capacity

curves. REFERENCES

Austroads 2011, AGPT, Part 5: Pavement Evaluation and

200
Deptb
400 Treatment Design. AGPT05-11, Austroads, Sydney,
into road
by layer
600 Australia.
800
(mm)
1000 COLTO. 1985. Structural design of interurban and rural
80 road pavements, TRH 4. COLTO, Pretoria, RSA.
60 COLTO. 1997. Flexible pavement rehabilitation investi-
gation and design. TRH12. COLTO, Pretoria, RSA.
Horak, E. 2008. Benchmarking the structural condition
:: .•lliii.JtllbliDlbltl: FWD Test Points at 20 m s acin
of flexible pavements with deflection. Journal, SAICE,
RSA.
Ioannides, A & Rohde, G. 1986, Theoretical Implications
Figure 14. Part of Illawarra 2 NSW Trial Section. of the AASHTO Method 2 for Pav’ Eval’. TRR 1307,
TRB.
200
ISOHDM 2004, HDM-4 (version 1.3), software and doc-
Deptb umentation. ISOHDM, PIARC, Paris, France.
400
into road
600
by layer
800
Loizos, A, Roberts, J & Crank, S. 2002. Asphalt pave-
(mm)
1000 ment deterioration models for mild climatic condi-
80
tions. 9th ICAP.
60 Muller, W & Roberts, J. 2013, ‘Revised approach to
40 assessing traffic speed deflectometer data. IJPE, vol.
20 14, no. 4.
QDTMR. 2005. Pavement Design manual. Transport
TSD Test Points at same 20m s acin as FWD
Technology Division, Brisbane, Queensland, Australia.
Roberts, J, Crank, S, Roper, R, et al. 2003. Road Infra-
Figure 15. Same Illawarra 2 NSW Trial Section tested structure Management System: User Guide. ARRB
by TSD. NTUA, Athens.
Roberts, J, Roper, R & Michel, N. 2010. RTA Procedures
5 TRAFFIC SPEED DEFLECTOMETER for Pavement Structural Condition and Performance.
(TSD) ARRB.
Salt, G. 1999. Determining the structural capacity
of unbound granular pavements in New Zealand.
A method was developed, Muller & Roberts (2013), Transfund.
for processing TSD data into FWD type deflection. TRL. 1993. A guide to the structural design of bitumen-
Test sections in NSW were tested by FWD at surfaced roads in sub-tropical countries, ORN31,
close spacing under research conditions. Data TRL, UK.
analysis gave a model per test point of pavement Ullidtz, P. 1987. Pavement analysis, Elsevier Science Pub-
structure and remaining life, as in Figure 14: lishing Company, Amsterdam, Netherlands.

709
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Determination of an optimum backcalculation cross-section for

flexible pavements

K.A. Tutu & D.H. Timm

Auburn University, Auburn, Alabama, USA

ABSTRACT: Backcalculation of in-situ moduli is essential for pavement analysis and design. However,
unconventional pavements (subgrade stiffer than granular base) present a challenge to backcalculation
due to reasons such as compensating layer effects. This study describes a methodology to determine an
optimum backcalculation cross-section for nine unconventional pavement test sections constructed at
the National Center for Asphalt Technology Test Track in Alabama, USA. The unbound materials also
exhibited different stress-sensitivity behaviour. Historically, the effect of the unconventional stiffness pro-
file and the different stress-sensitivity is that as-built cross-sections have yielded unacceptably low back-
calculated aggregate base moduli. Falling weight deflectometer testing provided 2916 deflection basins for
this study, which were backcalculated using EVERCALC 5.0. Nine cross-sections were evaluated based on
deflection basin fit and reasonableness of the backcalculation results. The study converged on pavement
structure consisting of the as-built asphalt layer thickness over a pseudo-base (as-built aggregate base
thickness plus 400 mm of subgrade soil) over the subgrade as the optimum backcalculation cross-section
that minimized the effects of the unconventional stiffness profile and the different stress-dependency
characteristics.

1 INTRODUCTION may not be a primary indicator of the validity of

results. Backcalculated moduli are sensitive to layer
Backcalculation of in-situ layer moduli using thicknesses (Rwebangira et al. 1987; Von Quintus
Falling Weight Deflectometer (FWD) measure- & Killingsworth, 1998). Lee et al. (1998) suggest
ments provides useful data for pavement analysis the thickness of the base layer must be at least 1.5
and design. Most backcalculation programs itera- times greater than that of the Asphalt Concrete
tively determine in-situ moduli. First, seed moduli (AC) layer in order to obtain reliable in-situ base
are assumed and layered elastic theory is used to moduli. Similarly, the recommended minimum AC
calculate deflections. The measured and calculated thickness is 25% of the FWD load plate diameter
deflections are compared and, if necessary, the (Ullidtz & Coetzee, 1995; ASTM, 2015). ASTM
seed moduli are adjusted and the process repeated (2015) notes that for adjacent unbound layers, a
until there is an acceptable match between the modulus ratio (upper layer modulus over lower
deflection basins. Moduli that correspond to the layer modulus) of less than 0.5 adversely affects
matching deflections represent the in-situ moduli. backcalculation results. The reason for these guide-
The process initiates at the outer edge of the deflec- lines is that relatively thin layers have relatively lit-
tion basin, whereby the moduli of the bottom pave- tle influence on the resulting deflections, which are
ment layers are determined first and they are used, often ignored by backcalculation routines (Irwin,
together with the deflections at the nearer sensor 2002). Unless the stiffness and/or thickness of a
offsets, to determine those of the upper layers. layer considerably impact deflections, backcalcu-
Although automation has improved the efficiency lation is unsuccessful in determining its modulus
of the backcalculation process, challenges such (Ullidtz & Coetzee, 1995). Bedrock, stress-harden-
as non-uniqueness of the solution, the sensitivity ing subgrade, and water table create a stiff bottom
of the backcalculation results to program inputs, layer, so it is recommended to account for these
proper pavement modeling, and compensating factors in backcalculation (Uddin et al., 1986;
layer effects all affect the backcalculation results. Mahoney et al., 1993; Newcomb et al., 1995).
Multiple sets of program inputs could produce Compensating layer effects arise from the
a good match between measured and calculated sequential nature of the iterative backcalculation
deflections even though the corresponding moduli process, which causes an interaction among the
may be invalid. Thus, small backcalculation error backcalculated moduli. Thus, an over-estimation of

711
one layer’s modulus leads to an under-estimation of and thus bedrock or groundwater was unexpected
an adjacent layer’s modulus. Since iterative backcal- at a shallow depth. Historically, the effect of the
culation programs calculate the subgrade modulus unconventional stiffness profile and the different
first, if it is under- or over-estimated, the algorithm stress-sensitivity is that as-built cross-sections have
compensates for the low or high subgrade modu- yielded unrealistically low backcalculated aggregate
lus by over- or under-estimating the base moduli to base moduli. Therefore, there is the need to deter-
match the deflections measured in the base layer. mine an optimum backcalculation cross-section to
For a stress softening subgrade, Ullidtz & Coetzee minimize the effects of the unconventional stiffness
(1995) explain the stress at the outer FWD sensor profile and the different stress-dependency charac-
location is lower than that beneath the load plate, teristics of the granular materials.
leading to a corresponding higher subgrade modu-
lus at the outer sensor location. The overall effect is
alternating layers that exhibit low or high stiffness. 3 OBJECTIVE AND SCOPE
It is important to address compensating layer effect
because it leads to incorrect moduli despite accept- This study sought to determine an optimum back-
ably small backcalculation error. calculation cross-section for nine unconventional
pavement sections (subgrade stiffer than aggregate
base). An optimum backcalculation cross-section
2 PROBLEM STATEMENT was considered as one that provided an acceptable
deflection basin fit as measured by the Root Mean
A unique characteristic of the pavement sections Square Error (RMSE) and reasonable moduli
at the National Center for Asphalt Technology based on empirical knowledge of the pavement
(NCAT) Test Track is the unconventional stiffness materials. To achieve this objective, FWD testing
profile (subgrade stiffer than the crushed granite was conducted at the NCAT Test Track to provide
base). Timm & Priest (2006) reported a pseudo- 2916 deflection basins for backcalculation using
base to subgrade stiffness ratio of 0.34. Taylor & EVERCALC 5.0, with nine potential backcalcula-
Timm (2009) also found a similar stiffness ratio. tion cross-sections evaluated.
The pseudo-base, in both studies, consisted of
crushed granite and subgrade soil, which were
combined for backcalculation purposes. Labora- 4 TEST FACILITY AND SECTIONS
tory resilient modulus testing (AASHTO T307)
showed that the subgrade soil was, indeed, stiffer The NCAT Test Track is a full-scale accelerated
than the aggregate base material, although addi- pavement testing facility located in Alabama,
tional testing over a wider range of stress produced USA. The 2.7 km oval track has forty-six 60 m
comparable moduli (Timm & Priest, 2006). long test sections. A full research cycle comprises
Stress-sensitive subgrade is a typical reason for an one year of construction/forensic study, followed
unconventional stiffness profile. However, the Test by two years of traffic application, during which
Track’s subgrade soil is stress-independent, accord- 10 million equivalent single axle loads are applied
ing to a study by Taylor & Timm (2009) in which by trucks operated by human drivers.
four constitutive models were used to evaluate the Seven of the nine test sections used for this
stress sensitivity of the unbound materials, utilizing study were constructed in the summer of 2015 as
both triaxial (NCHRP 1–28 A) and FWD test data. part of NCAT’s sixth research cycle. The sections
The models examined the relationship between resil- differed only in the surface mix; the label above the
ient modulus and (1) bulk stress, (2) deviatoric stress, sections in Figure 1 indicates the composition of
(3) bulk and deviatoric stresses, and (4) bulk stress the surface mix. The sections were designed to have
and octahedral shear stress. Both the laboratory and similar total AC and aggregate base thicknesses,
field models showed no major stress-dependency for though the figure shows natural construction vari-
the subgrade soil. However, the laboratory results ability. The AC base mixtures were dense-graded
indicated the aggregate base material was stress- Superpave mixes with highly polymer-modified
hardening, whereas backcalculation revealed an binder while the surface mixes used a variety of
opposite behaviour. The study found a good agree- binder contents and recycled materials. The granu-
ment between the laboratory- and field-measured lar base was designed as 150 mm of dense-graded
subgrade moduli; however, this was not the case crushed granite. The other two sections (N3 and
for the aggregate base material. Note that the inclu- N4) were constructed in a previous research cycle.
sion of a stiff bottom layer in the backcalculation Section N3 consisted of 150 mm of AC over Cold
cross-section yielded poor results (Timm & Priest, Central Plant Recycled (CCPR) base over 150 mm
2006; Taylor & Timm, 2009). This was not surpris- of crushed granite. Section N4 had a similar con-
ing since the Track was built on an embankment, struction except that the AC was 100 mm.

712
 150 mm IOO mm 35%R."J'
AC AC I S% RAP
CCPR AZGTR
CCPR

so
100
§
150
~-
u
J2u 200
:.a
1-<
~
«S
250
o-l

300

350

400

450
~ AC Surface m AC Base (llH CCPR Base ~Crushed Grani1e 0:] Subgrade

Figure 1. As-built pavement cross-sections.

5 FWD TESTING error approach was adopted; different subgrade

soil thicknesses were added to the aggregate base.
Deflection testing was conducted with Dynatest For instance, H0 represented the as-built cross-
Model 8000 FWD on October 19, October 26, and section; H100, a cross-section in which 100 mm of
November 11, 2015 when the sections were reason- subgrade soil was added to the aggregate base to
ably new in terms of traffic application. The FWD obtain a pseudo-base thickness of 250 mm, and
had nine sensors installed at offsets of 0, 200, 300, so on. Note that the backcalculated pseudo-base
450, 600, 900, 1200, 1500, 1800 mm from the center modulus was a composite material parameter.
of a 150 mm radius load plate. There were 12 test EVERCALC 5.0 was used for the backcalcula-
locations within each test section. Three replicates of tion. Extreme modulus limits were specified to avoid
27, 40, 53 kN FWD loads were applied at each test producing artificially-limited moduli. The limits
location, yielding a total of 2916 deflection basins. were: 550 to 20,680 MPa for the AC layer and 7 to
The large data set captured a wide range of material 689 MPa for both the pseudo-base and subgrade.
variability and stress-dependency behaviour. Seed moduli were selected, based on experience,
to minimize non-uniqueness of the backcalcula-
tion solution. The seed moduli were 3450, 140 and
200 MPa for the AC, granular base, and subgrade,
6 BACKCALCULATION METHODOLOGY
respectively. Poisson’s ratio for the AC, granular base
and subgrade soil were 0.35, 0.40, and 0.45, respec-
The AC layers were combined for backcalculation.
tively. Five iterations per deflection basin were speci-
Based on experience at the NCAT Test Track, it
fied, and the Root Mean Square Error (RMSE) and
was not necessary to account for a stiff bottom
modulus tolerance were each set at 0.1%.
layer. Nine trial backcalculation cross-sections con-
sisting of AC over a pseudo-base (aggregate base
plus subgrade) over the subgrade were evaluated.
7 RESULTS AND DISCUSSION
The pseudo-base was meant to mitigate com-
pensating layer effects. It was considered as an
7.1 Pseudo-base effect
effective approach to minimize the effects of the
unconventional stiffness profile and the different The pseudo-base effect on the backcalculated
stress-dependency characteristics of the granular moduli are shown in Figures 2 through 4. In
materials. A critical question regarding the pseudo- Figure 2, the AC moduli from the trial cross-sections
base was what thickness of the subgrade soil are plotted versus those obtained from the as-built.
should be added to the aggregate base? A trial-and- Similarly, Figure 3 plots the pseudo-base moduli

713
 • HIOO
• H200
12000 - H300
+ H400
o H550
=
0
10000
t:. H700
.B
0) + H850
~ 8000 o HIOOO
2 -Line of Equa lity
u
]
6000
t-
"'::l
-;
"0
0
4000
2
u
<(
2000

0 ~------,-------,-------,-------,-------,-------,-----~

0 2000 4000 6000 8000 10000 12000 14000

AC Modulus (As-Built Cross-Section), MPa

Figure 2. Backcalculated AC moduli.

250

200

"'
.;
"0
0
150 ....
2 • HIOO
• H200
- H300
• H400
o H550
50 "' H700
+ H850
o HIOOO
~ Line of Equality
0
0 50 100 150 200 250

Aggregate Base Modulus (As-Built Cross-Section), MPa

Figure 3. Aggregate base moduli.

against the as-built aggregate base moduli, whereas indicated the relative change in modulus from the as-
Figure 4 shows the subgrade moduli plotted simi- built condition. The goodness-of-fit was high; R2 for
larly. Increasing the pseudo-base thickness generally the AC curves ranged from 0.9992 (H100) to 0.9577
decreased the AC modulus and increased the pseudo- (H1000); for the pseudo-base, from 0.9899 (H100) to
base and the subgrade moduli to varying degrees. The 0.8438 (H1000); and for the subgrade, from 0.9993
pseudo-base modulus was most affected, followed by (H100) to 0.9938 (H1000). The slopes are presented
the AC modulus and the subgrade modulus. in Figure 5. As expected, the pseudo-base thicknesses
To quantify the effect of the pseudo-base, linear had the greatest impact on the pseudo-base moduli.
trend lines were fitted; the slope of the trendlines Greater pseudo-base thickness corresponded to higher

714
 400

~
0
350
~
c0
·~ 300
en
il
8
u
~ 250
E, • HIOO
"';:l a H200
:;
"0 200 - H300
0
~ • H400
I!)
"0 o H550
~
l;1J 150 "' H700
.D
;:l + H850
en
o HIOOO

100
100 150 200 250 300 350 400

Subgrade Modulus (As-Built Cross-Section), MPa

Figure 4. Subgrade moduli.

3.0 1================;------------,
• AC ~ Pseudo-base El Sub grade J

2.5

I!)
g. 2.0
Ci3
~;:l
~1.5
0
·~

~
~ 1.0

0.5

0.0
HIOO H200 H300 H400 H550 H700 H850 HIOOO

Trial Cross-Section

Figure 5. Linear regression curve slopes.

pseudo-base stiffness. For example, Figure 5 shows subgrade modulus increased less significantly, and the
that H100 (i.e. adding 100 mm of subgrade soil to pseudo-base modulus remained generally constant
the as-built base thickness) increased the average at an average increase of 138%. This suggested that
pseudo-base modulus by 55%. For H100 to H400, the addition of more than 400 mm of subgrade soil
both the AC and subgrade moduli were changed to the aggregate base did not considerably increase
less significantly as demonstrated by their slopes the pseudo-base modulus, but had a severe impact
nearly equal to 1.0. However, for H550 and beyond, on the AC and subgrade moduli. Essentially, above
the AC modulus decreased more significantly, the 400 mm, EVERCALC compensated for the increased

715
 100

90

80

Cl.l
C/)
70
lS
c
Cl.l
!:! 60
+ HO
~
Cl.l 50 • HIOO
.2:
l!:::: £ H200
E 40 - HJOO
8 oH400
30
OH550
20 ... H700
H 50
10 111000

0
0 2 4 6 8 10
Root Mean Square Error (RMS E) ,%

Figure 6. Distribution of root mean square error.

pseudo-base stiffness by calculating lower AC moduli their relatively lower RMSE, a major drawback of
in order to match measured and calculated deflections greater pseudo-base thicknesses was their potential
in the AC layer (Fig. 5). to induce compensating layer effects (Figs. 2–5)
which would adversely affect the backcalculated
7.2 Deflection basin fit AC and subgrade moduli.

Root mean square error (Eqn. 1) is typically used

to evaluate the goodness-of-fit between the meas- 8 OPTIMUM CROSS-SECTION
ured and predicted deflection basins.
The as-built cross-section yielded 80% accept-
1 ⎛ δ ci − δ mmi ⎞
n
2 able backcalculation solutions for a typical 3%
RMSE = ( ) ∑ (1) RMSE threshold. Based on the RMSE statistic,
n i =1 ⎜⎝ δ mmi ⎟⎠ the as-built cross-section may not be considered
as an optimum backcalculation cross-section. On
where RMSE = root mean square error (%); the contrary, H400 yielded backcalculation results
n = number of deflection sensors; δci = calculated sur- that had a mid-range error distribution (Fig. 6).
face deflection at sensor location i; and δmi = meas- Approximately 90% of the moduli had RMSE
ured surface deflection at sensor location i. below 3%. As Figure 5 shows, pseudo-base thick-
EVERCALC automatically computes the RMSE nesses greater than H400 resulted in a decreasing
for each deflection basin. These RMSE’s were then AC modulus and an increasing subgrade modu-
compiled into cumulative distributions to evaluate lus. Considering the total deflection basins, H400
the effect of the pseudo-base thickness on deflec- decreased the average AC modulus by 3.6%, but
tion basin fit. Figure 6 shows that greater pseudo- it increased the average pseudo-base modulus by a
base thickness yielded lower RMSE. Considering factor of 1.3 and the subgrade modulus by 0.92%,
an acceptable RMSE of 3%, three groups of RMSE respectively. The changes in the AC and subgrade
are evident in the figure. Cross-sections H0 to H300 moduli may be considered practically insignificant.
produced 75 to 85% acceptable backcalculation Despite the increase in the pseudo-base modulus
results. The as-built cross-section (H0) yielded 80% by a factor 1.3, its average ratio to the subgrade
acceptable backcalculation moduli. The second modulus was 0.38, compared with a value of
group, containing only H400, produced 90% accept- 0.34 in previous NCAT studies (Timm & Priest,
able results. The last group, which consisted of H55 2006; Taylor & Timm, 2009). The objective of the
to H1000, produced between 92 to 96% satisfac- current study was not to determine a backcalcu-
tory backcalculation moduli. Notwithstanding lation cross-section to reverse the unconventional

716
stiffness profile, but one that provided optimum Management and the Federal Highway Administra-
backcalculation results, given the unique char- tion also sponsored this research.
acteristics of the pavements. Thus, a three-layer
pavement structure comprising the as-built AC
thickness over the as-built granular base thickness REFERENCES
plus 400 mm of subgrade soil (pseudo-base) over
the subgrade may be considered as an optimum ASTM 2015. Standard Guide for Calculating in Situ
backcalculation cross-section. It would be neces- Equivalent Elastic Moduli of Pavement Materials
sary to use embedded instrumentation to confirm Using Layered Elastic Theory. ASTM Standard D
the selected optimum backcalculation pavement 5858-96 (Reapproved 2015). ASTM International,
West Conshohocken, Pa.
cross-section. For instance, measured and theo- Irwin, L.H. 2002. Backcalculation: An Overview and
retical stresses and strains corresponding to the Perspective. Presented at the Pavement Evaluation
selected cross-section could be compared to deter- Conference. Roanoke, Virginia.
mine the extent of agreement. Lee, S.W., Mahoney, J.P. & Jackson, N.C. 1988. Verifi-
cation of Backcalculation of Pavement Moduli. In
Transportation Research Record: Journal of the Trans-
9 CONCLUSIONS AND portation Research Board, No. 1806, TRB, National
RECOMMENDATIONS Research Council, Washington, D.C., pp. 85–95.
Mahoney, J.P., Winters, B.C., Jackson, N.C. & Pierce,
L.M. 1993. Some Observations about Backcalculation
This study described a methodology to determine and Use of a Stiff Layer Condition. In Transporta-
an optimum backcalculation cross-section for pave- tion Research Record: Journal of the Transportation
ments with stiffer subgrade than the aggregate base Research Board, No. 1384, TRB, National Research
as well as with different stress sensitivity of the gran- Council, Washington, D.C., pp. 8–14.
ular materials. The following conclusions and recom- Newcomb, D. E., Van Deusen, D.A., Jiang, Y. & Mahoney,
mendations are based on the results of this study: J.P. 1995. Considerations of Saturated Soil Conditions
in Backcalculation of Pavement Layer Moduli. In
• The as-built cross-section of an unconventional Transportation Research Record: Journal of the Trans-
pavement does not necessarily yield optimum portation Research Board, No. 1473, TRB, National
backcalculation results. Errors may be acceptable, Research Council, Washington, D.C., pp. 63–71.
but the corresponding moduli may be incorrect. Rwebangira, T., Hicks, R.G. & Truebe, M. 1987. Sen-
• Unconventional pavements are prone to com- sitivity Analysis of Selected Backcalculation Proce-
pensating layer effects so a well-formulated pave- dures. In Transportation Research Record: Journal of
ment model is key to successful backcalculation. the Transportation Research Board, No. 1117, TRB,
National Research Council, Washington D.C., pp.
• This study tentatively converged on a viable 25–37.
pavement cross-section for backcalculation. It Taylor, A. J. & Timm, D.H. 2009. Mechanistic Characteri-
is applicable to these test sections for the dura- zation of Resilient Moduli for Unbound Pavement Layer
tion of the 2015 NCAT research cycle, although Materials. NCAT Report 09-06. National Center for
the methodology could be adopted for other Asphalt Technology, Auburn University, Alabama.
investigations. Timm, D.H. & Priest, A.L. 2006. Material Properties of
• Comparing measured and predicted pavement the 2003 NCAT Test Track Structural Study. NCAT
responses is a viable approach to validate back- Report 06-01. National Center for Asphalt Technol-
calculation cross-section. It is recommended ogy, Auburn University, Alabama.
Uddin, W., Meyer, A.H. & Hudson, W.R. 1986. Rigid Bot-
to use embedded pavement instrumentation to tom Considerations for Nondestructive Evaluation of
validate the optimum backcalculation cross- Pavements. In Transportation Research Record: Journal
section. of the Transportation Research Board, No. 1070, TRB,
National Research Council, Washington, D.C., pp.
21–29.
ACKNOWLEDGEMENTS Ullidtz, P. & Coetzee, N.F. 1995. Analytical Procedures
in Nondestructive Testing Pavement Evaluation. In
The authors wish to thank the cracking group exper- Transportation Research Record: Journal of the Trans-
iment sponsors for their cooperation and support of portation Research Board, No. 1482, TRB, National
Research Council, Washington, D.C., pp. 61–66.
this research. The sponsors include the following Von Quintus, H. & Killingsworth. B. 1998. Analyses
departments of transportation: Alabama, Florida, Relating to Pavement Material Characterizations and
Illinois, Michigan, Minnesota, New York, North Their Effects on Pavement Performance. Publication
Carolina, Oklahoma, South Carolina and Wiscon- FHWA-RD-97-085. FHWA, U.S. Department of
sin. The Alabama Department of Environmental Transportation.

717
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Backcalculation of asphalt concrete moduli using field-measured strain

M.M. Robbins
Ohio University, Athens, Ohio, USA

D.H. Timm & K.A. Tutu

Auburn University, Auburn, Alabama, USA

ABSTRACT: Asphalt Concrete (AC) modulus is a critical component of mechanistic-empirical pave-

ment design. It is well known that AC is a time and temperature dependent material that is influenced
by vehicle speed and pavement temperatures in the field. However, current methods to characterize AC
modulus, such as dynamic modulus (|E*|) and backcalculation from deflection basins resulting from fall-
ing weight deflect to meter testing (FWD) do not accurately represent field conditions. As a result, the
backcalculated AC moduli from FWD testing do not accurately reflect the effect of vehicle speed on
modulus. One means to circumvent the limitations of |E*| and FWD backcalculated AC moduli is to use
measured strain values to backcalculate AC moduli. In this investigation, relationships between predicted
tensile strain and AC modulus that enabled the application of measured strain at the bottom of the AC
layer were determined to estimate in-place AC modulus. The estimated AC moduli values produced pre-
dicted tensile strains that closely matched measured strain values and were found to be an improvement
over strain predicted from FWD backcalculated AC moduli.

1 INTRODUCTION is used in |E*| testing to achieve strains within

the linear-elastic range of AC. Researchers have
It has long been known that asphalt concrete is a vis- explored both stress pulses (Loulizi et al., 2002)
coelastic material that is both time and temperature- and strain pulses in the field (Garcia & Thomp-
dependent. However, previous design systems have son, 2008). However, there has not been consensus
historically been empirically based and therefore did on the appropriate pavement response to investi-
not consider the effects of temperature and speed gate the relationship between laboratory and field
(time). With the advent of mechanistic-empirical measurements. In the laboratory, compressive load
pavement design, current design methodologies such pulses are applied in a continuous fashion for each
as the Mechanistic-Empirical Pavement Design frequency tested. However, load pulses are not
Guide (MEPDG) (now Pavement ME Design) con- continuous in the field due to axle configurations
sider the viscoelastic effects of AC through the use of and vehicle spacing. Additionally, |E*| testing is
the dynamic modulus (|E*|) of the Asphalt Concrete completed at a range of frequencies in the labora-
(AC) in combination with pavement temperature tory; however, loading is applied in the field at a
and vehicle speed. The Pavement ME Design soft- range of vehicle speeds. To relate field conditions
ware calculates |E*| at a given temperature and speed to laboratory conditions would require that the
from the master curve created from user-provided time of loading be determined in the field and con-
laboratory measured |E*| data or |E*| predicted from verted to loading frequency. Within the MEPDG,
the embedded predictive equations (ARA, 2004). the proposed time-frequency relationship is simply
Once predicted for the given conditions throughout t = 1/f, where f is loading frequency and t is loading
the pavement design life, |E*| is used within a lay- time (ARA, 2003). However, others suggest that
ered elastic system to determine critical pavement the correct conversion is t = 1/ω based on the field
responses such as tensile strain at the bottom of the of rheology, where ω is angular frequency (Van der
AC and vertical strain at critical depths. Poel, 1954; Schwarzl, 1975; Ferry, 1980), and this
Although |E*| enables mechanistic relationships can be expanded to read as t = 1/2πf, while oth-
to be used to predict critical distresses while con- ers propose alternative relationships (Christensen,
sidering the viscoelastic effects of speed and tem- 1982; Daniel & Kim, 1998; Schapery & Park, 1999;
perature, the testing conditions do not accurately Al-Qadi, 2008). At present, no research has been
represent field conditions. A haversine compres- found to corroborate these time-frequency rela-
sive stress pulse applied at a range of frequencies tionships under field conditions.

719
 An alternative to using dynamic modulus to of the 14 sections was instrumented with asphalt
characterize the AC modulus is to backcalculate strain gauges, earth pressure cells and temperature
moduli from deflections measured with a Falling probes. FWD testing was conducted multiple times
Weight Deflectometer (FWD) at a range of tem- per month to determine in-situ material properties
peratures. The deflection basin created by drop- of the unbound layers. A layered-elastic analysis
ping loads on the pavement surface enables the was utilized to calibrate AC modulus to field meas-
determination of in-situ moduli of the bound (AC) ured strain at the bottom of the AC layers.
and unbound (base and subgrade) layers. This is
typically done through the use of an iterative proc-
ess using a layered elastic system to converge on a 3 FACILITY AND TEST SECTIONS
modulus for the measured deflections and known
load. Although FWD testing is a relatively quick The NCAT Pavement Test Track is a 2.74-km
way to assess the in-situ material properties and closed-loop full-scale flexible pavement test facil-
condition of the pavement layers, FWD testing ity located in Alabama, US. The Test Track has
does not accurately represent traffic loading. FWD 46 sections that are subjected to accelerated load-
testing is completed with an impact load rather ing by triple-trailer vehicles over two-year research
than a dynamic load. Research at the National cycles. The sections are designed to address spe-
Center for Asphalt Technology (NCAT) Test cific research needs and many have been embedded
Track has shown that the impact load applied dur- with instruments to measure pavement response as
ing FWD testing is more representative of a truck described below.
traveling at 193 km/h (Leiva-Villacorta, 2012), an
unrealistic speed for truck traffic on US highways.
3.1 Pavement cross-sections
Therefore, there is a need to characterize AC
modulus under conditions that more accurately Test sections from both the 2006 (Fig. 1) and 2009
represent field conditions to take advantage of the (Fig. 2) test cycles were used for this evaluation.
known viscoelastic effects of vehicle speed and pave- Four sections newly constructed in the 2006 test
ment temperatures. One method of achieving this cycle were used for this analysis: N1, N2, N8 and
is to use measured strain at the bottom of the AC N9. The 2009 test cycle consisted of 16 structural
layer, as each strain value is unique to the conditions sections as shown in Figure 2 with various technol-
at the time of loading, such as type of loading, vehi- ogies and unconventional materials. For this inves-
cle speed and pavement temperature. A 2011 study in tigation nine of the ten newly constructed sections
Oklahoma utilized an instrumented section of I-35 were evaluated in addition to the N9 section left
to backcalculate AC moduli using a layered-elastic in place from the 2006 test cycle. These included
system to converge on AC moduli necessary to match N5, N6, N7, N10, N11, S9, S10, S11 and S12.
the strain values measured in the field (Solanki et al., In total, the sections represent a wide range of
2011). Although this study was limited to one section materials, support conditions and cross-sections.
with conventional AC mixtures, the process showed The subgrade used in all 10 newly constructed
promise particularly for facilities and open-roadways sections was consistent with the materials used in
that have instrumented cross-sections. previous research cycles and has been well-doc-
umented as an AASHTO A-4(0) metamorphic
quartzite soil obtained on-site (Taylor & Timm,
2 OBJECTIVE AND SCOPE OF WORK 2009). The aggregate base was a crushed granite
material often used by the Alabama Department
The objective of this research was to develop a of Transportation (Taylor & Timm, 2009).
method to backcalculate AC moduli considering
measured pavement responses under varying vehi-
3.2 Instrumentation
cle speeds and pavement temperatures within the
context of a layered-elastic system. To characterize pavement response, instrumentation
To meet this objective, a total of 14 test sec- was embedded within each structural section at the
tions from the 2006 and 2009 research cycles at time of construction, providing extensive informa-
the NCAT Test Track were evaluated. These sec- tion on temperature, stress and strain within the
tions covered a wide range of mix types including structures. Embedded within each of the structural
Stone Matrix Asphalt (SMA), Warm-Mix Asphalt sections was a pair of earth pressure cells, twelve
(WMA), high Reclaimed Asphalt Pavement (RAP), strain gauges, and four temperature probes. The
sulfur-modified WMA, RAP and WMA combina- strain gauges were installed in groups of three by
tions, rich bottom layers and conventional mixes. direction, oriented either longitudinally or trans-
Furthermore, these test sections ranged in AC versely. Within a group, one gauge was aligned
thickness from 146 to approximately 356 mm. Each with the centerline of the outside wheel path, and

720
 Florida Alabama & FHWA Oklahoma
(new) (left in-place) (new) Missouri Alabama
~(new)
~
(new)

Nl N2 N3 N4 N5 N6 N7 N8 N9 N10 Sll
0
50
100
150

~
200

~-- 250
..Q
300
:EE-
350
·::;
Ctl 400
< 450
500
550
600
c::::::J PG 67-22 !===:! PG 76-22 ~ PG 76-22 (SMA) ~ PG 76-28 (SMA)
- PG 76-28 rlDDI PG 64-22 ~ PG 64-22 (2% Air Voids) ~ PG 70-22
~ Limerock Base c:::::::::::J Granite Base rnmmm Type 5 Base ~ Track Soil ~Seale Subgrade

Figure 1. 2006 Structural study test sections (Timm, 2009).

~
-d
~ ~
u :gt:: !:<. V)

~ ~
N

~
Florida Alabama lhiopave
,--A-..
J Oklahoma
,--A-.. ~ ~ ~
0 8 <
~
Nl N2 N3 N4 N5 N6 N7 N8 N9 NIO Nil S8 S9 SlO Sll S12

5
100

!50

200
:RR=E~~IBimr
~ 250
:f 300
]
:a.... 350
~ 400
a
~ 450
<(
500

550
600 ~~~~~~~~LW~~~~~Wll~~

c=:::J PG 67-22 !===:! PG 76-22 ~ Kraton ~ PG 76-28 (SMA) - PG 76-28

IIIII:Il 40% Thiopave fR!IBI!II PG 64-22 ~ PG 64-22 (2% All" Voids) ~ 30% Thiopave
~ Limerock Base c:::::::::J Granite Base ~ Track Soil ~Seale Subgrade

Figure 2. 2009 Strucrtural study test sections.

the remaining two were offset two feet center to the were installed in both the longitudinal and trans-
right and left of the outside wheel path. Doing so verse directions to create redundancy in the sys-
helped to account for wheel wander, allowing the tem. Full explanation of the sensors, arrangement
best-hit to be captured. Two groups of strain gauges and installation has been previously documented

721
(Timm, 2009). Temperature probes were installed growth of any existing cracks. From this, crack maps
at various depths within the structure, creating a were created and maintained, enabling a very useful
complete temperature profile for each section. For and visual representation of formation and progres-
consistency, Campbell-Scientific model 108 tem- sion of cracking in a section relative to the traffic level.
perature thermistors were installed in each test cycle.
One thermistor was installed at the surface, mid-
depth and bottom of the AC layer and 76 mm into 4 METHODOLOGY
the base layer in each section (Timm, 2009).
The first step in developing the modulus-strain rela-
tionship was to determine strain from measured
3.3 FWD testing
pavement responses. Tensile strains were measured
Once construction was completed and trafficking at the Test Track through the use of embedded
commenced, FWD testing was conducted three strain gauges at the bottom of the AC layers. Ten-
days per month on each section during both test sile strain was measured in both the longitudinal
cycles using a Dynatest Model 8000 FWD with a and transverse directions in each of the 2006 and
150 mm radius split plate. Nine sensors were used 2009 structural sections. For this investigation, the
to measure pavement deflection and were spaced at maximum tensile strains were desired, as these are
0, 200, 300, 450, 600, 900, 1200, 1500, and 1800 mm the primary cause of bottom-up fatigue cracking.
from the center of the load plate. The testing was Previous studies at the track have found that maxi-
performed in the middle 75% (45 m) of each 60 m mum tensile strains were in the longitudinal direc-
test section, within this area, four Random Loca- tion and induced by the trailing single axles (Timm
tions (RL) were chosen for testing documented in & Priest, 2006). Therefore, only tensile strains meas-
detail elsewhere (Timm, 2009). Testing was con- ured in the longitudinal direction under single axles
ducted at four drop heights representing four load were utilized for this investigation. To be consistent
levels: 27, 40, 53, and 71 kN, with three drops con- with the Linear Elastic Analysis (LEA) program
ducted at each drop height. Pavement temperatures that was selected for this investigation, WESLEA
were recorded at the time of each test. version 3.0, strain was defined as the peak strain.
Backcalculation of the deflection basins meas- Peak strain, as shown in Figure 3, is the difference
ured from FWD testing was completed for each between the peak value and baseline value on the
section using EVERCALC 5.0 (WSDOT, 2005). strain trace.
A three-layer pavement section (AC over aggregate The best-hit was used in selecting strain, such
base over subgrade) was simulated. Surveyed layer that for a given section, the maximum peak strain
thicknesses at each offset and random location were was retrieved by surveying data from each of the
used in the backcalculation process; however, only working longitudinal strain gauges and selecting
the backcalculated moduli in the outside wheel path the maximum peak strain measured from all the
of the random location in the gauge array was con- passes of a given axle type. Therefore, on a given
sidered for this investigation to be consistent with date, the maximum strain was selected from at most
the location of the strain measurements. It should 450 strain measurements (5 trucks × 3 passes × 5 sin-
be noted that although four random locations were gle axles × 6 longitudinal gauges) in the longitudinal
established in both the 2006 and 2009 test cycles, the direction under a single axle. The best-hit approach
fourth random location was not established during ensured that the most representative strain meas-
construction of the 2006 test cycle. As a result, the urement was used for the analysis.
layer thicknesses at RL 4 were not surveyed; they To be consistent with the assumptions of a
were either taken from another RL in close prox- layered elastic system and for the resulting backcal-
imity, or interpolated from layer thicknesses at the
!50
surrounding random locations. It is important to
Peak
clarify that test section N9 in the 2006 research cycle
was considered to be a separate section from N9 in 100
the 2009 research cycle.
~c 50
·~
3.4 Performance monitoring
"' 0
The heavy triple trailer trucks were operated Tues-
day through Saturday, allowing for the Track to be
-50
shut down on Mondays, for performance moni- Time (seconds)
toring and other testing. Visual inspection was
performed manually, at which time any cracks that Figure 3. Illustration of peak tensile strain used for
were detected were marked as were the progressive investigation.

722
culated modulus estimates to be accurate, the sec- 3000 20684
tions used in this investigation must be undamaged
and the responses that are measured in the field 2500 17237
~
should be elastic. Strains associated with damage
~r:n
,.-...

were identified through the use of crack maps, rut ;g 2000 13790
<ll
depth and IRI measurements made weekly dur- ::l
:; 1500 .E
ing the research cycles and once identified, were "'C
0
~~ y = 618,330,000x- 1304
10342 .g
excluded from the analysis. Additionally, non-elastic ~ R2 = 0.999
0
u 1000 - 6895 ~
strains were identified and removed from the analy-
~ u
-<(
sis by deviation from observed temperature-strain ~
trends and through large variations of strain meas-
urements identified by the difference between the
maximum and 95th percentile strain values. Traffic
500

0
0
~
1000
.. ...
2000
3447

0
3000
at the Test Track operates at a target vehicle speed Strain (J. u::)
of 72 km/h. However, the heavy triple-trailer trucks
are manually operated, therefore, the speed is some- Figure 4. Modulus-strain relationship, S9.
what variable. For the 2006 and 2009 structural
studies, vehicle speed was calculated using the raw
strain traces to determine the actual vehicle speed that fell below the standard Root Mean Square
associated to measured strain values. More infor- Error (RMSE) threshold of 3% were used.
mation on the manual calculation of vehicle speed To simulate the single axles at the Test Track
has been previously documented (Robbins, 2012). which were found to induce the largest tensile
strains at the bottom of the AC layer (Timm &
Priest, 2006), a single axle was selected for the
4.1 Modulus-strain relationship
WESLEA evaluation. For each tire, the load was
To circumvent the shortfalls of backcalculation specified at 23.4 kN with a tire pressure of 689 kPa.
from FWD testing for AC layers, modulus-strain This was consistent with the average of the single
relationships were developed using a LEA pro- axles at the Test Track (93.4 kN).
gram. WESLEA version 3.0 was utilized to develop The modulus of the AC layer was varied while
modulus-strain relationships for each of the four keeping the dense graded aggregate base and sub-
sections from the 2006 structural study and the ten grade moduli constant at their respective (average)
sections from the 2009 structural study. Due to the backcalculated moduli. Tensile strain was meas-
limited number of layers (5) that can be simulated ured under vehicle speeds ranging between 53 and
in WESLEA the AC layers were simulated as com- 101 km/h, with at least 95% of the strains measured
posite AC layers. The composite AC modulus was at vehicle speeds between 68 and 84 km/h. Mid-depth
varied to determine the relationship between AC pavement temperatures associated with measured
modulus and resulting predicted tensile strains at tensile strains varied from 2 to upwards of 49°C.
the bottom of the AC layer. WESLEA was utilized Tensile strains were predicted at two locations at
to predict tensile strain at the bottom of the AC the bottom of the AC layer. The first was directly
layer for the known in-place properties of each below the center of the outside tire. The second
cross-section and loading. A wide range of AC location was 171 mm away from the center of the
moduli were selected to input into WESLEA to tire, which represents the center point between the
enable the prediction of tensile strain at the bottom two tires in a single axle. The maximum strain was
of the AC layer which would encompass possible selected from these two locations in the longitudinal
ranges of measured tensile strain for each section. (y-direction) direction. The varied AC moduli and
Simulations were completed such that the AC resulting strain predictions were plotted to deter-
layer was atop a granular base material on top of mine the modulus-strain relationships for each of
a subgrade material. For the granular base layer the 14 sections. An example of the modulus-strain
and subgrade layers, the surveyed lift thickness and relationships derived from the AC moduli input
backcalculated modulus specific to each section into WESLEA and predicted strain output by the
were used. Although backcalculation was com- program is shown for section S9 in Figure 4.
pleted for all load levels and dates, the average mod-
ulus for each the granular base and subgrade at the
40 kN load level was used for this investigation and 5 RESULTS AND DISCUSSION
are documented elsewhere (Robbins, 2012). The
40 kN load level best represents the 89 kN single The goal of this investigation was to develop a
axles on the triple trailer trucks used to traffic the method for backcalculation of AC modulus in a
sections. Additionally, only backcalculated moduli layered-elastic system that considered the effects

723
of varying vehicle speed and mid-depth pavement where Ecomp = composite AC modulus (psi);
temperatures on tensile strain, which in turn could c, d = regression terms; and ε = tensile strain at the
be used to improve tensile strain predictions in bottom of the AC layer.
design. To this end, modulus-modulus relation- To confirm the predicted tensile strains closely
ships were developed using WESLEA for each matched the tensile strains measured in the field,
of the 14 sections evaluated. A power model was the estimated AC moduli were then input into
fitted to each curve; an example is shown in Fig- WESLEA using the material properties for S9 to
ure 4 for section S9. The form of the power model again predict tensile strain. The predicted tensile
shown in Equation (1) fitted the data, with all of strains were compared with the strains measured in
the 14 sections having a coefficient of determina- the field at the bottom of the AC layer, as shown in
tion (R2) of 0.96 or higher as listed in Table 1. Figure 5. As illustrated in this figure, this method
The power model and the accompanying coef- results in very good predictions of tensile strains,
ficients shown in Table 1 determined through as the linear trendline forced through the origin
WESLEA iterations allowed for the determination shows a slight deviation from the line of equal-
of AC modulus for any measured longitudinal ten- ity with approximately a 4% overprediction of
sile strain at the bottom of the AC layer. Numer- measured strain values. The errors associated with
ous studies have shown strain at the bottom of the these strain predictions were relatively small with
AC layers to be a function of pavement temperature the largest percent difference found to be 4.73%
and vehicle speed (Chatti, 1996; Mateos & Snyder, occurring at the third highest mid-depth pavement
2002; Robbins & Timm, 2009). Specifically, previous temperatures (36°C). It should be noted that, in
studies at the Test Track related longitudinal tensile general, the percent error from the measured strain
strain at the bottom of the AC layers to measured values tended to be higher at increased mid-depth
mid-depth temperature and vehicle speed (Robbins pavement temperatures; however, the largest error
& Timm, 2009). Therefore, by using measured ten- was less than 5% of the measured strain.
sile strain to estimate AC moduli, the viscoelastic For comparison, backcalculated AC moduli
effects of speed and temperature are inherently from FWD deflections in section S9 were utilized
incorporated in the AC moduli estimates. in WESLEA to predict tensile strains. FWD test-
Eleven tensile strain measurements made in sec- ing and high-speed data collection (of measured
tion S9 throughout the course of the 2-year traf- strain) were not conducted on the same date, there-
ficking period were selected to compare predicted fore FWD backcalculated AC moduli values were
tensile strain values with field measurements. These selected such that they had pavement temperatures
tensile strain measurements covered a wide range of similar to those temperatures recorded at the time of
mid-depth pavement temperatures from 6 to 47°C tensile strain measurements. The highest mid-depth
and vehicle speeds between 66 and 74 km/h. Using temperature during FWD testing did not match that
the modulus-strain relationship shown in Equation of the measured strain values; therefore, there is one
(1) and the associated coefficients listed in Table 1 less strain prediction for the FWD backcalculated
for section S9, AC moduli were estimated. AC moduli. As shown in Figure 5, the predicted
strains resulting from the FWD backcalculated AC
Ecomp = cεd (1) moduli do not match measured strain values. The lin-
ear trendline shown in Figure 5 indicates that using
the FWD backcalculated moduli will result in a 53%
Table 1. Coefficients for Modulus-Strain Relationships. underprediction of tensile strain values. Absolute
percent differences between measured strain values
Year Section c d R2
1200
2006 N1 6.37E+08 −1.389 0.96 + Strain Backcalculated Moduli -'
2006 N2 −1.511 0.97
1.5E+09
~ LI FWD Backcalculated Moduli
2006 N8 2.7E+08 −1.256 0.99
"
·;; 900
2006 N9 9.9E+07 −1.213 0.99 ~
2009 N5 −1.257 0.99
3E+08
~
2009 N6 7.52E+08 −1.376 0.97 ]
600
2009 N7 1E+09 −1.302 0.99 "-
<
2009 N9 7.8E+07 −1.184 0.99 tll
...J 300
−1.245
2009
2009
N10
N11
3.43E+08
5.29E+08 −1.316
0.99
0.99 ~ -'
-'
2009 S9 6.18E+08 −1.304 0.99 0 -'

2009 S10 5.89E+08 −1.281 0.99 0 300 600 900 1200

2009 S11 5.54E+08 −1.270 0.99 Measured Strain (!'E)
2009 S12 5.95E+08 −1.277 0.99
Figure 5. Measured versus predicted tensile strain, S9.

724
and predicted strain values range from 32.3 to 57.9%. In examining the predicted strain values more
The low strain values are therefore consistent with closely, deviations from measured strains tended
Leiva-Villacorta’s (2012) finding that the impact load to be greater at slower speeds. Furthermore, the
used in FWD testing simulates a much higher vehi- largest absolute percent difference occurred at a
cle speed as it has been found that increased vehicle mid-depth pavement temperature of 27°C, the
speed results in decreased tensile strain at the bottom same temperature at which the percent differences
of the AC layer (Robbins & Timm, 2009). were also the largest for all three vehicle speeds. For
In the plot shown in Figure 5, tensile strains for mid-depth pavement temperatures of 16 and 27°C,
a very narrow range of vehicle speeds were uti- the backcalculated AC moduli overpredicted strain
lized. Therefore, tensile strains measured under a whereas at the highest temperature, 43°C, strain
broader range of vehicle speeds were also investi- was underpredicted.
gated to determine if the whole spectrum of viscoe-
lastic effects could be captured with this method.
A Test Track speed study was conducted during 6 CONCLUSIONS AND
the 2006 test cycle in which pavement responses RECOMMENDATIONS
were measured due to vehicle speeds of 24, 40, 56
and 72 km/h (Robbins & Timm, 2009). Measure- This paper described a method to backcalculate
ments at these speeds were made on several dates AC moduli considering viscoelastic effects in a lay-
to also capture pavement responses under varying ered-elastic system. Strain-modulus relationships
mid-depth pavement temperatures (Robbins & were developed which enabled the backcalculation
Timm, 2009). As a result of this study, the longitu- of AC moduli as a function of measured longitu-
dinal tensile strain at the bottom of the AC layer dinal strain at the bottom of the AC. As a result
was modeled as a function of the vehicle speed of utilizing measured strain, the viscoelastic effects
and mid-depth pavement temperature, as shown in of vehicle speed and pavement temperature were
Equation (2) (Robbins & Timm, 2009). incorporated. Based on this study, the following
conclusions are made:
ε t = −40.57 ln v + e 0 053T + 206.67 (2)
1. As illustrated, using backcalculated AC moduli
values from FWD deflections in a LEA system
where εt = longitudinal tensile strain under single results in predicted strain values significantly
axle loading (με); v = vehicle speed (mph); and less than those measured in the field. There-
T = mid-depth pavement temperature (ºF). fore, it can be concluded that backcalculated
In applying this model, strains at 16, 27 and AC moduli are not effective at predicting tensile
43°C and 24, 40, and 72 km/h were estimated for strain that closely matches field measured strain
the 2006 N9 section and used to calculate modu- when used in a layered-elastic system.
lus for the AC layer using Equation (1) and associ- 2. A close correlation between measured and pre-
ated coefficients listed in Table 1. The AC moduli dicted strain at a range of mid-depth pavement
were then used in WESLEA to predict maximum temperatures and vehicle speeds was found when
tensile strain at the bottom of the AC layer. using estimates of AC moduli based on measured
Figure 6 shows that the backcalculated AC moduli strain values. Based on this finding, it can be con-
closely predict the strains based on the previous cluded that the backcalculated AC moduli based
speed study with only a slight underprediction on measured strain values are able to account,
evident by the linear trendline forced through the in practical terms, for the viscoelastic effects of
origin. vehicle speed and pavement temperature within
the confines of a layered-elastic system.
500
Based on the results of this study, the following
-3
s:: recommendations are made:
-~ 400
00
C1) 1. Although this method for backcalculating AC
·oo 300
modulus results in accurate predictions of tensile
~
"'C strains at the bottom of the AC layer, measured
~ strains are necessary. Furthermore, modulus-
:.a 200
~
~ strain relationships necessary for this method
~
~
100 are unique to the cross-section and tensile strain
00
must be measured. Therefore, it is recommended
~ 0
0 100 200 300 400 500
that for future use in design the backcalculated
Measured Tensile Strain (f.lc) AC moduli be used to further develop relation-
ships for the prediction of AC composite mod-
Figure 6. Measured versus predicted tensile strain ulus from laboratory-derived properties, such
N9–2006 speed study. as |E*| of each AC mixture in the cross-section
725
 and other influential material properties. Once Garcia, G., & Thompson, M.R. 2008. Strain and Pulse
developed, AC moduli could be estimated from Duration Considerations for Extended Hot Mix
these known values and used in conjunction with Asphalt Pavement Design. In Transportation Research
a layered-elastic system or directly input into Record: Journal of Transportation Research Board,
No. 2087, Washington, D.C., pp. 3–12.
the MEPDG to more accurately predict tensile Leiva-Villacorta, F. 2012. Advanced Computing Tech-
strain at the bottom of the AC layer. niques in Structural Evaluation of Flexible Pavement
2. It has been found in previous research that the Using the Falling Weight Deflectometer. Doctoral Dis-
impact load is comparable to unrealistically sertation. Auburn University.
high truck speeds. This could, in part, account Loulizi, A., Al-Qadi, I.L., Lahouar S., & Freeman, T.E.
for the findings here, in which the use of AC 2002. Measurement of Vertical Compressive Stress
moduli backcalculated from FWD deflections Pulse in Flexible Pavements: Representation for
results in inadequate tensile strain predictions. Dynamic Loading Tests. In Transportation Research
Therefore, further investigations should be Record: Journal of the Transportation Research Board
No. 1816, Washington D.C., pp. 125–136.
made to explore the FWD test parameters to Mateos, A. & Snyder, M.B. 2002. Validation of Flexible
determine a testing scheme that more accurately Pavement Structural Response Model with Data from
represents traffic conditions. the Minnesota Road Research Project. In Transpor-
tation Research Record: Journal of the Transporta-
tion Research Board, No. 1806, Washington, D.C., pp
ACKNOWLEDGEMENTS 19–29.
Robbins, M.M. 2012. New Methods for Predicting Criti-
This project was co-sponsored by Kraton Per- cal Tensile Strains in an M-E Framework, Doctoral
formance Polymers Inc., Shell Oil Products, Lake Dissertation, Auburn University.
Robbins, M. & Timm, D. 2009. Effects of Strain Pulse
Asphalt of Trinidad and Tobago Ltd and the fol- Durations on Tensile Strain in a Perpetual Pavement.
lowing departments of transportation: Alabama, Proc. International Conference on Perpetual Pavement,
Florida, North Carolina, Oklahoma, South Caro- 2009. Columbus, OH.
lina, Tennessee. The Federal Highway Administra- Schapery, R.A. and Park, S.W. 1999. Methods of
tion also sponsored this research. Interconversion between Linear Viscoelastic Mate-
rial Functions. Part I—An Approximate Analytical
Method. In International Journal of Solids and Struc-
REFERENCES tures, Vol. 36, pp. 1677–1699.
Schwarzl, F.R. 1975. Numerical Calculation of Stress
Al-Qadi, I.L, Xie, W. & Elseifi, M.A. 2008. Frequency Relaxation Modulus from Dynamic Data for Linear
Determination from Vehicular Loading Time Pulse to Viscoelastic Materials. In Rheological Acta, Vol. 14,
Predict Appropriate Complex Modulus in MEPDG. pp. 581–590.
In Journal of the Association of Asphalt Paving Tech- Solanki, P., Zaman, M.M. & Muraleetharan, K.K. 2011.
nologists. Proc Technical Sessions, Vol. 77, pp. 739– Validation of Backcalculated Modulus Values from
772, Philadelphia, PA. Falling Weight Deflectometer Through Instrumenta-
ARA Inc., Eres Consultants Division. 2003. Develop- tion on I-35 in Oklahoma. In Transportation Research
ment of the 2002 Guide for the Design of New and Record: Journal of the Transportation Research Board,
Rehabilitated Pavement Structures; Appendix CC-3, No. 2227, Washington, D.C., pp. 79–86.
Updated Traffic Frequency Calculation for Asphalt Taylor, A.J. & Timm, D.H. 2009. Mechanistic Charac-
Layers, Final Document, NCHRP 1-37 A. terization of Resilient Moduli for Unbound Pavement
ARA Inc., Eres Division. 2004. Part 2, Chapter 2: Mate- Layer Materials. Report No. 09-06, National Center
rial Characterization. Guide for Mechanistic-Empirical for Asphalt Technology, Auburn University.
Pavement Design of New and Rehabilitated Pavement Timm, D.H. 2009. Design, Construction and Instrumenta-
Structures, NCHRP. tion of the 2006 Test Track Structural Study, Report
Chatti, K., Kim, H. B, Yun, K.K., Mahoney, J.P., & Mon- No. 09-01, National Center for Asphalt Technology,
ismith, C.L. 1996. Field Investigations into Effects of Auburn University.
Vehicle Speed and Tire Pressure on Asphalt Concrete Timm, D.H. & Priest A.L. 2006. Material Properties of
Pavement Strains. In Transportation Research Record: the 2003 NCAT Test Track Structural Study. Report
Journal of the Transportation Research Board, No. No. 06-01, National Center for Asphalt Technology,
1539, Washington, D.C., pp 66–71. Auburn University.
Christensen, R.M. 1982. Theory of Viscoelasticity, Aca- Van der Poel, C.A. 1954. General System Describing the
demic Press, New York. Viscoelastic Properties of Bitumens and its Relation
Daniel, J.S. & Kim, Y.R. 1998. Relationships Among to Routine Test Data. In Journal of Applied Chemis-
Rate-Dependent Stiffnesses of Asphalt Concrete Using try, Vol. 64, pp. 221–236.
Laboratory and Field Test Methods. In Transporta- Washington State Department of Transportation
tion Research Record: Journal of the Transportation (WSDOT) 2005. EVERCALC: User’s Guide Pave-
Research Board No. 1630, Washington, D.C., pp. 3–9. ment Analysis Computer Software and Case Studies,
Ferry, J.D. 1980.Viscoelastic Properties of Polymers. 3rd Washington State Department of Transportation,
Ed., John C. Wiley and Sons, New York. Olympia, WA., 2005.

726
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Optimization of deflection bowl measurements

V. Le Boursicaud, J.-M. Simonin & P. Hornych

IFSTTAR, MAST(LAMES), Bouguenais Cedex, France

ABSTRACT: In France, the curviameter and the deflectograph are widely used for pavement deflec-
tion measurements. However, some comparisons made with reference devices such as anchored displace-
ment sensors have shown that the rolling devices often measure different deflection bowls. The working
principle of these rolling vehicles and some simplifying assumptions explain the differences and lead to
measurement biases. Thus, the maximum deflection is often underestimated and moduli determined by
back-calculation are not very realistic. For this reason, measurements are mainly used to identify homo-
geneous zones.
An approach, based on a three parameter model of the deflection bowl under a single wheel (Jouve
and Martinez, 1990), has been developed to correct curviameter and deflectograph measurements. By
using this model and superimposing the loads, the pavement deflection under the complete vehicle can be
simulated. Then, the model parameters are optimized. Thanks to this model, a corrected deflection bowl,
under a single-wheel, can be deduced and used for back-calculation, or to calculate different indicators
similar to those used for the FWD.

1 INTRODUCTION due to their capacity to make practically “continu-

ous” measurements at network level. The deflecto-
The French road network has a total length of graph has a measuring speed of about 4 to 8 km
more than 1 million kilometers and is widely used per hour, and the curviameter a fixed speed of
for passenger and freight traffic. Its preservation 18 km per hour, and measurements are usually
is crucial and has an impact on the national econ- made every 5 meters.
omy. Budget cuts force road managers to optimize However, these methods also present several
maintenance funds and to program works over limitations. The first and main one is that in cur-
several years. Improvement of maintenance poli- rent practice, the exploitation of the results is
cies also requires a better knowledge of structural relatively limited. Generally, only the maximum
and surface characteristics of the road network. deflection and potentially the radius of curvature
are analyzed, and only a “qualitative” evaluation
of the pavement stiffness is obtained. No back-
2 CONTEXT calculation is performed. Other drawbacks are the
relatively high cost of the tests, and the slow meas-
First, pavement monitoring evaluates the struc- urement speed, which requires to close the lane
tural capacity, and then the surface characteristics. to traffic during the measurements. Due to these
The measurement of the deflection under a load is limitations, the use of deflection measurements for
one of the most appropriate and widely used Non- road inspection tends to decrease in France.
Destructive Test methods (NDT) to evaluate the The objective of the work presented in this
structural characteristics of pavement layers. How- paper is to improve the interpretation of deflec-
ever, deflection measurements can be performed tograph and curviameter measurements, in order
using different devices such as the deflectograph, to promote the use of these tests. The proposed
the curviameter, the Falling Weight Deflectometer approach consists in using the full deflection
(FWD), or more recently the Traffic Speed Deflec- bowl (which is recorded by the apparatus, but not
tometer (TSD). (Baltzer, et al., 2010). exploited in current practice), to perform back-
calculations, like for the FWD, to determine the
mechanical properties of the different pavement
2.1 Current use of deflection measurements
layers. However, to obtain exploitable results, the
In France, deflection measurements are performed measured deflection bowls must first be corrected,
mainly using the deflectograph and curviameter. to eliminate some experimental biases, associated
These rolling devices are preferred to the FWD, with each apparatus.

