José P.

Aguiar-Moya
Adriana Vargas-Nordcbeck
Fabricio Leiva-Villacorta
Luis G. Loría-Salazar Editors

The Roles of
Accelerated
Pavement Testing
in Pavement
Sustainability
Engineering, Environment, and
Economics
The Roles of Accelerated Pavement Testing
in Pavement Sustainability
José P. Aguiar-Moya Adriana Vargas-Nordcbeck
•

Fabricio Leiva-Villacorta Luis G. Loría-Salazar

•

Editors

The Roles of Accelerated

Pavement Testing
in Pavement Sustainability
Engineering, Environment, and Economics

123
Editors
José P. Aguiar-Moya Fabricio Leiva-Villacorta
Universidad de Costa Rica Universidad de Costa Rica
LanammeUCR LanammeUCR
San José San José
Costa Rica Costa Rica

Adriana Vargas-Nordcbeck Luis G. Loría-Salazar

Universidad de Costa Rica Universidad de Costa Rica
LanammeUCR LanammeUCR
San José San José
Costa Rica Costa Rica

ISBN 978-3-319-42796-6 ISBN 978-3-319-42797-3 (eBook)

DOI 10.1007/978-3-319-42797-3

Library of Congress Control Number: 2016947396

© Springer International Publishing Switzerland 2016

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made.

Printed on acid-free paper

This Springer imprint is published by Springer Nature

The registered company is Springer International Publishing AG Switzerland
Preface

Accelerated Pavement Testing (APT) provides pavement engineers and practi-

tioners with valuable information concerning the behavior, performance, and life
expectancy of pavement structures in a very short period of time, by means of
controlled loading conditions. This results in improved understanding of pavement
structural behavior and response under different loading and environmental con-
ditions, comparison between different design methods and evaluation of new
materials and specifications, and identification of construction method’s deficien-
cies and sources of variability.
The previous has been highly valued by several transportation agencies where
the development of an APT program has been preferred to the option of analyzing
pavement structures in the long term, under the understanding that an APT test does
not necessarily equal a long-term pavement performance study. Therefore, in most
recent years, several APT programs worldwide have been established or are in the
process thereof.
The data that can be generated by an APT facility is incredibly rich and has the
advantage that it very closely resembles the true process that pavements are sub-
jected to during their service lives. Consequently, more and more countries are
relying on their APT facilities to develop or calibrate their own pavement analysis
and design methodologies to ensure that the developed method and their expected
results are reliable and trustworthy, based on their local materials, traffic patterns,
construction practices, and environmental conditions.
Furthermore, with global increase in infrastructure construction prices, APT
capabilities have been identified to be very effective in discriminating between
different alternatives to ensure pavement performance, identifying the most
cost-effective methods, and as validation tools for pavement designs in new con-
struction and rehabilitation projects.
Consequently, at the National Laboratory of Material and Structural Models
of the University of Costa Rica we feel deeply honored to host the Fifth Edition
of the Accelerated Pavement Testing (APT) Conference. The conference is orga-
nized by the Transportation Research Board (TRB) through the Standing

v
vi Preface

Committee on Full-Scale Accelerated Pavement Testing (AFD40). The committee

is concerned with the full-scale testing of traditional and innovative pavement
systems that reflect various construction conditions and maintenance practices,
where testing can be performed using conventional or accelerated methods utilizing
mobile or fixed equipment under in-situ controlled conditions. The fifth edition
of the conference follows a very successful meeting in Davis, California, where
APT future trends and needs were discussed, based on the experience developed at
over 25 APT programs worldwide.
The Fifth Edition of APT Conference originally received 76 abstracts from 16
countries. After rigorous review of these abstracts and full contributions by a sci-
entific committee of experts experienced in APT operations, data analysis, and/or
implementation of findings from APT experiments, 58 papers were finally accepted
for presentation and discussion at the conference and publication in these
proceedings.
We would like to thank each of the authors for sharing their APT experiences,
and we would also like to express our gratitude to the APT2016 Conference
Scientific and Organizing Committees for reviewing the papers and helping to
maintain the conference series quality standard.

San José, Costa Rica José P. Aguiar-Moya

2016 Adriana Vargas-Nordcbeck
Fabricio Leiva-Villacorta
Luis G. Loría-Salazar
Conference Organization

Chairpersons

Luis G. Loría-Salazar, University of Costa Rica

Wynand JvdM Steyn, University of Pretoria
Imad L. Al-Qadi, University of Illinois at Urbana-Champaign
John Harvey, University of California Davis

Scientific Committee and Review Panel

José P. Aguiar-Moya, University of Costa Rica

Abubeker Worake Ahmed, National Road and Transport Research Institute,
Sweden
Kumar Anupam, TUDelft
Martin Arraigada, Empa
Jason Bausano, MWV Speciality Chemicals
Rongj Cao, JSTI Group
Armelle Chabot, IFSTTAR
Timothy R. Clyne, Minnesota Department of Transportation
Nicolaas F. Coetzee, Retired
Seosamh Costello, University of Auckland
Morris De Beer, CSIR, South Africa
Louw Du Plessis, CSIR, South Africa
Ahmed El Youssoufy, Laval University
Sigurdur Erlingsson, VTI, Sweden
Angeli M. Urbano-Gamez, University of Illinois at Urbana-Champaign
Navneet Garg, FAA
Daba Gedafa, University of North Dakota
James Howard Greene, Florida Department of Transportation

vii
viii Conference Organization

Erman Gungor, University of Illinois at Urbana-Champaign

Xiaolong Guo, Auburn University
Jaime Hernandez, University of Illinois at Urbana-Champaign
Pierre Hornych, IFSTTAR
Mustaque Hossain, Kansas State University
David Jones, University of California Pavement Research Center
Shahbaz Khan, CSIR, South Africa
Issam Khoury, Ohio University
Jeremy D. Lea, UC Pavement Research Center
Fabricio Leiva-Villacorta, University of Costa Rica
Liping Liu, Tongji University
Larry Lynch, US Army Engineer Research and Development Center
Rebecca S. McDaniel, Purdue University
John Metcalf, ARRB Group
Michael A. Moffatt, ARRB Group
Mahdi Nasimifar, University of Nevada, Reno
Julio C. Paniagua-Fernandez, University of California Pavement Research Center
Manfred N. Partl, EMPA
Raymond Buzz Powell, National Center for Asphalt Technology, Auburn
University
Lubos Remek, University of Zilina
Stefan A. Romanoschi, The University of Texas at Arlington
Milad Saghebfar, Louisiana Transportation Research Center
Aimin Sha, Chang’an University
Injun Song, SRA International, Inc.
Wynand JvdM Steyn, University of Pretoria
David H. Timm, Auburn University
Erol Tutumluer, University of Illinois at Urbana-Champaign
Adriana Vargas-Nordcbeck, University of Costa Rica
Michael C. Vrtis, Auburn University
Bastian Wacker, BASt, Germany
Benjamin J. Worel, Minnesota Department of Transportation
Zhong Wu, Louisiana Transportation Research Center

Organizing Committee

Luis G. Loría-Salazar
Wynand JvdM Steyn
Imad L. Al-Qadi
John Harvey
José P. Aguiar-Moya
Adriana Vargas-Nordcbeck
Fabricio Leiva-Villacorta
Conference Organization ix

Oscar Rodriguez-Quintana
Raquel Arriola-Guzmán
Gabriela Contreras-Matarrita
Rosa I. Cordero-Solano
Fabián Elizondo-Arrieta
Contents

Part I Overview of Accelerated Pavement Testing

A Brief History of Full-Scale Accelerated Pavement Testing
Facilities to 1962 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
John B. Metcalf
Four Decades of the Circular Test Track at the Institute
of Engineering UNAM Contributing to Pavement
Research in Mexico . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Noé Hernández, Alexandra Ossa and Francisco A. Rangel

Part II Establishment of New Accelerated Pavement

Testing Facilities
Design and Engineering Challenges in the Development
of the FAA’s Full Scale Accelerated Pavement Test
Facility NAPMRC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
Murphy Flynn, Navneet Garg, Wilfredo Villafane and David Traverzo
Heavy Vehicle Simulator and Accelerated Pavement
Testing Facility at Rowan University . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Ayman W. Ali and Yusuf Mehta

Part III General Accelerated Pavement Testing

A Link of Full-Scale Accelerated Pavement Testing to Long-Term
Pavement Performance Study in the Western Cape Province
of South Africa . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
J.K. Anochie-Boateng, W. JvdM Steyn, C. Fisher and L. Truter
Accelerated Pavement Testing in Slovakia: APT Tester 105-03-01 . . . . . 81
Lubos Remek, Jan Mikolaj, Matej Kyselica and Milan Škarupa

xi
xii Contents

APT with the Mobile Load Simulator MLS10 Towards

Non-destructive Pavement Structural Analysis . . . . . . . . . . . . . . . . . . . . . 95
Bastian Wacker, Frédéric Otto, Pengfei Liu, Dawei Wang, Markus Oeser,
Micha Buch and Ulf Zander
Evaluating Nonlinearity on Granular Materials and Soils Through
the Use of Deflection Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
Fabricio Leiva-Villacorta, Luis Loría-Salazar and Edgar Camacho-Garita
Full Scale Accelerated Pavement Tests to Evaluate the Performance
of Permeable and Skeletal Soil Block Pavement Systems . . . . . . . . . . . . . 131
Abubeker Ahmed, Fredrik Hellman and Sigurdur Erlingsson
In-Situ Validation of Three-Dimensional Pavement Finite
Element Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
Osman Erman Gungor, Imad L. Al-Qadi, Angeli Gamez
and Jaime A. Hernandez
Investigation of 3D-Move Responses Under Traffic Speed
Deflection Devices (TSDDs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
Mahdi Nasimifar, Raj S. Siddharthan, Elie Y. Hajj and Ramin Motamed
Long-Term Pavement Performance Monitoring and the Revision
of Performance Criteria for High Modulus Asphalt
in South Africa . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
Julius Komba, Joseph Anochie-Boateng, Johan O’Connell
and Benoit Verhaeghe
Optimum Properties of Geocell Reinforcement for Sustainable
Low-Volume Paved Roads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
B. Bortz and M. Hossain
Perspectives on Trends in International APT Research . . . . . . . . . . . . . . 211
W. JvdM Steyn and F. Hugo
Potential Benefits of APTF for Evaluation of Flexible Pavement
for Its Permanent Deformation Behaviour . . . . . . . . . . . . . . . . . . . . . . . . 227
M.N. Nagabhushana, Shahbaz Khan, Abhishek Mittal and Devesh Tiwari
Study of the Bearing Capacity of Swiss Standard Pavements
Under MLS10 Loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
Martin Arraigada, Andreas Treuholz and Manfred N. Partl
The Development of APT Methodology in the Application
and Derivation of Geosynthetic Benefits in Roadway Design . . . . . . . . . . 257
Jonathan Cook, Michael Dobie and David Blackman
Tire Type Effect on Pavement Responses; Accelerated Pavement
Testing Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
Mojtaba Ziyadi and Imad L. Al-Qadi
Contents xiii

Part IV Accelerated Pavement Testing Focused on Mechanistic—

Empirical Pavement Design Procedures and Models
Analysis of Dynamic Response of Asphalt Pavement in Heavy
Vehicle Simulator Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291
Aimin Sha, Jie Wang, Liqun Hu and Xiaolong Zou
Automating Mechanistic-Empirical Pavement Design Calibration
Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
Xiaolong Guo and David H. Timm
Calibration of ME Design Using a Combination of APT
and PMS Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
Jeremy D. Lea, Rongzong Wu and John T. Harvey
Key Concepts in Dynamic Signal Processing from Instrumented
Pavement Sections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
David H. Timm
Protocols for Accelerated Pavement Testing of Fully Permeable
Pavements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
D. Jones, R. Wu and H. Li

Part V Accelerated Pavement Testing on Asphalt Concrete

Pavements
Behavior Evolution on Performance of UV-Irradiation Aged Asphalt
Mixtures Under Reduced-Scale Accelerated Trafficking . . . . . . . . . . . . . 373
Jinting Wu, Fen Ye, Frederick Hugo, Yinting Wu, Feng Wang
and Xinhui Ding
Effects of Binder and Mix Properties on the Mechanistic Responses
of Fatigue Cracking APT Sections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393
Daba S. Gedafa, Mustaque Hossain, Stefan A. Romanoschi
and Mbakisya A. Onyango
Evaluating Resistance of Hot Mix Asphalt Overlays to Reflective
Cracking Using Geocomposites and Accelerated Loading . . . . . . . . . . . . 407
Mansour Solaimanian, Ghassan Chehab and Marcelo Medeiros
Optimum Tack Rate for Hot-Mix Asphalt Bonding . . . . . . . . . . . . . . . . . 419
A. Sufian, M. Hossain and G. Schieber
Rutting Resistance of Asphalt Mixes Containing Highly Modified
Asphalt (HiMA) Binders at the Accelerated Pavement Load Facility
in Ohio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 429
Issam Khoury, Shad Sargand, Roger Green, Benjamin Jordan
and Paul Cichocki
xiv Contents

Structural Study of Perpetual Pavement Performance in Ohio . . . . . . . . 441

Issam Khoury, Shad Sargand, Benjamin Jordan, Paul Cichocki
and Matthew Sheer
Study of In-Service Asphalt Pavement High-Temperature
Deformation Based on Accelerated Pavement Test . . . . . . . . . . . . . . . . . . 461
Liping Liu, Yu Yuan and Lijun Sun
Thickness and Binder Type Evaluation of a 4.75-mm Asphalt Mixture
Using Accelerated Pavement Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 475
James Howard Greene and Bouzid Choubane
Towards Understanding Tyre-Pavement Contact in APT Research
on Flexible Pavements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 491
Morris De Beer and Colin Fisher
Use of APT for Validating the Efficiency of Reinforcement Grids
in Asphalt Pavements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509
Martin Arraigada, Federico Perrotta, Christiane Raab, Gabriele Tebaldi
and Manfred N. Partl

Part VI Accelerated Pavement Testing on Airports

Estimation of In-Situ Shear Strength Parameters for Subgrade
Layer Using Non-destructive Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 525
Hadi Nabizadeh, Elie Y. Hajj, Raj V. Siddharthan, Sherif Elfass
and Peter E. Sebaaly
Understanding Airfield Pavement Responses Under High Tire
Pressure: Full-Scale Testing and Numerical Modeling . . . . . . . . . . . . . . . 539
Hao Wang, Navneet Garg and Maoyun Li
Use of Ground Penetrating Radar at the FAA’s National Airport
Pavement Test Facility (NAPTF) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 555
Injun Song and Albert Larkin

Part VII Relating Laboratory Tests to Performance Using

Accelerated Pavement Testing
Development and Calibration of Permanent Deformation Models . . . . . 573
Fabricio Leiva-Villacorta, Adriana Vargas-Nordcbeck,
José Pablo Aguiar-Moya and Luis Loría-Salazar
Laboratory Tack Coat Investigation of Factors Contributing
to Debonding Observed at the NCAT Test Track . . . . . . . . . . . . . . . . . . 589
Michael C. Vrtis and David H. Timm
Contents xv

Research Partnership Between the National Center for Asphalt

Technology (NCAT) and the Minnesota Department of
Transportation MnROAD Facility for a Nationwide Pavement
Performance Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 601
R. Buzz Powell and Ben Worel

Part VIII Instrumentation for Accelerated Pavement Testing

Development and Validation of a Nondestructive Methodology
to Measure Subgrade Moisture Contents at the NCAT Pavement
Test Track . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 621
R. Buzz Powell
Development of a New Pavement Strain Coil Measuring System
at CAPTIF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 633
F.R. Greenslade
Effects of Surface Condition on Non-uniform Contact Stresses
of APT Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 645
W. JvdM Steyn and J. Maina
Electronic Upgrade of a Standard Benkelman Beam to Enable
Capture of Full Bowl Deflections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 659
F.R. Greenslade
Evaluation of Weight in Motion Sensors on the IFSTTAR
Accelerated Testing Facility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 671
Pierre Hornych, Jean-Michel Simonin, Jean-Michel Piau,
Louis-Marie Cottineau, Ivan Gueguen and B. Jacob
Influence of Tire Footprint Area and Pressure Distribution
on Pavement Responses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 685
Fabricio Leiva-Villacorta, Adriana Vargas-Nordcbeck,
José P. Aguiar-Moya and Luis Loría-Salazar
Instrumentation of an Innovative Pavement Section
on Motorway A10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 701
Ngoc Son Duong, Juliette Blanc and Pierre Hornych

Part IX Accelerated Pavement Testing on Portland Cement

Concrete Pavements
Design, Instrumentation and Construction of Bonded Concrete
Overlays for Accelerated Pavement Testing . . . . . . . . . . . . . . . . . . . . . . . 717
Julio Paniagua, Fabian Paniagua, Angel Mateos, John Harvey
and Rongzong Wu
xvi Contents

Performance of Thin RCC Pavements Under Accelerated Loading . . . . 735

Zhong Wu, Moinul Mahdi and Tyson Rupnow
The Design, Construction and First-Phase Heavy Vehicle
Simulator Testing Results on Full Scale Ultra-Thin Reinforced
Concrete Test Sections at Rayton, South Africa . . . . . . . . . . . . . . . . . . . . 751
L. du Plessis, G.J. Jordaan, P.J. Strauss and A. Kilian
The Design, Construction and Heavy Vehicle Simulator Testing
Results on Roller Compacted Concrete Test Sections at the CSIR
Innovation Site and on a Full-Scale Test Road at Rayton . . . . . . . . . . . . 769
L. du Plessis, G. Rugodho, W. Govu, K. Mngaza and S. Musundi

Part X Accelerated Pavement Testing to Evaluate Functional

Performance
Assessment of Flexible Pavement Response During Freezing
and Thawing from Indoor Heavy Vehicle Simulator Testing . . . . . . . . . 787
Ahmed El-youssoufy, Guy Dore, Jean-Pascal Bilodeau and Fritz Prophète
Development of IRI Models Based on APT Data . . . . . . . . . . . . . . . . . . . 799
José P. Aguiar-Moya, Pablo A. Torres-Linares, Edgar Camacho-Garita,
Fabricio Leiva-Villacorta and Luis G. Loría-Salazar
Permanent Deformation of an Unbound Granular Pavement
Subjected to Accelerated Multiple-Axle Group Loads . . . . . . . . . . . . . . . 815
Michael A. Moffatt
Three Years’ Monitoring of IRRF Instrumented Test Road . . . . . . . . . . 829
Leila Hashemian and Alireza Bayat

Part XI Accelerated Pavement Testing to Evaluate Sustainable

Materials
Comparative Evaluation of Performance of Warm Mix RAP Asphalt
Under Accelerated Unidirectional Wheelload Trafficking . . . . . . . . . . . . 847
Yucheng Huang, Fred Hugo, W. JvdM Steyn, Haocheng Xiong
and Linbing Wang
Evaluation of the Optimum Percentage of RAP and RAS
in Asphalt Mixes in Texas Using Accelerate Pavement Testing . . . . . . . . 863
Stefan A. Romanoschi, Tom Scullion, Fujie Zhou and Reza Saeedzadeh
Contents xvii

Field Performance Evaluations of Large Sized Unconventional

and Recycled Aggregates for Subgrade Improvement . . . . . . . . . . . . . . . 877
Hasan Kazmee, Erol Tutumluer and Debakanta Mishra
Structural Assessment of the Effect of a Cement-Stabilized Base
Combined with a Cold Central-Plant Recycled Layer at the NCAT
Test Track . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 891
M.A. Díaz-Sánchez, D.H. Timm and B.K. Diefenderfer
 Part I
Overview of Accelerated Pavement Testing
A Brief History of Full-Scale Accelerated
Pavement Testing Facilities to 1962

John B. Metcalf

Abstract Full-scale accelerated pavement testing (APT) facilities developed

throughout the world since the AASHO Road Test 1958–62 were described by Metcalf
(Application of full-scale accelerated pavement testing. National Academy Press,
Washington, 1996). The purpose of this paper is to document facilities in operation
before the 1960’s. Some facilities were identified where details are very limited.

1 Introduction

Since Roman times the thickness of pavement structures decreased from about
900 mm to perhaps some 100 mm, before the appearance of the motorcar in the
early 1900s. Since then, as the number, gross mass, multi-axle configuration and
speed of vehicles has increased (to about 9 tonne on typical dual-tire truck axles,
tire pressures to 700 kPa) and speeds up to 100 km/h) so has the thickness, material
quality and thus the cost of pavements. When added to the cost of maintaining
pavements under heavy traffic with consequently high user costs, the need for a
much better understanding of the structural behaviour of pavements became evi-
dent, hence accelerated pavement testing was developed.
The definition of accelerated pavement testing (Metcalf 1996) used and used
here is:
Full scale accelerated pavement testing is defined as the controlled application of a pro-
totype wheel loading, at or above the appropriate legal load limit, to a prototype or actual,
layered, structural pavement system to determine pavement response and performance
under a controlled, accelerated, accumulation of damage in a compressed time period.

J.B. Metcalf (&)

Sydney, Australia
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 3

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_1
4 J.B. Metcalf

1.1 Early Installations

The development of APT tracks paralleled the development of test roads, on which
controlled traffic loading was applied, for example, the Bates Road Test (Older
1924) and, of course, many sections of experimental pavements were laid and
observed under ‘normal’ traffic. While these types of studies continue, they are not
addressed in this paper. Croney and Croney (1991) suggested that a machine to
asses the performance of stone setts existed in the 1840s but provided no details.
Boulnois (1919) referred to a linear machine designed in 1905, but not built,
perhaps that of RJ Kirwan, Fig. 1.

1.2 The Paving Determinator 1909

This facility, built in 1909 (Hines 1913) was the first APT meeting the definition
(Metcalf 1996). The facility was developed to test ‘the quality of paving materials
under heavy abrasive influences’ and was used to evaluate eight pavement sections
(Anon 1912, 1913).
The circular track had a trafficked strip with a radius varying from 2.6 to 3.3 m
with an eccentric drive mechanism which moved the wheels across the trafficked
area (Fig. 2). It was driven at speeds from 4.8 to 19.3 km/h but testing was usually
at about 11 km/h. Loading was applied using two wheels, one with varying widths
of (presumably steel) tyre and one with five plungers each carrying a horse-
hoof-shaped plate and a mechanism to simulate the ankle movement of a walking
horse. The wheels weighed 636 kg and the ‘horse shoes’ had a contact pressure of
1034 kPa.
The article claimed concrete performed best but noted ‘downstream’ effects
between the sections. A later test (Anon 1913) included a ‘macadam’ section which
failed under the ‘horse shoes’ but was found to be very well compacted after
trafficking. During later tests modifications were made to the machine, to better
simulate real traffic—the loading was reduced from five to three ‘horse shoes’.
A feature of the results was the importance of achieving an even surface finish.
Interestingly, Stoffels (1995) reported tests on a similar 4.9 m circular track to
evaluate the effects of horse-and-buggy traffic.

Fig. 1 Kirwan design (Boulnois 1919)

A Brief History of Full-Scale Accelerated Pavement Testing … 5

Fig. 2 General view of the ‘Paving Determinator’ ‘Concrete-Cement Age’ (Anon 1912, 1913)

1.3 National Physical Laboratory Road Machine 1911

In 1911, the placement of a number of trial road sections led to the suggestion that
similar information could be obtained more rapidly:
by means of a machine which should apply to samples of various surfacing material the
same sort of stresses as took place on the road… the result was… the Road Machine.
(Crompton 1913).

The ‘Road Machine’ was developed at the National Physical Laboratory (NPL), in
the United Kingdom in 1911–12 (Boulnois 1919; Speilmann and Hughes 1936).
The circular track (Fig. 3) was 10 m diameter loaded by eight steel-rimmed wheels,
each independently driven, on radial arms and operating at 5–13 km/h. The
intention was to simulate one year’s wear on a typical heavily-trafficked road in
24 h (Pyatt 1983). In 1916, the machine was temporarily retired until 1920 and was
replaced in 1933 (NPL 1911–33) by a 34 m diameter track with loading was
applied by a full-sized ‘captive’ lorry.
The original 1911 machine had features still found on current equipment: the
wheelpaths were varied by a cam mechanism to impart a transverse motion to the
wheels, the load on each wheel could be adjusted and a hot air duct system,
discharging in front of the wheels, could alter the road surface temperature.
Instrumentation was also developed to measure transverse and longitudinal profiles.
An overall conclusion from the early work is still valid:
An important finding from the road machine tests was the importance of achieving a low
voids content in asphalt surfacing, this was corroborated by experience with five full scale
6 J.B. Metcalf

Fig. 3 NPL machine 1911–33

experimental roads. It was also pointed out that inadequate bitumen content, the type of sand
and the initial compacted density were critical factors for good performance. (NPL Annual
Report 1911).

In 1930 the NPL machine was transferred to the Road Research Laboratory
(RRL), by 1934 a second full-scale circular road machine was in use.

1.4 Indonesian Test Circuit

The Dutch East India Roads Association built a test circuit, in the late 1920s at the
Bandoeng Technical College, loaded by ‘captive’ vehicles (Fig. 4). Ravensteijn and
Kop (2008) state:

Fig. 4 The Indonesian Test circuit (Ravensteijn and Kop 2008)

A Brief History of Full-Scale Accelerated Pavement Testing … 7

Different types of surfaces were tested on the 175 m long and 5.5 m wide rack. Two
‘vehicles’, a lorry and a two-wheeled goods cart (i.e. grobak) were used to test the road.

1.5 The United States

In the United States, accelerated trials were conducted from about 1914 by running
trucks over experimental road sections, culminating with the AASHO Road Test
1958–60 (AASHO 1961). The transition from test roads to test tracks appeared to
commence in about 1934, when a test road of 54 m mean diameter was built in
Michigan and loaded by a truck at a speed of about 19 km/h with a rear wheel load
of 2500 kg (Emmons 1934).

1.6 Bureau of Public Roads

Accelerated pavement testing at the Bureau of Public Roads (BPR) appears to have
begun in 1924 when an impact testing device (Fig. 5), with loading applied using a
solid tyred truck wheel, was developed (Teller 1924). It was first used to test
concrete slabs.
Later, Carpenter and Goode (1936) described a small circular track operating at a
BPR laboratory in Arlington, Virginia, from 1934. The track pavement was built in
a 450 mm wide concrete trough, with a mean diameter of 3.6 m, and 312 mm deep
including provision for flooding. Two wheels with low pressure tyres were used to
compact and test the pavement. The speeds ranged from 5.6 to 14.4 km/h with the
weight on each wheel being about 354 kg. The wheels could be distributed across
the width of the track to simulate traffic, but provision was made to fix the track and
thus concentrate the loading for faster testing. The tests reported in 1936 were on

Fig. 5 BPR Impact tester (Thompson 1928)

8 J.B. Metcalf

road-mix or oiled gravel surfacing. Vertical and horizontal displacements were

measured and it was reported that the forward surface movement was, in general,
proportional to the vertical deformation.

1.7 Kentucky

A circular track at the University of Kentucky (Fig. 12) (Anon 1946; Gregg 1945,
Cantrill 1942) was 3.6 m median diameter with standard truck wheels loaded to
707 kg per wheel at speeds to 40 km/h (Fig. 6), drawings from the BPR were used
in constructing the track. Transverse rutting was used as a measure of the perfor-
mance of sandstone and limestone aggregate bituminous concrete pavements.
Peed (1952) also described a machine located at the State Highway Materials
Research Laboratory, a paved area, 4.3 m in diameter, at the centre of a 600 mm
wide paved area, carried a pair of truck tyres on an eccentric mounting to cover the
whole surface with a skidding action which was used to test traffic paints. It is
possible that these reports all refer to the same facility.

1.8 Kansas

Benson (1937) talked of the use of a circular track similar to that of the BPR con-
structed at Kansas State College, Manhattan. He stressed the importance of temper-
ature control when testing asphalt paving but gave no details of the track or program.

Fig. 6 Kentucky track (Cantrill 1942)

A Brief History of Full-Scale Accelerated Pavement Testing … 9

Fig. 7 Illinois track detail (after Ahlberg and Barenberg 1963)

1.9 University of Illinois

Research at the University of Illinois circular track commenced in 1957 (Barenberg

1967). The pavements were constructed with an outer diameter of 7.5 m operating
with wheel loads to 1680 kg and speeds to 24 km/h (Ahlberg and Barenberg 1963)
(Fig. 7).

1.10 Indiana

A circular track of close to full scale was used in Indiana (Speer 1960) where
pavement sections were fitted into a 4.3 m diameter annulus loaded by four
164 × 305 mm mobile home wheels mounted on cruciform arms. The speed range
was 8–64 km/h, typically 40 km/h, with loads of 454 kg on each wheel at a tyre
pressure of c. 500 kPa. Some comparisons with the Ottawa, Illinois road test
were made. Much of the testing was directed at the relationship between asphalt
(bitumen) penetration, content and temperature.

1.11 Washington State

In 1964 the University constructed a full-scale circular test track, apparently fol-
lowing experimentation with a model scale (1:3) linear track (Ekse and LaCross
1957). The full-scale track (Mahoney and Terrel 1982) had a 13.6 tonne steel frame
and water tank rotating on a 25.6 m diameter circle (Fig. 8). Three dual wheels
10 J.B. Metcalf

Fig. 8 Washington University full scale circular track (Mahoney and Terrel 1982)

could each be loaded to 5040 kg at speeds up to 72 km/h. Krukar and Cook (1971)
reported tests at 32 km/h on three base configurations over a silt subgrade with an
asphalt surface.

1.12 United States Army Corps of Engineers

The US Army Corps of Engineers took the lead in flexible runway pavement design
to 1939 (Lynch et al. 1999), when aircraft with wheel loads of 1.7 tonne and
heavier were being developed. Investigators used loaded earth-moving equipment
with tyres comparable to those of aircraft to simulate wheel loads up to 27 tonne on
dual tyres. Drivers travelled along designated lanes on the pavement until failure
occurred or until a specific number of coverages was completed.
By the 1950s airfield pavement research had extended dramatically. The chal-
lenge of dealing with aircraft weight increases was compounded by the use of
high-pressure tyres and complex multiple-wheel landing gears. Pilot tests began in
1949 and 1950 with two test lanes, each 120 ft (36 m) long, inside a hangar. Towed
vehicles with gear assemblies from aircraft at various tyre pressures applied 2000
passes to the test lanes as the norm.

1.13 Europe

There appears to have been little test road or APT activity in Europe between 1930
and 1950, although, in the UK, a different approach was adopted: that of closely
monitoring a large number of experimental pavement sections incorporated into
existing roads and subject to normal traffic (Croney and Croney 1991). Whilst the
A Brief History of Full-Scale Accelerated Pavement Testing … 11

Fig. 9 Iasi Polytechnic Institute c. 1954 (PIARC 1967)

Road Research Laboratory ‘road machine’ existed, little activity was reported.
Design traffic levels for UK motorways increased by an order of magnitude from
1960 to 1980 (Lister and Porter 1982), whereas the traffic on road experiments had
only increased about fivefold. This led to the development of several APT facilities
in Europe which began to produce results in the 1960s.

1.14 Romania

Circular test tracks in Bucharest (1952) and at the Iasi Polytechnic Institute (1956)
(Fig. 9) were described in a Permanent International Association of Roads
Congress (PIARC 1967). The Bucharest track was 30 m in diameter, the applied
load was 3.3 tonne. The diameter of the Iasi facility was 10 m at the centreline of a
1.5 m wide pavement. A load of 2.15 tonne was applied at a speed to 20 km/h
(Preda et al. 1958).

1.15 Germany

During the 1950s the BundesAnstalt fur Strassen Wessen (BASt) adopted a repe-
ated pulse load approach in which the point of application of the pulse, the contact
area and the force applied could be varied to simulate wheel loading (Behr 1963)
(Fig. 10). BASt had earlier considered two other full-scale wheel loading facilities
but did not build either.
12 J.B. Metcalf

Fig. 10 Impact unit 2 1955 BASt (Behr 1963)

1.16 Czech Republic

A circular facility, 36–46 m diameter, operating at speeds up to 50 km/h with a load

to 10 tonne, existed at Brno (Czech National report 1967). The initial tests reported
were directed at the restoration of failed concrete slabs (Fig. 11).