727
2.2 Deflection measurement devices The deflection measured by the curviameter is
then obtainedby a double integration of Equation
This part describes the deflection measurement
(1). This requires the introduction of two bound-
devices studied along this work, the Curviameter
ary conditions: the deflection, and the slope of the
and the deflectograph. The Curviameter is used in
deflection bowl (derivative of the deflection) are
France mainly on the motorway network, due to
both assumed to be zero at a distance of 3 m from
its better accuracy for stiff pavements, presenting
the load. These assumptions have an important
low deflections. The deflectograph is used mainly
influence on the calculation of the deflection bowl.
for pavements for low to medium traffic, with a
They are acceptable for flexible pavements, but lead
granular or bituminous base. These devices are
to an underestimation of the deflection for rigid
used at network level, to check the homogeneity of
pavements with cement treated layers, for which the
road sections, and detect possible damaged zones.
width of the deflection bowl largely exceeds 3 m.
This study presents analyses which can be used at
network level.
2.2.2 The deflectograph
The deflectograph (Figure 2) measures the verti-
2.2.1 The curviameter
cal displacement under each dual-wheel of the rear
The curviameter (Figure 1) measures the velocity
axle of a heavy truck moving at constant speed
of the vertical displacement under the dual-wheel
(between 4 km/h and 8 km/h depending on the type
of the half rear axle of a heavy truck moving at
of deflectograph) using an automated Benkelman
constant speed (18 km/h), using geophones. The
beam. Four versions of the deflectograph have been
deflection is obtained by integration of the geo-
successively developed: D02, D03, D04 and deflec-
phone measurements. A chain system ensures the
tograph Flash. The accuracy of the measurements
placement of the geophones on the pavement sur-
has been improved with each version, by increasing
face, in front of the rear axle, and their retrieval,
the length of the truck, and the length of the meas-
after the axle has moved. The chain supports three
urement beam, to reduce the influence of the front
geophones spaced every 5 m. Each deflection bowl
axle load on the deflection. Depending on the truck
measurement begins 1 m before the dual-wheel
used, the spacing of the measurement points is
and ends 3 m after its passage, with a resolution of
between 3 and 6 m. The rear axle of the deflecto-
100 measurement points. The rear axle of the cur-
graph is loaded at 130 kN (Lacroix, 1963).
viameter is generally loaded at 130 kN, but can be
As for the curviameter, some assumptions have
adjusted between 80 and 130 kN (Paquet, 1977).
been made for the analysis of the measurements: it
To determine the deflection measured by the
is assumed that the deflection is zero at the start of
curviameter, the first step consists in calculating
the measurement and that two of the three support
the vertical acceleration incurred by the pave-
points of the beam are fixed (zero vertical deflection).
ment under the rear axle of the curviameter. It is
Figure 3 shows the geometry of deflectographs.
expressed as:
The calculation of the deflection basin (De bois-
soudy, et al., 1984) can be done using a deflection
dS (t ) t

γ (t ) a + bS (t ) c ∫S (t ) dt (1) function, depending on the geometry of the device

dt 0
and the displacements of points of application A,
B and C. The displacements of these points at the
where a, b and c are calibration parameters of the beginning and at a distance n of the load must be
geophone (sensitivity, depreciation and cut-off fre- known (dAn, dA0, dBn, dB0, dCn, dC0). Thus, for
quency of the geophone); and S(t) is the signal of a given beam, the deflection Wn at a distance n of
the geophone (speed). the load is expressed as:

Figure 2. Deflectograph flash with the courtesy of

Figure 1. Curviameter with the courtesy of BRRC. CEREMA.

728
 first bias, common to both apparatus, is due to the
loading. Indeed, the measurement truck has sev-
eral axles, and the deflection under a dual-wheel is
increased, by the influence of the loads of the sec-
ond dual-wheel and of the front axle. The second
bias depends on the device used:
• For the curviameter, when the signals of the
geophones are integrated, it is assumed that the
deflection and its derivative are both null at a
distance of 3 m after the load.
• For the deflectograph, two other biases are intro-
Figure 3. Geometry of the deflectograph.
duced. The first hypothesis is that the deflection
at the beginning of the measurement is null. The
second is that the points B and C supporting
the beam (Figure 3) remain fixed when the vehi-
cle moves. However, these support points are
affected by small vertical displacements, when
the pavement deforms under the vehicle loads.

3.1 Evaluation of the measurement error

As explained above, each deflection measuring equip-
ment introduces measurement errors, due to various
simplifying assumptions. To quantify these meas-
Figure 4. HWD with the courtesy of STAC. urement errors, a theoretical study was conducted,
for the curviameter, and the deflectograph Flash
(the most recent type) using the French pavement
design software ALIZE (Balay, 2013). This software
OC + OB is based on a Burmister’s model (infinite multilayer
Wn = * (dAn − dA
A) linear elastic). For each apparatus, the measurement
AB
+ (dC C )
error was evaluated in three steps. First, the deflec-
Cn − dC
(2) tion bowl under each complete truck was calculated
AB + OC + OB with ALIZE. Then, the mode of operation of each
−
AB apparatus was simulated, in order to calculate the
* (dB B)
Bn − dB deflection effectively recorded by each apparatus:
• For the deflectograph, the geometry of the
measurement beam, and its movement under the
2.2.3 FWD load of all the vehicle axles was simulated.
The FWD (Figure 4) measures the vertical displace- • For the curviameter, the vertical displacement
ment under an impulse load. Nine to 15 geophones of the geophones under all the axles of the vehi-
record the deflection bowl at different distances cle was calculated. Then, the geophone signals
from the load. This device is available in three were integrated, and the “boundary conditions”
versions: Light, “standard” and Heavy Weight (deflection and its derivative null at 3 m from the
Deflectometer. The distance between measurement load) were introduced.
increments depends on the operator but it is gener- • Finally, the simulated deflections under each
ally around 100 m. The load applied by the FWD apparatus were compared with the “reference”
varies between 30 and 130 kN for the FWD and can deflection obtained under a single standard
reach 240 kN for the HWD. (Broutin, 2010) dual-wheel axle, loaded at 65 kN. This load cor-
responds to the reference load of the French
pavement design method.
3 THEORICAL EVALUATION OF DEVICES
3.2 Example of measurement error for an
Paragraph 2.2 summarizes the working principles
asphalt pavement
of the curviameter and deflectograph. The main
function of these high-speed devices is the meas- The measurement errors induced by each appara-
urement of the pavement deflection under a single tus, calculated using ALIZE, for an asphalt pave-
axle. Nonetheless, each device introduces measure- ment, with 22 centimeters of asphalt materials are
ment biases related to its mode of operation. The illustrated on Figure 5. This figure compares:

729
 Distance frfm the load(~) The study has been conducted for the differ-
-1 0 1 ent deflectographs (02, 03, 04, Flash) and the
curviameter.
············ It has been noted that the higher the stiffness of
the pavement is, the more the “true” deflection is
underestimated.
§f ----~~~~----~~----~~----------~--~ This study also showed two accuracy levels:
} -~~,___~~........ • Measurements by deflectograph 04 (and 02 and
~ ------~~~--~~------~~~~------------ 03) induce large relative errors, which can exceed
§
·B -------+•H-~~--------~~-----------------­
60% (100% for 02) for deflection basins meas-
<1)
p;;:: ured on rigid pavements and even on an asphalt
0 ~~--~~~------~~---------------------- pavement at low temperature.
• Lower relative errors are obtained with the cur-
viameter and the deflectograph Flash. Errors
between the theoretical bowl and the basin
measured by these devices are usually between
- - • Absolute deflection under the curviameter
1% and 40% for flexible and asphalt pavements,
- - - Absolute deflection under the deflectograph Flash but can go up to 60% for rigid pavements.
• • • • • • Measured deflection by the curviameter As a conclusion, only the most accurate appa-
- - Measured deflection by the deflectograph Flash ratus the curviameter and the Flash deflectograph
were considered for correction.
- - - Absolute deflection under French standard dual-
wheel (65 kN)

Figure 5. Comparison of theoretical and effectively 4 PROCEDURE FOR DEFLECTION

measured deflections bowls—Asphalt pavement. BOWL CORRECTION

The analysis of the measurement errors has led to

• The absolute deflections calculated under the six
conclude that direct measurements obtained using
loads of each complete truck. The two deflection
the Curviameter and deflectograph present a signif-
basins are slightly different because the distance
icant bias, and cannot be used without correction
between the two axle loads are different and the
for such applications as back-calculation of pave-
speed of the trucks influences the bituminous
ment layer moduli. For this reason, it was decided
material properties.
to search for correction methods, which could
• The deflections bowls effectively measured by the
eliminate some of the measurement biases, such as
curviameter and the deflectograph Flash, taking
assuming that the deflection is null at 3 m from the
into account their respective measurement errors.
load, which is clearly not valid for stiff pavements.
• The absolute deflection under a French stand-
To correct the measurements, two methods were
ard dual-wheel half-axle (65 kN). This last
considered:
deflection bowl represents the “reference” that
each apparatus should measure. • The first possible method consists in defining a
theoretical model of the pavement structure, and
Figure 5 shows that due to their operating prin-
fitting it to the deflections measured by the curvi-
ciple, both apparatus introduce large measurement
ameter or deflectograph. Then, this “fitted” model
errors, and underestimate the “reference” deflec-
can be used to simulate the deflection under a ref-
tion under a French standard dual-wheel (65 kN),
erence, 65 kN, dual-wheel. However, this approach
that they should measure.
is not easy to apply, because the structural charac-
In order to better assess the relative error
teristics of the pavement (nature and thickness of
between the ‘theoretical’ maximal deflection under
the pavement layers) have to be known.
standard dual-wheel and the maximal deflection
• The second method is based on the use of a sim-
measured by each device, a parametric study was
ple mathematical function (with few parame-
conducted. For each pavement types (flexible,
ters), describing the shape of the deflection bowl
asphalt, semi-rigid and rigid), 25 different struc-
under a single-wheel. By taking into account
tures have been studied including:
the operating principle of each apparatus, the
• Five different types of subgrade with a Young’s deflection bowl measured by the complete vehi-
modulus ranging from 50 MPa to 120 MPa; cle is calculated. Then, the parameters of the
• Five temperatures of asphalt materials (10°C, function are optimized, by minimizing the dif-
15°C, 20°C, 25°C and 30°C). ference between the measured and simulated

730
 bowl. The fitted function is then used to calcu- 2Rd 2
late the deflection under a dual-wheel, which WA ( x ) = (8)
represents the corrected deflection. This second x 2 2Rd
method, which is simpler to apply (no informa-
tion about the pavement structure is needed), where d = Maximum deflection; and R = Radius of
has been chosen for the correction process. curvature at the top of the deflection bowl
Jouve and Martinez’s model, called here “J&M”,
(Martinez & Jouve, 1990) uses three parameters to
4.1 Choice of the deflection bowl shape function express the deflection bowl as:
Three models describing the shape of the deflec-
d
tion bowl were successively assessed. WJM ( x ) = n
(9)
⎡ 1 ⎛ x⎞ ⎤
2

4.1.1 Hogg model (Hogg, 1944) ⎢1 + ⎥

In this model, the pavement is considered as a thin ⎢⎣ 2 n + 1 ⎝ λ ⎠ ⎥⎦
plate of thickness h1, of Young modulus E1 and
Poisson’s ratio v1 resting on a subgrade, of infinite where d = Maximum deflection; λ = Abscissa of
depth, of Young modulus Es and Poisson’s ratio vs. the inflection point of the deflection bowl; and
The pavement deflection is expressed as: n = exponent related to the shape of the deflection
bowl.
0. *P
WH ( x ) = C ( x L ) (3) 4.1.3 Model comparison
L0 K
The three models describing the deflection bowl
have been evaluated, by comparison with deflec-
where P is the load; tion bowls calculated with the design software
K is the soil stiffness expressed as: ALIZE, assumed as the reference. The parameters
of each model were fitted by a least squares opti-
Es ( −
2E ) mization procedure.
K= s

( + s )( − s ) (4)
Figure 6 compares the deflection bowl under a
single-wheel calculated by Alizé (simulated meas-
urement) and those fitted with each model, in the
case of a flexible pavement structure. These four
L0 is the characteristic length defined by:
basins are very similar at short distance from the
load (0 to 2 m).
1/ 3
⎡D⎤ However, at a certain distance from the load,
L0 = ⎢ ⎥ (5) Hogg’s model and Autret’s model diverge from
⎣K ⎦
the reference deflection bowl. In fact, function
C(x/L0), defining Hogg’s model, becomes negative
with:

E1h13 ,, Distance from the load (m)

D= (6)
12 ( − )
~v

r I
.1- .1
.
~ ~ I ~

where C(x/L0) is the function characterizing the

"

-
"
}_. ~.,::. :::=----=--
shape of the deflection bowl, expressed as: ,... A'"~ .,
~ v

C ( x / L0 )
1 r /'
= ⎡⎣ 0. (x / L ) 2
− 0. ( x / L ) ⎤⎦8
..,,
~ v
I
--..,,.. I I
0 0
-.r

ln ( x / 2L
L0 ) + 1 − 0.08738 ( / L0 ) (7)
2

0.015621( / L0 ) + 0.0036748 ( / L0 )
.J V

− 0.015621
4 5 I"
35
− 0.00043392 ( / ) 0.000008447 ( x / L0 ) + …
6 8
0 • • • • • • Absolute deflection under a single-wheel (32,5 kN)
• Autret's model
- - • Hogg's model
4.1.2 Empirical models - - Jouve & Martinez's model
Two empirical expressions of the deflection bowl
have also been considered. Autret’s model (Autret, Figure 6. Comparison of the three optimized deflec-
1969) expresses the deflection as: tion bowl models.

731
when x/L0 is higher than 2.95. This leads to a nega- 2. Superimposition of the load effects to simulate
tive deflection. In Autret’s model, parameters R the complete truck and to obtain the complete
and d characterize the shape of the deflection bowl deflection basin.
(maximum deflection and radius) for x = 0, but 3. Application of the working principles of each
no additional parameter controls the shape of the device to obtain the simulated deflection bowl.
deflection bowl far from the load, which cannot be 4. Comparison of the simulated and measured
predicted very accurately. deflection bowls by calculating the index I in
The three models give good results at a short Equation 10.
distance from the load. However, for distances
The optimization can be validated if the index I
over 2 m, only J&M’s model estimates correctly
satisfies Equation 10:
the deflection bowl. Consequently, it is the only
model which can correctly estimate the deflection
induced by the front axle on the sensors move- =
∑S 2
≤ 5% (10)
ment. The two other models introduce an error on
the deflection of the front axle.
∑S
Reef
2

Based on these limitations, it was decided to With S the sum of the squared deviation
select J&M’s model for the correction of the curvi- between the simulated deflection bowl and the
ameter and deflectograph measurements. measured one.
With SRef the squared sum of the deflection
4.2 Correction method basin measured by the devices.
Then, I is minimized by an optimization algo-
The algorithm developed to correct the measured rithm to determine the three parameters df, λf and
deflection bowls is summarized on Figure 7. nf. To avoid abnormal solutions, constraints are
The first stage is the initialization of the param- applied on the parameters λ and n. A study carried
eters of the J&M’S model. They are set as d equal out in 1993 by Jouve and Martinez concluded that
to the maximal deflection of the basin measured the value of n should vary between 0.5 and 2, and
by the devices (actual measurement, or measure- that the inflection point λ should be located at a
ment simulated by the software Alizé), n = 1 and distance of more than 0.2 m from the center of the
λ = 1. deflection bowl (outside the area where the load
Then the processing includes four steps: is applied). These recommended boundaries were
1. Calculation of the deflection bowl under a sin- adopted in this study.
gle load (32,5 kN) using J&M’s model. The three final variables (df, λf and nf) are then
used to calculate the corrected deflection basin for
different load configurations (single-wheel for the
calculation of indicators and dual-wheel for back-
~ Initial values of parameters d0 , A0 , n 0 calculation for pavement design).

"' 4.2.1 Example of correction of deflection bowl

Simulation of the deflection basin .., To highlight the benefits of the correction method,
under the complete vehicle " three deflections basins were compared:
by load superposition
• The reference basin calculated with ALIZE
I
using dual-wheel load (65 kN).

~
Simulation of measurement principle • The measured basin from the devices, taking
of each device into account the multiple loads and the working
I principle of each device.
Optimization of parameters d, Aand n • The corrected basin deduced from the final
J&M’s parameters applied with the dual-wheel
~ Least squares method load. Final J&M’s parameters are obtained
"' using the optimization procedure of the meas-
ured bowl.
Application of the fitted model to Figure 8 and Figure 9 show the results obtained
calculate the corrected deflection for a flexible pavement and a thick asphalt pave-
basin for different load cases ment, with a PF2 subgrade (modulus 50 MPa).The
comparison of the three basins shows that:
Figure 7. Algorithm of deflection bowl correction • The corrected basins are much closer to the bowl
using J&M’s model. under a dual-wheel (65 kN) than those measured

732
 Distance from the load (m) zero at a certain distance from the load. In reality,
-2 -1 0 1 2 3 at this distance, the deformation of deeper layers is
still influential, and the distant part of the deflection
basin cannot be predicted accurately.
,.-._
8 ----------~~~~----~~--~-A~~-------

l--------~ffi-r-~~~~--------­ 4.3 Back-calculation

c To validate the correction procedure, it was also
.g decided to perform back-calculations of layer
u ------------~~-r..-~--------------------­
<1)

~ -------------b&-~F---------------------- moduli, using the deflections with and without

0 correction obtained for each device. The back-
calculations were performed using the dedicated
module of the ALIZE software. The calculations
• • • • • • Absolute deflection under a dual-wheel (65 kN)
were done assuming a dual-wheel load.
- - - Deflection bowl measured Curviameter
- - Deflection bowl measured Deflectograph Flash
To determine coherent solutions, Boundaries
• Deflection bowl corrected from the Curviameter were imposed for the modulus of each layer:
- - • Deflection bowl corrected from the Deflectograph Flash
• For the flexible pavement, the modulus of the
unbound granular material layer was assumed
Figure 8. Comparison between deflection bowls meas- to vary between 50 and 250 MPa.
ured by each device, with and without correction and
under a dual-wheel; flexible pavement.
• For the bituminous pavement, the modulus of
the second bituminous layer was assumed to
vary between 7000 and 12,000 MPa.
Distance from the load (m)
-1 0 1 2 3 These layers are marked with an asterisk in
-
-. -·······
/ ____ ...... Table 1 and Table 2. These tables indicate the mod-
\ ~~-- uli determined for each pavement structure and
'
C)---'~--~~·n~r---------~L~·'~~~~~~··-·-·_·_·~~-
~~ ,. the relative errors (in brackets). The original elas-
:3 ', / " • , •• ••• .,.,- - tic moduli of the pavement and those determined
l- ' '~ ,'/ / ..···· / . . . .
~'/ ..··· -"'
by back-calculation (with or without correction)
using the Alizé software are compared.
~:: ------~~~~/
~~,~~--~.,~---------------- .. Table 1 and Table 2 show that back-calculated
·~
0 -
~ .··
· ·· /
moduli based on uncorrected measurements of the
~ cr. - ..... • / deflectograph and curviameter are not realistic.
o ------~vtt•v~~~-~,--------------------------
Soil modulus is particularly overestimated due to
./ the low level of the maximum deflection.
Determined moduli based on corrected deflec-
• • • • • • Absolute deflection under a dual-wheel (65 kN) tion measurements are more realistic. Soil modulus
- - - Deflection bowl measured Curviameter
- - Deflection bowl measured Deflectograph Flash is well estimated. The maximum difference with the
• Deflection bowl corrected from the Curviameter reference value is only 15 MPa. The errors on the
- - • Deflection bowl corrected from the Deflectograph Flash
estimated moduli are usually less than or equal to
20%. These errors include a bias due to the back-
Figure 9. Comparison between deflection bowls meas-
ured by each device, with and without correction and Table 1. Relative errors between original Young modu-
under a dual-wheel; thick asphalt pavement. lus and modulus determined using back-calculation; flex-
ible pavement.
by the devices. The best results are obtained for
the flexible pavement. Original Curviameter Deflectograph Flash
• The basin shape is well restituted for both devices; Young
• A nearly constant difference in deflection appears Modulus Measures Correction Measures Correction
(MPa) (MPa) (MPa) (MPa) (MPa)
between the deflection bowl under a dual-wheel
and the corrected deflection bowl. The more rigid 7000 214 5625 11200 8179
the pavement is, the bigger the gap is,. (97%) (20%) (60%) (17%)
These conclusions are valid both for the curvi- 600 679 505 335 619
ameter and the deflectograph Flash. (13%) (16%) (44%) (3%)
This constant gap might come from the fact that 150* 136 122 52 96
(9%) (19%) (65%) (36%)
the two devices have a limited measurement length
50 121 45 141 58
(3 m for the curviameter and only 1.5 m for the deflec- (142%) (10%) (182%) (16%)
tograph Flash). Thus, the deflection is assumed to be

733
Table 2. Relative errors between original Young modu- Comparisons with theoretical deflection basins,
lus and modulus determined using back-calculation, calculated with the software ALIZE, have shown
thick asphalt pavement. that for flexible and thick asphalt pavements, the
corrected deflection basins are quite close to the
Original Curviameter Deflectograph Flash
Young
absolute deflection under a dual-wheel.
Modulus Measures Correction Measures Correction As a consequence, the corrected deflections can
(MPa) (MPa) (MPa) (MPa) (MPa) be used for back-calculation, and give layer moduli
in good agreement with their real values (relative
7000 7194 6386 8625 7105 error generally less than 20%). This opens new
(3%) (9%) (23%) (1.5%) perspectives, for a better exploitation of deflecto-
9000* 8248 7334 8529 10792 graph and curviameter measurements, for applica-
(8%) (18%) (5%) (20%) tions such as evaluation of pavement residual life,
9000 13502 7575 15026 9296 or design of pavement reinforcements.
(50%) (16%) (67%) (3%) In continuation of this work, it is planned to val-
50 92 65 84 38 idate further the proposed correction procedure by
(84%) (30%) (68%) (24%)
comparison with measurements on real pavements,
of known characteristics, and with FWD tests.
calculation process. Thus, the difference between the
measured deflection bowl and those obtained with
the back-calculated modulus is less than 4%. This ACKNOWLEDGEMENTS
value is lower than the reproducibility of the devices.
The authors are grateful to the Belgian Research
Road Centre for providing them valuable and
5 CONCLUSION accurate information about the FWD and the
curviameter.
Non-destructive devices, such as the curviameter
and the deflectograph, are widely used in France
for monitoring of pavement deflections. Some com- REFERENCES
parisons made with reference sensors (anchored
Autret, P., 1969. Utilisation du produit Rd pour
displacement sensors) have shown that the roll- l’auscultation des chaussées à couche de base traitées..
ing devices frequently tend to under estimate true Bulletin de Liaisons du Laboratoire des Ponts et
deflection bowls. The working principles of these Chausséees n°42, Décembre, pp. 67–80.
devices, and some simplifying assumptions in the Balay, J.-M., 2013. Manuel d’utilisation du logiciel
processing of the results lead to measurement ALIZE; Version 1.5. p. 184.
biases. The back-calculation of layer moduli with Baltzer, S. et al., 2010. Continuous bearing capac-
these bowls gives inconsistent results due to an ity profile of 18,000 km Australian road network in
underestimation of the maximum deflection and 5 months. 24th ARRB Conference.
a modification of the shape of the basin. There- Broutin, M., 2010. Assessment of flexible airfield pave-
ments using Heavy Weight Deflectometers. In: Paris:
fore, deflection measurements obtained with these Thèse de doctorat, Ecole Nationale des Ponts et
devices cannot be fully exploited, and are generally Chaussées, p. 370.
used only as “qualitative” indicators of the stiff- De Boissoudy, A., Gramsammer, J., Keryell, P. & Pail-
ness of the pavement, and of its evolution. lard, M., 1984. Appareils d’auscultation. Le déflectog-
To overcome these limitations, procedures for raphe 04.. Bulletin de liaison Laboratoires des Ponts et
correction of curviameter and deflectograph deflec- Chaussées n°129, janvier-février.
tion basins have been developed. These procedures Hogg, A., 1944. Equilibrium of a thin slab on an elastic
are based on the use of simple functions describing foundation of finite depth. Philosophical Magazine,
the shape of the deflection basin. After evaluation 35(243), pp. 265–276.
Lacroix, J., 1963. Déflectographe pour l’auscultation rap-
of 3 different functions, the Jouve and Martinez ide des chaussées. Bulletin des Laboratoires des Ponts
model, depending on three parameters, has been et Chaussées n°3, Septembre-Octobre, pp. 1–12.
selected to develop a correction procedure. Martinez, J. & Jouve, P., 1990. Rational determinination
With this correction, a deflection basin inde- of pavement deflection from Benkelman beam meas-
pendent of the device, which corresponds to a urements. Trondheim, 3rd Int. Conf. on the Bearing
single-wheel (useful for back-calculation or cal- Capacity of Roads and Airfields.
culation of various performance indicators) or to Paquet, J., 1977. Un nouvel appareil d’auscultation des
dual-wheels (standard equivalent axle load), can be chaussées: Le Curviamètre. Revue générale des Routes
defined. The correction method is validated if the et des aérodromes, Avril, pp. 79–107.
solution satisfies Equation 10.

734
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Evolution of asphalt modulus from falling weight deflectometer tests

and challenges associated with its interpretation and applications:
A case study using LTPP data

M. Oshone, M. Elshaer, E. Dave & J.S. Daniel

University of New Hampshire, Durham, NH, USA

ABSTRACT: Dynamic modulus is a key parameter used in mechanistic response and performance
models to characterize the linear viscoelastic properties of asphalt concrete. Typically, only the initial
mixture condition during mix design phase or plant production is characterized. However, it is known the
modulus of the asphalt layer varies during service as the mixture experiences traffic and environmental
loading as well as other physio-chemical changes (such as, oxidative aging). Although such information is
available in the Long Term Pavement Performance (LTPP) database in terms of modulus back-calculated
from Falling Weight Deflectometer (FWD), the challenges associated with the determination and nor-
malization of temperature and frequency inhibits its use in the performance models requiring viscoe-
lastic characterization. This paper presents an approach to tie and normalize in-service back-calculated
modulus values to a specific temperature and frequency. The study focuses on a test section in Oklahoma
and information needed for the study is obtained from the LTPP database. The modulus back-calculated
from the first FWD test on pavement section is compared to the as-built built complex modulus master
curve to yield the equivalent loading frequency which is assumed the same for subsequent FWD measure-
ments. The time temperature superposition principle is applied to modulus measurements for temperature
correction. Modulus indices determined by comparing initial reference modulus to subsequent measure-
ments are used to understand the evolution of in-service complex modulus master curves of mixtures
over time. The feasibility and limitations of the proposed analysis method and data collection scheme
featured in LTPP are also investigated. Finally, the variation in in-service modulus is correlated to field
distress data. The proposed method was found promising to establish a reference frequency and normal-
ize the temperature difference among FWD measurements and subsequently develop modulus indices.
The results also showed that the variation in environmental changes (temperature and moisture) can have
a profound effect on the degree to which the layered-elasticity and time temperature superposition prin-
ciple is a valid model for the pavement system based on the deflection encountered. To mitigate these
limitations associated to validity of the time temperature superposition principle, the authors recommend
establishing a test temperature range that results in a linear relationship during FWD testing.

1 INTRODUCTION AND BACKGROUND Currently, many state highway agencies use

the Falling Weight Deflectometer (FWD) test as
The Long Term Pavement Performance Database a tool for pavement performance monitoring.
(LTPP) encompasses in-service pavement perform- The test entails collection of deflection data for a
ance data from various test sections of asphalt pavement structure by dropping a weight on the
and cement concrete pavements (Elkins et al. pavement to understand its load response char-
2013). The data collection plan for LTPP envi- acteristics (Schmalzer 2006). Deflection measure-
sioned development of models that explain why ments, corresponding temperature profiles and
some pavements perform better than others. Test modulus values back-calculated from FWD are
sections with various structure, material, environ- available in LTPP for various test sections. The
ment, traffic, field distress and maintenance prac- temperature profile available in LTPP is collected
tice information are available in LTPP. This data at the same time as the FWD testing and the back-
has been used by researchers to link pavement calculated pavement layer modulus is calculated
related details to pavement performance and to from the FWD deflection measurements using
gain a better understanding on how these variables three programs namely EVERCALC, MODTAG
affect pavement performance (Mensching et al. and Modulus. The back-calculation for the
2016, Elshaer et al. 2017). study section (Oklahoma) is performed using the

735
EVERCALC program. The program performs linear elasticity assumption in the back-calculation
layer elastic analysis to iteratively adjust a set of process (Von Quintus et al. 2015).
trial layer modulus to match predicted and meas-
ured deflections within some tolerable error. The ATAF = 10 slope
l (Tr Tm )
(1)
general procedure for FWD deflections and back-
calculated asphalt concrete moduli is presented by Where,
Von Quintus et al. 2015. ATAF = Asphalt Temperature Adjustment Factor
The asphalt layer modulus is a key material Slope = Log (Mr) = intercept + slope T
property input to mechanistic response and per- Tr = Reference Temperature, °C
formance models. The property is temperature, Tm = Measured Temperature, °C
frequency and loading history dependent. During
the course of the pavement life, it is perceived that To mitigate the aforementioned limitations asso-
the modulus deteriorates due to traffic and envi- ciated with back-calculated modulus, this study
ronmental loading whereas aging causes the modu- aims to establish a reference frequency by initially
lus to increase. A cumulative effect of loading and linking FWD measurement to the as-built complex
aging can result in an overall increase or decrease modulus master curve. The as-built complex mod-
in stiffness which can be influenced by factors such ulus is constructed using as-built dynamic modu-
as mixture type, structure, loading and environ- lus values from LTPP and as-built phase angle
ment. A realistic simulation of this variation in which is computed from the slope of dynamic
modulus through the course of the pavement life in modulus master curves to characterize the mate-
the mechanistic response and performance models rial condition during construction. Furthermore,
leads to improved reliability in design and analy- the time-temperature superposition principle is
sis. This study presents a new analysis method for applied for temperature correction. This is done
determination of such information from FWD test assuming the FWD test is performed in the linear
data available in LTPP. The proposed method aims viscoelastic range based on the small strain level
to tie and normalize the modulus values back- observed (Airey et al., 2004): (Gardner et al. 1967).
calculated from FWD to laboratory measured When asphalt mixtures are in the linear viscoe-
dynamic modulus values with respect to tempera- lastic range, they generally exhibit thermorheo-
ture and frequency so that they may subsequently logically simple properties and a time-temperature
be used as an input to mechanistic response and superposition principle can be employed to hori-
performance models. A preliminary study is car- zontally shift results measured at different tem-
ried out on a test section from Oklahoma to evalu- peratures along the time or frequency axis (van der
ate the applicability and the challenges associated Poel, 1955). The amount of time or frequency shift
with the proposed analysis method as it relates to is called the time-temperature shift factor. In this
the data collection scheme in LTPP. case, the as-built time-temperature shift factors
The temperature profile of a pavement is are used to adjust the temperature of the reference
recorded during FWD testing hence, the tem- modulus to the same as any given FWD measure-
perature that corresponds to each back-calculated ments. The temperature correction to reference
modulus can easily be determined. However, the modulus is applied to obtain similar temperature
testing is conducted at different times (different conditions as the conditions in field during FWD
temperatures) over the course of the pavement life measurements. Different temperature values were
and this variation in temperature among different evaluated in this study: temperature at mid-depth
back-calculated modulus values should be normal- of asphalt layer and one-third depth from surface.
ized. Researchers (Ali et al. 1996); (Johnson et al. The evolution in material properties is also stud-
1992); (Ullidtz, 1987); (Kim et al. 1994 & 1995); ied by determining the modulus indices over time.
(Ali et al. 1996), (Chen 2000 et al.) have developed The modulus indices at different times are deter-
regression models on a regional basis to capture mined in terms of the ratio between modulus at
the modulus and temperature relationship and any time to the temperature corrected reference
use them as a tool for temperature correction. The modulus. The primary objectives of this study
LTPP recommends the use of a semi-logarithmic are to 1) Propose a time-temperature superposi-
relationship (Equation 1) for temperature adjust- tion based approach for temperature correction of
ment of back-calculated modulus (Lukanen et al. back-calculated modulus values from FWD tests;
2000). However, the inability to develop a unique 2) Assess the applicability and limitations of this
relationship that works well for all materials and approach as it relates to the analysis method and
climatic zones still presents a challenge. The deter- data collection scheme in LTPP; 3) Detect changes
mination of the equivalent loading frequency to the in pavement modulus throughout the pavement
back-calculated modulus also still remains a chal- life based on FWD measurements available in
lenge due to the utilization of the static load and LTPP to determine effects of aging and field con-

736
ditioning on modulus variation; and, 4) Investigate 6.E+03
the correlation between modulus evolution and
distresses. 0 0 0
0
0 0
E4.E+03 OOo O 0
2 MATERIAL AND METHODS "0"'0 8oce oo o
o OcfJ 0~
0
:2 ooo
A General Pavement Study (GPS) section from
"0
<l.) 0 rti og
CO f6 o 0
<;; o ~ 00 ~o o
o,;~~~~o
Oklahoma is selected for this preliminary study. -s2.E+03
u
The basis for the selection was the availability of a -.; 8
u

~o:l
significant number of FWD tests in LTPP for this
section. The test section was built in June 1984 and o Traffic Mid Lane ~- ~M O dl.
0 Outer Wheel Path 0~ o o0
the first FWD test was conducted in May 1990, six O.E+OO
years after construction. While a narrow time gap 0 10 20 30 40 50
between construction and the first FWD measure- AC Mid Depth Temperature, oc
ment is preferred for establishing initial modulus
values, the quantity and variety of information Figure 1. Back-calculated modulus and mid depth tem-
available after the sixth year made this section pref- perature variation during FWD testing (raw data from
erable for the study. From May 1990 to July 2003 LTPP).
(161 months or approximately 13 years), 144 FWD
tests were carried out on the section along the mid from FWD to a loading frequency. Further it
lane and outer wheel path along same transverse explains the time-temperature superposition
line. In several instances, three or four FWD tests principle as it is applied to normalize the variation
were conducted on the same day but at different in temperature among FWD tests conducted at dif-
times (different temperature profiles). The section ferent times. Modulus indices are used to track the
is out of study since December 2007. evolution of asphalt mixture stiffness in this study,
Information such as pavement structure, traf- these indices are also discussed in this section.
fic, as-built mixture volumetrics, as-built dynamic Asphalt layer dynamic modulus values predicted
modulus values, modulus back-calculated from from ANN prediction model using measured resil-
FWD measurements conducted at different times ient modulus at temperatures of −10, 4.4, 21.1,
and corresponding pavement temperature profiles 37.8, 54.4°C and loading frequencies of 25, 10, 5,
are obtained from the LTPP database in the Data- 1, 0.5, 0.1 Hz are obtained from LTPP. These will
Pave 3.0 software. Important information of the be referred to as as-built modulus as they represent
test section is summarized below. the mixture property at the time of construction.
The modulus at any other given time after con-
struction will be referred to as in-service modu-
3 RESEARCH APPROACH lus. The phase angle corresponding to the as-built
dynamic modulus is not available in LTPP and is
Figure 1 depicts modulus values back-calculated determined using two methods namely, the Hirsch
from FWD tests conducted along the mid traf- model (Christensen et al. 2003) and the slope of
fic lane and outer wheel path and their respective the dynamic modulus master curve (Oshone et al.
pavement mid depth temperature for the study sec- 2017). The phase angle computed from both meth-
tion from May 1990 (referenced as time 0) to July ods were comparable which led to the decision to
2003 (referenced as time 161) as collected from use the phase angle predicted from the slope of the
LTPP. The plot is presented to show the consid- dynamic modulus master curve. Figure 2 shows
erable amount of variation in temperature during the as-built dynamic modulus constructed repre-
FWD tests for the section and to emphasize the senting as-built mixture condition at the required
importance of the temperature adjustment proc- reference temperature. The time-temperature
ess. As described in the previous section, temper- shift factors for the as-built dynamic modulus
ature and frequency normalization is needed to master curve are also shown in Figure 2(b). For
be able implement the back-calculated modulus the as-built condition, a reference temperature
values in mechanistic response and performance that is same as the temperature during the first
models. Moreover, the normalization allows for a FWD measurement is used to avoid the tempera-
direct comparison to be made. Engineers can then ture correction for the initial modulus value. For
better understand how the asphalt mixture proper- example, the master curve shown in Figure 1(a) is
ties change over time. constructed at a reference temperature of 26.3°C
This section first describes the procedure imple- which is the temperature at which the first FWD
mented to associate the back-calculated modulus test is conducted. It is well known that there is

737
Table 1. Test section information. the first FWD measurement to a particular loading
frequency which will be assumed the same for sub-
General information sequent FWD measurements taken on the same
State code 40 (Oklahoma) asphalt layer. It should be noted that the method
SHRP ID 4165 is not used to determine the actual frequency of
GPS- (Lat., Long. (Degrees)) (36.39117, −98.2855) FWD measurement rather it is used to establish a
Functional class Rural principal reference frequency. The six-year time gap between
arterial
construction and the first FWD measurement as
Climatic zone Wet, Non-Freeze
well as the linear elastic and static load assumption
As-built dynamic modulus date 1-Jun-1984
during the back-calculation process will contribute
First FWD Meas. considered 31-May-1990
Last FWD Meas. considered 24-Jul-2003
to the deviation of the determined frequency from
First and last FWD 161 months (approx.
the actual value. In this process a reference is estab-
time duration 13 years) lished based on the first FWD measurement which
No of FWD tests in the 144 measurements is also considered as time zero and the modulus
time duration both along the mid indices for succeeding times are determined based
lane (F1) and outer
wheel path (F3)
(a)
Pavement structure 1.E+05
AC surface layer 69 mm o As-Built Modulus Master Curve at 26.3°C
(Hot mix asphalt dense graded) ,-., e Back-calculated modulus from FWD-1 (26.3°C
AC base layer 140 mm
~ '" ooo
E.,t.E+04 oo
(Hot mix asphalt concrete) oo
Subgrade (silty sand) Semi-Infinite "'
= Point A ~4! 0 0°

AC surface layer mixture

==
"0

:;; l.E+03 00
00
00

NMAS 9.5 mm 00
% Asphalt content 5.0 ·§ 00
00

% Air void 5.0 '"»

= oO
oo
VMA 15.8 Q l.E+02
VFA 68.4 I.E-03 I.E-0 I I.E+O1 l.E+03
Reduced Frequency (Hz)
Traffic (AADTT: Average Annual Daily Truck Traffic) 6
Range of AADTT (1984–2003) 290–327 E 4
(b) Time Temperature Shift Factor
'".:
.su 2
.s 0
a non-uniform temperature variation along the .;:
:c
pavement depth. Researchers have proposed repre- "' OJ)
-2
sentation of pavement temperature profile in terms ....0 -4
of effective temperature (11,12). Such representa- 0 20 40 60
tions include considering 1/3rd or mid depth of the Temperature, °C
pavement temperature as an effective temperature.
Both approaches are implemented in this study. (c) O Temp. adjusted modulus back-calculated
For pavement study section, the FWD test is from FWD-I (25°C)
performed at four load levels (4 drops per load) ~l.E+04 :~<Modulus back-calculated from FWD-2
test (25°C)
with a total of 16 drops (Schmalzer, 2006). Back- E..
calculated moduli from all load levels are available "'= [j]
in LTPP. However, for this study the average back-
calculated modulus corresponding to the four
==
"0

:;; 1.E+03
Point A from
Figure 1 (a)
~
w
40 kN load drops is used because it simulates axle ·§
loads that are closest to values used in majority
of pavement analysis and design approaches. The
'"»
=
Q
frequency that matches the initial back-calculated
modulus (time 0) is determined using the sigmoi- I.E+02
l.E-03 l.E-0 1 l.E+O1 l.E+03
dal equation fitted to the as-built dynamic modu- Reduced Frequency (Hz)
lus master curve. This is illustrated in Figure 2(a),
where the frequency associated with the first FWD Figure 2. (a) As-built dynamic modulus master curve
test (FWD-1, closed circle) is matched to the as- (b) time—temperature shift factors (c) temperature and
built dynamic modulus curve. This is done to link frequency normalization for modulus index computation.

738
on the ratio between the initial back-calculated alignment indicate measurements taken on the
modulus (reference modulus) and subsequent same day at different times to cover a range of pave-
back-calculated modulus values. To allow a direct ment temperatures. Modulus indices computed
comparison between the reference which is tied to from these measurements are first compared. This
the as-built condition and subsequent modulus step was vital to validate the process by observ-
values, a temperature correction has been applied ing the closeness or separation in estimated indi-
using time temperature superposition principle. ces from FWD measurements taken on the same
The as-built time-temperature shift factors are used day but different time (different temperature). The
to adjust the reference modulus to the temperature result showed a difference in computed modulus
for any FWD measurement, Figure 1(c). As an indices implying a noticeable change in modulus
example, Figure 1(c) shows the reference modulus within a day which is inaccurate. It is authors’
(FWD-1, closed circle) measured at 26.3°C shifted hypothesis that these differences could arise from
to 25°C (FWD-1, open circle) which is the same errors associated to FWD testing, back-calculation
as the second FWD test (FWD-2) mid depth tem- or temperature correction process. However due to
perature to normalize the temperature difference the lack of a large set of data within a day, further
between the measurements. investigation could not be made to as it relates to
The same frequency is assumed for subsequent each of the mentioned factors.
back-calculated modulus based on the assumption Figure 3(b) just shows the modulus indices for
that the FWD tests at different times are carried the time frame between 40 to 100 months to better
out a similar manner. The evolution in modulus evaluate the trends over a shorter time period. In
(modulus indices at different times) are determined general, for the overall considered time frame the
in terms of the ratio between modulus at any time modulus indices do not show a distinct incremen-
to temperature corrected reference modulus, this tal or decremental trend which is anticipated due
relationship is shown in Equation 1. Index val- to a cumulative effect of traffic loading and aging,
ues greater than one infer an increase in modulus rather a cyclic pattern is observed. Within cyclic
from the first referenced time whereas index values curve, the slightly higher concentration of points
less than one indicate a drop in stiffness. Reduc- with index values less than one (black line) dem-
tion in asphalt mixture stiffness, especially from onstrates a general stiffness deterioration from the
FWD back-calculation, is often associated with first referenced time. The deterioration indicates a
deterioration of the asphalt mixture due to pres- slightly dominant effect of traffic and environmen-
ence of distresses. The reduced moduli values are tal loading as compared to aging on the asphalt layer
often used to calculate remaining service life of stiffness. It is known that the change in dynamic
pavements. modulus due to seasonal variation (temperature)
follows a cyclic pattern. The cyclic (month to
Modulus index month) variation which is observed in Figure 3(b)
Modulus at any time t (point 2ini Fig 1c) seems to be attributed due such factors related to
=
Temp. correctedd refernce modulus (point 1in
i Fig 1c) seasonal variation rather than actual change in
modulus. However, in this process before comput-
(2)
ing the modulus indices, a temperature correction
has been applied to back-calculated modulus val-
ues to eliminate the effect of temperature variation
4 RESULTS AND DISCUSSION
among FWD measurements at different times.
Moreover, even if the cyclic pattern is the same
4.1 Modulus indices
as seasonal temperature variation, the correlation
The evolution in modulus in terms of modu- observed for the section is inverse. For instance,
lus indices for the Oklahoma section is depicted the modulus value in July is higher as compared
in Figure 3(a). The modulus indices shown are to December which is opposite to what we would
determined based on FWD measurements taken expect with respect to typical temperatures in July
along the mid lane. The computation is done for and December, see Figure 3(b). The above two rea-
the time frame between May 1990 (designated as sons indicate the following hypothesis: Firstly, the
0 Month) to July 2003 (designated as 161 Month) temperature used for modulus correction in this
which is approximately 13 years. As discussed in analysis is at mid-depth for asphalt layer, the use
the previous section, temperature correction (in of this temperature appears to have too severe of
this case assuming pavement mid-depth tempera- a correction that leads to a lower modulus index
ture as an effective temperature) has been applied for winter months and a higher one for summer.
and the same frequency value is tied to all measure- Secondly, the deviation in pavement temperature
ments to allow a direct comparison among them. can cause the behavior of the pavement system to
In Figure 3(a) the points on the same vertical deviate from the assumptions of layered elasticity

739
 10.00 !.E+Ol
(a) €1
~
~ 1.00
,
:g 0.10 0

~ 1 ONonnartzed Back·calculated Modulus I

0
0.01
150 200

, !.E-02
:go. Io 0 20 40 0 60
~
0.01 40
L_------------------~:-----~~----~9~0
50 60 MJ~ths 80
~--~100 AC Mid Depth Temperature, C

Figure 4. Modulus indices and AC mid depth tempera-

ture relationship.
Figure 3. Modulus Indices for the time frame between
(a) 0 to 161 months (b) 40 to 100 months.

It has to be also noted that the effect of seasonal

moisture variation has not been considered in this
and can introduce in inherent error associated with process. While the properties of the asphalt layer
the linear elastic assumption in the back calcula- do not vary with moisture, a significant variation
tion process. The extent of deviation from linear in unbound material stiffness is anticipated due to
elasticity depends on the amount of deflection seasonal moisture variation. Oklahoma is catego-
encountered in the asphalt concrete layer during rized as a no-freeze site and exhibits a cyclic varia-
the FWD test which in turn depends on the test tion in stiffness due to variation in moisture. This
temperature. This can be greater during the hot cyclic variation can produce variation in deflection
seasons (higher temperature) where the higher measurements which could affect the linear elastic
deflection measurement leads to a non-linear rela- assumption in the back-calculation process as well
tionship and can introduce more error with every as the linear-viscoelastic assumption in time tem-
temperature increment resulting in the cyclic vari- perature superposition process for temperature cor-
ation. Furthermore, the time temperature super- rection. While the back calculation of AC surface
position principle which is applied for temperature modulus depends merely on the deflection of the
correction assuming a linear viscoelastic relation- same AC layer, the effect of variation in unbound
ship can become invalid and can also introduce to layer moisture could contribute to the cyclic varia-
some extent to the month to month cyclic variation tion that is observed within a year due to the vari-
observed. Again here each increment of tempera- ation in deflection. Thus, month to month changes
ture can cause more deviation from the assumption observed in the back-calculated modulus may not
and can lead to the cyclic variation. Figure 4 shows directly be the result of environmental changes
the modulus index versus asphalt layer mid-depth (temperature or moisture) but instead could be tied
temperature over the course of the pavement life. to month to month changes in the degree to which
A loosely linear trend observed between log of layered elasticity, the time-temperature superposi-
modulus indices and temperature supports the tion principle and use of mid-depth temperature as
above hypotheses indicating the use of incorrect effective layer temperature are valid to model pave-
location for reference temperature, the linear elas- ment system deflection.
tic assumption in the back-calculation process and
linear viscoelastic assumption during temperature
4.2 FWD test location
correction could impact back-calculated modulus
in a consistent manner over range of temperatures. Figure 5 compares modulus indices computed
While one might argue that part of the linear incre- from FWD tests conducted along the traffic mid
ment could also be attributed from an increase in lane (F1) versus tests conducted along the outer
modulus over time, for this specific study such a wheel path (F3) on the same transverse line. This
distinct increment has not been observed (see is done to investigate the difference in modulus
Figure 3(b) for modulus indices over time). This changes due to prevailing variation in load appli-
leads the authors to conclude that use of the time- cation, distress type and other differences between
temperature superposition principle for tempera- wheel path and mid-lane. In general, the modulus
ture correction of back-calculated modulus using indices calculated from the two sets of data show
a single mid-depth temperature of asphalt concrete similar cyclic trends indicating the same effect
layer can introduce a significant cyclic error in the of temperature variation (deflection variation)
proposed process. on the back-calculation and time temperature

740
 10.00
(a) 10.00 30
0
It>
j 1.00 B. ~1.00

"
0
0
::g 0.10
a 0 ..s
0
!.R "' .....-:-:. .
::;: • Modulus Indices along Fl .,
-30.10 ...:···---·· ... :.....··········
____ .,
0,01
o Modulus lndices alon' F3
~ 0.01 . -. ' i \ ./~ \

·.:/
40 50 60
M1~ths 80 90 100
0
100 0 50 100 150
(b) I • Modulus indices-l /3rd Depth Temp
o Modulus indices-Mid Depth Temo- Months
Modulus Index ········Rutting
•
•• ,; ~ ---- Fatigue ·- Transverse Cracking

8 .r:P OCT.
0 Figure 6. Modulus indices and distress relationship.
0.1
40 50 60 70 80 90 100
Months
different distresses and modulus indices. In the
Figure 5. Effect on modulus indices due to (a) FWD future, collection of distress data along with the
test location (b) Effective pavement temperature. FWD test is suggested to help with development
of mathematical models that help to describe any
relationship among them.
superposition process. The indices are also com-
parable in magnitude indicating a similar stiffness
deterioration along the outer wheel path and the 5 CONCLUSION AND
mid traffic lane. RECOMMENDATION

The LTPP program sought to collect data believed

4.3 Effective temperature
to be most important in characterizing factors that
The modulus indices computation in the previ- affect pavement performance. Intensive effort has
ous section was done assuming the pavement mid been made to establish data collection procedures
depth temperature as an effective temperature. and protocols so that the database is effectively
To see whether a different assumption in effective useable by researchers. With these intent, this
temperature affects the magnitude or the previ- study proposes a new approach for establishing a
ously observed trends with respect to modulus reference frequency and normalizing the tempera-
indices, the modulus indices are again computed ture difference among back-calculated modulus
assuming the temperature at 1/3rd depth of AC values available in LTPP to be able use them in
layer as an effective temperature. Since the 1/3rd mechanistic response and performance models.
depth temperature measurements for the first and Furthermore, it allows engineers to understand
second FWD measurements were missing from how the stiffness of material changes over the
LTPP, the third FWD measurement is taken as a course of the pavement life. To accomplish these
reference for this comparison in both scenarios. goals, first step of this approach tied the initial
Therefore, the modulus indices computed for mid back-calculated modulus to a frequency using the
depth temperature are with reference to a differ- as-built dynamic modulus master curve. Then,
ent time than considered in the Figure 2 and the temperature normalization is done using the time
trends appear slightly different with respect to the temperature superposition principle. For normal-
referenced time. Figure 5 shows both temperatures izing the temperature of reference modulus to the
exhibited a similar cyclic trend and resulted in a temperature at the time of FWD measurement, the
similar interpretation of the transition in the stiff- as-built time temperature shift factors are used.
ness material property, albeit with a shift in the Evolution in modulus at different stages of pave-
index values due to the use of a different reference ment life is studied by computing modulus indices
temperature. Thus, previous hypothesis on use of (ratio of normalized modulus at any given time to
inappropriate reference temperature as a reason referenced modulus value).
for cyclic variations in modulus index was not fully The proposed method was found promising to
supported in the results presented here. establish a reference frequency and normalize the
temperature difference among FWD measure-
ments and subsequently develop modulus indi-
4.4 Modulus index and distress relationship
ces. Modulus indices developed for an Oklahoma
Figure 6 presents the relationship between vari- section depicted a cyclic month to month varia-
ation in modulus and different distresses. No tion rather than a distinct incremental or decre-
meaningful correlation was found between the mental trend for the time frame considered in the

741
study. It is hypothesized that the apparent cyclic Elkins, G.E., Schmalzer, P., Thompson, T., & Simpson, A.
trend observed could be largely attributed due to 2003. Long-Term Pavement Performance Information
deviation from the considered linear viscoelastic Management System Pavement Performance Database
assumption for application of the time tempera- User Reference Guide.
Elshaer, M., M. Ghayoomi, & J.S. Daniel. 2017. Method-
ture superposition process. The results indicate ology to Evaluate Performance of Pavement Structure
that a non-linearity can be increasingly introduced using Soil Moisture Profile. Journal of Road Materials
with an increase in temperature (AC layer) or mois- and Pavement Design, 1–20.
ture (unbound layers) by increasing the deflection Gardner, L., & Skok, E. 1967. Use of viscoelastic concepts
encountered in the AC layer during FWD testing. to evaluate laboratoty test results and field perform-
To mitigate these limitations associated to validity ance of some Minnesota asphalt mixtures. Paper pre-
of the time temperature superposition principle, sented at the Intl Conf Struct Design Asphalt Pvmt.
the authors recommend establishing a test tem- Johnson, A.M., & Baus, R.L. 1992. Alternative method
perature range that results in a linear relationship. for temperature correction of backcalculated equiva-
lent pavement moduli. Transportation Research
Otherwise non-linear material behavior should be Record, (1355).
used in back-calculation process to discard the Kim, Y., Hibbs, B., & Lee, Y. 1994. New temperature
error associated with the non-linearity. correction procedure for FWD deflections of flexible
The effect of FWD test location (mid lane versus pavements. Paper presented at the 4th International
outer wheel path), and pavement effective tempera- Conference, Bearing Capacity of Roads and Airfields.
ture (1/3rd depth of AC layer versus AC layer mid Kim, Y.R., Hibbs, B.O., & Lee, Y. 1995. Temperature
depth) on the modulus indices computation was correction of deflections and backcalculated asphalt
found minimal. Moreover, no meaningful correla- concrete moduli. Transportation Research Record,
tion was found between the different distresses and (1473).
Lukanen, E.O., Stubstad, R., & Briggs, R. 2000. Temper-
modulus indices. In the future, collection of dis- ature Predictions and Adjustment Factors for Asphalt
tress data along with the FWD test is suggested. Pavement.
The authors recommend for future study an Mensching, D.J., Jacques, C.D., & Daniel, J.S. 2016.
evaluation of the applicability of the proposed Applying the glover-rowe parameter to evaluate low-
methods based on other test sections from LTPP. temperature performance of hot mix asphalt LTPP
Furthermore, actual quantification of the error sections. Journal of Materials in Civil Engineering,
introduced due to the non-linearity should be con- 28(10), 04016096.
ducted. A data collection routine that helps with Oshone, M., E. Dave, & J.S. Daniel, G. Rowe. 2017.
capturing FWD measurements that can be effec- Prediction of phase angle from dynamic modulus data
and implications on cracking performance evaluation.
tively implemented in the proposed method also Journal of Association of Asphalt Paving Technolo-
needs to be investigated. gists, In press.
Schmalzer, P. 2006. LTPP Manual for Falling Weight
Deflectometer Measurements.
REFERENCES Ullidtz, P. 1987. Pavement analysis. Developments in civil
engineering.
Ali, H., & Lopez, A. 1996. Statistical analyses of Van der Poel, C. 1955. Time and temperature effects on
temperature and moisture effects on pavement struc- the deformation of asphaltic bitumens and bitumen-
tural properties based on seasonal monitoring data. mineral mixtures. Society of Petroleum Engineers
Transportation Research Record: Journal of the Trans- Journal, 47–53.
portation Research Board, (1540), 48–55. Von Quintus, H.L., Rao, C., & Irwin, L. 2015. Long-
Chen, D., Bilyeu, J., Lin, H., & Murphy, M. 2000. Term Pavement Performance Program Determination
Temperature correction on falling weight deflectom- of in-Place Elastic Layer Modulus: Backcalculation
eter measurements. Transportation Research Record: Methodology and Procedures.
Journal of the Transportation Research Board, (1716),
30–39.
Christensen, D.A., Pellinen, T., & Bonaquist, R.F.
2003. “Hirsch Model for Estimating the Modulus of
Asphalt Concrete,” Journal of Asphalt Technologists,
Volume 72. 97–121.

742
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Application of falling weight deflectometer for the estimation of

in-situ shear strength parameters of subgrade layer

H. Nabizadeh, E.Y. Hajj, R.V. Siddharthan, S. Elfass & M. Nimeri

University of Nevada, Reno, Nevada, USA

ABSTRACT: Falling Weight Deflectometer (FWD) has been routinely used to determine the in-situ stiff-
ness of pavement layers. However, FWD is not viewed as a useful tool in pavement failure investigations
where shear strength parameters (c and φ) of unbound layers, particularly subgrade, are required for the
evaluation of shear failure in subgrade layer under moving superheavy load. In this study, a novel FWD-
based methodology for estimating the shear strength parameters of the subgrade layer was investigated. To
validate the proposed methodology for different types of pavement structures, large-scale experiments have
been carried out on a variety of unpaved and paved pavement structures at University of Nevada, Reno.
It was found that the FWD surface measurements at multiple load levels in conjunction with the fitting of
hyperbolic stress-strain relationship can be used to extend the data obtained at FWD stress levels to the
failure state, facilitating a reasonable estimate of the subgrade’s shear strength parameters.