Fig. 11 Brno track slab cracking 1967 (Czech National report, PIARC 1967)
A Brief History of Full-Scale Accelerated Pavement Testing … 13

1.17 Other Installations

There were installations in Germany, in Karlsruhe and an impulse loading device at

the University of Hannover, in the early 1960s (Metcalf 1977), and in Poland
(Keller 1974) but no details are currently available. There was also a track in
Bratislava (Fig. 12) possibly dating from the late 1950s, and one at the
Commonwealth Scientific and Industrial Research Organization, Soil Mechanics
Section in Australia in 1961 of which little has been discovered except for a
photograph dated 1961 (Fig. 13).

Fig. 12 Bratislava track (date and origin unknown)

Fig. 13 Australian APT facility 1961(Metcalf—personal record)

14 J.B. Metcalf

1.18 Final Words

Hugo and Martin (2004) reported that:

a large number of APT tests were… useful for answering questions relating to the use of
new materials, composite materials and materials with complex physical characteristics…
systematically developing and evaluating performance models… answering questions
related to rehabilitation, construction and maintenance in the field.

Studies of economic benefits have been conducted on South African (Rust et al.
1997), Californian (du Plessis et al. 1994) and Louisianan (King and Rasoulian
2004) studies and on Australian programs Rose and Bennett (1994) reported
benefit/cost ratios from 1.4 to 12, suggesting an overall figure between 4 to 5.
Brown (2004) states:
Full-scale accelerated testing of pavements has become a powerful technique for assisting
with understanding pavement deterioration under realistic conditions and measuring
pavement response to moving wheel loads. It forms an essential bridge between laboratory
work and theory and the site situation. …Extensive experience in South Africa suggests a
(cost/benefit) ratio of about 1:10.

Thus any input to this study would be welcomed to complete the history of APT
because the value of the studies to pavement design and material selection tech-
nologies is amply demonstrated. A comprehensive study of the findings from the
various installations could also be of considerable interest.

References

AASHO. (1961). The AASHO road test, history and description of the project. Special Report
61A, Highway Research Board, Washington, DC.
Ahlberg, H. L., & Barenberg, E. J. (1963). The University of Illinois pavement test track—A tool
for evaluating highway pavements (pp. 1–28). Highway Research Record 13, Washington, DC.
Anon. (1912). Concrete: A superior pavement under hard traffic (pp. 52–54), Concrete-Cement
Age, December 1912.
Anon. (1913). A severe test of paving slabs, Concrete-Cement Age (pp. 303, 305–7), June 1913.
Anon. (1946). Proposed field and laboratory study of flexible pavements, Highway Materials
Research Laboratory. Commonwealth of Kentucky Department of Highways (Internal report).
Barenberg, E. J. (1967). Evaluating stabilized materials. NCHRP, Department of Civil
Engineering, Engineering Experiment Station, College of Engineering, University of Illinois.
Behr, H. (1963). Eine Anlage fur Dauerbelastungsversuche an Strassen. Strasse und Autobahn,
XIV(8), 281–286.
Benson, J. R. (1937). Bituminous research in Kansas. In Montana National Bituminous
Conference, Sep. 1937 (pp. 81–89). Montana State Highway Department and Western
Petroleum Refiners Association.
Boulnois, H. P. (1919). Modern roads. London: Edward Arnold.
Brown, S. F. (2004). Accelerated pavement testing in highway engineering. In Proceedings of the
ICE—Transport (Vol. 157, Issue 3), 01 August 2004.
Cantrill, C. (1942). The use of a circular track for testing bituminous pavement mixtures. In
Proceedings of the Association of Asphalt Paving Technologists (Vol. 13, pp. 69–83).
A Brief History of Full-Scale Accelerated Pavement Testing … 15

Carpenter, C. A., & Goode, J. F. (1936). Circular track tests on low-cost bituminous mixtures.
Public Roads, 17(4), 69–82.
Crompton, R. E. (1913). The mechanical engineering aspects of road construction. In Proceedings
Institute of Mechanical Engineers Part ¾ (pp. 1253), discussion 1310–1326.
Croney, D., & Croney, P. (1991). The design and performance of road pavements (2nd ed.). New
York City: McGraw-Hill.
Czech National Report 1967, XIIIth PIARC World Road Congress Tokyo, Question IV–V, IV–II,
p. 36.
du Plessis, L., Nokes, W. A., Mahdavi, M., Burmas, N., Holland, T. J., &. Lee, E.-B. (1994).
Economic benefits assessment of accelerated pavement testing research in California—Case
study (pp. 137–146). Transportation Research Record 2225, National Research Council,
Washington, DC.
Ekse, M., & LaCross, L. M. (1957). Model analysis of flexible pavement and subgrade stresses. In
Proceedings 26th Annual Meeting of the Association of Asphalt Paving Technologists (Vol. 26,
pp. 312–320).
Emmons, W. J. (1934). Stability experiments on asphaltic paving mixtures. Public Roads, 14(11),
197–218.
Gregg, L. E. (1945). An outline of prospective research for the highway materials research
laboratory. Kentucky Transportation Center Research Report, Paper 1375. http://uknowledge.
uky.edu.
Hines, E. N. (1913). Concrete roads of Wayne County. Washington, DC: Portland Cement
Association.
Hugo, F., & Martin, A. L. E. (2004). Significant findings from full-scale accelerated pavement
testing, NCHRP synthesis 325, Transportation Research Board, Washington, DC.
Keller, H. (1974). Die technische Entwicklung einer Rundlaufenlage zur Dauerbelastung von
Strassen. Strasse und Autobahn, 8(74), 283–288.
King, W., Jr., & Rasoulian, M. (2004). Experimental and operational progress with a benefit/cost
analysis for Louisiana’s pavement research facility. In 2nd International Conference on
Accelerated Pavement Testing. University of Minnesota, September 26–29, 2004.
Krukar, M., & Cook, J. C. (1971). Comparison of Washington State University Test Track
Experimental—Pavement Ring Nos. 2 and 3. In Proceedings of the 41st Annual Meeting of the
Association of Asphalt Paving Technologists (Vol. 40, pp. 1–30).
Lister, N. W., & Porter, J. (1982). Philosophy and experience of full scale pavement testing at the
Transport and Road Research Laboratory. In Proceedings International Colloquium Full Scale
Pavement Test (pp. 1–18). Institute for Strassen-, Eisenbahn und Felsbau an der
Eidgenossischen Technischen Hochschule, Zurich, Switzerland.
Lynch, L., Janoo, V., & Horner, D. (1999). US Army Corps of Engineers experience with
accelerated and full-scale pavement investigations. In Proceedings of the International
Conference on Accelerated Pavement Testing, Reno, NV.
Mahoney, J. P., & Terrel, R. L. (1982). Laboratory and field fatigue characterization for sulphur
extended asphalt paving mixtures. In Proceedings of the Fifth International Conference on the
Structural Design of Asphalt Pavements (pp. 831–843). University of Michigan, Ann Arbor.
Metcalf, J. B. (1977). Report on an overseas visit June/July 1977 to United Kingdom and Europe.
Internal Report AIR 000-78, Australian Road Research Board, Vermont South, VIC.
Metcalf, J. B. (1996). Application of full-scale accelerated pavement testing. Synthesis of Highway
Practice 235, National Academy Press, Washington, DC.
National Physical Laboratory (various). Reports of the National Physical Laboratory 1911, 1912,
1913–1914, 1914–1915, 1915–1916, 1921, 1930, 1931, 1933, various publishers.
Older, C. (1924). Bates experimental road or highway research in Illinois, Bulletin 21, Illinois
Department of Highways, original circa 1937, Reprint in Illinois Digital Archives.
Peed, A. C. (1952). Physical properties of traffic paints. Highway Research Board Bulletin, 57,
9–22.
PIARC. (1967). Proceedings XIIIth PIARC world road congress. Tokyo, Permanent International
Association of Road Congresses (World Road Association), Paris.
16 J.B. Metcalf

Preda, N., Atanasiu, D., Virlan, R., & Sesan, A. (1958). Asupra necesitatii studierii problemelor
rutiere in stadiu pilot, Buletinul Institutului Politehnic din Iasi, serie noua, Tomul IV (VIII)
fasc, pp. 3–4.
Pyatt, E. (1983). The National Physical Laboratory—A History. London: Hilger.
Ravensteijn, W., & Kop, J. (Eds.). (2008). For profit and prosperity. Zaltbommel/Leiden:
Aprilis/KITLV Press.
Rose, G., & Bennett, D. (1994). Benefits from research investment: Case of Australian Accelerated
Loading Facility pavement research program (pp. 82–90). Transportation Research Record
1455, TRB, National Research Council, Washington, DC.
Rust, F. C., Kekwick, S. V., Kleyn, E. G., & Sadzik, E. S. (1997). The impact of the heavy vehicle
simulator (HVS) test program on road pavement technology and management. In Proceedings
of the 8th International Conference on Asphalt Pavements (pp. 1073–1085). Seattle, USA.
Speer, T. L. (1960). Progress report on laboratory traffic tests of miniature bituminous highways.
In Proceedings of the 29th Annual Meeting of the Association of Asphalt Paving Technologists
(Vol. 29, pp. 316–361).
Speilmann, P. E., & Hughes, A. C. (1936). Asphalt roads, The Roadmakers Library (Vol. 5).
London: Edward Arnold & Co.
Stoffels, S. M. (1995). Mitigation of horseshoe-induced roadway damage (summary) (pp. 6–7).
Pennsylvania Transportation Institute 1994–5 Annual Report, Pennsylvania State University.
Teller, L. W. (1924). Impact tests on concrete pavement slabs. Public Roads, 5(2), 1–13.
Four Decades of the Circular Test Track
at the Institute of Engineering UNAM
Contributing to Pavement Research
in Mexico

Noé Hernández, Alexandra Ossa and Francisco A. Rangel

Abstract The circular test track at the Institute of Engineering UNAM was built
four decades ago, between 1970 and 1971. Since its construction, this device has
been very valuable for the development, research, and validation of flexible
pavement design in Mexico. To date, 40 studies have been conducted on the
circular track. It is common for each experiment to build three different sections of
pavement to maximize the number of variables to be analyzed. With the infor-
mation obtained from the experimental sections under service, data on the behavior
of sections in different locations throughout Mexico, and experimental results from
the circular track, a method for the structural design of flexible pavements was
developed. This research program was carried out by the Institute of Engineering,
UNAM. The latest version of the pavement design method was published in 2014,
called the Structural Design of Flexible Pavements (DISPAV-5 3.0), and is cur-
rently the most used method in Mexico for the design of such structures. This
resource has a mechanistic-empirical background and includes models to predict
asphalt fatigue and permanent deformation damage. However, the models are over
15 years old and are unable to incorporate current characteristics of materials and
traffic conditions. This paper describes the role and significance of the circular test
track in the development of flexible pavement design in Mexico, the characteristics
of the variables that influence pavement behavior and future directions proposed to
develop a method of pavement design that considers current needs.

N. Hernández (&)
A. Ossa
F.A. Rangel

Geotechnical Department, Institute of Engineering, UNAM, Edificio 6,
Fernando Espinosa, Circuito Exterior, Ciudad Universitaria, 04510 Mexico,
D.F., Mexico
e-mail: [email protected]
A. Ossa
e-mail: [email protected]
F.A. Rangel
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 17

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_2
18 N. Hernández et al.

Keywords Circular test track Pavement design Fatigue cracking Performance


models Rutting

1 Introduction

In 1961, the Ministry of Public Works (now the Secretariat of Communications and
Transport, SCT) launched a research program in low traffic volume roads through
the Engineering Institute of the Autonomous University of Mexico (UNAM) to
address the pavement problems of the time due to the importance of annual
investments made for preservation and the increasing trend of such investments.
These low traffic volume roads constituted the majority of the national road network
at that time. According to Corro (1970), the proposed research guidelines were:
• identify the most important variables that affect the behavior of flexible pave-
ments for low traffic volume roads,
• establish criteria to judge the pavement behavior and set levels of acceptance
and service periods,
• conduct experimental research to define test methods that have good correlation
with the behavior of the materials used in pavements, and
• review road tests carried out in other countries to determine particular tech-
niques that are applicable to Mexico.
The main objective of the program was to develop a design method for low traffic
volume roads, more precisely considering the particular conditions of the country,
such as materials and construction processes, traffic characteristics, and climatic
conditions. The research included the study of three specially constructed test roads,
analysis of the performance of typical road networks, and full-scale tests under
repeated loads in a laboratory, which was constructed as a part of the research.
One of the main achievements of the research was the development of a
structural design method for flexible pavements (currently called DISPAV-5). This
method was originally published in 1974 and has been updated several times to
reflect changes experienced by the roads throughout the years. The final update
consisted of the migration of the original code to a JAVA environment to run the
program independently of the operating system, creating DISPAV-5 version 3.0.
DISPAV-5 is the method most widely used for the design of new construction
and the rehabilitation of existing pavements. However, its models used to predict
damage are more than 15 years old and are unable to estimate the performance of
pavements using current designs with a good level of accuracy. The original per-
formance models were developed with different traffic conditions and materials than
Mexican roads currently have. To show these changes, the main variables that
influence the performance of pavements (traffic, environment, materials and pre-
diction models) are described. Finally, future directions in which to drive the
research efforts in the Engineering Institute are mentioned.
Four Decades of the Circular Test Track … 19

2 Circular Test Track of the Engineering Institute,

UNAM

The planning, design and construction of a laboratory with full-scale facilities for
testing roads and airfield structures were fundamentals aspects of the research
program that the Engineering Institute was commissioned to undertake. The main
facilities are a circular test track and two test pits for repeated loading studies.
According to the guidelines of the research program, the circular test track had two
main purposes (Corro et al. 1972):
• controlling the parameters evaluated in the experiment, using a set number of
load repetitions with a known magnitude, and changing some characteristics of
the pavement sections constructed in the circular test track to obtain statistically
significant conclusions; and
• solving specific problems, such as evaluating specified parameters in standards
to assess their suitability and analyzing new materials.
The facility was designed to test full-scale pavements placed over various types
of subgrades and embankments up to failure, and the facility has been in operation
since February 1971. To avoid uncontrolled effects due to environmental condi-
tions, the test track is within a laboratory that is 17.5 m wide, 34.40 m long and
6.50 m high. The majority of tests were performed during the day within an
eight-hour period to decrease temperature gradients; the temperature inside the
laboratory varies from 18 to 24 °C. The following section describes the main
features of the circular test track.

2.1 Test Pit of Circular Test Track

The pavement test sections are constructed in a concrete tank that confines the
bottom and sides of the sections. To control the humidity in the unbound layers and
the drainage conditions, a very rigid concrete pit was built that is sufficiently wide
to construct a road lane and the side slope of an embankment.
According to Corro et al. (1972), to define the geometry of the pit, technical and
economic aspects were considered. Too large of dimensions would increase the cost
of constructing the sections to be tested; in addition, aspects that would influence on
the test section performance such as the construction process and the boundary
conditions were also considered. Through the analysis of the pressure distribution
using the Boussinesq theory, the researchers concluded that the influence of
boundary conditions on the outer wall and the floor of the test pit were negligible
(Corro et al. 1972).
The pit has a 14 m outside diameter, and the central ring has a 3.20 m outside
diameter. The walls and floor slab are 0.5 m thick. The depth of the gap between
the two rings is 1.50 m and is the area where the experimental sections are
20 N. Hernández et al.

Fig. 1 Cross section of the circular test track

constructed (Fig. 1). The central ring is 4.0 m deep and contains a tunnel that is
1.5 m wide and 2.0 m high for access to the control area. The pit is used to confine
the experimental sections, accommodate instrumentation devices, place anchors to
build load frames, saturate the material to the water table depth, and support the
eccentricity mechanism, which allows planetary motion of the loading frame.

2.2 Load Frame

For the load frame design (Corro et al. 1972) the following characteristics were
considered: continuous operate, loads applied within a certain range of the surface
of the pavement sections, independent suspension of the different sets of wheels,
and planetary motion to reproduce the load distribution histogram in the pavement
cross section.
The result of the points described above was a rotating device with three metal
arms with dual wheels on the ends used to apply loads; the load level can be
adjusted between 80 and 100 kN per set of dual wheels. The radius between the
wheel center and the center of rotation is 5.0 m (Fig. 1). The arms are spaced at
120° apart. A plate system regulates the ballast loads. Using an eccentric device, the
rotation axis of the loading frame describes a planetary motion to simulate the
histogram of load applications on a road, covering an adjustable strip of pavement
from 0 to 1.50 m. The device has a ball joint and bearing, which allows rocking the
frame without vertical reactions in the axis of rotation. The velocity can be varied
between 4 and 40 km/h by three 30 kW variable velocity AC motors with 40 HP
Four Decades of the Circular Test Track … 21

coupled to five-speed gearboxes. Each 62 revolutions of the frame generate a full

cycle of wheel movement on the complete pavement surface.
The instrumentation of the study sections initially consisted of electrical strain
gauges, linear variable differential transformers (LVDT), pressure cells and other
transducers to determine the dynamic deflections, pressure, water content, and
temperature.

3 Pavement Research by the Engineering Institute

In 1961, the Ministry of Public Works (now the Secretariat of Communications and
Transport, SCT) launched a research program for low traffic volume roads through
the Engineering Institute of the Autonomous University of Mexico (UNAM). Its
main objective was to develop best practices for the design, construction and
maintenance of roads with low traffic volumes in Mexico considering the charac-
teristics of materials, traffic, weather, regional conditions, conservation practices,
specifications, construction procedures, safety factors and investment programs.
Since the inception of the research program, the Institute of Engineering has
developed various projects related to the study of pavements in Mexico.
The research was based on three complementary areas: performance assessment
of three test roads located in two different regions; studies on the behavior of
several test sections on the national road network; and full-scale laboratory tests
using the circular test track (Corro and Prado 1974). The main results of this
research program are mentioned.

3.1 Research on the Behavior of the Test Roads

To meet this objective, a limited number of experimental sections were designed

and built in 1964, located on roads that were under construction, taking advantage
of the construction equipment available. Three experimental sections located in two
climatic zones were defined. Road test A was located in a climate transition zone
between tropical and subtropical, and the traffic circulating on this road was low.
Sections B and C were built on the same road separated only by a transition section.
The objective was to compare the behavior of the same pavement structure for two
sections on embankments with different characteristics (exceeding the specification
parameters for road C and below these parameters for road B). The traffic was
medium volume, and the climate zone was the cold steppes. The details for the
sections are given elsewhere (Corro and Prado 1974; Corro 1978, 1982).
22 N. Hernández et al.

The three test roads were divided into 80 sections, each 30 m long. These
sections were designed factorially to distinguish the main variables that influence
performance. The specifications and procedures of the time were used, and rigorous
control was maintained to guarantee the uniformity of the characteristics for the
materials used in the pavement and earthworks.
The main objective of this study was to establish typical tendencies in the
behavior of the pavements tested under critical conditions of differing maintenance
(Corro and Prado 1974). A sealing coat was applied to roads B and C in 1967, and
no other maintenance work was undertaken until 1978. Test road A received a
sealing coat in 1972 due to routine maintenance in that zone, without any sign of
distress in the test section. In 1978, section B reached failure conditions, after
14 years of actual service and a cumulative number of standard axles close to one
million; the average terminal serviceability was 2.5. Test section C, along the same
road, had an average terminal serviceability of 2.8 for the same traffic and years of
service.
The differences between the performance of test section B, with earthworks
below specification, and test section C, with high quality earthworks, was not
significant because the geotechnical, climatic and traffic conditions were the same;
furthermore, the materials of the surface dressing, road base and sub-base courses
were the same quality. Consequently, the long-term analysis of pavements with the
same design but constructed upon radically different earthworks was possible.
Destructive testing of sections B and C was performed in 1978. Test section A
continued as part of the study; after 18 years of service, it had a serviceability level
of 2.8 and less accumulated traffic. Test sections B and C were reconstructed in
1979 to study typical maintenance practices for low traffic volume roads and to
analyze the performance of two types of road mixes for overlays.
It was considered that the experimental sections were representative of a high
percentage of roads in Mexico, in terms of materials, specifications, construction
procedures, design, weather and traffic. The measurements obtained on the exper-
imental stages showed a linear relationship between the damage accumulated and
the logarithm of the number of load applications. In addition, the results were
significant and constituted an adequate estimation of the behavior of pavements in
road tests within a period equivalent to the life of the project, as long as they were
properly preserved.
Based on the behavior of the experimental sections, it was concluded that the
charts for pavement design used at that time were not adequate because for low
traffic volume roads, pavements were over-designed and for high traffic volume
roads, pavements were under-designed. Among the recommendations from the
research program, the most important are the following: use unified test procedures
for characterizing materials and zoning the different climatic zones of the country;
adopt the concept of equivalent traffic; and adopt the new chart developed for
thicknesses design, which was developed based on the behavior observed in the
road test.
Four Decades of the Circular Test Track … 23

3.2 Assessment of the Highway Network Performance

Planning to define this stage began in 1965 with the aim of analyzing the trends of
the general behavior of the roads in Mexico and defining the most significant
variables for the experiment. The main variables were climate, subgrade quality,
bearing capacity of the structure, traffic volume and years of service.
Climate was included at three levels, following the Köppen-Geiger classification
system: A, tropical; B, steppe and C, sub-tropical. All the other variables were
considered at two levels: high and low. The factorial combination of the variables
suggested an experiment with 96 test sections (Corro and Prado 1974). All the
sections were 500 m long and one lane wide and were located in typical roads.
From 1972 to 1979, a program to systematically evaluate the 96 sections was
carried out. The evaluation surveys were performed once or twice a year to detect
seasonal changes. To avoid problems related to traffic operation and taking into
account budget limitations, in most cases, the studies were limited to
non-destructive testing of the road sections. The main problems found in these
surveys were lack of reliable construction and maintenance records for the specific
lengths of road within the experimental sections. Additionally, the wide scatter of
collection data made it necessary to perform a large number of tests along each
section.
The typical data obtained in each experimental section were serviceability rat-
ings, May’s roughness charts, photographs of the road and details of the type of
prevailing distress, 80 kN Benkelman beam and Dynaflect deflection basins, and
rut-depth measurements.
The information collected in this phase qualitatively verified the trends observed
in the test sections and the circular track. Among the experimental data of greatest
interest were the dynamic measurements of the resistance of the bearing capacity of
the road, where it was noted that seasonal variations in maximum deflections
registered with a Dynaflect device had little importance. Because it does not show
the shape of the basin of deflections, no significant correlation was found between
the maximum elastic deformations in the pavement surface and the resistance of the
bearing capacity of the pavement.

3.3 Tests Carried Out in the Circular Test Track

Tests performed on the circular track of the Engineering Institute have been fun-
damental in the research, development and validation of different phases of the
pavement design methods used in Mexico since it began operation in 1971. There
have been 40 tests conducted on this device; to maximize the number of variables
analyzed, each ring has three sections for different pavements that are 9 m long with
1.5 m long transitions, as measured along the axis of wheel movement; therefore,
approximately 120 pavement structures have been tested to date.
24 N. Hernández et al.

All of the tests have used simple dual axis wheels with 100 kN of weight,
540 kPa of inflation pressure in the tires, and a velocity of 10 km/h. Through
experimentation, the damage factors were obtained for converting the load appli-
cations of the circular track into equivalent single-axle loads of 80 kN: the number
of applications on the track are divided by 2.3 to obtain the number of equivalent
single-axle loads. Next, the most significant trials conducted in the circular test
track over four decades of research are described.
The first experimental stage of the circular track was from 1971 to 1973 and
tested 6 rings to structural failure, representing 18 different pavement structures.
One of the objectives of this first stage was to verify the operation of the test track
and determine the times required for the construction of each ring, load application,
data acquisition and construction of a new ring in the test track. According to the
results obtained in this first stage, a full study of a ring, to the failure of the section
or by the application of one million equivalent single axle loads (ESALs) of 80 kN,
was conducted in three months, including the construction of a new ring.
The main purpose of this stage was to evaluate the influence of the thickness and
quality of the subgrade layer on the pavement behavior with a surface treatment. In
this stage, 18 structural sections were tested (six test rings) with pavements com-
prising a base layer with a surface treatment on subgrades and earthwork of the
same clayey-loam material. However, the sections had different strength charac-
teristics, obtained by varying the compaction degree and test conditions, such as the
construction procedures: sections were made impermeable or saturated, flooding the
slope zone to establish the water table at a depth of 0.6 m beneath the ring surface.
Another important stage of experimentation that was carried out in the circular
track was assessing the behavior of thin, cold in-place asphalt mixes. To evaluate
their performance, tests were performed on rings 21 through 23 between the years
of 1980 and 1982. The variables analyzed in the asphalt mixes were type of
aggregate (crushed basalt and river aggregate) and grading of the asphalt mix
(coarse, dense and fine). Based on the specifications of the time, the first tests at
elevated temperature were performed (45–50 °C) compared with the temperatures
commonly performed (18–24 °C). The system used to control the heating of the
circular track sections consisted of infrared spotlights, which were lit according to a
predetermined cycle to control the temperature within the specified range; this
system had a capacity to heat the pavement to 60 °C.
Among the general recommendations of this stage of experimentation, it was
proposed to amend the upper limit of the grading zone because the asphalt mixes
with fine graduation had better performance in the test sections. Another recom-
mendation was to review the design criteria of asphalt mixtures with the aim to
perform tests that represent the mechanical behavior of asphalt mixtures.
The study of fatigue cracking in flexible pavements began in 1984, with testing
of ring 24. In this experiment, a hot-mix asphalt (HMA) produced by a plant near
the Engineering Institute was used. The main variable to evaluate was the thickness
of the HMA layer in pavement structures with similar maximum deflections to
maintain the tensile strain at the bottom of the HMA layer at similar levels. Because
strain gauges were not used in pavements sections, the tensile strain at the bottom of
Four Decades of the Circular Test Track … 25

Fig. 2 Pavement section after the test at the circular track

the HMA layer was estimated indirectly through surface deflections by layer elastic
analysis.
The sections of ring 24 were composed of three layers (HMA, granular base and
embankment of clayey-loam material); the HMA thicknesses were 30, 60 and
90 mm. The sections were designed for the pavement to fail within a short period of
load applications and with a predominant damage of fatigue cracking.
The test of the sections was performed at an average laboratory temperature of
21 °C (with ranges of ±3 °C). The initial cracking was observed simultaneously in
the three sections for a transit of 10,000 ESALs of 80 kN, showing a scattered fine
crack pattern. The test continued until 100,000 ESALs were applied, when the
cracking pattern was dense and dominated by fine and intermediate cracks. The
section that experienced more damage was the HMA layer 60 mm thick, and the
section with 90 mm thick HMA was the best performing. Figure 2a shows fatigue
cracking after the test, and Fig. 2b shows a cut of the section.
Extensive research was conducted continuously in the laboratory to characterize
the fatigue behavior of HMA used in the study, and diametral compression tests
were performed with cyclic loading on laboratory-compacted samples and cores
extracted from the layer placed on the circular track. At the time of the study, there
was no fatigue failure criterion for flexible pavements in Mexico, so the results of
the study showed the need to develop more rational criteria for the characterization
of materials and models to estimate fatigue damage in asphalt pavements.
The study of fatigue cracking continued until 1994, when ring 32 was tested. The
main objective was to verify the design method for asphalt pavements. These three
sections were designed for one million ESALs of 80 kN. A confidence level was
not applied to obtain structural failure at the end of the application of loads;
therefore, the confidence level was 0.5. According to the design, it was expected for
the sections to present fatigue failure at the finished load application. The pavement
structures were composed of HMA, granular base, sub-base and subgrade; the
materials of each of the layers complied with the specifications and standards.
The HMA thicknesses were 30, 60 and 90 mm, those of the base were 200 and
150 m, and those of the subbase were 390, 380 and 320 mm.
26 N. Hernández et al.

During the time of the study, different tests were performed to evaluate per-
formance under loads, including measuring the rut depth, measuring the deflections
with a Benkelman beam and Dynaflect device, and measuring the tensile strains in
the HMA using strain gauges.
Parallel testing was conducted of the dynamic modulus (ASTM D3497), the
unconfined compression under sinusoidal cyclic loading to evaluate the permanent
deformation, and indirect tensile stress under sinusoidal cyclic loading to assess
fatigue. At the end of the test, cores were extracted in sections to assess the actual
thickness and mechanical properties of the materials.
The three sections behaved in accordance with the estimates from the developed
design method. At the end of the test, it was observed that the depth of the rut was
equal or inferior to the permissible depth (12 mm, in 20 % of the section). During
the study, there was no fatigue cracking, although the tensile strain exceeded the
tensile strain estimated in design. During the study, no other structural defects were
observed. These results confirmed the suitability of the developed design method,
which had a theoretical and experimental basis.
The latest study on the circular track was in the year 2007, related to the use of
special concrete that included fragments of shredded tire rubber. The aim of the
study was to evaluate the behavior of this type of concrete for residential areas,
lowering the cost of concrete and solving ecological problems by using shredded
old tires (Corro and Rangel 2008). Nine sections were built in the circular track and
subjected to one million ESALs of 80 kN. The main variable analyzed was the type
of concrete used in the slabs: three sections of normal concrete, three sections with
small pieces of crushed rubber and three sections with medium fragments of cru-
shed rubber were used. All of the sections had a slab 150 mm thick. At the end of
the scheduled number of load repetitions, good behavior was observed in all of the
sections evaluated, without providing evidence of any damage in the slab sections.
Figure 3 shows the sections of concrete in the circular test track.

Fig. 3 Image of the slabs

with crushed tire rubber in the
circular track
Four Decades of the Circular Test Track … 27

3.4 Asphalt Pavement Design Method

The first pavement design guide (currently DISPAV-5) was presented in 1974 by
the Institute of Engineering, which was based on an experimental-theoretical cri-
terion for the design of flexible pavements. The experimentation carried out in road
tests, the evaluation of existing highways and testing in the circular test track during
1965–1973 were the foundation for this first design guide.
The proposed design method consisted of a series of design charts that con-
sidered variables, such as the damage coefficients for the axis type or type of
vehicle, equivalent cumulative traffic, and structural design of flexible pavements
for different types of roads (primary and secondary). It was recommended that these
variables were evaluated within a general framework to reach economic solutions.
In addition, the method proposed the standardization of procedures for laboratory
tests for characterizing materials.
The method employs the concept of bearing capacity in cohesive soil and
Boussinesq’s theory of the distribution of vertical stress (σz) (deduced for a circular,
static flexible plate uniformly supported at the surface of an elastic, homogeneous,
isotropic medium) in the particular case of a multilayered structure of uniform
relative strength, subjected to repeated loads of an equivalent single axle with a
defined static weight of 80 kN and with a constant impact coefficient (I). It is further
assumed that the California bearing ratio at the site (CBRz) is a good indicator of
the bearing capacity of the different layers (Corro 1978).
Failure in a layer at the surface of the highway is analyzed in light of the
hypothesis that there is a linear relation between the logarithm of resistance (log
CBRz) and the logarithm of the cumulative number of equivalent 80 kN single axle
loads (log ΣL). For any layer of depth z, the concept is generalized, multiplying the
resistance at the surface (log CBRz=0) by Boussinesq’s coefficient of influence (Fz),
assuming by definition a structure of constant relative strength.
Analyzing the data referencing this hypotheses, based on experimental evidence,
made it possible to establish the equations of the design charts for different degrees
of reliability for the minimum strength required in any layer so that the structure
might support a determined number of equivalent applications (ΣL) before surface
deterioration reached the level defined as functional failure of the highway.
The design charts presented are limited to the typical case of the structures
employed in Mexico, where the thickness of the asphaltic concrete surface rarely
exceeds 75 mm and where the others layers of the structure consist of granular
materials or fine soils mechanically stabilized by compaction. In the case of thick
asphalt pavements, the design hypothesis would vary, and it would be necessary to
take into account the radial stress that can cause failures due to fatigue in the HMA.
From 1994 to 1999, the DISPAV-5 design method underwent major upgrades
that established the current design method. In 1994, it was expanded to consider
fatigue cracking in the asphalt layers and thus consider the damage that appeared on
the roads of Mexico. In 1996, it developed an interactive tool programmed in an
MS-DOS operating system, which incorporated all the information on pavement
28 N. Hernández et al.

design developed so far. In 1997, the method was updated to consider the maximum
legal loads for trucks transport in Mexico.
Among the improvements made with respect to the original method, published
in 1974, was incorporating a mechanistic model to determine the tensile horizontal
strain that generates fatigue damage in asphalt mixtures based on experimental
studies on asphalt mixtures from 1985 to 1999. A new model was developed to
design structures of high-standard expressways (high traffic), considering a
mechanistic-empirical approach for the two main modes of pavement failures,
which include permanent deformation of the unbound pavement layers and fatigue
cracking of the asphalt-bound layers (DISPAV-5 version 2.0). Additional infor-
mation about this update can be found elsewhere (Corro and Prado 1997).
The last update was in 2012; the scope was defined in the source code migration
to a JAVA environment to run the program independently of the operating system.
The model created in the DISPAV-5 version 2.0 was maintained, but the interface
and presentation of the results were modernized. One of the improvements in the
new version (DISPAV-5 version 3.0) was updating the weights and dimensions of
heavy trucks (Corro et al. 2014).

4 Main Factors Affecting the Performance of Pavements

in Mexico

The DISPAV-5 pavement design method is the most commonly used design tool in
Mexico. Currently, a large number of newly built pavements and most road
reconstructions are designed using this method. However, according to the brief
description of the method presented earlier, the models incorporated in DISPAV-5,
have a number of limitations in representing the current condition of the pavements.
Therefore, this chapter describes the main factors influencing the performance of
flexible pavements, and the prevailing conditions in Mexico are particularly
described.

4.1 Traffic

One of the most significant improvements offered in the M-E Design Guide is the
methodology to account for highway traffic volumes and loads. This enhancement
proposes replacing the number of equivalent single axle loads (ESALs) by actual
axle load distributions per axle type per vehicle class. These distributions are often
referred to as axle load spectra. Traffic is indeed one of the most critical pavement
design inputs, determining how a pavement will perform and how long it will last.
Traditionally, traffic data are associated with a high level of uncertainty (Prozzi
et al. 2008).
Four Decades of the Circular Test Track … 29

100%
90%
80%
70%
60%
Percentile

C2 (Class 5)
50% C3 (Class 6)
40% T3S2 (Class 9)
T3S3 (Class 10)
30% T3S2R4 (Class 13)
U.S. Legal
20% Limit
Mexican Legal
10% Limit for this class

0%
0.0 200.0 400.0 600.0 800.0 1000.0
Gross Vehicle Weight, (kN)

Fig. 4 Gross vehicle weight from Las Rajas, México State (Central Zone)

In Mexico, the uncertainty is even greater because there is a high percentage of

truck vehicles with overload. The percentage of overload that exists on the roads of
Mexico can be observed in a study published by the Mexican Institute of Transport
(Mayoral et al. 2014). The study focuses on four main roads to Mexico City, and
the report shows the percentage by vehicle type and the percentage of overload for
each type of vehicle. The vehicles with the highest percentage of overload on the
maximum weight allowed were trucks such as T3S3 (six axis) and T3S2R4 (nine
axes); the percentage of overload for the second vehicle was between 27.6 and
50.7 %.
Figure 4 shows the gross vehicle weight obtained from “Las Rajas”
weigh-in-motion (WIN) station located in Mexico State. The articulated vehicle
average weight (more than five axles) exceeds 300 kN. An interesting case is the
T3S2R4 truck, which is similar to the USA class 13. This Mexican truck has an
average weight of 540 kN, which is more than 40 % over the Mexican legal load
limit. To show the percentage overload, the dotted green line shows the legal load
limit in the USA (363 kN) for all freight vehicles.
Garnica and Hernández (2013) recently published representative axle load
spectra from the Mexican Federal road network. All of these data were obtained
from different WIM stations from different regions in Mexico. They defined four
types of axle load spectra based on the observed overload level. However, it is
necessary to extend this study to define load spectra based on other factors and not
just based on the percentage of overload observed. One example is the methodology
defined by Lu and Harvey (2009), evaluating road characteristics (number of lanes,
functional classification, etc.) and socioeconomic factors (population density, land
use, etc.).
30 N. Hernández et al.

4.2 Climate

Environmental conditions have significant impact on pavement design and per-

formance. These conditions are represented as the effects of weather and climate on
the strength, durability and load bearing capacity of the pavement. In flexible
pavement, the surface layer is usually made of hot-mix asphalt (HMA), which is a
viscoelastic material, and its behavior is highly related to its temperature, i.e., HMA
responds similar to an elastic solid under low temperature and strain conditions;
conversely, it also acts as a viscous material at high temperature in the sense that the
deformation due to traffic loading cannot be fully recovered within a finite time
period under the unloading condition. Thus, an accurate prediction of the temper-
ature profile in the HMA layer is desired when selecting the asphalt binder and
predicting asphalt pavement responses under traffic and environmental loadings
(Wang 2010).
According to the National Institute of History and Geography (INEGI), Mexico
has a variety of climates: arid territory in the north, warm and sub-humid in the
south and southeast, and cold or temperate climates in the mountain geographic
regions. Because of this large variety of climates, it is important to consider the
influence of climate on the behavior of materials both for their selection (in the case
of asphalt grade PG) and pavement design as well as considering their variation
during different seasons in the structural design of the pavement.
In Mexico, there is little or no research on this subject; one of the main con-
straints is the lack of climate information to develop models to estimate the vari-
ation of the temperature or humidity in the layers of the pavement structure.
However, a recent study by Schwartz et al. (2015) presented new sources of climate
data henceforth unknown to pavement engineers. This data source, the Modern-Era
Retrospective Analysis for Research and Applications (MERRA), developed by the
National Aeronautics and Space Administration (NASA) for its own in-house
modeling needs, provides continuous hourly weather data starting in 1979 on a
relatively fine-grained uniform grid. MERRA is based on a reanalysis model that
combines computed model fields (e.g., atmospheric temperatures) with ground-,
ocean-, atmospheric-, and satellite-based observations that are distributed irregu-
larly in space and time. The result is a uniformly gridded dataset of meteorological
data derived from a consistent modeling and analysis system over the entire data
history. MERRA data are provided at an hourly temporal resolution and at 0.5
degrees latitude by 0.67 degrees longitude (approximately 55 km by 73 km at
mid-latitudes) spatial resolution over the entire globe (Schwartz et al. 2015).
With this database and the information generated by the National Water
Commission (CONAGUA), the models developed in other countries such as the
Enhanced Integrated Climate Model (EICM) can be calibrated and models for
Mexico can be developed to estimate the variation of climatic conditions in
pavement materials.
Four Decades of the Circular Test Track … 31

4.3 Pavement Materials

The characteristics of the materials is one of the most important input parameters
during pavement design. Within the extensive research conducted at the Institute of
Engineering, great efforts have been made to incorporate the mechanical properties
of the materials forming the flexible pavements; however, these practices were
never adopted as common practice during the design process. In particular, the
evaluation of the mechanical properties for the design of asphalt mixtures is a
procedure that has become widespread since 2008, due to implementing the design
methodology of high performance asphalt mixtures called “AMAAC protocol,”
based on the superior performing asphalt pavements (Superpave) system developed
by the Strategic Highway Research Program (SHRP).
Four levels are defined in this protocol, according to the traffic level. The first
level has the same Superpave requirements; in the next levels, performance tests are
specified: Hamburg Wheel Tracking is specified for the second level, stiffness tests
(Dynamic modulus, |E*|) are specified for the third level, and fatigue tests
(four-point bending beam) are specified for the fourth level.
Although there is some experience in this type of testing, so far, this procedure is
limited to the design of the asphalt mix, i.e., to determine the proportions of its
components without using the mechanical properties evaluated in the design of
pavements. This limitation is mainly due to the lack of a method in which to
incorporate these features. It is also important to calibrate or develop predictive
models of the mechanical properties of the materials to be used in projects that are
not feasible for laboratory tests.

4.4 Performance Prediction Models

Performance prediction models developed in the research conducted by the Institute

of Engineering were successful in their time and were the basis of the DISPAV-5
design method. The level of service demanded by users currently requires large
investments to maintain the road network under suitable conditions, so performance
models that adequately simulate real pavement performance are indispensable. This
required level of reliability is not achieved by the models included in the DISPAV-5
because technological development has generated a greater variety of materials (for
example, modified asphalts) and construction procedures (for example, mixtures
with RAP) with behavior that cannot be estimated from these models. Furthermore,
due to the age and lack of durability of roads,, the rehabilitation or reconstruction of
pavements in the road network has become increasingly common, a function for
which this method was not developed.
32 N. Hernández et al.

With the development and publication of the MEPDG guide, the way pavements
are designed is changing and is directed toward the development of
mechanistic-empirical pavement design methodologies. To take advantage of
research conducted in this area, the proposal is to evaluate emerging performance
prediction models in different design methods recently established to determine
which is more feasible to be used for the conditions of Mexico considering the
influence of the variables mentioned above.
In the second stage, the evaluation of the models considered more convenient
should be accompanied by experimentation in the circular test track to verify their
applicability to road conditions in Mexico.

5 Future Directions

Currently, the pavements lab at the Engineering Institute is undergoing a transition,

as it renews the research staff and garners increasing interest from graduate students
to conduct pavement research. Currently, the Institute is defining the future direc-
tions to be followed, giving more importance to the current problems in order to
generate solutions to these problems. The areas of current research focus are as
follows:
• characterization of asphalt used in road construction, including modified and
conventional asphalts through rheological properties and parameters represent-
ing service behavior to validate the performance-based specifications that are
currently in use;
• study of the mechanical properties of asphalt mixtures (dynamic modulus and
fatigue) and calibration and development of predictive models of these prop-
erties according to the materials used;
• characterization of granular materials and soil to incorporate their properties into
design and quality control of these materials in the performance of the
pavement;
• evaluation of new technologies such as mixtures with RAP, warm mix, and use
of construction residues to verify their feasibility in road construction;
• development of a new pavement design method that considers the current
condition of the roads in Mexico; and
• modernization and instrumentation of the circular test track for use in pavement
investigation.
Linking projects with highway agencies and the private sector is an important
step for carrying out the points above because without their financial support,
performing the research is more difficult. Furthermore, these groups can implement
the obtained results.
Four Decades of the Circular Test Track … 33

6 Conclusions

For four decades, research by the Institute of Engineering has contributed to the
development of pavements in Mexico; this research has generated important
products that even now contribute to pavement practices in Mexico. One of the
main objectives of the project that began in 1961 was the design and construction of
a device that would allow the laboratory testing of full-scale pavement sections. The
result was the circular test track, which began operation in 1971 and is currently still
in service. In this device, approximately 120 sections of pavement have been tested
and approximately 45 million equivalent axes of 80 kN have been applied.
The main outcome of the research was to develop a method of structural design
for flexible pavement known as DISPAV-5; this method has been updated at dif-
ferent times to incorporate the road conditions in pavement design throughout these
four decades. It is currently the most widely used method of design in Mexico for
new pavements and the rehabilitation of existing pavement.
The latest update of the performance models of DISPAV-5 took place more than
15 years ago, decreasing the reliability of the results obtained with this method.
Differences in traffic loads and materials currently used increases the uncertainty in
the results obtained with DISPAV-5 in the pavements design process, so it is urgent
to develop a comprehensive update of this method.
Clearly, this update cannot be conducted in the short term; therefore, the future
directions to be followed in the pavement lab must be the foundation for an upgrade
or development of a method for more reliable pavement design. A very important
aspect of the proposed projects is linking the research with the road agencies and
the private sector.

References

Corro, S. (1970). Diseño de pavimentos flexibles. Comportamiento de los tramos experimentales.

Instituto de Ingeniería, UNAM, Informe 240, México, DF.
Corro, S., Camacho, A., & Aguirre, M. (1972). Pista circular para estudio del comportamiento
estructural del pavimentos. Proyecto SA -649. Archivo histórico, Instituto de Ingeniería,
UNAM.
Corro, S., & Prado, G. (1974). Diseño estructural de carreteras con pavimento flexible. Instituto
de Ingeniería, UNAM, Informe 325, México, DF.
Corro, S. (1978). Recommendations for the structural design of flexible pavements. XVth World
Road Congress, Permanent International Association of Road Congresses, Question II,
Mexico, D.F. (oct 1975). Sobretiro Series del Instituto de Ingeniería, Informe E-35, Instituto de
Ingeniería, UNAM, México, DF.
Corro, S. (1982). Full scale testing applied to pavement design. International Colloquium: Full
Scale Pavement Tests, Federal Institute of Technology (ETH), Technical Report 50, Zürich
Corro, S., & Prado, G. (1997). Diseño estructural de pavimentos asfálticos, incluyendo carreteras
de altas especificaciones. Series de Instituto de Ingeniería, UNAM, Informe 598. México, DF.
34 N. Hernández et al.

Corro, S., & Rangel, A. (2008). Estudio del comportamiento estructural, a fatiga de pavimentos de
concreto normales y con agregados de hule. Informe interno Instituto de Ingeniería, UNAM,
México, DF.
Corro, S., Castillo, G., Ossa, A., & Hernández, J. (2014). Estudio del comportamiento estructural,
a fatiga de pavimentos de concreto normales y con agregados de hule. Informe interno
Instituto de Ingeniería, UNAM, México, DF.
Garnica, P., & Hernández, R. (2013). Manual de usuario IMT-PAVE 1.0. Instituto Mexicano del
Transporte. Documento Técnico No. 53. Sanfandila, Qro.
Lu, Q., & Harvey, J. (2009). Estimation of truck traffic inputs for mechanistic–empirical pavement
design in California. Transportation Research Record: Journal of Transportation Research
Board, 61–72 (1945)
Mayoral, E., Cuevas, A., & Mendoza, A. (2014). Criterios de ubicación de estaciones fijas
automatizadas para el control de peso, dimensiones y velocidades de los vehículos que
circulan por las carreteras federales, Instituto Mexicano del Transporte, Documento Técnico
No. 397, Sanfandila, Qro.
Prozzi, J., Prozzi, J., Villa, J., & Middleton, D. (2008). Integration and consolidation of border
freight transportation data for planning applications and characterization of nafta truck loads
for aiding in transportation infrastructure management: Second year. Texas: Texas
Transportation Institute, The Texas A&M University System College Station.
Schwartz, C., Elkins, G., & Rui Li, B. (2015). Evaluation of long-term pavement performance
(LTTP) climatic data for use in mechanistic-empirical pavement design guide (MEPDG)
calibration and other pavement analysis. Report No. FHWA-HRT-15-019. Federal Highway
Administration.
Wang, D. (2010). Analytical solutions for temperature profile prediction in multi-layered
pavement systems. Dissertation for the degree of Doctor of Philosophy in Civil Engineering in
the Graduate College of the University of Illinois at Urbana-Champaign.
 Part II
Establishment of New Accelerated
Pavement Testing Facilities
Design and Engineering Challenges
in the Development of the FAA’s Full Scale
Accelerated Pavement Test Facility
NAPMRC

Murphy Flynn, Navneet Garg, Wilfredo Villafane

and David Traverzo

Abstract Accelerated Pavement Test Facilities are not a common type of facility
for Architecture and Engineering (A/E) firms to have experience with. The Federal
Aviation Administration (FAA) commissioned the design of their APT Facility, the
National Airport Pavement and Materials Research Center (NAPMRC) in 2012.
Personnel from the FAA’s National Airport Pavement Test Facility near Atlantic
City, NJ USA worked with an A/E firm to develop a facility to utilize the FAA’s
Heavy Vehicle Simulator for Airfields (HVS-A). The HVS-A commissioned by the
FAA and built by Dynatest Corporation is the largest HVS ever constructed by
Dynatest. Challenges were encountered in working with designers to build pave-
ments “designed to fail” and to gather information from those pavements as they
fail. This included educating the designers to the build, test, fail and repeat concept
of accelerated pavement testing. The NAPMRC encompasses both indoor and
outdoor test pavements. Electric power required to run the HVS-A needed to be
provided throughout the facility to make full use of the six lanes of test pavement
being constructed. Data collection and communications networks that could deliver
pavement sensor information were designed with flexibility to handle a variety of
sensor for the original and future test pavement builds.

1 Introduction

In 2010 the FAA’s Airport Technology Research and Development Branch began
the procurement of a Heavy Vehicle Simulator for Airfields (HVS-A) from
Dynatest Corporation. The custom made HVS-A would ultimately be the largest
ever built by Dynatest with the capability of enhanced control systems, a larger

M. Flynn
N. Garg (&)
W. Villafane
D. Traverzo

Federal Aviation Administration, Airport Technology R&D Branch,
Atlantic City, NJ, USA
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 37

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_3
38 M. Flynn et al.

wander capability than similar highway models and the abilty to apply loads up to
100,000 lbs. The design build contract was awarded in August 2011 and delivered
in October 2013.
The as-built specifications of the FAA HVS-A are as follows:
• 125 ft. long, 16 ft. wide and 13 ft. high
• Wheel loads ranging from 10,000 to 100,000 lbs.
• Pavement temperatures up to 150 °F
• Test speeds—0.17 to 5 mph
• Single and Dual-Wheel configuration.
• Single wheel will be radial aircraft tires size 52 × 21.0R22
• Dual wheel assembly is designed to accommodate smaller tires (B-737-800)
• Wander Width—6 ft.
In 2010 the FAA began planning a new test facility, the National Airport
Pavement and Materials Research Center (NAPMRC) to be the home of the FAA’s
HVS-A. The facility would be located adjacent to the FAA’s existing National
Airport Pavement Test Facility (NAPTF), a fully enclosed instrumented test track
900 ft. long by 60 ft. wide with the capability of applying full scale aircraft gear
loads, which is now in its 16th year of operation. The NAPTF’s rail based test
vehicle can be configured to represent two complete landing gear trucks, each
having 1–10 wheels per truck. Wheel loads can be independently varied up to
75,000 lbs. per wheel (Fig. 1).
The climate in New Jersey is cold from October through April with snow
expected from late November though March. The HVS-A is designed to heat the
pavement while testing. Therefore the new facility would need to have multiple test
sections indoor as well as outdoor. Our initial conceptual plan included four cov-
ered indoor test strips and two outdoor test strips (Fig. 2).
Both the outdoor and indoor test areas would require modern data collection
networks capable of hosting a variety of sensors and have the flexibility to
accommodate potential sensors of the future.
Strict emissions laws in NJ require permits for running the HVS-A on diesel
fuel. With emissions permit restrictions and the need for the pavement testing to

Fig. 1 FAA HVS-A

Design and Engineering Challenges in the Development … 39

Fig. 2 NAPMRC conceptual plan

occur indoors for half the year, use of electric power is the preferred option. The test
areas would need sufficient electric service to power the HVS-A gear and the
electric heaters.
A small office area was required to house the operators of the facility. It was
envisioned that a staff of three to four employees would operate the facility full
time.

2 The FAA Technical Center Construction Approval

Process

The Airport Technology Research and Development Branch is located on the

campus of the FAA’s William J. Hughes Technical Center near Atlantic City,
New Jersey, USA. The FAA Technical Center is the nation’s premier air trans-
portation system laboratory. The Technical Center also happens to be located in one
of the nation’s most densely populated states and therefore is subject to tight
regulation regarding continued development.
40 M. Flynn et al.

Approval to build the NAPMRC facility began with a request submitted to the
Master Planning Board of the Technical Center. Conditional approval was granted
by the board pending the outcome of an Environmental Assessment study (EA) as
required by state and Federal law. To receive approval to build, the EA must result
in a Finding of No Significant Impact (FONSI). Preparation of the EA can take
several months followed by many more months of review by state and federal
regulatory agencies and is ultimately subject to public review and comment.
However, past experience suggested the chosen site had a high likelihood of
gaining reaching a FONSI and ultimately approval. The Environmental Assessment
for development of the approximately five acre site began in March of 2012 and
concluded in February of 2013 with the necessary FONSI.

3 Design

Concurrent with the conduct of the EA, the scope of work was developed for design
of the NAPMRC facility. The design scope of work was completed in May of 2012.
However, award of the design contract was withheld pending the initial results of
the EA to ensure there were no issues that may prevent development (Fig. 3).
The FAA Technical Center has two methods to award a design contract: (1) The
request for design can be advertised for competitive bids under a firm fixed price
contract; or (2) the design contract can be negotiated under the Technical Center’s
Indefinite Delivery/Indefinite Quantity (ID/IQ) contract for Architectural and
Engineering services, a 5 year A&E services contract with a single company. Under
the ID/IQ contract the FAA has the ability to meet with the designers prior to
submission of their proposal to ensure a thorough understanding of the work
required. Since accelerated pavement test facilities are not typical for a designer to
have experience with, it was decided the ID/IQ process was the best option for
design of the NAPMRC. In August of 2012 the contract for design of the
NAPMRC was awarded.

Fig. 3 Major project tasks

Design and Engineering Challenges in the Development … 41

The process of building research pavements for testing to failure, removal,

rebuild and repeat was new a challenge for the design team. To educate the
designers the FAA required the team to visit to other HVS sites in the United States.
Those sites included:
• U.S. Army Corps of Engineers Engineering Research and Development Center
(ERDC) at Vicksburg, Mississippi (the only other HVS for airfields).
• Florida Department of Transportation Accelerated Pavement Testing site in
Gainesville, Florida.
• University of California at Davis, Institute of Transportation Studies, Pavement
Research Center.
In addition, the designers had the benefit of being able to visit our National Airport
Pavement Test Facility on a regular basis.

4 Outdoor Test Area

The FAA Technical Center surrounds the public drinking water supply of Atlantic
City, NJ. Both groundwater wells and a surface water reservoir are downslope and
within a half mile of the NAPMRC site. Therefore collection of storm water is
highly regulated. State and Federal regulations required the capture of all surface
run off as part of the design. To direct the flow of storm water the entire test
pavement was sloped at 1 % lengthwise from South to North. To contain the flow a
3 in. high perimeter curb surrounds the test pavement on the sides and downslope
end. The curb is designed with a broad gentle transverse slope such that a vehicle or
the HVS-A could easily traverse it. The up slope end of the pavement was left open
to allow construction and paving equipment in.
The HVS–A has two 475 gallon diesel tanks on board as well as many
mechanical points that require grease, both of which are potential surface and
ground water contaminants as well as oils from the asphalt surface itself. At the
down slope end of the test pavement a 12 in. trench drain runs transversely across
the test lanes and parallel to the perimeter curb the full width of the test pavement to
capture storm run-off. Water flows from the trench drain into a 2000 gallon
oil-water separator before finally discharging to a bio-retention basin on site.
When constructing test sections to represent airport pavements, strict adherence
to FAA construction standards (AC 150-5370-10G) is required. FAA specifications
require compaction of the aggregate base and subbase layers to 100 % of modified
proctor (ASTM D 1557) when subjected to heavy aircraft loading. Proper aggregate
moisture during construction is critical to achieving this level of compaction.
A 3 in. PVC water line with multiple hose outlets was installed at the NAPTF
along 1100 ft. of its overall length to provide water for construction, concrete
curing, or pavement surface wash down. The water line was installed above ground
after construction of the NAPTF and is exposed to potential freezing and therefore
must be drained after every use during the coldest months of the year.
42 M. Flynn et al.

Fig. 4 NAPMRC test

P-401 HMA/WMA SURFACE
pavement cross section 5 inches (127 mm)

P-209 CRUSHED STONE BASE

12 inches (305 mm)

P-154 SUBBASE
12 inches (305 mm)

SANDY SUBGRADE
CBR 15

Table 1 Binder grade by test lane

Test lane 1 2 3 4 5 (inside) 6 (inside)
Binder 76–22 64–22 76–22 64–22 76–22 76–22
(PG Grade) Warm mix Warm mix Hot mix Hot mix Hot mix Warm mix

An underground water line was included in the NAPRMC design. The water line
features three non-freeze ground hydrants recessed below finished grade and evenly
spaced along the length of the test area adjacent to the eastern curb. This water
system does not need to be drained during the colder months of the year.
A total of six test lanes, four outside and two inside, were constructed with the
same cross section (Fig. 4). The asphalt concrete surface course for each test lane
called for a 1 in. maximum aggregate size aggregate blend meeting gradation 1 of
FAA AC 150-5370-10G, P-401 Hot Mix Asphalt Pavements. The gyratory design
criterion for aircraft weighing over 60,000 lbs. (75 gyrations) was used for all test
lanes. Mix types and PG grading of binders for each test lanes are shown in Table 1.
Each test lane contains three distinct instrumented test zones; North, South and
Center. The North and South test zones have a significant number of sensors. The
center test zone contains minimal sensors and is intended as a backup test area if
needed.

5 Indoor Test Area

The FAA intends to use the HVS-A to test pavements at high temperatures (130–
150 °F). The requirement for a covered test area large enough to keep the HVS-A
occupied during the colder months from about mid-October to mid-April is
Design and Engineering Challenges in the Development … 43

necessary for efficient use of the HVS-A. While our HVS-A is equipped with an
insulated chamber, testing outside all year greatly reduces the amount of time the
pavement can be tested at the desired high temperatures while at the same time
increasing the cost to heat the pavement with snow and wind acting upon it.
The FAA’s conceptual plan for NAPMRC included four covered indoor test
lanes and two outdoor test lanes. During the design phase it was determined that a
building large enough to cover four test lanes (270 ft. long × 150 ft. wide) would
cost nearly four times a building half the width and would likely exceed the
available budget. The decision was made to reduce the width of the building to
approximately 75 ft. with just two covered test lanes while increasing outside test
lanes to four.
When approaching the building choice for NAPMRC, the FAA’s choice was
heavily influenced by two of our existing buildings. The NAPTF is a 1200 ft. ×
100 ft. × 47 ft. high pre-engineered metal building. Even with 42 skylights it can
be still be too dark without significant secondary lighting. In 2008 the FAA erected
a fabric skinned metal arch building for processing drying and mixing subgrade and
aggregates for NAPTF test pavements. The building has a white fabric skin over the
majority of the roof span. The translucent fabric allows enough light that no
overhead electric lights needed to be installed. The arch style frames of the building
were erected on a raised foundation in order to provide sufficient clearance for
operation of heavy equipment throughout the interior. The cost of the building was
lower than a similar sized pre-engineered metal building. The fabric skin is war-
rantied for 15 years.
For the benefits of reduced lighting requirements and cost described above, the
FAA’s preference was for another fabric skinned building. The design used a
similar approach as our subgrade processing building. A 72 ft. wide × 270 ft. long
arch style building with a nominal internal clearance of 26 ft. 6 in. was chosen as
the basis for design. The construction contractor actually supplied a manufacturer’s
stock size building with an internal clearance from base to inside peak of
28 ft. 4¾ in. The building sits atop a raised foundation with a minimum height
of 3 ft. 1 in. providing a minimum internal ceiling height of 31 ft. 5 in. at the
center of the arch.
The pavement surface inside the building slopes longitudinally south to north
with a slope of approximately 1 %. Similar to the outdoor test area, there is a
transverse trench drain at the bottom of the slope to capture any water used to wash
the pavement surface. However since the potential volume of water is much less
than outside, the run-off is channeled to a much smaller 57 gallon capacity oil and
sand interceptor prior to flowing to the sanitary sewer.
The building design called for a single overhead coiling door at the south end to
provide clearance for moving the HVS-A in and out of the building. The clear
opening dimensions are 48 ft. wide × 20 ft. high (Fig. 5). During construction it
was determined that the tubular arches of the membrane building could not support
such a large door, both in weight and wind loads, due to its large surface area.
Therefore the frame of the door had to be self-supporting and independent of the
building. The shop drawings called for 16 in. × 12 in. × ½ in. thick steel box
44 M. Flynn et al.

Fig. 5 NAPMRC fabric arch building and large coiling overhead door

columns with a 16 in. × 16 in. × ½ in. overhead box beam. The columns were
anchored upon 14 ft. × 9 ft. × 3 ft. thick steel reinforced spread footers set 4 ft.
below finished grade. The opposite end of the building has two 20 ft. × 20 ft.
overhead coiling doors. These doors were small enough that no special footers or
columns were required.
The test pavement building is neither conditioned nor ventilated. It was felt the
total of 1760 ft.2 of door opening at the building ends would provide sufficient
ventilation during construction. If additional ventilation is required such during
paving, power outlets are provided throughout the building for supplemental fans.

6 Maneuver Areas

The FAA HVS-A is 125 ft. long. It can move under its own power and is controlled
by a remote control. Since the NAPMRC facility was designed with both indoor
and outdoor test areas, the site had to be designed with enough clear space to move
the HVS-A from inside to the outside and back again. At the time of the facility
design, the HVS-A had yet to be delivered. The exact details of the HVS-A’s
maneuverability and control under its own power were an unknown. Despite the
unknown, site limitations allowed only 130 ft. of clear space at the end of the indoor
test area building for moving the HVS-A straight out of the building. It was
unknown if the HVS-A could be “crabbed” out of the end door using its limited
steering capability.
The first choice of the designers was to pave nearly the entire site to allow the
HVS-A to roll around virtually unimpeded. However, this greatly increased the
need for storm water retention as well as not providing a very aesthetically pleasing
Design and Engineering Challenges in the Development … 45

Fig. 6 GrassPave2 stabilized

turf

site. At the request of the FAA, the designer was tasked with researching a stabi-
lized grass surface for as much of the proposed maneuver area as possible. The
engineers chose GrassPave2 to provide a stabilized turf surface for the HVS-A to
maneuver on. GrassPave2 consists of a layer of plastic ring cells connected by a
web backing. Provided in rolls, it is laid over a high quality crushed stone base. The
rolls of cells are connected to each other and filled with washed coarse sand. The
entire structure is then covered in sod or hydro-seeded to produce a grass covered
but porous surface that can function as a paved surface (Fig. 6). This design choice
resulted in a 23,400 ft.2 reduction of asphalt paved surface.

7 Control Building

The FAA required office space to act as a “control room” for operation of the
NAPMRC. The building was to house the manager of the facility and one or two
operators of the HVS-A and associated data systems. A storage and work space was
also desired for spare parts, tools and a small workshop area. The concept plan
estimated a minimum footprint at about 24 ft. × 60 ft. with the actual building
being a modular building that could be built off site and set on a foundation
constructed on-site. This limited size of the office space was mostly budget driven at
this point, with the majority of the estimated project costs going to the pavement
test areas and the large membrane building.
The designers specified a pre-engineered metal building 24 ft.
6 in. × 60 ft. 10 in. The building included an open floor plan area as the control
room, two enclosed offices, a unisex restroom, electrical room, and data/telecom
46 M. Flynn et al.

Fig. 7 Control building floor plan

room. Included is a combination storage and workspace approximately 11 ft.

6 in. × 23 ft. with a doorway to the covered test area and an overhead coiling door
to the outdoor test area (Fig. 7). The metal building was positioned centered on the
west side and mated to the membrane building. Inside the control room three large
windows provide a generous view of the inside test area.

8 Data Acquisition Infrastructure

Developing the data collection and pavement sensor infrastructure to survive the
process of pavement reconstruction was a joint effort between the FAA and the
design team. The pavements at the NAPTF which are tested to failure deep within
the pavement structure, sometimes several feet deep to the subgrade level prevent
the installation of any permanent sensors or infrastructure directly beneath the
pavement. The sensors and wires are usually destroyed by the end of the pavement
testing or destroyed during the removal of the failed pavements. The intended
design of the pavements structures at NAPMRC is for the failure to be contained in
the pavement surface layer only. Therefore some permanent installation of the data
collection infrastructure could remain in place.
To accomplish this, the concept was put forth that sensors would be fed to a
central point within the pavement test area and connected to permanent data
infrastructure. A series of concrete vaults were installed underground, each with
access from the surface through a heavy duty manhole (Fig. 8). The heavy duty
concrete vaults are AASHTO H-20 load rated capable of withstanding very heavy
vehicles with airport grade cast iron manhole lids and risers. The vault dimensions
are 38 in. long × 62 in. wide × 42 in. high.
Each vault supports the adjacent test pavement strips on either side. From each
instrumented pavement test zone (three zones per test lane) sensors are fed into the
sides of the vaults. The vault walls facing the adjacent pavement test lanes were
Design and Engineering Challenges in the Development … 47

Fig. 8 Location of vaults and data cabinets—vaults are shown as data connection pull boxes. The
data cabinets are shown as data collection box

Fig. 9 Concrete vault manhole showing side access

cored to fit three—3 in. diameter conduit stubs approximately 2 in. long protruding
from each side at 20 in. below the finished pavement surface. A heavy duty cor-
rugated plastic pipe is cut in half and positioned over the stubs (Fig. 9) to prevent
the aggregate layers from filling them in. This allows the access to the vault con-
tinuously as the pavement layers are added and sensors installed in each layer. This
concept was borrowed from the NAPTF where the conduits run through the side
foundations for the sensor data acquisition system.
48 M. Flynn et al.