1 INTRODUCTION Although the laboratory testing which is

destructive, provides the most reliable measure-
During the past two decades, the need for mov- ments for shear strength parameters, time con-
ing non-standard Super Heavy Loads (SHLs) on suming process of sample collection, testing, and
US highways, which is a vital economic necessity associated cost are objectionable limitations. On
for many important industries such as chemi- the other hand, use of engineering judgement can
cal, oil, electrical and mining etc., has drastically lead to an inaccurate determination of in-situ shear
increased. Many of such superheavy components strength parameters. Hence, application of differ-
are much larger in size and weight and may some- ent Non-Destructive Testing (NDT) techniques
times approach 9000 kN. Consequently, the risk for reasonable estimation of in-situ shear strength
of instantaneous shear failure in pavement layers parameters of unbound pavement layers is a desir-
due to the application of such a heavy load needs able undertaking.
a detailed investigation. It means that the common Falling Weight Deflectometer (FWD) as a non-
pavement analysis methods which only relies on the destructive device has been routinely used in the
elastic stiffness properties of pavement layers are pavement industry for the purpose of in-situ evalu-
not able to consider all the required factors when ation of structural capacity of in-service pavements.
the risk of instantaneous shear failure in pavement The FWD simulates pavement responses under a
layers under a SHL move is of concern (Dong et al. moving truck wheel though it applies a stationary
2014, Chen et al. 2013, Jooste & Fernando 1994). impact load to the pavement surface. The measured
The 3D Mohr-Coulomb yield criterion has been applied load in conjunction with the vertical surface
implemented in previous researches to investigate displacements at different radial distances from the
the possibility of shear failure in the pavement load application (called deflection basin) used as
layers subjected to the SHL move (Chen et al. the input for backcalculation algorithms to estimate
2013, Jooste & Fernando 1994, Fernando 1997). the in-situ resilient stiffness of the various pavement
Although the linear elastic theory in conjunction layers. The backcalculated layer stiffnesses have been
with the backcalculated stiffness parameters can be used as an indicator of the pavement structure’s ade-
employed to reasonably estimate the load induced quacy subjected to the standard traffic loading.
state of stresses within the pavement layers, shear Despite the fact that the FWD device has been suc-
strength parameters (cohesion, c, and friction cessfully employed to estimate stiffness of pavement
angle, φ) at the point of interest are also required layers and subsequent load carrying capacity of the
for the yield strength calculation. In the aforemen- pavement structure, the applicability of this NDT
tioned studies, these parameters were determined device to estimate the in-situ shear strength param-
by means of laboratory tests or specified based on eters of pavement layers, especially for unbound
the engineering experiences. materials, has not been explored. In this paper an

743
innovative FWD-based methodology to estimate the 2002). Accordingly, if the stress-dependency of
in-situ shear strength parameters of subgrade layer subgrade material behavior under multiple FWD
is proposed. This is to be accomplished by focusing load levels is represented by a hyperbolic stress-
on the stress dependent behavior of subgrade mate- strain relationship, the data obtained at lower
rials under multiple FWD load levels. stress levels can be extended then to the failure
The validity of proposed approach has been state which enables the estimation of maximum
explored by means of numerical simulation of deviatoric stress (asymptotic value). The following
FWD data recently (Nabizadeh et al. 2016). In steps outline the proposed methodology.
order to further validate the proposed methodol- Step 1: The backcalculated layer moduli values
ogy, FWD data collected under multiple load lev- at each load level are used with a layered Linear
els from large-scale experiments at University of Elastic Program (LEP) to compute the stress ten-
Nevada, Reno large-scale facility have been utilized. sor (σij) at a representative element in the subgrade
layer. An element located at the depth of B/2 (B
is the diameter of the FWD plate) from the top
2 PROPOSED METHODOLOGY of subgrade can be treated as the representative
element to determine the load-induced stresses.
The triaxial compression tests have been tradi- The representative element (at B/2 from top of
tionally used as a laboratory method to determine subgrade surface) is bounded by shearing zones in
the shear strength of soils. A typical stress-strain the subgrade and experiences the largest vertical
relationship in a triaxial compression test is repre- strain under circular loaded area (Schmertmann
sented by a hyperbola as presented in Equation 1. et al. 1978). It should be mentioned that the
assumptions used in the backcalculation proce-
ε1 (1) dures and the LEPs are consistent (e.g., static and
σd = axisymmetric loading condition).
1 ε
+ 1 Step 2: The calculated induced stress tensor (σij)
Ei σ df at the representative element are “transformed”
into an equivalent laboratory triaxial stress test-
where σd = triaxial deviator stress; ε1 = axial strain ing conditions by the use of stress invariants which
corresponding to deviator stress; Ei = initial tan- have been used to compare the stress states in pre-
gent modulus; and σdf = deviator stress at failure. vious studies (Hajj et al. 2010, Nabizadeh et al.
A widely accepted procedure is used to deter- 2016). Stress invariant values are the same regard-
mine the deviator stress at failure (σdf) by rewriting less of the orientation of the coordinate system
hyperbolic equation in a linear form, see Equation chosen. The octahedral normal (σoct) and shear
2. Hence, the inverse of the slope of ε1/σd versus ε1 is (τoct) stresses, which are invariants, are used to con-
equal to σdf (Kondner 1963, Duncan & Chang 1970). vert the stress tensor computed in the representa-
tive subgrade element under the FWD loads to
ε1 1 1 deviator (σd) and confining (σc) stresses in a triaxial
= ε1 + (2) testing set-up using Equations 3 through 6. In
σd σ df Ei
these equations, σ1, σ2, and σ3 are the major, inter-
mediate, and minor principal stresses, respectively.
In the triaxial compression test, the desired
stress states and corresponding axial strains are
measured during the test and fitting a line through σ oct =
1
3
(σ 1 + σ 2 + σ 3 ) (3)
available measured data enables the extrapolation
of the stress-strain relationship to the near failure
− σ 2 ) + (σ 1 − σ 3 ) + (σ 2 − σ 3 )
1
(σ
2 2 2
level. A major step in this study is to determine τ oct 1 (4)
3
multiple datasets of σd and ε1 as a function of
confining pressure for a representative element in 3
σd τ oct (5)
the subgrade layer using the FWD measurements, 2
even though those measurements are not expected
σd
to induce failure or permanent deformation. σc σ oct − (6)
It is commonly accepted that the resilient modu- 3
lus of an unbound materials such as crushed aggre-
gate base and subgrade soils is a function of the Step 3: Calculated σd values at each load level
stress condition. Such a stress-dependent behavior can be used to compute the corresponding axial
of the unbound material is usually reflected in the strain (ε1) as expressed by Equation 7 where (Mr)
FWD backcalculated moduli values when multiple SG is the backcalculated subgrade resilient modulus
load levels are applied (Von Quintus & Simpson at the corresponding load level.

744
 σd structure in the first experiment, designed as Pave-
ε1 = (7)
( ) SG
Box-CAB, consisted of 167.6 cm of SubGrade (SG)
and 15.2 cm of Crushed Aggregate Base (CAB). For
Step 4: The inverse slope of ε1/σd versus ε1 is equal the experiment 2 (PaveBox-AC), 12.7 cm of Asphalt
to deviator stress at failure (σdf), according to the Concrete Layer (AC) was supported by the similar
hyperbolic stress-strain relationship (see Equation recompacted CAB and SG layer (see Figure 2).
2). It should be mentioned that in the case when The subgrade materials used in the PaveBox
the backcalculated subgrade modulus increases experiments have been classified as Clayey Sand
with increasing the load level (i.e., dominant hard- with gravel (SC). The result of unsaturated triax-
ening behavior), the inverse slope of ε1/σd versus ε1 ial tests indicated φSG = 38.2° and cSG = 11.7 kPa.
is negative which can be seen as a limitation for the Since estimation of shear strength parameter of
estimation of shear strength parameters. subgrade is the main objective of this paper, the
Step 5: Based on the determined σdf and com- detailed information regarding the AC and CAB
puted σc-avg which is the average of σc at each load material have not been presented here. Note that
level (Equation 6), it is possible to establish one the selected materials for the base and asphalt layer
single Mohr circle of failure. Accordingly, the satisfied the Nevada Department of Transporta-
shear strength parameter, cohesion values can be tion (NDOT) specifications.
estimated using Equation 8 by assuming an accept- In these experiments, cycles of dynamic pulse
able range for the friction angle, based on the soil loading using a hydraulic ram were applied on the
classification of the subgrade. It may be noted that 300 mm Dynatest FWD plate. FWD load level up
reliable data exists in the literature for selecting an to 31 kN and 120 kN were applied on the pavement
appropriate range of friction angle based on the surface constructed in the PaveBox-CAB and Pave-
soil classification (Das & Sobhan 2014). Box-AC, respectively. As illustrated in Figure 1 and
Figure 2, the corresponding pavement surface dis-
σc avg
avg+ σ df = σ c avg 2
( 45 φ / ) placements were measured by surface Linear Variable
+ 2c ⋅ tan( 45 + φ / ) (8) Differential Transformers LVDTs at 0.0 (LVDT1),
203 (LVDT2), 305 (LVDT3), 610 (LVDT4), 914
It may be noted that using linear elastic the- (LVDT5), 1219 (LVDT6), and 1524 mm (LVDT7)
ory in conjunction with FWD measurements is away from the center of the FWD plate. In addi-
the common practice to evaluate in-situ resilient tion to the surface deflection measurements, Earth
moduli values of the pavement layers. Subsequent Pressure Cells located in the base and subgrade layer
determination of state of stress also used the same captured the load induced vertical stresses during the
set of assumptions (i.e., stationary static axi- experiments. A set of routinely-used initial process-
symmetric loading). Since FWD backcalculations ing steps were utilized for all recordings to identify
used displacements from a number of FWD loads, and separate the proper load-induced response sig-
the proposed approach inherently incorporates the nals from the measured data.
stress-dependency of subgrade material behavior.

3.1 PaveBox-CAB experiment

3 VALIDATION OF PROPOSED Table 1 presents maximum surface vertical dis-
METHODOLOGY placements from LVDT measurements at different
load levels in PaveBox-CAB experiment. Since high
In order to experimentally assess the validity of
noise to signal ratio have been recorded by LVDT4,
the proposed methodology, two large-scale experi-
LVDT5, LVDT6, and LVDT7, only three surface
ments that included FWD testing were conducted
measurements were available. The surface displace-
on the unpaved and paved pavement structures
ment at the locations away from the center are influ-
at University of Nevada, Reno large-scale facil-
enced by interior layers due to the development of
ity. As shown in Figure 1, the unpaved pavement

PC1
Note: Dimensions Not to Scale

Figure 1. Pavement structure in PaveBox-CAB Figure 2. Pavement structure in PaveBox-AC

experiment. experiment.

745
Table 1. Measured surface displacement in PaveBox-
CAB experiment.

Maximum surface displacement

(mm)
Average applied
load level (kN) LVDT1 LVDT2 LVDT3

13 0.18 0.10 0.06

18 0.29 0.13 0.09
22 0.40 0.17 0.11
27 0.50 0.21 0.14
31 0.58 0.27 0.16

0
30 35

Figure 4. Backcalculated subgrade modulus in the

'E PaveBox-CAB experiment.
E
~0.4
Q)
E adopted, in which the subgrade layer was subdi-
Q)
~ 0.3 vided into two sublayers with the thickness of each
c..
(/)
sublayer determined using “Depth to an Apparent
i:S 0.2 Rigid Layer Method” (Rohde & Scullion 1990). It
is customary to expect a stress-softening behavior
0.1 in the subgrade (indicated by a decrease in modu-
lus with an increase in FWD load level). On the
other hand, the CAB is expected to show stress-
30 35 hardening as FWD load level increased.
Figure 4 shows the backcalculated modulus of
subgrade layer 1 within the influence zone. This
Figure 3. Evaluation of nonlinearity in the PaveBox- figure reveals that an increase in FWD load level
CAB experiment using Load-Response Characteristic. results in a reduction in backcalculated subgrade
modulus, indicating the expected softening behav-
ior. To corroborate the correctness of the backcal-
stress bulb under impact loading. The influence culation exercises, the backcalculated layer moduli
zone from the surface load is confined to a region values that correspond to the load levels under
directly under the load and therefore the displace- consideration were used in 3D-Move analysis
ments at further away locations were smaller. software (Siddharthan et al. 1998) to compute the
Since the proposed FWD-based methodology surface displacements at the location of LVDTs
relies on the stress dependency of the subgrade as well as load induced vertical stresses measured
material, the existence of nonlinearity in the by Earth Pressure Cells under different load lev-
pavement structure should be investigated first. els by assuming static loading conditions. Figure 5
According to the study conducted by Von Quin- and Figure 6 show a good agreement between the
tus & Simpson (2002) on the FWD measurements predicted versus measured surface deflections and
at the LTTP test sections, the load-response char- vertical stresses, respectively. These observations
acteristics can be divided into three categories; reveal the appropriateness of the steps that have
Linear elastic, Deflection-hardening, and Deflec- been undertaken in the backcalculation procedure.
tion Softening. As shown in Figure 3, the nega- Subsequently, based on the backcalculated mod-
tive intercept of the “Center Displacement-FWD uli values for the base and subgrade layers at each
Contact Stress” relationship implies the evidence of the load levels, the stress tensor (σij) at 150 mm
of deflection softening in the pavement structure. (i.e., B/2 where B is the plate diameter) below the
As mentioned earlier, to calculate the state of subgrade surface was computed by assuming static
stress in the pavement layers, layers’ moduli val- loading conditions using 3D-Move analysis soft-
ues are required. For handling the backcalculation ware. The stress conditions was converted to stress
process, BAKFAA open source software was uti- invariants and corresponding triaxial testing con-
lized. In addition, a trial and error procedure was ditions as outlined before.

746
 0.7 Radial Distance (mm)
~ LVDT 1 0 250 500 750 1000 1250 1500 1750
I 0.6 - . . - LVDT2 0.00
E -+- LVDT3
~ 0.5 • •• •••• Line of Equality
Q)
~
E'o.5o
ro 5
g. 0.4 c Q)
i5 E
-o
Q) ~ 1.00
:2
-o
0.3
g.
ro
~ i5
a.. ~ 40 kN
~ 0.2 ~ 1.50
0
:2 ~ ···•·· 53 kN
::I
~ 0.1
C/)
• - + - 71 kN
2.00
0.0
..··*· - . - 95 kN
- :::« - 121 kN
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Measured Displacement (mm) 2.50

Figure 5. Comparison between measured and 3D-Move Figure 7. Measured surface displacement in PaveBox-
predicted surface displacements in the PaveBox-CAB AC experiment.
experiment.

2.50
120 I I I
-ro )( ..··
./..:.··· ,•
y =0.0194x- 0.3305

,,
~ 1 00 2.00 1----
R2 =0.999
en
en ..;:.···
~
Ci5 80 ;;..:.·· E'
~ 1.50
I l ,~
TI /_.'···· I,
,.,
r:::
:e x.··
Q)

Q)
>
-o 60 /.... !
E
Q)

~
'5
.. :f-:.··
i5
1.00
,
Q)

a: 40 ~ PC 1
, rl
,,• I
Q)
> ~ • PC2 0.50
0
:2 ~ - PC3
20
0M -• - PC4
• •• •• •• Line of Equality 0.00
0 0 25 50 75 100 125 150
0 20 40 60 80 100 120 Applied Load (kN)
Measured Vertical Stress (kPa)
Figure 8. Evaluation of nonlinearity in the PaveBox-AC
Figure 6. Comparison between measured and experiment using Load-Response Characteristic.
3D-Move predicted vertical stresses in the PaveBox-CAB
experiment.

The inverse slope of ε1/σd versus ε1 is equal as unavoidable, even though special attention was
to deviator stress at failure (σdf) which is to be given during the compaction and construction of
122.2 kPa in this experiment. By assuming an the pavement layers to achieve similar densities.
acceptable range for the friction angle based on the
identified soil classification corresponding cohe-
3.2 PaveBox-AC experiment
sion can be calculated (Equation 8). In the cur-
rent analysis, the cohesion value, cSG, is estimated As previously stated, in the PaveBox-AC experi-
as 20.0 kPa, if φSG was assumed to be 38.2°. This ment, multiple FWD load levels were applied
set of shear strength parameters can be compared on top of the full pavement structure. Figure 7
with the measured values for the subgrade mate- presents the maximum pavement surface vertical
rial based on the triaxial tests (cSG = 11.7 kPa and deflections at different load levels.
φSG = 38.2°). The slight difference in shear strength The stress dependency in the large-scale pave-
parameters can be attributed to the difference in ment structure was examined first using the meas-
subgrade layer densities. Such a discrepancy is seen ured deflection basins. As displayed in Figure 8,

747
 140,000 350
ro-
~ 120,000 :;3oo
"'·~···
, _.·, ..
Vl Vl

~ 250
::::J
"'5
"'C
0
100,000 C/) ,..,....
:2 ~ "' .··
"' "'..·.·
(1)
"'C
ro 80,000 ~200
~"' .. ··
t:n
..0
::::J
C/)
"'C
(1)
60,000
>
"'C
Q)

ti 150
'5
Q)
,"' ..
, • •• ··_/ ·
· ·, -
~ PC 1
-ro
"3
a: 41 .:;;,.x -~ ·- PC2
Q 40,000 ~ 100 ~.·. PC3
~
..::.:: :2
0
- • - PC4
Q
ro 20,000 ~ 50 -~ PC5
CD -~- PC6
• • • • • • • Line of Equality
0 0
0 25 50 75 100 125 150 0 ~ 100 1~ 200 2~ 300 3~
Applied Load (kN) Measured Vertical Stress (kPa)

Figure 9. Backcalculated subgrade modulus in the Figure 11. Comparison between measured and
PaveBox-AC experiment. 3D-Move predicted vertical stresses in the PaveBox-AC
experiment.
2.5
~ LVD T 1
-~· LVD T2 Table 2. Estimated subgrade shear strength parameters
E'
-~ LVDT3 in PaveBox-AC experiment.
:22.0 - .& - LVDT4
c:::
~ - LVDT5
Q)
E Typical Measured Estimated
~ - • LVDT6
~ • values of shear shear
~ 1 .5 LVDT7
Soil friction strength strength
•• • •• •• Line of Equality ,
i:S ,, "
"'C
Q)
classification angle (°) parameters parameters
~Q)
1.0 Clayey sand 35 to 40 φSG = 38.2° φSG = 37.5°
a:
Q)
with gravel cSG = 11.7 cSG = 17.9
> kPa kPa
~ 0.5
0C"?
0.0 ';t[l.,;:,..__~_ _ _ _ . . _ _ _, _ _ _ __ _ _,___ _ _

controls on the variability of the elastic modulus

0.0 0.5 1.0 1.5 2.0 2.5 revealed that the subgrade layer should be subdi-
Measured Displacement (mm) vided to two sublayers using the aforementioned
“Depth to an Apparent Rigid Layer Method”
Figure 10. Comparison between measured and (Rohde and Scullion, 1990). Figure 9 shows the
3D-Move predicted surface displacements in the PaveBox- reduction in the backcalculated subgrade modulus
AC experiment.
with increasing the applied load levels.
To investigate the correctness of backcalculation
deflection softening behavior was identified in process in PaveBox-AC, surface displacements and
PaveBox-AC structure according to the inspection vertical stresses under different load levels were cal-
steps outlined above. Since asphalt concrete materi- culated. As presented in Figure 10 and Figure 11,
als does not generally behave as a stress-dependent the satisfactory difference between the measured
material and the crushed aggregate base materials and predicted responses implies the validity of
exhibits either elastic or hardening characteristic, assumptions made in the backcalculation process.
the observed deflection softening can therefore Subsequently, results of stress tensor at 5.91
be directly attributed to the softening behavior of inches below the subgrade surface was computed
subgrade material. Such a behavior of subgrade and converted to the corresponding triaxial testing
materials will be reflected in the backcalculated sub- conditions as outlined above. Table 2 presents the
grade moduli values obtained at different load levels. result for the shear strength parameters estima-
The surface deflection basins at different load tion using the calculated deviator stress at failure
levels were employed in the backcalculation (σdf). As before, by assuming an acceptable range
analysis using the program BAKFAA. Repeated for the friction angle based on the identified
attempts at the backcalculation process with many soil classification (clayey sand with gravel),

748
corresponding cohesion values were estimated. the subgrade layer when the material exhibited a
The results of the cSG and φSG can now be com- dominant softening behavior under the applied
pared against the measured values for the subgrade FWD stress levels. It is concluded that the linear
material. A close match between measured and elastic theory in combination with a backcalcula-
predicted soil strength parameters seen in Table 2 tion exercise that appropriately accounts for the
indicates that the proposed FWD-based methodol- stress dependency of unbound materials along
ogy for estimating strength parameters of subgrade with the hyperbolic stress-strain relationship can be
is applicable. used successfully to estimate the subgrade’s shear
strength parameters.

4 CONCLUSION
REFERENCES
Substantial increase to move Super Heavy Loads
(SHLs) on US highways have led to the need for Chen, X. & Lambert, J.R. & Tsai, C. & Zhang, Z. 2013.
investigating the risk of instantaneous shear failure Evaluation of Superheavy Load Movement on Flex-
in pavement layers. Such an analysis requires the ible Pavements. International Journal of Pavement
shear strength parameters (c and φ) of pavement Engineering 14(5): 440–448.
Das, B.M., & Sobhan, K. 2014. Principles of Geotechni-
layers, particularly the weakest one, i.e. subgrade cal Engineering. Connecticut: Cengage Learning.
layer. Although laboratory triaxial test can provide Dong, Q. & Huang, B. & Shu, X. & Zhou, C. & Max-
the most precise measurement for these param- well, J. 2014. Use of Finite Element Analysis and
eters, destructive and time consuming nature of Fatigue Failure Model to Estimate Costs of Pave-
this test may sometimes be not economical and ment Damage Caused by Heavy Vehicles. Transpor-
practical. In this study, a novel methodology was tation Research Record: Journal of the Transportation
investigated to estimate these parameters for the Research Board 2455: 54–62.
subgrade layer using non-destructive FWD test- Duncan, J.M. & Chang, C.Y. 1970. Nonlinear Analy-
ing undertaken at multiple load levels. The widely sis of Stress and Strain in Soils. Journal of the Soil
Mechanics and Foundations Division 96(5): 1629–1653.
accepted hyperbolic stress-strain relationship was Fernando, E.G. 1997. Guidelines for Evaluating Super-
used to extend the data obtained at the FWD- heavy Load Routes. TX -98/3923-S.
induced stress levels to the failure state for the Hajj, E.Y. & Ulloa, A. & Siddharthan, R.V. & Sebaaly,
estimation of the deviatoric stress at failure (i.e., P.E. 2010. Estimation of Stress Conditions for the
asymptotic σdf value) for the subgrade material. Flow Number Simple Performance Test. Transporta-
This methodology consisted of the following tion Research Record: Journal of the Transportation
steps: (1) Obtain the stress tensor (σij) at a repre- Research Board 2181: 67–78.
sentative location in the subgrade using a layered Jooste, F.J. & Fernando, E.G. 1994. Victoria Superheavy
linear elastic program, e.g., 3D-Move static analy- Load Move: Report on Route Assessment and Pave-
ment Modeling. FHWA/TX-94/1335-1.
sis, for each of the FWD load levels; (2) Use the Kondner, R.L. 1963. Hyperbolic Stress-Strain Response:
FWD load-induced stress tensors to compute Cohesive Soils. Journal of the Soil Mechanics and
the stress conditions associated with triaxial tests Foundations Division 89(1): 115–144.
(σd and σc) for each of the FWD load levels; (3) Nabizadeh, H. & Hajj, E.Y. & Siddharthan, R.V. &
Compute the triaxial test datasets of deviatoric Elfass, S. & Sebaaly, P.E. 2016. Estimation of In-Situ
stress (σd) and axial strain (ε1) in the representative Shear Strength Parameters for Subgrade Layer Using
subgrade element; (4) Fit a hyperbolic relation- Non-Destructive Testing. The Roles of Accelerated
ship to the triaxial responses to obtain the ultimate Pavement Testing in Pavement Sustainability: 525–538.
deviator stress (σdf); and (5) Use Mohr-Coulomb Rohde, G.T. & Scullion T. 1990. MODULUS 4.0: Expan-
sion and Validation of the MODULUS Backcalcula-
failure envelope equation to estimate the shear tion System. FHWA/TX-91/1123-3.
strength parameters of the subgrade by assuming Schmertmann, J.H. & Brown, P.R. & Hartman, J.P. 1978.
an acceptable range for the friction angle. Improved Strain Influence Factor Diagrams. Journal
In order to assess the applicability of the pro- of Geotechnical Engineering 104(8): 1131–1135.
posed approach, two large-scale experiments Siddharthan, R.V. & Yao, J. & Sebaaly, P.E. 1998. Pave-
that included FWD testing were conducted on ment Strain from Moving Dynamic 3D Load Distri-
the unpaved and paved pavement structures at bution. Journal of Transportation Engineering 124(6):
University of Nevada, Reno large-scale facility. 557–566.
Analysis of FWD measurements from these two Von Quintus, H.L. & Simpson, A.L. 2002. Back-
Calculation of Layer Parameters for LTPP Test Sec-
experiments revealed that the proposed FWD- tions, Volume II: Layered Elastic Analysis for Flexible
based methodology was able to reasonably esti- and Rigid Pavements. FHWA-RD-01-113.
mate the shear strength properties (c and φ) of

749
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

A model to adjust the falling weight deflections due to temperature

variations

J.C. Pais & P. Pereira

University of Minho, Guimarães, Portugal

ABSTRACT: The evaluation of the bearing capacity of the pavement that is used to assess the stiffness
modulus of its layers and is usually carried out by the measurement of the pavement deflections due to the
application of a known load through a falling weight deflectometer. These measurements are influenced
by the pavement temperature and the obtained stiffness moduli can only be compared if they correspond
to the same temperature. Thus, this work develops a model to adjust the pavement deflections to a refer-
ence temperature based on the characteristics of the pavement. The development of the work is divided
into four parts: i) a laboratory evaluation of the stiffness modulus of asphalt mixes at different tempera-
tures and the proposal of a model for the stiffness modulus as function of the temperature; ii) assessment
of the pavement deflections for different testing conditions; iii) development of a deterministic model for
the adjustment of the pavement deflection; and iv) validation of the model.

1 INTRODUCTION perature to a reference temperature. There are var-

ious methods for this adjustment such as graphics,
Pavement deflection is affected by factors that models with logarithmic relations and linear rela-
include loading factors (applied load, radius of cir- tions, among others. The model developed by
cular load plate and plate contact pressure), struc- the American Association of State Highway and
tural factors (number of layers, layer thickness, Transportation Officials (AASHTO, 1993) is now
stiffness modulus and Poisson’s ratio), temperature presented because of its simplicity, where it is only
of the asphalt layers, discontinuities and variation needed to intercept the average of pavement tem-
of the pavement structure (SHRP, 1993). perature with the type of pavement (from A to G)
As indicated by FHWA (2000), the stiffness of to obtain the adjustment factor (Figure 1).
asphalt mixtures is highly sensitive to temperature Other models can be enumerated such as the
changes that may occur both in long periods of time model developed by the Asphalt Institute (Asphalt
(seasonal alterations) and in short periods (time Institute, 1983) based on graphics, the model devel-
changes). With the increase of pavement tempera- oped by Kim & Park (2002) in which the adjust-
ture, deflection increases also, as long as other fac- ment is made through a logarithmic model and the
tors remain the same (Minhoto et al., 2008). Thus, model developed by Lukanen et al. (2000) which
deflections measured in the summer are greater introduces parameters such as deflection basin
than deflections measured in colder periods. With shape factors, pavement temperature, latitude of the
regard to changes in temperature in depth (vertical pavement section and thickness of the asphalt layer.
gradients), its influence in the state of stress of the
layers is more significant at higher thickness.
Given the influence of temperature on asphalt
mixtures behavior, it is important to know the pave-
ment temperature when measuring the deflection,
either by using thermocouples or through the use
of models to predict temperature at various depths
(Minhoto et al., 2005). As indicated by Fernando
et al. (2001), once the characteristics of the pave-
ment are known, the results of deflection obtained
with a falling weight deflectometer to project appli- o- I

cations, pavement evaluation and other applica- T EMPERATU RE ADJU STMENT FACTOR
tions should be adjusted to a reference temperature.
Methods to adjust pavement deflection aim the Figure 1. Deflection-temperature adjustment factor
adjustment of deflections measured at test tem- (AASHTO, 1993).

751
 Thus, the goal of this work is the development I I Ff'.V
of a model to adjust deflections obtained with fall-
ing weight deflectometer at a test temperature to H
- o- AC14

- x - AC16

_ ..,. _ AC20
t-
+ 'I,
-r- f-+ ;:: r-
a reference temperature, on a pavement with one
asphalt layer and one granular layer. The model is ·-
developed from three distinct phases:
/ ~r
·, 'II
• Laboratory evaluation: In this phase stiffness -, ~
modulus tests are performed with AC14, AC16 0 I I 1 1

0.01

and AC20 mixes at temperatures of −10, 0, 10, Sieveopening (mm)

20 and 30ºC, and a model to define the stiffness Figure 2. Aggregate gradation curves of the asphalt
modulus is presented. mixes.
• Numerical modeling: Using the software JPav
(linear elastic analysis of pavements), developed
by the first author, expected deflections for pave- 20000

ments with arbitrated characteristics at different

temperatures are obtained. These results allowed
~a16000
18000
I ·-. . I V ~ -388.49H 13656
! 14000 R~ " 0.9884

the development of the adjust model. 3

~ 12000

• In situ evaluation: In this phase in situ tests are ]

E
1())(()
..
carried out with falling weight deflectometer in ~ 8000 ··.
~ 6000
I ···. . . •
a pavement with known constitution to validate .;;; 4000

the developed model. 2000

o+-----~----~----~----~----~----~
- 2<} -10 10 20 30 40
Temperature (!!C)
2 STIFFNESS MODULUS OF ASPHALT
MIXES Figure 3. Stiffness modulus @ 10 Hz for all asphalt
mixes.
The adjustment of the pavement deflections due to
the temperature requires the knowledge of a model
to define the stiffness of the asphalt mixes with the Each gradation was used with three different
temperature. This type of model needs two differ- air-void contents, representing a low, an average
ent values: the stiffness at a reference temperature and a high value, typical for each aggregate gra-
and a law that defines the variation of the stiffness dation, represented respectively by F1, F2 and F3,
with the temperature. ranging from 4% up to 8%. The produced asphalt
This variation of the stiffness is relatively well mixes had the designation of AC14-F1, AC14-F2,
known through some models developed in the past AC14-F3, AC16-F1, AC16-F2, AC16-F3, AC20-
all over the world while the stiffness modulus for F1, AC20-F2 and AC20-F3. All asphalt mixes
a reference temperature depends of the volumetric were produced with a 35/50 pen asphalt.
composition of the asphalt mix and the stiffness of The asphalt mixes used in this work were pro-
the asphalt at that given temperature and frequency duced and compacted in laboratory and from the
(Sousa et al., 1998). Also for the stiffness of the slabs compacted with a steel roller, specimens with
asphalt mixes at a given temperature there are some 38.1 cm long × 5.1 cm thick × 6.3 cm wide were cut
models that can be applied as a first approach. for laboratory tests to obtain stiffness modulus in
To take into account the local conditions for four point bending tests at temperatures of −10, 0,
the adjustment of the pavement deflections due to 10, 20 and 30ºC and at frequencies of 0.1, 0.2, 0.5,
the temperature variation, it is important to know the 1, 2, 5, 8 and 10 Hz.
typical values of the stiffness modulus as well as its Because in the pavement design it is usual to
variation with the temperature and frequency. Thus, a consider the stiffness modulus around at 10 Hz, the
laboratory testing program was carried out to evalu- development of a model to adjust the deflection
ate the stiffness modulus of typical asphalt mixes, with was carried out by considering the stiffness modu-
different aggregate gradation curves, and subjected to lus of the asphalt mixes given by the mean values
different testing temperature and frequency. To com- obtained in the laboratory tests for a frequency
prise the typical asphalt mixes used in pavements, of 10 Hz that are presented in Figure 3 with the
the stiffness evaluation included three asphalt mixes linear relationship that best represent these val-
used in wearing coarse, binder and base layer, with ues, indicated in Equation 1, where E is the stiff-
nominal aggregate gradation of 14, 16 and 20 mm, ness modulus of the asphalt mix (MPa) and T is
designated as AC14, AC16 and AC20. The aggregate the temperature (ºC). This equation is only valid
gradation curves are represented in Figure 2 where it for this study and the application for other studies
can observed the differences among them. must be should be carried out very carefully.

752
E 390 × T + 13600 (1) 4 DEFLECTION RATIO

For the development of the model the pavement

deflections obtained for with the simulations
3 PAVEMENT MODELING defined in the previous chapter were used to estab-
lish a Deflection Ratio (DR) according Equation 2
For the development of the model to adjust the where defT20 is the pavement deflection at 20°C and
pavement deflection due to the temperature it is defT is the deflection at a temperature T.
necessary: i) to obtain the pavement deflections
due to the application of a load for several pave- def
efT 20
ment conditions and different temperatures; ii) the DR = (2)
def
efT
calculation of the deflection ratio that transforms
the pavement deflections at a given temperature to This DR value allows to adjust the measured
another temperature. deflection ( fT ) at a temperature T for a tem-
The pavement deflections were obtained for a dj
perature of 20°C ( fT 20 ) according to Equation 3.
pavement with 4 layers as follows: i) Asphalt layer;
ii) Granular layer; iii) Subgrade; and iv) Rigid def dj
efT 20 effTmeas × DR
ddef (3)
layer.
The stiffness modulus of the pavement layers Having obtained the DR for all the data set, the
were chosen as follows: study of the influence of each dependent variable
on DR is needed. The first analysis is the evolution
• Thickness of the asphalt layer (H) ranging from
of DR with changes in the distance (d) between
0.10 m to 0.40 m, with variation of 0.025 m;
the sensor and the load. This can be observed in
• Thickness of the granular layer set as 0.20 m;
Figure 4 where the deflection ratio is plotted as
• Thickness of the subgrade set as 1 m.
function of the distance for different temperatures
In term of the stiffness of the pavement layers, (−10°C up to 30°C) for a pavement with 0.15 m
they were chosen as follows: of asphalt layer thickness and a subgrade stiff-
ness of 60 MPa. The evolution of the deflection
• Stiffness of the asphalt layer was defined func-
ratio follows a linear relationship with the distance,
tion of the selected Temperatures (T) equals to
although some deviation from the straight line is
−10, 0, 10, 20, 25 and 30ºC, according the values
observed depending on the asphalt layer thickness
indicated in Table 1;
and subgrade stiffness combination. However, a
• Stiffness of the granular layer set as twice the
straight line can be used to define the relationship
stiffness of the subgrade;
between the deflection ration and the distance.
• Stiffness of the subgrade (E) ranging from
The influence of the asphalt layer thickness on the
20 MPa to 120 MPa with a variation of 20 MPa;
deflection ratio is expressed in Figure 5 where it can
• Stiffness of the rigid layer set as 1000 MPa.
be observed that a logarithmic law can be used to
The magnitude of the load was considered define this relationship as observed for a pavement
equal to 40 kN and the radius of circular load was with a 100 MPa subgrade stiffness and an asphalt
0.15 m. The calculation of the pavement deflection layer at 10°C. This deflection ratio corresponds to
was carried out at 10 distances (d), namely at 0, a distance of zero, meaning that the measurement
0.20, 0.30, 0.45, 0.60, 0.90, 1.20, 1.50, 1.80 and 2.10 is just in the center of the load. For the other pave-
meters from the center of the load. The combina- ment combinations and distances, the deflection
tion of all values taken by these variables yielded ratio can be defined by a logarithmic law although
468 data sets to perform the numerical modeling.
1.8
- - -10
1.6
1.4 -- o
0
1.2
Table 1. Stiffness modulus for the asphalt layer. ~ - - 10
c
0
........ 20
Temperature (ºC) Stiffness modulus (MPa) ~ 0.8
r; 0.6 - - 25
0

−10 17500 0.4

- - 30
0.2
0 13600 0
10 9700 0 0.2 0.4 0.6 0.8
Distance (m)
20 5800
25 3850
30 1900 Figure 4. Typical influence of the distance on the
deflection ratio.

753
 1.34 1.80
1.32 1.60
1.30 1.40
0
·;
Ill:
1.28
1.26
!
0
1.20
y = 0.11011n(x)+ 1.428 6
s 1.24 R2 = 0.9984 "i
1.00
0.80
~ 1.22
'tQ 0.60 y = -0 .0003x 2 - 0.0192x + 1.4885
llQ 1.20
0 .40 R2 = 0.999
1.18
0 .20
1.16
0.00 - 1 - - - - - - , - - - - - - , - - - - . - - - - - -
0.1 0.15 0.2 0.25 0.3 0.35 0.4
-10 10 20 30
Asphalt thickness (m) Temperature (2C)

Figure 5. Typical influence of the asphalt layer thick- Figure 8. Typical influences of the temperature on the
ness on the deflection ratio. deflection ratio.

0.70

0.68 it produces a better adjustment compared to the

0
·; 0 .66
logarithmic law.
Ill:
0
The influence of the asphalt layer temperature
·~
Gl
0.64 on the deflection ratio is presented in Figure 8
llQ y = 0 .04421n(x) + 0.4753 for the case of a pavement with 0.20 m of asphalt
0 .62 R2 = 0 .9998
layer, a subgrade with 20 MPa of stiffness and a
0 .60 distance of 0.30 m. A second degree order equa-
0 50 100 150 tion produces a perfect adjustment of the deflec-
Subgrade stiffness (MPa ) tion ratio for all cases defined in this work.
Figure 6. Logarithmic influence of the subgrade stiff-
ness on the deflection ratio. 5 DEVELOPMENT OF THE MODEL

1.29 For the development of the model, the dependent

1.28 variable considered is the Deflection Ratio (DR),
1 .27 given by the ratio between deflection at 20°C (ref-
0
·; 1 .26 erence temperature) and deflection at test tempera-
Ill:
0
1.25 ture. Having obtained the DR for all the data set
·~
Gl
1.24 and the influence of each dependent variable on
ll
Q
1 .23 DR, namely a linear relationship for distance and
1 .22 subgrade stiffness, a logarithmic for thickness of
1 .21 the asphalt layer and quadratic polynomial for the
0 so 100 150 temperature, the model to predict the DR is con-
Subgrade stiffne ss (MPa) stituted by the multiplication of the laws for each
independent variable, as presented in Equation 4,
Figure 7. Linear influence of the subgrade stiffness on where βi are coefficients statistically defined, d is
the deflection ratio. the distance in meters, H is the thickness of the
asphalt layer in meters, E is the stiffness of the sub-
grade in MPa, and T is the temperature in °C.
some cases where the error obtained by this law is
relatively large.
The influence of the subgrade on the deflec- ⎧( β1 + β 2 (T ) d) ×
⎪( β 3 + β 4 (T − 20 ) log( H )) ×
tion ratio follows a logarithmic and a linear law =⎨ (4)
( β + β (T − 20 ) E ) ×
depending on the thickness of the asphalt layer. ⎪ 5 26
If this thickness is at least 0.25 m, the relationship ⎩( β 7 ⋅ Ti + β 8 ⋅ T + β 9 )
between the deflection ratio and the subgrade stiff-
ness changes from logarithm to linear, as indicated The coefficients of the Equation 4 were deter-
in Figure 6 for a pavement with 0.15 m of asphalt mined with the least squares method to deter-
layer at 30ºC for a distance of 0 m, and in Figure 7 mine the parameters β to minimize the sum of the
for a pavement of 0.25 m of asphalt layer at 10°C squares of the errors made in the results of every
for a distance of 0.20 m. equation. The best approximation is presented in
For the development of the model, the linear Table 2. Notice that each group of coefficients β is
variation of the deflection ratio is used because related to a distinct independent variable.

754
Table 2. Coefficients β of the adjust model.

β1 β2 β3 β4 β5 β6 β7 β8 β9

−3.460E-01 −3.957E-03 −2.112E-01 2.318E-03 4.937E+01 2.006E-03 8.021E-05 −1.342E-02 5.136E-01

2.00 sible to conclude, with this value, that 99.7% of the

values of DR estimated by the model are explained
a:: 1.50 by the independent variables thickness of AC layer
.,Q (H), stiffness of the subgrade (E), test Temperature
~ 1.00 (T) and distance from the load application (d).
E
·~ In this work the evaluation of RMSEA for dif-
UJ 0.50
ferent distances from the point of load application
is of great importance. This analysis is presented
0.00 0.50 1.00 1.50 2.00
in Figure 10.
Actual DR
In Figure 10 it is possible to see that the higher
RMSEA is observed for distance d10 (2.10 m from
Figure 9. Actual and estimated DR for d1. load application) followed by distance d9 (1.80 m
from the load application). The best approxima-
tion is made for the distance d3 (at 0.30 m from the
load application point), with RMSEA = 0.029. This
result means that the mean difference between real
and estimated values of DR at distance d3 is 0.029.

6 VALIDATION OF THE MODEL

After the development of the model, its validation

dl d2 d3 d4 d5 d6 d7 dB d9 dlO through in situ conditions is important. For this part
Sensor
of the work, tests were performed in situ with falling
Figure 10. RMSEA for each distance from load weight deflectometer and the asphalt layer tempera-
application. ture was measured with thermocouples at the mid-
dle of the asphalt layer. Tests started in the morning
and ended when asphalt layer temperature reached
In this study was possible to obtain p-value of 30ºC. The section of pavement is constituted by an
each coefficient β, working p-value as a measure asphalt layer with 0.20 m and 0.20 m of granular
of significance for decision making process in material. In Table 3 the pavement temperature and
hypotheses tests. The p-value obtained for all vari- the deflections obtained with a falling weight deflec-
ables was equal to 0, which means that all the coef- tometer are indicated, normalized for a load of
ficients β are indispensable to the model. 40 kN with 0.15 m of radius. The resulted deflection
The comparison between real DR and DR esti- basin of each test is represented in Figure 11.
mated by the model is presented in Figure 9, where In Figure 11 it is possible to verify that the dif-
is possible to see that the discrepancy between val- ferences among deflection basins are localized in
ues is not significant, both for d1 (distance = 0). the first sensors meaning that only the stiffness of
This conclusion is the same for the rest of dis- the asphalt layers changed during the tests. This
tances of load not presented here.
Having the development of the model been done it Table 3. Temperature and deflections (mm/1000).
is now necessary to perform the analysis of residuals
and the appropriateness of the data used for its devel- Distance (m)
opment. The evaluation of the quality of the model Temp.
can be made through the analysis of the coefficient of Time (ºC) 0 0.3 0.45 0.6 0.9 1.2 1.5 1.8 2.1
determination (or r-squared) R2 and the Root Mean
09:19 19 129 104 92 79 58 41 30 24 18
Squared Error of Approximation (RMSEA).
10:21 22.5 136 109 94 81 58 41 30 22 19
Obtaining r-squared makes possible the quanti-
11:12 23 143 113 99 84 60 42 31 22 17
fication of the total variation of the response varia- 12:15 26 155 120 103 88 61 42 30 22 17
ble that is explained by the developed model. In this 13:38 30 178 130 111 92 62 43 30 22 18
study it was obtained R2 = 0.997. Thus, it is pos-

755
 Distance(m)
A parameter named Deflection Ratio (DR) was
defined to correlate the measured pavement deflec-
20
tions at any temperature with the deflections at
20ºC. The multiplication of the measured deflec-
tion by the deflection ratio gives the deflection at
20ºC. This deflection ratio was plotted against dis-
tance of the sensors to the load application point,
asphalt thickness, subgrade stiffness and tempera-
ture to study the influence of these variables on the
deflection ratio. These influences were used to con-
Figure 11. Deflection basin obtained with falling weight struct the model to adjust the pavement deflection
deflectometer.
to 20°C.
The quality of the model was ensured by the
Dista nce(m)
analysis of p-value, the comparison between the
actual and estimated deflection ratios and the R2
parameter. All analysis proved the quality of the
20
model as well as the application of a real case where
g 40

'E
in situ measurements were adjusted with the model.
.s
ACKNOWLEDGEMENTS

160 The first author would like to express the support

of Portuguese national funding agency for science,
Figure 12. Deflections adjusted to T = 20ºC. research and technology (FCT) through scholar-
ship SFRH/BSAB/114415/2016.

fact is only attributed to the temperature because

the traffic was reduced during the test period. REFERENCES
To the validation of the adjust model, the char-
acteristics of the pavement were applied to Equa- AASHTO, 1993. Guide for design of pavement struc-
tion 4, with the subgrade modulus obtained from tures American Association of State Highway and
back-analysis, allowing the calculus of DR and Transportation Officials, ed., Washington, D.C.
deflection at reference temperature. The results of Asphalt Institute, 1983. Asphalt Overlays for High-
way and Street Rehabilitation, Maryland: Asphalt
this study are presented in Figure 12.
Institute.
As shown in Figure 12 the model developed Fernando, E., Liu, W. & Ryu, D., 2001. Development of
for the adjustment of the pavement deflection a procedure for temperature correction of back cal-
works perfectly in spite of some small differences culated AC modulus, Texas Transportation Institute,
obtained in several sensors inherent to any type of The Texas A&M University System, College Station,
model developed through a statistical process. Texas, USA.
FHWA, 2000. LTPP Manual for Falling Weight Deflec-
tometer Measurements, Maryland.
7 CONCLUSIONS Kim, Y.R. & Park, H., 2002. Use of Falling Weight
Deflectometer Multi-Load Data For Pavement
Strength Estimation. Report No. FHWA/NC/2002-
This paper developed a model to adjust deflections 006, Raleigh, USA.
obtained during falling weight deflectometer tests Lukanen, E.O., Stubstad, R. & Briggs, R., 2000. Temper-
to a reference temperature based on the asphalt ature Predictions and Adjustment Factors for Asphalt
layer thickness, subgrade stiffness, testing tempera- Pavement, Publication N. FHWA-RD-98-085, Fed-
ture and distance from the measuring point to the eral Highway Administration, VA, USA.
load application point. Minhoto, M.J.C., Pais, J.C., Pereira, P.A.A., 2008. The
Laboratory tests were carried out to assess the temperature effect on the reflective cracking of asphalt
stiffness of typical asphalt mixtures represented of overlays, Road Materials and Pavement Design. vol 9,
nº 4, p. 615–632.
wearing course, binder and base layers. For each
SHRP, 1993. SHRP Procedure for Temperature Correc-
of these asphalt mixes, three different air-void tion of Maximum Deflections, Washington, DC.
contents were considered. The laboratory tests Sousa, J.B., Pais, J.C., Prates, M., Barros, R., Langlois, P.,
allowed to obtain typical values of the asphalt Leclerc, A.M., 1998. Effect of aggregate gradation on
mixes as function of the temperature to be used in fatigue life of asphalt concrete mixes, Transportation
the model for the deflections adjustment. Research Board, p. 62–68.

756
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Development of a frequency temperature correction model

for FWD back-calculated moduli based on frequency-temperature
superposition principle

Julius Marvin Flores, Phuc Le Van, Chang Kyu Park, Wonjae Kim & Hyun Jong Lee
Department of Civil and Environmental Engineering, Sejong University, Seoul City, South Korea

ABSTRACT: A frequency temperature correction model for Falling Weight Deflectometer (FWD)
back-calculated modulus was developed with the consideration of FWD loading and pavement tempera-
ture effect. Indirect tensile dynamic modulus and FWD testing were conducted to determine laboratory
and field dynamic modulus, respectively. Frequencies of 33 and 16 Hz were evaluated as suggested by
other researchers as the FWD representative frequency. It was found that the reduced frequency at the
specific depth represents the FWD frequency based on the back-calculated and laboratory measured
modulus. In addition, mid depth temperatures were found to be the most suitable pavement temperature
in representing FWD back-calculated modulus. A regression equation of frequency and temperature was
developed as a function of shift factor, slope of master curve and frequency. The correction model was
verified using several frequencies and temperatures. It was found that temperatures and frequencies that
are closer to field condition give lesser modulus prediction error. The frequency temperature correction
model was validated and compared with other correction models; it was found that the developed regres-
sion equation gives the least prediction error of 8.2%.

1 INTRODUCTION back-calculated modulus (Baltzer and Jansen,

1994, Kim et al., 1995, Lukanen et al., 2000, Chang
Evaluation of the current condition of the Asphalt et al., 2002 and Appea, 2003). These models were
Concrete (AC) pavement is essential in pavement developed by regression of field data only in which
rehabilitation. Currently there are several pave- calibration and validation is needed. In addition,
ment evaluation techniques being practiced by these models only considered temperature effect
several agencies. The Falling Weight Deflectom- without frequency.
eter (FWD) testing is one of the most common Moreover, some researchers tried to interpret
tools used in evaluating existing pavement condi- FWD frequency with the laboratory frequency.
tion especially in Seoul City. The FWD equipment Due to the FWD loading time of 0.03 s (Loulizi
applies an impulse load at the surface of the pave- et al., 2002), some researchers suggested a 33 Hz
ment in which deflections of each layer are meas- laboratory frequency (Kim and Lee, 1995). On the
ured by evenly spaced geophones attached into contrary, Witczak et al described the FWD loading
the equipment. From these pavement deflections, as highway speed with a frequency of 10 Hz while
modulus of asphalt concrete, aggregate subbase Lytton et al. proposed a 16 Hz reduced frequency
and subgrade can be back-calculated. from the asphalt surface. There are still ongoing
Asphalt concrete is a viscoelastic material in discussions on which is the real FWD representa-
which dynamic modulus is dependent with fre- tive frequency.
quency and temperature. During FWD testing, Hence, a new correction model must be devel-
temperature may vary in different testing sections oped not only based from field data regression
or location and even within the asphalt pavement analysis but also from laboratory testing. In addi-
layer. Since FWD testing is conducted on a specific tion, the correction model must consider not only
temperature and frequency from an instantaneous temperature but also frequency effect to evaluate
loading, back-calculated modulus must therefore vehicle loading.
be corrected to a reference temperature so that In this study, FWD testing was conducted at six
evaluation of the pavement condition for a certain (6) regions wherein two to three 150 mm cores were
rehabilitation length is similar. taken. Twenty-nine (29) samples were cored spe-
Several researchers proposed temperature cifically where the FWD testing was made. Indi-
correction models to properly evaluate FWD rect Tensile (IDT) dynamic modulus testing was

757
conducted to cored samples to measure dynamic 2.2 Laboratory testing
modulus at several temperature and frequency in
Cores 150 mm in diameter were taken directly
the laboratory. Measured deflections of the pave-
under where the FWD testing was conducted to
ment during FWD testing were back-calculated to
properly relate the dynamic modulus. Due to the
determine the existing pavement modulus. In addi-
observed pavement thickness variation on each sec-
tion, specific pavement temperatures and frequen-
tion, cored samples in each section varies depend-
cies in the field were estimated through prediction
ing on the pavement thickness.
models. A frequency and temperature correction
Shown in Table 1 are each pavement description
model for asphalt pavement modulus was devel-
including location, pavement life and correspond-
oped based on frequency-temperature superposi-
ing Seoul Pavement Index (SPI).
tion principle.
The cored samples were cut into 50 mm thick
laboratory specimen which is the standard thick-
ness for IDT dynamic modulus testing (Kim et
2 TESTING APPROACH
al., 1995). Due to thickness variation, some have
two specimens from one cored sample while others
2.1 Field testing
have only one also shown in Table 1.
The FWD is a nondestructive testing device used in The specimens were subjected to IDT dynamic
the simulation of pavement surface deflection due modulus testing (Kim et al., 1995) at different fre-
to traffic. It is usually attached to a vehicle in which quencies (25, 10, 5 and 1 Hz) and different tem-
it applies a pulse load to the pavement by dropping peratures (–10, 5, 20 and 35°C). The dynamic
weight creating deflections under the surface which modulus of each specimen was determined using
is measured by geophones as shown in Figure 1. the sigmoidal equation:
Six (6) regions in Seoul city were surveyed
wherein twenty-nine (29) road sections were tested
for FWD and field coring as shown in Table 1. The Table 1. Pavement test sections.
FWD sensors were spaced at 0, 20, 30, 45, 60, 90,
and 150 cm wherein plate radius is 15 cm. Four Overlaylife No. of
load drops were applied which are approximately Region Route (years) SPI specimens
22, 44, 53, and 71 KN wherein surface deflections
and pavement temperatures were measured by Kangseo MO63 13 5.13 2
MO65 13 5.13 2
FWD sensors.
CH73 13 5.56 2
From the FWD testing, pavement deflections
Nambu SI82 5 5.05 1
were used to back-calculate existing pavement’s
SI83 5 5.05 1
modulus. The FWD test results were back-calcu-
SI84 5 5.05 2
lated using the Modulus 6.0 program. Pavement
HO92 4 2.80 2
was considered to have three (3) layers: asphalt, HO93 4 2.80 2
aggregate base and subgrade layer. The bedrock HO94 4 2.80 1
option was used to consider its effect to FWD Seobu GA143 5 2.80 1
deflections; also, specific pavement layer thick- GA145 5 2.80 1
nesses and temperature were used in the back-cal- SO182 9 3.02 2
culation program. A synthetic database of FWD SO183 9 3.02 2
back-calculated modulus was then generated SO184 9 3.02 1
which was used for further analysis. Bukbu SA202 5 5.79 1
SA203 5 5.79 1
JO231 11 5.05 2
JO233 11 5.05 2
JO235 11 5.05 2
JO242 14 3.97 2
JO244 14 3.97 2
Dongbu NG253 5 4.95 2
NG255 5 4.95 1
Seongdong NGO314 3 5.54 2
KO322 4 4.37 1
KO324 4 4.37 2
KO325 4 4.37 2
MA332 4 6.01 2
MA335 4 6.01 1
Figure 1. FWD cross section diagram.

758
 2 ,---------------------------------, mate effective loading time at any depth within AC
layer in pavement design to accurately determine
y = 0.0793x - 1.6667
R 2 = 0.9962
pavement strength. In this study, the representative
FWD frequency was determined by evaluating 33
and 16 Hz frequencies as suggested and proposed
by other researchers as FWD representative fre-
quency. In addition, an effective reduced frequency
from the 33 Hz FWD surface frequency to a spe-
cific depth was also evaluated. To consider this,
loading time at any point within the asphalt layer
was calculated. Loading time depends on vehicle
speed, tire contact radius and AC effective depth
-3
-20 - 10 0 10 20 30 40 as shown in the following equation proposed in
ARA, 2004:
Temperature (°C)

Figure 2. Shift factor curve of specimen SI84.

t=
(
2 * ar deff ) (2)
277.8 *V
I 00000 -.---------------------------------,
wher et is the loading time (s), ar is the tire contact
Reference Temp =20 °C
radius (mm), deff is the effective depth (mm), and V
is the vehicle speed (kph).
.-1_0000 Using 33 and 16 Hz as the representative fre-
"'
"-
6 quency based from the studies several research-
ers, the back calculated and laboratory measured
*w
-1000 modulus were compared as shown in Figure 4. As
seen in the figure, the dynamic modulus consider-
ing 33 Hz is underestimated with an error of 13%
100 +--~~..,._~~..,._~~..,._~~"1-~~
whileat16 hz frequency is overestimated with an
error of 21%.
0.0001 0.01 100 10000 I 000000
Further analysis was conducted consider-
Reduced Frequency (Hz)
ing the reduced frequency to a specific depth
Figure 3. Dynamic modulus curve of specimen SI84.
from a 33 Hz surface frequency using equation
2. It was found that at this frequency, the error
α between the back-calculated and measured mulus
log E * = δ + (1) is 7% suggesting that at reduced frequency, FWD
1 + e β +γ lloggξ back-calculated modulus is more accurately
where E * is the dynamic modulus, δ, α, β represented.
and γ arefitting parameters, and ξ = the reduced
frequency.
Master curves at 20°C reference temperature were
then generated for each specimen. Figures 2 and 3 20,000 , - - - - - - - - - - - - - - - 7 1
• 33 Hz effective loading time at specific depth
show a sample shift factor and dynamic modulus 033 Hz
master curve graph generated for one specimen 6 16Hz
(SI84). Each specimen yields different shift factor ~ 15,000
coefficients and sigmoidal parameters based on ~

the results of IDT dynamic modulus testing using :g"'

~ 10,000
equation 1. Regression coefficients and fit parame-
ter values for all specimens were collected and were
.,
used for further analysis. ~
~ 5,000

3 FREQUENCY AND
0-l"----------,-------.----------,----------j
TEMPERATURE EFFECT
0 5,000 10,000 15,000 20,000
Backcaluated Modulus (MPa)
3.1 Frequency effect evaluation
The stiffness of asphalt mixture varies with load- Figure 4. Evaluation of the relationship of laboratory
ing frequencies. It is important to accurately esti- to FWD frequency.

759
3.2 Temperature effect evaluation perature can represent the FWD back-calculated
dynamic modulus accurately.
Asphalt concrete mixtures are viscoelastic materi-
als in which it behaves not only based on frequency
but also temperature. During FWD testing, the
4 FREQUENCY-TEMPERATURE
weight is released from a certain height to the sur-
CORRECTION MODEL
face in which sensors attached to the equipment
measures the atmospheric and surface temperature
It is well known that the behavior of asphalt con-
necessary in FWD back-calculation procedure.
crete is a function of temperature and loading
However, temperature within the asphalt layer
time or frequency wherein dynamic modulus at
varies at different depths.
different temperature can be shifted relative to a
In this study, the effect of temperature was
reference frequency. A single master curve is con-
evaluated by considering the specific temperature
structed based on the alignment of various curves.
during the time of FWD testing at a certain depth.
Based on the frequency temperature superposition
Temperature of AC layer at a certain depth from
principle, asphalt dynamic modulus is a function
the surface is predicted using Bells equation with
of the reduced frequency defined by:
coefficients developed by Lukanen et al. (2000) as
shown:
f
ξ (ξ ) = log ( ) − ( ) (4)
Td 2.8 0.894 * IIR ⎡⎣logg (d ) . ⎤⎦ * [ . * IR
I at
+ 0.77 * ( − day) + 3.763 763 * ( h 18) ⎤⎦
where ξ is the reduced frequency, f is the frequency
i ( hr − 188) ⎤⎦ * [0.474 0.031* IR ]
+ ⎣⎡ sin (3) and at is the shift factor.
Meanwhile, the shift factor as a function of tem-
perature, at (T), can be defined by a linear relation-
where Td is the pavement temperature at depth
ship shown in the following equation:
d in °C, IR is the Pavement surface temperature
(°C), d is the depth at which temperature is to be
log at = a1T a2 (5)
predicted (mm), 5-day is the 5 day mean air tem-
perature (°C) and hr is the time of FWD testing in
24 hour format. where T is the temperature (ºC) and a1, a2 are
To consider the effect of pavement temperature regression coefficients.
within depths, IDT dynamic modulus at differ- Considering the linear range of the dynamic
ent depth ratios and AC thickness (0.3, 0.4, 0.5, modulus master curve as shown in Figure 6, the
0.6, and 0.7) were measured and were compared frequency-temperature correction model can be
to back-calculated dynamic modulus as shown described as follows:
in Figure 5. Moduli at mid-depth AC layer were
found to have the lowest average error of 4.6% log E*=k1 logξ +k
+ 2 (6)
compared to other layers. Hence, mid-depth tem-
Back-calculated dynamic modulus, |E*|back, and
reference dynamic modulus, |E*|ref at reference

15,000 ~-----------------,

4.5 1.===~~------------~
0 Measured IE* I
0.7 - -Fit
"'
0.. 4.0
?
;:I
10,000
'<;;'
3.5
- IE*lback
- IE*Iref
:; 0..
"0

~ 6 3.0
"0

~ 5,000
*t8_ 2.5
Oil
0
"' ....l
:2 2.0

1.5
0~------.-------~------~
0 5,000 I 0,000 15,000 1.0 +--------,---..liL,---L-----,-----1
Backcaluated Modulus (MPa) -3.0 - 1.0 1.0 3.0 5.0
Log Reduced Frequecy (Hz)
Figure 5. Evaluation of the effect of analysis depth to
fwd back-calculated modulus. Figure 6. Sample dynamic modulus master curve.

760
temperature and frequency, (tr, fr) were substituted Considering the frequency, comparison of
into equation 6, resulting to equations 7 and 8: measured and predicted modulus is shown in
Figure 7. Reference frequencies of 10 Hz, field
log E refe
*
l gξrefeef
k1log k2 (7) reduced frequency and 25 Hz have an average error
of 15%, 8.2% and 9.1% respectively when estimat-
ing AC modulus.
l gξt f
log|E|*back = k1 log k2 (8) This shows that the further the reference fre-
quency deviate from field frequency, the higher
Equation 8 is then subtracted from equation 7 error it gives. This result is totally suitable for
to obtain equation 9: frequency temperature superposition princi-
ple. Hence, when reference frequency is same as
k1 field frequency, field modulus can be predicted
| |*refe ⎛ ξrrefe ⎞ accurately.
=⎜ ⎟ (9)
⎝ ξt , f f ⎠
*
E back Three reference temperatures (10°C, 20°C,
and 30°C) were used to also further validate the
study. The comparison of measured and pre-
The reduced frequency in Equation 9 can be dicted modulus using different reference tempera-
represented by the normal frequency and tempera- tures is shown in Figure 8. Average error of 5.8%,
ture using equations 4 and 5. 8.2%, and 25%, were found for 10°C, 20°C, and
Finally, frequency and temperature is intro- 30°C reference temperatures respectively. Refer-
duced in the equation as:
k1
12,000 .---------------~

Ere*ef ⎛ f ⎞
= ⎜ r * 10 a1 (t −tr ) ⎟ (10) 10,000
*
Eback ⎝ ff ⎠
8,000
where E*ref is the dynamic modulus at reference cA
temperature and frequency, E*back is the back-cal- 6,000 D 0 t::.
culated dynamic modulus at specific temperature
and frequency, fr is the reference frequency, ff is the 4,000
frequency at specific depth, a1 is the slope of the
log shift factor versus temperature curve, and k1 is 2,000
the slope of the log modulus versus log reduced
frequency curve. 0 ~--,--,--,---,--,,-~
Using the average shift factor coefficients and 0 2,000 4,000 6,000 8,000 I0,000 12,000
sigmoidal parameters from the dynamic modulus Measured Modulus at Reference Temp (MPa)
database obtained from laboratory testing of the
thirty-six (36) specimens, it was found that α1 is Figure 7. Evaluation of the effect of frequency to the
0.075 and k1 is 0.211. These were used to obtain proposed correction model.
the final form of the equation as shown below:
0.211 12,000 - . - - - - - - - - - - - - - - - - - - ,
Ere*ef ⎛ f ⎞
*
Eback
= ⎜ r * 10 0.075(t tr ) ⎟
⎝ ff ⎠
(11)
~ 10,000 D

l "
0::
8,000 D

~ 6,000
5 VALIDATION OF THE MODEL
:g"
~ 4,000
The remaining eleven (11) cored specimens
obtained from road sections in Seoul city were "0

t 2,000
used in the validation study. The dynamic modu-
lus of specimens were measured at a reference ~ 0 *----.----.----.----.----.--~
temperature (20°C). Pavement temperature at spe- 0 2,000 4,000 6,000 8,000 I 0,000 12,000
cific depth and frequency in the field are predicted Measured Modulus at Reference Temp (MPa)
using equations 2 and 3. Three reference frequen-
cies (10 Hz, field frequency and 25 Hz) were then Figure 8. Evaluation of the effect of temperature to the
used to predict the AC modulus. proposed correction model.

761
ence temperature at 10°C provided lowest error field temperature range, the closer the reference
because FWD testing in eleven road sections was temperature is from field (FWD) temperature, the
done at low temperature (–2°C to 20°C). Since lower error in estimating AC modulus. Therefore,
the proposed model considers log modulus versus the proposed model can predict modulus more
reduced frequency as linear form at reference and accurate if reference temperature is closer to field
testing temperature.
~ 12,000 - , - - - - - - - - - - - - - - - - - - - - , . , Furthermore, four (4) similar AC modulus tem-
perature correction models were compared with
i 10,000 the proposed model as shown in Table 2. The dif-
~ ferent existing temperature correction models were
~ 8,000 applied to eleven road sections for error compari-
"""'
~
"' 6,000
~
• D Baltzer&Jansen (5)

I> Kim el al.(6)

son. Using these temperature correction factors,
backcalculated modulus at different temperature
were converted toa reference temperature.
:§0 4,000 x Lukanen et al.(7) To compare the proposed model to other exist-
::E o Chang el al. (8)
ing models, the field frequency was assumed to be
] 2,000 the same with the reference frequency reducing the
<.)

'i3 • Proposed Model equation to temperature factors only. With this,

£ 0 ~----.-----.-----.-----.-----.---~ the coefficient of the proposed model is –0.0157 as
0 2,000 4,000 6,000 8,000 I0,000 12,000 shown in the table.
Measured Modulus at Reference Tempt. (MPa)
Modulus error in Table 2 indicates that model
Figure 9. Comparison of the proposed frequency tem- developed by Kim et al. (error of 8.8%) based on
perature correction model to other existing models. regression analysis of field data is similarwith the
proposed model (error of 8.2%) which is based on
Table 2. Review and evaluation of similar temperature shift factor and master curve slope. Meanwhile,
correction models. models developed by Baltzer and Jansen (1994)
and Lukanen et al (2000) provided smaller modu-
Temperature Average lus compared to the models proposed by Chang
Author models error (%) et al. (2002) which has significant modulus error that
E refe is maybe due to several parameters such as specific
Baltzer and 10.1
= 10 −0.018 ( − Tac )
pavement locations, mixture types, existing pave-
Jansen, 1994 E AC ment conditions, backcalcultion program, and etc.
Kim et al., 1995 E68 8.8
= 10 −00.0153(68 −T ) 6 CONCLUSION
ET
Lukanen et al., 2000 ATAF = 10 slope l (Tr Tm ) 12.3
Based on the findings of this study, the following
Chang et al., 2002 Er 29.9 conclusions can be made:
= 10 −0.02822 (25 −T )
E FW
WD
• A temperature and frequency correction fac-
Proposed Model E refe 8.2
= 10 −0.0158 (Treef T ) tor was proposed using frequency temperature
E FW
WD
superposition principle. The frequency tem-
perature correction factor is a function of tem-
Note: The following are the parameters used in each tem- perature, frequency, the slope of log shift factor
perature correction model. versus temperature curve, and slope of log mod-
Eref is the asphalt modulus at reference temperature ulus versus log reduced frequency curve. Using
(20°C), EAC is the backcalculated modulus at 1/3depth this model, asphalt dynamic modulus at refer-
AC temperature ence temperature can be estimated accurately at
E68 is the asphalt modulus at reference temperature 68°F varying loading speeds or frequencies for asphalt
ET is the backcalculated asphalt modulus at mid depth pavement design.
temperature of AC layer, T • Reduced frequency at specific depth can be used
ATAF is the asphalt temperature adjustment factor to accurately determine the dynamic modulus
Slope is the slope of the log modulus curve, recom- from FWD loading with impulse loading dura-
mended as –0.0195 in wheel path and –0.21 for mid
lane
tion of 0.03 seconds.
Tr is the reference temperature • Mid-depth temperature can represent the asphalt
Tm is the mid depth temperature layer in predicting AC dynamic modulus accurately.
Er is the adjusted modulus to 25°C • Reference frequency and temperature closer
Eo is the measured modulus at temperature, TC to field frequency and temperature gives much
Tc is the mid depth asphalt pavement temperature accurate modulus prediction.

762
• Regression equation for asphalt frequency tem- Pavement in Taiwan Using the Falling Weight Deflec-
perature correction factor was developed based tometer. International Journal of Pavement Engineer-
on average slope of log shift factor versus tem- ing, vol. 3.2002, pp. 131–141.
Kim, Y. R., and Y. C. Lee. Interrelationships Among-
perature curve. The equation was validated and
Stiffnesses of Asphalt-Aggregate Mixtures. Journal of
compared to current similar models for temper- Association of Asphalt Paving Technologists, Vol. 64,
ature correction factor using FWD testing in the 1995, pp. 575–609.
field. Results indicated that the proposed model Kim, Y. R., B. O. Hibbs, and Y. C. Lee. Temperature Cor-
can be capable of estimating the dynamic modu- rection of Deflections and Backcalculated Moduli. In
lus value accurately. Transportation Research Record: Journal of the Trans-
• Based on the limitions of available data, it is portation Research Board, No. 1473. National Research
recommended to further evaluate the equation Council, Washington, D.C., 1995, pp. 55–62.
on different pavement conditions in the field Kim, Y. R, Y. Seo, M. King, and M. Momen. Dynamic
Modulus Testing of Asphalt Concrete in Indirect
to be able to validate and calibrate the findings
Tension Mode. In Transportation Research Record:
obtained in this study. Journal of the Transportation Research Board, No.
1891, Transportation Research Board of National
ACKNOWLEDGEMENT Academies, Washington, D.C., 2004, pp. 163–173.
Leiva-Villacorta, F. and D. Timm. Falling Weight Deflec-
tometer Loading Pulse Duration and Its Effect on
This research was partially supported by a grant Predicted Pavement Responses. In Transportation
(15TLRP-C099511-01) from Transportation & Research Record: Transportation Research Board 92nd
Logistics Research Program (TLRP) funded by the Annual Meeting, No. 2171. Transportation Research
Ministry of Land, Infrastructure, and Transport Board of National Academies, Washington, D.C.,
of Korean Government and the Seoul Metropoli- 2013, pp. 13–2171.
tan Government. Loulizi, A., I. L. AI-Qadi, S. Lahouar, and Th. E.
Freeman. Measurement of Vertical Compressive
Stress Pulse in Flexible Pavements. In Transporta-
REFERENCES tion Research Record: Journal of the Transportation
Research Board, No. 1816, Transportation Research
Appea, A. K. Validation of FWD Testing Results at the Board of National Academies, Washington, D.C.,
Virginia Smart Road: Theoretically and by Instrument 2002, pp. 125–136.
Responses. Ph.D. dissertation. The Virginia Polytech- Lukanen, E. O., R.N. Stubstad, and R. Briggs. Tempera-
nic Institute and State University, Blacksburg, Vir- ture Predictions and Adjustment Factors for Asphalt
ginia, USA, 2003. Pavement, Report No. FHWA-RD-98-085. FHWA,
ARA, Inc., ERES Consultants Division. Guide for U.S. Department of Transportaion, 2000.
Mechanistic–Empirical Design of New and Reha- Lytton, R. L., F. P. Germann, Y. J. Chou, and S. M. Stof-
bilitated Pavement Structures. Final report, NCHRP fels. Determining Asphaltic Concrete Pavement Struc-
Project 1–37A. Transportation Research Board of the tural Properties by Nondestructive Testing. National
National Academies, Washington, D.C., 2004. http:// Cooperative Highway Research Program Report 327,
www.trb.org/mepdg/guide.htm. Transportation Research Board, Washington, D. C.,
Baltzer, S., and J. M. Jansen. Temperature Correction of 1990.
Asphalt-Moduli for FWD-Measurements. In Proc., Witczak, M. W., K. Kaloush, T. Peillinen, M. El-Basy-
4th International Conference on Bearing Capacity of ouny, and H. Von Quintus. NCHRP Report 465:
Roads and Airfields, Vol. 1, Minneapolis, Minn., 1994. Simple Performance Test for Superpave Mix Design.
Chang, J. R., J. D. Lin, W. C. Chung, and D. H. Chen. Transportation Research Board of National Acad-
Evaluating the Structural Strength of Flexible emies, Washington, D.C., 2002.

763
Structural evaluation
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Evaluation of soil bearing capacity by plate load test

Qasim Al-Obaidi
University of Technology, Baghdad, Iraq

Ali Al-Shamoosi & Azad Ahmed

Andrea Engineering Laboratory, Baghdad, Iraq

ABSTRACT: The accurate estimation of soil bearing capacity plays an important role in the design
of a stable foundation for road pavement and airfields projects. Plate load test is one of the quick, direct
and economic methods for soil bearing capacity determination. This method is useful to solve the design
problems for subgrade and subbase layers, where application of high wheel loads is expected due to heavy
traffic during construction and service life of the pavement. This study is interesting to investigate, analyze
and discuss the geotechnical behavior of soil layers selected from two locations south of Iraq.
On the other hand, the comparison is made between the values of bearing capacity, modulus of sub-
grade reaction and elasticity of soil determined by carrying out field plate load tests and conventional
laboratory tests. Furthermore, the effect of stress distribution and test condition on the shape of elasticity-
stress curve and modulus of subgrade reaction are discussed in detail.

1 INTRODUCTION areas (Coduto, 2001). On the other hand, the out-

put of the plate load test is commonly relative to
The soil layer underneath the engineering struc- direct measurement of modulus of subgrade reac-
tures such as rigid pavement and airfields projects tion (Ks) where the Ks parameter is used as a pri-
is commonly designed for carrying the applied mary input for rigid pavement design. Typically,
loads placed upon it. the modulus of subgrade reaction is determined
Application of soil mechanics in the design of from field plate load test where there is no direct
subgrade and subbase layers is necessary to study laboratory method to measure it. However, the Ks
the stress distribution and geotechnical properties value can be estimated from some stiffness tests as
of sub-soil material. indirect procedure.
Furthermore, the accurate estimation of soil Some studies in Iraq investigated the bearing
bearing capacity plays an important role in the capacity of the soil calculated from laboratory and
design of a stable foundation during construction field tests and tried to find an empirical relation-
and service life of pavements. ships between them such as:
Since damaging may result from foundation Al-Obaidi (2008) investigated the effect of sub-
failure (i.e. collapse) as well as from excessive base layer thickness and relative compaction on
settlement, the bearing capacity could be evalu- the values of bearing capacity and modulus of sub-
ated from one of the following main methods: grade reaction. The study found that the increases
(1) Based on geotechnical soil properties according of the layer thickness and relative compaction lead
to Terzaghi equation with modification suggested by to smoothly increase in both of bearing capacity
Meyerhof (1963) & Terzaghi et al. (1996), (2) From and modulus of subgrade reaction.
Standard Penetration Test (SPT), and (3) Direct Ahmad (2009) concluded that the bearing
method using field plate load test (see section 3). capacity values obtained from laboratory test
The plate load test method is used to avoid using Terzaghi equation and from field plate load
comprehensive geotechnical exploration of high test seems to be similar for shallow soil layer.
cost and time especially for light load structures.
However, geotechnical engineers can be used the
plate load test results for quick estimation of the 2 TEST PROGRAMME AND
soil bearing capacity. The plate load test is still METHODOLOGY
useful for design problems of shallow depth, such
as those involving wheel loads on pavement sub- The test program of this work is divided into two
grades, where the service loads act over smaller parts: Laboratory tests and Field plate load test.

767
The soil investigation was made for two sites, the plate to an accuracy of at least 0.01 mm is
first one named as West Qurna (w) and the second used.
site named as Faw (F). Both sites are located in • Miscellaneous Apparatus: Such as loading
Basra Governorate, about 550 km south of Bagh- columns and steel stands.
dad, Iraq. The soil investigation consists of drilling
three boreholes to a depth of 3–4 m for each site.
Several tests were conducted on the undisturbed 3 THEORIES FOR RESULTS ANALYSIS
soil sample such as classification, compaction and
consolidation tests. The ultimate and allowable bearing capacity (qu
The standard method for a field load test (PLT) and qa), Elasticity (Es), Modulus of subgrade
given by the American Society for Testing and reaction (Ks) and other geotechnical parameters
Materials (ASTM-D1194-1997) was followed in of the soil samples are calculated according to fol-
this work. The field Plate Load Test (PLT) consist lowing equations as mentioned in Bowles (1997):
of doing 3 test points for each site. The test was
carried out at depth equal to 0.5 m from natural qu CN
CNc qNq 0 5Bγ Nγ (1)
ground surface and at dry condition.
The width of the test pits were at least four times Where: qu = Ultimate bearing capacity (kPa);
the width of the bearing plate used for this test. The Nc, Nq, and Nγ = Bearing capacity parameters (−);
bearing plate is placed on the soil at the bottom of C = Soil cohesion (kPa); q = Overburden pressure
the pit, and an incremental load on the bearing plate (kPa); B = Foundation width (m) and γ = Soil den-
is applied. After the application of an incremental sity (kN/m3).
load, enough time is allowed for settlement to occur.
Ks
When the settlement of the bearing plate becomes qa = (2)
negligible, the next incremental load is applied. In 40 FS
this manner, a load-settlement plot is obtained.
The results of elasticity and modulus of sub- Where: qa = Allowable bearing capacity (kPa);
grade reaction obtained from consolidation test FS = Factor of safety taken as 3, and Ks = Initial
and plate load test were calculated based on the tangent modulus of subgrade reaction (kPa/m).
theories mentioned in the paragraph of “Theories The equation (2) above based on total deforma-
used for results analysis”. tion settlement of ΔH = 25 mm, for ΔH = 6, 12 and
In this work, the following equipment and test 20 mm the constant 40 can be adjusted to 160, 83
conditions are used as shown in Figure 1: and 50 respectively.
• Bearing Plate: Circular steel bearing plates, not Δσ
Ks = (3)
less than 2.5 cm thickness, 2000 cm2 in area and ΔSettelment
50.4 cm in diameter is used.
• Hydraulic Jack with Pressure Gauge: Hydraulic Where: Δσ = difference between two successive
jack with pressure gauge of sufficient capacity vertical stress (kPa), and ΔSettlement = difference
is used to provide a maximum vertical load of between two successive vertical displacements (m).
300 KN.
• Settlement-Recording Devices: Three dial gauges, Δσ
Es = (4)
capable of measuring settlement of the test Δ∈

Figure 1. Plate Load Test: (Left) Loading system and (Right) Steel circular plate with three settlements.

768
Table 1. Summary of laboratory tests results, (AETL 2012, 2015).

Site Name *Laboratory tests

West γd W
Qurna (natural) (natural) L.L P.L P.I qu Pc
(W) g/cm3 % % % % Gs kPa eo kPa Cc Cr

1 1.39 34.5 52 21 31 2.71 24 0.95 105 0.25 0.03

2 1.45 30 46 24 22 2.68 56 0.85 105 0.24 0.04
3 1.4 29 42 22 20 2.71 73 0.94 90 0.24 0.04
Faw (F)
1 1.44 31 41 22 19 2.7 62 0.87 95 0.2 0.03
2 1.43 29 44 21 23 2.69 55 0.88 130 0.34 0.04
3 1.42 30 44 23 21 2.69 63 0.89 125 0.29 0.04

*Where γd: natural dry density, w: natural water content, L.L: liquid limit, P.L: plastic limit, P.I: plasticity index, Gs:
specific gravity, qu: unconfined compressive strength, eo: initial void ratio, Pc: pre-consolidation pressure, Cc: compres-
sion index and Cr: rebound index.

Table 2. Summary of field plate load tests results, 0.8

(AETL 2012, 2015).

Site Name **Field plate load test

West Qurna Ks Es qa :::!: 0.7

(W) kPa/m kPa kPa 0
·~
1 13098 6723 109 ...
"0
2 85545 5594 178 ·::;
3 62673 6723 131 >
0.6
Faw (F)
1 4131 8000 34
2 8734 8000 73
3 24671 7273 51
0.5
**The parameter Ks calculated from equation 3, Es calcu- 10 100 1000
lated from equation 4 and qa calculated from equation 2. Vertical stress Gv [kPa]
(a)
Where: Es: Secant modulus of elasticity (kPa), 0.8
and Δ∈: difference between two successive strains. -+-F l
-X-F2
Es
Ks = (5)
B( − ) 0.7

Where: B = width of foundation (m), in this

paper taken as 0.504 m corresponding to width of
the steel plate, and μ = Poisson’s ratio (−), taken
as 0.4 for the saturated silty Clay according to
Table (2–8) in Bowles (1997).

4 TEST RESULTS AND DISCUSSION

0.5
In this section, the laboratory tests and field plate 10 100 1000
Vertical stress G v [ kPa J
load tests results with detailed discussion will be
presented. The summary of the laboratory and (b)
field tests results are shown in Table 1 and 2 respec-
tively. A discussion of the soil tests results which Figure 2. Consolidation test results: (a) West Qurna
were derived from both laboratory and field tests (W) site and (b) Faw (F) site.

769
is as follows: From the results of the physical, clas- void ratio resulting more dense structure of the soil
sification analysis and consolidation tests shown mass. Moreover, the uniform variation of elasticity
in Table 1 and Figure 2, the two soil samples can and modulus of subgrade reaction can be related
be described as medium silty Clay with high com- to that the homogenous soil specimen is subjected
pressibility due to relatively high initial void ratio. to uniform vertical pressure under saturated state
dial gauges. of two way drainage condition.
The relationships between the modulus of sub- From Figure 4 it can be seen that the PLT
grade reaction (Ks) and Elasticity (Es) with respect results showed non uniform trend for the modu-
to vertical stress estimated from both Consolida- lus of subgrade reaction and elasticity with respect
tion Test (CT) and Plate Load Test (PLT) can be to applied pressure, where the Ks and Es values
seen in Figures 3 to 6 respectively. From the con- are high at the beginning of the loading stage
solidation test of the West Qurna (W) site shown in (i.e. below the pre-consolidation pressure) for test
Figure 3, it can be clearly observed that the values points 2 and 3. After that sharp reduction in both
of both Ks and Es are smoothly increases with the Ks and Es was occurred especially when the value
increase of applied vertical pressure. This behavior of applied pressure passed the value of pre-consoli-
is expected due to the reduction in the value of the dation pressure, and then increases. This behavior

40000
-+-Wl-CT-Ks 250000
- - - W2-CT-Ks -0-Wl-PLT-Ks
~ -o- W2-PLT-Ks
_,._W3-CT-Ks ~
=
-~
30000
=
200000 -cr- W3-PLT-Ks
"'"... ~
" ""'~ 150000
..
OJ-

~ ~20000 ~5
,_
""'~
~
... "
~-
....0 ~~ 100000
:{)
....0
=
'0
0
10000 ]
'0 = 50000
~ 0
~
0
0
10 100 1000
10 100 1000
Vertical stress "• [kPa I Vertical stress "• [kPa[
(a)
(a)
100000
-0-Wl-PLT-Es
-D-W2-PLT-Es
_80000 -Cr-W3-PLT-Es
"
~
~ 60000
~
·;::;
.::
~40000

20000

0
0
10 100 1000
Vertical stress "• [kPal 10 100 1000
Vertical stress "• [kPa]
(b) (b)

Figure 3. Variation of Modulus of subgrade reaction Figure 4. Variation of Modulus of subgrade reaction
and Elasticity with vertical stress of West Qurna (W) site and Elasticity with vertical stress of West Qurna (W) site
from Consolidation Test (CT). from Plate Load Test (PLT).

770
can be attributed to the type, transformation and due to stability of the soil structure and uniform
distribution tendency of the applied stresses on the interaction of the soil particles within the load dis-
large scale of partially saturated and non homoge- tribution zone of bubble shape with a depth equal
neous subgrade soil layer. However, the rigid reac- to the double of the foundation width below the
tion of the soil layer upon instantaneous applied ground surface. Furthermore, with incremental
pressure is recorded due increases of bulk density increases of the applied pressure and loading time,
of the soil in short time duration. the soil mass recovers its ability to resist the verti-
Thus the soil particles and pore water demon- cal stress and then the Ks and Es magnitudes are
strated a strongly react upon the applied pressure significantly increases.
at the first time of loading. After that, successive For the Faw (F) site, the variation of Ks and Es
deformation in the soil mass with rearrangement over the vertical stress for the consolidation test in
of the soil particles associated with pore water Figure 5 showed the same trend of those values for
dissipation were occurred and cause consider- West Qurna (W) site.
able reduction in the Ks and Es values. Moreover, On the other hand, the loading procedure of the
the Ks and Es values become relatively constant Faw (F) site divided into two cycles: Loading (L)

40000 40000
-+-Fl-CT-Ks - ¢ - F l-PL T(L)-Ks
-X-F2-CT-Ks - X - F2-PLT (L)-Ks
~ -+- F3-CT-Ks ~ - + - F3-PLT (L)-Ks
= 30000 ~
= 30000
+ - ~ - F I-PLT(RL)-Ks
-lK--· F2-PLT (RL)-Ks
Q

.."'...
'=.... ..
....
~ \ X
-+- F3-PLT (RL)-Ks
\
\
"'~ \
~_§
~... :§20000
.. \

s.
~20000
+ \\
2'~
= =-
..C.:t: \

x
-"'
Q
~
....."'Q

=
"'"' 10000
it \
\

=
~

'C
Q
::;
10000

+
'C

::;
Q
X-- ',,
o-----0
x' '
\~
'!!-
0
10 100 1000 0
10 100 1000
Vertical stress av [kPa]
Vertical stress av [kPa]
(a) (a)
16000 16000
-+-Fl-CT-Es
- ¢ - FI-PLT(L)-Es
------ F2-CT-Es - x - F2-PLT (L)-Es
- + - F3-PLT (L)-Es
12000 12000 - -<>- Fl-PLT (RL)-Es
--lK--· F2-PLT (RL)-Es
- -+- F3-PLT (RL)-Es

+~)(',,
+ ',
<l '-::.:
4000 4000
-k \\
X--X',',
<>------¢ '~
0
0
10 100 1000
10 100 1000
Vertical stress av jkPa] Vertical stress av jkPa]
(b) (b)

Figure 5. Variation of Modulus of subgrade reaction Figure 6. Variation of Modulus of subgrade reaction
and Elasticity with vertical stress of Faw (F) site from and Elasticity with vertical stress of Faw (F) site from
Consolidation Test (CT). Plate Load Test (PLT).

771
with vertical stresses of 50 and 100 kPa and great effect on the value bearing capacity of the
Reloading (RL) with vertical stresses of 50,100 and soil.
200 kPa as shown in Figure 6. The values of Ks 6. The magnitudes of modulus of subgrade reac-
and Es increases with applied vertical stress for first tion and elasticity of the soil layer obtained by
loading cycle, while they decreases with applied direct method using plate load test are more
vertical stress in the second cycle of loading. reliable and accurate than those obtained by
In general, in first stage of loading (i.e. before indirect method using consolidation test.
the pre-consolidation pressure value), the modulus
of subgrade reaction and elasticity estimated from
consolidation test exhibits higher values than from REFERENCES
field plate load test.
This behavior can be related to the way of the Ahmad A.A., Al-Obaidi, Q.A.J. & Al-Shamcy A.A.J.,
pressure distribution which is in two directions in 2009. Evaluation of Bearing Capacity from Field and
case of consolidation test and in three directions in Laboratory Tests. Engineering and Technology Jour-
case of plate load test. After the pre-consolidation nal, Vol. 27, No. 3, 2009, pp. 445–453, University of
Technology, Baghdad, Iraq.
stage, the void ratio for both soils become almost Al-Obaidi, Q.A.J., 2008. Study the Effect of Layer Thick-
constant and their structures are more stable. ness on Modulus of Subgrade Reaction and Bearing
Capacity of Soil Estimated by Plate Load Test. 7th
Scientific Conference for Engineering Researches in
5 CONCLUSIONS Basra, 19-20.03.2008, pp. 1–8, College of Engineer-
ing, Al-Basra University, Basra, Iraq.
1. Plate load test is quite useful for estimating the American Society for Testing and Materials (1997).
bearing capacity of subgrade and subbase lay- Annual Book of Standards (ASTM D1194). Vol. 04.08,
ers and to solve the design problems of the rigid Philadelphia, PA, ASTM, USA. Copyright, ASTM
International, 100 Barr Harbor Drive, PO Box C700,
pavement. West Conshohocken, PA 19428–2959, United States.
2. The values of both elasticity and modulus of Andrea Engineering Tests Laboratory, AETL (2012,
subgrade reaction estimated from consolidation 2015). Geotechnical investigation reports. Report Nr.
test, are smoothly increases with the increase of 1734, May 2012 & report Nr. 1963, August 2015,
applied vertical pressure. Baghdad, Iraq.
3. The plate load test results shows high values Bowles, J.E. (1997). Foundation Analysis and Design. 5th
of the elasticity and subgrade reaction at the edition, Mc Graw-Hill Book Co., New York.
beginning of the loading stage, after that sharp Coduto, Donald P. (2001). Foundation Design Principles
reduction in these values were occurred until and Practices. 2nd Edition Book. ISBN 0-13-178461-7.
USA.
the applied pressure passed the value of pre- Meyerhof, G.G. (1963). Some Recent Research on the
consolidation pressure then increases. Bearing Capacity of Foundations. CGJ, vol. 1, no. 1,
4. At first stage of loading (i.e. before the pre- Sept, pp. 16–26.
consolidation pressure value), the elasticity and Terzaghi, K., Peck, R. & Mesri, G. (1996). Soil Mechan-
modulus of subgrade reaction estimated from ics in Engineering Practice. ISBN 978-0-471-08658-1,
consolidation test exhibits higher values than 3rd edition, John Wiley & Sons, Inc., USA.
from field plate load test.
5. The boundary conditions of the test and pres-
sure distribution within the soil mass have a

772
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Variability of Light Weight Deflectometer during laboratory and field

applications for subgrade soils under variable moisture conditions

Mehran Mazari
Department of Civil Engineering, California State University, Los Angeles, CA, USA

Gerardo Garcia, Imad Abdallah, Jose Garibay & Soheil Nazarian

Center for Transportation Infrastructure Systems (CTIS), The University of Texas at El Paso, TX, USA

ABSTRACT: The stiffness-based nondestructive devices have become popular for quality control pur-
poses during construction of earthwork and unbound aggregate materials. Light Weight Deflectometers
(LWDs) have been gaining popularity as a tool for quality control. The variability and uncertainty associ-
ated with their results should be evaluated under various field conditions and for different sources of geoma-
terials. Such variability could be either device-related, material-related or operator-related. This paper aims
at studying the variability of two LWD devices. A number of laboratory and field studies were performed to
estimate the variability of the LWD with respect to soil properties and moisture conditions. The sources of
variability were then identified and correlated to moisture conditions. The differences in moisture contents
at the time of compaction and quality control testing seems to affect the uncertainty of LWD test results.
Furthermore, the operational mechanism of the device measurements impacts the estimated moduli.

1 INTRODUCTION the sources of LWD device variability could also

help in understanding its results during the quality
Performance of pavement structures mostly control process. This paper aims at investigation of
depend on the structural behavior of unbound the LWD variability under various moisture con-
granular layers and most of the pavement deterio- ditions from laboratory and field evaluations of a
rations could be traced to the insufficient strength number of subgrade soils.
of these materials. The current state of practice
for Quality Control and Quality Assurance (QC/
QA) during the construction process is the estima- 2 BACKGROUND
tion of density and sometimes the moisture con-
tent using a Nuclear Density Gauge (NDG). On A number of studies have been dedicated to evalu-
the other hand, the main parameter used for the ating the application of LWD and its variability
mechanistic-empirical design of flexible pavements for quality control purposes. Fleming et al. (2004)
is modulus. Consequently, there is a missing link reported the variability of LWD device in terms of
between the design and QC/QA processes which Coefficient Of Variation (COV) of field measure-
reflects the need for a modulus-based construction ments of fine-grained subgrades in the range of 25%
quality control approach. to 60%.
The Light Weight Deflectometer (LWD) has Alshibli et al. (2005) performed an extensive lab-
been gaining popularity as a modulus-based device oratory evaluation of LWD and other in-situ tests
for quality management purposes. LWD is the to estimate the stiffness of different compacted geo-
portable version of the Falling Weight Deflectom- materials. They reported wide scatter and poor reli-
eter (FWD) which is well-known among the pave- ability of LWD deflections on weak subgrade soils.
ment professionals. However, the uncertainties in Nazzal et al. (2007) studied the repeatability of
the application of LWD for quality control process LWD by estimating the COV of five measurements
may affect the interpretation of the results. Such taken on various test sections. They reported that
variability could be device- or operator-related, COVs of LWD moduli ranging from 2% to 28%.
and furthermore, can be associated with the spa- They also noted an inverse correlation between the
tial variation of geomaterials. Some index proper- variability of the measurements and LWD moduli.
ties such as moisture content could be the source Hossain and Apeagyei (2010) reported spatial vari-
of non uniformity in compacted layers. Studying ations of 22% to 77% for in-situ LWD moduli.

773
 Stamp and Mooney (2013) studied the effect of 150-mm-thick geomaterial was compacted on top
LWD characteristics on the measured moduli with of the subgrade layer to study the behavior of the
two types of LWD devices on a number of test sec- composite unbound aggregate structure under the
tions. They concluded that the configuration of LWD tests.
LWD significantly impacted the reported moduli. The appropriate size of the specimen to mini-
Volovski et al. (2014) developed statistical limits mize the impact of the boundaries was determined
for the use of LWD during construction quality through Finite-Element (FE) modeling (Amiri
assurance in Indiana. They suggested to increase et al., 2009). The FE model comprised of the
the number of acceptance tests from three to bounding tank, loading plate and pavement geo-
seven, based on a number of analyses of variance materials. The elements of the tank body and
(ANOVA) of field test results. geomaterial in the FE model were assumed to be
Even though a number of investigations have bilinear, axisymmetric and continuum elements.
been undertaken to evaluate the variability of Due to the small-scale nature of the test specimens,
LWD devices, there is still a need to extend those there was no need to place a mesh to avoid the mix-
studies for a wider range of geomaterials as well as ing of the geomaterial layers.
considering index properties of target geomaterials. To study the impact of moisture variation on the
This paper briefly investigates the impact of mois- variability of the LWD tests, three sets of specimens
ture condition on the variability of LWD during a were prepared. The geomaterials for the top layer
limited number of laboratory and field evaluations. were placed and compacted at 20% wet of their cor-
responding OMCs for the first set of specimens and
at their corresponding OMCs for the second set.
3 METHODOLOGY The LWD tests were performed upon completion
of compaction process and were repeated while the
Four types of common subgrade soils were used specimens were drying. Another set of specimens
in this study. The basic characteristics of these was compacted and tested at 20% dry of OMC.
soils and their classification according to the uni-
fied soil classification system are summarized in
Table 1. The Optimum Moisture Content (OMC) 4 RESULTS AND DISCUSSION
and Maximum Dry Unit Weight (MDUW) for
each soil were obtained as per standard laboratory Two sets of ASTM standards are available for con-
Proctor tests. The fifth soil type in Table 1 repre- ducting LWD tests: E2583 (applicable to Dynat-
sents the geomaterials used during field evaluation. est LWDs) and E2835 (pertinent to Zorn LWDs).
A number of 0.9-m-diameter cylindrical labo- Figure 1 illustrates the two LWDs during the labo-
ratory specimens were prepared under controlled ratory evaluation. Both standards require three
moisture and density conditions. These specimens seating-drops followed by three measurement
contained a 75-mm-thick layer of pea gravel at the drops to estimate the modulus of the compacted
bottom to facilitate the saturation of specimens for layer. The variability of LWD could be related to
other aspects of study that are beyond the scope of the variability of device readings among the meas-
this paper. A 400 mm-thick layer of common sub- urement drops and the spatial variation of test
grade soil was placed and compacted in 50-mm- points along the test section. The LWD tests were
thick lifts using a hand compactor. An additional repeated at three different locations on top of the
compacted geomaterial layer to investigate the spa-
Table 1. Index properties of soils.
tial variability of the results.
The process proposed by Burati et al. (2003) was
Moisture density followed to evaluate the variability of the devices.
Atterberg limits summary
USCS1
Soil MDUW6,
Classification PI2 LL3 PL4 OMC5, % kg/m3

CH 53 86 33 25.9 1533
CL 14 27 13 10.0 2002
ML Non Plastic 9.4 1996
SC 12 23 11 11.4 1945
CLa 27 41 14 16.7 1714
1
Unified Soil Classification System, 2Plasticity Index,
3
Liquid Limit, 4Plastic Limit, 5Optimum Moisture Con- Figure 1. Evaluation of laboratory specimens using
tent, 6Maximum Dry Unit Weight, a field evaluation. LWD.

774
This process defines each triplicate measurement performed on top of the compacted geomaterials
as a “lot.” The variance of test results at each test at three different locations. Figure 4 illustrates the
moisture content was calculated using the follow- COVs of the Zorn LWD measurements at differ-
ing equation: ent moisture contents. Some of the test results for
the CH geomaterial were not available due to the

V=
∑ [( n 1) SD
S ]
(1)
excessive cracking of the specimen surface dur-
ing the drying cycles (see Figure 4a). Since lim-
p m ited number of soil types were used in this study,
a global correlation could not be drawn between
where SD = standard deviation of each lot, the moisture content and COV of the device meas-
p = total number of measurements, n = number of urements. However, isolating the results of each
test replicates, and m = number of lots. geomaterial would provide some insight about the
Figures 2 and 3 illustrate the variability of LWD impact of the moisture content on the reliability of
measurements for the soil types used in this study. the device measurements. As an example, the CL
A rigorous and comprehensive statistical analy- soil placed at OMC seems to show higher variabil-
sis of LWD data in quality control process is pre- ity during the drying cycle when compared to the
sented in Mazari et al. (2013). The uncertainties in specimens compacted at the wet of OMC.
the measurements with the Dynatest LWD seem to Figure 5 summarizes the results of the same
be more material dependent compared to the Zorn process for the Dynatest LWD. Similar trends to
LWD. The variability of the Dynatest device on top the Zorn LWD can be observed. The COVs of the
of the high-plasticity clay seems to be greater com- Dynatest LWD seem to be slightly higher than the
pared to silt, sand and low-plasticity clay. However, Zorn LWD. This could be due to the functional dif-
the variability of measurements using Zorn device ferences between the two devices. Another plausi-
seems to be almost geomaterial independent. At ble conclusion could be that the geomaterials with
80% confidence limit, the COVs of the Dynatest higher OMC (i.e. CH soil in this experiment) yielded
LWD measurements were about 9%, 12%, 14% and higher variability inthe LWD measurements. Again,
28% for CH, ML, SC and CL soils, respectively. At these observations are based on limited number of
the same confidence level, the COVs for the Zorn soil types investigated under the laboratory condi-
LWD measurements were about 20%, 16%, 15% tions. This process should be extended to a wider
and 17% for CH, CL, ML and SC soils, respectively. range of geomaterials to explore the correlations
As mentioned earlier, the top geomaterial layer between the moisture content, or other index prop-
was placed at wet of OMC and OMC followed by erties, of compacted soil layer and the variability
a drying period to study the impact of moisture
content variation on the uncertainty of the test 3~ ~~----~~--~---, 3~ ~------~--------,
a) Compacted at Wet of OMC b) Compacted at OMC
results. During these drying cycles, LWD tests were
"" ""
,.. 2~

§15% X

·~ 40% ~ CH
'" "!~
~ L---------------==~
~
---cL

l 20%

o% +-~~~~-------.----.-----.----.----~-~K~-~·c~
····ll··· Ml
10 20
Testing Moisture content,%
30 10 20
Testing Moisture content,%
30

S% 10% 20% 30% 40% SO% More

Coefficient of Variation, % Figure 4. Variability of Zorn LWD measurements at
different moisture conditions.
Figure 2. Variability of Zorn LWD measurements for
different soil types. 3~ ,---- - - - - - - - - - - - - - - - - - - ,
3~ ~~--~~~--~~­
a) Compacte d at wet of OMC
•• b) Compacted
atOMC
25%
""
,.. 2~

§1s% X

·~
•a
'" IJ.ML 1
xsc
~ L-------------~===
S% 10% 20% 30% 40% 50% More 10 zo 30 10 zo 30
Coefficient of Variation, % Testing Moisture Content, % Testing Moisture Content, %

Figure 3. Variability of Dynatest LWD measurements Figure 5. Variability of Dynatest LWD measurements
for different soil types. at different moisture conditions.

775
of the LWD measurements. Such uncertainties .... ~~~~~~~,-----------------------~--,
A Compacted at Wet of OMC:
0 Compacted at OMC
I
could be operator or device related as extensively e Compacted at Dry of OMC

discussed in Mazari et al. (2013), Nazarian et al. •• oD

(2014) and Tirado et al. (2016). ••

Figure 6 summarizes the variability of the LWD
measurements on the specimens compacted at the
10%

"" ~----------------------------------~~----~
•
10
..
12 14
0

16 18
dry of OMC. The two devices yield results that are Oven Moisture Content, %
not significantly different except for the ML soil.
Overall, the ranges of the LWD variability on the Figure 8. Variation of moisture content during field
specimens compacted at the dry of OMC were lower evaluation.
than those compacted at OMC or wet of OMC.
To further evaluate the variability of LWD with
respect to different moisture conditions, a field the robustness of modulus-based devices, such as
test section was selected during the placement of a LWD, during the quality control process.
clayey subgrade geomaterials in south-central Texas. Only a Zorn LWD was available for this part of
To study the moisture variations, three test strips study. The COVs of the LWD measurements with
were placed and compacted at wet of OMC, OMC respect to moisture content along the test strip for
and dry of OMC. Soil samples were extracted after each of the three sections (wet of OMC, OMC and
the compaction of the geomaterials to estimate the dry of OMC) are summarized in Figure 8. Despite
oven-dry moisture contents in the laboratory. some scatter in the collected field data, the COV
Spatial variability of the compacted geomaterials of the LWD measurements along the wet section
is inevitable due to the construction conditions and reached as high as about 37% while the COVs on the
the inherent variability of the material source. Fur- optimum and dry sections did not exceed 26%. Fur-
thermore, the distribution of the moisture content thermore, the COVs for the dry and near-optimum
along the compacted section is also uncertain due moisture conditions show less scatter as compared
to the uneven distribution of the moisture through- to the results from the wet section. However, these
out the section. Such variations were observed dur- observations are based on limited materials and
ing the field activities in this study as summarized they could not be extended to other soil types and
in Figure 7. Despite the effort to prepare the geo- moisture conditions. The other sources of variability
materials at OMC, wet of OMC and dry of OMC, in the geomaterials under various field conditions
the actual moisture contents of the compacted geo- should be studied to investigate the impact of mois-
materials were different than the target values. Such ture variation on the uncertainty of the LWD results.
spatial variabilities of soil moisture contents affect

5 CONCLUSIONS
"'" a) Zorn LWO .,, ''"' b) Dynatest LWD .,,
2"' AML
"'' AMl A number of laboratory specimens were prepared
xsc xsc
;;<
2'"'
e cH ;;<
2'"' from four common subgrade geomaterials to study
e cH
8~ 15% '----- 8~15% '----- the variability of the LWD measurements with
respect to material type and moisture content. Two
1'"'
• 1'"'
sets of specimens were prepared at wet of OMC
5% Ax • 5% ""
A
and OMC conditions. Measurements were under-
"" 10 20 30
0%
10 20 30 taken periodically while the specimens dried with
Testing Moisture Content,% Testing Mofsture Content,%
time. The spatial variability of the measurements
was obtained by repeating LWD tests at three
Figure 6. Variability of LWD measurements on geoma- different locations on top of each specimen. The
terials compacted at dry of OMC.
COVs of the Dynatest LWD measurements seem
to be more material dependent compared to the
20
-
0AttualfleldMoi$WreContent
Tarcet Field Moslture Content
Zorn device. Since the Dynatest LWD reports the
imparted load as well as the deformation of the
soil surface, they are more in line with the assump-
tions made for the data reduction using the Bussi-

~
Ory ofOM C
Section
WetofOMC
Section
OMCSectlon
nesq theory whereas the Zorn LWD measures the
deflection of the LWD plate.
The high-plasticity clay showed higher variabil-
ity compared to the low-plasticity clay. The drying
Figure 7. Variation of moisture content during field cycle under the OMC conditions seem to be associ-
evaluation. ated with higher variability compared to the con-

776
ditions where the geomaterials are compacted at research record: J of the Transportation Research
the wet of OMC. Limited extension of this study Board., Vol. 2004(9): 80–87.
to the field conditions showed that the LWD meas- Hossain, M. S., and Apeagyei, A. K. (2010). “Evaluation
urements on a low plasticity clayey soil, compacted of the Lightweight deflectometer for in-situ determi-
nation of pavement layer moduli.” Federal Highway
at the wet of OMC, tend to exhibit slightly higher Administration (FHWA), Report No. FHWA/VTRC
variability compared to other moisture conditions. 10-R6, Charlottesville, VA.
However, extending the field evaluations to differ- Mazari, M., Garcia, G., Garibay, J., Abdallah, I., and
ent types of soils at various moisture contents are Nazarian, S. (2013). “Impact of Modulus Based
desirable for further estimation of device variability. Device Variability on Quality Control of Compacted
Geomaterials Using Measurement System Analysis.”
Transportation Research Board, 92nd Annual Meet-
ACKNOWLEDGMENTS ing, No. 13-3192, Washington, D.C.
Nazarian, S., Mazari, M., Abdallah, I. N., Puppala,
A. J., Mohammad, L. N. and Abu-Farsakh, M.
This study was carried out as part of the NCHRP (2014). “Modulus-based construction specification
Project 10–84. The contents of this paper reflect for compaction of earthwork and unbound aggre-
the authors’ opinions, not necessarily the policies gate.” National Cooperative Highway Research Pro-
and findings of NCHRP. The authors are grateful gram (NCHRP) Project No. 10–84, Transportation
to Dr. Ed Harrigan and the study panel for their Research Board, Washington, D.C.
help and advice throughout this study. Nazzal, M., Abu-Farsakh, M., Alshibli, K., and Moham-
mad, L. (2007). “Evaluating the light falling weight
deflectometer device for in situ measurement of
REFERENCES elastic modulus of pavement layers.” Transportation
Research Record: J. of the Transportation Research
Board, Vol: 2016(2): 13–22.
Alshibli, K. A., Abu-Farsakh, M., and Seyman, E.
Stamp, D. H., and Mooney, M. A. (2013). “Influence of
(2005). “Laboratory evaluation of the Geogauge and
lightweight deflectometer characteristics on deflec-
light falling weight deflectometer as construction con-
tion measurement.” Geotechnical Testing Journal,
trol tools.” J Mater Civil Eng., Vol. 17(5): 560–569.
Vol. 36(2): 216–226.
Amiri, H., Nazarian, S., and Fernando, E. (2009). “Inves-
Tirado, C., Rios, K.Y., Mazari, M., Nazarian, S. (2016).
tigation of impact of moisture variation on response
“Simulation of Light Weight Deflectometer Measure-
of pavements through small-scale models.” J Mater
ments Considering Nonlinear Behavior of Geomate-
Civil Eng., Vol. 21(10): 553–560.
rials.” Transportation Research Board 96th Annual
Burati, J. L., Weed, R. M., Hughes, C. S., and Hill, H. S.
Meeting, No. 17-03576, Washington, D.C.
(2003). “Optimal procedures for quality assurance spec-
Volovski, M., Arman, M., and Labi, S. (2014). “Develop-
ifications.” Federal Highway Administration (FHWA),
ing statistical limits for using the Light Weight Deflec-
Report No. FHWA-RD-02–095, Washington, D.C.
tometer (LWD) in construction quality assurance.”
Fleming, P., Frost, M., and Lambert, J. (2004). “Review
Joint Transportation Research Program, No. FHWA/
of lightweight deflectometer for routine in situ assess-
IN/JTRP-2014/10. West Lafayette, IN.
ment of pavement material stiffness.” Transportation

777
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Comparing CBR values obtained from field with laboratory test results

S. Karahancer, E. Eriskin, O. Sarioglu, M. Saltan & S. Terzi

Suleyman Demirel University, Isparta, Turkey

ABSTRACT: California Bearing Ratio (CBR) values are not always easily accessible, especially when
the highway is already constructed and therefore, CBR values obtained from laboratory test results may
need to be utilized. However, the laboratory test results do not provide the same results as the field tests
due to external factors. To utilize the CBR values from laboratory tests, they need to be corrected with a
coefficient to correspond to the in situ values. For the current investigation, Dynamic Cone Penetrom-
eter (DCP) measurement from the field have been collected and samples have been compacted with the
same field material and the same in situ density and laboratory CBR tests have been carried out. A trend
analysis was performed between the results from field and laboratory test values and a resulting correction
coefficient is calculated.

1 INTRODUCTION In the current study, we perform DCP tests

on three random points of a subgrade. Then the
All the pavement structures are constructed over physical conditions (water content, specific grav-
unbound granular layers. So, the unbound layer ity, sieve analysis, density) of these test points are
should provide the needed strength for safe flow- determined. Afterwards, the base materials are re-
ing traffic (Yohannes et al. 2014). Before construc- compacted in the laboratory to achieve the same
tion of the pavement layers, the strength and the physical conditions as the in situ material. Then the
stability of the unbound base, subbase or sub- CBR tests are performed on these compacted base
grade should be determined and reinforced when materials. CBR values obtained from DCP tests
needed. Physical properties such as density, par- (in situ) and CBR tests (in the laboratory) are then
ticle size and thickness are used to determine the compared.
strength and stability (Chen et al. 2005, Yohannes
et al. 2014, Fleming et al. 2015). However, proce-
dures for determining the physical properties are
2 MATERIALS AND METHOD
time-consuming. So, determining the strength
value of the unbound layer in situ would save a
2.1 Subgrade
lot of time.
Loadbearing capacity is very important strength DCP tests are performed on a subgrade which is
parameter of the base layers. An often test for the compacted naturally. A sieve analysis of the sub-
load-bearing capacity is the California Bearing grade is shown in Figure 1.
Ratio (CBR) test which is usually performed in the The soil classification, water content and Atter-
laboratory. One of the biggest handicaps of this test berg Limits are shown in Table 1.
it is difficult to take a sample to the laboratory with
the same conditions as in the field. Therefore, meas- 100
uring the loadbearing capacity of the base in situ is
necessary. 80
The load bearing capacity can be calculated
from the resistance of field penetration value '*oii 60
c
·;;;
which is obtained from the Dynamic Cone Pen- ~ 40
etrometer (DCP). There are several advantages for a_

using DCP such as simple transportation of the 20

device, its ease of use, and it’scost-effective. A cor-
0
relation coefficient, however, is needed to use DCP 0.01 0.1 1 10 100
test results to obtain the CBR value of the base Sieve Size, mm
layers. Multiple researchers have performed stud-
ies concerning the correlation coefficient. Figure 1. Sieve analysis of the subgrade.

779
Table 1. Physical properties of the subgrade.

Soil Classification CL-ML 8 kgwetglllt

Water Content,% 13.22

Saturated Density, g/cm3 1.68
Liquid Limit,% 24.8
Plastic Limit,% 18.9
Plasticity Index %5.9

193Smm 575mm
2.2 Dynamic Cone Penetrometer (DCP) test
The DCP is a fast, simple and economical test
method for determining the strength of the sub-
grade. In addition, it can be used for quality control
of the constructed subgrade or pavement. It is user-
friendly and can be used rapidly for multiple tests.
It consists of two parts; (i) a weight of 8 kg with a
575 mm sliding road, and (ii) a cone either 300 or 600 Rule
which is installed on a tube with 20 mm in diameter. A
schematic plan is shown in Figure 2. Initially, the DCP
is placed on the ground to be tested. Then, the weight
is lifted up and released to fall down. From the impact
of each fall, the lower tube with the cone sinks into Con.e
the tested material. The penetration depth after each
fall is then measured and recorded. Recorded penetra- Figure 2. A schematic plan of DCP test equipment.
tion depths are used to draw the penetration depth vs
blow count graph. This graph is utilized later to deter-
mine the penetration resistance of the testedmaterial. machine should be 1.25 mm/min and it should
Therefore, the gradients of each line occurred on have a minimum of 5000 kg load capacity.
the graph according to blow counts and penetration During the test procedure, differentpenetra-
depths are calculated. When the gradient is small, the tiondepths (0.625, 1.25, 1.825, 2.5, 5, 7.5, 10 and
ground has more resistance and could be called “stiff” 12 mm) readings should be recorded. The results
and when the gradient is larger, the ground has less of the readings are used for drawing the graph to
resistance and could be called “soft”. determine if there are any problem with the results.
The penetration depth is measured after each Usually, CBR values are calculated from 2.5 and
weight impact. However, when the measured depth 5 mm readings. The calculated CBR is the ratio of
is smaller than 20 mm, there are several recom- penetration of standard piston to the penetration
mendations as follow; (i) one reading-off at every of a standard material.
two impacts, with penetration from 10 to 20 mm, CBR test results vary according to factors such
(ii) one reading-off at every five impacts, with pen- as density, water content and tested soil type.
etration from 5 to 9 mm, (iii) one reading-off at Therefore, it is very important to this study to set
every ten impacts, with penetration from 2 to 4 mm the test materials to the correct conditions.
(iv) penetration depth smaller than one mm and
exceeds twenty impacts, in which case the test is not
considered applicable (Tosovic & Vujanic 2010). 3 RESEARCH FINDINGS

In the current study, a subgrade is selected and DCP

2.3 California Bearing Ratio (CBR) test
tests are performed on three random points of the
The CBR test is a penetration test for subgrade, subgrade (Figure 3). The penetration depth of each
subbase and base layer which shows the strength weight drop is known as the Penetration Index (PI)
of the layer. The CBR value is generally used as a and expressed in mm/blow. The PI of the material
parameter in the design of roads. can be used for estimating shear strength, modulus
Specimens with a diameter of 150 mm of and loadbearing using empirical relationships. The
unbound base material are compacted to perform PI is determined using blow numbers and penetra-
the CBR test. A loading machine is needed with a tion depths. However, before calculating the PI, ini-
penetration piston 50 mm in diameter and a length tial seating drops should be performed. Generally,
of at minimum 100 mm. The speed of the loading three seating drops and five to ten (according to

780
 r··
procedures. According to the calculations, 23 blows
10 15 20
for each layer are made to achieve target density.
E -100 •.•
E 0 ·a Afterwards, the compaction is controlled by meas-
~ -200 5 ··~ Iy = -35.122x ±35.991 1 uring the density of the compacted specimen.
QJ

~ -300
•••
Only one compaction is performed for the CBR
...
..•
test because of the similarity of the DCP test
:;:;
g -400
0 ••••· •.. results. Compacted specimens have loaded with
:, 1y = -26.964x + 11.607 1 •·.... CBR test equipment at a speed of 1.25 mm/min.
......
OJ
~ -500 .
"-
Load reading for specific penetration depths is
-600
Blow Count
shown in Table 2.
When the CBR readings were done, recorded
(i) values were divided by the piston area which is
19.53 cm2. To obtain the CBR value of the tested
material, depth readings of 2.5 and 5 mm should be
E -1000 b······-...•
~ 5
10 15 20
taken into consideration. Calculated pressure val-
£
c. -200 ues should be divided into 70.31 and 105.46 kg/cm2
QJ
·····e.
.•.. I y = -34.58x + 35.175 1
for 2.5 and 5 mm readings, respectively.

..
0
c -300
There are some correlations to the computation
0
:;:;
g -400 •··.•·· ..... of CBR values from DCP test results, depending
QJ

~ -500
:1y = -25.964x + 11.036 1
.... on the type of material and the cone. Most of the
"-
-600
Blow Count
·• correlations are logarithmic with the general form
of Equation 1 (Tosovic & Vujanic 2010).

(ii) log10 ( ) l g10 ( PI )

log (1)
0 fT.e where CBR is the laboratory CBR value, K is con-
~ i ······! 10 15 20

£
c. -200
-100
.....
·~ Iy = -34.545x + 28.106 1
stant, C is the coefficient and PI is the penetration
index (mm/blow). Equation 1 is used to generate
~ •..•..
~
c -300
0
...•·... Table 2. CBR test results.
QJ

~ -500
"-
-400
Iy = -28.25x + 14.321 1 ......... Reading
depth, mm
Upside
load reading, kg
Downside
load reading, kg
-600
Blow Count Section i 0.625 13.86 18.9
1.25 27.72 30.24
(iii) 1.825 34.02 46.62
2.5 45.36 73.08
Figure 3. DCP test results of the subgrade’s three ran- 5 105.84 182.7
dom points. 7.5 167.58 252
10 236.88 304.92
12 291.06 337.68
the material type) measurement drops are recom-
Section ii 0.625 15.12 20.16
mended (Siekmeier et al. 2009).
1.25 28.98 32.76
As shown in Figure 3, the first five drops’ PI val-
1.825 32.76 47.88
ues are different and can be defined as the seating
2.5 42.84 76.86
drops. Therefore, the first five drops are neglected
5 103.32 176.4
in the study. The remaining drop values are used
7.5 161.28 248.22
to calculate the PI value of the subgrade material. 10 229.32 301.14
Calculated PI values are obtained as 34.58, 35.122 12 283.5 332.64
and 34.545.
Section iii 0.625 12.60 16.38
After performing DCP tests on the selected sub-
1.25 25.20 26.46
grade, material from deeper than 200 mm was dug,
1.825 28.98 45.36
extracted and taken to the laboratory fo testing.
2.5 42.84 70.56
The material was re-compacted in the laboratory
5 108.36 178.92
according to conditions shown in Table 1. The com- 7.5 171.36 245.70
paction was done in a CBR mould with a diameter 10 241.92 293.58
of 150 mm and height of 120 mm. A Proctor ham- 12 298.62 328.86
mer, in three layers, was used for the compaction

781
the relationship between DCP and CBR values. PI
∑ (Y Y ′)
2
and CBR test results are shown in Table 3.
σ est = (3)
Using the general form of the correlation, n
Equation 1, the coefficient and constant values
are calculated for a 600 cone and subgrade mate- where σest is the SEE value, Y is the measured CBR
rial with the physical properties shown in Table 1, value, Y’ is the calculated CBR value and n is the
the correlation equipment has been obtained number of pairs of CBR values. As long as the
(Equation 2). smaller the σest is, the more accurate are the predic-
tions made (Lane 2017). The result calculated from
log10 (CBR ) 2.56
56 1. log10 ( PI ) (2) Equation 3 is 0.22 which is relatively small.

The determined Equation 2 doesn’t seem to sig-

nificantly vary from previously developed relation- 4 CONCLUSIONS
ships. There are only slightly different coefficients.
So, our study is in accordance with the previous lit- This paper tries to correlate DCP test results
erature. The difference in the coefficients is thought obtained from the field with the CBR results from
because of the difference of the subgrade material. the laboratory. Therefore, DCP tests have been
Therefore, the effect of the subgrade materials’ performed on a subgrade material. Afterwards,
physical properties and the gradation on the coef- the physical properties have been obtained in the
ficients could be studied for the future works. laboratory and CBR tests have been performed.
Correlation between CBR values obtained from As a result, by the help of the logarithmic equa-
laboratory CBR test equipment and CBR values tion given in this paper, PI obtained from DCP test
calculated from PI values with the help of Equa- results represents CBR results with a high correla-
tion 2 is shown in Figure 4. tion as 88.31%.
As seen in Figure 4, the correlation coefficient The equation obtained from this test result
is relatively high. should be tried for 300 cone and for further DCP
In addition to the R2 value of the data, standard test results on different base materials as further
error of the estimate (SEE) value is calculated as work before it can be used precisely.
well to demonstrate the accuracy of the developed
Equation 2. To calculating the SEE value, Equa-
tion 3 has been used. REFERENCES

Chen, D.H., Lin, D.F., Liau, P.H. & Bilyeu, J.A. 2005.
Table 3. PI and CBR results. Correlation between dynamic cone penetrometer val-
ues and pavement layer moduli. Geotechnical Testing
Section PI, mm/blow CBR,% Journal, 28(1): 42–49.
Fleming, P., Frost, M. & Lambert, J. 2015. Review of
i 35.122 8.56 lightweight deflectometer for routine in situ assess-
ii 34.58 8.47 ment of pavement material stiffness. Journal of the
iii 34.545 8.41 Transportation Research Board.
Lane, D.M. 2017. Standard Error of the Estimate.
Last accessed: 09.02.2017. Available online at http://
onlinestatbook.com/lms/regression/accuracy.html.
9.05 Siekmeier, J., Pinta, C., Merth, S., Jensen, J., Davich, P.,

~
Camargo, F. & Beyer, M. 2009. Using the Dynamic
*"'
2
Q)

"'>
c::
9

8.95 ···..... ···......•···............ Cone Penetrometer and Light Weight Deflectom-

eter for Construction Quality Assurance, Minnesota
Department of Transportation, MN/RC 2009–12,
E
0
.;::
"':l
Q)

'iii
>
8.9

8.85

8.8
+
.........
R' =o.8~iii··· . Minnesota, 244p.
Tosovic, S. & Vujanic, V. 2010. C.B.R. Testing with
Dynamic conical penetrometer in the process of road
rehabilitation and construction control. 10. Sloven-
skiKongres O CestahInPrometuPortorož, October
"'"'u 8.4 8.45 8.5 8.55 8.6
20–22.
CBR values from test equipment,% Yohannes, B., Tan, D., Khazanovich, L. & Hill, K.M.
2014. Mechanistic modelling of tests of unbound
granular materials. International Journal of Pavement
Figure 4. Comparison of the CBR values. Engineering, 15(7): 584–598.

782
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Contribution to lateritic soils calibration using the dynamic

penetrometer PANDA

Y. Gansonré, C. Bacconnet & P. Breul

Institut Pascal, Université Blaise Pascal, Clermont Ferrand, France

M. Benz, P. Moustan & R. Gourvès

Sol Solution, Riom, France

S. De-Maistre
VECTRA, Paris, France

ABSTRACT: The degradation of roads network in tropical countries is a major concerned nowadays.
Many analysis link this problem to the specific environment of roads in their areas, but also to the limited
knowledge on the real behavior of lateritic soils, mainly used for these roads construction and mainte-
nance. To understand more this problematic in order to develop an appropriate and simple methodology
of road control, we led a study to characterize laterites from Burkina Faso in a laboratory by using the
penetrometer PANDA. This article presents the process of calibration of the lateritic soils and the differ-
ent results obtained through this investigation.