Fig. 10 Data cabinet

A total of six—3 in. diameter PVC conduits encased in concrete run from each
vault to the nearest data collection cabinet. With two vaults assigned to each
cabinet, a total of 12–3 in. conduits run to each of the three data cabinets (Fig. 10).
Trunk lines run to the data cabinets entering through the conduits at the base of
the cabinets. In Fig. 10 the Mini-Power Center is a step down transformer to take
the 480 V required to power the HVS-A down to 120 V for the three GFI electric
outlets inside the data cabinet for use by the data acquisition components.
Ultimately the transformer and the GFI outlets were removed from within the
enclosure by change order to the construction contract. The purpose was to remove
sources of electro—magnetic interference from the data network. The only
remaining source of alternating current in the cabinets is the cooling fan.
Fortuitously this allowed more room inside the data cabinet as the final sensor
terminal arrangement took more space than planned.
Design and Engineering Challenges in the Development … 49

9 Data Acquisition Systems

The FAA took full responsibility for the design and installation of the data
acquisition systems. This included all components for communications, server
computers, sensor network arrays, data acquisition (DAQ), power sources, camera
systems, video/graphical presentation system, data collection software, and a robust
data storage system.
Six test lanes were constructed at NAPMRC with subterranean conduits and
manholes to support the sensor arrays. The sensor arrays consist of the sensors
themselves and all low voltage wiring and cabling that connect the DAQ to each
embedded pavement, aggregate or subgrade sensor. During demolition and
reconstruction it is inevitable for sensors embedded in the pavement, base or
sub-base layers to be damaged or removed altogether. For these expendable sen-
sors, a “trunk cable” solution was implemented. The trunk cables are permanently
installed from the vaults to the data cabinets. Two 12 pair trunk cables are com-
bined and attached to circular 55 pin connector to provide a permanent connection
point to the DAQ avoiding the time and cost of installing new cables all the way to
the DAQ cabinets each time the pavement undergoes reconstruction. In general,
approximately eight trunk cables were required per test lane: four cables for full
bridge sensors; one cable for pressure cells; one cable for quarter bridge sensors;
one cable for thermocouples; and one cable for moisture sensors.
The data trunk lines terminate in the data cabinet at another 55 pin connector.
From the back of the connectors the sensor wires are split up, individualized and
distributed to terminal strips. Phoenix Contact terminal strips were chosen as each
contact has an individual knife switch cutout enabling technicians to easily dis-
connect a sensor for troubleshooting or to remove if it is shorting out. From the
terminal strips the sensor signal is distributed to the DAQ. National Instruments
data acquisition components were chosen based on these features:
• Designed for applications in rugged environments.
• Can measure up to 256 channels of electrical, physical, mechanical, or acoustic
signals.
• Can choose from more than 50 hot-swappable I/O modules with integrated
signal conditioning.
• Ability to run up to seven hardware-timed analog I/O, digital I/O, or
counter/timer operations simultaneously.
• The NI systems are programmed using LabVIEW software.
Each cabinet has its own DAQ system consisting of the following components
(Fig. 11):
• National Instruments compact DAQ (cDAQ)—The cDAQ’s are rugged DAQ
devices that allow for dynamic sensor response measurement at different scan
rates according to requirements. Dynamic measurement is any measurement that
requires greater than 1 sample per second scan rates. Sensors requiring dynamic
measurements include strain gages and pressure cells. The NI cDAQ devices are
50 M. Flynn et al.

Fig. 11 Data cabinet interior

computer dependent in that they cannot store data or be custom programmed.

However, they can be externally triggered if properly configured. They are
capable of multi-sensor measurements via signal-specific modules. Each of
these modules can provide signal conditioning such as filtering, voltage auto
ranging, dynamic response, serial communications, etc. Data is acquired and
transmitted via TCP/IP using Cat6 Ethernet to a Network Attached Storage
(NAS) device. The cDAQ’s require a computer running LabVIEW software to
operate. The computer is housed in the control building and connected via the
network. Each test lane has two dedicated cDAQ’s; one for each test zone within
a lane. A total of four cDAQ’s are in each data cabinet. The cDAQ’s handle all
the dynamic data from the asphalt strain gages (ASG’s), pressure cells (PC’s)
and other sensors.
• National Instruments compact Reconfigurable Input/Output (cRIO)—The
cRIO’s are rugged DAQ devices that allow for static sensor response mea-
surement. Sensors requiring static measurement include thermocouples, ther-
mistors, moisture sensors, strain gages, and crack detection sensors. These
cRIO’s offer a real-time platform that can be custom programmed and opens the
way to performing tasks additional computer systems. The cRIO device is a
computer with the same capabilities as the cDAQ and more. In addition to its
real time measurement capabilities, it has data storage, communication busses,
and can implement a field-programmable gate array (FPGA) backplane
The FPGA can be utilized to provide customization for features that may require
parallel processing for speed and power. Because this device is a computer, it
Design and Engineering Challenges in the Development … 51

can also host web servers, FTP servers, etc. Currently, the FPGA will be used to
manage the data cabinet power usage turning off nonessential devises during a
power failure to conserve backup battery power.
• Custom Uninterruptable Power Supply (UPS) from TSi Power—This custom
built UPS was designed specifically for the NAPMRC data systems. The UPS
provides direct DC power to the entire DAQ system, network switches, and
sensor excitation voltages, as well as acting as a UPS. The UPS is located
outside the data cabinet so no conversion from AC to DC occurs within the
cabinets, reducing the possibility of electronic noise.
Output from the cDAQ’s and cRIO is converted from Ethernet to fiber optic signals
before being sent to the data server in the control building. Once in the control
room, the fiber signal is converted back to Ethernet via rack mounted patch panels
for distribution to the server and PC’s for analysis. A Microsoft Windows server
(Main Server) is the central point for all network connections. Housed in the
NAPMRC control room, the server becomes a bridge between the existing network
types.
Three networks centrally connected at the Main Server: the Data Net, the
HVS-A Net and the Video Net. During testing wheel speed and position data are
collected from the HVS-A Net and these parameters are used to trigger the cor-
responding DAQ on the Data Net to start or stop data collection according to stored
trigger parameters. This program will display the data in a Graph or Chart in real
time on the video net as it is read by the DAQ. Data files are received by the Main
Server are stored on the rack mounted Network Attached Storage (NAS).

Fig. 12 Aerial view of NAPMRC

52 M. Flynn et al.

The main server is also linked to FAA’s IT network so the information can be
accessed by others at the FAA Technical Center, primarily the engineers at the
NAPTF just across the street.

10 Summary

The FAA’s NAPMRC Facility took nearly 4 years from inception to completion of
construction. The total cost of construction was approximately $4.4 Million.
The FAA benefited greatly from experience gained from operating the NAPTF for
12 years prior to the development of NAPMRC. The Facility will immediately be
used to advance the testing of asphalt technologies for airfield pavements such as
warm mix asphalt, polymer modified binders, stone matrix asphalt, RAP mixes and
full depth rehabilitation (Fig. 12).
Heavy Vehicle Simulator and Accelerated
Pavement Testing Facility at Rowan
University

Ayman W. Ali and Yusuf Mehta

Abstract The use of full-scale accelerated pavement testing is gaining more

prominence in recent years due to (1) the several advantages such tests offer and
(2) the development of portable and non-portable accelerated pavement testing
equipment that facilitate conducting these tests. For instance, conducting acceler-
ated pavement testing offers the advantage to simulate long-term traffic conditions
in a short period of time that ranges between three to six months. It can also help in
better simulating field conditions simply because full-scale sections are constructed
using paving equipment and procedures utilized by contractors in the field.
Therefore, it can be argued that full-scale accelerated pavement testing can provide
results that are more representatives of field conditions that does laboratory testing.
To keep up with current transportation and pavement research trends, Rowan
University is currently in the process of constructing an accelerated pavement
testing facility. In fact, the mission of the Civil and Environmental Engineering
Department at Rowan University is to grow the program to become one of the
nationally recognized research programs in the areas of transportation and pave-
ment engineering. To accomplish this mission, the department along with the
College of Engineering at Rowan University, the State of New Jersey (NJ), and the
United State Department of Defense (USDoD) have established a state-of-the-art
transportation research center known as the Center for Research and Education in
Advanced Transportation Engineering System (CREATEs). CREATEs houses an
AMRL (AASHTO Materials Reference Laboratory) certified Rowan University
Construction Materials Laboratory (RUCOM) and the Rowan University
Accelerated Pavement Testing Facility (RUAPTF). Therefore, CREATES is envi-
sioned to provide states and local agencies as well as the pavement industry in the
northeastern region of the United States (US) with both laboratory and full scale
accelerated pavement testing capabilities. This paper documents the design and
construction processes currently being implemented for building RUAPTF.

A.W. Ali
Y. Mehta (&)

Department of Civil and Environmental Engineering, Center for Research and Education
in Advanced Transportation Engineering Systems (CREATEs), Rowan University,
Glassboro, NJ 08028, USA
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 53

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_4
54 A.W. Ali and Y. Mehta

1 Introduction

The use of full-scale accelerated pavement testing is gaining more prominence in

recent years due to (1) the several advantages such tests offer and (2) the devel-
opment of portable and non-portable accelerated pavement testing equipment that
facilitate conducting these tests. For instance, conducting accelerated pavement
testing offers the advantage to simulate long-term traffic conditions in a short period
of time that ranges between three (3) to six (6) months. It can also help in better
simulating field conditions simply because full-scale sections are constructed using
paving equipment and procedures utilized by contractors in the field. Therefore, it
can be argued that full-scale accelerated pavement testing can provide results that
are more representatives of field conditions that does laboratory testing.
To keep up with current transportation and pavement research trends, Rowan
University is currently in the process of constructing an accelerated pavement
testing facility. In fact, the mission of the Civil and Environmental Engineering
Department at Rowan University is to grow the program to become one of the
nationally recognized research programs in the areas of transportation and pave-
ment engineering. To accomplish this mission, the department along with the
College of Engineering at Rowan University, the State of New Jersey (NJ), and the
United State Department of Defense (USDoD) have established a state-of-the-art
transportation research center known as the Center for Research and Education in
Advanced Transportation Engineering System (CREATEs). CREATEs houses an
AMRL (AASHTO Materials Reference Laboratory) certified Rowan University
Construction Materials Laboratory (RUCOM) and the Rowan University
Accelerated Pavement Testing Facility (RUAPTF). Therefore, CREATES is envi-
sioned to provide states and local agencies as well as the pavement industry in the
northeastern region of the United States (US) with both laboratory and full scale
accelerated pavement testing capabilities.
This paper documents the design and construction processes currently being
implemented for building RUAPTF. This paper will also present currently active
projects that will be completed at CREATEs. It is believed that the experiences
learned through the construction of RUAPTF will be of great importance to other
institutions interested in accelerated pavement testing.

2 Objectives

The overall goal of this paper is to present to the transportation/pavement engi-

neering and accelerated pavement testing communities the design and construction
processes implemented by Rowan University for constructing RUAPTF. To
accomplish this goal, the following set of specific objectives were identified:
Heavy Vehicle Simulator and Accelerated Pavement Testing … 55

– Discuss the specification of the Rowan University Heavy Vehicle Simulator

(HVS).
– Discuss the design considerations and layout of the RUAPTF.
– Discuss the potential for CREATEs to be a resources to state and local agencies.
– Discuss currently active projects using the resources available at RUAPTF.

3 Specifications of Rowan University Heavy Vehicle

Simulator

Rowan University, through a Cooperative Agreement with the United States Army
Corps of Engineers (USACE) and the United States Department of Defense
(USDoD), has recently acquired a Dynatest Heavy Vehicle Simulator (HVS).
This HVS (Fig. 1) offers the ability to test full-scale pavement sections and subject
them to the equivalent of twenty (20) years of traffic loading in as few as three to six
(3–6) months.
Generally speaking, the specifications of Rowan University’s HVS are those of
the Dynatest Mark IV model. This model can be described as an automatic, diesel
operated pavement loading device that can apply at least 26,000 loading passes in
bidirectional configurations and 13,000 loading passes in unidirectional configu-
rations. All of those loading passes are applied within a period of one (1) day or
twenty-four (24) hours. Rowan University’s HVS also has the ability to simulate
truck tires and single aircraft tires simply because the device allows for these tire
configurations to be mounted on the loading carriage.

Fig. 1 A picture of Rowan University’s Heavy Vehicle Simulator (HVS)

56 A.W. Ali and Y. Mehta

In terms of loading magnitude and speed, Rowan’s HVS can apply wheel loads
ranging between 30–100 kN (7–22.5 kips) with the ability to apply up to 200 kN
(45 kips) for short-duration during special testing needs. The speed at which the
loading is applied (i.e., wheel speed) can also be varied and controlled using
Rowan’s HVS. For instance, in the case of specialty testing (wheel loading higher
than 100 kN) the wheel speed is typically maintained at 12 ± 3 km/h (7.5 ± 2
mph) while in the case of regular testing (i.e., wheel loading magnitude lower than
100 kN) the speed of the HVS is typically maintained to a maximum speed of
6.5 ± 2 km/h (4 ± 1.25 mph) with the ability to use lower speeds. Rowan’s HVS
also has the ability to simulate wander in traffic by applying passes over the width
of a pavement section. Typically, the wander distance from a section’s centerline is
approximately 0.75 m (20.5 inches).
The linear distance of wheel loading that can be covered by Rowan’s HVS is
shown in Fig. 2. As can be seen from this figure, the length of this loading area can
be at minimum 6 m (19.5 ft.) over which an approximately constant wheel load is
applied using an approximately constant speed. The schematic in Fig. 2 also shows
that about 1.5 linear meters (5 ft.) are required for the wheel loading to accelerate
and about the same linear distance is required to decelerate until reaching a full
stop. Therefore, a typical sections that can be tested using Rowan’s HVS is
approximately 9 m long.
Rowan’s HVS can be operated at wide range of temperatures; thus, allowing for
continuous operations even in adverse weather conditions. The temperatures at
which Rowan’s HVS can be used to conduct pavement testing range between
−5 °C to +40 °C (5–105 °F). The equipment is also equipped with a heater that is
typically used to heat the surface of pavement sections. The heater is powered using
electric power. In addition, Rowan’s HVS is controlled using a computer that
allows for manual and/or automated operations. In a typical test, Rowan’s HVS is
operated using the controlling computer to automate stages from the start of
operations to shutting down of the HVS. The controlling computer also monitors
the applied wheel load every 100 ms using pressure transducers and records the
average wheel load after completion of every pass. The average tire pressure and

Fig. 2 Schematic of the

dimensions of a typical
pavement section for testing
using Rowan’s Heavy
Vehicle Simulator (HVS)
Heavy Vehicle Simulator and Accelerated Pavement Testing … 57

tire temperature also recorded after every pass by the controlling computer. The
computer also records two time stamps that define the start and end of a particular
pass. This information is typically essential to synchronizing the applied loading
passes to the measurements obtained from sensors installed within the pavement
section.
In summary, Rowan’s HVS has a variety of specifications that allow it to be a
state-of-the-art tool for testing, in an accelerated fashion, full-scale pavement sec-
tions. It can be utilized to evaluate pavement responses due to truck/aircraft wheel
configuration and magnitude, due to a variety of loading speeds, and due to various
wander patterns. All of these evaluations can be completed and automated using the
controlling computer.

4 Rowan University Accelerated Pavement Testing

Facility (RUAPTF)

In order to successfully use the HVS, Rowan University has started to design and
build an accelerated full-scale pavement testing facility. This facility is named as
Rowan University Accelerated Pavement Testing Facility (RUAPTF) and will be
housed in the South Jersey Technology Park. RUAPTF will include space for up to
twenty (20) full-scale pavement sections. The facility will have a fabric building in
which two (2) full-scale pavement sections can be constructed. This fabric building
will also serve as the storage facility in which the HVS will be stored when not
operational. The following subsections provide a discussion of the facility’s design
considerations, layout, and instrumentation capabilities.

4.1 Design Considerations

Several design considerations were taken into account during the design stage of
RUAPTF. These considerations summarized as follows:
– Ease of HVS Maneuverability.
– Infrastructure for Instrumentation.
– Requirements of Data Acquisition System (DAQ).
– Future expansion of RUAPTF.
The HVS is a heavy equipment that is difficult to maneuver and move from one
location to another. While Rowan’s HVS has the capability to propel itself from one
section to another, it typically requires approximately 4 h to move about 30 m
(100 ft.). Therefore, Rowan’s HVS requires a significant amount of time to be
moved. In order to alleviate some of the maneuverability issues associated with
Rowan’s HVS, RUAPTF was designed to be of linear layout. In other words, the
58 A.W. Ali and Y. Mehta

sections in RUAPTF were laid out in such a way that the HVS requires minimal
maneuvering.
Another consideration that was taken into account during the design of RUAPTF
was the required infrastructure needed to ensure successful instrumentation of the
various pavement sections at the facility. The objective here was to build the
infrastructure in a fashion that allows for continuous connectivity between the
fabric building and the sensors placed in all full-scale pavement sections. For this
purpose, RUAPTF was designed to have an underground linear conduit. In this
conduit, the power needed to operate a portable DAQ is provided along with
internet connectivity to ensure continuous monitoring of the data collected by the
system. A series of manholes connecting the underground linear conduit system to
ground surface was also designed and put in place. This manhole system allows for
placing the portable DAQ system in close proximity to the various full scale sec-
tions. The manholes were placed strategically within RUAPTF in order to ensure
the distance between the instruments installed in the pavement sections and the
portable DAQ is less than 100 ft. long. This was essential in order to optimize the
cost of the DAQ without compromising the data quality. The main reason for such a
design was attributed to the fact that using an instrument cable that is longer than
100 ft. long might result in significant amounts of noise in the collected data. It is
worth noting that there exists DAQ systems that compensate for instruments’ cable
lengths; however, these systems are very expensive and were not considered due to
budget constraints.
In addition to the instrumentation infrastructure needs, RUAPTF design team
took into consideration the requirements of the DAQ system. The team decided to
use a portable DAQ system for reasons discussed above. The DAQ system was also
designed to have the capability to collect approximately 5000 data points per
second from all dynamic sensors installed in any particular pavement section being
tested. Such a high rate of data collection is required because the loading wheel
travels over the footprint of each sensor in a matter of milliseconds. Internet con-
nectivity to the DAQ system was also of paramount importance. Such connectivity
allows for continuous monitoring of the data being collected by the DAQ in
real-time and for ensuring that all testing is being conducted without problems. If
any problems in the data collection process arise, real-time monitoring of collected
data facilitates making an educated decision on whether testing should be stopped
or continued.
Finally, the research team designed RUAPTF to account for future expansion
needs. This was accomplished through allocating space for future sections that are
anticipated to be constructed in the future. During the current construction phase,
only eight (8) of the potential twenty-two (22) sections were constructed. However,
instrumentation infrastructure for all sections was put in place and the locations of
was identified and preserved. To be a little more elaborate, all sections were
instrumented with temperature sensors, pressure cells, deflection sensors, and
moisture sensors in both compacted subgrade and base layers. Additional infor-
mation can be found in the following sections.
Heavy Vehicle Simulator and Accelerated Pavement Testing … 59

4.2 Layout and Instrumentation Capabilities of RUAPTF

The layout and instrumentation plans for RUAPTF are illustrated in Fig. 3. As can
be seen from Fig. 3a, RUAPTF will include eight (8) full scale pavement sections
that will be constructed during the current construction phase. Space was also
allocated for a total of twelve (12) additional future test sections to be constructed
and instrumented. Land space is also available to construct additional test sections if
the need arises.
Figure 3c shows the different sensors that will be used in any particular pave-
ment section constructed at RUATPF. Generally, all RUAPTF sections will be
instrumented with two types of sensors; dynamic and static. The dynamic sensors,

Fig. 3 Layout and Instrumentation of Rowan University Accelerated Pavement Testing Facility
(RUAPTF)
60 A.W. Ali and Y. Mehta

which record data as the HVS’s wheel applies loading on top of the pavement
section, include pressure cells, longitudinal and transverse strain gages, and vertical
deflection sensors. These dynamic sensors are typically sampled using high sam-
pling rate in order to ensure fully characterizing section’s response due to a par-
ticular loading pass. The static sensors, which do not necessarily record data when
testing is in progress, include temperature sensors and moisture sensors. Those
instruments are typically sampled using a low sampling rate due to the fact they are
independent of loading patterns.

5 CREATEs as a Research Resource to the North Eastern

Region of the United States

Rowan University has established the Center for Research and Education in
Advanced Transportation Engineering Systems (CREATEs) in order to merge in
the currently available Rowan University AMRL (AASHTO Materials Reference
Laboratory) certified Construction and Materials Laboratory (RUCOM) and
RUAPTF under one leadership. The center was established through collaboration
between Rowan University and the State of New Jersey (NJ) and is envisioned to be
a nationally recognized leading center in transportation engineering education and
research.
CREATEs will provide great opportunities to State Departments of
Transportation (State DOTs) in the northeastern region of the United States.
CREATEs offers these agencies state-of-the-art resources to conduct research and
answer state specific pavement materials problems. CREATEs business model is
offer these State DOTs the opportunity to “own” certain sections and use them to
conducted accelerated pavement testing research. Those sections can be tested
using Rowan’s HVS and the data can be analyzed by the experts at Rowan
University. A report can also be prepared to document the results and
recommendations.
CREATEs also offers the pavement industry in the region the opportunity to
conduct materials research using RUAPTF and RUCOM. The industry can use
CREATEs resources to test and validate, using a third unbiased party, their
new/innovative products. The results of these evaluations can be used to enhance
the marketing pitch of these products for state and local funding municipalities. All
of these opportunities will be available to the pavement industry in the northeastern
region of the United States through CREATEs Industry Consortium (CIC). The
members of the consortium will be charged a fee for conducting the research. Other
non-members will also be allowed to conduct testing at CREATEs; however, they
will lose the discounts offered by CIC.
Heavy Vehicle Simulator and Accelerated Pavement Testing … 61

6 Research Projects at CREATEs

Several research projects, funded by the New Jersey Department of Transportation

(NJDOT); Rhode Island Department of Transportation (RIDOT); New York State
Department of Transportation (NYSDOT), Wisconsin Department of
Transportation (WisDOT); Federal Aviation Administration (FAA); and Federal
Highway Administration, were completed or are currently in progress at CREATEs.
The focus of this section will be on presenting a current research study funded by
NJDOT and utilizes resources available at RUAPTF. The following subsections;
therefore, provide a summary of a study titled Heavy Vehicle Simulator Evaluation
of Thin Asphalt Overlays over Composite Pavements.

6.1 HVS Evaluation of Thin Asphalt Overlays

Over Composite Pavements

The primary goal of the study is to identify and predict the expected life of thin
asphalt overlay treatments used for rehabilitating and preserving Portland Cement
Concrete (PCC) pavements in the NJ. To accomplish this goal, a two-phase work
plan (Fig. 4) was prepared and currently being executed by researchers at
CREATEs. In Phase I (Literature Review and State of Practice), which includes
Tasks 1 through 3, the research team will document, synthesize, prioritize and
conduct gap analyses on national and international state of practice pertaining to
thin asphalt overlays. This includes determining the major factors affecting the
performance and service life for a variety of thin asphalt overlay mixes and treat-
ments. In addition, completing these tasks will facilitate the identification of the
current PCC pavement conditions in NJ. This information will be valuable when
constructing PCC pavement that reflect actual conditions in NJ.
Phase II (Evaluation of Thin Asphalt Overlays through Full-Scale Accelerated
Testing), which consists of Tasks 4 through 8, includes a systematic effort to
establish a methodology for examining the performance of thin asphalt overlays and
their expected service life. This will facilitate identifying the major performance
concerns that significantly influence the service life of up to four thin asphalt
overlay treatments and recommend methods to predict their expected service life.
The final report produced at the conclusion of Phase II will summarize findings,
draw conclusions, document results, and present the guidelines, practices, and
suggested changes in thin asphalt overlay mix specifications for NJDOT in the
appropriate NJDOT format.
Six full-scale pavement sections will be constructed at RUATPF as a part of this
study. Figure 5 illustrates the layout of these sections along with the thin asphalt
overlay mixes that will be evaluated. As can be seen from this figure, a total of three
thin asphalt overlay mixes will be evaluated. These mixes include the typical
Superpave 9.5 mm Nominal Maximum Aggregate Size (NMAS), a 12.5 mm
62 A.W. Ali and Y. Mehta

Fig. 4 Work plan prepared for completing HVS thin asphalt overlay evaluation study

Fig. 5 Layout of thin asphalt overlay mixes in RUAPTF

NMAS Stone Matrix Asphalt (SMA) mix, and a NJ specialty High Performance
Thin Overlay (HPTO) mix. The various mixes will be laid in RUAPTF with and
without using a Binder Rich Intermediate Course (BRIC) to also evaluate the
benefits of using BRIC courses to alleviate reflective cracking. It is noted that all
Heavy Vehicle Simulator and Accelerated Pavement Testing … 63

thin asphalt overlays and BRIC mixes will be placed on top of an 8-inch thick PCC
layer.
Rowan’s HVS will be utilized to apply loading on the constructed pavement
sections. Dual-tires representing one-half of a standard single axle truck will be
used to apply the loading. These tires will be pressured to a 100-psi tire pressure.
Loading will be applied using unidirectional configurations while wandering across
a width of about 3 ft. at a constant speed of 7.5 mph. During and after the appli-
cation of loading surface deformations will be monitored. In situ moduli values will
also be collected periodically through utilizing Falling Weight Deflectometer
(FWD) testing. In addition, forensic evaluations will be conducted to establish the
condition of the six pavement sections after applying a specific number of loading
passes. It is anticipated that after completion of the study the service life of each of
the three thin asphalt overlay mixes will be determined.

7 Summary and Closing Remarks

This paper presented a discussion of the design and construction processes

implemented by Rowan University during the construction of RUAPTF. The paper
also included a discussion of design considerations and the instrumentation capa-
bilities of RUAPTF and the reasons behind selecting specific design elements for
RUAPTF. In addition, this paper presented a discussion of how CREATEs can be
of value to State DOTs and the pavement industry in the northeastern region of the
US along with a brief documentation of a study funded by NJDOT and currently
being executed at CREATEs and RUAPTF.
Based on the discussion presented in this paper, the following closing remarks
summarize the gist of this paper:
– Rowan University has acquired a state-of-the-art Dynatest Mark IV HVS
through a collaborative agreement with the United States Army Corps of
Engineers (USACE).
– Rowan HVS has several capabilities that include: (a) applying approximately
20 years of traffic loading in as few as three to six (3–6) months; (b) simulating
both truck and aircraft loading; (c) heating pavement sections to facilitate testing
various temperatures; and (d) automatically controlling testing from start to
finish.
– RUAPTF is currently being constructed by Rowan University to house Rowan’s
HVS. This facility was designed while taking into consideration several factors
that include: (a) ease of HVS maneuverability; (b) infrastructure needed for
simple pavement instrumentation; (c) Data Acquisition System
(DAQ) portability; and (d) future expansion needs.
– Rowan University has established CREATEs with partnership with NJDOT to
be the go to transportation and materials engineering research center in the
64 A.W. Ali and Y. Mehta

northeastern region of the US. CREATEs is envisioned to become one of the

nationally leading research centers in the country.
– CREATEs offers State DOTs and pavement industry the opportunity to utilize
resources available at RUAPTF and RUCOM.
– Currently, a research study funded by NJDOT is currently being executed at
CREATEs. The outcomes of this study will facilitate determining the service life
of three different thin asphalt overlay mixes.
 Part III
General Accelerated Pavement Testing
A Link of Full-Scale Accelerated Pavement
Testing to Long-Term Pavement
Performance Study in the Western Cape
Province of South Africa

J.K. Anochie-Boateng, W. JvdM Steyn, C. Fisher and L. Truter

Abstract The development of a new mechanistic-empirical pavement design

method by the South African National Roads Agency Ltd (SANRAL) is at its final-
ization stage. The new design method, referred to as the South African pavement
design method (SAPDM) will heavily rely on Long-Term Pavement Performance
(LTPP) and Accelerated Pavement Testing (APT) data to calibrate performance
models for cracking and rutting. This paper discusses more than 10 years APT and
LTPP data established from LTPP sections in the Western Cape province of South
Africa. In addition, the paper presents updated field results of an expanded LTPP
program that links both field and laboratory data to pavement performance. In com-
parison, the rut data from the LTPP section are similar to or higher than the HVS rut
data depending on the type of pavement (i.e. surfacing material, layer materials, layer
thickness and so forth). Similarly, deflection data could be similar or varied for
different pavement types. Based on the conclusions from this paper, it is envisaged
that the LTPP and APT data will provide the basis for the calibration of cracking and
rutting models for the SAPDM. This would require in some cases, an introduction of
shift factors to adjust APT data to LTPP (real life) data.

Keywords Accelerated pavement testing Long-term pavement performance



Heavy vehicle simulator rutting Cracking

J.K. Anochie-Boateng (&)
C. Fisher

CSIR Built Environment, P O Box 395, Pretoria 0001, South Africa
e-mail: [email protected]
C. Fisher
e-mail: cfi[email protected]
J.K. Anochie-Boateng
Tshwane University of Technology, Private Bag X680, Pretoria 0001, South Africa
W.J. Steyn
University of Pretoria, Lynnwood Road, Hatfield, Pretoria 0002, South Africa
e-mail: [email protected]
L. Truter
Department of Transport and Public Works, Western Cape, Cape Town, South Africa
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 67

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_5
68 J.K. Anochie-Boateng et al.

1 Introduction

The Department of Transport and Public Works (DTPW) in the Western Cape
province is funding a Long-Term Pavement Performance (LTPP) program in South
Africa. The DTPW LTPP study is the most comprehensive LTPP program in South
Africa, and has successfully been linked with APT and laboratory studies.
Accelerated Pavement Testing (APT) provides an accelerated view of major
structural behavior to be expected from tested pavements, and while various
environmental conditions can typically be simulated during APT, it is important to
link the results obtained from the APT to real world pavement behavior to enable
outputs from APT to be calibrated for use in pavement design and analysis. On the
other hand, LTPP focuses on evaluation of the in-service performance of pave-
ments, incorporating normal traffic loads and typical environmental conditions
affecting the pavement in a normal way. The need to link APT with LTPP is
therefore, essential. Metcalf (1996) and Hugo and Martin (2004) emphasized that a
comprehensive APT program would enable the transposition of APT data to
real-world applications. Both programs generate relevant data for pavement design
and analyses and complement each other.
Between 2005 and 2011, two LTPP sites were established and monitored in the
Western Cape province of South Africa. These sites were located on national
highway N7 (high traffic volume road), and a regional road MR538 (low traffic
volume road). These LTPP sites were located near APT sites with the main
objective of linking the data from both studies. Detailed findings and recommen-
dations of the combined study of the LTPP and APT are presented by Steyn et al.
(2012). The Western Cape LTPP program aims at collecting extensive performance
data from the field and laboratory for future link to APT data.
This paper compares data established from an APT and two LTPP (N7 and
MR538) sites in the Western Cape province, and discusses current data of the LTPP
program. The paper provides the background on the development of the APT and
LTPP programs, as well as the methodologies used for an expanded LTPP moni-
toring and data collection that includes extensive field and laboratory testing. Only
updated field data for N7 section are presented as monitoring of the MR538 was
discontinued in 2011.