1 INTRODUCTION 1989). PANDA is a light dynamic penetrometer

developed by Sol Solution (Gourvès, 1992).
The construction of roads is a necessity for each After presenting generalities on lateritic soils,
country and an obligation for developing countries the article highlights the process of calibration by
to improve their transportation systems. Many using this dynamic penetrometer to evaluate dry
economists and politics show the benefits of road density in a compaction control process through
management for the economy and the develop- the measurement of the cone resistance.
ment. The United Nations (2015) has identified the
resilient infrastructures as a support of economy
development and human well-being. However,
2 LATERITIC SOILS
in a large part of Africa, in tropical and equato-
rial countries, it is possible to note the low quality
2.1 Definition and origin of laterites
level of the road networks. The degradation rate
of these roads is about 60%. As it is the case, any- In Burkina Faso and in large part of tropical coun-
where in the world, the construction of resilient tries, the economical and available materials are
networks in these countries involves taking into “laterites”. It can be found in Africa, Asia, Oceania
account different parameters. Some of them are and America and presents 12% of emerged soils
related to the specific environment (climate, traf- (Bagarre, 1990). The term “laterite” which means
fic, etc.) of road and the process of construction brick, was suggested for the first time by a Scottish
and maintenance; others are linked to the knowl- scientist named Buchanan (1807) to designate a
edge of the characteristics and the behavior of the red material for construction during his mission in
materials. In tropical countries, the economical India. It is important to note that this is a material
and available material is “lateritic gravel”, result- resulting from physical and chemical weathering in
ing from the alteration of the parent rock in tropi- tropical climates of the parent rock. It is composed
cal climates Buchanan (1807). Otherwise, the use of alumina and iron oxide (Al2O3 + Fe2O3) in
of this material in the construction requires certain addition to silica. From engineering point of view:
performances, especially mechanics. To control this “it is a vacuolar structural material, with moder-
performance, there is a large range of technologies. ated and varying colors from the yellow to the red
PANDA is one of them, widely used for materials more or less dark and even black, constituted of
and roads construction specific control (Gourvès, more or less continuous crust, variable thickness

783
and hardness, showing many vacuums and slaga Table 2. Classification of the samples.
spects, or also oolitic and pisolitic isolated con-
cretions with more or less great resistance and an Classes’ Numbers
argillaceous fraction” Autret (1983).
B2 3
GTR (France) B4 219
2.2 Variability and classification [IFSTTAR, 1992] B5 1
B6 438
The most important for these materials is their use A1-a 11
in civil engineering and then their geotechnical AASHTO (USA) A1-b 9.0
characteristics. As much of materials, the identifi- [HRB, 1945] A2–4 88
cation of lateritic soils is done by the knowledge A2–6 414
of the basic parameters as the sieve analysis, A2–7 139
Atterberg limits, etc. Based on a prospection led SLA 299
in Burkina Faso (LNBTP, 2005) about research of Brazilian method SLP 359
road-making materials for tracks maintenance, the [Rodriguez, 2010] SLF 3
statistical analysis of a sample of 661 surveys gives
the results below Table 1.
This analysis allowed to sum up the following Histogram (80 11m)
characteristics for the two samples of laterites 0 .25
studied.
0.2
– the percentage of fines is less than 35% with an
average around 15.1%
– the passing to 2 mm is between 10.5% and 89% .,::>~ 0.15
with an average around 29.8% .,.
~ 0.1
– the plasticity index in the range of 4% and 40%
with an average around 17% 0.05
– the dry density (MPO) is in the range of 18.9 and
29.9 kN / m3 with an average around 22 kN /m3 0
– The soaked CBR at 95% of MPO is between 11 10 15 20 25 30 35

and 110 with an average around 48. S0 11m

This study shows that the characteristics of lat-

eritic soils are different. But the classification of
Histogram (PI)
these materials through three international stand-
ards: GTR, AASHTO and Brazilian method high- 0.3
lights less groups for each classification standard -
0.25
(Table 2). ;--
According to the GTR classification, the later- 0.2
ites of Burkina Faso are classified in B6 and a few ~
c ;--
of them in B4 group. That is in conformity with ~ 0.15
.,. -
the literature which shows that the lateritic materi- .t
als are generally classified B6 (Millogo 2008, Bohi
0.1
-
2008 & Ndaye 2013). It is also possible to note that 0.05

Table 1. Basic statistics of the sample characteristics. 0 J

10 20
r--1----.
30 40 50

Parameters MinMax Aver Var(n−1) SD(n−1) PI

Passing 80 μm% 3.0 32.2 15.1 29.7 5.5

Passing 2 mm% 10.5 89.0 29.8 109.2 10.5 Figure 1. Variability of passing to 80 μm and plasticity
Liquidity limit% 7.0 69.5 33.6 100.4 10.0 index for B6.
Plasticity index% 4.0 40.0 17.0 42.4 6.5
w (MPO)*% 4.7 12.0 7.4 1.0 1.0
there is a little correlation between these classes
and the geo-climatic mapping of the country. The
γd (MPO)*kN/m 318.9 29.9 22.0 1.5 1.2
CBRi (95% MPO) 11.0 110.0 47.6 252.8 15.9
B4 material is mainly present in the South-west
CBRi (98% MPO) 17.0 160.0 75.1 671.9 25.9
which is the rainiest area and close to the sedimen-
tary rocks of the country. The distribution of the
* Water content for modified Proctor optimum. passing to 80 μm and the plasticity index for the
* Dry density for modified Proctor optimum. B6 material is respectively shown in the Figure 1.

784
 Compared to the B6 loans: the identification of lateritic soils and the differ-
ences of their properties and behaviors compared
– more than 80% of the samples have a percentage
to temperate and colder climatic ones.
of fine lower than 25%.
For the notation Gxy, G means particle size
– more than 90% of the samples have a plasticity
analysis, x the number of the test and y the number
index between 10 and 30.
the sample.
Although, the sample n°1 contains more fines
compared to the sample n°2, it is observed that the
3 CHARACTERIZATION OF TWO
various curves of each sample keep the same form.
LATERITIC SAMPLES
However, the grading curves of the sample n°2 are
almost uniform whereas the grading curves of the
3.1 Selection of samples
sample n°1 are more or less extended.
According to the results of our analysis and the The notation of the legend for Figures 4 and 5 is
literature studies, we decided to focus on the lat- in the same logic as previously.
eritic soils of Burkina Faso. We choose two sam-
ples of lateritic gravels (6 tons for each one) in the
~
great urban area which shelters Ouagadougou, the ~
capital. Each of them are has been submitted to 1'\'
the identification tests in order to confirm and also
establish the geotechnical nature of the material.
~

N.,
I:"'
3.2 Identification of laterites 1-

All of the samples are submitted to a basic charac- SievH [rnm.]

qm qaxn

terization in order to determinate their nature, their Samp le n"2: - G12 - G22 - G32 - G42 - GS2 - G62
Sample n•t: - Gll - G21 - G31 - G41 - GSl - G61
physical and chemical properties. This campaign of
test includes particle size analysis by sieving and
sedimentation, Atterberg Limits and the Meth- Figure 3. Particle size analysis.
ylene Blue test, normal and modified Proctor, un
soaked and soaked CBR. The tests are realized in 24,00

one hand in France and in other hand in Burkina

Faso. The results for particle size analysis, modified "·"
Proctor and soaked CBR are presented from the "·"
Figures 3 to 5. The execution of the tests has fol- 21,00

lowed the international standards (ISO or ASTM)

and are repeated at least 5 times. The choice in the "·"
presentation of these tests can be explained because 19,00

they are the most practical tests in tropical coun- 18,00

tries. For Cissé (2007), the fundamental criterion 17.002,00

'---- -4,00- -- - -6,00- -- - -12,00- -- - -16,1!0- -
for the material selection for base layer is the CBR 6,00 1(,1,00 14,00 16,00

Watercontent w ('!l.)
index which must be at least 80. Some authors like Burkina Fr.mce
Camapum and al. (2015) shown the difficulties in Sample n"2:
S.m ple n"1.: - •
Poly. (I'M12) - • • Pott. (PM22)
Poly.(Piol 11) - - Pot,r.(F'M21) -
Poly. (Piol32)
PDiy. (Pt.lll) -
Potf. [PIU2)
Pot,r.(PI.I<41) -
Poly.[PI.I52)
Poly[Piol51)

Figure 4. Modified proctor of the two samples.

160

140

80

60

40

20

Figure 2. Laterites site or career. Figure 5. Soaked CBR index of the two samples.

785
 The particle size analysis shows that in most
cases, the percentages of fines of the two materi-
als are between 9 and 25%. Beside the results of
Atterberg limits and Methylene Blue test allow to
conclude that the both of them tend to B6 accord- 38cm
ing the French GTR classification or A2–6 accord-
ing to AASHTO classification.
Concerning the modified Proctor and soaked
CBR index, we found respectively around
2.10 t/m3 for the sample n°1 and 2.30 t/m3 for the
sample n°2. The soaked CBR index correspond-
ing to the maximum density can increase beyond Mould Panda2
100 for the two materials. But the most important
for both is that the CBR index decrease when we Figure 6. Calibration instruments.
pass from un soaked CBR to soaked CBR. And
in addition for the sample n°2, with the strong
energy of compaction, the CBR index decreases
also. A situation which can be explained by the
uniformity of the grains, but also by the fact that,
these grains break under relatively important
strength of compaction. About the crushing of
the grains, the percentages before and after the
compaction, vary respectively from 14 to 25%
for the sample n°2 and from 20 to 30% for the
sample n°1.
This situation drove us to study more deeply
the grading curve evolution of the materials.
With strong compaction energy, the grains of
the material broke, thus reducing its bearing
capacity.
Figure 7. Compaction device.
4 PROCESS OF CALIBRATION
Then the soil is introduced in the mould and com-
4.1 Principle and devices pacted at known water content and dry densities
The objective of the calibration is to establish a by a static compaction Chaigneau, Bacconnet &
relationship between the dry density and the cone Gourvès (2000). Each tube is made in two (02)
resistance with reference curves for in situ control. layers. The device of the compaction is shown in
Then the material is compacted and submited the Figure 7.
to the penetration tests to determinate the cone
resistance. So the process uses mainly a mould for 4.3 Penetration tests
compaction, a penetrometer for penetration tests
and also a geo-endoscope. The geo-endoscope is The process of calibration uses the penetrometer
used to supervise the morphology evolution of the Panda. The penetrometer Panda was created in
grains. 1989 by R. Gourvès (1991). It is a light dynamic
penetrometer with variable energy. The idea of
this technology consist to insert using a hammer,
4.2 Compaction test
a train of rods (diameter 14 mm), with a 2 cm2
The compaction of the soil is realized in oedo- cone (Chaigneau 2001& Benz 2009). Confer to the
metric conditions. It uses a system of loading Figure 8. The speed of the impact and the depth
and a mould large enough to avoid lateral fric- are recorded after each blow. The cone resistance
tion between the soil and the mould surface, asit according to the depth is obtained thanks to a
is observed by some authors (Van der Berg 1994, formula called Dutch’s formula (equation 1). The
Lunne et al. 1997, Chaigneau 2001, Ahmadi et results are recorded automatically and are transfer-
Robertson 2004). In this work the speed of the able on a microcomputer to establish the penetro-
loading is 5 kN/s and the dimensions of the gram. Dutch’s formula dates since 1820 (Lowery
mould are respectively 80 cm high and 38 cm et al, 1968). Its use for Panda is mentioned by
diameter as initially proposed by Chaigneau. Suhua Zhou in 1997.

786
 respectively to 6.5%; 9% and 11%. For each soil and
each water content, the material is compressed to
five levels of density, from bulk to maximum den-
sity (100%) of modified Proctor. This maximum
density corresponds to 110% of standard Proctor.
Based on the set of these 15 tests the calibration
curves highlight a relationship between each of the
three calibration parameters (Zc, qd0, qd1) and the
dry density for each of the 3 water contents. Here,
we choose to present some results about the sam-
ple n°2. The construction of the curves (Figures 10
to 12) are obtained by regression. They are in the
following form:

γ d = al (qqdd1 ) b (3)

γd is the dry density, qd1 the cone resistance below

the critical depth, a & b are parameters depending
of the soil nature.
It is shown that the relationship between each
parameter and the dry density depends also of the
moisture content. For the same dry density, the
critical depth and the cone resistance increase from
wet to dry conditions. It exists a good correlation
between these parameters.

Figure 8. Penetration test.

1E M
qd = (1)
A ε (M P )

E: Energy of the impact (J)

M: Weight of the striking mass (kg)
A: Section of the cone (2 cm2)
ε: Residual depth (m)
P: Weight of the struck mass (kg)
The so-obtained penetrogram can be related to
the dry density and the water content.

5 RESULTS OF CALIBRATION

5.1 The penetrogram and its characteristics

The penetrogram is the curve of distribution of the
dynamic cone resistance according to the depth. It
is characterized by three important parameters: 0.6
the critical depth (Zc), the cone resistance at sur-
face (qd0) and the cone resistance below the criti-
cal depth (qd1). In the homogeneous soil, without 0.7
lateral friction, the cone resistance increases from
surface to stabilize itself starting from the critical
depth, with a certain value. 0.8

5.2 Curves of calibration 09~

- ~~~~~~~~

All the soils have been prepared with three water

contents (dry, medium and wet). That corresponds Figure 9. Example of penetrogram.

787
 24
The most important is the low variability of the
23 nature of soils. Their characteristics of identifica-
22
tion show a few variation and the international
standards of classification highlight very few
21
groups of lateritic soils. That explains the need for
¥20 a better classification of these materials to improve
;- their use in road engineering.
" 19

In addition, the mechanical behavior of materi-

16
-
-
Log. {Zc (AO))
Log. {Zc(A1))
R'"' 0,9368
R'" 0,9452
als depends of the moisture content and the state
17 - Log. (Zc(A2)) R" " 0 ,8137 of density. The dry density and the bearing capac-
16 ity (CBR), although they increase with the energy
0,06 0,07 0,06 0,09 0, 1 0, 11 0,12 0,13
Zc(m) of compaction, for a soil constituted of too coarse
grains, the bearing capacity fall with strong ener-
Figure 10. Calibration curve and relationship between gies of compaction. For the calibration of laterites
the dry density and the critical depth for three water using the penetrometer PANDA, it exist a good
contents.
correlation between the dry density and the three
24
parameters of the penetrogram. The critical depth,
the cone resistance at the surface and the cone
23
resistance after the critical depth. For compac-
tion control purpose, the in situ dry density can be
expressed in function of the cone resistance (qd1)
value and the moisture content.

ACKNOWLEDGEMENTS

The authors acknowledge the contributions from

business and technical managers, engineers and
25
qdO(MPa) technicians of SOGEA SATOM and LNBTP
of Burkina Faso. Mainly, Mr Robin TESTOT-
Figure 11. Calibration curve and relationship between FERRY and Kalsibiri KABORE and MrYenté-
the dry density and the cone resistance in surface for maOusmane YONLI for their technical assistance.
three water contents.
24
REFERENCES
23

22 Ampadu, S.I.K and al. 2005. The influence of water

21 content on the Dynamic Cone Penetration Index of
a lateritic soil stabilized with various percentages of a
quarry by-product.
Autret, P. 1982. Latérites et graveleux latéritiques,
16 ISTED/LCPC, Paris, France.
17 - log.(qd1 (AO)) R""0,9363
Bacconnet, C. & Breul, P. 2006. Banque de données de
-
-
Log. (qd1 (A1))
Log. (Qd1 (A2))
R" : 0,9905
R' : 0,9943
matériaux, Rapport d’étude MS-PB4, Contrat de
16
Recherche LGC/Sol Solution, document interne
1 5 ~L_------------------------------------------, Bagarré, E. 1990. Utilisation des graveleux latéritiques en
0 10 20 30 40 50 60
qd1(MPa) technique routière, ISTED/CEBTP, Paris, France.
Benz M. (2009). Mesures dynamiques lors du battage
Figure 12. Calibration curve and relationship between du penetrometre Panda 2, Université Blaise Pascal,
the dry density and the cone resistance in depth for three Clermont Ferrand, France.
water contents. Bohi, Z.P.B. 2008. Caractérisation des sols latéritiques
utilisés en construction routière: Cas de la région de
6 CONCLUSION l’Agneby (Cote d’Ivoire); ENPC, Paris, France.
Breul, P. 1999. Caracterisation endoscopique des milieux
In conclusion of this work and according to the dif- granulaires coupléa l’essai de pénétration, Université
Blaise Pascal, Clermont Ferrand, France.
ferent tests from identification to the calibration, it
Breul, P.; Benz, M.;Gourvès, R.&Saussine, G. 2009. Pen-
is important to note that lateritic soils present clas- etration test modeling in acoarse granular medium.
sical characteristics, even if, it is current to meet International conference on micromechanics of
some difficulties related to the homogeneity of the granularmedia, Powder and Grain, AIP Conf. Proc.,
material during the process of characterization. vol 1145, pp. 173–176.

788
Breul, P.; Haddani, Y. & Gourvès, R. 2004. Diagnostic domaine routier. Institut National Polytechnique de
des ouvrages urbains en interactionavec le sol par cou- Lorraine, France.
plage de techniques rapides complémentaires, XXI- Haddani, Y. 2004. Caracterisation et classification des
Iiemes Rencontres Universitaires de Génie Civil. milieux granulaires par geoendoscopie, Université
Buchanan, F. 1807. Journey through the countries of Blaise Pascal, Clermont Ferrand, France.
Mysore, Canara and Malabar, London, British. LNBTP, 2005. Etude de recherche de matériaux de via-
Camapum de Carvalho, J. and al. 2015. Tropical soils for bilité dans le cadre des travaux d’entretien des routes
highway construction: Peculiarities and considerations. en terre; Ouagadougou, Burkina Faso.
Chaigneau (2001). Caractérisation des milieux granu- Millogo, I. 2008. Etude géotechnique, chimique et
laires de surface à l’aide d’un pénétromètre, Université minéralogique de matières premières argileuse et
Blaise Pascal, Clermont Ferrand, France. latéritique du Burkina Faso améliorées aux liants
Chaigneau, L.; Bacconnet, C.; Boissier, D. & Gourvès, hydrauliques: application au génie civil (bâtiment et
R. 2000. Penetration test coupled with a geotechni- route), Ouagadougou, Burkina Faso.
cal classification for compaction control, GeoEng Morvan, M. and al. 2016. Optimisation of in-situ dry
2000 An international conférence on geotechnical & density estimation, Institut Pascal, Clermont Ferrand,
geological Engineering, Melbourne, Australia. France.
Cissé, I.K. 2007. Caractérisation physico-chimique des Ndaye. M. 2013. Contribution à l’étude de sols latéri-
couches de base en latérite-ciment. 14ème CRA MSG, tiques du Sénégal et du Brésil, Université Paris Est et
Yaoundé, Cameroon. Cheick AntaDiop de Dakar.
Fall, I. 1993. Identification et caractérisation mécanique
des graveleux latéritiques du Sénégal: application au

789
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Using QC LWD data for improving pavement design and management

in Chile and Peru

J.R. Marcobal
Sacyr Construcción, Madrid, Spain

F. Cerrolaza
Sacyr Construcción Chile, Madrid, Spain

J. Arias
Sacyr Concesiones Peru, Lima, Peru

M.A. Moreno
Sacyr Concesiones, Madrid, Spain

ABSTRACT: In South America the systems most commonly used for designing pavements and ana-
lyzing their life cycle costs are based on subgrade resilient modulus. For example, performance models
from the World Bank’s Highway Design and Maintenance Standards Model are based on the adjusted
structural number that is obtained from the structural number of the pavement and the resilient modulus
of the corresponding highway section. In existing roads, in order to calculate the expenses of major main-
tenance, subgrade resilient modulus is frequently obtained through FWD data back calculation, and then
used in calculation software as input data for distress performance models. Likewise, during road con-
struction, quality control data is analyzed obtained directly from LWD measurements on the subbase and
the subgrade, which allows for a detailed characterization of the resilient modulus throughout granular
layers construction and therefore, providing a more reliable adjusted structural number. The main experi-
ences acquired in Chile and Peru during the implementation of this QC procedure is explained.

1 INTRODUCTION − Pavement Evaluator: System developed during

the mid 90’s by the FICEM, during work on the
The PPP Highway concessions in Chile and Peru HDM-IV in which the HDM-III deterioration
have by contract, an obligation to define a main- models in bituminous pavements and the new
tenance plan, which must be updated annually. models for concrete of the HDM-IV that were
Investment in pavements is one of the most impor- being developed, were combined.
tant of its chapters. The systems most commonly − HDM-IV: System developed during the 90’s in
used to analyze the life cycle costs of pavements the ‘International Study of Highway Develop-
are based on the deterioration models from the ment and Management’ (ISOHDM) in which
World Bank’s Highway Design and Maintenance the World Bank and other organizations and
Standards Model and its further developments. countries participated. It includes models in
In order to implement the systems in a practical bituminous and concrete pavements, and is cur-
manner, various systems have been used, those of rently the most used system.
which are listed below from the oldest to the most
recent: Within the HDM-IV, Initiation of All Structural
Cracking Model for bituminous pavements with a
− HDM-III: System developed during the 80’s by granular base layer and the rutting model (initial
the World Bank with deterioration models in densification and structural deformation compo-
bituminous pavements. nents), a parameter is included which characterizes
− GIMP (Pavement Management and Mainte- the bearing capacity of the pavement, the adjusted
nance System): System developed during the Structural Number of the Pavement (SNP), that is
mid 80’s by the Highway Agency of the Minister obtained from the structural number of the pave-
of Public Works (MOP) in Chile, and was based ment and the characteristics of the subgrade. This
on the deterioration models of the HDM-III. value is calculated in the following way (equation 1):

791
SNPs = SNBASUs + SNSUBGs + SNSUBAs (1) Table 1. Reduction coefficient of the back-calculated
modulus.
where SNPs = adjusted structural number of the
Subgrade “C” Value
pavement for season s; SNBASUs = contribution
of surfacing and base layers for season s (equa- Fine soil 0,33
tion 2); SNSUBAs = contribution of the sub-base Granular soil 1,00
or selected fill layers for season s (equation 3); Granular embankment 0,33 + 0,33 × h 0<h<2m
SNSUBGs = contribution of the subgrade for sea- above fine soil
son s (equation 4). Fine soil embankment 1–0,33 × h 0<h<2m
above granular soil
n
N U S = 0 0394 ⋅ ∑ ais hi
SNBASU (2) *h = Height of embankment (m). For h > 2 m, use
i =1 C = 0,33 ó 1 as guided.

SNSU
N UBG GS = (b b exp( −bb zm ))(exp( − b3 zm ))
(3)
× [3.. log10 CBR
C Rs − 0. (log
o 10 CBR Rs )2 .43] An alternative method is a deflection survey
once finished the construction on the road surface
and to back calculate subgrade resilient modulus
SNSU
N UBA
AS indirectly. AASHTO 93 guide indicates that the
⎧⎛ b0 b3 z j ) ⎞ ⎫ value obtained this way doesn’t correspond directly
⎪⎜ ⎟ ⎪ to the laboratory resilient modulus that is included
⎪⎜ − b3
⎟ ⎪ in the AASHTO equation and should be reduced
⎪⎜ b1 b2 + b3 )z j ) ⎟ ⎪
m ⎪⎪⎜⎝ + ⎟⎠ ⎪⎪ according to the type of soil and other factors. In
(b2 b3 )
= 0.0394 ⋅ ∑ a js ⎨ ⎬ (4) the AASHTO 93 method, it is indicated that this
j =1 ⎪ ⎛ b0 − b3 z j −1 ) ⎞⎪ coefficient is 0.33. In Chile this correction is pro-
⎪ ⎜ − b3 ⎟⎪ vided in the Highway Manual using the following
⎪− ⎜ b − b + b z )
⎟⎪ expression below (5):
⎪ ⎜⎜ + 1 2 3 j −1 ⎟
⎟⎠ ⎪⎪
⎩⎪ ⎝ (b2 + b3 ) ⎭ Mr = C × EFWD (5)

where N = number of base and surfacing layers where: Mr = Resilient Modulus for the AASHTO
(i = 1, n); Ais = layer coefficient for base or sur- equation; EFWD = Resilient Modulus obtained
facing layer I for season s; hi = thickness of base by back-calculation and C = Coefficient Corrector
or surfacing layer i, in mm; m = number of sub- indicated in the following Table 1.
base and selected fill layers (j = 1, m); ajs = layer Values obtained by this procedure are used to
coefficient for sub-base or selected fill layer j for define pavement homogenous sections and to
season s; z = depth parameter measured from the achieve greater precision in the input values of the
top of the sub-base (underside of base), in mm; deterioration models of HDM-IV.
zj = depth to the underside of the jth layer (z0 = 0), In the article, the work carried out is described
in mm; and CBRs = in situ subgrade CBR for sea- using an alternative procedure based on character-
son s. izing the structural contribution of the subgrade
It is necessary to mention that the indicated through deflection measurement by applying
models are based on an adjusted structural number load directly on the subgrade surface during the
that considers that the contribution of the sub- construction. This method, unlike the measure-
grade by means of its CBR is affected not only by ments on the wearing course, has the advantage of
the material own characteristics, but also by the obtaining the subgrade elastic modulus directly by
construction conditions. As a result the term ‘in application of the Boussinesq equation. These val-
situ’ is used to define the subgrade CBRs. ues are more reliable and therefore do not require
In common practice, as a first approach of the any correction. Their use as input values for the
project phase, design theoretical CBR is used. deterioration models of HDM-IV, reflects in the
Given that it is usually obtained in laboratory test- most reliable way, the structural contribution of
ing material samples, this value can be more or less the subgrade. Additionally, results gathered during
representative of the material that is finally used. construction later become part of the quality con-
However, it doesn’t ensure the real conditions of trol process and allow for the corrective action to
the CBR after construction which will be deter- be taken on the subgrade should it be considered
mined by other parameters such as the percentage necessary at locations failing to meet specification
of its maximum dry compacted density. before placement of next lift.

792
2 EXPERIENCE GAINED IN CHILE IN 2. Higher loads can be applied so measurements
THE IN SITU CHARACTERIZATION OF can be done on upper layers.
THE SUBGRADE BEARING CAPACITY − LWD:
1. Flexibility to locate the device on any part of
During a year, requirements and specifications the construction site.
compliance of the resilient modulus has been 2. Easy to handle and operate
monitored for a section of highway, with its main 3. Given that cost is much lower than those of
objective being to check construction processes and the FWD, it may be easier to get it as a work
reduce the risks in obtaining the design parameters. site laboratory device. In this way, the device is
Below, the most important points which have been always available and can be used at any given
experienced during these works are presented. time, independently of the works production
rate. That is more difficult with the FWD,
2.1 Measurement devices used which if it’s available, must be moved from
Santiago.
To measure the deflection of a section of highway
on the granular subgrade, two types of devices With regards to the maintenance of the LWD
have been used: device, operated by the concessionaire, it is nec-
essary to indicate that intensive usage has caused
− Falling Weight Deflectometer (FWD). This
various breakdowns to appear, which has shown
device is the most common for this type of
that the device is less robust than the FWD. With
measurement. The measurement procedure with
a sustained advance in measurements with 6,000
this device is described in detail in the stand-
controlled points registered, a first problem with
ard ASTM D4694: Standard Test Method for
the device was found. The LWD has a trolley sys-
Deflections with a Falling Weight Type Impulse
tem which is used to transport the device within
Load Device. Measurements were carried out
the work area. However, the axle is used as a pull-
by an external company with a Dynatest type
ing handle what should have generated a deforma-
FWD.
tion in this main axle. A decision had to be taken to
− Light Weight Deflectometer (LWD). This device
send an important part of the device to Denmark
is designed specifically to be used in the control
(to be repaired) which caused a significant delay
of subgrade and granular base layer. The main
for the studies being carried out. Furthermore, the
difference between the LWD and the FWD is
analysis that was being carried out lost its conti-
that, due to the fact the LWD is controlled man-
nuity. Once the part had returned, another fault
ually, the maximum load that can be applied is
appeared in one of the geophones connections.
15 kN, much less than what can be applied with
As a result of this incident, it was decided that for
the FWD (120 kN). This measurement method
intensive quality control it was reasonable to meas-
is described in the standard ASTM E2583-07:
ure without the auxiliary geophones and only use
Standard Test Method for Deflections with a
them when more detailed information was needed.
Light Weight Deflectometer (LWD). In this
study the model used was the LWD 3031 by
Dynatest. This type was selected because uses 2.2 Measurement methodology
the same sensors as the Dynatest FWD and
that should improve comparisons. Measure- The methodology set out to collect field data,
ments were taken by the concession own staff. was focused on the work developed by the opera-
For the operator selection process, a certain level tor and the device and the arrangement of the
of training was needed, in order to evaluate in test points where the deflections were to be
situ soil or granular material characteristics and measured.
condition. This technical profile of the opera-
tor allowed for solving in situ doubts or to get 2.2.1 Testing rate
clear explanation of any other incident involv- Points were selected so they were located equi-
ing measurements. distantly and depending on the area given by the
earth works units in each sector, which varied in
Both devices provide satisfactory results from a length. For this reason, in sections longer than
practical point of view, and advantages from each 1,000 m tests every 100 m and lane were done. In
side can be appreciated in different aspects. shorter sections, test points were selected to match
− FWD: FWD test locations. This ensured a minimum of
1. Long experience gained by companies with 10 points per section and lane at distances between
the device, operators and technicians who points ranging from 50 to 100 m. Transverse layout
analyze deflection data. of test points was focused on controlling specially

793
the working platform sides to look for any com- 2.2.4 Calculation of elastic modulus
paction problem. Deflection measurements recorded allows for the
With this arrangement, control points to test elastic modulus of the subgrade to be calculated
subgrade deflection were between 2–7 daily per using the following expression below (6).
road section, covering the total length of the high-
way (180 km) with one device. 2 (1− 2
).σ 0 .a
E0 = (6)
d0
2.2.2 Drop number
The number of load drops at each station was
determined according to the experience gained where E0 = surface modulus (MPa); ν = Poisson’s
and considering prior information gathered from ratio, usually 0.35; σ0 = maximum contact stress
literature. (kPa); a = plate radius (m); d0 = maximum deflec-
One of the first issues observed was that the tion (mm).
subgrade deflection changed with additional load
drops showing a compaction effect in spite of the 2.2.5 Other subgrade characteristics
fact that the required density target had already FWD test points were located on subgrade that
been achieved (95% of the DMCS). So the modu- was controlled for complying with the density
lus calculated based on deflection measurements specification (95% DMCS) and with the moisture
increased with additional drops, which can be content closed to optimum.
observed in Figure 1 (below). Temperature and humidity readings were incor-
As established in some works, first, second and porated into the measurement data set with the
third drops are seating drops (compaction effect) aim of studying possible variables that could affect
and the three following drops gives representative deflection values and the layer elastic modulus.
values that should be averaged. From results of
more than 15.000 test points, also fourth, fifth and
2.3 Evaluation procedure
sixth drop values were considered representative
avoiding the compaction effect that affects first Subgrade resilient modulus defined in the pave-
values. Applying more drops than six didn’t get ment design of the highway project was 200 MPa
lower deflections values. for a total traffic of 15 msa that together resulted
Based on the above, it is advisable to apply at in a structural number required above subgrade of
least six consecutive drops at each test point. 6.9 cm.
During this study, several options were studied:
2.2.3 Target load
− Following the AASHTO criteria, defining
Stress specified for the test ranged from 100 to
homogenous sections and comparing the aver-
150 KPa with a peak load of 10–12 KN. This is
age measured modulus obtained from the deflec-
achieved adjusting the height from where the load
tions with the design value of resilient modulus.
is released, with a maximum load weight (20 kg).
− requiring that all of the points comply with the
This stress is within the range that the layer would
value of design modulus.
support when traffic loads are applied on the road
− Applying a statistical processing similar to that
surface.
of the Minister of Public Works in Chile, remov-
The plate size used was 300 mm in diameter, in
ing values of extreme percentiles.
order to make easier a comparison between the
FWD and the LWD results. In this work, data obtained in the first set of
measurements met specification in almost every
section, applying any of the previous criteria.
Given the complexity of the variables that affect
the LWD modulus measurements, it wasn’t possi-
ble to define a reliable methodology with an accept-
ability criteria to apply in sections with a modulus
lower than 200 MPa. Neither was it possible to
define a procedure that considered former design
studies that showed the variability of the material
as the distribution of the subgrade modulus in the
i
lOO rehabilitated carriageway has CV between 28%
and 47%.
In cases when section doesn’t meet specifica-
Figure 1. Elastic modulus values from multiple LWD tion value, earthwork unit was requested for a
load drops. recompaction in order to increase the modulus.

794
These sections increased their modules effectively explanation. For this reason, it has been decided
however did not generate a homogenous result as that during the next measurements, a thorough
shown in Figure 2, which shows modulus of some analysis of the moisture content for each testing
points before and after recompacting. Once the point will be performed. In this work, analyzing
recompaction was done and the compliance with several sections which had been tested, an average
a new test was verified, authorization was given increase of 30% of the modulus value was obtained.
for the following granular base layer. It should be
remarked that density of these sectors was ade-
2.5 Comparison with back calculated resilient
quate (above 95%).
modulus
These results have made complicated the pro-
posal and implementation of corrective actions, Direct measurements on the subgrade were com-
as results has not shown clearly whether the rec- pared with values of modulus bakckcalculated
ompaction process of re-compacting is effective from deflection tests on the wearing course. The
or not. latter were measured using a FWD and followed
the back calculation procedure set up by the Labo-
ratorio Nacional de Vialidad de Chile (National
2.4 Evolution of Mr values with time
Laboratory of Highways in Chile) in the Highway
Another issue that made difficult to understand Manual (8.5020.5 and 8.502.6). A big difference
some aspects of this type of control was that some between the direct values measured on the sub-
sections that were initially rejected, without any grade during construction and the results back cal-
recompaction improved their results after a period culated from deflections on the wearing course was
of time as shown in Figure 3. It is understood that observed. In Figure 4, the result of direct measured
in some cases, these sections have been exposed to modulus (represented with dots) is compared with
construction traffic which could have produced the range of the backcalculated resilient modulus
compaction but only in wheel paths. Further- values, which depends on the reduction coefficient
more, the loss of moisture content may be another which ranges from 0.33 to 1.
In Figure 4 several issues should be mentioned.
This section seems to be made up of three subsec-
tions according to their subgrade resilient modu-
lus. Considering different modulus values in this
range, for example 200 and 350 MPa, the way in
which subgrade sectioning influences can be seen
in Figure 5, the results of the HDM-4 all cracking
initiation model.
Furthermore, it is possible to see in Figure 4,
that the maximum reduction coefficient used in
0 the Chilean Highway Manual, C = 1, looks rea-
0
sonable in the majority of the sections analyzed;
however, it is a little conservative in some subsec-
tions. Taking into consideration all these parame-
Figure 2. Resilient modulus before and after ters and information in the analysis, it is clear that
recompaction. there is not enough evidence to establish a clear
relationship. Data collecting processes should
go on in order to advance in this field and have
clearer conclusions.

0~
.- i ·~ ~:-~_·~c~-~~,~·:~··~·~~~··~'~/~··' ~·~--~---·~:~:~'~:~7.~-·_-_>'_.._--~;~·',·-_-_-"_~_~,~-­
~
·'
,
/.~.~-J~~--f. ----.. \ .."\...._-~ , . FWDimo)
~I
j <• l---ei-"- __:,· ,_·~··.'---=~~----,:i-"
. -:.·':.--'- ·. '--'"--'"'-----'-
"~ ·'~~··-----1
-- -·~ --.,..-·:_ ____ . . ---i - --- - -- ·: .--r;.;;:-
----
0 2807'2010

Figure 3. Increase in MR after a period of time without Figure 4. Comparison between direct modulus and
additional recompaction. back-calculation.

795
 .~xl¢ 1oading(:-..·WSAI. ~·tnr)
Figure 6. Results from LWD.

Figure 5. Result of the model of initial MR cracking.

GROWTH TN RESTLTENCE :\·IODULT (in da:-·s)
270

3 EXPERIENCE IN PERU

2!0·-J
:no
One of the most interesting experiences was located
in a project in the Andean region of Peru between
1600 to 2700 m.a.s.l. The road was a second-class
road with 91 km long. The pavement thickness was li •)

37.5 cm, with an asphalt layer of 7.5 cm and the 150

granular layer of 30 cm. For the control and moni- ~

·- "" ~ -

- ~
- ~ ~
c; .. ! c
~ ~ ~
,~' ]' l '
~
J'
toring of the pavement construction, the LWD
device was used extensively on granular layers.
Later, after laying the wearing course, final pave- Figure 7. Increase in Resilient moduli.
ment was surveyed using the FWD.
It is important to take into consideration that in
the design traffic was 3.4 Million ESALs (10 years value increased in a period of time, 2 to 4 days,
design period). The study was developed in granu- after finished the layer executed, and if this growth
lar layers and subgrade level. in Module is because the loss of humidity or road
traffic application.
3.1 Evaluation procedure The data collected in some controlled layers
Whenever the subgrade did not meet the quality showed that if there is an increase of the Mod-
requirements of the Technical Specifications, it ule in time, even after a short time (2–4 days from
had to be replaced in its upper part by a layer of compaction). Although a small difference, it is
granular material with improved properties. considered relevant to the purpose of the study
As a pavement layer is completed, the Resilience (Figure 7).
Moduli (Mr) of that layer is controlled using the This growth is because the loss of humidity and
LWD device. This measured Mr value is compared construction traffic over compaction Due to the con-
to the design value in the project. struction process, measurements could not be made
In every section where subgrade layer didn’t exactly on the same section as these were covered.
need to be improved, the value of the Resilience
Moduli should be 250 MPa; on the other hand, 3.3 Comparison with back calculated resilient
in the case of a section where the subgrade, given modulus
the low qualitiy, had to be replaced, the required
Resilience Moduli was considered acceptable over After laying the asphalt concrete in the sections of
a value of 180 MPa. Almost all the subgrade sec- this study, an evaluation with the FWD was made
tions surveyed during the study included replac- by means of back calculating the values of Resil-
ing layers of poor quality material. With this ient Moduli for the subgrade.
consideration, Mr values obtained with the LWD Figure 8 compares the Mr values obtained from
show an average of 180 to 200 MPa. (Figure 6). the LWD and FWD. In general, values obtained
from the back calculation of FWD measurements
on the wearing course are above of those obtained
3.2 Evolution of Mr values with time with the LWD directly on the subgrade. The dif-
A detailed study was developed to verify the evolu- ference between both results is between 25 and
tion of Resilience Moduli values with time, if this 50 MPa.

796
 planned in some control section but problems arose
during the construction that made difficult to repeat
measures at the same points as some layers were
covered before measurement could be made.
Differences were found between the value
obtained from measurements with LWD and FWD.
Some of the studies show difference between 25 to
50 Mpa or more than a 25%. The LWD Resilience
Module values were, in general, below the Mr values
KILOMETERS IN THE ROAD
obtained by the FWD. It seems that depending on
the stress sensitivity of the soil, these values scat-
tered all over the sections.
Figure 8. Comparison FWD vs LWD. Given that moisture content is one of the vari-
ables that is thought to influence results, further
studies should be designed to develop moisture/
4 CONCLUSIONS moduli curves that allow establishing this relation-
ship and later setting thresholds approval criteria.
In general, the bearing capacity input param- Data collected and Information obtained from
eters of some of the deterioration models of the LWD is considered quite reliable. However, since
HDM-IV, depend partly on the CBR in situ of it is a manual equipment, its reliability will also
the Sub-base. This characteristic is selected using depend on the training of the operator (how to set-
project information or by indirect back calcula- tle the equipment plate, how to select the measure-
tion procedures starting with measurements on the ment point, among others) and the availability of
road surface. These results obtained indirectly are moisture and material density at the time of meas-
at the same time, corrected by generic coefficients urement, for example.
(AASHTO C) which have a direct influence on the
model outputs and the maintenance plans.
REFERENCES
The experience gained in this work has been
useful to better understanding that the technology De Solminihac, H. 1998. Gestión de infraestructura vial.
allows to evaluate the bearing capacity of the sub- Santiago de Chile: Ediciones Universidad Católica de
grade and granular layers though direct measure- Chile.
ments in situ. Implementation of this technology George, K.P. 2003. Falling Weight Deflectometer for esti-
to the subgrade is relatively new and requires fur- mating subgrade resilient moduli. Mississippi: Uni-
ther work focused on understanding the variations versity of Mississippi.
and results obtained during the study which have Osorio, A. 2008. Metodología de evaluación in-situ del
been, on occasions, surprising. modulo resiliente de capas granulares de pavimen-
tos flexibles con el deflectómetro de impacto liviano.
Results showed that construction traffic and
Puerto Varas: Provial 2008.
water content of surveyed layers strongly influenced Puppala, A. 2008. NCHRP synthesis 382: Estimating
values measured, although a significant dispersion stiffness of subgrade and unbound materials for pave-
was found with no clear pattern. Most Mr increased ment design. Washington: Transportation Research
over time by an average of 13%. Measurement were Board of the National Academies.

797
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

E* prediction algorithm for pavement quality control assessment

K. Georgouli & A. Loizos

Laboratory of Pavement Engineering, National Technical University of Athens, Athens, Greece

ABSTRACT: The determination of pavement original condition immediately after construction is of

critical importance, since concerns arise as to whether the constructed pavement corresponds to the pave-
ment and mix design requirements. Focusing on the asphalt layers, the HMA dynamic modulus (E*)
could be a key parameter. E* determination through laboratory testing is time consuming and needs
specialized equipment. Therefore, it would be of great importance to be able to estimate the E* of the
asphalt mixes through an algorithm, based on the experimental results of common practice testing during
the mix design procedure. The objective of the present research study is to investigate the utilization of E*
prediction algorithm in the quality control procedures of a newly constructed pavement. Analysis showed
that E* is a critical parameter for the assessment of the HMA quality control procedures. The possibility
of estimating the E* through prediction algorithms makes the proposed methodology easily applicable.

1 INTRODUCTION modulus is also considered in the frame of the

quality control procedures (NCHRP 2004).
The determination of pavement original condition In light of the above, the HMA dynamic modu-
immediately after construction is of critical impor- lus (E*) could be a key parameter (Cominsky et
tance, since concerns arise as to whether the con- al. 1998, Bonaquist et al. 2003). E* is fundamental
structed pavement corresponds to the pavement property of a viscoelastic material in the frequency
and mix design requirements. The importance of domain. It can be determined from sinusoidal load
determining the pavement mechanical characteris- applied at different frequencies to capture the lin-
tics shortly after the construction lies on the fact ear viscoelastic properties of the Hot Mix Asphalt
that it is those characteristics that will define to a (HMA) mixture.
great extend the evolution of the pavement behav- The E* which can be utilized for mixture rank-
ior on the long term. ing and characterization purposes and mix design,
Focusing on the asphalt layers, traditional qual- is a primary stiffness property used as an input
ity control procedures are based determining the parameter in Mechanistic-Empirical Pavement
properties of asphalt mixtures (density, air voids Design (MEPDG) and analysis processes for the
and permeability) measured either using in situ prediction of fatigue and rutting damage.
methods at the surface of an asphalt layer or tak- The master curve generated at a reference tem-
ing cores for laboratory measurements. Although perature based on dynamic modulus test results
these traditional quality control procedures play is utilized to calculate the dynamic modulus, and
an important role in ensuring high quality pave- thus describe the HMA performance, over a wide
ments, they employ only the volumetric proper- range of temperatures and frequencies.
ties but not actual mechanical properties such as Through E* master curves comparative evalu-
modulus, despite the fact that it is a key parameter ation of the designed mix and the mix that was
for the pavement design. The consideration of the actually constructed, more generalized and safe
same mechanical characteristics in the quality con- conclusions can be drawn. The master curves slope,
trol procedures as in the pavement and mix design, maximum and minimum E* values, whether they
allows a more accurate estimation of the effect that are parallel or non to each other, as well as their
a material of inferior or superior quality may have point of intersection can provide valuable infor-
on the pavement performance. Hot Mix Asphalt mation of the HMA performance under various
(HMA) modulus is the dominant characteristic temperatures and frequencies. Taking into consid-
that determines not only the asphalt layers behav- eration that during operation the pavement will be
ior but also the behavior of the pavement, due to subjected to different environmental conditions
the contribution of the asphalt layers in reducing and vehicle traffic speeds, the critical importance
the stresses transferred to the underlying layers. of asphalt mix quality control over a wide range of
Consequently, according to recent studies HMA loading conditions emerges.

799
 It is often the case that HMA stiffness data,
and especially E* values, are not available. The
Marshall mix design aims mainly to the determi-
nation of the optimum asphalt content and air HMA
void content. In addition, the asphalt mixes that layers
will be constructed are not always available dur-
ing the design of the pavement. In light of the
above, it would be of great importance to be able
to estimate the E* of the asphalt mixes through
an algorithm, based on the experimental results
of common practice testing during the mix design
procedure (aggregate gradation and binder charac-
teristics). Therefore, E* could be accessed with no
additional testing as part of the common practice
mix design procedures. E* prediction algorithms
can be further utilized to estimate the E* of the
mix that was actually constructed. The necessary
input data can be obtained from extracted cores
shortly after construction.
The objective of the present research study is to Figure 1. Pavement design cross section.
investigate the utilization of E* prediction algo-
rithm in the quality control procedures of a newly Table 1. HMA aggregate gradation, Section A.
constructed pavement. More specifically, a newly
developed E* algorithm will be utilized to estimate Asphalt binder Asphalt base
the modulus of the asphalt mixes during the pave-
ment design process. E* of these mixes will be also Sieve Passing (%)
determined through laboratory testing. Estimated
19.00 mm (3/4 in.) 100 91
and determined E* values will be compared to
9.5 mm (3/8 in.) 78 65
those that were actually achieved in situ.
4.75 mm (No. 4) 52 44
For the purpose of the present study an in-
75 μm (No. 200) 4 3.4
situ road experiment was held at two newly con-
structed asphalt pavements. More specifically,
upon the pavement design and prior to the con-
Table 2. HMA aggregate gradation, Section B.
struction of the asphalt layers, loose asphalt mix
was collected for the preparation of asphalt mix Asphalt binder Asphalt base
specimens according to the mix design. Asphalt
mix specimens were tested for the determination Sieve Passing (%)
of their E* in the laboratory. E* values were also
estimated through a prediction algorithm. Upon 19.00 mm (3/4 in.) 100 100
completion of the pavement construction, cores 9.5 mm (3/8 in.) 82 61.8
were extracted and transferred to the Labora- 4.75 mm (No. 4) 53.1 43.1
tory of Pavement Engineering of NTUA for the 75 μm (No. 200) 5.1 4.5
determination of the E*. The quality control was
accomplished through comparative evaluation of
the designed (both estimated through algorithm the pavement design cross section, common for
and determined through testing) and constructed both pavement sections.
asphalt mixes E* master curves. Fatigue cracking HMA layers consist of the asphalt binder layer
was also estimated considering the E* values of the and the asphalt base layer. The mixture of the
designed and constructed asphalt mixtures. asphalt layers was produced using 30–45 grade
asphalt penetration. Modulus of the base/subbase
layer is considered equal to 200 MPa, while sub-
grade modulus equal to 80 MPa.
2 EXPERIMENTAL STUDY
Aggregate gradation of asphalt mixes used for
the construction of sections A and B asphalt lay-
2.1 Experimental road sections—asphalt mixes
ers are shown in Tables 1 and 2, respectively. Test
The experimental study was performed at two results on asphalt, Penetration index (PEN) and
newly constructed pavement sections, 890 m (Sec- softening point (TBA), are given in Tables 3 and 4,
tion A) and 790 m (Section B) long. Figure 1 shows for HMA mixes of sections A and B, respectively.

800
Table 3. Binder characteristics, Section A. loading and proceeded to the lowest. At the begin-
ning of testing, specimens were preconditioned
Asphalt binder Asphalt base with 200 cycles at 25 Hz. The applied loading was
o of a certain magnitude for each testing tempera-
PEN (at 25 C) 36.3
ture so as to obtain axial strains between 50–150
TBA(oC) 56.3
microstrain.
E* was estimated through a newly developed
Table 4. Binder characteristics, Section B. algorithm (Georgouli 2016) which has proven to
have good applicability to these mixtures. The
Asphalt binder Asphalt base mathematical expression of the algorithm is given
in equation 1.
PEN (at 25oC) 35 35
TBA(oC) 56 55
8.103 0.830 p200 + 0.0573 ( p ) 2
log E *
⎛ Vbeff ⎞
−0.0209 p4 − 0.0608Va − 0.4774
7 ⎜ ⎟
Table 5. Volumetric composition. ⎝ Vbeff Va ⎠
6.7703 +0
+ 0.00763 p4 − 0 3705 p38 + 0.00663((pp38 )2 0 0344 p34
Specimens Va (%) Pb (%) +
1 e p ( 1.01849 − 0.4966 log(( f ) − 0.6480 log( ))
Asphalt binder layer—Section A 4.6 4.2 (1)
Asphalt base layer—Section A 4.4 4
Asphalt binder layer—Section B 3.7 4.3
Asphalt base layer—Section B 4.3 4.2
where Ε*: dynamic modulus of mixture (psi), η:
viscosity of binder (106 poise), f: loading frequency
(Hz), Va: air voids (% by volume), Vbeff: effective
The compaction of the specimens was per- binder (% by volume), ρ34: cumulative percentage
formed in the NTUA Laboratory of Pavement retained on 3/4 inch (or 19 mm) sieve, ρ38: cumu-
Engineering with a gyratory compactor according lative percentage retained on 3/8 inch (or 9.5 mm)
to EN standards (EN 12697-31 2007). Gyratory sieve, ρ4: cumulative percentage retained on No. 4
compaction is the compaction method required for (or 4.75 mm) sieve, ρ200: percentage passing No. 200
the preparation of specimens subjected to dynamic (or 0.075 mm) sieve.
modulus testing (AASHTO T 342-11). Two to E* values were estimated for the same tempera-
three specimens were compacted from each asphalt tures and loading frequencies that the laboratory
mix and the air void content target was according testing for the determination of E* was performed.
to the mix design. In Table 5 the mean value of the E* of the constructed HMA layers was deter-
air void (Va) and asphalt content of the specimens mined through laboratory testing on cores
are presented (Pb). extracted from the pavement shortly after construc-
tion. In total 10 cores were taken (5 from section A
and 5 from section B). The total HMA thickness
2.2 HMA Dynamic modulus (E*) was in any case greater than 15 cm, thus enabling
E* testing of the compacted specimens was per- the E* testing according to AASHTO standards
formed at the NTUA Laboratory of Pavement (T 342–11).
Engineering according to AASHTO standards
(T 342-11). A controlled sinusoidal (harvesine)
2.3 Viscoelastic analysis
compressive loading was applied to each specimen
for a range of loading frequencies and tempera- The viscoelastic analysis of the under investiga-
tures. Due to difficulties obtaining reliable results tion pavement section was completed by 3D-Move
for the temperature of −10°C and 54°C required analysis program. The finite-layer approach three-
in related AASHTO standards, testing was con- dimensional moving load analysis (3D-Move) has
ducted at five temperatures 4, 10, 20, 25, 37°C and been developed by the University of Nevada, Reno
six frequencies, 25, 10, 5, 1, 0.5 and 0.1 Hz. Other (UNR) (Siddhartan et al. 2000) and treats each
researchers (Bennert & Williams 2009, Georgouli pavement layer as a continuum and uses the Fou-
et al. 2016) have also noted these difficulties. rier transform technique. The finite layer method
Each test specimen, individually instrumented is much more computationally efficient than the
with LVDT brackets, was tested for each of the moving load models based on the finite element
30 combinations of temperature and loading method (Huhtala & Pohlajamaki 1992). In addi-
frequency starting with the lowest temperature tion, rate-dependent material properties (viscoelas-
and proceeding to the highest. Testing at a given tic) can be accommodated by the approach (Ulloa
temperature began with the highest frequency of et al. 2013).

801
 For the fatigue analysis, the approach to calcu- layer, C1,2,3 are transfer function regression con-
late the allowable number of axle load applications stants; C3 = 6000; C1 = 1.00; C2 = 1.00, C*1 = −2C*2,
needed for the incremental damage index accord- C*2 = −2.40874–39.748(1 + 39.370 HHMA)−2.856.
ing to NCHRP 1-37A is shown in equation 2. For the purpose of the present research a reli-
ability level (R) of 95% was considered according
βf 2
Nf HMA . k f (C )(CH ) β f (ε t )k f to the recommendations by (AASHTO 2008) for
(E HMA)
H
kf β f 3
(2) primary roads.
Fatigue cracking was calculated at the end of
where Nf-HMA is the allowable number of axle-load the design life of the current pavement structure
applications for a flexible pavement, εt is the ten- taking into consideration both, the E* values
sile strain at critical locations calculated by the measured in the laboratory and the ones estimated
structural response model, in./in, E*HMA is the using the four prediction models, under the same
dynamic modulus of the HMA, psi, kf1, kf2, kf3 temperature conditions and loading frequencies.
are global field calibration parameters (from the It is noted that calibration of the fatigue crack-
NHCRP 1-37A calibration; kf1 = 1, kf2 = −3.9492 ing model in order to increase the accuracy of the
and kf3 = −1.281), βf1, βf2, βf3 are local or mixture pavement design procedure, may be considered.
specific field calibration constants; for the global However, in the framework of the present research
calibration effort, these constants were set to 1.0. study it is used for comparison purposes in order
for the sensitivity analysis to be performed.
C = 10 M (3)
⎛ Vbeff ⎞
M = 4.84 ⎜ − 0.69⎟ (4) 3 DATA ANALYSIS
⎝ a
V V beff
f ⎠
3.1 E* master curves and induced compressive
where Vbeff is the effective asphalt content by vol- strains
ume, percentage, Vα is air void percent in the HMA Given that cores extracted from the pavements
mixture, CH is the thickness correction term consist of two asphalt layers (asphalt binder and
asphalt base course) with different mechanical
1 characteristics, the E* of the constructed HMA
CH = (5)
0.003602 layers determined through laboratory testing refers
0.000398 +
1+ exp ( − MA )
H HM to the composite E*. In order for the evaluation
to be enabled, E* data of the laboratory prepared
HMA mixtures are combined with the thickness
where HHMA is the total HMA thickness, in. data, for the calculation of the composite E* of
The cumulative Damage Index (DI) for the the design HMA mix. The E* of a complex mate-
design life of the pavement (a design life of 20 years rial (in the sense that it consists of two layers) can
was selected for the simulation) was calculated at be calculated through equation 8.
critical location by summing the incremental dam-
age indexes.
( (( h / h ) ))
3
* /
Ecomp i comp
m E i* (8)
⎛ n ⎞
DI = ∑ ( DI ) = ∑ ⎜ ⎟ (6) where hi: the thickness of the “i”th layer, E*i:
⎝ N f HMA ⎠
dynamic modulus of the “i”th layer, i: 1 to n, where
n is the number of HMA layers, hcomp: the total
where n is the actual number of loads expressed in HMA layers thickness.
ESALs within a specific time period. In Figures 2 and 3 E* master curves of the
Once the cumulative damage index was pre- asphalt mix layers and the induced tensile stains
dicted, transfer functions were used to convert at the bottom of the asphalt layers are presented
data into alligator cracking using equation 21. for sections A and B, respectively. The three
E* master curves correspond to the E* of the
⎛ C3 ⎞ design mix determined through laboratory testing
Cbottom = 0.18 ⎜
FC (7)
)) ⎟⎠
(E*mix_lab) and estimated through the algorithm
⎝ 1 + exp (C C + C C (DI × (E*mix_pred), and to the E* of the cores (that
were taken shortly after construction) determined
where FCbottom is the area of alligator cracking through laboratory testing (E*core). The reference
that initiates at the bottom of the HMA layers, temperature for the construction of the E* master
percentage of total lane area, DI is the cumula- curves was 20°C. The induced tensile strains at the
tive damage index at the bottom of the HMA bottom of the asphalt layers was calculated tak-

802
Figure 2. E* master curves and induced tensile strains at the bottom of the asphalt layers—Section A.

ing into consideration the E*mix_lab, E*mix_pred E*mix_pred values slightly deviate from the
and E*core values. E*mix_lab values and RMSE values range from
According to Figures 2 and 3, the estimated E* 4.9 to 13.4%. In almost every case E*core values
values are close enough to the ones measured, pro- deviate widely from those of the designed mix.
ducing evidence of the good applicability of the The maximum value of the induced tensile
algorithm. Core master curves in general are not strain at the bottom of the asphalt layers consider-
parallel to the ones of the design mix and as such ing the E*mix__pred values, deviate from −2.7 to
under some loading conditions values are larger 3.6%, with regards to those calculated considering
and under other conditions values are smaller. More the E*mix_lab values. Deviations range from −14.1
specifically, for high temperature and/or low fre- to 14.5% in case of E*core values are considered.
quency E* values of the constructed asphalt mixes
are lower or equal to those of the designed mixes. In
case of lower E* values deviations are not signifi- 3.2 Fatigue analysis
cant. On the contrary, at lower temperatures and/or In Figures 4 and 5 the Damage Index (DI) and the
high frequency E* values of the constructed asphalt FCbottom at reliability level of 95% that were calcu-
mixes are much higher compared to those of the lated by taking into consideration the E*mix_lab,
design mixes. Table 6 presents the RMSE (Root E*mix-pred and E*core values are presented for
Mean Square Error) values expressed in percentage. sections A and B, respectively.

803
Figure 3. E* master curves and induced tensile strains at the bottom of the asphalt layers—Section A.

Table 6. RMSE (%) of E* values. According to the Figures 4 and 5, the deviations
that were observed between the E* values of the
RMSE (%) design mixes and those of the constructed ones,
have an impact in the performance indicator which
A1 A2 A3 A4 A5
is the FCbottom. Regarding Section A the FCbottom
E*mix_lab—E*mix_pred 5.1 13.3 13.4 5.4 4.9 estimated at the end of the design period taking
E*mix_lab—E*core 43.7 33.5 27.7 35.5 41 into consideration the E*core values is higher
E*mix_pred—E*core 41.6 26.3 20.6 37 38.8 compared to the FCbottom estimated by considering
the E* values of the designed mix (E*mix_lab and
RMSE (%) E*mix_pred).
On the contrary, in Section B deviations in the
B1 B2 B3 B4 B5 estimated FCbottom are smaller, while in some cases
FCbottom may be equal or even smaller. The final
E*mix_lab—E*mix_pred 9.4 8.2 8.7 8.8 9.5
acceptance of the in situ mixes can be assessed
E*mix_lab—E*core 14.2 21.7 21.3 26.2 22.7
based on the desired upper limit of the FCbottom
E*mix_pred—E*core 17.9 19.8 26.2 27.8 23.1
value at the end of the design period. Besides,

804
 • DI(%) • FCbuttum(%) a performance indicator in the frame of the com-
16.0
mon practice design and quality control proce-
14.0
dures, with the aim to ensure the quality of both
12.0 HMA and pavement deign.
"
'E IO.O

i
ii 8.0
REFERENCES
~ 6.0
Q
4.0
AASHTO. 2008. Mechanistic-Empirical Pavement Design
Guide, Interim Edition: A Manual of Practice. Ameri-

• can Association of State and Highway Transportation

2.0

0.0
Officials, Washington, D.C.
AASHTO T342-11. 2011. Standard Test method for the
Determining Dynamic Modulus of Hot Mix Asphalt
Figure 4. DI and FCbottom, Section A. (HMA). American Association of State Highway
and Transportation Officials, Washington, D.C.
Bennert T. and Williams S.G. 2009. Precision of
• D I {% ) • FCbottom(%)
AASHTO TP62-07 for use in mechanistic—empirical
5.0 pavement design guide for flexible pavements. Trans-
4 .5 +-- - - - - ---t-- - - - - - - - - - - - - - portation Research Record: Journal of the Transporta-
4.0
tion Research Board, No 2127, Washington, DC, pp.
115–126.
~ 3.5
Bonaquist R., Christensen D.W. and Stump W. 2003.
i'o +-------- NCHRP Report 513. Simple Performance Tester of
~ 2.5
Superpave Mix Design: First-Article Development and
i:'·o+-- - - - - - Evaluation. Transportation Research Board of the
Ct.s H · - -- - - National Academies, Washington, DC.
1.0 Cominsky R.J., Killingsworth B.M., Anderson R.M.,
0.5
Anderson D.A. and Crockford W.W. 1998. Quality
0 .0
Control and Acceptance of Superpave Designed Hot
Mix Asphalt. Project D9-7, NCHRP Report No. 409,
Transportation Research Board, Washington, DC.
Figure 5. DI and FCbottom, Section B. EN 12697-31. 2004. Bituminous mixtures—Test methods
for hot mix asphalt—Part 31: Specimen preparation by
gyratory compactor. European Standard.
these analysis results can also be a useful tool for Georgouli K, Loizos A, Plati C. 2016. Calibration of
Dynamic Modulus Predictive model. Construction
the timely planning of maintenance activities. and Building Materials 102: 65–75. doi:10.1016/j.
conbuildmat.2015.10.163
Georgouli K. 2016. Development of prediction algorithms
4 CONCLUSIONS for the estimation of asphalt mix dynamic modulus.
PhD Thesis, National Technical University of Athens.
Overall, the results presented in this study showed Huhtala M, Pohlajamaki J. 1992. New Concepts on Load
that E* is a critical parameter for the assessment Equivalency Measurements. In Proceedings of 7th
of the HMA quality control procedures. HMA E* International Conference on Asphalt Pavements. Not-
provides useful information of the HMA behavior tingham, UK.
NCHRP. 2004. Final Report: Guide for Mechanis-
under a wide range of temperature and frequency tic-Empirical Design of New and Rehabilitated
conditions. Pavement Structures, NCHRP 1–37A Project,
The laboratory determination of the E* is Transportation Research Board National Council,
time consuming and demands specialized equip- Washington D.C.
ment. However, through the newly developed Siddhartan R, Krishnamenon N, Sebaaly P. 2000.
E*algorithm that was presented in this study, an Pavement Response Evaluation Using Finite-Layer
estimation of the HMA behavior can be assessed. Approach. Transportation Research Record: Jour-
The required parameters are aggregate grada- nal of The Transportation Research Board, No 1709,
tion, HMA volumetric composition and basic test Washington, DC, pp 43–49.
Ulloa, A., Hajj, E. Y., Siddharthan, R. V., and P. E.,
results on asphalt binder. All these parameters are Sebaaly. 2013. Equivalent Loading Frequencies for
determined as part of the common practice asphalt Dynamic Analysis of Asphalt Pavements. Journal of
and pavement design procedures. Materials in Civil Engineering, ASCE, Vol. 25, No. 9,
Therefore, E* prediction algorithm can be September 1, 2013, pp. 1162–1170, DOI: 10.1061/
utilized for the introduction of the HMA E* as (ASCE)MT.1943-5533.0000662.

805
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Implementation of modulus-based quality control testing for pavement

construction in Thailand

A. Sawangsuriya & P. Jitareekul

Department of Highways, Bangkok, Thailand

Y. Taesiri, S. Sirisak & W. Lawanwisut

IMMS Co. Ltd., Bangkok, Thailand

ABSTRACT: This paper presents a series of modulus-based quality control tests that were conducted
in a field trial section as part of highway construction project in Thailand. The compaction quality of
compacted earth fill embankment was evaluated using the conventional moisture-density approach
adopted by Thailand Department of Highways in companion with a series of alternative quality control
tests including the Soil Density Gauge (SDG), the Light Weight Deflectometer (LWD), the Soil Stiffness
Gauge (SSG), and the Dynamic Cone Penetrometer (DCP). The densities and moistures determined
from the SDG were compared with those determined from the conventional approach. Three alternative
modulus-based quality control tests e.g. LWD, SSG, and DCP were conducted in companion with the
moisture-density tests. The effects of compaction, different devices, and moisture conditions on modulus-
based quality control tests were discussed when interpreting the results. Finally, the proposed criteria
for implementing modulus-based and compaction quality control testing for pavement construction in
Thailand were recommended.

1 INTRODUCTION This study presents a series of modulus-based

quality control tests that were conducted in a
Thailand Department of Highways (DOH) has field trail section as part of highway construc-
adopted the sand replacement method for a statis- tion project in Thailand. Three alternative mod-
tical evaluation of compaction quality for several ulus-based quality control tests e.g. LWD, SSG,
decades. However, such method is generally time and DCP were conducted in companion with the
consuming, labor intensive, less cost-effective, and moisture-density tests. The effects of compaction,
considered destructive. A simple, rapid, and direct different devices, and moisture conditions on mod-
structural property testing in conjunction with ulus-based quality control tests were discussed.
moisture-density testing which can be conducted Finally, the proposed criteria for implementing
independently and safely by the inspector without modulus-based and compaction quality control
interference with the construction process is antici- testing for pavement construction in Thailand
pated to increase test coverage, to improve statisti- were recommended.
cal basis of evaluation, and to reduce variability.
Alternative quality control tests for in-place
assessment of mechanical properties and quan- 2 MODULUS-BASED QUALITY CONTROL
titative evaluations of construction practices TESTING
and materials such as Light Weight Deflectom-
eter (LWD), Soil Stiffness Gauge (SSG), Dynamic 2.1 Light Weight Deflectometer
Cone Penetrometer (DCP) are currently available
in the transportation geotechnics and pavement A Light Weight Deflectometer (LWD) is a portable
engineering communities (Meehan et al. 2012, device used to measure the in situ modulus of defor-
Umashankar et al. 2016). They provide in-place mation of pavement layers. The LWD used in this
stiffness and strength assessment of pavement study is the Prima 100 (Figure 1a). It consists of a
materials and gain popularity over time as alter- 10 kg drop mass with a drop height ranged from 0.1
native means for Thailand highway construction to 0.9 m. The impact force imparted by the drop
quality control (Sawangsuriya et al. 2009, Taesiri mass is buffered by elastomeric pads. The applied
et al. 2009, Wachiraporn et al. 2010, 2013, Sawang- force is measured by a load cell while the vertical
suriya et al. 2016). surface deflection is measured by a geophone in

807
 measurements are made in accordance with ASTM
D6758.

2.3 Dynamic Cone Penetrometer

Dynamic Cone Penetrometer (DCP) (Figure 1c) is
relatively simple, rugged, and economical device that
can provide the in situ strength index of pavement
(a)LWD (c) DCP
layers. The DCP is used for measuring the mate-
(b) SSG
rial resistance to penetration while the cone of the
Figure 1. Modulus-based quality control tests. device is being driven into the pavement structure.
The number of blows during operation is recorded
with depth of penetration. The slope of the rela-
tionship between number of blows and depth of
direct contact with the ground through varied load penetration (in millimetres per blow) at a given
plates of 100, 200, and 300 mm in diameter. The linear depth segment is recorded as DCP penetra-
Prima LWD software uses the stress distribution tion index (DPI), which can be used to estimate the
factor (f) of 2, which implies a uniform plate- shear strength characteristics of pavement materials
ground contact stress distribution (e.g. flexible in accordance with ASTM D6951 (Sawangsuriya &
plate) and Poisson’s ratio (ν) of the soil to esti- Edil 2005). The measured DPI from the DCP can
mate elastic modulus (ELWD) of pavement materials be used to calculate the CBR and the corresponding
(Stamp & Mooney 2013). elastic modulus (EDCP) of pavement materials (Web-
ster et al. 1992, Powell et al. 1984).
f r −ν 2)
r(
WD =
o
E LW (1)
so log(CBR)) .46
. 1. log( ) (3)
where σo is the peak applied stress, so is the peak E DCCP 1 6C
CBR 0.64
(4)
vertical deflection, and r is the radius of the load
plate. The Prima LWD tests are made in accord-
ance with ASTM E2583.
3 MATERIAL AND FIELD TESTING
SCHEME
2.2 Soil stiffness gauge
Soil Stiffness Gauge (SSG), which is currently mar- 3.1 Material properties and classification
keted as the Humboldt GeoGaugeTM (Figure 1b), Samples were collected near the shoulder of the
is a portable, non-nuclear testing device that pro- roadway to determine the index properties, soil
vides simple and non-destructive means of directly classification, and compaction characteristics.
and rapidly measuring in situ soil stiffness. The Table 1 summarizes the properties of earth fill
SSG weighs about 11.4 kg, is 28 cm in diameter,
25.4 cm in height, and rests on the soil surface via a
ring-shaped foot. The SSG measures near-surface
stiffness by imparting small dynamic force to the Table 1. Material properties and classification.
soil though a ring-shaped foot at 25 steady state
frequencies between 100 and 196 Hz. Based upon Passing
the force and displacement-time history, stiffness 50.0 –
is calculated internally as the average force per 25.0 –
unit displacement over the measured frequencies 19.0 –
and reported. A measurement takes only about 9.5 100.0
1.5 minutes. The measured soil stiffness from No. 4 90.4
the SSG (KSSG) can be used to calculate the elas- No. 10 54.7
tic modulus (ESSG) of pavement materials at near No. 40 37.2
surface (Sawangsuriya & Edil 2005). Knowing the No. 200 25.5
measured KSSG, the ESSG can be calculated as: LL 37.3
PI 12.7
K SSG ( − 2 )
ESSG = (2) OMC (%) 8.4
1.77 ⋅ R γdmax (g/ml) 2.16
CBR (%) 14.3
where R is the outside radius of the ring-shaped Soil Classification A-26
foot of the SSG (∼57.2 mm). The SSG stiffness

808
embankment and its classification. The compaction 4 RESULTS AND DISCUSSION
test was conducted in accordance with the standard
compaction effort described in ASTM D698. 4.1 Modulus-based and compaction quality
control test results
3.2 Field testing scheme Three alternative modulus-based quality control
tests e.g. LWD, SSG, and DCP were found to be
A series of modulus-based and compaction quality
simple and can be operated by at most two persons.
control tests were conducted in a field trial section
Both LWD and SSG devices are found to provide
as part of highway construction project in Thai-
quick test results on elastic modulus of compacted
land. The compaction quality of compacted earth
earth fill embankment of the field trial section.
fill embankment was evaluated using the conven-
Results of modulus-based and compaction quality
tional moisture-density approach adopted by Thai-
control tests are illustrated in Figure 3. The Prima
land Department of Highways in companion with
the non-nuclear Soil Density Gauge (SDG 200)
by Transtech Systems, Inc, conformed to ASTM Une 1

D7830. The field trial section was located at the t.MO

. lWDH~~rnm

new highway route: Pattaya-Map Ta Phut motor- " Q lWDH500mm

1,000 ·o
way, the Department of Highways, Thailand. ~ 2( LW0Hl~Omrn

Figure 2 illustrates a typical layout of field •

trial section, which had a total length of about
500
r • e •"'
lij
"' 1t. SSG

100 m (from STA 6 + 450 to STA 6 + 550) with f ...',. ...,. ,..,.,.""'
l l

about 7.0 m width. After the compaction process, 6 +4!10 6+H6

STA
the LWD, SSG, DCP, and SDG were conducted
instantaneously at ten test locations along two (a) Modulus tests along line 1
measurement lines (e.g. lines 1 and 3) as shown
Line 3
in Figure 2. The sand replacement tests (ASTM . lWOH~OrTm

.:
1.~0 -r······················
D1556) were conducted at five test locations along Q LWDHSOOI'l'm
the centerline (e.g. line 2) of the section. At most
.!~
1.000

five SSG readings were made first, followed by ~ "0 D

"
.. ?A LWOHI501M1

six LWD readings, five SDG readings, and a sin-

gle DCP measurement, respectively per one test iU
500

•• .• 0

••" i
o~.SSO

•""'
location in each measurement line. The moisture 6 +4~ 6 +4 7~ 6~00 6~26 6+:.!10
samples were also taken from those test locations STA

in order to determine the moisture content in soil (b) Modulus tests along line 3
by oven drying method. Every measurement was
carefully made at the adjacent location to avoid 112

disturbances. In addition to these tests, a single ~

110 1
~ • SOG(l1net)

sand replacement method was performed per one g t •

"'' .lt. 5nd"•plat~nwni(Line2)

test location in a separated measurement line.

It should be noted that the LWD characteristic
! "''
1D• l •
• SDG(l•ne3)

used in this study consisted of the drop mass of j 102

approximately 10 kg with varied 250, 500, 750-mm 100

6 +415 6+525
drop height for the 200 mm plate diameter. In 6 +450 6+600 6+550

STA
addition, the weighted average DPI was calculated
over 150-mm DCP penetration depth, which was (c) Relative compaction along lines 1-3
considered as the most representative depth for
modulus comparison purpose. . 0

l
~
~
. e
g il
• SDG(Lmt1)

Q Ov..,..OntdS.-..pie(l..,~

• SOG(llne3)
IJ

O...en-DnedS;Imj)IC(l.ot~ClJ

-~ '• D

l
~

t · --------· -------· --------· --_!l"-L_e---

8------r 6 ---------!:s---------8---_!l"'-L__6 ---
·----~----------·--------·--_!l'!_L_____
'1
6+450 6+475 6+500

STA
6+525 6+550

STA6+550 STA6+450
(d) Moisture content measurements along lines I and 3
• lWU, SSG.OCP,SOG,M~stureSample

Figure 3. Modulus-based and compaction quality con-

Figure 2. Typical layout of field trial section. trol test results.

809
LWD tended to provide the highest modulus, while
the SSG registered the smallest modulus as shown
in Figures 3a and 3b. The elastic modulus of com-
pacted earth fill embankment was ranged from
700 +-~~~~~~~~r-~~~~~~~~~~~~
approximately 150 to 1,400 MPa. The ELWD values i Goo ~
r-~~~~~~~~~."
~~~~~~~~~~---

were found to vary from 220 MPa to 1,400 MPa ~500 ~48~3~M~PaL_~~~~~~~~~~~~~~~~~---
~ ---- -- ----- -- --- ---- ---- ----------~

along test locations and measurement lines. The i400 +-~~~~~~~~~~~!~~~~

~~~~~~~
ESSG values were closed to the EDCP values. They
~ 300 +-~~~~~~~~~~-~~~~~~-=~~~

varied between 150 and 350 MPa. y=32.26x· 0· 99

R2 = 1.00
Results of relative compaction (e.g. the ratio
between dry unit weight, γd and maximum dry unit
weight, γdmax) and moisture content measurements 0.065mm:

were found to vary much larger along line 1 than

line 3 as shown in Figures 3c and 3d. Considering Deflection (mm)

all the modulus-based quality control tests con-

ducted on compacted earth fill embankment, the Figure 5. Elastic modulus estimated from the FEM.
LWD tests at varied drop heights tended to capture
those variations in the compaction quality along
line 1. Comparing between lines 1 and 3, the DCP form solution were performed to validate the LWD
tests also exhibited such variation. The modu- measurements and compared in Figure 4. In order
lus variations between lines 1 and 3 are possibly to determine elastic modulus from the Finite-
caused by the effects of moisture condition and Element Model (FEM), the input applied stress
density state imparted during compaction process. must be specified. In this study, the applied stress
These findings confirmed the fact that, although was averaged from the measured LWD stress
the conventional approach of moisture-density induced by a 10 kg drop mass with a 500-mm drop
control satisfied a compaction quality control height for the 200 mm plate diameter, which was
requirement based on density testing, large vari- equal to 178.9 kPa. According to the measured
ability in soil texture, microstructure, and thus its vertical deflection of approximately 0.065 mm, the
modulus was still observed. elastic modulus from the FEM was then estimated
from such applied stress (Figure 5). Similar input
parameters were used for the Boussinesq solution.
4.2 Comparison between modulus-based quality
Both moduli obtained from the FEM and Boussi-
control tests
nesq solution were identical and equal to 483 MPa.
Average moduli obtained from the LWD, SSG,
and DCP were compared as shown in Figure 4.
The average moduli from the LWD were calculated 5 CONCLUSIONS AND
based on a single 500-mm drop height only because RECOMMENDATIONS
it represented the mid-height for the reference value.
The average moduli were 184, 256, and 555 MPa A series of modulus-based and compaction qual-
for the SSG, DCP, and LWD, respectively. The 2D ity control tests were conducted in a field trial sec-
finite-element analysis and the Boussinesq closed- tion as part of highway construction project in
Thailand. In this study, the compaction quality
of compacted earth fill embankment based on the
conventional moisture-density approach adopted
by Thailand Department of Highways (DOH) and
the non-nuclear Soil Density Gauge (SDG) was
evaluated and compared with a series of alterna-
~ LWD
tive modulus-based quality control tests including
~ SSG
the Light Weight Deflectometer (LWD), the Soil
1111 DCP
Stiffness Gauge (SSG), and the Dynamic Cone
[JFEM
Penetrometer (DCP). Three alternative modulus-
• BoussinesqSolution
based quality control tests e.g. LWD, SSG, and
DCP were conducted in companion with the mois-
ture-density tests by sand replacement method,
oven drying method, and SDG. The effects of
compaction, different devices, and moisture condi-
Figure 4. Comparison between modulus-based quality tions on modulus-based quality control tests were
control tests. observed herein.

810
 The comparison between modulus-based Sawangsuriya, A., Sramoon, W. & Wachiraporn, S. 2009.
and compaction quality control tests suggested Innovative tools for highway construction quality con-
that conventional moisture-density testing can- trol. Research and Development Journal of the Engi-
not describe variability in the modulus of the neering Institute of Thailand, Vol. 20, No. 3, pp. 36–42.
Sawangsuriya, A., Wachiraporn, S. & Sramoon, W.
compacted earth fill embankment. Since the 2016. Proposed performance criteria for earthwork
non-uniformity of modulus is directly related to construction quality control. 5th International Con-
progressive failures and life-cycle cost, a simple, ference on Geotechnical and Geophysical Site Char-
rapid, and direct modulus testing which can be acterisation, ISC’5, Queensland, Australia.
conducted independently and in conjunction with Stamp, D.H. & Mooney, M.A. 2013. Influence of light-
conventional moisture-density testing without weight deflectometer characteristics on deflection
interference with the construction process is nec- measurement. Geotechnical Testing Journal, ASTM,
essary. Such modulus-based test is anticipated to Vol. 36, No. 2, pp. 216–226.
increase test coverage, to improve statistical evalu- Taesiri, Y., Sawangsuriya, A., Wachiraporn, S. & Sra-
moon, W. 2009. Assessment of in-situ tests for stiff-
ation, and to reduce variability, thus substantially ness and strength characteristic of pavement materials.
enhance construction quality control of the entire 13th REAAA Conference Korea 2009, Icheon, South
earthwork. Finally, the proposed criteria for earth Korea.
fill embankment and pavement construction qual- Umashankar, B., Hariprasad, C., & Kumar, G.T. 2016.
ity control testing in Thailand shall integrate both Compaction quality control of pavement layers using
modulus-based and compaction quality control LWD. Journal of Material in Civil Engineering,
tests in parallel with the development of future Vol. 28, No. 2.
performance specifications. Wachiraporn, S., Sawangsuriya, A. & Sramoon, W. 2013.
Mechanistic based quality control during highway
construction. Proceedings of the 18th Southeast Asian
Geotechnical Conference, Singapore, pp. 605–610.
REFERENCES Wachiraporn, S., Sawangsuriya, A., Sunitsakul, J., & Sra-
moon, W. 2010. Stiffness and strength based in-place
Meehan, C.L., Tehrani, F.S., & Vahedifard, F. 2012. A evaluation of compacted unbound materials. Proceed-
comparison of density-based and modulus-based in situ ings of Sessions of GeoShanghai 2010, Geotechni-
test measurements for compaction control, Geotechni- cal Special Publication No. 203, Shanghai, China,
cal Testing Journal, ASTM, Vol. 35, No. 3, pp. 387–399. pp. 347–354.
Powell, W.D., Potter, J.F., Mayhew, H.C. & Nunn, M.E. Webster, S.L., Grau, R.H., & Williams, T.P. 1992.
1984. The structural design of bituminous roads. TRRL Description and application of dual mass dynamic
Laboratory Report 1132, Transportation and Road cone penetrometer. Instruction Report GL-92–3, U.S.
Research Laboratory, Crowthorne, Berkshire, 62 pp. Army Engineers Waterways Experimental Station,
Sawangsuriya, A. & Edil, T.B. 2005. Evaluating stiffness Vicksburg, MS, 17 pp.
and strength of pavement materials. Proceedings of
the Institution of Civil Engineers-Geotechnical Engi-
neering, Vol. 158, No. 4, pp. 217–230.

811
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

A practical non-destructive testing based approach to improve the

quality of the asphalt compaction process

P. Georgiou & A. Loizos

Laboratory of Pavement Engineering, National Technical University of Athens, Athens, Greece

ABSTRACT: Asphalt mixture compaction has been well recognized as one of the most critical factors
affecting the pavement behavior. This implies that effective compaction strategies should be established to
achieve the desired in-situ compaction degree. In practice, contractors seek to establish the rolling pattern
that ensures the quality acceptance specifications are met through the installation of a trial section, with-
out taking into consideration dominant factors affecting compact ability, such as mixture temperature
and the variability of construction conditions. In this study, a practical methodology was developed to
model the compaction density to enhance the asphalt compaction process by integrating a closed formula
and the implementation of emerging Non-Destructive Testing (NDT) technologies. The fitted model
describes the compaction density evolution with respect to mixture temperature cooling and number of
passes; by controlling these two parameters it is expected to become easier to achieve the desired compac-
tion density. The experimental results used for this investigation verified the prediction performance of
this approach.

1 INTRODUCTION the laboratory. However, these studies highlighted

the key parameters affecting asphalt mixture com-
Asphalt mixture compaction has been well recog- pact ability, such as mix temperature; compaction
nized as one of the most critical factors affecting the energy (effort) and methodology; and lift thickness.
pavement behavior, providing all desirable mixture In practice, achieving on-site the target com-
design characteristics are met. Inadequate compac- paction degree is still a heuristic process. Prior to
tion has been associated with poor asphalt pavement commencement of permanent works, road agen-
performance. On the contrary, the good or proper cies specify the installation of a trial section so that
compaction of asphalt mixture increases fatigue evaluation of the contractor’s construction tech-
life, decreases permanent deformation or rutting, niques is possible. Within this process, the contrac-
reduces the amount of oxidation or aging, decreases tors seek, among others, to establish the rolling
moisture damage or stripping, increases strength pattern that ensures the quality acceptance specifi-
and internal stability and optimizes the surface tex- cations are met. To accomplish this goal, contrac-
ture (Crispino et al. 2007, Brown 1990, Linden et al. tors generally try several different combinations
1989, Bell et al. 1989). Therefore, it is vital that the of equipment, and numbers of passes with each
desired in-situ compaction degree is achieved dur- combination, to determine the most effective roll-
ing construction and the Quality Control/Quality ing pattern. In-place density, de-fined by nuclear/
Assurance (QC/QA) specifications are met. This non nuclear density gauges, is the measure of
further implies that effective compaction strate- rolling pattern effectiveness. With this procedure,
gies, appropriate to the site conditions, should be compaction curves relating density and number
established. of passes are developed, which enable to de-fine
Although it would be essential to design in the the required compactive effort to achieve the tar-
laboratory the on-site compaction strategies, sev- get density. Although this procedure concerns a
eral previous studies have demonstrated that the general guide described, in some instances, in the
in-situ compaction characteristics cannot be reli- construction specifications the outcome of this
ably predicted from laboratory testing (Schmitt procedure reflects only the site conditions met dur-
et al. 2009, Leiva and West 2008a, b). This may ing the installation of the trial section. It is worth
be attributed to several factors affecting field mentioning that this procedure lacks dominant
compaction such as environmental, lay down site factors affecting compact ability, such as mixture
and confinement conditions, compaction equip- temperature. In this respect, for the scenario that
ment and others, which cannot be simulated in the mixture temperature after placement departs

813
from its relevant value during the installation of performance. For this purpose, experimental
the trial section, it is likely that a different number compaction data were collected. The results were
of roller passes to achieve the target density will then analyzed, discussed and a methodological
be needed, which however cannot be assessed. This approach to improve the quality of the asphalt
implies that a more effective methodology shall be compaction process is proposed.
followed to establish the rolling pattern, thus mak-
ing due allowance for the variability in construc-
tion conditions. 2 EXPERIMENTAL INVESTIGATION
In recent years, NDT technologies have emerged
seeking to improve compaction quality and proc- 2.1 Test site and materials
ess by measuring asphalt pavement characteristics
The experimental data used for this investigation
during compaction. For instance, the Intelligent
was obtained during the construction of a full-
Compaction (IC) technology has recently been
scale test section, which involved the paving of a
implemented in the USA and Japan. IC produces
50 mm thick new asphalt overlay surface course,
massive amounts of geospatial data with 100%
namely the AC12.5 surf 50/70, over the existing
coverage of the compacted area in real-time, which
flexible pavement structure, as shown in Figure 1.
needs to be effectively analyzed and managed for
The test section was divided into two subsections
quality control and acceptance applications for
subjected to different compaction modes, namely
HMA pavement (Xu et al. 2012, Commuri et al.
vibratory and a combination of static-vibratory
2011, Furuya et al. 2010). Although this methodol-
mode using a 10-ton steel double-drum roller. To
ogy offers significant benefits, inconsistent trends
fulfill the aim of this study, five spots were located
found between IC measurements and in-situ den-
across the mat width, at a fixed distance of 0.8 m,
sity indicate that further research is needed for
and measured immediately after finisher paving
this technology to become mature for widespread
and successively after each roller pass (a pass cor-
implementation.
responds to one forward or backward motion).
Also, the Infrared Thermography (IRT) tech-
Monitoring of roller operations was conducted
nique has been implemented for the evaluation of
by means of non-nuclear density and Infrared
pavement construction operations by means of
Thermography testing given their practicability.
temperature profiling of the pavement mat surface
Although surface temperature is generally lower
(Stroup-Gardiner et al. 2004). IRT allows continu-
than the in-asphalt temperature, it can be accepted
ous real-time data collection during both the lay
as a reliable indicator of mat asphalt temperature
down and compaction process, which provide use-
(Delgadillo & Bahia 2008). In that respect, the
ful insight for the compaction quality (ter Huerne
infrared-based temperature was used for the pur-
et al. 2009, Elsafey et al. 2009). It has been dem-
pose of this study analyses.
onstrated that temperature differentials in the mat
during HMA paving can result in substantial den-
sity differentials (Rahman et al. 2013, Henault & 2.2 Infrared thermography
Larsen 2006, Willoughby et al. 2002). Research
Infrared thermography makes use of the infrared
in this area has emphasized the identification of
spectral band and Planck’s law, which claims that
cooler areas in the mat and its relationship with
all objects with temperatures above absolute zero
final HMA density. Evaluation of the final density
(−273.15°C or −459.67°F) emit infrared radiation.
has also indicated that areas experiencing over-
all large mat temperature differentials exhibited
higher air void contents when compared to areas
with more normal temperature differentials. Con-
sidering the above, asphalt mix density-growth
curves, with respect to the IRT-based mat tem-
perature evolution, need to be established for the
enhancement of pavement compaction quality.
Based on the above, in this paper it was investi-
gated the potential of modelling the in-situ com-
paction density by integrating the implementation
of emerging NDT technologies for measuring in
real-time key compaction characteristics. On this
basis, a closed formula found in literature which
quantifies the relationship between density and
number of roller passes, taking into consideration
the temperature evolution (cooling) with respect to
roller passes was verified in terms of its prediction Figure 1. The experimental test section.

814
The magnitude of the radiation emitted is heavily
dependent on the objects’ temperature. Based on
this, it is possible to produce images of that radia-
tion and consequently the objects’ temperature,
called thermographs (Arndt 2010).
The systems used to measure and image the
emitted infrared radiation from an object are
based on complex infrared or thermal cameras.
When viewing an object, the camera receives radia-
tion not only from the object itself, but it also col-
lects radiation from the surroundings reflected via
the object surface. Both of these radiations are
affected to some extent by the atmosphere in the
measurement path (Maldague 2001).
On the purposes of this study, the NTUA ther-
mographic system was used. The NTUA ther-
mographic system consists of a Thermo Vision
A325G infrared camera developed by Flir Systems
and a conventional digital camera. The Thermo-
Vision camera provides a 320 × 240 pixel image,
temperature measurements ranging from –20 to Figure 2. PQI gauge.
350°C with an accuracy of ±2°C, operating in the
far infrared spectrum (7.5–13 μm). The system
collects data as scans of temperature, is portable
and is mounted on a properly modified vehicle, as 2.4 Closed-form equation for density prediction
shown in Figure 1. During the experimental inves- As previously demonstrated, temperature influ-
tigation, the surface mat temperature was meas- ences significantly the compaction process (Plati
ured during the laying and compaction of the test et al. 2014). Building on this knowledge, a closed
subsections previously described. The collected formula found in the literature (Huh & Nam 2000)
data were then pre processed and analyzed with which quantifies the relationship between density
suitable software. and number of roller passes, taking into consid-
eration the temperature evolution (cooling) with
2.3 Non-nuclear density method respect to roller passes is verified in terms of its
prediction performance, using the experimental
Non-nuclear electromagnetic density gauges have compaction data. This equation is expressed as
entered the market targeting to replace nuclear follows:
density gauges and the coring process. These non-
nuclear devices use electromagnetic signals to
⎧ ⎡ ⎛ −W ⎞ ⎤ ⎫
measure in-place density (ASTM 2010). The use ρ ρ0 + ( ρ ρ0 ) ⎨1 exp ⎢ −V exp
⎝ T ⎠ ⎥⎦ ⎬⎭
N (1)
of electromagnetic signals has the advantages of ⎩ ⎣
completely eliminating nuclear source licensing,
training and certification of operators, specialized where ρ = mixture density; ρo = initial density value
storage, and risks associated with devices that use obtained with no roller passes; ρm = upper limit of
a radioactive source, while also being non-destruc- density obtained with an infinite number of roller
tive. The most popular of these devices is the passes; T = mixture temperature and N = number
Pavement Quality Indicator (PQI), commercially of roller passes.
available by Trans-Tech Systems Inc. (Fig. 2). The closed-form Equation 1 describes den-
The PQI operation is based on a novel toroidal sity changes between the obtained finisher-based
electrical sensing field that is established in the precompaction ρ0 (with no roller passes) and the
material to be measured (e.g. HMA) via a flat sens- upper limit obtained with an infinite number of
ing plate. Density, or compaction degree, is meas- roller passes ρm as a function of both roller passes
ured by the response of the PQI’s electrical sensing and temperature. Note that solidification by cool-
field to changes in electrical impedance of the ing is represented by a temperature-dependent fac-
material matrix, which in turn is a function of the tor [exp(−W/T)] and that consolidation by rolling
composite dielectric constant of the paving mate- compaction is denoted by a temperature independ-
rial and the air trapped in the voids of the mate- ent factor V. This equation can be simplified insert-
rial. Once calibrated, respectively to the nuclear ing one more relationship expressing the cooling of
method, direct density readings can be consistently mix temperature with respect to roller passes, for
obtained. example, in terms of a quadratic form. Then, the

815
compaction density can be expressed by a single 2400

variable of roller passes. •

2350
"'• • ..
3 RESULTS
..
~E 2300
.:.:
~ 2250
liE

"' •
..
liE
"'* • .
0.6
1.4
2. 2
3.0
3.8

Figure 3 shows the effect of mixture cooling on li

Q
2200 ' •
compaction density evolution with respect to the 2 150
field roller passes for the static-vibratory com-
paction mode. The counterpart results related to 2100
0 10
the vibratory compaction mode are illustrated in
140
Figure 4.
120
As shown in Figures 3–4, the compaction den-
sity of paving layer increases rapidly at the begin- E 100
..
ning. With the increase of rolling passes and ~ • 0.6

mixture cooling, the compaction density increases

slowly. Generally, when the mat surface tempera-
3
~
c.
80
60
~
"' ~ • •
• . 1.4
2.2
3.0
E 40
ture falls below 80°C it seems that additional com- "'
1- 3 .8
20
paction causes insignificant mat density increase.
The experimental compaction data, in terms of 0
0 10
the average values of the five test spots, were then Number of passes
fit into the closed-form (Eq. 1) to obtain the model
parameters related to the mixture material prop- Figure 4. Effect of mixture cooling on compaction den-
erties. Figures 5–6 show the predicted compaction sity for vibratory compaction mode.
data for each subsection, respectively.
From these figures, it is demonstrated a good 2400 r-------------------------------------.
fit to the experimental data. Moreover, the high • Measured data
R2 values, of the order of 0.9, as well as the low 23 50 Predicted data

values of Mean Absolute Error (MAE), ranging

..
~~ 2300 .t -··· ··--:
from 8.3–11.1 kg/m3 and Mean Absolute Percent-
age Error (MAPE), of the order of 0.03%, indicate
that the Equation 1 can successfully predict den-
sity evolution with respect to roller passes.
.:<
:;:: 2250
·;;;
Q
~ 2200
..• •....,..-·····
p- 2186+ 269{ 1-exp[-0. 158 exp((-33.929)fr] N}
2150 R'~o.85 I
2400
2100
2350 0

..
~ 2300
e *I ••
• I
..•
140
120
•
IT-0.695N2-~47N+ I l9.88 l
I I
..
0.6 R - 0.983
"" 2250
;
·;;;
liE

.. •
. 1.4
2.2
E
~
.::
100

80
----· ----.---.·-----· ----·----
"'"
2200 3.0
Q
3.8 i: 60
2150 "c.E
40
2100 ~
0 2 20

140 0
0
120 ' 'umber of passes
E •...
.::"'
~
100

80
• •. I Figure 5. Fitting of experimental compaction data
~
Q.
E
60
40
. .
0.6
1.4
2.2
(static-vibratory mode).

"'
1-
20 3.0 Also, Figure 7 summarizes the compaction curves
3.8 related to each subsection and compaction mode.
0
0 The results indicate that the vibratory compaction
Number of passes mode is more efficient compared to the static-vibra-
tory mode, considering that for lower compaction
Figure 3. Effect of mixture cooling on compaction den- temperature range the compaction density achieved
sity for static-vibratory compaction mode. is superior for the same number of roller passes.

816
 ~ 2300
c.
2350
• Measured data

... ...·
·
.
...
Predicted data
. . . .
2400 r----------------------------------,

······•···
Trial section
Formulation of dcnsih

1:-:
•
grmYth YS mixhue coOling

·•··•··•
• !9-··T·
- ~, ' '~:....:..·
Dctcrmiuatiou of required

""'~ 2250 -~---~-~·

compaction etlOrt to compl~
Infrared
ca me ra f ::: •• wilh dcnsity specifi cations

5 2200
Q

2 150
p:o2 197+ 258(1-exp[-0.638 exp((-101.822)ff)] N}
R 2=0.934
I;I.~:~ ·-. :. -
-~-·

2100
0 10

140

120 IT=-0.205N: · I.442N+86.87S I

E 100
I R·- 0.754 I
Figure 8. NDT-based methodology to establish the
• -- ·· ·-- · ---· ---. -- - ~ - -.! _
"
~
80 rolling pattern during installation of the trial section.
~ Q.

E 40
60
-.
~
20
ing may be further enhanced commencing rolling at
0
0 10 a) typical and b) the lowest allowable mixture tem-
Number of passes perature on delivery at site, which will thus enable to
define necessary modifications to be conducted due
Figure 6. Fitting of experimental compaction data to the diversity of mixture temperatures on delivery
(vibratory mode). at site. Hence, the proposed methodology is superior
to the procedure commonly followed in practice by
offering the potential to capture the density growth
2400 in respect to the temperature drop curve, which is
2350 crucial to define an effective rolling pattern.
The measured in-situ data may be further utilized
~ 2300 to describe mathematically the compaction process.
c.
""'
; 2250
A closed formula, as described in section 2.4 and
verified with our experimental data, was found to
~ 2200 very effective for such analysis. The fitted model
-+- Static-Vibratory
2150 ..._ Vibratory describes the compaction density evolution with
respect to mixture temperature cooling and number
21oo L---------------------------------~
0 of passes; by controlling these two parameters it is
Number of passes expected to become easier to achieve the desired
compaction density. This comprehensive approach,
Figure 7. Effect of compaction mode on asphalt mix- which offers the possibility to make due allowance
ture compactability. for the variability of construction conditions, ena-
bles to establish not only the most effective com-
paction strategy but also the required compaction
4 PROPOSED NDT-BASED APPROACH TO effort to ensure compliance with density speci-
IMPROVE THE ASPHALT COMPACTION fications. It is therefore believed to offer essential
information for contractors to enhance the asphalt
Based on the above findings, an improved and compaction process.
practical methodology is developed to establish
effectively the rolling pattern and thereby maximize
the benefits from the installation of a trial section. 5 CONCLUSIONS
This methodology, which is shown schematically
in Figure 8, involves, during the installation of the In this paper, it was investigated the potential of
trial section, the implementation of modern NDT modelling the in-situ compaction density towards
techniques, namely Infrared Thermography and improving the quality of asphalt compaction proc-
PQI, and roller pass-by-pass measurements of mat ess. Based on the results of this study, a practical
temperature and in-place density, respectively, for methodology was developed to model the compac-
each compaction strategy to be evaluated. tion density with respect to the roller passes and
With the implementation of these methods a clear mixture temperature evolution, by integrating a
understanding of the impact of asphalt mixture tem- closed formula and the implementation of emerg-
perature on densification is gained. This understand- ing NDT technologies (Infrared thermography

817
and PQI) for measuring in real-time key compac- of Sixth International Conference on Maintenance and
tion characteristics, during the installation of a Rehabilitation of Pavements and Technological Control
trial section. The experimental data used for this MAIREPAV6, Turin, Italy.
investigation verified the prediction performance Furuya, H., Tsukimoto, Y., Koseki, H., Mansell, T., Gal-
livan, V. & Uchiyama, K. 2010. Innovative QC/QA
of this methodology. Thus, it can be stated that compaction method for HMA pavement using Intel-
the mathematical formulation of the compaction ligent Compaction (IC) technology. In: Proceedings of
process on the basis of a closed form equation 11th International Conference on Asphalt Pavements,
which makes due allowance for the variability of Nagoya, Japan.
construction conditions, in lieu of sophisticated Henault, J. & Larsen, D. 2006. Thermal imaging of hot-
but complex computational models (Masad et al. mix asphalt paving projects in Connecticut. Transpor-
2016), may be an effective tool for suggesting the tation Research Record 1946: 130–138.
number of roller passes required to achieve the Huh, J.D. & Nam, Y.K. 2000. Study of compaction den-
desired compaction level. sity in closed form. Transportation Research Record
1712: 66–73.
Considering that the quality of paving and Leiva, F. & West, R. 2008a. Analysis of HMA field
compaction conditions of individual asphalt mix- compactability using the Accumulated Compaction
ture layer has a major impact on the life-time and Pressure (ACP) concept. In: Proceedings of Transpor-
durability of pavements, it is believed that with this tation Research Board 87th Annual Meeting, Wash-
methodology a proper pavement’s initial condition, ington, DC.
according to design criteria, will be ensured. In Leiva, F. & West, R. 2008b. Relationships between labo-
this case, the positive influences through enhanced ratory measured characteristics of HMA and field
field asphalt compaction will therefore be not only compactability. Journal of the Association of Asphalt
a substantial decrease in needs for maintenance Paving Technologists 77: 183–220.
Linden, R., Mahoney, J. & Jackson, N. 1989. Effect of
works and thereby in the total life cycle cost, but compaction on asphalt concrete performance. Trans-
also less disruption to the road users. portation Research Record 1217: 20–28.
Maldague, X. 2001. Theory and practice of infrared technol-
ogy for nondestructive testing. Canada: J. Wiley & Sons.
REFERENCES Masad, E., Scarpas, A., Rajagopal, K.R., Kassem, E.,
Koneru, S. & Kasbergen, C. 2016. Finite element
Arndt, R.W. 2010. Square pulse thermography in fre- modelling of field compaction of hot mix asphalt.
quency domain as adaptation of pulsed phase ther- Part II: Applications. International Journal of Pave-
mography for qualitative and quantitative applications ment Engineering 17(1): 24–38.
in cultural heritage and civil engineering. Infrared Plati, C., Georgiou, P. & Loizos, A. 2014. Use of infrared
Physics and Technology 53(4): 246–253. thermography for assessing HMA paving and com-
ASTM 2010. Standard test method for density of bitumi- paction. Transportation Research Part C: Emerging
nous paving mixtures in place by the electromagnetic Technologies 46: 192–208.
surface contact methods. ASTM D7113/D7113M, Rahman, M., Grenfell, J., Arulanandam, S. & Ianakiev,
West Conshohocken, Pa. A. 2013. Influence of thermal segregation on asphalt
Bell, C., Hicks, R. & Wilson, J. 1989. Effect of percent pavement compaction. Transportation Research
compaction on asphalt mixture life. ASTM, STP829 Record 2347: 71–78.
Placement and Compaction of Asphalt Mixtures, Schmitt, R., Johnson, C., Bahia, H. & Hanz, A. 2009.
107–130. Effects of temperature and compaction effort on field
Brown, R. 1990. Density of asphalt concrete—How and lab densification of HMA. Journal of the Associa-
much is needed? NCAT Report No. 90-3, Auburn, tion of Asphalt Paving Technologists 78: 171–208.
Alabama. Stroup-Gardiner, M., Nixon, J. & Das, P. 2004. Auto-
Commuri, S., Mai, A.T. & Zaman, M. 2011. Neural mated temperature profiling during hot-mix asphalt
network–based intelligent compaction analyzer for construction. Transportation Research Record 1900:
estimating compaction quality of hot asphalt mixes. 41–49.
Journal of Construction Engineering and Management Ter Huerne, H.L., Doree, A.G. & Miller, S.R. 2009.
137(9): 634–644. Monitoring hot mix asphalt temperature to improve
Crispino, M., Rampini, R. & Pozzi, M. 2007. An experi- homogeneity and pavement quality. In: Proceedings
mental analysis of the effects of compaction on asphalt of Sixth International Conference on Maintenance and
pavement macrotexture. In: Proceeding of International Rehabilitation of Pavements and Technological Control
Conference on Advanced Characterisation of Pavement MAIREPAV6, Turin, Italy.
and Soil Engineering Materials, Athens, Greece. Willoughby, K., Mahoney, J., Pierce, L., Uhlmeyer, J. &
Delgadillo, R. & Bahia, H. 2008. Effects of temperature Anderson, K. 2002. Construction-related asphalt con-
and pressure on hot mixed asphalt compaction: field crete pavement temperature and density differentials.
and laboratory study. Journal of Materials in Civil Transportation Research Record 1813: 68–76.
Engineering 20(6): 440–448. Xu, Q., Chang, G. & Gallivan, V. 2012. Development of
Elsafey, M., Herry, C., Abd El Halim, O., Goubran, R. & a systematic method for intelligent compaction data
Goodman, S. 2009. Utilization of infrared imaging as analysis and management. Construction and Building
a tool of hot and warm asphalt mixes. In: Proceedings Materials 37: 470–480.

818
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Predicting roller-compacted concrete properties from

mixture proportions

J. LaHucik
Airfield Pavement and Materials Engineer, Tigerbrain Engineering, Inc., University of Illinois at
Urbana-Champaign, Urbana, IL, USA

J. Roesler
University of Illinois at Urbana-Champaign, Urbana, IL, USA

ABSTRACT: Modern Roller-Compacted Concrete (RCC) pavements are being increasingly specified
in the U.S. because of their construction expediency, material cost savings, and early opening to traffic.
RCC mixture proportioning and adjustments are not as well understood as conventional concrete mix-
tures. The objective of this study is to provide guidance in selecting the initial RCC mixture proportions
and adjustments as needed to achieve target properties. A statistically-based experimental design was
developed to determine the combined effect of three independent, mixture variables on RCC properties:
cementitious content, sand percentage, and fly ash dosage. Predictive response equations and contour
plots were derived to predict moisture-density and strength based on the RCC mixture proportions and
constituents. Significant increases in sand percentage in the RCC mixture was found to have a negative
effect on density and strength of RCC.

1 INTRODUCTION tious content, and fly ash dosage. The influence of

these three variables individually on RCC proper-
Use of Roller-Compacted Concrete (RCC) pave- ties has been previously studied in the literature,
ments in the U.S. has increased significantly in which will be briefly summarized next. However,
recent years because of its economic and construc- their interactions on the RCC properties has not
tion benefits including early opening to traffic, been described adequately in the literature.
rapid construction schedule, potential material
cost savings, and more favorable sustainability rat-
2.1 Percent fine aggregate
ing. The laboratory mixture design phase is crucial
in facilitating a successful construction process, While it is typically specified that the combined
which leads to a well-performing RCC pavement. gradation for RCC pavements follow the maxi-
RCC mixture proportioning is not as commonly mum density line (ACI, 1995; Harrington et al.,
understood by engineers as conventional Port- 2010; ACPA, 2014), limited research regarding the
land Cement Concrete (PCC) and thus, mix con- effects of aggregate gradation on RCC properties
stituents and proportions may need to be adjusted exists. Previous research regarding the relationship
initially to achieve the desired RCC properties for between combined aggregate gradation and RCC
construction and long-term performance. The strength has shown that coarse-to-fine aggregate
objective of this lab study is to supplement existing ratio and strength are positively correlated (Qas-
literature regarding the influence of mixture pro- rawi et al. 2005; LaHucik and Roesler, 2015).
portion adjustments on key RCC properties, such Williams (2013) found that combined gradations
as moisture-density relationship and strength, and following the 0.45 power curve maximum density
to provide guidance to engineers in determining line yielded greater densities than gradations fur-
initial lab mixture proportions for the target RCC ther from the maximum density line but still within
pavement properties. the recommended gradation limits. LaHucik
(2016) found for certain aggregate sources that
gradations following the maximum density line
2 LITERATURE REVIEW did not necessarily produce the highest strength
or fracture properties. This was primarily for RCC
The three variables investigated in this lab study are mixes with low fines content (less than 1% passing
percent fine aggregate (i.e. percent sand), cementi- 0.075 mm).

819
2.2 Cementitious content sive strength) accounting for the three independ-
ent variables and their interactions. The general
As expected, strength has been shown to increase
form of the response model is shown in Equation
with increasing cementitious contents (Amer et al.
1 where βi are regression constants, αC (Equation
2004 and Vahedifard et al. 2010).
2) is the coded coefficient for cementitious content,
αF (Equation 3) is the coded coefficient for fly ash
2.3 Fly ash dosage dosage, and αS (Equation 4) is the coded coefficient
By product Supplementary Cementitious Materials for sand percentage. All coded coefficients (Equa-
(SCM) are commonly employed in concrete pave- tions 2, 3, and 4) are valid in the range of -1.68 to
ments for a variety of reasons including durability 1.68. Outside of this range, the experimental design
improvement, workability, and economics. Fly ash model predictions are not valid. Only those terms
is the most commonly applied SCM and its impact shown to be statistically significant, p-value less
on RCC properties is well documented. Fly ash than 0.05, are included in the response equations.
has been shown to reduce compressive strengths at
earlier ages, less than 28 days but can increase or Response = β1 + β 2α C + β α F + β 4α S + β α C2
decrease at later ages depending on mixture propor- + β 6α F2 + β α S2 + β 8α C ⋅ α S + β 9α C
tions, constituents, and source materials. Similarly, ⋅α F β α S ⋅α F (1)
split tensile and flexural strengths at an age of 28
days were found to either increase or decrease, rela-
tive to RCC without fly ash, depending on the fly
ash content. As expected, Tangtermsirikul et al. αC =
Cementitious Content kg ( m ) − 281.7
3
(2)
26.5
(2004) found that the relative effect of fly ash on
RCC strength is dependent upon chemical compo- l Ash Dosage (
Fly ) −1 .5
sition and shape properties of the fly ash particles. αF = (3)
7.43
Sand Percentage ( ) − 50
3 EXPERIMENTAL DESIGN αS = (4)
5.95
3.1 Design of experiment
A statistics-based experimental design was devel- 3.3 Mixture proportions
oped to determine the effects of the three independ- The mixture proportions for the 20 experi-
ent variables (percent sand, cementitious content, mental design points were calculated using the
and fly ash dosage) on the strength and moisture-
density properties of RCC. The circumscribed Box-
Wilson model (Box and Wilson, 1951) was chosen Table 2. Mixture proportions.
for its ability to handle cross-interactions of three
independent variables. A total of 15 different mix- Type I/II Oven-Dry
Design Class C Fly Cement Aggregate Water
tures with the central point being repeated 6 times, Point Ash (kg/m3) (kg/m3) (kg/m3) (kg/m3)
resulted in a total of 20 experimental design points.
The range of each variable was chosen such that 1 35.2 246.5 2110.0 157.5
it encompasses the values typically found in RCC 2 40.8 285.4 2037.9 147.9
pavement mixture designs as shown in Table 1. 3 50.9 204.4 2139.0 142.4
4 0.0 281.7 2101.0 153.5
3.2 Analysis of responses 5 35.2 246.5 2105.5 148.8
A response surface model was generated for 6 35.2 246.5 2042.7 154.3
each dependent variable (e.g., 28 day compres- 7 12.9 242.3 2046.2 157.3
8 15.6 292.5 2103.4 152.0
Table 1. Test range of independent variables. 9 15.6 292.5 2011.8 151.2
10 50.9 204.4 2103.4 154.9
Independent variable Range tested Points tested 11 61.4 246.7 2050.7 147.2
12 35.2 246.5 2105.6 154.0
Percent sand (%) 40–60 40.00, 44.05, 13 35.2 246.5 2105.6 154.0
50.00, 55.95, 14 29.7 207.6 2153.3 152.8
60.00 15 35.2 246.5 2125.4 144.4
Cementitious content 237.2–326.2 237.2, 255.2,
16 61.4 246.7 2103.2 145.9
(kg/m3) 281.7, 308.1,
326.2 17 70.4 211.3 2090.3 144.9
Fly ash dosage (weight% 0–25 0.00, 5.07, 18 35.2 246.5 2105.6 154.0
of cementitious 12.50, 19 12.9 242.3 2159.0 148.6
materials) 19.93, 25.00 20 35.2 246.5 2105.6 154.0

820
 100 effect of sand percentage and cementitious con-
E
90
tent on MDD and OMC are shown in Figures 2
80
1'
-~ 70
and 3, respectively. The statistical insignificance of
;:_
60
cementitious content on MDD is seen in Figure 2.
1
.!:
50
40
.~
10 30
4.2 Compressive strength
~ 20
Compressive strength testing was conducted at
a 10
three separate ages for all experimental design
0.01 0.1 10 100 points: 1, 7, and 28 days. Compressive strength
Sieve Size (mm)
was tested according to ASTM C39 (2012) on trip-
licate 100 × 200 mm cylinders for each testing age
Figure 1. Combined aggregate gradations (values in
and experimental design point. The response equa-
legend indicate sand percentage).
tions for 1, 7, and 28-day compressive strength, σC,
are shown in Equations 7, 8, and 9 respectively.
circumscribed Box-Wilson model from the values in
The standard errors for Equations 7, 8, and 9 are
Table 1 and Modified Proctor testing according to
3.1 MPa, 5.4 MPa, and 3.3 MPa, respectively.
ASTM D1557 (2012). The mixture proportions are
shown in Table 2; note the fly ash is presented in
kg/m3 rather than percent as shown in Table 1. The 1 Day
a C( ) = 222 9 + 4 03 C 2.63α F (7)

C( ) 45 3 + 5 76 C 4.84α S
combined aggregate gradations for each sand per- 7 y (8)
centage are shown in Figure 1. Since conventional
PCC aggregates were used, all RCC mixtures had 28 − Day
a C ( ) = 57 9 4 93α C − 3.21α S (9)
low fines content in the gradation.
As expected, cementitious content is the main
factor for compressive strength at all three ages. Fly
4 PROPERTY PREDICTIONS ash dosage and 1-day compressive strength were

4.1 Moisture-density relationship

MOD
The moisture-density relationship using Modified (kg/111"3)
< 2320
Proctor compaction (ASTM D1557, 2012) was • 2320 - 2340
determined for each experimental design point. • 2340- 2360
• 2360 - 2380
The resultant parameters are maximum wet den- • 2380 - 2400
sity, Maximum Dry Density (MDD), and Opti- • > 2400

mum Moisture Content (OMC). The MDD and

OMC response equations are given by Equations 5
and 6, respectively.

MDD kg ( m3 )= 2382 − 31.66α S 7 43 2

S + 14.54α Sα F
(5)
o/o Sand
57.0 59.5

Figure 2. Effect of sand percentage and cementitious

OMC ( % ) 6.35 + 0.204α S − 0 115α F − 0.112α Cα S content (12.5% fly ash dosage) on MDD.
(6)

The response equation for MDD, Equation 5,

has a standard error of 12.2 kg/m3 whereas the
response equation for OMC, Equation 6, has a
standard error of 0.1%. Sand percentage is the most
influential independent variable on the moisture-
density relationship parameters. The maximum
dry density is inversely related to the sand percent-
age whereas the optimum moisture content is posi-
tively related to the sand percentage. RCC strength
and durability properties are directly related to
density (Pittman, 1989; Shihata, 2000; Delatte and % Sand
Storey, 2005; Harrington et al. 2010) and therefore,
strength would be expected to decrease as sand Figure 3. Effect of sand percentage and cementitious
percentage increases. Contour plots reflecting the content (12.5% fly ash dosage) on OMC.

821
inversely related as supported by literature (Cao points. Split tensile strength was tested according
et al. 2000; Tangtermsirikul et al. 2004; Atis, 2005; to ASTM C496 (2011) on triplicate 100 × 200 mm
Nili and Zaheri, 2011; Mardani and Ramyar, 2013). cylinders for each testing age and experimental
However, fly ash dosage was not statistically signifi- design point. The response equations for 1 and
cant in the 7 and 28-day compressive strength equa- 7-day split tensile strength, σSP, are shown in Equa-
tions. Figure 4 shows the significant effect of fly ash tions 10 and 11, respectively. The range of 28-day
dosage and cementitious content on 1-day compres- split tensile strengths was insufficient to yield a sta-
sive strength; the range being more than 20 MPa. tistically significant response equation. The stand-
At ages of 7 and 28 days, the cementitious con- ard errors for Equations 10 and 11 are 0.49 MPa
tent and sand percentage variables are the only and 0.38 MPa, respectively.
statistically significant parameters. As noted in
Section 4.1 and shown in the literature (Qasrawi 1 Da
ay SP( ) = 2 82 + 0 37α C (10)
et al. 2005; LaHucik and Roesler, 2015), sand per-
centage and compressive strength were found to be 7 y SP ( ) 3 78 − 0 26α S (11)
inversely related. Contour plots showing the effect
of sand percentage and cementitious content on 7 The key factor affecting 1 and 7-day split tensile
and 28-day compressive strength are shown in Fig- strength changes with concrete age. Cementitious
ures 5 and 6, respectively. The range of compres- content is positively correlated with 1-day split ten-
sive strength shown in Figures 5 and 6 is between sile strength whereas sand percentage is inversely
20 and 30 MPa, demonstrate for these materials correlated with 7-day split tensile strength. Similar
and proportions a significant effect of cementi- to compressive strength, the effect of sand percent-
tious content and sand percentage exist. age is not significant until a concrete age of 7 days.
Contour plots showing the effect of sand percentage
and cementitious content on 1 and 7-day split tensile
4.3 Split tensile strength strength are shown in Figures 7 and 8, respectively. As
Split tensile strength testing was conducted at three shown in Figures 7 and 8, cement content increases
ages (1, 7, and 28 days) for all experimental design resulted in higher 1-day split tensile strength while
increased sand content reduced the 7-day split ten-
sile strength. The range of split tensile strengths in
Compressive Figures 7 and 8 is approximately 1.5 MPa.
Strength
(MPa)
< 15
15 - 20 4.4 Flexural strength
20 - 25
25 - 30
Flexural strength of RCC beams was tested at an age
30 - 35 of 28 days according to ASTM C78 (2010) with trip-
> 35 licate specimens per experimental design point. Beam
specimens, 100 × 100 × 400 mm, were fabricated
according to mold and compaction plate proposed
by LaHucik (2016). The 28-day flexural strength,
MOR, predictive response is given by Equation 12
Cemc nlitious Conle nl with a standard error of 0.43 MPa. Similar to com-
pressive and split tensile strength, sand percentage is
Figure 4. Effect of Fly Ash Dosage and Cementitious inversely related to flexural strength.
Content (50% Sand) on 1-Day Compressive Strength.
Split Tensile
Compressive Strength
Strength (MPa)
(MPa)
< 3.00
< 25
- 30
- 35
40 ,
c
- 45
""c
e
u

Ofo Sa nd
o/o Sand

Figure 5. Effect of Cementitious Content and Sand Figure 6. Effect of Cementitious Content and Sand
Percentage (12.5% Fly Ash Dosage) on 7-Day Compres- Percentage (12.5% Fly Ash Dosage) on 28-Day Com-
sive Strength. pressive Strength.

822
 necessary adjustments of RCC given the mixture
design of RCC differs significantly from that of
conventional PCC. The contour plots and response
equations for MDD and strength are intended to
provide initial estimates of mixture proportions
and to reduce iterations of trial batching in order
to achieve the desired properties.
When early-age strength is a concern for opening
to traffic, several mixture adjustments can be made
such as reducing the fly ash content or increas-
ing cementitious content. Achieving density dur-
ing field compaction is critical to RCC pavement
performance. If difficulties in achieving the speci-
Figure 7. Effect of cementitious content and sand percent- fied density are encountered during construction,
age (12.5% fly ash dosage) on 1-Day Split Tensile Strength. increasing the sand percentage will reduce aggre-
gate interlock as well as the maximum dry density,
increasing the likelihood of achieving the specified
density in the field. However, this increased sand
percentage will result in a strength reduction, which
should be checked against the specified strength.
The contour plots and response equations pre-
sented are based on specific materials used in this
study: Type I/II cement, class C fly ash, dolomitic
limestone coarse aggregates, and natural sand. The
results of this study are only valid within the ranges
shown in Table 1 for each independent variable. The
interaction of the three variables was not signifi-
cant for RCC strength prediction equations. RCC
Figure 8. Effect of cementitious content and sand percent- strength was a linear function of the three independ-
age (12.5% fly ash dosage) on 7-Day Split Tensile Strength. ent variables. With the exception of flexural strength,
there was little contribution of second-order effects
to the RCC strength prediction equations.

6 CONCLUSIONS

There is an increase in interest, application, and

construction of Roller-Compacted Concrete (RCC)
for pavements. The mixture design of RCC is sig-
nificantly different from that of PCC pavements.
Therefore, this study investigated the effects of
sand percentage (40–60%), fly ash dosage (0–25%),
and cementitious content (237.2–326.2 kg/m3) on
RCC strength and moisture-density properties.
Figure 9. Effect of cementitious content and sand per-
centage (12.5% fly ash dosage) on 28-Day Flexural Strength. Based on a 15 point experimental design, two
main variables were found to control RCC prop-
MOR = 6.61 − 0.43α S − 0 28α C2 (12) erties: cementitious content and sand percentage.
Fly ash dosage did not have a significant effect on
The effect of sand percentage and cementitious most RCC properties except decreasing early age
content on flexural strength is shown in Figure 9. (1-day) compressive strength with increasing fly
Figure 9 indicates an optimal range of cementi- ash content. As expected, cementitious content
tious content to maximize flexural strength for was found to be positively related to compressive
these constituents. and split tensile strength. Flexural strength results
indicated an optimum cementitious content of
approximately 280 kg/m3.
5 APPLICATIONS AND LIMITATIONS Sand percentage had an inverse impact on all
RCC properties in this study as it alters the com-
This study was undertaken to develop recom- bined aggregate gradation of RCC. Increasing
mendations for initial mixture proportioning and sand percentage reduced the maximum dry density,

823
compressive strength, split tensile strength, and flex- point loading). ASTM International, West Consho-
ural strength. The predictive response equations for hocken, 4 p.
moisture-density and strength values are intended ASTM D1557. 2012. Standard test methods for labora-
to provide a starting point for RCC mixture propor- tory compaction characteristics of soil using modified
effort. ASTM International, West Conshohocken, 14 p.
tioning and adjustments. Results indicate that suf- Atis, C. 2005. Strength properties of high-volume fly
ficient compressive and flexural strengths for typical ash roller compacted and workable concrete, and
specification compliance can be achieved with rela- influence of curing condition. Cement and Concrete
tively low cementitious contents and even higher fly Research, 35 (6), 1112–1121.
ash contents. Increasing the sand percentage will Box, G. and Wilson, K. 1951. On the experimental attain-
facilitate achieving field compaction, however it will ment of optimum conditions. Journal of the Royal Sta-
come at the expense of strength reduction. tistical Society. Series B (Methodological), 13 (1), 145.
Cao, C., Sun, W., and Qin, H. 2000. The analysis on
strength and fly ash effect of roller-compacted con-
ACKNOWLEDGEMENTS crete with high volume fly ash. Cement and Concrete
Research, 30 (1), 71–75.
Delatte, N. and Storey, C. 2005. Effects of density and
This publication is based on the results of ICT- mixture proportions on freeze-thaw durability of
R27–149, which was conducted in cooperation roller-compacted concrete pavement. Transporta-
with the Illinois Center for Transportation, the tion Research Board: Journal of the Transportation
Illinois Department of Transportation Division of Research Board, 1914, 45–52.
Highways, and the U.S. Department of Transpor- Harrington, D., Abdo, F., Adaska, W., Hazaree, C., Cey-
tation Federal Highway Administration. The con- lan, H., and Bektas, F. 2010. Guide for roller-com-
tents of this report reflect the view of the authors, pacted concrete pavements.
who are responsible for the facts and the accuracy LaHucik, J. 2016. Selecting material constituents and
of the data presented herein. The contents do not proportions for specific roller-compacted concrete
mechanical properties. Masters Thesis, University of
necessarily reflect the official views or policies of Illinois at Urbana-Champaign.
the Illinois Center for Transportation, the Illinois LaHucik, J. and Roesler, J. 2015. Low fines roller-compacted
Department of Transportation, or the Federal concrete. In International Airfield and Highway Pavements
Highway Administration. This report does not Conference, ASCE, Miami, Florida, 441–452.
constitute a standard, specification, or regulation. Mardani-Aghabaglou, A. and Ramyar, K. 2013. Mechan-
The authors would like to thank Joshua Cheung, ical properties of high-volume fly ash roller compacted
Sachindra Dahal, Juan Pablo Ricardo Mendez concrete designed by maximum density method. Con-
Ruiz Fernandez, Omar Jadallah, Douglas De struction and Building Materials, 38, 356–364.
Andrade Neves, and Quang Tran for their assist- Nili, M. and Zaheri, M. 2011. Deicer salt-scaling resist-
ance of non-air-entrained roller-compacted concrete
ance with specimen fabrication and testing. pavements. Construction and Building Materials, 25
(4), 1671–1676.
Pittman, D. 1989. The effects of the construction process
REFERENCES on selected fresh and hardened properties of roller-
compacted concrete (rcc) pavements (No. WES/MP/
Amer, N., Storey, C., and Delatte, N. 2004. Roller- GL-89–22). Army Engineer Waterways Experiment
compacted concrete mix design procedure with gyra- Station, Vicksburg, MS Geotechnical Lab.
tory compactor. Transportation Research Record: Jour- Qasrawi, H., Asi, I., and Wahhab, H. 2005. Propor-
nal of the Transportation Research Board, 1893, 46–52. tioning rccp mixes under hot weather conditions for
American Concrete Institute (ACI). 1995. ACI 325.10R- a specified tensile strength. Cement and Concrete
95 State-of-the-art report on roller-compacted con- Research, 35 (2).
crete pavements. ACI, Farmington Hills, Michigan. Shihata, S. 2000. Strength and density of laboratory-
American Concrete Pavement Association (ACPA). 2014. prepared rcc specimens: effect of compaction proce-
ACPA guide specification: roller-compacted concrete dure. Cement, Concrete and Aggregates, 22 (1), 1–9.
pavements as exposed wearing surface. ACPA, Rose- Tangtermsirikul, S., Kaewkhluab, T., Jitvutikrai, P. 2004.
mont, Illinois. A compressive strength model for roller-compacted
ASTM C1435. 2008. Standard practice for molding concrete with fly ash. Magazine of Concrete Research,
roller-compacted concrete in cylinder molds using a 56 (1), 35–44.
vibrating hammer. ASTM International, West Con- Vahedifard, F., Nili, M., and Meehan, C. 2010. Assessing
shohocken, 5 p. the effects of supplementary cementitious materials
ASTM C39. 2012. Standard test method for compressive on the performance of low-cement roller compacted
strength of cylindrical concrete specimens. ASTM concrete pavement. Construction and Building Materi-
International, West Conshohocken, 7 p. als, 24 (12), 2528–2535.
ASTM C496. 2011. Standard test method for splitting Williams, S. 2013. Comparison of the Superpave gyra-
tensile strength of cylindrical concrete specimens. tory and proctor compaction methods for the design
ASTM International, West Conshohocken, 5 p. of roller-compacted concrete pavements. Transporta-
ASTM C78. 2010. Standard test method for flexural tion Research Board: Journal of the Transportation
strength of concrete (using simple beam with third- Research Board, 2342, 106–112.

824
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Bearing capacity assessment of a flexible pavement subjected

to seasonal effects

Gabriel Bazi
Lebanese American University (LAU), Beirut, Lebanon

Steve Saboundjian
Alaska Department of Transportation and Public Facilities, USA

Robert Briggs
Dynatest North America, Inc. Chicago, USA

Per Ullidtz
Dynatest International A/S [Former Affiliation], Gladsaxevej, Søborg, Denmark

ABSTRACT: Alaska’s Parks Highway is an asphalt paved and predominantly two-lane roadway that
extends over 323 miles (520 km), and links its two largest cities, Anchorage and Fairbanks. The Alaska
Department of Transportation and Public Facilities (DOT&PF) places annual springtime load restric-
tions on this primary haul route. To investigate the possibility of structurally upgrading the roadway to
eliminate the springtime load restrictions, the DOT&PF assessed the bearing capacity of this flexible
pavement by performing Falling Weight Deflectometer (FWD) testing along the highway after the thaw
period, in conjunction with continuous Ground Penetrating Radar (GPR) measurements. In addition,
periodic FWD testing was carried out weekly during springtime at three temperature data probe sites
along the highway to assess springtime bearing capacity and develop seasonal adjustment factors for the
different pavement layers. Using a 20-year design period, this study assessed bearing capacity and quan-
tified the required structural overlay with the currently imposed springtime weight restrictions and the
additional pavement damage due to the springtime load restrictions removal. Based on the assumptions
made in this study, results showed that two to six inches (5 to 15 cm) of overlay may be required and the
additional overlay needed to remove the weight restrictions is minimal.

Keywords: Falling Weight Deflectometer (FWD), Ground Penetrating Radar (GPR) Springtime load
restrictions, Spring thaw, Parks Highway

1 INTRODUCTION result in accelerated fatigue damage in the bound

layer (Raad et al. 1998, Raad et al. 1992, Doré and
The bearing capacity of flexible pavements is Zubeck, 2008). This can be illustrated by observing
highly sensitive to environmental conditions. In the maximum and minimum air temperatures dur-
cold regions, pavement layers freeze in the winter, ing the month of April at one location along the
thaw in the spring, and recover in the summer and Parks Highway at Milepost (MP) 208 as shown in
fall seasons. Figure 1; the maximum air temperature during the
During the spring, the accumulated frozen water day is above 32°F (0°C) leading to thawing, whereas
in the unbound layers melt. The thaw initiates at the the minimum air temperature at night drops below
top of the pavement and progresses rapidly into the 32°F (0°C) leading to the stiffening of the asphalt
granular base and sub base layers. The released water concrete surface layer while the unbound base and
in the granular layers can be temporarily trapped by sub base layers are saturated and weak. The stiffen-