2 Background of the Western Cape Province APT

and LTPP Program

The protocol for the establishment and monitoring of LTPP sections in conjunction
with Heavy Vehicle Simulator (HVS) experiments was completed on behalf of the
Gauteng Department of Public Transport, Roads and Works (GDPTRW) in 2002. In
2003, a study was commissioned to layout LTPP sections at three HVS experiments
in Gauteng on two provincial roads, carry out assessments and establishes databases
A Link of Full-Scale Accelerated Pavement Testing to Long-Term … 69

for each. Although committed funding had not been guaranteed for the South
African HVS/APT program, the need for a protocol to standardize the methodology
used for establishing and monitoring LTPP sections in conjunction with APT sec-
tions was identified, with a view to initiating such a study. The protocol was finalized
in 2004 (Jones and Paige-Green 2003) and a paper was presented on it at the 2nd
International Conference on Accelerated Pavement Testing (Jones et al. 2004). The
protocol covers management responsibilities, section location and establishment,
instrumentation, data collection and reporting criteria.
In 2004, funding to conduct LTPP studies adjacent to two recently completed
APT studies was provided by the Gauteng Department of Transport and Public
Works to “test” the protocol. Annual monitoring of these two sections continues
and interim reports have been published (Steyn et al. 2007). In 2005, additional
funding was provided by the Department of Transport and Public Works (DTPW)
of the Provincial Administration of the Western Cape for an additional two sites to
assess the performance of rehabilitated pavement sections of N7 and MR538 after
HVS tests were completed on a number of test sections on the two roads. The
selection of the LTPP sites was based on similar pavement types and environmental
areas, as well as areas where detailed traffic count and traffic characteristics
affecting performance of the pavement could be obtained. The studies on these two
roads form part of the discussion in this paper.
Both long term structural and functional performance of the two sections were
monitored, with the primary goal of providing appropriate, adequate data, infor-
mation, and products (e.g., rutting and cracking models) to better understand
pavement performance.

3 Methodology and Current LTPP Program

In 2011, the DTPW APT program and monitoring of the MR538 LTPP site were
discontinued due to lack of funding. In the same year, the combined program of
APT/LTPP was restructured in order to only focus on LTPP study. A total of six
sites including the N7 site were included in the expanded program. The new sites
include a national highway N1 and regional highways M5, R46, and MR559 (two
sites). The expanded LTPP program includes monitoring and laboratory evaluation
of field samples essentially, to establish a comprehensive database for the experi-
mental sections. The initial plan to evaluate pavement base and subbase layer
materials in the laboratory as part of the program was shelved in order to establish
additional road sections for monitoring. Collection of quantitative data on cracking,
which has not been part of any LTPP programme in South Africa at the time, was
included in the current program. In view of this development, the existing LTPP
protocol (Jones and Paige-Green 2003) was reviewed, and a new LTPP guideline
was developed to allow the engineering properties of asphalt concrete cores from
the LTPP sites to be determined in the laboratory (Anochie-Boateng and O’Connell
2013).
70 J.K. Anochie-Boateng et al.

The current program is the most comprehensive LTPP program in South Africa
with six monitoring sites. The ultimate intention is to use the LTPP data for
pavement design and pavement management systems in the Western Cape pro-
vince, and to develop simple models to represent the behaviour of the asphalt
materials in the LTPP sites over time. The laboratory testing program designed for
asphalt materials was based on the test methods developed for the new
mechanistic-empirical South Africa pavement design method (Anochie-Boateng
et al. 2010). These tests are presented below:
• Binder testing:
– Binder characterization using specification (standard) test procedures;
– Binder characterization using Dynamic Shear Rheometer (DSR) procedures,
and
– Development of interim binder ageing models.
• Aggregate testing:
– Grading;
– Bulk and Apparent relative densities;
– Water absorption, and
– pH, and flakiness index.
• Asphalt concrete testing:
– Dynamic modulus test;
– Indirect tensile test for resilient modulus;
– Indirect tensile strength test, and
– Permanent deformation test.
Monitoring of the six LTPP sites is carried out approximately every six months
to ensure that data are collected at the end of the dry and wet seasons. Data
collected from the sites include visual assessment, rut depth, pavement temperature,
density and moisture content, as well as deflection measurements. The visual
assessment is done based on the standard visual assessment manual for flexible
pavements in South Africa (TMH9 1992). Logistics do not always allow for precise
intervals in these data collection, and gaps may thus exist in the data.
The structural conditions of the LTPP sites are assessed in the field by means of
Falling Weight Deflectometer (FWD) and Dynamic Cone Penetrometer
(DCP) tests. The primary objective of these non-destructive testing is to establish
whether or not failures of the sites are related to the specific properties of the
surfacing and the conditions of the substrata.
After completion of the non-destructive testing (deflection), cored samples of the
asphalt surfacing are extracted from the LTPP sites to determine the physical and
engineering properties including density, stiffness, tensile strength, resistance
against permanent deformation, grading and water permeability. The conditions of
the binder recovered from the cored samples are also assessed by means of various
physio-chemical tests. In the laboratory, standard test methods are followed to
A Link of Full-Scale Accelerated Pavement Testing to Long-Term … 71

collect data from the asphalt cores extracted from the LTPP sites. The laboratory
test results are correlated with asphalt mix design data, as-built data and perfor-
mance data and to identify causes of failure and attributes of performance taking
into consideration all available data on the LTPP sites.
Traffic counts are performed to distinguish between light, medium, heavy, and
very heavy vehicles. The damage caused by the different vehicle types on the
pavement is usually expressed in an equivalent of a standard 80 kN axle load (E80).
As part of the traffic survey, axle configurations are also identified for all heavy
vehicles and the percentage of heavy vehicles are also identified for analysis
purposes.

4 Discussion of the LTPP and APT Results

Detailed analysis of calibrated results of the full scale APT/HVS testing data using
LTPP data were presented by Steyn et al. (2012). The analysis consisted of a
comparison of the surface rut and surface deflection measured for the MR538 and
N7 sections. The average, as well as the average ±1 standard deviation of the rut
and deflection data, were presented and discussed for both the LTPP and APT data,
to demonstrate the variability in pavement properties. The pavement structures for
the two sites are shown in Fig. 1.

4.1 Comparison of Rutting and Deflection Data

The rut comparisons for MR538 and N7 are shown in Fig. 2, whereas deflections
for the two sites are compared Fig. 3. These comparison re based on data obtained
from the HVS test and that from the LTPP sites from 2005 to 2011. Traffic on the
roads since HVS testing was completed amounted to the equivalent of 120,000
E80s (MR538) and 2,440,000 (N7). In terms of the HVS tests conducted on the two

Fig. 1 Nominal pavement structures of the two LTPP sections studied (Steyn et al. 2012)
72 J.K. Anochie-Boateng et al.

Fig. 2 Rut from the HVS and LTPP sites

roads, the current traffic relates to 31.8 per cent of the test duration on the N7
(Theyse 2004a, b), Long and Brink 2004) and 3.8 per cent of the test duration on
road MR538 (Theyse 2004a, b). The traffic volume on MR538 and N7 were 440
(low volume) and 13 115 (high volume), respectively.
Figure 2a, b shows the average rut measurements from the LTPP and HVS
sections. A vertical scale with a maximum of 20 mm has specifically been used as
this relates to the typical failure/terminal level for surface rutting in South Africa.
The data indicate that the rut on the N7 LTPP sections is similar to the HVS rut,
whereas the rut on the MR538 LTPP section is higher than the HVS rut. Thus, the
rut on the low volume road (M538) showed significantly high values for the LTPP
when compared with the APT data. This results is similar to previous work in which
it was indicated that for low volume roads, the environment may have a more
significant role in the performance of the pavement than the actual applied loads
(Steyn and Sadzik 1998). On the other hand, both the rut and the deflection data for
the LTPP and APT sections on N7 (high traffic volume) compare well. The damage
exponent (used to calculate the equivalent traffic for the APT section on the N7
section) was 4.2 for both rut and deflection.
The elastic surface deflection data are shown in Fig. 3. For the N7, the data
indicate that the deflections measured on the LTPP sections are comparable to those
measured on the HVS section. For MR538, both the deflections measured from the
LTPP section are generally lower that the deflection data for the HVS section.
It is interesting to see that the LTPP deflections on the N7 section are much
lower than the HVS section, as the rut relationship was relatively close. Thus, the
LTPP-measured deflections on the low volume road (M538) were similar to the
higher values recorded towards the end of the APT testing. As indicated earlier, this
observation supports the theory that environment often has a bigger influence on the
performance of low volume roads than traffic.
A Link of Full-Scale Accelerated Pavement Testing to Long-Term … 73

Fig. 3 Surface deflection from the HVS and LTPP sites

4.2 Pavement Temperature Data

The temperature of a bituminous layer (as determined by the climate), in con-

junction with the grade and age of the binder, plays a pivotal role in determining the
stiffness characteristics of the asphalt layers in the pavement. The stiffness, in turn,
determines such behaviors as the rutting performance, fatigue performance, and low
temperature cracking performance in the bituminous layers. In the LTPP pro-
gramme, the pavement temperature is monitored using i-buttons installed under the
surfacing (about 20 mm from surface of asphalt layer). A summary of the tem-
peratures measured for the MR538 and N7 sites is provided in Fig. 4. Generally, the
maximum and minimum temperatures for the two sites were comparable (ap-
proximately 55 °C maximum, and 5 °C minimum). As expected, the maximum
temperatures were recorded during the summer months, whereas the minimum
temperatures are recorded during the winter months. The gaps in the MR538 results
are due to some damages in i-buttons as this site (a seal road) has relatively very
thin bituminous layer.

5 LTPP Performance Data for MR538 and N7

5.1 Visual Assessments

Photographs showing aspects of the functional performance of the two sites is

presented in Fig. 5. The rut and FWD data collected from the two sites indicate that
rutting and deflection in the wheel paths is the major concern. Similar trends were
observed on the other LTPP sites that are not included in this paper. The major
deterioration on the MR538 section was found to be aggregate loss and rutting,
whereas the N7 section experienced cracking and rutting during the 10 year period
of assessment. No raveling was noted and the binder condition of the two sites was
rated as fresh. Indistinct discoloration was evident in the wheel tracks of the N7 site,
74 J.K. Anochie-Boateng et al.

Fig. 4 Pavement temperature results for LTPP sites

attributed to minor wear (polishing) of the pre-coating on the surfacing chips. No

major deterioration was recorded over the period of assessment, thus that the skid
resistance and to some extent the surface drainage were all rated as very good. In
addition, all shoulders were rated as good and side drainage remained clear of
vegetation. Some potential seasonal variation in terms of moisture content and its
influence on density and deflection were apparent, although additional monitoring
was necessary before any firm conclusions could be drawn (Steyn et al. 2007).
In terms of structural assessment, increasing areas of cracking were noted. There
were signs of pumping from these cracks. Rut depth generally increased signifi-
cantly in the inner wheel tracks when compared with the outer wheel tracks. No
potholes, patches or edge break were noted. A moisture profile of the outer wheel
track indicated an increase in moisture over time (Anochie-Boateng and Fisher
2010). Seasonal influences were not clear cut during the time of monitoring.
Density measurements remained relatively constant and consistent over the length
of most sites. Densities were marginally higher in the outer wheel tracks compared
with the inner tracks and between wheel tracks, again supporting the rut and
deflection data observations (Anochie-Boateng and Fisher 2010).

5.2 Rutting Measurements

A two-meter long straight edge and a wedge are used to measure rut depths of the
LTPP sites in accordance with procedures set in TMH9 (1992). Rut depth is defined
as the maximum permanent deformation measured under the two-meter straight
edge. The cross sectional profile was measured from the yellow line on the shoulder
of the road towards the centre of the pavement. If there was shoved asphalt, then in
most cases, the cross sectional profile was measured by placing the straight edge on
top of the shoved asphalt so that a measurement of zero (0) mm could be obtained
on the shoulder.
A Link of Full-Scale Accelerated Pavement Testing to Long-Term … 75

Fig. 5 Selected visual conditions of MR538 and N7

Fig. 6 Average Rut depths on N7 Section (2005–2015) in outer and inner wheel paths

Figure 6 shows the total amount of rut measured in approximately 5.5 and
10 years for MR538 and N7, respectively. Generally, rutting in the inner wheel
tracks were found to be higher than those of the outer wheel tracks as depicted in
Fig. 6. The average rut on the MR538 section in the period is about 1.6 mm,
whereas the rut on N7 in the same period is about 4 mm. This means that within
this period of time rutting was not a major concern as the two sites were in sound
condition (i.e. rut depth <10 mm) in accordance with South Africa specifications.
76 J.K. Anochie-Boateng et al.

Fig. 7 Average FWD deflection on N7 (2005–2015) in outer and inner wheel paths

It should be mentioned that the MR538 section was re-sealed after three years,
which could possibly show the decrease in rut depth especially on the outer lane.
The slight decrease in rut depths (<0.5 mm) in the inner lane of MR538 and both
lanes of N7 could be attributed to errors in the manual measurement methods used
in this study (Fig. 7).

5.3 DCP and FWD Measurements

DCP assessments at the two sites were difficult due to the very stony, tightly bound
nature of the base material, and dry, tightly bound and possibly naturally cemented
condition of the subbase sand and subgrade. Penetration was impossible in the upper
150 mm at any location without damaging equipment. Pre-drilling through this
upper layer allowed penetrations in some locations through to 800 mm. At least 500
blows were required to penetrate between the bottom of the drill hole and 800 mm.
No deflection measurements were taken during the initial assessment due to
unavailability of equipment. For the May and November 2006 assessments, FWD
deflection measurements remained consistent, with slightly higher measurements
recorded at the end of the wet season (November) compared with those recorded at
the end of the dry season (May). The data revealed that deflections were slightly
higher in the outer wheel path compared with the inner wheel path for MR538,
whereas deflections were comparable in both wheel paths for the N7 site. Pavement
and ambient temperatures for the two assessments were 22/22 °C (May) and
37/27 °C (November) respectively.

5.4 Cracking Measurements on N7

Cracking was a major problem observed on most LTPP sites in South Africa. There
are usually signs of crack development in the inner wheel and outer wheel tracks,
A Link of Full-Scale Accelerated Pavement Testing to Long-Term … 77

Fig. 8 Crack propagation on N7 during four cycles of measurements

and an increase in surface cracks. However, more signs were observed mainly in the
outer wheel track. Cracking was prevalent on the N7 section. Hence, the expanded
LTPP program recommended the inclusion of crack width and length measure-
ments during the 2014–2017 contract period.
Figure 8 shows the results of crack measurements that included block and
longitudinal cracking. The “other cracks” are those cracks that could not be well
defined. The slow lane of the dual carriage of the N7 section showed more cracking
and pumping when compared with the fast lane. As indicated in Fig. 8, increase in
crack lengths on the N7 section within six months appears to be substantial.

6 Conclusions

Comparisons between HVS and LTPP data were undertaken for two LTPP sections
(i.e. MR538 and N7). The rutting data from the N7 LTPP section are similar to the
HVS rutting data, while the MR538 rutting data are higher that observed from the
HVS. The deflection data for the MR538 LTPP section are similar to the HVS data
while the N7 LTPP data are consistently lower than the HVS data. However, the
period over which the LTPP data have been collected is still relatively short
compared with the period that the HVS tests were conducted and more data will
need to be collected before any clear trends start to emerge from the comparisons.
The following conclusions can be drawn:
78 J.K. Anochie-Boateng et al.

• For the low volume road it was found that the APT rutting data needs to be
shifted by a factor of five in order to obtain similar LTPP (real life) rutting data,
while the surface deflection should be increased by 50 % from the APT data to
get an equivalent LTPP data.
• The time factor used for the M538 elastic deflection data is required to ensure
that the effects of season and time on the low volume pavement be attributed
correctly to the APT data, as the APT data excludes the effect of season and
time, and these factors affect low volume (thin) pavements more significantly
than high volume (thick) pavements.

Acknowledgments The authors would like to appreciate the Department of Transport and Public
Works of the Western Cape province in South Africa for their continued support in funding this
project since 2007.

References

Anochie-Boateng, J., & O’Connell, J. (2013). Guideline for long-term pavement performance
(LTPP) data evaluation in Western Cape. Pretoria: CSIR.
Anochie-Boateng, J., Denneman, E., O’Connell, J., & Ventura, D. (2010). Hot-mix asphalt testing
for the South African pavement design method. In Proceedings of 29th Southern Africa
transportation conference, Pretoria.
Anochie-Boateng, J. K., & Fisher, C. (2010). Technical memorandum: Monitoring of two LTPP
experimental sections in the Western Cape. (Contract Report CSIR/BE/IE/IR/2010/0021/B).
Hugo, F., & Martin, A. E. (2004). Synthesis of highway practice 325: Significant findings from
full- scale accelerated pavement testing. Washington, DC: Transportation Research Board,
National Research Council. 201 pp.
Jones, D., & Paige-Green, P. (2003). A protocol for the establishment and operation of LTPP
sections. Pretoria: CSIR-BE. (Contract Report CR-2003/11).
Jones, D., Paige-Green, P., & Sadzik, E. (2004). The development of a protocol for the
establishment and operation of LTPP sections in conjunction with APT sections. In
Proceedings 2nd International Conference on Accelerated Pavement Testing. Minneapolis,
MN: University of Minnesota.
Jones, D. J., Steyn, W. J.vdM., & Fisher, C. (2007). Technical memorandum: 2006/2007
monitoring of four LTPP experiments in association with HVS tests in Gauteng.
CSIR/BE/IE/IR/2007/0108/B, CSIR BE, Pretoria, South Africa.
Long, F. M., & Brink, A. C. (2004). 2nd level analysis of the HVS data for the southbound
carriageway of the N7 (TR11/1). Pretoria: CSIR-Transportek. (Contract Report CR-2004/12).
Metcalf, J. B. (1996). Synthesis of highway practice 235: Application of full-scale accelerated
pavement testing. Washington, DC: Transportation Research Board, National Research
Council. 117 pp.
Pavement Management Systems. (1992). Standard visual assessment manual for flexible
pavements. Pretoria: Department of Transport. (Technical Methods for Highways - TMH9).
Steyn, W. J. vdM, Anochie-Boateng, J. K., & Fisher, C., Jones, D., & Truter, L. (2012).
Calibration of full-scale accelerated pavement testing data using long-term pavement
performance data. In 4th International Conference on Accelerated Pavement Testing, Davis,
California, USA, September 19–21.
A Link of Full-Scale Accelerated Pavement Testing to Long-Term … 79

Steyn, W. J. vdM, Jones, D., & Fisher, C. (2007). Technical memorandum: Monitoring of two
LTPP experimental sections in the Western Cape. (Contract Report CSIR/BE/IE/IR/2007/
0109/B).
Steyn, W. J. vdM., & Sadzik, E. (1998). Evaluation of superlight pavements under accelerated
traffic. Journal of the Transportation Research Board (Transportation Research Record (TRR))
1639, pp. 130–139. Washington, DC. ISBN 0-309-065119. ISSN 036-1981.
Theyse, H. L. (2004). First level analysis report: HVS testing of the foamed-bitumen-treated
crushed stone base on the slow lane of the southbound carriageway of the N7 near Cape Town.
Pretoria: CSIR-Transportek. (Contract Report CR-2003/23).
Theyse, H. L. (2004). HVS and laboratory testing of a light pavement structure on Main Road 538
between Leipoldtville and Lamberts Bay. Volume 1: Report. Pretoria: CSIR-Transportek.
(Contract Report CR-2004/36).
Accelerated Pavement Testing in Slovakia:
APT Tester 105-03-01

Lubos Remek, Jan Mikolaj, Matej Kyselica and Milan Škarupa

Abstract This article presents the APT facility constructed and operated by
University of Zilina. The machine is called APT tester 105-03-01. The article
describes its technical properties, operational capabilities, sensory equipment
embedded in the pavement and data gathering procedures. The device has several
unique design solutions that make it stand out from similar facilities in the world:
(1) The principle of fixed linear APT facility which loading unit is not positioned in
a fixed frame, but instead moves itself held only by a guiding rail to better simulate
traffic loading. (2) Loading unit consists of fixed and movable frame held by
support connected by joints, better simulating suspension like those found on truck
axles. (3) The construction of the electric motor, gear box and frequency inverter
and its mounting system on the loading unit. Frequency converter controlled
acceleration and deceleration ramps and speed during movement. (4) Hydraulic
stabilization system stabilizing the movable frame, since the load tends to tip the
loading unit in the acceleration and deceleration stage. (5) Autonomous hydraulic
system placed on the outer frame able to lift the loading unit and allows for free
manipulation without burdening the pavement.

1 The History of Accelerated Pavement Testing

in Slovakia

The history of Accelerated Pavement Testing in Slovakia started in 1984 by con-

struction of the Circular Test Track (CTT) by the VUIS-CESTY, Ltd. research
institute. The machine was constructed to perform tests of pavements on a 1:1 scale.
The machine elicited load with an actual vehicle axle weighing 115 kN (the EU

L. Remek (&)
J. Mikolaj
M. Kyselica
M. Škarupa

University of Zilina, Univerzitna 8215/1, 0101 Žilina, Slovakia
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 81

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_6
82 L. Remek et al.

Fig. 1 Circular Test Track 1984–2006

standard for a permissible axle load). The basis of the CTT mechanical part con-
sisted of three loading vehicles each of which has a driving axle. The vehicle was
attached to the arms of a medium anchor clapper. The CTT’s electric part served to
drive each vehicle by means of electro-motors and to control the entire CTT
mechanical facility (Fig. 1).
The facility allowed a transversal shift of the loading vehicles of ±950 mm. The
vehicles moved on two doubled wheels equipped with 11.00R—20 tires at a
0.7 MPa inflation pressure. The speed of the vehicle was limited to 60 km/h, but the
mean operational speed was 30 km/h. The pavement testing track was 100 m long
and 6 m wide. The facility was used to test new pavement types, materials and
other road equipment—rail crossing for instance in a 1:1 scale. The facility was in
operation till 2006 (Fonód 2005).
The most important outputs were pavement degradation curves used in
Pavement Management System of Slovakia from 1996. As an example, in Fig. 2
we present the trend line for degradation of longitudinal unevenness for specific
pavement used in Slovakia’s pavement management system. The functions are
shown in Fig. 3. In this Figure, n is the number of Standard Axle Loads repetitions
from the start of service to the expected date of pavement repair; N is the total
number of Standard Axle Loads repetitions needed to achieve the load limit value.
In this case, the parameter B 2.0 was derived for flexible pavements and parameter
B 3.0 was derived for semi-rigid pavements. A warning value is at 60 % difference
between initial and limit value. Critical value is 40 %.
Accelerated Pavement Testing in Slovakia: APT Tester 105-03-01 83

Fig. 2 Design of Circular Test Track 1984–2006

Fig. 3 Degradation functions of longitudinal unevenness for flexible pavements B = 2.0 and
semi-rigid pavements B = 3.0

2 APT Tester 105-03-01

Strategically focused, needs driven, pavement research programs have been shown
to be most successful when carried out using a combination of laboratory, APT
(accelerated pavement testing) and LTPP (long term pavement performance mon-
itoring) data collection studies in conjunction with standard pavement management
system monitoring procedures.
At present LTPP is carried out on selected road sections of Slovakia’s road
network using results of measurements carried out for the needs of Slovak Road
Inventory. However, given the large number of input variables and conditions, and
thus the large dispersion of the results, demand for a new experimental test facility
84 L. Remek et al.

arose. Design, construction and early operation of such facility were the objective
of project “Independent research of Civil Engineering construction Elements
Effectiveness” supported by the Research & Development Operational Programme
funded by the ERDF.

2.1 Design of APT Tester 105-03-01

The facility was designed based on the requirements of researchers at the University
of Zilina at the Department of Construction management. Design and construction
was supplied by CEIT, as. The first designs were the usual fixed linear APT’s held
by a frame construction. Figure 4 shows the tester in an early design stage.
The subsequent design efforts were aimed to make the tester a semi-mobile
linear APT facility, which could be fairly easily dissassembled, transported and
reasembled somewhere else. The frame was changed for a rail system leading the
vehicle along its path. This however lead to several design problems with sus-
pension of the vehicle as well as material fatigue incurred by sharp accelerations
and decelerations of the tester. These problems were solved by encasing the vehicle
in outer frame and inner frame which could freely tilt, raise and sink within the
outer frame. This arrangement is lead along a leading rail, moveable frame is
suspended in the outer frame by suspension system similar to suspensions of axles
in heavy trucks. The inner frame is also stabilised by hydraulic pistons which
tighten in acceleration and deceleration providing support of the inner frame and
loosen during constant movement enabling the vehicle to better copy pavement

Fig. 4 The first design of APT tester 105-03-01-visualization

Accelerated Pavement Testing in Slovakia: APT Tester 105-03-01 85

Table 1 Main dimensions of APT tester 105-03-01

Dimensions
Length 9042 mm
Width 5178 mm
Width with open doors 2452
Height 2452

Table 2 Technical parameters of APT tester 105-03-01

Technical parameters
Facility Accelerated pavement testing facility
Construction Semi-mobile, linear
Type 105-03-01
Maximum velocity 2.22 m s−1
Load 57.5 kN
Max acceleration 2 m s−2
Max deceleration 5 m s−2
Location Indoor
Operational temperature 10–40 °C
Operational humidity 30–80 % without condensation
Engine power 45 kW
Transition CLP HC VG 320, MOBIL SHC GEAR 320
Energy requirements 3 + N + PE, AC, 50 Hz, 230/400, V, TN-S

unevenessess. The vehicle is powered by an 45 kW electric engine. The transition is

an CLP HC VG 320, MOBIL SHC GEAR 320. Breaking is done by the electro-
motor itself, the power is recuperated and in form of heat lead to radiators. Main
dimensions of the facility are shown in Table 1, technical parameters are then listed
in Table 2.
The final design was the result of many person months spent by academic
personnel and practitioners, in which financial limitations and high requirements on
such tester resulted in probably the best value for money that could be achieved.
Figure 5 shows the final design of APT tester 105-03-0. Figure 6 shows the main
constituents of the tester.

2.2 Pavement Design and Sensory Equipment

The pavement test field has a length of 6 m and a width of 2.2 m. The pavement
structure was designed as a pavement for a road with traffic load class TLC III. It is
86 L. Remek et al.

Fig. 5 Final design of APT tester 105-03-01

a flexible pavement with bitumen concrete surfacing. The wearing base layer is
made of asphalt concrete (AC) 11; CA 35/50; 40 mm thick. The base course layer is
made of asphalt concrete (AC) 16 P, CA 35/50; 80 mm thick. These layers are
connected by penetrating coating PS; 0.5 kg/m2. The road base is a mechanically
bound aggregate MSK 31.5 GB; 180 mm thick. Sub-base is gravel ŠD; 31.5
(45) GC; 200 mm thick. Conformity of all supplied materials has been confirmed
by tests affirming the quality elaborate supplied by the constructor; quality of
particular layers was confirmed through quality tests performed during the con-
struction as prescribed in the test plan (Fig. 7).
The earth work construction –subgrade, beneath the pavement is simulated with
the Elaston rubber layer, the equivalent modulus of this layer lying on a concrete is
extremely similar to an earthwork embankment (Zgutova et al.2012).
Sensor equipment for tensile tangential stress measurement composes 21 ten-
zometers with strain gauges based on the wӧhler bridge principle. These strain
gauges had to withstand the embedding process—temperatures of up to 200 °C
combined with high pressure and vibrations. The body of the tenzometer is hard-
ened plastic, the sensitive parts are coated with Teflon. The range of measurement is
3000 microstrain. The accuracy of measurement is a ratio of input voltage and
existential voltage ≈1.3 mVout/Vex. The strain gauge composes 2 tension gauges
with resistance of 300–400 Ω. Apart from vertical tenzometers, additional 4 vertical
strain tenzometers are embedded in the pavement at the bottom of the surfacing.
Accelerated Pavement Testing in Slovakia: APT Tester 105-03-01 87

Fig. 6 Visualization of final design of the Tester 105-03-1. 1 Wheels, 2 Buffers, 3 Control panel,
4 Fuselage box, 5 Brake heat exchangers, 6 Supporting rail, 7 Guiding rail, 8 Load, 9 Rubber
buffers, 10 Hydraulic liquid container, 11 Hydraulic system, 12 Motor for lifting of the loading
unit, 13 Gear box for lifting of the loading unit, 14 Main electromotor, 15 Container for gear box
oil, 16 Gear box, 17 Electromotor for lifting of the loading unit, 18 Loading unit

Complementary set of sensors is embedded in the pavement for measurement of

humidity and temperature within the pavement. At present, there is only limited
capability to control the temperature of the pavement, however, the tester was build
88 L. Remek et al.

Fig. 7 Pavement test field construction

Fig. 8 Visualization of tenzometer placement in the pavement

with the prospect to install infrared heater devices on the frame of the machine to
simulate summer seasons. Two thermometers are embedded in the pavement cap-
able of temperature measurement within −30 to +80 °C interval with an accuracy of
±0.5 °C. Duo of humidity sensors are embedded in the subbase with the premise
that an experiment involving waterlogging of the pavement should be required.
Generally, many sources of information were used prior to the sensory equipment
design (Willis 2008). The full sensor suite is shown on Fig. 8.
The measured results are to be evaluated in conjunction with material research,
namely rheological and deformational parameters of asphalt mixtures as well as
fatigue of the asphalt construction (Schlosser 2014) (Fig. 9).
Sensors are connected to a PXI platform supplied by National Instruments
company. Data analysis is done via NI Diadem software. Figure 10 shows the
NI PXI platform.

2.3 Early Operation 105-03-01

The tester was finished in October 2014. Figure 11 shows the tester in the con-
struction stage, and Fig. 12 shows the finished tester in the early operation phase.
Early operation phase focused on safety checks, training on operation and
Accelerated Pavement Testing in Slovakia: APT Tester 105-03-01 89

Fig. 9 Sensor equipment in the embedding phase

Fig. 10 Left—NI PXI-1078; right—encased in a protection rack

maintenance personnel and sensor calibration. After first readings, the project team
found that the survivability of embedded sensors is 93 %. Several issues were
addressed and solved during this period, as of January 2015 the tester is in an
operation stage.
90 L. Remek et al.

Fig. 11 Construction stage

Fig. 12 Early operation stage

Accelerated Pavement Testing in Slovakia: APT Tester 105-03-01 91

3 Data Collection and Readings

Pavement performance data gathering during its whole life cycle is the main focus
at this stage of the project. If the measured data are archived thoroughly, they may
serve as a basis for the creation of important advisory models which are known as
degradation or prediction models. We comprehend the deterioration of the pave-
ment as gradual degradation of individual properties deteriorating under a variety of
influences. The aging of the material and its fatigue characteristics have considerate
impact on deterioration. For surfacing materials in particular, it is the transition
from a flexible to a plastic state continuing until reaching the limits of the infraction.
Degradation model characterizes expected changes of particular pavement
parameter in relation to time, or repeated loading. Standard practice for estimation
of pavement degradation and definition of such model is to repeatedly measure and
asses pavement characteristics in given time intervals on particular road section,
and, after a statistical processing of the data, ascertainment of the relationship
between the parameter and loading/time, in most cases, with traffic load, time, or
other factors affecting a given parameter. The results gained from the operation of
APT tester 105-03-0 needs to be compared against long term pavement perfor-
mance monitoring data conducted by the Slovak Road Databank.

3.1 Unevenness

The geometrical shape is measured with stationary High-Accuracy Laser Scanner;

the scanner places one raster point on each square millimetre. This precision is more
than enough to calculate the international roughness index as well as rutting.
Figure 13 shows the raster net as well as longitudinal section through the network
in the wheel track.

3.2 Macrotexture and Unevenness

The macrotexture is measured with hand held high precision laser scanner; the laser
places 100 raster points on a square millimetre. This precision is adequate for
macrotexture measuring. Additionally, from the final raster net, unevenness can be
measured as auxiliary reading to the stationary High-Accuracy Laser Scanner. The
scanning process and raster net is shown in Fig. 14.
92 L. Remek et al.

Fig. 13 Raster network of stationary High-Accuracy Laser Scanner; Down- longitudinal section
in the wheel path

Fig. 14 Scanning with the hand held high precision laser; the laser and the raster network

3.3 Pavement Strain Measurement

Reading from sensor described in Sect. 2.2 is shown in Fig. 15, in this case, the
reading is from static loading, i.e. the loading was steadily increased through the
wheel of the machine.
At the start, the wheel was completely lifted of the pavement; the full weight of the
axle (6 ton) was lowered in 35 s. The presented reading is from sensors directly under
the axle. Strain can be easily recalculated to stress by applying of the Hook’s law.
Accelerated Pavement Testing in Slovakia: APT Tester 105-03-01 93

Fig. 15 Sensor reading: 0–6 ton load in 35 s

Fig. 16 Sensor reading: 30 passes in less than 20 min

Figure 16 shows a reading of 50 passes of the axle over one sensor.