the underlying frozen layer causing a near-satura- ing of the surface layer and softening of the sup-
tion condition in these layers. This moisture increase porting layers lead to accelerated fatigue damage
can significantly reduce the bearing capacity of whenever loading is applied. The hard caramel
the unbound layers. Poor support from the under- layer on top of the soft custard base in the “Crème
lying granular layers during this condition may Brûlée” dessert is a perfect analogy of this behavior.

825
 Cantwell, AK (MP 208) 90
68 Weight restriction s tarting da te (Month/ Day)
80 3/14
56 c
......
G:'
..
·~
70

~t
60
32 ~

e 20
"fo 50

40
~ 8 ~
~ 0
30
4/ 19

~
-4
20
-16 10
Jan Feb. Mar Apr. May Jun. Jul. Aug. Sep. Oct. Nov. Dec
Month 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
....... Maximum ...... Minimum
Year

Figure 1. Monthly air temperatures at MP 208 of the Figure 2. 10-year history of Parks Highway weight
Parks Highway, Alaska. restriction days and starting dates.

The thaw will further progress into the subgrade

soil resulting in potential water saturation and pos- The Parks Highway crosses two out of the three
sible build-up of pore-water pressure. High pore- Alaska regions. It starts in the Central region near
water pressure changes the effective stress state and Anchorage at the junction with the Glenn High-
reduces subgrade shear strength (Simonsen et al. way at MP 35, reaches the Northern region at Lit-
1999). tle Coal Creek at MP 163, and ends in Fairbanks
Significant decreases in resilient moduli of sub- at the junction with Richardson Highway at MP
grade soils when the excess-water during thawing 362. Alaska DOT & PF’s Central and Northern
releases have been reported. Several studies have regions are responsible for imposing weight restric-
reported spring thaw to be the most significant tions within their jurisdictions. Figure 2 shows a
deterioration factor for these pavements (White 10-year history of the weight restriction days and
et al. 1990, Janoo et al. 1990). In pavements sub- starting dates along the Highway. The springtime
jected to freezing, the relative load-induced damage weight restriction is shown to range from 30 to
during spring thaw could be 1.5 to 3 times higher 80 days, and it could start around the second week
than the average annual damage (St-Laurent et al. of March to the third week of April. 75 to 85%
1995). restrictions are typically imposed, with the latter
To protect roadways during spring thaw, high- being the most common.
way agencies impose weight restrictions to reduce
pavement damage. These weight limits are applied
systematically during the spring period to restrict 2 PAVEMENT NON-DESTRUCTIVE
truck operations to lower than normal weights. TESTING

Alaska DOT&PF performed non-destructive test-

1.1 Alaska’s Weight Restrictions
ing using a Dynatest Falling Weight Deflectometer
The Alaska DOT&PF has developed an effec- (FWD) in August 2010 at about 0.1-mile (160 m)
tive seasonal weight restriction program that uses intervals from MP 72 to 356. Two sections from
Temperature Data Probe (TDP) profiles as one MP 72 to 90 and MP 287 to 305 were not tested
tool to issue fact-based weight restriction notices. due to rehabilitation plans or construction at the
TDP sensors deployed at strategic locations pro- time of the testing. The FWD was setup with the
vide a vertical temperature profile in the six feet small plate (≈ 12-in (300 mm) diameter) and seven
(1.8 m) layer below the pavement surface. geophones located at 0, 12, 18, 24, 36, 48, and
Seasonal weight restrictions on Alaska’s road- 60-in (0, 30, 45, 60, 90, 120, and 150 cm) radial off-
ways are based on historical data, buried thermis- sets from the center of the plate. Three drops were
tor data, weather forecasts, and field observations recorded at each test point. FWD loads ranged
by maintenance personnel. from approximately 8,000 to 11,000-lbs (35.5 to
Seasonal weight restrictions apply to all vehicles 49 kN). For analysis purposes, the last drop (drop 3)
over 10,000-lbs (44.5 kN) gross vehicle weight. The at each test point was evaluated. A summary of
weight restrictions are imposed as a percentage the average normalized center pavement deflec-
of the legal allowable weights, except for steering tions and the surface temperature during testing
axles. 100% represents the legal maximum allow- is shown in Figure 3. The air temperature during
able weight. Restrictions are based on 85, 75, or testing ranged from 46 to 79ºF (8 to 26ºC); while
50% of the maximum allowable axle/axle group the pavement surface temperature ranged from
weight. 32 to 89ºF (0 to 32ºC).

826
 50 90 20,000
i~tP i\fP
" ~5 : \0~ ~ ~ : l$~.6:<
.§ ~0 7 ~ 16.000 ' '
~
c: ~
35 72; ~ 12,000
'I I \\1)
:<-inttnii":
'I

0~ 30 66 .:
~ !. 25 60 ll e
~ ~ 20 5 ~ E"
.::
5
~
8.000

.. ·~
~ :~ .p ~ 4,000

~ s "
36 :n
z 0 30 0
50 100 150 200 250 300 350 ~00
so 100 ISO ~00 ~so 300 350 100

~IP
o FWD Center DeOcctions • urf.1cc Temper:~ture
Figure 5. AADT along the Parks Highway and FWD
Figure 3. FWD normalized center deflections and sur- testing locations.
face temperature.

225
15 75
200
c 0

"' 12
0
60 c 175
~ ~ ISO
.., 0
45 j !-
0

.
125

II
,3_ ...:
~ .: :c -'=' 100
30 !-
~~ ~0
~
75
u 15 ::E
50

"-a< """ 25

50 100 150 200 250 300 350 400

10 II 12
MP
Mont h
o Asphalt Concrete t:. Aggregate Base
- Minimum -+-Maximum - Average

Figure 4. Pavement layer thicknesses determined from Figure 6. AADT monthly variations.
GPR testing.

Ground Penetrating Radar (GPR) testing was show the seasonal traffic variations. Figure 6 shows
performed in July 2012 and reported at 0.005-mile on average that the monthly ADT exceeds the aver-
(26.4 ft., 8 m) intervals from MP 72 to 362. Fig- age annual daily traffic from May through Septem-
ure 4 shows the GPR-determined thickness of the ber, which is expected.
asphalt concrete and base layers. Automatic vehicle classification counters are
All the data was georeferenced and the GPS used at several locations along the highway to
coordinates were used to select the closest GPR classify the vehicles into the FHWA’s thirteen (13)
record to every FWD test point; this task was vehicle classes or into groups of classes. Classes 1
accomplished using a Visual Basic for Applications through 3 correspond to motorcycles, passenger
(VBA) code that calculated the ‘as-the-crow-flies’ cars, pickup trucks and vans; Class 4 corresponds
distance between points using the haversine for- to busses; classes 5 through 7 correspond to single
mula and selected the GPR record with the closest unit trucks; and classes 8 through 13 correspond to
distance to each FWD test point. single—and multi-trailer combination trucks.
Based on the FWD and GPR testing limits, At one location in the Central region near Wil-
complete data is only available from MP 90 to MP low (MP 72), the vehicles are classified into the
287 and from MP 305.45 to MP 355.65. thirteen classes, as opposed to the groups of classes.
This data is used to obtain the proportions of the
individual classes within every group as shown
3 TRAFFIC in Table 1. For example, 35.2% of the group that
contains vehicle classes 8 through 13 corresponds
Figure 5 shows the Average Annual Daily Traf- to class 9 (5-axle single-trailer combination truck,
fic (AADT) collected and reported by Alaska typically referred to as the 18-wheeler truck).
DOT&PF in 2013 from MP 90 to MP 356. The Classes 1 through 3 are typically ignored in the
AADT is relatively low and does not exceed 4,000 analysis due to the insignificant damage relative to
vehicles per day except near Fairbanks (∼MP 355). the larger and heavier trucks and buses. Classes 7,
The monthly Average Daily Traffic (ADT) is 11 and 12 were ignored in this analysis based on
calculated at several locations along the highway to the data from Table 1.

827
Table 1. Vehicle classification data at Willow. 4.1 Asphalt concrete seasonal variations
Group FHWA Vehicles Proportion with The air temperatures were obtained for several
(Class) Class (%) in each group weather stations along the highway from the
Western Regional Climate Center (WRCC 2016)
1–3 1 2.0 2.4 database over periods ranging from 6 to 25 years.
2 46.3 55.7 The weather information was grouped into fiv-
3 34.8 41.9 ezones by milepost with similar air temperatures
4 4 0.8 100.0 (Table 3). The temperature range (maximum
5–7 5 8.8 87.4 and minimum difference) is increasing with an
6 1.2 12.2 increase in latitude (or MP from south to north).
7 0.0 0.4 The air temperatures were used to estimate the
8–13 8 0.7 11.2 Asphalt Concrete (AC) temperatures for pave-
9 2.1 35.2 ment modeling.
10 1.2 19.7
11 0.0 0.0 4.1.1 Asphalt concrete seasonal temperature
12 0.0 0.0
variation
13 2.1 33.8
The seasonal variation of the AC temperature is
assumed to be sinusoidal in accordance with the
following equation:
Table 2. Axles weights and proportions.
T1 T2 T1 T2
Axles Proportion T= +
Representative weights of Total 2 2
⎛2 ×π π⎞
Class vehicle (kips)(1) Traffic (%) × sin ⎜ × (W W)+ ⎟ (1)
⎝ 52 2⎠
4 12/38 2
5 12/20 10 where:
T = Pavement temperature in week number W
6 12/38 1 T1 = Maximum pavement temperature during
the year
8 12/20/20 1 T2 = Minimum pavement temperature during
the year
9 12/38/38 3 W = Week number counted from January 1st
W0 = Week number corresponding to maximum
10 12/38/42 2 temperature
T1 and T2 were obtained using the air tem-
13 12/38/38 3 peratures. T1 was magnified by 20% to account
/20/20
for the maximum asphalt temperature being
1)
Axles or axle groups weights are shown from left to higher than the air temperature during the sum-
right. mer (Barker et al. 1977); this assumption was
found to be reasonable after reviewing the air
and pavement temperatures for two Long-Term
Finally, the axle weights for the classes used in Pavement Performance (LTPP) sections. Week
the analysis before weight restriction are shown in number 26 corresponds to the maximum tem-
Table 2 along with their proportions relative to the perature (W0).
total traffic.
A 13-inch (33 cm) dual wheel spacing, 36-inch
(90 cm) axle spacing and 110 psi (760 kPa) tire con- Table 3. Average air temperatures.
tact pressure were used in the analysis.
A growth rate of 1.5% is calculated based on Air Temperature (°F)
historical records from 1996 through 2014. From To
MP MP Minimum Maximum Range

90 163 2 68 66
4 PAVEMENT MODELING 163 210 (Cantwell) −2 66 67
210 249 (Healy) −8 69 77
The seasonal modeling of the various pave- 249 305 (Nenana) −13 72 86
ment layers is a critical step in the Mechanistic- 305 355 (Fairbanks) −16 74 90
Empirical (ME) analysis.

828
4.1.2 Asphalt concrete temperature during where:
fwd testing εAC R = Tensile strain at bottom of AC, microstrain
The average asphalt temperature at the time of the N = Number of load repetitions in millions
testing was determined in accordance with ASTM E = Modulus of AC, ksi
D7228 using the “BELLS” equation which uses The permanent deformation model for the base
as input the infrared surface temperature which and subgrade materials was derived from an anal-
is measured automatically, in real time, while the ysis of the WASHO and AASHO road tests and
FWD tests are being conducted; the time of day; from CBR designs by J.M. Kirk (Kirk 1973). A
the thickness of the asphalt-bound layer; and the regional factor of 4 was used in the model which is
previous day average air temperature. recommended for spring thaw regions. This model
is similar to the model found in the Alaska Flexible
Pavement Design Manual (ADOT&PF 2004).
4.2 Unbound layers seasonal variations
C
Seasonal monitoring was performed at three sites ⎛E⎞
along the Parks Highway; Cantwell (MP 211, σ 1, p = 0.0135 × N −0.307
×⎜ ⎟ (3)
⎝ E0 ⎠
northernmost), Little Coal Creek (MP 163.2) and
Chulitna (MP 117, southernmost). The seasonal
where:
modular ratios were determined by analyzing the
σ1,p = Vertical stress on top of unbound layer, ksi
pavement deflection measurement son 12 dates
N = Number of load repetitions in millions
between March 8 and July 6, 2011, at each of the
E = Modulus of material, ksi
locations (Bazi et al. 2015).
E0 = 23.2 ksi
Based on the three sites, the average subgrade
C = 1.16 for E < E0
and aggregate base seasonal modular ratios were
C = 1 for E ≥ E0
calculated as 0.36 and 0.66 relative to the sum-
mer conditions, respectively. They occurred
about 5 days after the Spring thawing initiation. 5 PAVEMENT EVALUATION
Figure 7 shows the modular ratios used in the
modeling. All layer moduli were back calculated for each deflec-
tion basin using the Dynatest ELMOD software
4.3 Deterioration models (ELMOD 2016) through a reverse, layered analysis
technique using the Odemark-Boussinesq deflection
Two deterioration models were used in the ME basin fit option. It should be noted that, in general,
pavement evaluation, a fatigue cracking model for most of the measured magnitudes of deflection are
the AC layer and a permanent deformation model due to the response of the subgrade. It is therefore
for the unbound layers. very important that the subgrade modulus is accu-
The fatigue cracking model was used in the rately determined. A small error in the subgrade
development of the Asphalt Institute’s MS-1 Pave- modulus will lead to large errors in all layer moduli.
ment Design Manual. The asphalt concrete layer moduli are adjusted
to a reference temperature of 77°F (25°) for report-
−0.259
⎛ E ⎞ ing purposes. The subgrade moduli are shown to
ε AC = 261 × N −0.304 × ⎜ (2)
⎝ 435 ⎟⎠ be relatively low with high base moduli from MP
90 through 155 (Figure 8).
1,000.0

100.0 ,. 10.000
.g I
I

"'
~ I
I 1.000
10.0
"'
~0 I Reference g
:2
1.0
~ ~
;:::
100

§· 10
0.1
01 /01 03/02 05/02 07/02 09/01 11/01 01 /01
Date (MNUDD) 50 ISO 200 250 300 350 400
:\II'
- - Subgrade - - - Base o Asph:tlt Concrete A Aggregate Base o Subgr.tde

Figure 7. Subgrade and base modular ratios. Figure 8. Backcalculated layer moduli.

829
5.1 Overlay calculation Table 4. Weight restrictions effect on overlay.
The FWD deflection data was analyzed using the Weight restriction
ELMOD software package for a 20-year design during spring season (%)
period using the traffic mix with no weight restric-
tions. Structural overlay requirements for the traffic MP 100 (Legal) 85 75 50
mix are shown graphically in Figure 9. The calcu-
lated structural overlay is shown to range from 2 90–155 2.0 1.7 1.6 1.4
to 6-inch (5 to 15 cm) on average for the four sec- 155–287 2.2 2.2 2.1 2.1
305.45–325.2 5.6 5.5 5.4 5.3
tions that were identified based on relatively simi-
325.2–355.65 1.6 1.6 1.5 1.5
lar performance (MP 90 to 155, 155 to 287, 305.45
All 2.4 2.3 2.2 2.1
to 325.2 and 325.2 to 355.65). The section of the
highway with the highest overlay requirement from
MP 305 to 325 is currently scheduled for rehabilita-
tion. A more detailed project level analysis would
6 SUMMARY AND CONCLUSIONS
typically be required to evaluate various alternatives
along with a life cycle cost analysis for the highway
FWD and GPR testing performed along the Parks
sub-sections.
Highway were used to assess the bearing capacity
of its flexible pavement. Data from weather stations
5.2 Springtime load restrictions and subsurface temperature data probe sites were
used to model the seasonal behavior of the differ-
To quantify the damage, or the additional over-
ent pavement layer. Using a 20-year design period,
lay needed, during the weight restriction period,
the ELMOD software was used to determine the
the analysis was repeated by applying 85, 75 and
required AC structural overlay with the currently
50% weight restriction on all axles except for the
imposed springtime weight restrictions and the addi-
steering axles. An 85% weight restriction indicates
tional pavement damage due to the springtime load
that the weight of the axle is reduced by 15%. It
restrictions removal. Based on the assumptions made
was assumed that the weight restriction starts on
in this study, results showed that two to six inches of
April 1 for 50 days and the traffic during that
overlay may be required and the additional overlay
period accounts for 13% of the total annual traffic;
needed to remove weight restrictions is minimal.
it should be noted that the traffic is not uniform
The methodology described and used in this
throughout the year.
study can be applied to other highway pavements
The average damage calculated using the Miner’s
provided detailed data from nondestructive tools
rule during the spring season accounted for 27, 21,
(FWD, GPR) are acquired and analyzed in con-
18 and 11% of the total damage for the 100, 85, 75
junction with accurate traffic and temperature
and 50% weight restrictions scenarios, respectively.
data obtained along the highway corridor.
Due to the large pavement damage, even with
the 50% weight restrictions, Table 4 shows minimal
variations in the calculated overlays for all cases.
ACKNOWLEDGMENTS
For the four scenarios, the Parks highway
requires a 2 to 3-inch (5 to 7.5 cm) overlay on
The work represented herein was the result of
average, with one section requiring about 6-inch
a team effort. The authors would like to thank
(15 cm) overlay.
ADOT&PF personnel for providing the data for
analysis, and Infrasense for collecting and analyz-
15
MP90 -1 55 MP 155-287 t-.·IP 305 .45 - 325.2
ing the ground penetrating radar data.
Ave. "'2.0" Ave. = 2.2" ~ - = 5.6"
0
12
0
31 0

DISCLAIMER

The contents of this paper reflect the views of the

authors, who are responsible for the accuracy of
the data presented herein. The paper’s contents
do not necessarily reflect the views or policies of
Alaska DOT&PF or any local sponsor. This work
50 100 150 200 250 300 350 400
MP does not constitute a standard, specification, or
regulation. Alaska DOT&PF does not endorse,
Figure 9. Calculated overlay thickness without weight support or favor any product, equipment, technol-
restrictions. ogy, software or procedure cited in this paper.

830
REFERENCES [8] Raad, L., G. Minassian, and S. Gartin. Charac-
terization of Saturated Granular Bases Under
[1] Alaska Flexible Pavement Design Guide, Report Repeated Loads. Transportation Research Record:
No. FHWA-AK-RD-03-01, ADOT&PF, April Journal of the Transportation Research Board, No.
2004. 1369, Transportation Research Board, Washington,
[2] Barker, W.R., W.N. Brabstone, and Y.T. Chou. A D.C., 1992, pp. 73–82.
General System for the Structural Design of Flex- [9] Raad, L., E. Johnson, D. Bush, and S. Saboundjian.
ible Pavements. Proceedings of the Fourth Inter- Parks Highway Load Restriction Field Data Analy-
national Conference on the Structural Design of sis Case Study. Transportation Research Record:
Asphalt Pavements, Ann Arbor, 1977. Journal of the Transportation Research Board, No.
[3] Bazi, G., R. Briggs, S. Saboundjian, and P. Ullidtz. 1615, Transportation Research Board, Washington,
Seasonal Effects on a Low-Volume Road Flexible D.C., 1998, pp. 32–40.
Pavement. Transportation Research Record: Jour- [10] Saint-Laurent, D. and Roy, M. (1995). Structural
nal of the Transportation Research Board, No. Evaluation of Flexible Pavements in a Northern
2510, Transportation Research Board, Washington, Context: A Study Using the FWD, Proceeding of
D.C., 2015, pp. 81–89. DOI: 10.3141/2510-10. the 30th Annual Conference of AQTR, Association
[4] Doré, G. and Zubeck, H.K. (2008). Cold Regions Québécoise du Transport et des Routes, Quebec,
Pavement Engineering. USA: McGraw-Hill Canada.
Profes-sional. [11] Simonsen, E. and Isacsson, U. (1999). Thaw Weak-
[5] Evaluation of Layer Moduli and Overlay Design ening of Pavement Structures in Cold Regions-State
(ELMOD) Software, version 6.1.79. Dynatest of the Art. Cold Regions Science and Technology,
International, Naverland 32, DK – 2600 Glostrup, Volume 29, Issue 2, pp. 135–151.
Denmark, 2016. [12] Western Regional Climate Center Database. http://
[6] Janoo, V.C. and Berg, R.L. (1990). Thaw Weakening www.wrcc.dri.edu/coopmap/. Accessed October 17,
in Seasonal Frost Areas, Transportation Research 2016.
Record: Journal of the Transportation Research [13] White, T.D. and Coree, B.J. (1990). Threshold
Board, No. 1246, Transportation Research Board, Pavement Thickness to Survive Spring Thaw, Pro-
Washington, D.C., pp. 217–233. ceedings of the Third International Conference on
[7] Kirk, J.M. Revideret metode til dimensionering Bearing Capacity of Roads and Airfields. Norwe-
af bituminøse befæstelser (Revised method for gian University of Science and Technology, Trond-
the structural design of bituminous pavements, in heim, Norway, pp. 41–51.
Danish), Asfalt nr. 42, 1973.

831
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Evaluation of pavement load bearing capacity comprised of insulation

layers during thaw season

L. Hashemian, N. Tavafzadeh & A. Bayat

Department of Civil and Environmental Engineering, University of Alberta, Edmonton, Alberta

ABSTRACT: One of the strategies for minimizing the negative effects of freezing on frost-susceptible
subgrade is placing insulation layers on top of the subgrade. This technique helps to mitigate the forma-
tion of ice lenses and frost heave in the subgrade and subsequently reduces the associated damage. Using
insulation layers prevents subgrade strength reduction that results from the existence of excess water
formed from melted ice lenses during thaw season.
This paper evaluates the effect of using bottom ash, a recently introduced by-product of power gen-
eration, as an insulation layer, as well as the effect of commonly used polystyrene boards on subgrade
resilient modulus variations during thaw season at the University of Alberta’s IRRF test road facility in
Edmonton, Canada.
To evaluate the subgrade strength, Falling-Weight Deflectometer (FWD) testing was conducted at
10-m intervals along the test road during thaw season. The back-calculated moduli from deflection basins
were used to determine the resilient modulus of the insulated sections and the control section.
The study results revealed that using polystyrene boards as insulation layers protected the subgrade
soil from freezing and thawing effect; however, it also reduced the pavement bearing capacity. The bottom
ash layer was affected by freeze-thaw, but it could properly protect the subgrade soil without having an
adverse effect on pavement bearing capacity.

1 INTRODUCTION Using waste and recycled materials such as

bottom ash as an insulation layer can be an eco-
One of the major challenges of pavement per- nomical and environmentally friendly solution.
formance in cold regions is frost-heave, which can Bottom ash is a waste by-product material pro-
negatively affect the frost-susceptible subgrade soil duced from burning coal in power plants and has
and, consequently, the pavement riding quality. In recently been used as an insulation layer. Alberta
addition, after the start of the thaw season, a large generates almost 60 percent of its electricity from
amount of excess water will be available in the burning coal, which produces a huge amount of
pavement system for a short time as a result of the ash (CAPA, 2015). Research has shown that bot-
frozen water melting. This extra water can reduce tom ash is a non-frost-susceptible material that
the strength of the subgrade soil (Sebaaly et al., can effectively reduce the freeze-thaw depth in cold
1994; Hein and Jung, 1994; Ovik et al., 1999) and areas (Tavafzadeh et al., 2014; Field et al., 2011,
bearing capacity of the whole pavement, which can Havukainen, 1987).
lead to an accumulation of fatigue in the upper Although using insulation layers can effectively
pavement layers (Dore and Zubeck, 2009). protect the subgrade soil from the effects of freez-
Insulation layers have been used as one of the ing, using these soft layers may decrease the pave-
strategies for protecting frost-susceptible sub- ment bearing capacity (Dore and Zubeck, 2009).
grade soil from freeze-thaw damages. Different The objective of this study is to investigate the
materials such as sawdust, shredded tires and pol- effect of two different insulation layers, namely
ystyrene boards have been evaluated as insulation polystyrene board and bottom ash, on subgrade
layers (Dore et al., 1995). Polystyrene boards were protection as well as pavement bearing capacity
shown to be effective at reducing heave in a study during the thaw season in a test road in Edmon-
conducted in Colorado (Hayden and Swanson, ton, Alberta, Canada. Edmonton is located in
1972), and another study done in Alaska showed a freeze-dry climatic region, where the average
they reduced the freezing depth (Esch, 1972). Freezing Index (FI) based on 30 years of his-
Two studies in Edmonton, Canada, showed that toric weather data is 1,365°C days (Environment
polystyrene board can reduce the frost depth by Canada, 2013). Frost heave and thaw weakening is
at least 40 percent (Tatarniuk and Lewycky, 2011; a challenging problem that local highway agencies
Tavafzadeh et al., 2014). deal with each year.
833
1.1 IRRF test road 100

II 1\ 1'~1 1 I
--E!-GBC
The IRRF’s test road is an access to the Edmon- 80
1:,. I I- · Suhgrade Soil
ton Waste Management Centre (EWMC), located
I I
~ \I I
......,...Bottom Ash
on the eastern edge of Edmonton. Construction ~0 I I
I
of the road started in May 2012, the first stage 0

of paving was completed in August 2012, and the

,.§
C40 I JI' I I I I

second stage was finished in October 2013. The ~

test road has been opened since mid-October
ct
20
'ls I\ I I ......
1~
2015; the two-lane, 500-m long road is currently
carrying approximately 1,000 garbage trucks per 0
I I I ~I I
100.000 10.000 1.000 0.100 0.010 0.001 0.000
lane each day. Grain size. D (mm)
The IRRF road pavement consists of 25 cm
of dense-graded Hot Mix Asphalt (HMA) and Figure 2. Grain size distributions of different materials
45 cm of Granular Base Course (GBC). The three used in the pavement.
insulated 20-m long sections are comprised of
a 100-cm thick bottom ash layer (section B.A.),
a 10 cm (section P-10) and a 5 cm (section P-5) 1.2 Test road material
polystyrene board. All the insulation layers are
Two different HMA mixes, 16-cm thick at the bot-
located under the base layer and on top of the
tom and 9-cm thick on top, consisted of granu-
subgrade. The adjacent section constructed with
lar aggregates with the maximum nominal size of
the same pavement and no insulation layer acts as
12.5 mm and 25 mm, respectively. The GBC layer
a Control Section (CS).
was comprised of well-graded gravel (GW) based on
To investigate the temperature in the pavement
the Unified Soil Classification System (USCS) with a
depth, the B.A. and P-10 sections were instru-
maximum particle size of 19 mm. The subgrade soil
mented using 109 AM-L thermistors, and the CS
classified the subgrade as clayey sand (SC) with a liq-
was also instrumented using CS650 time domain
uid limit of 25 and Plastic Index (PI) of nine percent.
reflectometers (TDRs). The probes are installed
The bottom ash was mainly composed of silica,
on top of the subgrade, 1.7 m and 2.5 m below the
alumina, iron and less than five percent non-com-
pavement surface. A CR1000 datalogger was pro-
busted coal particles, and it was free of large lumps
grammed to collect data at 15-minute intervals from
and impurities. Optimum moisture in the bottom
all of the sections. The datalogger was equipped
ash layer was calculated as 35 percent, and during
with a spread spectrum Model RF401 radio with an
construction, the layer was wrapped in geotextile
antenna Model L14221 installed at an onsite trailer,
to avoid mixing with natural soil. The particle size
where a computer transmitted the data regularly to
distributions of all the granular layers are illustrated
the University of Alberta. Figure 1 shows the cross-
in Figure 2. Closed-cell Styrofoam Highload 100
sections and thicknesses of layers, as well as the
extruded polystyrene boards with a compressive
locations of probes.
strength of 690 kPa and a minimum flexural strength
of 585 kPa were used in other insulated sections.

<m
0
9 2 THAW WEAKENING DURATION
"
70
Figure 3 shows the temperature variations in dif-
ferent seasons (1) on top of the subgrade in the CS,
(2) at the top of the bottom ash layer in the B.A.
section and (3) underneath the polystyrene layer in
the P-10 section. The data shows that the CS and
I I
B.A. sections were frozen during the winter time
l _____ l while the temperature below the polystyrene layer
170 Q TDR-CS-2 0 . TH-B.Ash-2 . TH-Poly-2 never goes below zero. That means the P-10 section
250 is fully protected from freeze. Figure 4 shows that
the subgrade under the bottom ash layer is also
350 protected from freezing effects, while the subgrade
O TDR • TH: Tbermi.slor in the CS section is frozen in the winter time at the
same depth. Figure 5 shows the thaw depths for
Figure 1. Cross-sections and as-built depth of thermis- the CS and B.A. sections. According to the figure,
tors and moisture probes. thaw starts in the CS on March 8th and continues

834
 Table 1. FWD testing dates and temperature.
V:O<..o~ev:~~~~

Test ID April-21 May-5 June-5 July-4

, Temperature ('C)

Test Date 21-Apr-15 5-May-15 5-Jun-15 4-Jul-15

Temperature 18.0 19.0 30.0 37.0
(°C)

Jan, IS

Mar. !5
Feb, JS
Table 2. Maximum (D0) and minimum (D9) deflections
Figure 3. Temperature in different sections on top of of the sections from FWD tests.
the subgrade layer.
21-Apr 5-May 5-Jun 4-Jul
Test
ID D0 D9 D0 D9 D0 D9 D0 D9
--NN""<.o>
~

I
I
~ 0

i1 I

CS 161.4 41 179.9 43 251.4 42.4 362 40.8

0

Fe~ 1515 1 1 1 ~/j' I I :5~

Temperatu re ("C)

B.A. 158.5 40 172.2 40 239.4 42.6 384.2 41.7

* '"
~

P-10 215.5 34 254.3 35 364.5 37.2 552.9 37.9

0 ~

P-5 177.4 38 195.3 39 341.8 41 574.9 41

I
0
~

I
•
--

I
0 ~

Aug. 13 I

Oct., 13

Dec, 14

Apr, 15
Dec, 13

~w
-n
~~

•w
Nov, 14
Scp, 13

~~

-~

-~
-~ ~~

Jrut,\5
-~

-~

1500, and 1800 mm from the centre of the load

M~r.

plate was used for the testing. Four FWD tests

Figure 4. Temperature in different sections 1 m below were conducted in thaw weakening duration as
the subgrade level. shown in Table 1. The load was applied in the mid-
dle of each section to both outer and inner wheel
paths. The stress levels of 40.0 kN was chosen for
=
•

•
.,•:i::

•:i::
i:
0~

"'~
~

investigation purposes. Table 2 shows the maxi-

::5 ::: ;5 ;,g g:;

mum deflection under the plate after the FWD

T h awOepl h (c m )

'f
~

loading (D0) and also the deflection measured

~

Te m p. (C)
w

by the farthest sensor (D9). It is clear that in all

g

v
lil

sections, D0 is going up by increasing the testing

:;

temperature. The maximum deflection is always

0
\
g

~ ~ c

allocated to the P-10 section, and the minimum D0

is for the B.A. section. Comparing D9 values of
l h:

-
"t
g ..

-!
n"
I~J
~].

JJ.
~

~ ;?

"
~ ~

~
~

5- ~

different sections shows that the P-5 and P-10 sec-

.. ....

tions always have lower values compared to the CS

Figure 5. Start and duration of thawing in CS and B.A. and B.A. sections.
sections.

to March 14th, whereas in the B.A. section, it starts 4 ANALYSYS OF THE RESULTS
on March 10th and ends on March 20th. In conclu-
sion, the thawing procedure lasts four days longer 4.1 Subgrade resilient modulus
in the B.A. section than in the CS.
To back-calculate subgrade resilient modulus using
To estimate the thaw weakening duration, this
FWD data, the AASHTO method was used; resil-
study used the MEPDG design guide. According
ient modulus can be calculated from Equation 1
to the guideline, the recovery period for this type
(Transportation Officials, 1993).
of subgrade soil with PI = 9% and P-200 = 46.9%
was deemed to be approximately 120 days (ARA
0.24 × P
2004). Consequently, the thaw weakening period Mr = (1)
would continue to mid-July. Dr r

In Equation 1, Mr is the resilient modulus of the

3 FALLING WEIGHT DEFLECTOMETER subgrade soil, P is the FWD target load, Dr is the
(FWD) TESTING normalized deflection under load P at distance r,
and r is the distance of the geophone to the load
A Dynatest 8000 composed of nine sensors with plate. As it is explained is this method, the selected
configuration at 0, 200, 300, 450, 600, 900, 1200, geophone should be far enough from the centre

835
of the load plate so as not to be affected by other
layers. To satisfy this criterion, the selected deflec- 1800 1343 1400
tion location should be compared to 0.7 × ae, which 1600
1200
is calculated from Equation 2. 1400
1000
1200
'i 1000 800
⎡ ⎛ Ep ⎞ ⎤
2 6 Q.
800 600
ae ⎢a 2 + ⎜ D3 ⎟ ⎥ (2) "' 600
⎢ Mr ⎠ ⎥
400

⎣ ⎝ ⎦
400
200
200

cs BA P- 10 P ·S
In Equation 2, ae is the radius of stress bulb - 21 -Apr - 5-May - 15-Jun ~ 15 -Jul - Ave
at the subgrade-pavement interface, a is the load
plate radius, and D is the total thickness of pave-
ment layers above the subgrade, which in this study Figure 7. Pavement effective modulus in different
is a total of 70 cm comprising of 25 cm of HMA sections.
and 45 cm of GBC layers. The deflection observed
at the ninth geophone or D9 located at 1,800 mm ⎧ ⎡ ⎤⎫
from the loading plate centre met the abovemen- ⎪ ⎢ ⎥⎪
tioned requirement for all sections. ⎪ ⎢ 1 ⎥⎪
⎪ ⎢ 1 3 ⎥⎪
Figure 6 shows the Mr variation for different ⎪ ⎢ 1+ ⎛ D⎞ ⎥ ⎪
sections in different months. The lowest value for ⎪⎪ 1 ⎢⎣ ⎝ a ⎠ ⎥⎦ ⎪⎪
the Mr is for the CS, and the highest value is for d0 . Pa ⎨ + ⎬
the P-10 section. Comparing these results with ⎪ ⎛ D3 Ep ⎞
2 Ep ⎪
Figure 3 reveals that since the subgrade soil of the ⎪ Mr 1 + ⎜ ⎟ ⎪
P-10 section never goes through freeze-thaw con- ⎪ ⎝ a Mr ⎠ ⎪
ditions, the resilient modulus of the subgrade is ⎪ ⎪
⎪ ⎪
always higher compared with other sections. The ⎪⎩ ⎪⎭
subgrade soil of the CS is more affected by thawing
effect and shows lower resilient modulus compared (3)
to other sections. The subgrade soil of the B.A.
section shows lower resilient modulus compared Figure 7 compares effective pavement modu-
to polystyrene sections but higher than the CS. As lus for different sections in thaw season, which is
Figures 3 and 5 show, the bottom ash layer is frozen an indicator of pavement bearing capacity. It is
in winter time. Although the subgrade below this clear that the Ep off all sections is very dependent
layer is never frozen (Figure 4), the excess water as a on pavement temperature. As the pavement tem-
result of the bottom ash layer thawing in spring may perature increases from 18°C in April to 37ºC,
have affected the subgrade resilient modulus. the Ep decreases significantly. Since the HMA
thickness of the pavement is high (25 cm), it can
be concluded that the weakening of the pavement
4.2 Effective pavement modulus due to temperature increase is a result of decreas-
The obtained Mr for each section was utilized to ing the HMA modulus. The lowest Ep values
calculate the effective pavement modulus (Ep) based were calculated for the P-10 section that had the
on the AASHTO method (Equation 3) (Transpor- highest deformation under FWD loading, as
tation Officials, 1993). shown in Table 2. As it was shown in Figure 6,
this section had the highest Mr value compared
to other sections. Hence, it can be concluded that
180 .5
140.0
using a soft polystyrene board under the base
160
140 120.0
layer affected the bearing capacity of the section.
120 100.0
As it is shown in Figure 7, by reducing polysty-
"(; 100 80.0 rene board thickness from 10 cm to 5 cm, the Ep
~ 80
60.0 increases 30% on average. The highest bearing
~ "' 40.0 capacity is for the bottom ash layer that has the
40
20 20.0 higher Mr compared to the CS.
0.0
cs BA P-10 P ·S

5 SUMMARY AND CONCLUSIONS

Figure 6. Subgrade resilient modulus in different This paper investigated the effect of using different
sections. insulation layers on subgrade modulus variation and

836
pavement bearing capacity during thaw season in REFERENCES
the IRRF test road located in Edmonton, Alberta,
Canada. The instrumented test road is comprised ARA, Inc. ERES Consultants Division. 2004. Guide
of three insulated sections including 1 m of bot- for Mechanistic-Empirical Design of New and Reha-
tom ash and 5 and 10 cm of polystyrene boards. bilitated Pavement Structures. Final report, NCHRP
The control section was located beside the insulated Project 1-37A. Transportation Research Board of the
National Academies, Washington, D.C.
sections for comparison purposes. Different FWD Canadian Association of Physicians for the Environ-
tests were conducted during the thaw season, and ments (CAPA). 2015. “Backgrounder: Phasing Out
back-calculation results were used to calculate the Alberta’s Coal Plants”, Toronto, Canada.
subgrade modulus and effective pavement modulus Doré, G. and Zubeck H.K. 2009. Cold Regions Pave-
of different sections. The results of this study are ment Engineering. 1st edition New York: McGraw-
summarized as follows: Hill Professional.
Dore, G., J.M. Konrad, M. Roy, and N. Rioux. 1995.
1. Monitoring the temperature data showed that “Use of alternative materials in pavement frost protec-
the subgrade below the 10-cm thick polystyrene tion: Material characteristics and performance mod-
board and 1-m thick bottom ash layer is fully eling.” Transportation Research Record 1481: 63–74.
protected from the effects of freezing. Environment Canada, 2013. http://weather.gc.ca/
2. All the sections showed greater deformation canada_e.html, Accessed Jan. 1.
under FWD testing in higher temperatures. The Esch, D.C. 1972. Control of Permafrost Degradation
polystyrene sections had the greatest deforma- Beneath a Roadway by Subgrade Insulation. State of
Alaska Department of Highways.
tion between all the sections. Field, D., T. Ruban, A. Johnston, D. Nixon and D.
3. The subgrade back-calculated Mr showed its Lewycky. 2011. Edmonton Experience with Bottom
maximum value in the 10-cm thick polystyrene Ash and Other Insulating Materials for Mitigation of
sections, followed by the P-5, B.A. and the control Frost Heave Induced Damage in Pavements. In Pro-
section. ceedings of the 2011 Annual Conference of the Trans-
4. By increasing the pavement temperature, the portation Association of Canada. September 11–14.
HMA modulus decreased, which resulted in Edmonton, Alberta.
decreasing the calculated effective pavement Havukainen, J. 1987. The Utilization of Coal Ash in Earth
modulus reducing in all sections. Works. In Advances in Mining Science and Tech-
nology, Vol. 2: Reclamation, Treatment and Utiliza-
5. Although the polystyrene board insulated sec- tion of Coal Mining Wastes, pp. 245–252. Elsevier,
tions had the strongest subgrade in thaw season, Amsterdam.
they also showed the lowest pavement bearing Hayden, R.L., and H.N. Swanson. 1972. Styrofoam
capacity as a result of the low strength of the Highway Insulation on Colorado Mountain Passes.
polystyrene board. Colorado Department of Highways.
6. Bottom ash could protect the subgrade from the Hein, D.K., and F.W. Jung. 1994. Seasonal variations in
effects of freezing, and at the same time, it did pavement strength. In 4th International Conference,
not have any adverse effect on pavement bear- Bearing Capacity of Roads and Airfields, vol. 1.
ing capacity. Ovik, J., B. Birgisson, and D.E. Newcomb. 1999. Charac-
terizing seasonal variations in flexible pavement material
The results of this study came from monitoring properties. Transportation Research Record: Journal of
the test road for one season before opening the road the Transportation Research Board 1684.1, 1–7.
to traffic. The monitoring should continue so as to Sebaaly, P.E., P. Schoener, R. Siddharthan, and J. Epps.
evaluate the pavement performance of different sec- 1994. Implementation of Nevada’s Overlay Design
Procedure. In 4th International Conference, Bearing
tions under real traffic loads and several freeze-thaw Capacity of Roads and Airfields, vol. 2.
cycles. Tatarniuk, C., and D. Lewycky. 2011. Case Study on the
Performance of High Load Polystyrene as Roadway
Insulation in Edmonton, Alberta. In 2011 Conference
ACKNOWLEDGMENTS and Exhibition of the Transportation Association of
Canada. Transportation Successes: Let’s Build On
The authors would like to thank Alberta Trans- Them. 2011 Congress et Exhibition de l’Association
portation, the City of Edmonton, and Alberta des Transports du Canada. Les Succes en Transports:
Recycling for their financial and in-kind support Une Tremplin vers l’Avenir.
Tavafzadeh, N., S. Nassiri, M. Shafiee, and A. Bayat.
of this project. In addition, thanks is extended 2014. Using Field Data to Evaluate Bottom Ash as
to ISL Engineering, Land Services, and DeFord Pavement Insulation Layer. Transportation Research
Contracting for coordinating the construction and Record: Journal of the Transportation Research
instrumentation activities of the IRRF’s test road Board 2433, no. 1: 39–47.
and to Ms. Sheena Moore for technical writing Transportation Officials, 1993. AASHTO Guide for
support. Design of Pavement Structures. Vol. 1. AASHTO.

837
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Evaluation of cement-bitumen treated material bearing capacity

by in-situ tests

M. Meocci & F. La Torre

Department of Civil and Environmental Engineering, University of Florence, Florence, Italy

ABSTRACT: Cement-Bitumen Treated Materials (CBTM) are widely used for the rehabilitation of
existing pavements and for the construction of new pavements as they offer several structural and envi-
ronmental benefits.
The structural performance of CBTM combines the advantages of ductility, typical of bituminous
mixtures, with the high resistance typical of cement mixtures. Furthermore there is the advantage of pro-
ducing a cold mixture using reclaimed asphalt in percentage that can reach 80–90% of total aggregates.
Although CBTMs have been used with great success for several years, their structural performance has
not been investigated in sufficient detail, and their structural design is mainly based on the experience of
practitioners. For this reason, an extensive experimental study has been conducted to evaluate the struc-
tural characteristics of this materials.
The experimental site of the “Pedemontana Lombarda” motorway in Lomazzo allowed to evaluate the
CBTM bearing capacity evolution over time. For this purpose the dynamic elastic modulus was measured
by means of LWD and FWD tests carried out during a 90 days curing period.
The evolution of the dynamic modulus over time for LWD tests can be described by means of alogistic
function. The model provides a good prediction of the dynamic modulus evolution in time with a R2 of
the fitted functions greater than 0.9. The validation of the model on a different dataset confirm the valid-
ity of the proposed model and excluded over fitting issues.

1 INTRODUCTION AND OBJECTIVES can reach also the 100% of the aggregates weight
in the production of cold bituminous mixture
Nowadays the road construction and maintenance (treated materials with foamed bitumen or bitu-
requires increased efforts to ensure high structural minous emulsion). This type of material, usually
and environmental pavement performances. defined cement bitumen treated material (CBTM),
The first solution of this problem is represented is produced, both on-site and in establishment.
by the replacement of the granular layers with The structural performance of CBTM is not
hydraulically stabilized layers generally made with affected by the use of the reclaimed asphalt. The
virgin aggregates. Sometimes base or sub-base layers production technology allows to obtain an excel-
can be realized using limited percentage of recycled lent mixture that combines the strength of the
aggregate (generally with cement stabilization in- cement to the ductility of the bitumen, limiting
situ), even though the quality of the different com- pavement distresses over time.
ponents could be rather poor and inhomogeneous. Literature on CBTM provides limited infor-
Unfortunately, during road construction and mation about mechanistic analysis and design
maintenance a large amount of material from dem- despite the great success of the technology across
olition of existing pavement remains unused. There- the world. Consequently, up to 10 years ago, pave-
fore, the need to achieve environmentally compatible ments realized with a layer of CBTM were based
and structurally performing materials addressed the only on the experience of practitioners and did not
research towards the production of a material that consider the equations governing the evolution of
was able to meet both these requirements. the materials performance over time (Liebenberg &
To date, Reclaimed Asphalt (RA) is currently Visser 2004).
used in the production of bituminous mixtures CBTMs are currently widely used in mainte-
in different percentages. The RA is used in lim- nance as well as in the construction of new infra-
ited percentage (generally less than 30% by weight structures, like in the “Pedemontana Lombarda”
on the mixture) in the production of traditional motorway pavement. The material, produced by
asphalt concrete (ANAS 2009). Thanks to the cold adding cement and over-stabilized bituminous
recycling technique, considered as a common con- emulsion to RA, was used in place of the tradi-
struction/maintenance technique, the use of RA tional cement treated material as sub-base layer.
839
 The research conducted was aimed at providing method. Literature research showed that the ITT is
a comprehensive characterization of the structural the most common test used for the evaluation of
performance of the mixtures through an extensive the stiffness modulus of cylindrical specimens for
in-field test campaign. The mechanical properties of different types of asphalt concrete materials manu-
the mixture and the evolution of the bearing capac- factured with a gyratory press (Thomas 2007).
ity of the CTBM sub-base course over time were Kim et al. (2008) assessed the dynamic modulus
analyzed. The design mixture was laid down in two and the static creep of the CBTM by means of the
trial sections realized in the Pedemontana Lombarda repeated load test in order to characterize the dif-
motorway work zone in May 2014. The two trial sec- ferent effect of load intensities and tests tempera-
tions (20 cm and 15 cm thick) were laid down directly tures. As in the ITTs, the tests were carried out in
on the cement stabilized and compacted subgrade. 150 mm diameter specimens gyratory compacted.
The two structures have been characterized by In Perez et al. (2013) a state of the art concern-
means of numerous in-situ and laboratory tests. ing the mechanical properties of the bituminous
In order to evaluate the evolution of the bearing emulsion stabilized material was carried out. The
capacity in time, a large number of Light Weight research aimed at reviewing the mechanical prop-
Deflectometer (LWD) and Falling Weight Deflec- erties and parameters used in the structural design
tometer (FWD) tests were conducted. The results due to the limited knowledge in this topic. The
achieved with the in-field tests have been used to CBTMs were analyzed according to the AASHTO
develop a model describing the evolution of the TP-62 test protocol in order to define a master
bearing capacity of CBTM over time. curve of the material.
To date, several laboratory experiments that
were aimed at evaluating the fatigue performances
2 BACKGROUND of the CBTMs have been conducted in recent years
(Bocci et al. 2012, Del Campo 2003, Gulyas et al.
In the last years, cement-bitumen treated materials 2006, Perez et al. 2013, Thomas, 2007). Studies have
represented a great solution to solve environmental been conducted to determine the elastic modulus
and structural issues concerning road pavements of the material carried outwith the ITT (Meocci
construction and maintenance. et al. 2016, Grilli et al. 2016) and to define the
This success has been achieved by road engineers fatigue curves or the compressive strength, flexural
by adding small quantities of cement and bitumen strength, CBR, Dynamic Creep and wheel tracking
emulsion to recycled aggregate (RAP) to obtain a in lab cured specimens (Baghini et al. 2015).
good quality material for road bases and sub-bases. Researches focused on the evaluation of the
This technology has been used earlier in the structural performances of the CBTM by means
rehabilitation of existing pavement and later for of in-situ tests performed directly on the road
new constructions (Bocci et al. 2011) but the struc- pavements as light weight deflectometer and fall-
tural performance of the resulting materials has not ing weight deflectometer tests are still missing.
been fully investigated (Liebenberg & Visser 2004).
The first literature research conducted between
1975 and 1990 aimed at defining transfer functions 3 MATERIAL
for CBTMs. Initially the structural design of the
mixture was conducted treating the material simi- 3.1 Aggregates
lar to the hydraulically stabilized materials (Otte & The cement bitumen treated mixture is produced
Marais 1979). Later, the structural design of CBTM by using both Reclaimed Asphalt (RA) and virgin
has been regarded as traditional asphalt concrete aggregate.
materials due to the high percentage of bitumen The recycled aggregate used in this trial was com-
emulsion added into the mixture (Santucci 1977). posed by two different types of Reclaimed Asphalt
A transfer function was developed by De Beer Pavement (RAP). Complying with EN 13108-
& Grobler (1993) by means of a Heavy Vehicle 8:2016 (EN 2016), the RA was defined as RAP 0/20
Simulator (HVS). In (Liebenberg 2002) and in and RAP 20/40. Since both of them were lacking in
(Robroch 2002) the HVS test has been identified the fine component, sand and filler were added in
as the ideal procedure to assess the bearing capac- order to obtain the correct design mixture grada-
ity of CBTMs. However, the HVS test is usually tion. In Figure 1 an example of gradation of each
limited to a few repetitions because of the long aggregate component used in the mixture is shown.
time and the high cost required to perform each
test. Laboratory test are then used to improve the
results obtained from in-field tests. 3.2 Bituminous emulsion
Kavussi & Modarres (2010a,b) estimated the An over-stabilized non modified bituminous emul-
stiffness modulus of materials by means of the Indi- sion was used to realize the sub-base course.
rect Tensile Test (ITT) according to ASTM D4123- The over-stabilization allows to delay the emul-
82 (ASTM International 1995) standard testing sion breaking and to act as a lubricant during the
840
 0,01 0,1 I
Sicvesize[mmj
Figure 2. Composition of the design mixture.
100 ~c===~7.=~==~~~,---~~TII--~~~
Figure 1. Gradation of different component of the - P reliminary gradation
90 analysis
mixture. - - In-site gradation analysis
80
• • • Design gradation for
70 i ~-r~C~B~T~M~su~b~-b~a~~co~ur~se~ 1
Table 1. Characteristics of the bituminous emulsion. ~ 60
~
~50
Standard Unit of
Parameters reference Values measure ·~ 40
~ 30
Polarity EN 1430 + – 20

Bitumen content EN 1428 ≥58 % 10

Settling tendency EN 12874 ≤10 % 0 +--L~~~--~~~+-~~~~~--~~

0,01 0,1 10 100

@7 days
Sieve size (mm]
Breaking index EN 13075-1 >170 –
Mixing stability EN 12848 <2 g
Efflux time EN 12846 15–70 s Figure 3. Preliminary mix design gradation
(2 mm@40°C) characterization.
Efflux time EN 12846 – s
0.5 . . . - - - - - ; - - - - - - - : - - - - - ; - - - - - ,
(4 mm@40°C)
0.45
0.4
Specification limit + 0.36
0.35 ---------------- ~----------
Table 2. Characteristics of extracted bitumen. 'i' 0.3 0.33 0.34
~
2; 0.25
Standard Unit of ;; 0.2
Parameters reference Values measure t:: 0.15
0.1
Polarity EN 1426 ≤100 dmm 0.05
Bitumen content EN 1427 ≥ 43 °C
0.5 1.5 2.5 3.5 4.5
Bituminous emulsion content (%]

compaction process. These features allow to obtain Figure 4. Preliminary ITS test results.
along workability time and therefore to achieve a
good stability for the sub-base course realized. The The selection of the proper design mixture was
technical characteristics of bituminous emulsion divided in to two phases. The first one aimed at deter-
used are summarized in Table 1. mining the gradation of aggregates to maximize the
In the Table 2 the characteristics bitumen RA component in a mix compliant with the Design
extracted from RA are summarized. Specification; the second one was aimed at determin-
ing the optimum content of bitumen, cement and
3.3 Cement water. The specimens were manufactured according
According to the UNI EN 197-1 (UNI EN 2011) to the gyratory compactor (GC) test protocol (EN
a Portland cement type II/B-LL, 32.5 R class 12697-31:2007) and tested according to the Indirect
strength was selected to prepare the CBTM to be Tensile Test protocol (EN 12697-23:2003).
used as a sub-base layer. In the preliminary mix The different content of each type of aggregate
design different cement contents (2.0%−3.0%) were is shown in Figure 2. In Figure 3 the preliminary
evaluated. mixture gradation is shown.
The preliminary characterization (solid curve)
shows a gradation slightly lacking in fine materials
3.4 Mix design despite the filler addition. In-site gradation analy-
A preliminary laboratory analysis was carried out sis shows instead the full compliance with design
in order to establish the composition of the mixture. gradation (dashed curve).
841
 300 mm diameter). A load cell placed on the top of
the load plate measures the load transferred to the
pavement surface.
Deflection sensors (geophones), placed radially
from the center of the load plate, measure the pave-
ment deformation in response to the load applied
on the surface. Offsets used in this trial are: 0 mm,
200 mm, 300 mm, 450 mm, 600 mm, 900 mm,
Figure 5. Paving and compaction. 1200 mm and 1500 mm from the plate center.
The FWD tests have been performed in accord-
To satisfy the volumetric and mechanical prop- ance with the ASTM D4694-96 “Standard test
erties defined in the design specification for a cold- method for deflections with a falling weight type
recycled sub-base course, 2.5% of cement and 3.5% impulse load device” (ASTM International 2003)
of over-stabilized bituminous emulsion have been and ASTM D4695-03, “Standard guide for general
selected. In Figure 4 the ITS values obtained on pavement deflection measurements” (ASTM Inter-
the specimens tested according to the EN 12697-23 national 2015).
(EN 2003) are shown. Only the specimen manufac- The stiffness values of each layer of the
tured with 3.5% of bituminous emulsion satisfies pavements were obtained by means of the
the design specification equal to 0.35 MPa. back-analysis performed with BAKFAA
v.2/2012software (FAA 2015) by the “DIMMS
Control S.p.A.”
4 TRIAL SECTION CONSTRUCTION

Two layers of cement-bitumen treated material 5.2 The Light Weight Deflectometer test
were laid down in work zone located in Lomazzo A Light Weight Deflectometer (LWD) is a port-
(Italy) directly on the cement stabilized and com- able falling weight deflectometer (Dynatest 2014b).
pacted subgrade. The subgrade was characterized The LWD is a lighter and faster nondestructive
by a 160 MPa equivalent dynamic modulus meas- testing device used for the evaluation of the pave-
ured with LWD. ment bearing capacity. The LWD determines the
The CBTM was manufactured directly on-site strength and the stiffness of pavement materials
using a semi-mobile mixer. The RAP from the by measuring the material’s response (deflections)
demolition of existing pavements was character- under the impact of a 20 kg load dropped from a
ized in terms of gradation and corrected with the height of 100 cm on a plate placed on the surface
addition of fine material (about 20%). The differ- of the pavement.
ent components of the mixtures (over-stabilized The LWD tests have been performed in
bituminous emulsion, cement and recycled aggre- accordance with ASTME2583-07” Standard Test
gates) were then mixed and laid down by a tradi- Method for Measuring deflections with a Light
tional paver (see Figure 5 – left). Weight Deflectometer (LWD)” (ASTM Inter-
Two sub-base trial sections were realized: the national 2015) by applying astressof approxi-
first one was 20 cm thick and the second one mately 70 kPa with load pulse durationof about
was 15 cm thick. The CBTM was immediately 30 msec.
compacted with pneumatic tyre compactors and The deflections were measured using three geo-
vibratory rollers (see Figure 5 – right). The curing phones placed in the center of the plate and 20 cm
period was about 90 days. and 30 cm apart from it.

5 TEST METHODS 5.3 Experimental program

The LWD measurements (ASTME2583-07:2015)
5.1 The Falling Weight Deflectometer test and the FWD measurement (ASTM D4694-96:203
The Falling Weight Deflectometer test (FWD) is a and ASTM D4695-03:2015) were performed in the
non-destructive test used to evaluate the physical trial sections with the frequency summarized in
properties of pavements. FWD is primarily used Table 3.
to estimate pavement structural capacity (Dynatest During the in-situ tests the mixtures were also
2014a). characterized by means of a series of labora-
The FWD is designed to apply a load pulse tory tests in order to evaluate volumetric proper-
(ranging from 10 kN to 120 kN) to the pavement ties, stiffness modulus, indirect tensile stress and
surface in order to simulate the load caused by a fatigue resistance in specimens manufactured with
rolling vehicle wheel. The load, produced by drop- the gyratory press according to the EN 12697-23
ping a large weight, is transmitted to the pavement (EN 2003) and UNI EN 12697-26 (EN 2012) as
by means of a circular load plate (typically with a described in Meocci et al. (2016).
842
Table 3. Characteristics of extracted bitumen. Table 4. Dynamic modulus @20°C –15 cm thick.
Tests Tests frequency N1 E1 [MPa] E2 [MPa] E3 [MPa] Average [MPa]
1 2
LWD 1h 3.5h 24h 7d 14d 21d 30d 60d 90d 1h 133 133 – 133
FWD 29d 90d 3.5h 149 149 – 149
24h 257 294 237 263
1
h = hours; 2d = days.
7d2 607 578 565 583
14d 808 730 754 764
6 TEST RESULTS 21d 891 848 872 870
30d 994 949 970 971
6.1 LWD test results 60d 1098 980 996 1025
The LWD tests were performed approximately in 90d 1103 983 1001 1029
the same areas of each trial section.
The elastic modulus of CBTM strongly depends
1
h = hours; 2d = days
on the air temperature, therefore the values
Table 5. Dynamic Modulus @20°C –20 cm thick.
obtained from LWD measurements (referred to the
average temperature of the sub-base course) have N1 E1 [MPa] E2 [MPa] E3 [MPa] Average [MPa]
been translated to a reference temperature equal to
20°C with the empirical relationship described in 1h 148 165 – 157
the equation (1) (Bocci et al. 2011). 3.5h1 167 – – 167
24h 354 399 322 359
Ereef Et × 10[ −00.00815(Trerefef T )]