Accumulation of strain in the pavement and subsequent relaxation of the pavement
can be observed on this picture.
The measurements haven’t been thoroughly analyzed and put into the context of
pavement design and pavement performance analysis, but the implication of the firs
readings are quite obvious, the APT tester 105-03-01 is a worthy successor of the
Circular Test Track and the results show premise for research in the ambit of
pavement design, Pavement Management System and Life Cycle Analysis.
94 L. Remek et al.

Acknowledgments The research is supported by the European Regional Development Fund and
the Slovak state budget for the project “Research Centre of University of Žilina”, ITMS
26220220183.

References

Fonód, A. (2005). Degradation of the bearing capacity of asphalt pavements. Slovak Journal of
Civil Engineering, 35–39. ISSN 1210-3896 (print version), ISSN 1338-3973 (electronic
version).
Kováč, M. (2013). Morphology of pavement surface in regards to its serviceability. Thesis at
Department of Highway Engineering, University of Zilina.
Schlosser, F., Šrameková, E., Šrámek, J. (2014). Rheology, deformational properties and fatigue of
the asphalt mixtures. In: Advanced materials research (Vol. 875–877, pp. 578–583), ISSN
1662-8985 (web), ISSN 1022-6680 (print).
Willis, J. R. (2008). A synthesis of practical appropriate instrumentation use for accelerated
pavement testing in the United States. In Proceedings international conference on accelerated
pavement testing, Spain, Madrid, 2008.
Zgutova, K., Decky, M., & Ďurekova, D. (2012). Implementation of static theory of impulse into
correlation relations of relevant deformation characteristics of earth construction. In SGEM
2012: 12th international multidisciplinary scientific geoconference, Albena, Bulgaria
(pp. 107–115). ISSN 1314-2704.
APT with the Mobile Load Simulator
MLS10 Towards Non-destructive
Pavement Structural Analysis

Bastian Wacker, Frédéric Otto, Pengfei Liu, Dawei Wang,

Markus Oeser, Micha Buch and Ulf Zander

Abstract In 2014 a research program has been started about non-destructive test
methods to evaluate the structure of pavements. This task has been given to two
research groups—first research group is led by RWTH Aachen University
(Rheinisch-Westfälische Technische Hochschule) and the second by University of
Siegen. This paper focuses on the initial findings of the running research program.
The assessment of the existing infrastructure and its condition will be one of the
main tasks during the next years in order to use the available budget for mainte-
nance accurately and efficiently. Therefore, it is necessary to identify possible
damages and examine their effects on the road construction. BASt (Federal
Highway Research Institute) is using the Mobile Load Simulator MLS10 for
accelerated pavement testing (APT) on different types of pavements. In addition to
non-destructive test methods, sensors are applied to measure structural impacts. The
overall objective of this research program is to develop a non-destructive test
method that allows the calculation of the remaining life time and load cycles of
pavements. To simulate realistic wheel loads in a short period of time the MLS10
on German full scale standard pavement constructions has been used. The first
pavement test section was loaded with 3 × 106 50 kN wheel loads while the sec-
ond, thinner pavement test section was loaded with 3 × 105 50 kN wheel loads.
Both loads are equivalent to the pavement design load. Three different strategies
have been used to analyze and monitor structural changes. The innovative mea-
surements have been realized by the two research groups to collect data for their
models. The RWTH Aachen collected data with twelve geophones aligned in a row
parallel to the wheel path. The geophones measure the entire vertical deflection
basin of the pavement surface that exists due to the passing real truck wheels. These
measurements were done for different truck speeds and at different transverse

B. Wacker (&)
BASt, Aachen, Germany
e-mail: [email protected]
F. Otto
P. Liu
D. Wang
M. Oeser
RWTH Aachen, Aachen, Germany
M. Buch
U. Zander
Uni Siegen, Siegen, Germany

© Springer International Publishing Switzerland 2016 95

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI: 10.1007/978-3-319-42797-3_7
96 B. Wacker et al.

distances to the wheel path. The University of Siegen collected data by using
acceleration sensors on the surface of the road construction. After recording the data
they were integrated into displacement signals and evaluated. Additionally to those
measurements BASt used conventional equipment to monitor the pavement struc-
ture and surface characteristics. The measurements and evaluation tools used for the
innovation program have a high potential to validate APT programs in the future.
Based on this research it is possible to start further research activities to push the
non-destructive evaluation of pavements structures—not only in APT—into an
improved direction.

Keywords Mobile load simulator MLS10
Non-destructive testing
Structural

analysis Remaining life time

1 Introduction

Roads are the main transport system for loads of personal- and freight-traffic.
Because of the rising traffic loads throughout Europe, it is important to keep the
infrastructure inacceptable conditions. The assessment of the existing road infras-
tructure will be one of the main tasks for the next years in order to use the available
budget for maintenance accurately and efficiently. Therefore, it is necessary to
identify possible damages at an early stage and to examine their effects on the road
construction. The necessity to assess the substance of a road construction on a
network level demands the use of a measurement system that is able to calculate
structural parameters of sections with sufficient accuracy and, above all, in a proper
interpretable manner. In order to meet those requirements, more research is nec-
essary. This research can be done with the help of the Mobile Load Simulator
MLS10 for accelerated pavement testing at 1:1 scale under laboratory conditions.
Besides non-destructive methods, sensors are applied into the test sections to
measure structural effects as well.
In order to support research that deals with this important challenge, the German
Ministry of Transport and Digital Infrastructure started a research program called
‘National Innovation Program’ to support new and innovative ideas. In August
2014 two research groups started with their activities to collect and evaluate
non-destructive data to get more information about the remaining service life and
about the calculated loading cycles (structural substance). During the research
program BASt operated the MLS10 for defined loading periods. At the end of every
loading period each research group has gotten 1 week to produce data with
non-destructive test methods. To validate the results destructive methods have been
used at the end of the APT.
During different measurement periods (after defined loading periods) the inno-
vative test methods have been applied by the research groups on test sections at the
BASt indoor asphalt test track. The out coming results had to be evaluated and if
necessary advanced for the next measurement period. The main focus was on the
APT with the Mobile Load Simulator MLS10 Towards Non-destructive … 97

deeper layers of the structure but for a complete analysis the surface characteristics
were relevant as well. To validate the results of the non-destructive testing,
destructive methods have been used at the end of the APT.
It is important to know that all results are focused only on the impact of loading
with 50 kN wheel load (half standard axle-load) under controlled and monitored
conditions. Because of the indoor situation the air temperature is constantly
between 18 °C (64 °F) and 20 °C (68 °F). During the loading time the temperature
is raising because of the APT unit itself. This fact has been analyzed and evaluated
in a separate research project. The results can be found in Wacker (2015). The
surely important influence of deep tempertures could not be tested during this test
program. This is one of the reasons why this research program is a first step and
needs further research activities during the next years.
The APT-program has been developed for two different test sections out of the
German guidelines RStO 01 (Guidelines for the standardization of the pavement)
(FGSV 2001). One section was designed for 3 × 106 loading cycles and the other
one for 3 × 105 loading cycles. Both sections were constructed in 2003 and have not
been under load since the start of this project. The construction of the first section can
be summarized as followed: 4 cm (1.6 in) surface course SMA 8, 5 cm (2.0 in)
binder course AC 16 and 13 cm (5.1 in) asphalt base course AC. The second con-
struction is thinner and can be summarized with: 4 cm (1.6 in) surface course
SMA 8 and 10 cm (5.1 in) asphalt base course AC 22. In order to get comparable
data from both research groups it was necessary to create a timeline for the relevant
non-destructive test methods. This timeline includes the measurement period before
starting the APT and the measurement period after 3 × 106 loading cycles for the
first section or 3 × 105 loading cycles for the second section. Due to the many
loading cycles on the first section a loading break was created after 1.5 × 106 loading
cycles in order to give both research groups the opportunity for measurements.
The test program started in 2014 and is going to end in 2016. Therefore it is not
possible to present final results in this paper. The paper focuses on the accelerated
pavement testing-program (APT-Program) and the different measurement programs
by the research groups. First results and discussions of the data are presented as well.

2 Accelerated Pavement Testing—The Innovative

Program

Accelerated Pavement Testing (APT) is one important part to evaluate new materials
and pavement structures before using them in practice. Therefore, it is important to
get as close as possible to real loading conditions. So the thickness of the pavement
structure and the loading with loading wheels ought to be the same as in reality. With
the help of the MLS (Mobile Load Simulator) technology, it is possible to accelerate
the loading without changing the loading conditions. Only the frequency of the
loading is different in relation to the frequency of the real traffic. Instead, the
environmental conditions cannot be influenced accelerated with the machine.
98 B. Wacker et al.

2.1 Mobile Load Simulator MLS10

The MLS technology was registered as a patent in 1991. The first full-scale APT
machine with the MLS technology was tested with the Texas Department of
Transportation (TxDOT) shortly after the construction. Further projects have pro-
vided new research projects and it was necessary to build a smaller APT machine.
This machine was a third-scale machine (mmls3) so that it was possible to transfer
the APT research into laboratories and to in-service roads. The next generation of
full-scale machines was called MLS10 and MLS66. The MLS10 was constructed
with four loading wheels and the MLS66 with six loading wheels. In 2006 the first
MLS10 was constructed and tested in Mozambique before the machine was
delivered to EMPA in Switzerland (Arraigada 2009). After the first test results,
more APT machines were build and delivered to China and Germany. The MLS10
arrived at BASt at the end of 2012 and started its work in research projects at the
beginning of 2013.
With a shift of the ownership of MLS-Systems (ZA) to PaveTesting (UK) the
descriptions of the different machines were changed and unified in 2014. After the
transfer the mmls3 is called Pave®MLS11, the MLS10 is called Pave®MLS30 and
the MLS66 is called Pave®MLS66. That makes it possible to identify the constant
loading length of each machine out of the description. For example the
Pave®MLS30 has a minimum constant loading length of 3.0 m (9.8 ft). For sim-
plicity the machine is still called MLS10 in this paper.
The dimensions of the 40 t MLS10 are 12.5 m (41 ft) in length, 3.3 m (10.8 ft)
in width, and 3.5 m (11.5 ft) in height (up to 4.5 m (14.8 ft) in inspection position).
With the use of the MLS it is possible to apply the most uni-directional, straight
loading setups to the test section. To be able to react to special experimental
situations (e.g. change of the position for measurements with other systems) the
MLS10 uses transport wheels at the front and the back. For long distances it is
necessary to use a heavy load truck to transport the MLS10 to the next location. It is
possible to operate the MLS10 with high-voltage current or with a diesel generator.
However, high-voltage current is preferred in order to reduce the pollution and the
noise at test sites.
To operate the MLS10 it is necessary to lift up the transport wheels and lower
the machine onto four corner jacks. Thereby, one of the loading wheels inside the
MLS10 is set down to the test section. During the test program only one of the four
loading wheels is on the ground at a time. All wheels are moving with the same
distance to each other in a closed chain through the machine. To hold the loading
wheels, four Bogies are placed inside the MLS10 (Fig. 1). It is possible to use the
MLS10 with different configurations. First of all the loading wheels can be changed
between twin and super-single tires. Secondary there is the option to adjust the load
between 40 and 75 kN. This is possible by using a hydraulic system (gas-oil
mixture) for every Bogie. The standard adjustment at BASt is 50 kN (standard
design wheel load in Germany) with a tire pressure of 8.5 bar (123 psi). Using this
load it is possible to operate the MLS10 with a maximum speed of 22 km/h
APT with the Mobile Load Simulator MLS10 Towards Non-destructive … 99

Fig. 1 MLS10 (Pave®MLS30) [left]; one of four Bogies (PaveTesting 2015) [right]

(13.6 mph). That leads to 6.000 uni-directional loading cycles per hour. All impacts
are applied to a 3.5 m (11.5 ft) long test section in the middle of the MLS10.
Each Bogie (Fig. 1) has six axles with steel wheels on its sides. These steel
wheels direct the Bogie through the machine. With 24 linear motors the MLS10 is
moving the four Bogies. Reaction plates on every Bogie are installed to connect it
to the linear motor.
For special research questions it is possible to move the MLS10 sideways. It is
possible to create shearing forces and simulate the tracking with the realization of a
lateral movement of about 500 mm (19.7 in) in each direction.
To reduce the noise emission during the operation time, the MLS10 is covered
and enclosed. On top of the machine there are four ventilator systems installed in
order to avoid heat accumulation during and after the loading time.

2.2 Measurement Program Done by BASt

The main task of BASt was to operate the MLS10 during the test program. That
included daily operation (around 8 h/day) and the regular inspection work with the
MLS10. In addition, BASt created its own measurement program with
non-destructive test methods. This measurement program was independent from the
measurement periods for the research groups and has been running in parallel to the
entire program.
During the loading cycles BASt was using strain sensors at the bottom of the
asphalt layer in the center of the loading area and the sensor sent data every 15 min.
The temperatures of the indoor asphalt test track (air and construction) were
monitored during the entire research program.
In different intervals the Falling Weight Deflectometer (FWD) has been used
with three 50 kN drops for two different measurement concepts. The main concept
included three measurement points in the center-line of the loading area (wheel
track) and the extended concept had three to four more points transversal to the
100 B. Wacker et al.

center line. During the FWD measurements the sensors inside the construction were
active as well.
In order to measure the transverse evenness a profilometer was used. For each
test section five fixed profiles were measured. For the first section measurements
were realized every 5 × 105 loading cycles. For second test section the measure-
ment was realized at the beginning and end of the loading program. At the same
time the surface of the test section was observed and pictures were taken.
For the second section it was decided to use Ground Penetration Radar (GPR) as
well. Because of the sensors inside the construction it turned out to be hard to detect
the layers and structures. Therefore a measurement setup with two longitudinal and
five transverse profiles has been developed. The observation of the test section was
realized before the beginning and at the end of a loading cycle with MLS10.

2.3 Measurement Program Done by RWTH

Aachen/HELLER

The measurement program by the research group RWTH Aachen and HELLER
consisted to deflection measurements at the surface created by a passing 40 t
semitrailer (truck) with a known axle load distribution as well as a measured tire
pressure. These measurements were carried out on the two test sections after
described loading cycles with the MLS10. For the first test section the measurement
were taken: at the beginning (no previous load applied), after 1.5 × 106 load cycles
by the MLS10 and the final measurement after 3.0 × 106 load cycles. For the
second test section only the initial and the final measurement (after 3.0 × 105) were
used. Due to the geometrical restriction of the test sections, the speed for the
passing truck was limited to 30 km/h (18.6 mph). The left wheels of five axles were
considered in the back-calculation in order to simplify the model. The geometry and
the distribution of the loads are illustrated in Fig. 2.
Geophones of type RTC-4.5 Hz-395 produced by R.T. Clark Companies were
used to measure the deflection of the test sections under traffic loads. One of the
geophones used in this study is shown in Fig. 3 [left]. Its natural frequency is
4.5 Hz with a tolerance of ±0.5 Hz and its sensitivity is 23.4 V/m/s with a toler-
ance of ±10 %.
On the test track surface a total of 12 geophones were arranged parallel to the
driving direction of the passing truck. Those were distributed equidistantly on the
test section over a length of 3300 mm. Geophones measure vibration speeds of the
test section surface caused by the truck passing by. Using a laser-distance sensor,
the position of the truck and therefore the distance between the wheels and the
geophones could be detected at any time during the measurement and used for the
evaluation. The measurements were carried out multiple times at different distances
between the wheels and the geophones, but with constant driving speed, permitting
an estimation of the transversal shape of the surface deflection. After several
APT with the Mobile Load Simulator MLS10 Towards Non-destructive … 101

0,9 MPa 0,75 MPa 0,9 MPa

Firestone TORQUE Firestone
385 / 65 R 22.5 315 / 80 R 22.5 385 / 65 R 22.5
220 mm
260 mm 260 mm
560 mm

2440 mm
2380 mm

2440 mm

132 mm
118 mm 176 mm

3650 mm
6600 mm

7900 mm

9250 mm

Fig. 2 Geometrical and loading data of the truck (Gohl 2006)

Fig. 3 Geophone used in this study [left]; the field setup of passing track and geophones [right]

adjustments, the data were converted into the surface deflection of the test section.
Due to a measuring error, the data of the first geophone were disregarded, i.e., only
11 geophones were used in this study. The field setup is shown in Fig. 3 [right].

2.3.1 The Algorithm of Back-Calculation

A back-calculation program combining the finite element program SAFEM and

artificial neuronal networks (ANN) to evaluate the bearing capacity of the test
section has been developed. The flowchart of the back-calculation is provided in
Fig. 4. The elastic modules E of different pavement layers have been generated as a
collection within a reasonable range in order to get all possible combinations
102 B. Wacker et al.

Geometrical Parameters Material Parameters

Boundary Conditions Pavement length (known) E-modulus (a range)
Moving loads etc. (known) Pavement width (known) Poisson ratio (known)
Layer thickness (known) Density (known)

input
SAFEM

output
Computational Deflections (database)

training / validation
input
Measured Deflections ANN

output
Back-calculated E-modulus

Fig. 4 Flowchart for the back-calculation

encountered in the pavement. The other input parameters such as boundary con-
ditions, geometrical parameters, and material parameters except for E-modules are
known. The database including the pairs of E-modules of each pavement layer and
corresponding surface deflection of the pavement is computed by SAFEM and then
fed into the ANN which is trained to set up the relationship between the E-modules
and the surface deflection. Hereafter, the deflections from field measurements are
transformed into the trained ANN and the corresponding E-modules are
back-calculated.
In order to simplify the computation, the five real layers of the test section were
merged whereby the three asphalt layers were merged to an equivalent asphalt layer
and the two bottom layers to an equivalent unbound layer in the back-calculation.
The interlayer behaviour was defined as being partially bound. The parameters of
the simplified test section structure are listed in Table 1.

Table 1 Thicknesses and material properties of the first test section used in back-calculation
Real layer Equivalent Length Width Thickness E (Mpa) μ Density
layer (mm) (mm) (mm) (–) (t/m3)
Surface course Equivalent 20,000 3750 220 4000–15,000 0.30 2.356
Binder course asphalt interval: 200
layer total: 56
Asphalt base course
Frost protection layer Equivalent 20,000 3750 2120 40–160 0.49 2.400
Sub-grade unbound interval: 20
layer total: 7
APT with the Mobile Load Simulator MLS10 Towards Non-destructive … 103

2.4 Measurement Program Done by Uni Siegen/TU

Dresden/HS Mittweida

In the research project “Evaluation of the Structural Substance and Calculation of

Residual Value of Asphalt Road Pavements” (Grohs 2012) funded in the earlier
‘National Innovation Program for Roads’, the Falling Weight Deflectometer
(FWD) has been used. In order to facilitate an improved formation of homogeneous
sections for the assessment of the structural substance the FWD is the most
widely-used load capacity measuring procedure in Germany. The measuring device
was modified for recording data and the evaluation process was redeveloped
entirely.
This evaluation procedure does not interpret the deformations formed on the
asphalt surface through the falling weight. Instead, it analyses the surges running
through the pavement produced by the FWD during an impulse. Initially, the
different single frequencies of the impulse are separated by a Fast Fourier analysis
and then allocated to certain depth positions by using physical principles and
empirical cohesions. As a basic principle, the low-frequency surges pass into the
deeper layers of the pavement structure while the high-frequency surges are allo-
cated to the stiffer layers of the asphalt construction. Speeds can be calculated for
the relevant surges using the time difference between a defined pair of sensors. If
the speeds are characteristic for the stiffness of the asphalt material, the modulus of
elasticity of the individual layers could be calculated from this. Many measure-
ments with the FWD produced the result that only surges with frequencies that pass
into the unbound layers of the road construction and sometimes into the asphalt
base course. Therefore, tests with a shaker to receive more detailed measurement
data were carried out. A shaker has the advantage, that higher and more defined
frequencies can be initiated.
The current innovative program is devided in two parts within the research
project: First, different measurements are carried out on the two test sections at the
asphalt test track of BASt. The measurements with the aim of recording and
defining parameters to describe the substance took place before loading, during the
interruption phase and after loading with the MLS10. Secondly, basic investigations
into the development of the stiffness of asphalt under loading cycles are carried out.
With the help of laboratory tests the cohesion between stiffness and state of fatigue
will be analysed.
The first part is a completely new approach. Various procedures of surge initi-
ation, measurement and evaluation have been investigated. To transmit vibrations
into the road construction a continuous vibration simulation by a shaker is used.
Additionally, hammers and falling weights were used to transmit vibrations as using
an impulse input. In addition to these two types of vibration simulation, the MLS10
of BASt was used to initiate vibrations. However, from measuring period to
measuring period the concept was modified. The modification was done by elim-
ination of not working measurement setups.
104 B. Wacker et al.

All measurement setups are developed in such a way that a structural parameter
to describe the structural substance should be found. The stiffness of the attached
and unattached layers is seen as the primary structural parameter of the investigated
asphalt pavement. However, these stiffness are calculated by using different
pavement simulations. Different vertical deformation of the structure and the speed
of the induced horizontal surges are relevant as well to the structure on a secondary
level. First, both parameters are provoked with the MLS10 on the basis of the
loading cycles. Secondly, a shaker is used that supplies another opportunity to
evaluate the surge speeds.
No information about the structural substance or the remaining life span of a
pavement construction can be made on the basis of the asphalt stiffness alone. In
order to get more information about the cohesion between stiffness and fatigue the
second part of the project has been created. Laboratory asphalt test specimens have
been investigated parallel to the measurements on the test section. Fatigue crack
tests were carried out to calculate the stiffness modulus temperature function and
the fatigue function of the asphalt test specimens at an unloaded stage as well as at
different degrees of damage.
The entire approach of the research project is designed to calculate the principles
for a non-destructive measuring procedure to evaluate the structural substance. To
evaluate the structural substance it is necessary that the MLS10 produce detectable
substance damage during the project. First estimations about the structural failure
have been carried out at the beginning of the project. The information necessary
was provided out of laboratory trials on previous core drillings of the indoor asphalt
test track. During the further progress of the investigations, it is attempted to
readjust the models through theoretical prognoses on the progress of the fatigue and
through the results coming from the measurements on the asphalt test track.

3 Structural Pavement Analysis

The chapter ‘Structural Pavement analysis’ focuses on the first experiences with the
three above described measurement programs and presents first results. The final
results will be available at the end of the project, estimated at the end of 2016.
Because of the running evaluation process, most of the results are coming from the
first test section. The other test section as well as more detailed information will be
presented and analyzed in further publications.

3.1 Analysis Done by BASt

The focus of this chapter is on first results of the surface observation. For this
purpose the transverse evenness and the surface damages are presented. Every
5 × 105 loading cycles both parameters have been monitored and collected.
APT with the Mobile Load Simulator MLS10 Towards Non-destructive … 105

Fig. 5 Surface characteristics; old setup [top]; new modification (for next APT-programs)
[bottom]

The surface inspection, direct at the pavement as well as on high resolution

images, shows that no cracks or raveling have developed after 3 × 106 loading
cycles with the MLS10. Due to experiences of this APT-program a new camera
position has been chosen in order to eliminate the angle shown in Fig. 5. The new
modification is going to be used for further APT-programs.
Based on the transverse evenness measurements it is possible to visualize the
changing at the surface as a consequence of the loading induced by the MLS10. The
top graph of Fig. 6 shows the results of the evenness measurements done in the last
third of the test section. The wheel track of the twin-tires is easy to identify and
different deformation types can be seen. Within the wheel tracks a ‘deformation
valley’ effect can be seen. Beside and between the path of the twin-tires the
deformation builds a ‘deformation hill’. Obviously, the deformation between the
twin-tires is noticeable with 2 mm (0.07 in).
From the chronological sequence (Fig. 6, bottom graph) it is possible to get a
first idea where the deformation is coming from. The important ‘deformation val-
ley’ of about 3 mm (0.12 in) was reached after 1.5 × 106 loading cycles and no
more changes were observed so far. Since then the level of deformation stays
constant between 2.5 and 3 mm (0.1–0.12 in). These results suggest that the first
loading cycles up to 1.5 × 106 loading cycles compacted the test section. The
chronological sequence in Fig. 6 shows clearly that the ‘deformation hill’ is
behaving differently and goes up constantly. One reason is that the MLS10 has been
106 B. Wacker et al.

Fig. 6 Transverse evenness

first test section; profile
measurement [top] and
chronological sequence of
maximum deformation
[bottom]

used on a straight line without lateral movements. Another reason could be the
higher temperature of the surface which is beneficial for higher deformation
(Wacker 2015).
The main evaluation process is the combination of different measurements to
show a possible relationship between these results. For that purpose data out of the
FWD measurements and other available data are going to be analyzed and com-
pared to each other. For example, if the construction was compacted during the first
1.5 × 106 loading cycles it could be that this effect can be found at the FWD data as
well. Another focus of the evaluation process will lay on the FWD data itself and
the different measurement concepts.

3.2 Analysis Done by RWTH Aachen/HELLER

The analysis done by the research group RWTH Aachen/HELLER focuses on the
first test section and will be continued on the second test section.

3.2.1 Back-Calculation of Bearing Capacity of the Test Section

The surface deflections derived from the three measurements, performed after
different loading cycles on the first test section with the 40 t truck, were included
into the ANN as input values. In order to ensure the comparability of the results,
specific conditions were used to retrieve the deflection values from the measure-
ment (when the second axle of the truck was passing the fourth geophone). As
shown in Fig. 7, the minimum peak value of the deflection can be found at (or
around) the fourth geophone for each measurement. The values of the x-axis rep-
resent the position of the geophones on the test section, with the positive orientation
of the coordinate axis corresponding to the driving direction of the truck and the
APT with the Mobile Load Simulator MLS10 Towards Non-destructive … 107

Fig. 7 The surface 0,00

Measurement of state 1 (N=0)
deflections derived from -0,01 Measurement of state 2 (N=1.5 Mio.)
mearsurements Measurement of state 3 (N=3.0 Mio.)
-0,02

-0,03

Deflection [mm]
-0,04

-0,05

-0,06

-0,07

-0,08

-0,09
5500 6000 6500 7000 7500 8000 8500 9000
X coordinaten in the simulation [mm]

reference point (x = 0) defined at a fixed point before the test section. The same
geometric dimensions were used in the SAFEM model for the database setup and
for the ANN training.
Following the procedure shown in Fig. 4, the E-modules of the equivalent
asphalt and unbound layers at the three measurement periods were back-calculated
and shown in Fig. 8.
Both E-modules of the equivalent asphalt and the unbound layers, will decrease
after 1.5 × 106 loading cycles, compared to the initial measurement (see Fig. 8). As
a result, the minimum peak value of the deflection decreases as well, which is
shown in Fig. 7. However, after 3.0 × 106 loading cycles, the E-modules of the
unbound layer suddenly increases, whereas the E-Modulus of the asphalt continues
to decrease. This effect manifests itself in the shape of the deflection (see Fig. 7), of
which the minimum peak value, as well as the neighboring values, are visibly
higher than their respective corresponding values of the previous measurements.
This particular increase of the E-modules in the unbound layers might be caused by
post compaction related processes, which occurred during the loading cycles of the
MLS10.
With the back-calculated E-modules, the tensile strains eel at the bottom of the
equivalent asphalt layer were computed by SAFEM under MLS10 loading condi-
tions. In order to verify the reliability of the prediction of the back-calculation, drill
cores of some specimens were taken from the asphalt base layer of the test track to
test the fatigue resistance with the indirect tensile test based on German rules and
regulations (FGSV 2009) which was at the same temperature and frequency as
applied for MLS10. The relationship between the number of loading cycles to the
failure of the indirect tensile test and tensile strain was determined with Eq. (1):

Nind ¼ 4:221
ðeel
103 Þ3:474 ð1Þ

The calculated loading cycles correlate well with those produced by MLS10.
The correlation is shown in Fig. 9.
108 B. Wacker et al.

E-modulus [MPa] 20000 120

E-modulus [MPa]
15000 110

10000 100

5000 90

0 80
0 1 2 3 4 0 1 2 3 4
Loading cycles by MLS10 [Mio.] Loading cycles by MLS10 [Mio.]

Fig. 8 The relationship between the loading cycles and E-modulus of the equivalent asphalt
layers [left]; the relationship between the loading cycles and E-modulus of the unbound layers
[right]

Fig. 9 The correlation

Loading cycles to the failure of the

40
between the loading cycles by
MLS10 and loading cycles to
indirect tensile test[10 3 ]

the failure of the indirect 30

tensile test
y = -5,2532x + 24,897
R² = 0,9699
20

10

0
0 1 2 3 4
Loading cycles by MLS10 [Mio.]

The number of loading cycles to the failure of the indirect tensile test decreases
with the increase of loading cycles applied by the MLS10, which confirms the
decrease of the bearing capacity. The correlation between the back-calculated
number of loading cycles to the failure of the indirect tensile test and the loading
cycles of MLS10 on the test section is established as a linear dependency with very
good correlation.

3.3 Analysis Done by Uni Siegen/TU Dresden/HS Mittweida

The types of load impact (shaker, FWD, hammer, and MLS10) were inspected
critically after the first measuring period. A selection out of those types was done
for the second measurement period as part of the research project. During data
inspection, it emerged that the vibration simulation of the MLS10 is inappropriate
APT with the Mobile Load Simulator MLS10 Towards Non-destructive … 109

in order to assess the structural substance with the equipment chosen. For further
evaluations the vibrations triggered by the shaker were used as the main type of
load impact.

3.4 Interpretation of All Analysis Done by the Two Research

Groups and BASt

The transverse evenness measurements done by BASt and the back-calculated

E-modules done by RWTH Aachen/HELLER have shown interesting aspects. It
seems to be possible to detect the point of time where the compacting process
stopped and the whole energy of the loading worked into the entire pavement
construction. Independently from the measurement or calculation system an
abnormality was identified around 1.5 × 106 loading cycles. Especially the
E-modulus of the unbound layer increased and the level of the maximum defor-
mation stayed stable. In order to verify this observation further APT-programs
should be done regarding this aspect.

4 Summary and Outlook

The Accelerated Pavement Testing (APT) for the current ‘National Innovation
Program’ has been completed but the evaluation is still running. The main focus of
the research activities was on the structural analysis with non-destructive test
methods and evaluation processes. For this purpose two research groups (RWTH
Aachen/HELLER and University of Siegen/TU Dresden/HS Mittweida) got the
opportunity to do innovative measurements on full-scale pavements at different
damage levels in order to gather non-destructive data for the evaluation process.
For inducing the loading to the two test sections the Mobile Load Simulator
MLS10 was used. The first test section was loaded with 3 × 106 loading cycles of
50 kN, the second test section was loaded with 3 × 105 loading cycles of 50 kN.
During the loading periods different measurement programs were realized by BASt
and the research groups. First results of these measurements have been presented in
this paper. The main focus of this paper was on the description of the measurement
programs.
During the realization of the loading program some improvements for further
APT-programs could be identified. First results are available regarding the behavior
of the pavement. Some measurements with geophones and other sensors were
realized while surface characteristics and transverse evenness measurements were
done as well. It is impressive that two different systems were able to detect a
changing in the first test section construction after 1.5 × 106 loading cycles. The
‘deformation-valley’ was increasing constantly up to the described loading cycles.
110 B. Wacker et al.

After reaching the maximum deformation it stagnated and remained stable while the
E-modulus of the unbound layer increased. At the end of the research program more
interesting results are to be expected based on laboratory results, non-destructive
measurement results (e.g. out of different FWD measurements) and final analytical
results.
A new APT-program has been started on the second test section using the
second loading lane. By use of the MLS10 the construction is going to be loaded
until failure. During this APT-program the measurement method of RWTH Aachen
could be used under the MLS10 in order to detect the actual deflection of the
MLS10 loading. Until March 2016 the test section has been loaded with 1.8 × 106
loading cycles. Impacts of periods without loading and the controlled environment
inside the test hall have to be considered when transferring the results to in service
roads in a second step.

References

Arraigada, M., Kalogeropoulos, A., Hugenschmidt, J., & Partl, M. N. (2009). Pilotstudie zur
Evaluation einer mobile Grossversuchsanlage für beschleunigte Verkehrslastsimulation auf
Strassenbelägen. Pilot-Study for the evaluation of a Mobile Full-Scale Accelerated Pavement
Testing equipment, EMPA (in German).
FGSV. (2009). Arbeitsanleitung zur Bestimmung des Steifigkeits- und Ermüdungsverhaltens von
Asphalten mit dem Spaltzug-Schwellversuch als Eingangsgröße in die Dimensionierung.
Instruction sheet for identification of asphalt stiffness and fatigue for design, AL Sp - Asphalt
09. FGSV, Köln 9 (in German).
FGSV. (2001). Richtlinien für die Standardisierung des Oberbaus von Verkehrsflächen. Guidelines
for the standardization of the pavement, FGSV, Köln (in German).
Gohl, S. (2006). Vergleich der gemessenen mechanischen Beanspruchungen der Modellstraße der
BASt mit den Berechnungsergebnissen ausgewählter Programme. TU Dresden, Professur für
Straßenbau, Diplomarbeit (in German).
Grohs, S., Lindemann, J., Hübelt, J., & Zander, U. (2012). Beurteilung der strukturellen
Restsubstanz von verkehrsflächenbefestigungen in Asphaltbauweise auf der grundlage
impulsförmiger Schwingungsanregung. Evaluation of the structural substance and calculation
of residual value of asphalt road pavements, Siegen, Institut für Straßenwesen (in German).
PaveTesting. (2015). http://www.pavetesting.com/.
Wacker, B. (2015). Accelerated pavement testing program with the mobile load simulator MLS10
—Temperature analysis. In 6th International Conference Bituminous mixtures and Pavements,
Greece.
Evaluating Nonlinearity on Granular
Materials and Soils Through the Use
of Deflection Techniques

Fabricio Leiva-Villacorta, Luis Loría-Salazar

and Edgar Camacho-Garita

Abstract The nonlinear behavior of soils and granular materials is widely known.
Its proper evaluation is essential for the accurate determination of the stiffness of
pavement layers and their structural evaluation. If this phenomenon is not con-
sidered appropriately, it can lead to significant errors in estimating the pavement
responses. As part of the investigation of accelerated tests on full-scale pavements
by the National Laboratory of Materials and Structural Models of the University of
Costa Rica, four instrumented pavement sections were built and evaluated using a
Heavy Vehicle Simulator. The wheel load was used at an average speed of
10 km/h, with various loading levels, at an average temperature of 23 °C and with a
lateral wandering of 10 cm. The deflection profile of each pavement was studied by
means of Multi-Depth Deflectometer sensors (MDDs), the Road Surface
Deflectometer (RSD), and; the Falling Weight Deflectometer (FWD). This paper
summarizes a comparison among the differences in the deflection basin measure-
ments, and the related estimated moduli. Additionally, the pavement response
modeling was performed by means of Multi-layer Elastic (MLE), Finite Element
(FE) and Linear Visco-Elastic (LVE) methods. The objective of this study was to
compare the results of different deflection measuring methods and their corre-
spondent backcalculated moduli. The analysis showed similar results in the
deflection curves and hence, back-calculated moduli obtained by the MDDs and the
RSD. Both devices captured, as expected, nonlinear behavior of the granular
materials as well as the subgrade. The FWD was not able to report such behavior.

F. Leiva-Villacorta (&)
L. Loría-Salazar
E. Camacho-Garita

National Laboratory of Materials and Structural Models (LanammeUCR),
University of Costa Rica, P.O. Box 11501-2060, San José, Costa Rica
e-mail: [email protected]
L. Loría-Salazar
e-mail: [email protected]
E. Camacho-Garita
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 111

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_8
112 F. Leiva-Villacorta et al.

1 Introduction

It is generally known that unbound materials have a non-linear elastic stress-strain

behavior. Because the unbound base courses have a substantial influence on the
load carrying capacity of the pavements, proper characterization of the mechanical
response of unbound aggregate materials is a crucial factor. In flexible pavements
granular layers play an important role in the overall performance of the pavement.
Consequently, to establish more rational pavement design and construction criteria,
it is essential that the response of unbound layers under traffic loading be thor-
oughly understood.
Accurate stiffness (modulus) characterization of the unbound layers is critical for
accurate pavement layer thickness design. For any mechanistically based pavement
design, an accurate knowledge of the strains at critical locations in the pavement
structure are necessary to give a quality prediction of the design life of that pavement.
Often, the stiffness of the unbound materials is quantified as the resilient mod-
ulus. Resilient modulus is the ratio of deviatoric stress (from a triaxial compression
test) to recoverable strain. This term is utilized since unbound material behavior is
not completely elastic; these materials usually experience some permanent defor-
mation after each load application. Unbound material resilient moduli can be
determined either in the laboratory using the triaxial testing apparatus or in the field
using a variety of non-destructive testing methods. The laboratory resilient modulus
test calculates resilient modulus by directly measuring the load and deformation of
the test specimen under repeated loadings that simulate the quick loads imparted on
the pavement structure by moving traffic. One of the most common field tests for
resilient modulus is done with a falling weight deflectometer (or FWD).
The resilient moduli of unbound paving materials often exhibit non-linear stress
dependent behavior with varying stress-states within the material (Irwin 2002). This
behavior can either be stress-hardening (increasing stiffness with increasing stress)
or stress-softening (decreasing stiffness with increasing stress) (Irwin 2002).
Research into unbound material performance through laboratory and field testing
has yielded several constitutive relationships relating resilient modulus to
stress-state. These models contain a wide range of terms that quantify the
stress-state of the unbound materials, such as: bulk stress (θ), deviatoric stress (σd),
and octahedral shear stress (τoct).
In order to effectively characterize the stiffness behavior of unbound materials,
several factors must be considered. A better representation of material behavior can
be given by either laboratory or field resilient modulus testing or both. The unbound
material is stress-sensitive; therefore, a model that best quantifies the behavior of
that material should be selected. Taking these factors into account could allow for
the most accurate quantification of resilient modulus for pavement design and more
accurate modeling of pavement design life. For many researchers and transportation
agencies, these factors can be analyzed through full-scale accelerated pavement
testing (APT).
Evaluating Nonlinearity on Granular Materials and Soils Through … 113

One such facility is the Costa Rican APT program call PaveLab. A technical and
economical study was performed and aided in determining that the Heavy Vehicle
Simulator (HVS) was the best fit for the medium and long term pavement perfor-
mance assessment. Specifically, the HVS met the following mobility, accelerated
pavement evaluation, application of real loads and comparable results from similar
equipments (Coetzee et al. 2008). Part of the testing program at the Pavelab
involves characterization of the various paving layer materials in both the labora-
tory and the field, making this facility ideal for a study of this nature.

1.1 Pavelab Test Tracks

For the first stage of accelerated tests in Costa Rica the construction of 4 experi-
mental sections was performed in May 2013 (Fig. 1). The objective of this phase
was to perform a structural comparison in terms of thickness of the asphalt concrete
layer and base material type (granular vs. cement treated) (Aguiar-Moya et al.
2012). Table 1 shows the characteristics of the 4 sections with their respective layer
thicknesses obtained from Ground Penetrating Radar (GPR) measurements and
backcalculated layer moduli based on Falling Weight Deflectometer results. The top
layer consists of an asphalt concrete mixture with nominal maximum aggregate size
of 19.0 mm with an optimum binder content of 4.9 % by total weight of mixture.
The cement treated base (CTB) was designed to withstand 35 kg/cm2 with an
optimum cement content of 1.7 % by volume of aggregate and with a maximum
density of 2013 kg/m3. The base material and granular sub-base were placed at a
maximum density of 2217 kg/m3 with an optimum moisture content of 8.6 %. The
sub-base material had a CBR of 95 %. Finally, the subgrade material was con-
structed for a maximum density of 1056 kg/m3 with an optimum moisture content
of 52 % and CBR of 6.6 %.

Thick Asphalt Layer Thick Asphalt Layer

Cement Treated Base Granular Base
Granular Subbase Granular Subbase

Thin Asphalt Layer Thin Asphalt Layer

Cement Treated Base Granular Base
Granular Subbase Granular Subbase

Fig. 1 Test track distribution

114 F. Leiva-Villacorta et al.

Table 1 Test tracks in-place properties

Properties\section AC1 AC2 AC3 AC4
Asphalt concrete thickness (H1) (cm) 5.1 6.3 13.2 13.2
Base thickness (H2) (cm) 21.9 21.2 31.0 24.9
Subbase thickness (H3) (cm) 30.1 30.1 30.1 30.1
HMA modulus (E1) @ 25 °C (MPa) 3800 3800 3800 3800
Base modulus (E2) (MPa) 1200 170 170 1200
Subbase modulus (E3) (MPa) 140 140 140 140
Subgrade modulus (E4) (MPa) 70 70 70 70

1.2 Instrumentation

The experiment included not only the instrumentation integrated with the HVS
system but also embedded instrumentation in all four test sections. HVS onboard
sensors can record the applied load, tire pressure and temperature, position of the
load and the velocity of the load carriage. Embedded instrumentation include
asphalt strain gauges (PAST model sensors), pressure cells (SOPT model sensors),
multi depth deflectometers (MDD), moisture and temperature probes. These sensors
were chosen based on previous HVS owners experience (Baker et al. 1994).
Additionally, the HVS was equipped with a laser profiler that can be used to create
a tridimensional profile of the section and a Road Surface Deflectometer is added to
the testing equipment to obtain deflection basins at any location along the test
section (Leiva-Villacorta et al. 2013, 2015).
Figure 2 shows the instrumentation array used for the first series of experimental
sections. The PAST sensors were placed at the base/HMA layer interface and were
placed in the longitudinal or traffic loading direction and in the transverse direction.
MDD sensors were installed at 4 different depths to cover all 4 structural layers. As
for the thermocouples, these were placed at 4 depths: surface, middle depth of the
HMA layer, at the PAST sensors depth and 5 cm into the base layer. In the case of
AC1 and AC3 sections the same gauge array was used while excluding PAST

30 cm 60 cm 90 cm

MDD MDD
Thermocouple

Section Length = 6.0 m

HMA
GB/CTB

Subbase
Concrete Slab
PAST Sensors
Subgrade

Fig. 2 Sensor array

Evaluating Nonlinearity on Granular Materials and Soils Through … 115

sensors. The Road Surface Deflectometer (RSD) was used to measure deflections at
other locations along each section.

1.3 Surface Modulus as Indicator of Non-linear Behavior

The surface modulus is the ‘weighted mean modulus’ of the semi-infinite space
calculated from the surface deflection using Boussinesq’s equations (Ullidtz 1987).
The surface modulus at a distance ‘r’ roughly reflects the surface modulus at the
same equivalent depth z = r. If the subgrade is a linear elastic semi-infinite space,
the surface modulus should be the same at varying distances. If a stiff layer is
present, the surface modulus at some distance should become very large. The
surface modulus (SM) directly under the point of loading at maximum deflection
D0 is calculated with Eq. 1, the general formula for surface modulus (SM) at any
point away from the point of maximum deflection is calculated with Eq. 2.
 
a
SM ¼ 2r0 ð1  l2 Þ ðr ¼ 0Þ ð1Þ
d0
 

a2
SM ¼ r0 1  l2 ð2Þ
rdr

where,
SM = Surface modulus at a distance r from centre of loading plate (MPa)
σ= Contact stress
µ= Poisson’s ratio, usually chosen as 0.35
a= Radius of the loading plate
d(r) = Deflection at distance r
r= Radial distance from the centre of loading where r > 0.
Ullidtz (1987, 2005) determined that the gradient of the surface modulus
(SM) plot over more or less the radial distance away of the deflection bowl can be
used to identify whether the subgrade has stress softening, stress hardening
behavior or whether it is exhibiting linear elastic behavior. Figure 3 shows what it
can be defined as linear elastic behavior (on the left) and non-linear elastic
behavior/rigid layer effect. The “kick” in the surface modulus plot (Fig. 3—right)
could be caused by a shallow rigid layer (bedrock) or nonlinear (stress softening)
materials such as clay will cause a similar effect in the deflection basins.
Leiva-Villacorta et al. (2015) performed an analysis of deflection basins using
surface modulus theory. Their results indicated that before APT test loads were
applied to one test track, FWD deflection basins most likely reflected an elastic
linear behavior of the lower layers while RSD and MDD reflected a moderately
non-linear elastic behavior. Furthermore, a similar surface modulus analysis per-
formed at the end of APT testing, indicated that FWD results consistently reflect an
elastic linear behavior of the lower layers, while RSD and MDD reflect a more
116 F. Leiva-Villacorta et al.

Surface Modulus E (MPa) Surface Modulus E (MPa)

0 100 200 300 400 500 600 700 0 50 100 150 200 250 300 350 400
0 0
Radial Distance r (mm)

Radial Distance r (mm)

200
500 400
600
1,000 800
1,000
1,500 1,200
1,400
2,000 1,600
1,800
2,500 2,000

Linear elastic behavior Non-linear elastic

Fig. 3 Typical surface moduli plots for pavement structures

intensified non-linear behavior of the lowers layers and could also exhibit the
presence of the test pit concrete support layer (shallow rigid layer). Therefore, a
more in depth study was recommended in order to explain the differences in the
observed behavior based on the different deflection measurement devices.

1.4 Objective

The objective of this study was to analyze and compared measured and modeled
surface deflection basins to try explain the linear and non-linear behavior of the
unbound layers captured by different devices.
In order to achieve the objective of this study, Multi-layer Elastic (MLE), Finite
Element (FE) and Linear Visco-Elastic (LVE) modeling was used to compute
surface deflections based on linear elastic properties of the four pavement structures
and non-linear elastic properties of the unbound layers. On the other hand, FWD
deflection basins at different load levels that were performed on the four test tracks
located at Lanamme’s APT facility were evaluated. Deflections obtained from
surface MDD sensors and the RSD testing at different load levels and locations
were analyzed as well. Finally, computed and measured results were compared and
the linear elastic/non-linear behavior was defined based on surface modulus plots
and benchmark parameters.

2 Pavement Modeling

The pavement response modeling was performed by means of Multi-layer Elastic

(MLE), Finite Element (FE) and Linear Visco-Elastic (LVE) methods. The pave-
ment structure was firstly simulated considering the subgrade as semi-infinite layer
and secondly considering the actual 40 cm concrete slab located at a depth of
2.8 m. The simulated elastic modulus of concrete slab was 8 000 MPa and it was
expected to behave as a shallow bed rock. A single circular load with uniform
Evaluating Nonlinearity on Granular Materials and Soils Through … 117

pressure distribution was used to compute surface deflections at the same locations
of the FWD sensors and the FWD load configuration (plate diameter = 300 mm).
Two loads were simulated 40 and 60 kN for all four sections.
The surface modulus differential (SMD) defined as the difference between the
surface modulus at 600 mm and that at 1200 mm was used to discriminate between
linear and non-linear behavior. This concept was introduced by Horak (2008).
Horak (2008) defined ranges of the SMD as benchmark of the unbound layers
response as either stress softening behavior (SMD < −20), stress stiffening
behavior (SMD > 20) or linear elastic behavior (−20 to 20). In addition, the con-
cept of “kick” value was introduced in this study to perform further analyses. This
value was defined as the difference between the surface modulus at 600 mm and
that at 1800 mm. In theory, if the “kick” value was similar to the SMD the linear
elastic behavior in the surface modulus plot can be confirmed, if the “kick” value
was significantly different from the SMD a non-linear behavior would be plausible.

2.1 Multi-layer Elastic Modeling

Layer moduli were simulated using the backcalculated results exhibit in Table 1 for
all four sections. The Multi-layer Elastic software PitraPave developed by
LanammeUCR was used to estimate surface deflections. Figure 4 exhibits the
computed deflections for each section and the associated surface modulus.

Deflection Location, mm Surface Modulus, MPa

0 500 1000 1500 2000 0 100 200 300 400 500 600 700
0 0
Deflection Location, mm

200
200
400 40 kN AC1
Deflection, µm

40 kN AC1
600 60 kN AC1
400
60 kN AC1 800 40 kN AC1 w/CS
600 40 kN AC1 w/CS 1000 60 kN AC1 w/CS
60 kN AC1 w/CS 1200 40 kN AC2
800
40 kN AC2 1400
60 kN AC2
60 kN AC2 1600
1000 40 kN AC2 w/CS
40 kN AC2 w/CS 1800
60 kN AC2 w/CS
1200 60 kN AC2 w/CS 2000

Deflection Location, mm Surface Modulus, MPa

0 500 1000 1500 2000 0 100 200 300 400 500 600 700
0 0
Deflection Location, mm

200
200
400 40 kN AC4
Deflection, µm

600 60 kN AC4
400 40 kN AC4
800 40 kN AC4 w/CS
60 kN AC4
600 1000 60 kN AC4 w/CS
40 kN AC4 w/CS
60 kN AC4 w/CS 1200 40 kN AC3
800
40 kN AC3 1400
60 kN AC3
60 kN AC3 1600
1000 40 kN AC3 w/CS
40 kN AC3 w/CS 1800
60 kN AC3 w/CS
1200 60 kN AC3 w/CS 2000

Fig. 4 Computed deflection basins and surface moduli

118 F. Leiva-Villacorta et al.

As expected, the section with the lowest structural capacity (AC2) presented the
highest overall deflection and the section with the highest structural capacity (AC4)
presented the lowest overall deflection. It was also expected to obtain the same
surface modulus at any load level and to reflected an elastic linear behavior of the
lower layers. However, the inclusion of the 40 cm concrete slab not only produced
lower deflections but also produced a “kick” in the surface modulus plot which
tended to be inversely proportional to the structural capacity of the pavement
section. As expected, lower surface moduli near the center of the applied load
indicated a lower supporting capacity of the upper layers.
The surface modulus differential (SMD) and the “kick” parameter were calcu-
lated for each section under the semi-infinite subgrade structure and the inclusion of
the concrete slab. These results along with the maximum deflection (D0) and
surface modulus at two different locations are shown in Table 2. The surface
modulus at D0 represents the structural capacity of the upper layers while the SM at
D1800, if a linear elastic behavior is observed, represents the modulus of the
subgrade. This was the case for the semi-infinite subgrade simulated structure
which yielded surface moduli around the assumed modulus of the subgrade
(70 MPa). Moreover, when following the SMD criteria it was observed that all the
sections but AC4 reflected a linear elastic behavior. In this case, section AC4
presented a stress stiffening behavior. The SMD and “kick” values were identical
and both followed the same trend for the semi-infinite subgrade structure while a
significant difference between these parameters was observed for the structure with
the concrete slab. Based on the SMD criteria all sections but AC4 followed a
non-linear behavior (stress softening). In this case, AC4 presented a linear elastic
behavior based on the SMD concept. However, the “kick” value indicates that all
sections presented a non-linear behavior (stress softening) as observed in Fig. 4.

2.2 Finite Element Modeling

Mechanistic analysis was performed using a non-linear finite element program

called MichPave (Harichandran et al. 1990). The mechanistic analyses performed
by MichPave includes the effect of gravity and lateral stresses arising from the
weight of the materials. Displacements, stresses and strains due to a single circular
wheel load are computed. Due to the assumptions used, the problem is reduced to
an axisymmetric one. Layer elastic moduli were simulated using the backcalculated
results exhibit in Table 1 for all four sections for the bounded layers. The resilient
modulus for the granular base was specified in terms of the bulk stress through the
k-θ model (Eq. 3) and the resilient modulus for the cohesive soil was specified in
terms of the deviatoric stress through the bilinear model (Eq. 4).
Table 2 Surface modulus analysis under MLE modeling
Section Semi-infinite subgrade With concrete slab
D0 SM@D0 SM@D1800 SMD Kick D0 SM@D0 SM@D1800 SMD Kick
(mm) (MPa) (MPa) (MPa) (MPa) (mm) (MPa) (MPa) (MPa) (MPa)
AC1 424.0 351.3 70.4 12.5 11.0 350.8 424.7 248.9 −20.7 −130.2
AC2 749.5 198.8 73.2 6.4 5.5 676.2 220.3 285.5 −41.2 −172.4
AC3 484.7 307.4 72.2 7.5 7.1 413.9 359.9 255.0 −34.5 −142.3
AC4 314.6 473.6 68.3 23.0 25.4 245.0 608.1 208.8 2.3 −65.8
Evaluating Nonlinearity on Granular Materials and Soils Through …
119
120 F. Leiva-Villacorta et al.

MR ¼ 11:795  ðhÞ0:342 ð3Þ

where,
MR = Resilient Modulus (MPa)
θ= Bulk Stress (kPa)

MR ¼ 82:5 þ 0:310½48  rd  when ðrd  48Þ

MR ¼ 82:5  0:192½rd  48 when ðrd [ 48Þ ð4Þ

where,
MR = Resilient Modulus (MPa)
σd = Deviatoric Stress (kPa).
Figure 5 exhibits the computed deflections at a 40 kN load for each section and
the associated surface modulus. The level of deflections and form of the surface
modulus plots were consistent with the previous analyses. It was also expected to
observe a non-linear trend of the simulated pavement structures with the
semi-infinite subgrade layer. However, the plots seemed to reflect a linear elastic
behavior despite the use of unbound material properties in the model. The inclusion
of the 40 cm concrete slab not only produced lower deflections but also produced
the “kick” in the surface modulus plot which tended to be inversely proportional to
the structural capacity of the pavement section as well.
The results of the deflection and surface modulus analysis are shown in Table 3.
In the case of the semi-infinite subgrade simulated structure, the simulation yielded
surface moduli around the assumed modulus of the subgrade (70 MPa) for sections
AC1 and AC4 and slightly higher values for sections AC2 and AC3. When fol-
lowing the SMD criteria it was observed that all the sections reflected a linear
elastic behavior. The SMD and “kick” values were slightly different for sections
AC2 and AC3 suggesting a low effect of the non-linear behavior of the unbound
layers. Once again, a significant difference between these parameters was observed
for the structure with the concrete slab. Based on the SMD criteria all sections but
AC4 followed a non-linear behavior (stress softening). In this case, AC4 presented

Deflection Location, mm Surface Modulus, MPa

0 500 1000 1500 2000 0 100 200 300 400 500 600 700
0 0
Deflection Location, mm

100 200
AC1
400 AC1 w/CS
200
Deflection, µm

AC1 600 AC2

300 800
AC1 w/CS AC2 w/CS
400 AC2 1000 AC3
500 AC2 w/CS 1200 AC3 w/CS
AC3 1400 AC4
600
AC3 w/CS 1600 AC4 w/CS
700 AC4 1800
800 AC4 w/CS 2000

Fig. 5 Computed deflection basins and surface moduli @ 40 kN load

Table 3 Surface modulus analysis under FEA modeling
Section Semi-infinite subgrade With concrete slab
D0 SM@D0 SM@D1800 SMD Kick D0 SM@D0 SM@D1800 SMD Kick
(mm) (MPa) (MPa) (MPa) (MPa) (mm) (MPa) (MPa) (MPa) (MPa)
40 kN 401.3 371.3 74.4 9.7 5.4 331.0 450.1 281.9 −25.7 −168.5
AC1
60 kN 595.5 375.2 74.3 10.2 6.2 490.5 455.6 280.0 −24.9 −165.3
AC1
40 kN 743.8 200.3 90.6 −12.0 −16.5 732.3 203.4 447.4 −94.2 −361.8
AC2
60 kN 1112.0 200.9 80.2 −11.8 −14.9 1002.5 222.9 437.7 −89.4 −350.4
AC2
40 kN 579.5 257.1 80.8 −8.1 −13.9 509.0 292.7 412.0 −71.6 −322.8
AC3
60 kN 844.3 264.7 80.5 −7.3 −12.9 738.3 302.7 402.9 −69.5 −312.4
AC3
Evaluating Nonlinearity on Granular Materials and Soils Through …

40 kN 322.3 462.3 72.9 17.8 17.0 255.5 583.0 239.0 −8.8 −107.4
AC4
60 kN 480.5 465.0 72.8 18.2 17.6 380.8 586.9 238.0 −8.2 −105.5
AC4
121
122 F. Leiva-Villacorta et al.

a linear elastic behavior based on the SMD concept. However, the “kick” value
indicated that all sections presented a non-linear behavior (stress softening). In the
case of MLE modeling, the estimated surface moduli were the same for the two
load levels. However, in this case a small change in the estimated values were
obtained due to the non-linear response of the unbound layers to different stress
states.

2.3 Linear Visco-Elastic Modeling

A single circular load with uniform pressure distribution was modeled using 3D-
Move Analysis_V2 (Asphalt Research Consortium 2014). The tool accounts for
moving traffic loads with complex contact stress distributions of any shape, vehicle
speed, and viscoelastic properties of asphalt concrete layers to calculate pavement
responses using a continuum-based finite-layer approach. A dynamic analysis
(10 kph) was performed using viscoelastic properties for the asphalt concrete layer
obtained from dynamic modulus test results. For the remaining materials, only a
linear elastic behavior can be modeled and the respective layer moduli are shown on
Table 1. The dynamic modulus master curve is shown in Eq. 5. The laboratory
results were obtained following AASHTO TP 79-11.

3:526
logjEj ¼ 0:1787 þ ð5Þ
1þe 1:444 þ 0:5139flog x þ 19:14714
197674
½ðT1 ÞðT1r Þg

where,
E* = Dynamic modulus (ksi)
ω = Frequency (Hz)
T = Analysis temperature (296.15 K)
Tr = Reference temperature (294.15 K).
Figure 6 exhibits the computed deflections at a 40 kN load for each section and
the associated surface modulus. It was expected to observe a linear elastic trend of

Deflection Location, mm Surface Modulus, MPa

0 500 1000 1500 2000 0 100 200 300 400 500 600 700
0 0
Deflection Location, mm

100 200
AC1
200 400 AC1 w/CS
Deflection, µm

AC1 600 AC2

300
AC1 w/CS 800 AC2 w/CS
400
AC2 1000 AC3
500
AC2 w/CS 1200 AC3 w/CS
600
AC3 1400 AC4
700 AC3 w/CS 1600 AC4 w/CS
800 AC4 1800
900 AC4 w/CS 2000

Fig. 6 Computed deflection basins and surface moduli @ 40 kN load

Table 4 Surface modulus analysis under VEA modeling
Section Semi-infinite subgrade With concrete slab
D0 SM@D0 SM@D1800 SMD Kick D0 SM@D0 SM@D1800 SMD Kick
(mm) (MPa) (MPa) (MPa) (MPa) (mm) (MPa) (MPa) (MPa) (MPa)
40 kN 449.4 331.5 102.9 0.6 −27.8 403.1 369.6 245.2 −21.3 −148.5
AC1
60 kN 671.8 332.6 108.8 0.6 −31.1 604.6 369.6 249.7 −22.9 −153.0
AC1
40 kN 820.0 181.7 130.7 −15.1 −57.3 774.0 192.5 428.8 −59.7 −335.1
AC2
60 kN 1270.4 175.9 121.4 −13.2 −51.0 1161.2 192.4 428.8 −61.4 −336.8
AC2
40 kN 567.7 262.4 106.6 −8.3 −39.8 500.3 297.7 257.8 −36.3 −173.3
AC3
60 kN 811.8 275.3 114.9 −9.5 −44.5 750.5 297.7 257.8 −36.3 −173.3
AC3
Evaluating Nonlinearity on Granular Materials and Soils Through …

40 kN 360.0 413.8 85.9 11.9 −4.3 296.2 502.9 151.5 3.8 −43.8
AC4
60 kN 540.0 413.8 85.9 11.9 −4.3 444.3 502.9 151.5 3.8 −43.8
AC4
123
124 F. Leiva-Villacorta et al.

the simulated pavement structures with the semi-infinite subgrade layer. However,
the plots seemed to reflect a low intensity non-linear behavior despite the use of
linear elastic modulus for the unbound layers. The inclusion of the 40 cm concrete
slab not only produced lower deflections but also produced the “kick” in the surface
modulus plot which tended to be inversely proportional to the structural capacity of
the pavement section as well.
The results of the deflection and surface modulus analysis are shown in Table 4.
In the case of the semi-infinite subgrade simulated structure, the simulation yielded
surface moduli higher than the assumed modulus of the subgrade (70 MPa) for all
sections. When following the SMD criteria it was observed that all the sections
reflected a linear elastic behavior. The SMD and “kick” values were significantly
different for all sections but AC4. These results suggested a non-linear behavior of
the unbound layers. Once again, a significant difference between these parameters
was observed for the structure with the concrete slab. Based on the SMD criteria all
sections but AC4 followed a non-linear behavior (stress softening). AC4 presented
a linear elastic behavior based on the SMD concept. However, the “kick” value
indicated that all sections presented a non-linear behavior (stress softening). In this
case, small changes in the estimated values were obtained between load levels due
to the viscoelastic response of the asphalt concrete layer to different stress/strain
states for all sections excluding AC4.

3 FWD Test Results

FWD testing was performed on each section, at three different locations, prior to the
application of the accelerated loading. Deflection basins were obtained at three
different load levels, approximately 40, 53 and 70 kN. Figure 7 exhibits the average
of the three measured deflection basins for each section and the associated surface
modulus. The surface modulus plots seemed to reflect a linear elastic behavior
despite the use of unbound material with non-linear properties, despite the existence
of the concrete slab and the apparent non-linear behavior induced by viscoelastic
properties of the asphalt concrete.
The results of the deflection and surface modulus analysis are shown in Table 5.
When following the SMD criteria it was observed that all the sections reflected a
linear elastic behavior; however, section AC4 followed a non-linear behavior (stress
stiffening). The SMD and “kick” values were very similar in all cases and both
followed the same trend thus confirming that the FWD was not able to report the
expected non-linear behavior.
Evaluating Nonlinearity on Granular Materials and Soils Through … 125

Deflection Location, mm Surface Modulus, MPa

0 500 1000 1500 2000 0 100 200 300 400 500 600
0 0

Deflection Location, mm
200 200
AC1 L1
400
Deflection, µm

400 AC1 L1 AC1 L2

600
600 AC1 L2 800 AC1 L3
800 AC1 L3 1000 AC2 L1

1000 AC2 L1 1200 AC2 L2

1400 AC2 L3
1200 AC2 L2
1600
1400 AC2 L3
1800
1600 2000

Deflection Location, mm Surface Modulus, MPa

0 500 1000 1500 2000 0 200 400 600 800
0 0

Deflection Location, mm
100 200
AC3 L1
200 400
Deflection, µm

AC3 L1 600 AC3 L2

300
AC3 L2 800 AC3 L3
400
AC3 L3 1000 AC4 L3
500
AC4 L1 1200 AC4 L2
600
1400 AC2 L3
AC4 L2
700 1600
800 AC4 L3
1800
900 2000

Fig. 7 Measured deflection basins and surface moduli from FWD results

Table 5 Surface modulus analysis for FWD results

Section Load (kN) D0 (mm) SM@D0 (MPa) SM@D1800 (MPa) SMD (MPa) Kick (MPa)
AC1 40.0 294.3 511.2 122.7 16.0 16.9
AC1 53.5 421.6 476.4 116.0 13.4 15.7
AC1 70.1 593.0 444.2 111.0 9.9 10.6
AC2 40.1 846.4 176.7 83.9 3.5 5.8
AC2 52.8 1123.0 175.3 82.6 2.8 4.0
AC2 67.5 1481.7 169.7 80.7 1.2 1.5
AC3 39.9 451.9 330.4 104.6 −8.5 −5.4
AC3 51.6 604.8 318.6 102.6 −8.2 −5.9
AC3 68.1 833.2 305.2 100.0 −10.2 −8.3
AC4 40.2 199.4 755.9 139.2 37.9 47.2
AC4 53.1 289.2 690.6 131.6 30.5 38.6
AC4 70.4 412.3 642.1 130.6 24.1 28.5

4 Multiple Depth Deflectometer and Road Surface

Deflectometer Results

MDD sensors located at the surface of the pavement were used to measure
deflection basins at two load levels (40 and 60 kN) prior to the actual load testing.
RSD testing was performed on each section, at two different locations and at 40 kN,
126 F. Leiva-Villacorta et al.