(1) 7d2 673 716 686 691
14d 970 1009 1054 1011
The dynamic elastic moduli at reference temper- 21d 1189 1128 1209 1175
ature are summarized in the Table 4 and in Table 5 30d 1230 1279 1308 1245
for 15 cm thick and 20 cm thick trial sections 60d 1313 1297 1301 1298
respectively. The values in the tables are measured 90d 1319 1312 1312 1305
in the different drop E1, E2 and E3, the correspond-
ing average values are given in the last column. 1
h = hours; 2d = days.
In Figure 6 dynamic elastic modulus increase
over time is shown for the trial section 15 cm thick 1600 -r;===c:======c==:==:======:=c===~=:==;c:==;c:====;-==---:-=--,
• Dynamic elastic modulns - trial section #I - 20 em I

and for the 20 cm thick respectively (average values F I4oo • Dynamic elastic modulns - trial section #2 - 15 em
i
of the 3 drop). ; 1200

The two structures show a similar trend. The ~

:;,;
1000 • + •

increase of the stiffness modulus is character- ,., 800

ized by two phases: in the first phase, extended [;:j 600

-~ 400
up to 30–40 days, a steady growth occurs. In the s
second phase, the modulust ends to stabilized to 200

the final value of approximately of 1300 MPa and lh 3.5h 24h 7 days 14 days 2 1 days 30 days 60 days 90 days
1000 MPa for trial section #1 and for #2 respec- Time (days]

tively. The modulus growth is faster in the trial sec-

tion with greater thickness.
Figure 6. Evolution of the LWD modulus over time.
Therefore, the modulus increase over time can
be described by means of a logistic function (2).
The generalized model proposed shows a very
1 good fit to the dynamic elastic modulus evolution
E (t ) = (2) over time for this type of sub-base layers. The model
1 + a bt
is characterized by a coefficient of determination R2
In the equations, E is the dynamic modulus at greater than 0.9. The parameters of the statistical
the curing time t [days] and a and b are experi- analysis performed are summarized in Table 6.
mental constant depending on the materials
characteristics.
In order to develop a function able to describe 6.2 Model validation
the evolution of the dynamic elastic modulus On September 2014 two new trial sections (15 cm
independently of the final value and therefore and 20 cm thick respectively) have been realized in
applicable to the different structures, data were the same work zone.
normalized towards the maximum value assumed The mixtures were characterized by a 2.5%
by the dynamic modulus. of cement type II/B-LL 32.5 R (EN 197-1) and
In Figure 7 the model proposed for the normal- 3.3% of over-stabilized bituminous emulsion.
ized dynamic elastic modulus is shown. The performances of these mixtures are similar to
843
 U
~
L2 tr=~===7================~,---~--.

l
2000 1" Dynamic elastic modulus- trial section #1 _ 20 em
• Dynamic elastic modulus- newtrial section # l -20cm
1F -
---r-t-
~1
1800 • Dynamic Elastic Modulus - trial section #2- 15 em
~ • Dynamtc Elastic Modulus- new trial section #2 - 15 em I I
~ ~ 1600 I I T
~ 08
'C
i= :::: 1 •
I :L· I. I ·I
~ 06
-~
§
g.oA
l~::: _JEt::r~
~
· I
• 1-

::!
-~ 6oo rr;·. .I I II
• I

I
I

';0 .2

~
s 400
t. I I I I II
z 20: I II
0.1 10 40 60
100
Tlme[days]
Time[days]

Figure 8. Comparison between LWD test results in the

Figure 7. Evolution of the elastic dynamic modulus
different test campaign.
over time.

Table 6. Statistical parameters of regression models. ~I CZJEXP-OBSModulus ~I [EXP-OBSModul

2 us J
1 .. . . . 1 . . . . ..
Parameters Logistic model
1 .
0 ~

1 ..
0

Equation E(t) = [1+3.037(0.890t)]–1 z • z •

0 . 0
R2 0.973 0 Normallzed Expected data I 0 Norma lized Expecled dala

Standard Error 0.599

F(1,16) 581.003 Figure 9. Comparison between expected and observed
p-value 0.05 data.

those of the first CBTM mixture and satisfy the The chi-square result is summarized in Table 7.
requirement of the design specification (ITS test). The model is therefore statistically significant at
The results obtained by LWD in this test cam- a 99% confidence interval.
paign were translated at a reference temperature of
20°C and subsequently compared with those already 6.1 FWD test results
obtained and analyzed (see § 6.1.). In Figure 9 the The experimental data evaluated in this section
comparison between the two different LWD test refer to the FWD tests conducted on the trial sec-
campaigns is shown. The black indicators show the tions 29 and 90 days after paving. The values shown
result of the first test campaign while the grey indi- in this section were kindly provided by the Dimms
cators show the results of the second test campaign. Laboratory (DIMMS Control S.p.A. 2014).
No significant differences are observed for the The on-site characterization provided the meas-
15 cm thick trial sections (black and grey dots). ure of the deflection of the sub-base course in 10
The two series of modulus values align perfectly points aligned along the center-line of both trial
along the same curve (Figure 8). Some differences sections. The FWD deflection measurements range
between the measured modulus values can be from 3 m to 5 m. The load pulse was defined as a
observed instead for the 20 cm thick trial sections function of the layer thickness of the trial sections.
(black and grey squares). The maximum value is 100 kN.
The new data collected were used as a validation In the first test campaign, conducted at 22°C,
sample in order to evaluate the goodness of fit of the deflection measured in the geophone2500 mm
the model proposed. Unfortunately the validation far from the center plate converges approximately
is referred only to the first period of curing. at 50 mm. The second test campaign, carried out
Therefore, the new normalized sample was fit- after 90 days of curing time at 26°C, shows very
ted with the logistic model. For the evaluation of similar results. Overall, the performance of thesub-
goodness of fit two different tests was conducted: grade is therefore homogenous over time.
R2and chi-square. A back-analysis was performed to determine
The R2 values obtained are equal to 0.76 and 0.73 the elastic modulus of the sub-base course and the
for 15 cm and 20 cm trial sections respectively. More subgrade at the test temperature. The process was
than 70% of the variance is therefore explained by carried out by means of BAKFAA v.2.0 (FHA
the models developed with the first dataset. 2012). The tool allows evaluating the elastic modu-
In Figure 9 a comparison between expected and lus by best-fitting process based on the minimum
observed data referred to the proposed models is mean square error.
shown (left for 15 cm trial section and right for The results obtained were translated by means
20 cm trial section). The values are spread around of the equation given in (1) at a reference tempera-
the diagonal of the graph confirming that the ture of 20°C. In Table 8 and in Table 9 the CBTM
models show a good fit to the validation samples. modulus obtained from the back-calculation
844
procedure are summarized. The bold values cor- Table 9. Results of back calculation – 20 cm thick trial
respond to the test deflection measures out of the section.
trend, therefore they were not considered in the
Modulus @20°C Modulus @20°C
back-analysis process carried out for the definition 29 days of curing 90 days of curing
of the bearing capacity of the CBTM layers. Position [MPa] [MPa]
The average dynamic modulus of the 15 cm
thick CBTM after 29 days of curing was about 1 5044 3211
5200 MPa, while the average value after 90 days was 2 4681 5482
about 8000 MPa. The dynamic modulus increases 3 4600 8322
of around 50% with the increase of the curing time. 4 5232 8239
A similar evolution was recorded for the 20 cm 5 7769 10186
6 6660 6936
thick layer. The average value of the dynamic elas-
7 6078 6891
tic modulus after 29 days of curing was about
8 4186 6443
5600 MPa. After 90 days of curing the value 9 – 9410
increased of about the 40%. Moduli obtained from Average 5601 7739
the back-analysis process are therefore homogene-
ous in the two different trial sections.
The subgrade modulus obtained in the back-cal- 7 CONCLUSIONS AND
culation process is consistent with those obtained RECOMMENDATIONS
in the LWD test and it’s about 180–200 MPa.
Unfortunately, the FWD values were referred This paper describes an extensive in-situ character-
only to29 and 90 days of curing. Therefore, the lim- ization concerning the evaluation of CTBM bear-
ited number of FWD measurements does not allow ing capacity. Two trial sections have been realized
to find a modelable to describe the growth trend on a cement stabilized subgrade. The first one was
of dynamic modulus as a function of curing time. 15 cm thick and the second one was 20 cm thick.
It should be noted that the moduli obtained Several LWD tests were performed in order to
from LWD and FWD tests cannot be compared. evaluate the evolution in time of the performance
The FWD test provides the value of dynamic of this type of mixture. The tests were carried out
modulus of each layer determined by means of in the same areas at different curing time aimed at
back-analysis. The LWD test, instead, provides the defining a numerical model describing the evolu-
equivalent dynamic modulus measured on the sur- tion of the elastic modulus over time.
face of the pavement (composite modulus of the The model proposed shows a good fit to the
CBTM and the subgrade). calibration dataset. The validation procedure based
applied the LWD measurements in two different
Table 7. χ2 goodness of fit test. trial sections confirms the goodness of fit of the pro-
posed model also when applied to a different dataset.
Parameters Values LWD test result showed that the dynamic mod-
Degree of Freedom (DoF) 23
ulus increases with the increase of curing time with
Level of confidence 99% 10.20 a logistic trend and after 30–40 days of curing the
χ2 0.917 stiffness tends to the final value (1000–1500 MPa)
Null hypothesis is accepted if 0.917 < 10.20 The two FWD tests performed after 29 and
90 days of curing show that even after 29 days
from the construction time the stiffness of the
Table 8. Results of back calculation – 15 cm thick trial
section.
material keeps growing. The increase is approxi-
mately equal to 40–50% of the 30-days value. The
Modulus @20°C Modulus @20°C maximum value is around 8000 MPa.
29 days of curing 90 days of curing Unfortunately, the limited number of FWD tests
Position [MPa] [MPa] performed on the trial sections did not allow to
define a predictive model to evaluate the long-term
1 2795 9320
modulus evolution. Long term FWD test should
2 4604 10860
be conducted to develop also models from this type
3 5696 10173
4 4729 7817
of test allowing to consider the single layer rather
5 2977 6445 than the composite moduli.
6 5603 6857
7 6728 6336
8 7000 6361 ACKNOWLEDGEMENTS
9 – 6182
10 – 7291
The authors wish to thank Iterchimica, Auto-
Average 5158 8118 strada Pedemontana Lombarda, STRABAG and
DIMMS Control SpA for providing the materials,
845
the analysis of the mixtures, the data collected in EN 13108–8:2016, Bituminous mixtures. Material specifica-
the LWD and FWD tests. tions. Reclaimed asphalt, (2016).
A special thanks to the 3TI Progetti S.p.A. for sup- Federal Aviation Administration (FAA), BACK FAA
porting the research on CBTM characterization. v.2/2012, http://bakfaa.software.informer.com/, last
access 30.01.2015.
Grilli, A., Cardone, F. & Bocci, E., Mechanical behavior
of cement-bitumen treated material containing differ-
REFERENCES ent amounts of reclaimed asphalt, European Journal of
ANAS S.p.A.- Capitolato Speciale di Appalto – Norme Environmental and Civil Engineering, Published on-line
Tecniche, ANAS S.p.A, (2009). 10.08.2016, (2016). doi.org/10.1080/19648189.2016.121
ASTM D4123–82 (1995), Standard Test Method for Indi- 9972.
rect Tension Test for Resilient Modulus of Bituminous Gulyas, A., Boromisza, T., Torocsik, F., & Williams, Expe-
Mixtures (Withdrawn 2003), ASTM International, West rience of Cold In-place Recycling in Hungary, Proceed-
Conshohocken, PA, 1995, www.astm.org. ings of transportation research board 85th annual meeting
ASTM D4694–96 (2003), Standard Test Method for (CD-ROM), 0412, Washington, D.C., (2006).
Deflections with a Falling-Weight-Type Impulse Load Kavussi, A., & Modarres, A., A model for resilient modu-
Device, ASTM International, West Conshohocken, PA, lus determination of recycled mixes with bitumen emul-
2003, www.astm.org. sion and cement from ITS testing result, Construction
ASTM D4695–03 (2015), Standard Guide for General and Building Materials, 24, 2252–2259 (2010b). doi.
Pavement Deflection Measurements, ASTM Interna- org/10.1016/j.conbuildmat.2010.04.031.
tional, West Conshohocken, PA, 2015 www.astm.org. Kavussi, A., & Modarres, A., Laboratory fatigue model for
ASTM E2583–07 (2015), Standard Test Method for Meas- recycled mixes with bitumen emulsion an cement, Con-
uring Deflections with a Light Weight Deflectometer struction and Building Materials, 24, 1920–1927 (2010a).
(LWD), ASTM International, West Conshohocken, PA, doi.org/10.1016/j.conbuildmat.2010.04.009.
2015, www.astm.org. Kim, Y., Lee, H., & Heitzman, M., Laboratory Evaluation
Baghini, M.S., Ismail A., & Bin Karim, M.R. Evaluation of Cold In-place Recycling Mixtures using Foamed
of cement-treated mixtures with slow setting bitumen Asphalt Based on Dynamic Modulus and Repeated
emulsion as base course material for road pavement, Dynamic Load Tests, Proceedings of transportation
Construction and Building Materials, pp. 323–336, research board 87th annual meeting (CD-ROM), Wash-
94(2015). doi.org/10.1016/j.conbuildmat.2015.07.057. ington, DC: Transportation Research Board (2008).
Bocci, M., Grilli, A., Cardone, F., & Graziani, A., A study Liebenberg, J., The Influence of Various Emulsion and
on the mechanical behavior of cement-bitumen treated Cement Contents on an Emulsion Treated Ferricrete
materials, Construction and Building Materials, 25, 773– from the HVS Test Sections on Road P243/1, Trans-
778 (2011). doi.org/10.1016/j.conbuildmat.2010.07.007. portek, CSIR, Contract Report, CR-2001/77, (2002).
Bocci M., Grilli A., Cardone F., & Ferrotti G. Full-depth Liebenberg, J.J.E. & Visser, A.T., Towards a mechanistic
reclamation for the rehabilitation of local roads: a case structural design procedure for emulsion-treated base
study, International Journal of Pavement Engineering, pp. layers, Journal of South Africa Institution of Civil Engi-
191–201, 15(2014). doi.org/10.1080/10298436.2012.6 neers, 46, 2–8 (2004).
57196. Meocci, M., Grilli, A., La Torre, F., & Bocci, M. Evaluation of
De Beer, M., & Grobler, J.E. ETB’s: Heavy vehicle simula- mechanical performance of Cement-bitumen treated mate-
tor (HVS) evaluation of the Heilbron sections, Service rials through laboratory and in-situ testing, Road Materials
Contract Report PR92/2/1, South African Roads Board, and Pavement Design, Published on-line: 17 august 2016,
Pretoria (1993). (2016). doi.org/10.1080/14680629.2016.1213506.
Del Campo, J.F., Recycling in Road pavement, Presented at Otte, E. & Marais, G.P., Plaveiseleksperimentop pad S12,
the 3rd International Symposium on maintenance and 3rd Conference on Asphalt Pavements, Durban, South
rehabilitation of pavement and technologies control, Africa (1979).
Guimaraes, Portugal, (2003). Perez, I., Medina, L., & del Val, M.A. Mechanical prop-
Dimms Control S.p.A., Relazione Tecnica – Indagini erties and behaviour of in situ materials which are
deflettometriche ad alto rendimento con FallingWeight- stabilized with bitumen emulsion, Road Materials and
Deflectometer su “MCAD”, (2014). Pavement Design, Vol. 14, No. 2, 221–238, (2013). doi.
Dynatest ® Pavement Engineering Specialist and Equip- org/10.1080/14680629.2013.779301.
ment, Falling Weight Deflectometer (FWD), http:// Robroch, S., Laboratory Testing on Foamed Bitumen and
www.dynatest.com/equipment/structural/fwd.aspx, last Cement Treated Materials from the HVS Test Sections
access 27.11.2014a. on Road P243/1, Transportek, CSIR, Contract Report,
Dynatest ® Pavement Engineering Specialist and Equip- CR-2001/69, (2002).
ment, Light Weight Deflectometer (LWD), http:// Santucci, L.E., Thickness design procedure for asphalt and
www.dynatest.com/equipment/structural/lwd.aspx, last emulsified asphalt mixes, Proceedings of the 4th Inter-
access 27.11.2014a. national Conference on the Structural Design of Asphalt
EN 12697–23:2003, Bituminous mixtures. Test methods for Pavements, Ann Arbor, Michigan, pp. 424–456 (1977).
hot mix asphalt. Determination of the indirect tensile Thomas, T.W., & May, R.W., Mechanistic-empirical design
strength of bituminous specimens, (2003). guide modelling of asphalt emulsion full depth reclama-
EN 12697–26:2012, Bituminous mixtures. Test methods for tion mixes, Proceedings of transportation research board
hot mix asphalt. Stiffness, (2012). 86th annual meeting (CD-ROM), Washington, DC:
EN 12697–31:2007, Bituminous mixtures. Test methods for Transportation Research Board (2007).
hot mix asphalt. Specimen preparation by gyratory com- UNI EN 197–1:2011, Composizione, caratteristiche e
pactor, (2007). criteri di conformità per i cementi comuni, UNI EN
(2011).
846
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Quality assurance of traffic-speed structural condition surveys

A. Wright, S. Brittain, D. Gershkoff & P. Werro

TRL Limited, Crowthorne, Berkshire, UK

R. Fairclough
Highways England, Manchester, UK

ABSTRACT: Since 2010 the structural condition of the English strategic road network has been assessed
using a Traffic-Speed Deflectometer (TSD) owned by Highways England. Initially, TSD surveys were car-
ried out by the Transport Research Laboratory (TRL) following their evaluation and development of the
prototype equipment on behalf of Highways England. More recently TSD surveys have been carried out
by a commercial operator with TRL acting as supervisor and auditor. In order to ensure that the surveys
deliver robust and reliable data to Highways England, operational procedures and a comprehensive qual-
ity assurance process have been developed for the surveys. This includes accreditation and re-accreditation
procedures for both the survey equipment and the operators together with 1st and 3rd party audit proce-
dures for the survey process. The paper discusses the development and application of these methodologies
and the challenges experienced in their implementation.

1 INTRODUCTION to be used as part of the assessment of mainte-

nance need for identified schemes, in conjunction
Highways England is responsible for managing with more detailed in-depth investigations.
and maintaining the English Strategic Road Net- However, in 2010 Highways England reintro-
work (SRN), encompassing over 55,000 lane km. duced routine assessments of structural condition
The structural condition of the network has been under the annual TRASS (Traffic-speed Structural
one of the primary indicators used by Highways Surveys) contract, now undertaken at traffic-speed
England to determine structural maintenance using the Traffic Speed Deflectometer (TSD)
requirements, to both plan and budget for the (Ferne et al, 2009).
annual maintenance programme, and to periodi- This paper provides a background to the Traffic
cally estimate the value of the asset. Speed Deflectometer and the principles of Accred-
Regular surveys of the structural condition of itation and Quality Assurance of surveys. This
the SRN commenced in 1984, initially over a three is then followed by a discussion of the approach
year interval (Hawker and Ferne, 1992) using the implemented for TRASS, and the experience
Deflectograph. The results of these Deflectograph gained from the implementation.
surveys could be used to provide an annual sum-
mary of the structural residual life of the network
and any required strengthening. Schemes identified 2 THE TRAFFIC-SPEED
for maintenance were typically investigated fur- DEFLECTOMETER (TSD)
ther, with more detailed in-depth techniques such
as coring and trial pits and, from the late 1980 s, Attempts to develop a method of measuring road
using the Falling Weight Deflectometer. deflection at traffic-speed were started as early
In 2000 routine network wide Deflectograph as the 1970 s. One of the earliest devices, which
surveys were discontinued as they were considered almost achieved this objective, was the Curviam-
costly, hazardous to operators and road users, and eter which measures deflection bowls at a survey
caused too much traffic disruption due to their speed of around 20 km/h (Van Geem, 2010). Many
slow survey speed (2.5 km/h). In parallel with this, other devices have been developed since then as
regular comprehensive surface condition surveys summarised by Flintsch et al (2012).
of the SRN were introduced by Highways England In 2005 TRL, on behalf of Highways England,
under the TRACS (Traffic-speed Condition Sur- identified the High Speed Deflectograph, origi-
veys) contract, which have been undertaken rou- nally developed in Denmark by Greenwood Engi-
tinely since 2000. Deflectograph surveys continued neering A/S (Hildebrand and Rasmussen, 2002),

847
as being capable of performing deflection surveys behalf of Highways England, TRL implemented
at traffic—speed. TRL were then commissioned to a Ground Penetrating Radar (GPR) system on
procure and develop the second prototype High Highways England’s TSD for the measurement of
Speed Deflectograph (subsequently renamed as pavement thickness, to be reported as the Major
the Traffic-Speed Deflectometer (TSD)) into a Material Thickness (essentially the overall thick-
survey tool capable of providing routine estimates ness of the bound layers). The GPR system uses
of structural condition of the SRN (Ferne et al, 1.5 GHz and 400 MHz dipolar antennas along
2013). The TSD is shown in Figure 1. with a 4 GHz horn antenna mounted inside a poly-
The TSD contains a complex array of instru- propylene enclosure suspended beneath the TSD
ments and recording equipment within an insu- chassis along the line of the left-hand wheel path.
lated steel container which is mounted on a single
rear axle trailer assembly having a rear axle load of
approximately 10 tonnes. The basic functionality is 3 ACCREDITATION AND QA OF SURVEYS
designed to measure the pavement response under
the rear wheels, in the nearside wheel-path by using Network survey data is typically used to support-
four velocity sensing lasers positioned on a rigid decisions on the need for scheme level surveys and/
steel beam. Three of the lasers are spaced out at or maintenance. It is important that the condition
100 mm, 300 mm and 756 mm in front of the rear data is both accurate and reliable. Inaccurate con-
wheel assembly. The fourth acts as a reference laser dition assessment data could, on one hand, lead to
and is located 3.6 m in front of the rear axle. unnecessary and costly works (or scheme level sur-
Prior to use, the TSD went through a pro- veys) causing needless traffic congestion. On the
gramme of assessment to determine if the device other hand, if maintenance is overlooked, it could
was suitable to support research into the measure- lead to poorly maintained carriageways presenting
ment of structural condition and, ultimately, rou- a risk to road users and inefficient use of mainte-
tine network surveys. The device has since been nance funds.
used to undertake research into the measurement In the UK, the continued reliability of routine
of deflections at traffic-speed, and in 2010 it began survey devices is monitored and checked using a
undertaking network surveys of Highways Eng- combination of Accreditation exercises and on
land’s SRN under their first TRASS (Traffic-speed going quality audits (QA) between Accreditations.
Structural Surveys) contract. These devices include both surface and structural
There are now more than eight TSD devices condition monitoring equipment, such as traffic-
operating worldwide with more expected to be speed profile and image collection equipment (in
delivered in 2017. Greenwood Engineering have the UK under the TRACS and SCANNER con-
continued to develop the TSD methodology, and tracts), SCRIMs, Deflectographs and FWDs. The
therefore the majority of TSD are second genera- process is overseen by an independent Auditor,
tion devices, with additional velocity sensing lasers which is a company appointed for the role by the
and other design improvements. However, the end user of the survey data. Further details of the
basic measurement systems are the same on the accreditation and QA processes applied (includ-
Highways England TSD as with the others. ing the criteria applied) can be found on the UK
In order to provide a more comprehensive Roads Liaison Group website UKRLG (2016).
assessment of structural condition in 2016, on The Accreditation and QA process brings ben-
efits to the end user of the data as it mitigates
against the risk of poor condition data being deliv-
ered. The appointment of an external independent
Auditor means that the necessary expertise does
not need to be maintained by the end-user organi-
sation. This process also helps to ensure that value
for money is obtained with a correctly targeted
maintenance programme.
The process also provides benefits to the survey
contractor. The accreditation process gives the con-
tractor confidence that the equipment is working
correctly at the commencement of the survey, and
the QA process provides confidence that the device
continues to work correctly throughout the survey.
The QA process can also help to identify potential
Figure 1. Traffic Speed Deflectometer. Photo supplied issues before they adversely affect the data or to allow
by FugroGeoServices. them to be fixed earlier and more cost effectively.

848
 The appointment of an independent auditor the differences is undertaken to allow the Audi-
can also provide support to both the end user of tor to identify whether they believe the device to
the data and the survey contractor, to help resolve be performing consistently during the survey sea-
conflicts. son. This process may also be supported using data
from the QA procedure.

4 THE TRASS APPROACH

4.2 Survey contractor QA
Under the TRASS contract model the TSD is The QA undertaken by the survey contractor is
owned by Highways England, maintained by TRL split into three parts, Primary site checks, Second-
(who are also the Auditor and Technical Advisor) ary sites checks and Daily checks. Primary site
and operated by FugroGeoServices Limited. This checks are undertaken close to the survey contrac-
contractual model has proved successful for intro- tor’s base of operations and are used to monitor the
ducing structural surveys to the English strategic ongoing performance of the device since the last
road network. The model defines the roles and Accreditation. Secondary sites are used to check
responsibilities of each party, to avoid unneces- the performance of the device when it is conduct-
sary disagreements and ensuring fairness of costs ing surveys on the network. Daily checks provide
allocations. a check on the general condition of the device and
A programme of Accreditation and QA has also a simple check on the ongoing performance. TRL
been implemented in TRASS to maintain robust have produced software and tools to aid the survey
and reliable survey data. This is carried out via contractor to conduct the check, assess the survey
a monthly Accreditation process supported by data, and determine the performance before leav-
ongoing QA carried out by the survey contractor ing the test site. These checks are further described
and the Auditor. below.

4.2.1 Primary site

4.1 Accreditation
As part of the survey contractor’s QA process the
During the development of Highways England’s contractor is required to set-up a site which is con-
TSD, TRL established a number of road sites to venient for them to survey on a weekly basis. The
enable monitoring of the long term performance purpose of the Primary Site check is to provide a
of the device and provide a means of accrediting robust, in depth, and regular check that the device
it. is performing consistently since the last Accredita-
The first location chosen was a closed test track tion. Reference data is therefore collected on the
site, which allows the testing of the TSD in control- site using the TSD very soon after the successful
led conditions (varying survey speed, accurate test completion of an Accreditation test.
lines etc.). However, this site is limited in size and Each week the survey contractor is required to
is not necessarily representative of the deflection undertake a survey of the Primary site and com-
levels experienced on the network. Therefore this pare the data to the Primary sitereference survey
site has been supplemented by a network route of data to make sure that it is consistent, via a differ-
around 4 km in length. The motorway site selected ence analysis. If it is consistent it is then deemed
for the network route was chosen as it is local to acceptable for continued use. If it is not consistent
TRL, includes several types of road construction then further investigations are undertaken. This
with varying deflection responses, and is used for can involve repeat surveys of the Primary site (in
the Accreditation of other survey devices. This case it was an anomalous run) or returning to the
means that TRL have good knowledge of the site Accreditation sites for more detailed investigation.
and any maintenance that could affect the survey
measurements is quickly spotted and factored into 4.2.2 Secondary sites
the analysis. The Secondary site check has been included in the
To undertake the Accreditation process the TRASS QA to offer the potential for implementing
two sites are surveyed (the test track at a range of a virtually continuous QA assessment process. A
speeds and the network route at a suitable traffic- set of Secondary check sites has been established
speed). The data is then processed, averaged over throughout the network. The slope data collected
50 m lengths and data from the velocity sensing from the previous year’s survey of each of these
lasers (referred to as the ‘slope data’) compared sites has been obtained from the TSD data and
against reference data to make sure that it falls stored on the vehicle as the Secondary site refer-
within certain required criteria. The reference ence data. Each site is defined locationally using
data used is a historical dataset (e.g. the survey the start and end coordinates of the site. At the
at the start of the survey season). An analysis of completion of any survey day the survey crew uses

849
a software tool provided on the TSD to deter- esses. Undertaking QA checks on GPR data is
mine if that day’s survey passed over a Secondary challenging. Tests on the output values (e.g. mate-
site (Figure 2). If it did then the tool extracts the rial thicknesses) require full processing of the data
survey data from that day’s survey, automatically and can be subjective as it requires the interpreta-
aligns it with the Secondary site reference data and tion of the data by a human assessor. Although
compares the two datasets, again using an analysis this would be practical for accreditation testing
of the differences. This provides the survey crews a (where the assessment is carried out externally by
method of checking the performance of the device the Auditor) it would not be practical for carry-
when surveying away from the primary site. If the ing out on-board the vehicle because of the time
examination of the data flags up a potential issue and specialist skills required. Therefore a bespoke
with the device, this data is then provided to the approach has been developed for the TSD survey,
Auditor for investigation. During the investigation which focusses on the raw data.
the survey contractor may be required to return to The accreditation tests have been developed to
the Primary site and or the Accreditation site(s) to assess the two way travel time of the GPR signal
provide additional data for the investigation. over a known distance in air (using a metal plate
reflector). As assessment is then made of the centre
4.2.3 Daily checks frequency and bandwidth of each GPR antenna.
The QA programme also includes a requirement The data is compared, by the Auditor, with ref-
for a series of Daily checks which can provide a erence data established on the commissioning of
regular check on the general condition and meas- the equipment to check that the performance and
urements of the device. behaviour has not changed.
The first part of the daily check is a visual For quality assurance tests a Primary check
inspection of the device to confirm that it is opera- process has been implemented in which the
tional and correctly configured for surveying. TRASS contractor is provided with software
The second part of the daily check is to perform which can carry out an analysis of the GPR data
a “before and after” check on the data provided by which assesses the centre frequency and bandwidth
the device. To achieve this, the contractor selects of each GPR antenna and also assesses the gen-
a site close to the planned testing program for the eral content of the raw GPR data against previous
day and conducts a survey on the site at the begin- survey data collected on the same site. The soft-
ning and the end of the survey day. This data is ware provides the contractor and auditor with a
then checked for consistency. This test provides a performance report for each antenna.
backup for Secondary checks if the day’s survey
did not cover a Secondary check site.
4.3 QA checks conducted by the Auditor
4.2.4 Addition of GPR The objective of independent Auditor role is to
The addition of GPR equipment to the TSD’s monitor, advise and arbitrate, with the contractor
capabilities has required that checks on the GPR taking responsibility for the planning, undertak-
data be added to the accreditation and QA proc- ing and delivering QA checks. It is expected that
this approach will minimise the time from when an
issue occurs to when it is identified and resolved.
However, due to the prototype nature of the
TSD device, we have found that the Auditor must
take a proactive role in the analysing of the data
from the QA checks, and use the lessons learnt to
develop and improve the process. Over time it is
anticipated that this requirement will reduce and
the Auditor will concentrate more on monitoring
the process undertaken by the contractor to make
sure that it is being undertaken correctly, and pro-
vide advice and arbitration where required.

4.4 Survey validation

In addition to the fundamental checks on the
slope data and the GPR data discussed above, the
Figure 2. Secondary check tool showing secondary Accreditation and QA checks include additional
sites identified within the day’s survey (contains OS data validation checks. These include checking that
© Crown copyright and database right 2016). the laser data rates (the proportion of data deliv-

850
ered that is considered as valid) are sufficient, and
that the survey speed is within the required range.
Monitoring is also carried out on the location data
(coordinates), TSD equipment and road surface
temperature etc. This enables surveys (or parts of
surveys) to be marked as invalid.
It is therefore necessary for the Auditor to review
the survey data collected by the survey contractor
to identify the length of data which is valid and
will be usable by the end user of the survey data.
This process serves two roles; the first is to notify
the end user the length of data which is valid for Chainage along site (m)
invoicing purposes. The second is to provide the
survey contractor details of gaps in the survey so Figure 3. Example of bias between repeat TSD datasets
they can target future surveys to fill in these gaps. on the same site.

The identification of the cause of the bias, and a

5 EXPERIENCE IN IMPLEMENTATION means to predict it, are a priority in improving the
quality of the survey data. As there is currently no
The TSD has been found to be quite repeatable practical means to remove this variable bias from
during each single visit to a test site (e.g. repeat sur- the network survey data, the data from the device
veys carried out on the same day). However, at an can currently only be used to provide an indica-
early stage in the introduction of the Accreditation tion of areas of poor structural condition, and
and QA process it was found that, although the improvements will be needed before it can replace
“shape” of the slope data was consistent between Deflectograph surveys entirely.
visits, the offsets, or bias, between the datasets (the It is currently not known how much of the bias
difference in the average values)can be significant arises from changes or drift in the TSD instru-
in comparison to the planned criteria, resulting in mentation, and how much is due to environmen-
failure of the QA criteria. tal effects e.g. pavement temperature. The bias
This made it difficult to assess the ongoing per- seen appears to be consistent along the length of
formance of the TSD as there are two parts to its a survey despite significant changes in pavement
variability. The first is the offset in the data, and construction, i.e. both composite and fully flexible
the second is the variability in the shape of the construction. A bias caused by changes in tempera-
data. It was therefore decided that the data should ture in the pavement would be expected to affect the
be normalised to remove the influence of bias. composite sections rather differently to the flexible
This is achieved by calculating the average slope sections. This suggests that at least a proportion of
values for matching lengths in the two runs and the bias is due to the configuration of the TSD (e.g.
subtracting one from the other. The resulting bias temperature of the equipment or possible unknown
value could then be applied to the survey being calibration aspects). Initial work into investigating
assessed so that the average value of the normal- this found a weak relationship with beam tempera-
ised slopes would match the average value on the ture, but this was not consistent. Tighter control
reference survey. Both the bias and the differences on the beam temperature or, using the temperature
in the shape of the profile are then examined and measurements to correct/estimate the bias, may
assessed. An example showing two separate runs help in reducing the influence of the bias.
with a bias between them is shown in Figure 3. The Highways England TSD is a first genera-
The process implemented for examining the bias tion device. The design has been subject to ongo-
independently from the shape of the data improves ing improvement over the last five years, with new
the robustness of the Accreditation and QA. As we devices employing an updated design and addi-
are currently not able to predict the bias, it is pos- tional sensors. It is possible that second and future
sible that changes in the performance of the device generations of these devices may be more stable
in the routine network survey could occur without with respect to bias. However, research is ongoing
it being picked up in the QA process. However, if to understand this issue.
the device begins to produce anomalous slope data
it is likely that the shape of the slope data would
also change which should be picked up by the cur- 6 SUMMARY AND CONCLUSIONS
rent QA process. In addition, as noted above, other
parameters (e.g. data rates) are also checked, which In 2010 routine surveys of the English SRN began
may provide indications of developing faults. under Highways England’s TRASS (Traffic-speed

851
Structural Surveys) contract. A programme of Ferne, B. W., Langdale, P., Wright, M. A., Fairclough, R.,
Accreditation and QA has been implemented in &Sinhal, R. 2013. Developing and implementing traf-
TRASS to maintain robust and reliable survey fic-speed network level structural condition pavement
data. The process is based on a set of checks car- surveys. Proceedings of 9th Int. Conf on the Bearing
Capacity of Roads, Railways and Airfields, Trond-
ried out before the survey begins (accreditation), heim, Norway, June 2013. Trondheim: Akademia
fundamental checks during the survey (Primary Publishing.
checks) and day to day checks on the collected Flintsch, G. W., Ferne, B. W., Diefenderfer, B., Katicha,
data (Secondary and Daily checks). S., Bryce, J., & Nell, S. 2012. Evaluation of traffic
The process has been successfully applied since speed continuous deflection devices. Paper 12–1670,
2010, with ongoing continuous improvement, 91st Annual Meeting of Transportation Research
including the addition of checks on GPR meas- Board. Washington DC, January 2012. Washington
urements which were introduced to the TRASS D.C.: Transportation Research Board.
survey in 2016. However the process is affected Hawker, L. G. & Ferne, B. W. 1992. Routine assess-
ment of the English trunk road network. Alan Brant
by an unexplained bias that is present in the TSD Memorial Workshop, Institution of Highways and
survey data with regards to slope values. The pres- Transportation, April 1992. London: Institution of
ence of the bias is currently accommodated in the Highways and Transportation.
Accreditation and QA process but work is ongoing Hildebrand, G. & Rasmussen, S. 2002. Development of
to obtain a better understanding of its cause and a high speed Deflectograph. Danish Road Directo-
to better compensate for its effect on network level rate Note No. 117, June 2002. Roskilde: Danish Road
structural condition data. Directorate.
UKRLG2016. Accreditation and Quality Assurance of
Deflectograph Survey Devices, http://www.ukroad-
sliaisongroup.org/en/asset-condition/road-condition-
ACKNOWLEDGEMENTS information/data-collection/deflectograph.cfm.
UKRLG2016. Accreditation and Quality Assurance
This paper has been produced by TRL Limited as of Dynamic Plate Test Survey Devices, http://www.
part of a contract placed by Highways England. ukroadsliaisongroup.org/en/asset-condition/road-
Any views expressed in it are not necessarily those condition-information/data-collection/dynamic-
of Highways England. plate-test-devices-dpt.cfm.
UKRLG2016. Accreditation and Quality Assurance of
Sideway-force Skid Resistance Survey Devices, http://
REFERENCES www.ukroadsliaisongroup.org/en/asset-condition/
road-condition-information/data-collection/skid-
resistance.cfm.
Ferne, B., Sinhal, R. & Fairclough, R. 2009. Structural
Van Geem, C. 2010. Overview of interpretation tech-
assessment of the English strategic road network—
niques based on measurement of deflections and
latest developments. Proc. of 8th Int. Conf on the
curvature radius obtained with the Curviameter.
Bearing Capacity of Roads, Railways and Airfields,
Proc. Of 6th European FWD User’s Group Meeting,
Champaign, Illinois, USA, June/July 2009. Leiden,
Sterrbeek, June 2010. Brussels: BRRC.
The Netherlands: A.A. Balkema Publishers.

852
Pavement serviceability condition
 Bearing Capacity of Roads, Railways and Airfields – Loizos et al. (Eds)
© 2017 Taylor & Francis Group, London, ISBN 978-1-138-29595-7

Non-destructive crack identification for concrete pavements:

A case study

L.S. Salles
University of São Paulo, São Paulo, Brazil

L. Khazanovich
University of Pittsburgh, Pittsburgh, Pennsylvania, USA

J.T. Balbo & A. Cargnin

Universityof São Paulo, São Paulo, Brazil

ABSTRACT: Crack location in concrete pavement has always been a major focus for maintenance and
rehabilitation. However, most cracking evaluation methods emphasize in well-developed surface cracking, not
giving information about bottom-up or half-height cracks still in development. This paper addresses visible
and non-visible (incipient) crack identification in an experimental section of a short Continuously Reinforced
Concrete Pavement (CRCP) constructed in Brazil. As the pavement presented an unusual cracking pattern
(less visible cracks at the slab surface than traditional CRCP), there was a need to verify if most cracks were
not yet fully developed. The ultrasonic testing procedure is described as well as a novel signal interpretation
method, employed to identify freeze and thaw and alkali-silica reaction distresses in concrete slabs in previous
studies. This technique provides a concrete quality indicator for each ultrasonic measurement. Results show
a 100% correlation of the indicator in a continuous scan survey with the visible cracks’ location and indicate
the potential location of 19 cracks not yet apparent at the slab surface. Four potential incipient cracks, firstly
discovered with the indicator, were visually confirmed as surface-visible cracks in later crack surveys.

1 INTRODUCTION methodologies for concrete pavements regard-

ing cracking are solely based on fully developed
Transverse cracking in concrete pavements is cracks showing in the slab’s surface. Several inter-
a primordial point of interest from design and esting studies based on pavement surface image
construction to rehabilitation. In Jointed Plain data provide only information on evident surface
Concrete Pavements (JPCP) transverse cracks are cracks [Huang et al., (2014), Zalama et al. (2014)
induced in the contraction joint and cracking out- and Ying and Salari (2010)]. Therefore, locations
side joints is a sign of early-age failure. Conversely, where the pavement is about to present a crack are
for Continuously Reinforced Concrete Pavements not computed and are neither treated nor included
(CRCP), transverse cracks are part of the struc- in rehabilitation plans. This happens because
ture, as crack spacing and width are important pavement management systems are mostly inter-
inputs for its design while also being performance ested in surface distresses with a greater focus on
indicators. Nevertheless, cluster and meander- rehabilitation than on prevention.
ing cracks are undesired features on CRCP due In order to address this issue, the case study pre-
to their potential for punch out development as sented in this paper shows a simple and reliable pro-
mentioned in Zollinger and Barenberg (1990) and cedure to identify incipient cracks within a concrete
1 Selezneva
3.87 1 2 .8§ 1et3.00
al (2003).
1 However, most
24.03evaluation
m slab using
1 a non-destructive
6 .65"' 1 ultrasonic
9 .60 m device.

I f··t·· l.,.
1 3.87 1 2 .8§ 1 3.00 1 24.03 m
.,) 1
.,,1
6 .65"' 1 9 .60 m

I f··t·· l.,. .,) .,,1

Figure 1. Crack map (June, 2015).

855
2 EXPERIMENTAL CRCP: TESTING SITE where, ôi , k = the image reflectivity of each position
inside the ROI; T = number of transducers positions;
The testing was conducted in one section of a e, r = index of the emitting and receiving transducers;
CRCP constructed in September, 2010 in Brazil. V = shear wave velocity; i, k = index of the ROI’s hor-
As to simulate bus stops, the section was designed izontal and vertical positions; st = the response mag-
with a 50 m concrete slab, short in comparison to nitude at the evaluated time for given transducers;
traditional CRCP. Due to its short length and lack A = apodization coefficient, which is related to the
of anchorage, the short CRCP presented a different various incident angles of the signal (Balmer, 1992).
crack pattern (Figure 1) than traditional CRCP (less For some applications, the SAFT B-scans have
surface cracks and thus a greater crack spacing) as space limitations when the area of interest is big-
seen in Salles et al. (2015). Cracks are numbered ger than that presented by a single B-scan (360 mm
according to the time in which they were visible at wide) which can create misunderstandings if the
the slab surface. The section was designed with a device is not placed directly above the distress or
longitudinal steel percentage of 0.57. More design design feature desired. To improve the SAFT evalu-
and initial performance details can be found in ation, a reconstruction method of more than one
Salles et al. (2015). overlapping B-scan was introduced. Two or more
B-scans are stitched together maintaining the verti-
cal dimensions of the scan while increasing the hori-
3 NON-DESTRUCTIVE TESTING zontal dimensions with the precise spacing between
the overlapping MIRA readings summarized in
The ultrasonic device called MIRA was applied for Equation 3 (Hoegh and Khazanovich, 2015).
the non-destructive ultrasonic testing in the short SAFT panoramic (SAFT-Pan) was applied in
CRCP section regarding the crack map presented this study to search for incipient cracks in the CRCP
in Figure 1 (five years after construction). The non- section.
visible cracks evaluation was performed in panels
between two cracks or between crack and trans-
versal edge. The device was placed longitudinally
m
ôPAN ( x, z )
at the lane center and readings took place every ⎧ m
ôPAN ( x, z ) , if x D
⎪
90 mm. Calibration was performed for every panel. = ⎨maxa (ôPAN ( x, z ) , ôI ( x, z )) , if D x xPAN
m 1 m 1

Also, readings were taken at the visible cracks (F2.1 ⎪ ôIm ( x, z ) , if x xIm (3)
⎩
to F2.5) to check for typical crack features. Scans
at the visible cracks were performed at the crack
where, m = index of the current SAFT scan;
center moving the device five times in the longitudi-
I = individual scan combined to form the new ROI;
nal direction with a scan spacing of 45 mm.
D = distance from the global origin.
3.1 Data processing 3.1.1 Hilbert transform indicator
Ultrasonic readings were processed using the Each individual MIRA measurement results in 45
technique developed by Hoegh et al. (2011)where data signals ranging in transducer spacing of 40
a much more detailed approach can be found. In to 360 mm. Freeseman et al. (2015), analyzing and
brief, the ultrasonic time-history data provided comparing signals from control and freeze-thaw
by MIRA’s 45 transducer pairs at various inci- damaged specimens developed an indicator that
dent angles was processed using the instantane- represents the difference from the undamaged speci-
ous amplitude-based Synthetic Aperture Focusing men to the ones presenting distresses. The research
Technique (SAFT). The information received by noted a shape variation and an excessive oscillation
transducers pairs at spatially diverse locations with damaged specimens’ signals. The indicator was
allowed for reconstruction of the relative reflectiv- able to identify and quantify the level of damage in
ity amplitude within the Region of Interest (ROI) the freeze-thaw specimens. A new study was also
creating a B-scan of the area when applying the able to verify alkali-silica reaction in similar concrete
Equations 1 and 2 (Hoegh et al., 2011). slabs (Clayton et al., 2016).
The signal instantaneous amplitude process-
∑ ( ) ( , zk ) ing through a Hilbert Transform results in a sin-
T 1 T
ôi , k r , e, i, k e ,r i (1)
e 1 r e 1 gle numerical value which is representative of the
signal characteristics. This value was called the
ψ e ,r ( xi zk )
Hilbert Transform Indicator (HTI) and is shown
in mathematical form in Equation 4:
⎛ 2 ⎞
= st xr , xe ⎛ zk2 + ( xi − xe ) + zk2 + ( xi xr ) ⎞ ⎟
1 2

⎝ V ⎝ ⎠⎠ 500 HT (t )
T =∫
HTI d
dt (4)
(2) 0 max ( HT (t ))

856
where HT(t) is the Hilbert transform envelope.
Each transmitting and receiving pair allows
for the calculation of one HTI. The average HTI
of the 45 pairs represents the concrete condition
under one MIRA scan. As such, a higher HTI
value (above 90) represents a damaged concrete,
while a low value indicates a more sound concrete.
The 500 microsecond time window was selected in
order to ensure that the signal’s direct arrival and
subsequent scattering was observed.

4 HTI AT VISIBLE TRANSVERSE CRACKS

The analysis of HTI from the cracks measurement

disclose a high value for all visible cracks, as seen
in Table 1. This investigation was the first time
that the HTI was utilized at transversal shrinkage
cracks. As the concrete presents a physical discon-
tinuity on those locations, it is expected that the
signal shows differences (more attenuation) from
sound continuous concrete signal (Figure 2).
SAFT-Pan reconstructions of scans at cracks
resulted in two major kinds of features, a bro-
ken slab/base back wall (210 mm) with a circular Figure 3. SAFT-Pan reconstructions of transverse
shaped inclusion located at 50 to 100 mm from the cracks (a) F2.2 and (b) F2.1.
surface (Figure 3a) or with continuous slab/base
backwall with another reflection around 150 mm
followed by the same circular shaped inclusion
(both highlighted) described above (Figure 3b).
Both reconstructions in Figure 3 are a pano-
ramic (SAFT-Pan) of five MIRA readings spaced
45 mm.

Table 1. Visible cracks HTI.

Crack HTI

F2.1 112
F2.2 128
F2.3 110 Figure 4. SAFT-Pan reconstruction of panel in between
F2.4 109 edge and crack F2.1.
F2.5 138

4.1 Non-visible cracks

SAFT-Pan reconstructions were performed for
every panel. Figure 4 presents an example (Panel
between transverse edge and F2.2). Two regions
(circled) present similar inclusions as that of trans-
versal cracks as discussed before. A crack visual
survey performed one month after the MIRA test-
ing disclosed a new crack (visible at the slab sur-
face) at the first region indicated by the panoramic.
It is expected that the other region will present a
new crack in the following surveys.
However for some panels the non-visible cracks
Figure 2. Signal amplitude differences in sound and observation was not so easily done. Some mid-
cracked concrete in the short CRCP section. depth reflections can be misleading (possible

857
transverse bars or delamination). For a more com- than the average HTI of the 5 prior and subsequent
prehensive analysis of the panels, the HTI was measurements.
also considered. Figure 5 shows the HTI variation
through the section’s length (50 m) along with the 4.2 Updated crack pattern
crack map.
As can be seen, there is a rise in HTI value at every Figure 6 displays the HTI survey now with the crack
visible crack (F2.1 to F2.5). Therefore, it was estab- map containing the potential non-visible cracks.
lished that whenever a point exhibited a peak of HTI Visible cracks are noted by Fx.y and non-visible
the B-Scan of such point was analyzed for having cracks by Ix.y. Where x is the section’s number and
any of the crack features exhibited in Figure 3. If the y is the crack number according to traffic direc-
crack feature was found the location was considered tion. A visual crack survey done one year after
a non-visible crack. The measurement was consid- the ultrasonic survey disclosed non-visible cracks
ered as an HTI peak when it presented a value higher I-2.14, I2.15 and I2.19 (arrows in Figure 6) as vis-
than the panel’s average HTI and a value higher ible cracks at the slab surface proving the method’s

150

140 F2 5
130 F2.2
120
F2.4 F21
110

..
~ 100

80

..
70

50
10 15 10 15
Lenath(m)
,. . •• 50

3.87 1 218 1 309 1 24Q3m 185m I 180m

Figure 5. HTI survey with June, 2015 crack map.

150

140
12.1
12.19

~
..
~ 100

80

..
70

50
10 15 10 15 lO l5 40 45 50
Lensth (m)

Figure 6. Updated crack map with incipient (I) cracks.

858
 100% a continuous ultrasonic survey indicated the loca-
.,w tion of more 19 cracks in the slab showing the
"•c
80%
technique potential to discover incipient cracks as
w
u
~ 60%
,...J well as to evaluate cluster and Y cracking before
c. the distress becomes apparent in the slab surface.
w Four of those potential crack locations later devel-
~ 40%
•
'3
oped a surface crack confirming the method. The
E 20% updated crack map with all cracks presents a crack
~
u spacing more similar to that of traditional CRCP.
0% It must be noted that the method for finding
0 12 15 18 21 24 27 30 incipient cracks presented in this case study has
Crack spacing (m) productivity limitations for concrete pavement
in highways since the spacing for each continu-
-visiblecracks -o-AIIcracks ous scan is set at 90 mm. Also this predetermined
scan spacing must be strictly followed in order to
Figure 7. Cumulative crack spacing of all and only vis- locate the incipient cracks position correctly which
ible cracks. requires a well-trained operator. Moreover, it is
still not clear if other distresses could create an
HTI peak being misinterpreted as incipient cracks.
ability to indicate the position of transverse cracks Further investigation will be performed to over-
prior to their development to the slab surface. As come these limitations and to evaluate the ultra-
mentioned, crack I2.1 was disclosed one month sonic signal differences between visible cracks and
after the ultrasonic survey. The average difference incipient cracks in order to analyze the crack’s ori-
in location (length) between the real crack location gin and propagation (bottom-up or top-down).
and the HTI incipient crack indication for these
four new cracks was 75 mm. A more severe control
on the ultrasonic survey spacing can improve this ACKNOWLEDGMENTS
variance.
Moreover, the new crack map seems more in The authors gratefully acknowledge the Coordena-
accordance with traditional CRCP. There are only cao de Aperfeicoamento de Pessoal de Nivel Supe-
three panels with length over 5 m. Additionally, rior (CAPES—Brazil Ministry of Education)
several panels present a very small crack spac- for the scholarship granted to the first author
ing (less than 0.9 m). If these incipient cracks are through international program PDSE (Process
individual cracks or part of a “Y” crack remains - 99999.003858/2015-04).
to be seem. Nevertheless, this technique reveals
potential for evaluation of cluster cracking or “Y”
crack areas before the problem is visible at the slab REFERENCES
surface.
The updated cumulative crack spacing for all AMERICAN ASSOCIATION OF STATE HIGH-
cracks and for only surface-visible cracks is shown WAY AND TRANSPORTATIONS OFFICIALS
in Figure 7. The inclusion of the non-visible cracks (AASHTO) 1993.1993. Guide for design of pavement
results in a much more similar crack spacing as that structures. AASHTO, Washington D.C., USA.
found in technical CRCP literature. Now the sec- Balmer, R.A. 1992. Comparison of Range-Doppler and
tion complies with the crack spacing recommenda- Wavenumber Domain SAR Focusing 544 Algorithms.
Geoscience and Remote Sensing, IEEE Transactions,
tion – 0.9 to 2.4 m (straight lines in Figure 7) – of vol. 30: 706–545.
AASTHO (1993). However, a greater potential for Clayton, D.; Khazanovich, L. & Salles, L. 2016. Lin-
cluster cracking is also observed. ear Array Ultrasonic Test Results from Alkali-Silica
Reaction (ASR) Specimens. Report No ORNL/
TM-2016/159. Oak Ridge National Laboratory. Elec-
5 FINDINGS trical and Electronics Systems Research Division, US