Deflection Location, mm Surface Modulus, MPa

0 500 1000 1500 2000 0 200 400 600 800
0 0
200

Deflection Location, m
100
40 kN AC1
200 400
60 kN AC1
Deflection, µm

40 kN AC1
600 40 kN AC2
300 60 kN AC1
800 60 kN AC2
400 40 kN AC2
1000 40 kN AC3
500 60 kN AC2
60 kN AC3
40 kN AC3 1200
600 40 kN AC4
60 kN AC3 1400
60 kN AC4
700 40 kN AC4 1600
800 60 kN AC4 1800
900 2000

Multiple Depth Deflectometer

Deflection Location, m Surface Modulus, MPa
0 500 1000 1500 2000 0 200 400 600 800 1000
0 0
100 200

Deflection Location, m
200 400 AC1 L1
Deflection, µm

AC1 L1 AC1 L2
300 600
AC1 L2 AC2 L1
400 800
AC2 L1 AC2 L2
500 AC2 L2 1000
AC3 L1
600 AC3 L1 1200 AC3 L2
700 AC3 L2 1400 AC4 L1
800 AC4 L1 1600 AC4 L2
900 AC4 L2 1800
1000 2000

Road Surface Deflectometer

Fig. 8 Measured deflection basins and surface moduli from MDD and RSD

also prior to the application of the accelerated loading. Figure 8 exhibits the average
measured deflection basins from three replicates for each section and the associated
surface modulus. Both devices captured, as expected, nonlinear behavior of the
granular materials as well as the subgrade. However, the RSD was able to report the
expected non-linear behavior in a lower intensity compared to the MDDs.
The results of the deflection and surface modulus analysis are shown in Table 6.
In the case of the MDD results, a small change in the surface moduli between load
levels were obtained. This can be explained by the non-linear response of the
unbound layers to different stress states and testing variability. Based on the SMD
criteria, it was observed that all the sections reflected a linear elastic behavior.
The SMD and “kick” values suggested a strong non-linear behavior of the unbound
layers for sections AC1, AC2 and AC3. In the case of section AC4 the SMD criteria
indicated a linear elastic response but the “kick” value and the surface modulus plot
indicated a non-linear response.
A significant difference in the measured deflections were obtained for the RSD
on sections AC1 and AC4 due to construction variability. Overall, the SMD and
“kick” values suggested a mild non-linear behavior of the unbound layers for
sections AC2 and AC3 and a mixed linear-elastic/non-linear behavior for section
AC1.
Evaluating Nonlinearity on Granular Materials and Soils Through … 127

Table 6 Surface modulus analysis for MDD and RSD results

Section D0 (mm) SM@D0 (MPa) SM@D1800 (MPa) SMD (MPa) Kick (MPa)
MDD
40 kN AC1 257.3 579.1 263.8 −35.4 −125.3
60 kN AC1 374.6 596.5 293.2 −35.6 −150.1
40 kN AC2 681.6 218.5 594.8 −86.4 −445.5
60 kN AC2 810.3 275.8 393.1 −74.2 −286.7
40 kN AC3 438.8 339.5 450.9 −102.1 −357.4
60 kN AC3 630.6 354.4 256.9 −45.7 −168.0
40 kN AC4 210.7 707.0 217.6 −5.3 −64.0
60 kN AC4 352.5 634.0 173.8 2.8 −44.9
RSD
AC1 L1 354.0 420.8 172.1 −14.8 −47.9
AC1 L2 226.7 657.1 149.1 2.2 −8.9
AC2 L1 894.0 166.6 132.0 −32.4 −44.7
AC2 L2 800.1 186.2 106.7 4.1 −23.2
AC3 L1 506.6 294.0 223.1 −52.0 −141.6
AC3 L2 619.0 240.7 152.1 −24.6 −78.6
AC4 L1 278.4 535.0 145.2 −1.7 −22.7
AC4 L2 164.6 904.9 193.4 10.7 −10.2

5 Summary and Comparison of Results

A summary of the SMD values obtained from measurements and mechanistic

analysis and the tendency with respect to the linear or non-linear behavior can be
observed in Fig. 9. The linear elastic behavior was set between −20 and 20 based
on the SMD criteria, below −20 and above 20 was defined the non-linear behavior

40
Stress Stiffening
20
Linear Elastic
0
AC1 AC2 AC3 AC4
SMD, MPa

-20

-40
Stress Softening
-60

-80

-100

-120 FWD MDD RSD1 RSD2 MLE MLE CS FEA FEA CS VEA VEA CS

Fig. 9 Summary of SMD results

128 F. Leiva-Villacorta et al.

region. In general, FWD results reflect an elastic linear behavior for sections AC1,
AC2 and AC3 and reflect a stress stiffening behavior for section AC4. In agreement
with this are the results from MLE considering a semi-infinite subgrade layer. MDD
results reflect strong non-linear behavior for sections AC1, AC2 and AC3 and in
agreement with this are the results from FEA when considering the concrete slab
located at 2.8 m deep. RSD results reflect a mild non-linear behavior for sections
AC2 and AC3 and in agreement with this are the results from MLE and VEA when
considering the concrete slab located at 2.8 m deep. Finally, based on both, mea-
sured and predicted responses, it was determined that sections AC1 and AC4
presented mostly a linear elastic behavior and sections AC2 and AC3 followed,
mostly a non-linear elastic behavior (stress softening) which was augmented by the
presence of the concrete slab.

6 Conclusions

• The results observed with the MLE modeling suggested that the “kick” in a
surface modulus plot could be caused by the shallow rigid layer (40 cm concrete
slab) providing an apparent non-linear behavior.
• The results observed with the FEA modeling suggested that the “kick” in a
surface modulus plot could be caused by the non-linear (stress
softening/stiffening) behavior of the unbound materials. This non-linear
behavior was intensified by the inclusion of the concrete slab.
• The results observed with the VEA modeling suggested that the “kick” in a
surface modulus plot could be caused not only by a shallow rigid layer (bed-
rock) or nonlinear (stress softening/stiffening) materials but also by the vis-
coelastic behavior of the asphalt concrete layer. This non-linear behavior was
also intensified by the inclusion of the concrete slab.
• The results observed with the modeling of the 40 cm concrete slab located at a
depth of 2.8 m indicated a significant change in the deflection basins and the
surface modulus plots for all sections. This results suggest that a non-linear
behavior should be observed when performing non-destructive testing on the
evaluated pavement sections.
• FWD results consistently reflect an elastic linear behavior of the lower layers,
while RSD and MDD reflect a more intensified non-linear behavior of the
lowers layers and could also exhibit the presence of the test pit concrete slab
(shallow rigid layer).
• Finally, the new concept “kick” value defined as the difference between the
surface modulus at 600 mm and that at 1,800 mm can be used as complement of
the SMD value in order to discriminate between linear and non-linear behavior.
Evaluating Nonlinearity on Granular Materials and Soils Through … 129

References

Aguiar-Moya, J. P., Corrales, J. P., Elizondo, F., & Loría-Salazar, L. (2012). PaveLab and heavy
vehicle simulator implementation at the National Laboratory of Materials and Testing Models
of the University of Costa Rica. In Advances in pavement design through full-scale accelerated
pavement testing, APT 2012.
Asphalt Research Consortium webpage. http://www.arc.unr.edu/Software.html. Date Accessed
03-20-2014.
Baker Harris, B., Buth Michael, R., & Van Deusen David, A. (1994). Minnesota road research
project: Load response instrumentation installation and testing procedures. Minnesota
Department of Transportation.
Coetzee, N., et al. (2008). The heavy vehicle simulator in accelerated pavement testing: Historical
overview and new developments. In 3rd International Conference APT.
Harichandran, R. S., Yeh, M.-S., & Baladi, G. Y. (1990). MICH-PAVE: A nonlinear finite element
program for the analysis of flexible pavements. Transportation Research Record, 1286,
123–131.
Horak, E. (2008). Benchmarking the structural condition of flexible pavements with deflection
bowl parameters. Journal of the South African Institution of Civil Engineering, 520(2), 2–9.
Irwin, L. H. (2002). Backcalculation: An overview and perspective. In Proceedings of the
Pavement Evaluation Conference. Roanoke, VA.
Leiva-Villacorta, F., Aguiar-Moya, J. P., & Loria-Salazar, L. G. (2013). Ensayos acelerados de
pavimento en Costa Rica. Infraestructura Vial, 15(26), 32–41.
Leiva-Villacorta, F., Aguiar-Moya, J. P., & Loría-Salazar, L. G. (2015). Accelerated pavement
testing first results at the Lanammeucr APT facility. In Transportation Research Board 94th
Annual Meeting, Washington DC.
Ullidtz, P. (1987). Pavement analysis, development in civil engineering (Vol. 19). Amsterdam:
Elsevier.
Ullidtz, P. (2005). Modelling flexible pavement response and performance. New York: Elsevier.
ISBN 8750208055.
Full Scale Accelerated Pavement Tests
to Evaluate the Performance of Permeable
and Skeletal Soil Block Pavement Systems

Abubeker Ahmed, Fredrik Hellman and Sigurdur Erlingsson

Abstract The increasing proportion of paved surface due to urbanization means

that the conditions for urban trees and vegetation to survive have deteriorated.
Factors such as air pollution, poor drainage, and the lack of usable soil for root
growth contribute to the short life expectancy of urban trees. To meet this chal-
lenge, several permeable and “structural” or “skeletal soils” have been developed as
alternatives to the typical compacted soil required to bear the weight of vehicular
traffic in urban areas. The main objective of this study is to evaluate the resistance to
permanent deformation of permeable and skeletal soil pavement structures based on
full scale accelerated pavement tests (APT) using a heavy vehicle simulator (HVS).
Interlocking paving stones of various types were used as permeable surface layer
for the test structures. The results demonstrated that the permeable test structures
exhibited higher permanent deformation than the corresponding impervious struc-
tures. The skeletal soil with bituminous base layer, however, produced performance
comparable to the impervious reference test structures.

Keywords Heavy vehicle simulator
Permeable pavement
Skeletal soil
Block

pavement

1 Introduction

The proportion of impervious land surface is steadily increasing due to densification

of buildings associated with urbanization. This leads to poorer local climate and
poorer growing conditions for trees and vegetation. In addition, the climate change

A. Ahmed (&)
F. Hellman
S. Erlingsson

Pavement Technology, National Road and Transport Research Institute,
VTI, Linköping, Sweden
e-mail: [email protected]
S. Erlingsson
Faculty of Civil and Environmental Engineering,
University of Iceland, Reykjavik, Iceland

© Springer International Publishing Switzerland 2016 131

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI: 10.1007/978-3-319-42797-3_9
132 A. Ahmed et al.

likely leads to more intense rainfall and increased annual precipitation.

Consequently, due to the increasing portion of paved surfaces without drainage
properties, existing sewerage system will be overloaded due to large amounts of
surface water during intense rainfall. Therefore, the surface water must be drained
to protect urban area from flooding.
Furthermore, increased urbanization or paved surface means that the conditions
for urban trees and vegetation to survive have deteriorated (Hellman 2014). Several
factors such as air pollution, poor drainage, and damage by people or vehicles
contribute to the short life expectancy of urban trees. However, the most limiting
factor in the growth of urban trees is the lack of usable soil for root growth (Craul
1992). This is due to the fact that the subgrade and base layers of a pavement
sections must be compacted to meet engineers’ specifications for load-bearing. This
often leads to soils that are too dense for root penetration.
Therefore, innovative approaches to build pavement systems are needed to
improve the local climate and to address climate change while fulfilling their
economical, structural, and functional requirements. This imposes high require-
ments on the design and performance of pavements because, in many cases,
pavements structures in urban areas are subjected to heavy and repeated traffic loads
such as at bus stops, roundabouts, and traffic lights. This places huge demands on
materials, design, and construction.
To meet this challenge, several permeable and “structural” or “skeletal soils”
have been proposed as alternatives to the typical compacted soil required to bear the
weight of vehicular traffic in urban areas. Permeable pavements are consisted of
engineered base and subbase materials that allow the movement of storm water
through the surface. Skeletal soils, on the other hand, have been developed to
improve street tree establishment and growth in paved areas. Skeletal soils are
designed planting media called structural soils or stone matrix soils (Bassuk et al.
1998; Gamstetter 1998; Kristofferson 1998; Grabosky 1999, 2001; Buckstrup and
Bassuk 2000). When designed correctly, skeletal soils form a rigid stone matrix
with large pores to allow root growth without sacrificing the load-bearing capacity
of the soil. The soil found between the stones is non-compacted, leaving room for
air exchange, water holding and root growth.
Several field and accelerated pavement tests have showed that permanent
deformation along the wheel path, also known as rutting, is one of the main distress
types affecting the performance of pavement, whether asphalt, cement concrete or
block pavements (Clifford 1984).
This study has conducted full-scale accelerated pavement tests (APT) using a
heavy vehicle simulator (HVS) to study the rutting resistance of two variants of
permeable pavements (with 10 and 20 % infiltration levels, respectively), two
variants of impervious pavements and two variants of skeletal soil pavement
structures. Interlocking paving stones of types Sienna Eco, Uni-Ecoloc, and granite
stone blocks were used as surface layer of the test structures.
Full Scale Accelerated Pavement Tests … 133

2 Test Structures

Full Scale Accelerated Pavement Tests (APT) using a Heavy Vehicle Simulator
(HVS) have been conducted to measure the development of permanent deforma-
tion. The tests were performed using a 3 m deep test pit testing facility at the
Swedish Road and Transport Research Institute, Sweden (Wiman and Erlingsson
2008; Saevarsdottir et al. 2014).
Three APTs, each with two adjacent test structures, were conducted using the
HVS to evaluate the rutting performance of various block pavements systems. The
first APT was performed on test structures with 20 and 10 % infiltration levels,
respectively. The structures consisted of concrete pavers over a bedding permeable
sand, crushed rock base course of maximum aggregate size of 32 mm, crushed rock
subbase with maximum aggregate size of 90 mm and fine sand subgrade layer over
a rigid bottom. The surface layers were made of concrete pavers of type Sienna Eco
for 20 % infiltration and Uni-Ecoloc for 10 % infiltration level.
The second APT test was carried out on two reference impervious test structures
(structures 3 and 4) made of Uni-Ecoloc concrete pavers and granite stone block
pavers, respectively. The material used as bedding sand, base course layer and
subbase are similar to the test structures 1 and 2 but with differing grain size
distribution so as to make the structures impermeable.
The third APT test was conducted to study the performance of impermeable
pavement systems with skeletal soil structure, Structure 5 and Structure 6. In this
study the skeletal soil layer was placed under the subbase layer and made of very
large rocks having size of between 100 and 150 mm. Two variants of the skeletal
soil pavements were tested in this study, one with crushed rock base course layer
and the other with bituminous base layer. Figures 1, 2 and 3 show the cross sections

Structure 1 Structure 2
0 0 Sienna-Eco
Uni-Ecoloc
100 100
Bedding permeable sand 2/5
145 145
Crushed rock base 4/32

Crushed rock subbase 4/90

Fine sand subgrade

373 373

1058 1058

z (mm) z (mm)

Fig. 1 Test structures 1 and 2 with permeability of 20 and 10 %, respectively

134 A. Ahmed et al.

Structure 3 Structure 4
0 0
Natural stone

100 80 Uni-Ecoloc

188 188 Bedding sand 0/4

Crushed rock base 0/32

Crushed rock subbase 0/90

Fine sand subgrade

393 393

1059 1059

z (mm) z (mm)

Fig. 2 Reference test structures 3 (Uni-Ecoloc pavers) and 4 (natural stone pavers), reference
structures

Structure 5 Structure 6
0 0
Uni-Ecoloc

100 100 Washedbeddingsand0/4

130 135 Asphalt concrete

Crushed rock base 0/32

248 214
Crushed rock subbase 32/64

Skeletal soil 100/150

403 402
Geotextile
Fine sand subgrade
979 978

z (mm) z (mm)

Fig. 3 Test structures made of skeletal soils (structures 5 and 6)

of the test structures. Note that for skeletal soil structures, the water and air which
are needed for the root growth of the trees are provided through establishments
surrounding the trees, such as through wells.
A 160/200 penetration grade bitumen with maximum aggregate size of 16 mm
was used as a bituminous base layer for the skeletal soil with bituminous base layer
(Structure 6).
Full Scale Accelerated Pavement Tests … 135

(a) (b)

(c) (d)

(e) (f)

Fig. 4 a Test structure 1 with permeability of 20 %; b test structure 2 with permeability of 10 %;

c a closer view of Sienna-Eco blocks, d a closer view of Uni-Ecoloc pavers; e impervious
reference test structure 3 using Uni-Ecoloc pavers and f impervious test structure 4 using granite
stone blocks

Figure 4a, b demonstrate the finished surface of the test structures with
Sienna-Eco and Uni-Ecoloc pavers which are shown in closer view in Fig. 4c, d. The
surface layers of the impervious reference test structures are shown in Fig. 4e, f.
Figure 5 shows the grain size distributions of the bedding sand, crushed rocked base,
and subbase course layers for the permeable structures.
136 A. Ahmed et al.

100
Sand 2/5
80 Crushed rock base 4/32
Precent passing

Crushed rock subbase 4/90

60

40

20

0
0.01 0.1 1 10 100
Sieve size (mm)

Fig. 5 Grain size distribution curves for the bedding sand, crushed base course and crushed
subbase course layers of the permeable structures 1 and 2

2.1 Test Programs

2.1.1 Accelerated Pavement Tests (APT)

The test structures were subjected to wheel loads of magnitude 30, 40, and 60 kN
using a Heavy Vehicle Simulator (HVS). The 30 kN wheel load was applied for
preloading of 500 passes. The 40 and 60 kN wheel loads were used to study the
permanent deformation behaviour of the test structures. A total of 140,000 load
cycles were applied for the permanent deformation tests. The testes were started
with a wheel load of 40 kN and load passes of 26,000. Then the wheel load was
increased to 60 kN and load cycles of about 48,000 were applied. After a total of
74,000 load cycles, the ground water level was raised to a level 30 cm below the
top surface of the subgrade and a wheel load of 60 kN was applied for another
70,000 load cycles. The permanent deformation on the surface of the test structures
was measured with the help of laser beam mounted on the edges of the test
structures. Table 1 summarizes the load levels, tire pressure, speed, and ground
water table adopted for the tests.
Falling weigh deflectometer (FWD) tests were carried out at various stages of the
constructions of the test structures. Surface FWD tests were also conducted before
and after the HVS tests to evaluate the effect of compaction due to wheel loading.
Furthermore, the relative hydraulic conductivities of the permeable structures
(Test structures 1 and 2) were measured according to the European standard, EN
12697-40: 2012. The method gives a relative permeability level of the test struc-
tures. The measurement principle is to measure the time it takes to drain 4 L of
water through a cylinder which is placed on the surface of the test object. In the
Full Scale Accelerated Pavement Tests … 137

Table 1 Test parameters

Number of Wheel load (dual) Ground water table Speed Tire pressure
passes (kN) (km/h) (kPa)
0–500 30 Natural 12 800
500–26,200 40 Natural 12 800
26,200– 60 Natural 12 800
73,768
73,768– 60 30 cm below the top of the 12 800
140,000 subgrade
140,000– 80 30 cm below the top of the 12 800
160,000 subgrade

experiment, the cylinder was placed across the joints of the pavers (Sienna Eco and
Uni-Ecoloc) with the drain hole placed at the middle of the cylinder. Ten different
permeability measurements were conducted on each surface.

3 Results and Discussions

The subsequent sections present the APT, FWD, and permeability test results of the
various test pavement structures.

3.1 FWD Tests

Figures 6, 7, and 8 show the FWD measurements of the test structures. The FWD
measures were conducted before and after the HVS test using a plate loads of 30
and 50 kN. The FWD results in the figures represents the average of the FWD
deflection values measured at nine different locations.

Distance (mm) Distance (mm)

(a) 0 400 800 1200
(b)
0 400 800 1200
0 0

200 200
Deflection (µm)

Deflection (µm)

400 Seinna-Eco 400 Uni-Ecoloc

30kN before HVS 30kN before HVS

600 600
30kN after HVS 30kN after HVS

800 50kN before HVS 800 50kN before HVS

(a) 50kN after HVS 50kN after HVS

1000 1000

Fig. 6 FWD measurements of permeable test structures (structures 1 and 2)

138 A. Ahmed et al.

Distance (mm) Distance (mm)

(a) 0 400 800 1200 (b) 0 400 800 1200
0 0

200 200
Deflection (µm)

Deflection (µm)
400 400 Granite stone blocks
Uni-Ecoloc

30kN before HVS 30kN before HVS

600 600
30kN after HVS 30kN after HVS

800 50kN before HVS 800 50kN before HVS

50kN after HVS 50kN after HVS

1000 1000

Fig. 7 FWD measurements of the impervious reference test structures (structures 3 and 4)

Distance (mm) Distance (mm)

(a) 0 400 800 1200 (b) 0 400 800 1200
0 0

200 200
Deflection (µm)

Deflection (µm)

Skeletal soil without Skeletal soil with

400 400
bituminous base bituminous base

600 600 30kN before HVS

30kN before HVS
30kN after HVS 30kN after HVS
800 50kN before HVS 800 50kN before HVS
50kN after HVS 50kN after HVS
1000 1000

Fig. 8 FWD measurements of the skeletal soil test structures (structures 5 and 6)

From the difference between measured FWD deflections before and after the
HVS tests in Figs. 6 and 7, it is evident that the two permeable test structures
exhibited more compaction due to the HVS wheel loading than the corresponding
impervious reference structures. This might be due to the open graded base and
subbase layers of the permeable structures. A similar trend was observed for the
skeletal soil pavement structure (Structure 5). The skeletal soil with bituminous
base layer (Structure 6) demonstrated the least maximum deflection, however, the
effect of compaction due to HVS loading was observed.
Figures 9 and 10 show the hysteresis loops for D0 obtained from the deflection
time history data of the FWD tests conducted on structures 5 and 6. As shown in
the figures, the maximum deflection for Structure 6 is less than Structure 5 which
indicates the positive effect of the bituminous base layer. Examination of the areas
within the hysteresis loops after the HVS tests (i.e., after vehicular compaction)
Full Scale Accelerated Pavement Tests … 139

60 60
(a) (b)
50 50

40 40

Load (kN)
Load (kN)

30 30

20 20

10 Structure 5 30 kN 10 Structure 6 30 kN
Structure 5 50 kN Structure 6 50 kN
0 0
0 200 400 600 800 0 200 400 600 800
D0 (µm) D0 (µm)

Fig. 9 FWD hysteresis for skeletal soil a Structure 5 and b Structure 6 after HVS tests for 30 and
50 kN loads

60

50

40
Load (kN)

30

20 Structure 5 before HVS

Structure 5 after HVS
10 Structure 6 before HVS
Structure 6 after HVS
0
0 200 400 600 800 1000
D0 (µm)

Fig. 10 FWD hysteresis for skeletal soil structures 5 and 6 before and after HVS tests for 50 kN
load

revealed that Structure 6 has slightly greater area (4.56 Nm of dissipated energy)
than Structure 5 (4.50 Nm) this might be due to the viscoelastic nature of the
bituminous base layer. The areas within the hysteresis loops before the HVS test, on
the other hand, revealed a larger area for Structure 5 (6.70 Nm) than Structure 6
(5.62 Nm) which might indicate that energy loss is mainly due to compaction than
viscoelastic effects.
140 A. Ahmed et al.

Load cycles
0 40000 80000 120000 160000
0
2 Uni-Ecoloc - 10% infiltration
Average Rut Depth (mm)

4 Sienna Eco - 20% infiltration

6
Ground water table 30 cm
8 below the top of the subgrade
10

60 kN
40 kN
30 kN

60 kN

12
14
16
Fig. 11 Measured permanent deformation of the permeable test structures (structures 1 and 2)

3.2 HVS Test

Figures 11, 12, and 13 show the measured permanent deformation of the test
structures. As expected the permeable test structures exhibited higher permanent
deformation than the reference test structures. The difference in measured

Load cycles
0 40000 80000 120000 160000
0
60 kN

80 kN

2
Average Rut Depth (mm)

4
Ground water table 30 cm
6 below the top of the subgrade

8
10
12
30 kN
40 kN

60 kN

Uni-Ecoloc
14
Granite Stone Blocks
16
Fig. 12 Measured permanent deformation of the reference impervious structures (structures 3
and 4)
Full Scale Accelerated Pavement Tests … 141

Load cycles
0 40000 80000 120000 160000
0
Skeletal Soil
2
Average Rut Depth (mm)

Skeletal Soil with Bituminous Base

4
6 Ground water table 30 cm
below the top of the subgrade
8
10
12
30 kN

60 kN
40 kN

60 kN
14
16
Fig. 13 Measured permanent deformation of the skeletal soil test structures (structures 5 and 6)

permanent deformation between the two permeable structures is attributed to types

of pavers used, i.e., Sienna-Eco and Uni-Ecoloc that are associated with the 20 and
10 % infiltration levels, respectively, as shown in Fig. 11. The relative hydraulic
conductivities of the permeable structures measured according to the European
standard, EN 12697-40: 2012, are also shown in Table 2. Table 2 shows the
average of 10 measurements of time required to drain 4 L of water through a
cylinder which was placed on the joints of the pavers.
The interlocking nature of the pavers might have also contributed to the dif-
ference in performance of the two permeable structures. The term interlocking
refers the bonding mechanism within the pavement matrix (Clifford 1984) which
might have affected the load transferred to the underlying layers. Furthermore, the
tests conducted on the impervious reference test structures made of Uni-Ecoloc and
natural granite stone pavers demonstrated an improved performance of the
Uni-Ecoloc pavers compared to the natural stone pavers, Fig. 12.
The permanent deformation test of the skeletal soil test pavements, Fig. 13,
showed an improved performance compared to the permeable test structures. Note
that, the skeletal soil test structures were not permeable and the Uni-Ecoloc pavers
were arranged to make impervious surface as in Fig. 4e. The water and air which

Table 2 Relative hydraulic conductivities of the permeable structures

Material Hydraulic conductivity (s−1) Permeability (l/m2/s)
Sienna Eco 0.089 5.2
Uni-Ecoloc 0.063 3.4
142 A. Ahmed et al.

Permanent deformation (mm) 4.5

4.0
after 500 cycles
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Sienna Eco - Uni-Ecoloc - Uni-Ecoloc Granite Stone Skeletal Soil Skeletal Soil
20% 10% Blocks with Asfalt
infiltration infiltration
Test structure
Fig. 14 Initial permanent deformation after 500 load cycles

are needed for the root growth of the trees are obtained through establishments
surrounding the trees, such as through wells. Furthermore, it was demonstrated that
using bituminous base layer in the skeletal soil construction significantly improved
the rutting resistance, indicating that pavements with skeletal soils if designed
correctly produces better bearing capacity while providing large pores to allow root
growth.
Note that, all the test structures exhibited most of the permanent deformation in
the first 1000 load cycles. Figure 14 shows the initial permanent deformation after
the first 500 load cycles, the permeable structures exhibited the most initial per-
manent deformation. The flattening of the curves of rut depth against time shows
improvements in the structural properties occur after the initial settling-in period.
This is not peculiar to block pavements as shown by various APT tests on other
pavement types. The rate of growth of permanent deformation of the test structures
is shown in Fig. 15. The permeable structures exhibited the highest rate followed by
the skeletal soil structures.
Finally, previous APT test have revealed that raising the ground water table
increases the rate of development of rutting (Erlingsson 2010), however, none of
the test structures has shown a clear impact on development of permanent defor-
mation to the change in ground water level that was applied after load cycles of
about 73,000, as shown in Figs. 11, 12, and 13. This is perhaps due to the thick
structures considered in this study, i.e., the water table was located 1.3 m below the
top surface of the test structures thus the influence of the location of the ground
water might be insignificant.
Full Scale Accelerated Pavement Tests … 143

Rate of permanent deformation 0.6

0.5
mm/10000 cycles

0.4

0.3

0.2

0.1

0
Sienna Eco - Uni-Ecoloc - Uni-Ecoloc Granite Stone Skeletal Soil Skeletal Soil
20% 10% Blocks with Asfalt
infiltration infiltration
Test structure
Fig. 15 Rate of permanent deformation in mm per 10,000 load cycles at 60 kN wheel loading

4 Summary and Conclusions

This study has conducted full-scale accelerated pavement tests (APT) using a heavy
vehicle simulator (HVS) to study the rutting resistance of permeable, impervious
and skeletal soil test pavement structures. Interlocking paving stones of types
Sienna Eco, Uni-Ecoloc, and granite stone blocks were used as surface layers of the
test structures. The performance to permanent deformation or rutting of the different
test structures were evaluated using HVS and the Falling weight deflectometer
(FWD) tests. The test results indicated that the permeable structures exhibited lower
resistance to rutting compared to the corresponding impervious reference structures.
The permeable structures exhibited a higher initial permanent deformation which is
mainly due to compaction and showed a higher rate of growth of permanent
deformation after the initial settling-in period. The skeletal soil structures, on the
other hand, showed an improved performance compared to the permeable structures
both in terms of the initial permanent deformation and the rate of growth of per-
manent deformation after the initial compaction. Furthermore, the introduction of a
bituminous base layer under the bedding sand for skeletal soil test structure sig-
nificantly improved the resistance to rutting.
This study was limited to full scale accelerated pavement tests, in addition, the
main focus of the study was the rutting performance of the test structures, therefore
future studies will encompass field performance evaluation of such structures and
other types of distresses that might affect block pavement systems.

Acknowledgments The authors are grateful for the financial support of VINNOVA, the
Sweden’s Innovation Agency, and SBUF, the Construction Industry’s Organisation for Research
144 A. Ahmed et al.

and Development. We would also like to acknowledge the STARKA for providing the concrete
pavers, EMMABODA GRANITE for providing the granite stone blocks and NCC Road AB for
providing the unbound materials and performing the constructions. Other companies that have
been involved in the project are Benders, Stenindustriförbundet, Stockholms Trafikkontor,
Cementa, SP—Technical Research Institute of Sweden, and CBI—Swedish Cement and Concrete
Research Institute.

References

Bassuk, N., Grabosky, J., Trowbridge, P., & Urban, J. (1998). Structural soil: An innovative
medium under pavement that improves street trees. In Proceedings of the ASLA annual
meeting. (pp. 183–185), Portland, OR, October 1998.
Buckstrup, M., & Bassuk, N. (2000). Transplanting success of balled-and-burlapped versus
bare-root trees in the urban landscape. Journal of Arboriculture, 26(6), 298–308.
Clifford, J. M. (1984). A description of “Interlock” and “Lock-up” in block pavements. In Second
International Conference on Concrete Block Paving, Delft, 10–12 April 1984.
Craul, P. J. (1992). Urban soil in landscape design. New York: Wiley.
Erlingsson, S. (2010). Impact of water on the response and performance of pavement structure in
an accelerated test. Road Materials and Pavement Design, 11(4), 863–880.
Gamstetter, D. (1998). Designing the right place for the right tree. Arborist News, 7(3), 9–12.
Grabosky, J. (1999). Growth response of three tree species in sidewalk profiles. Doctoral
dissertation. Ithaca, NY: Cornell University.
Grabosky, J. (2001). Shoot and root growth of three tree species in sidewalk profiles. Journal of
Environmental Horticulture, 19(4), 206–211.
Hellman, F. (2014). Grågröna systemlösningar för hållbara städer: Spårdjupbildning och styvhet i
konstruktioner av betong-och natursten. http://www.greenurbansystems.eu.
Kristofferson, P. (1998). Designing urban pavement sub-bases to support trees. Journal of
Arboriculture, 24(3), 121–126.
Saevarsdottir, T., Erlingsson, S., & Carlsson, H. (2016). Instrumentation and performance
modelling of heavy vehicle simulator tests. International Journal of Pavement Engineering, 17
(2), 148–165. doi:10.1080/10298436.2014.972957.
Wiman, L. G., & Erlingsson, S. (2008). Accelerated pavement testing by HVS—A trans-national
testing equipment. Transport Research Arena Europe, Ljubljana, 21–24 April, CD-ROM.
In-Situ Validation of Three-Dimensional
Pavement Finite Element Models

Osman Erman Gungor, Imad L. Al-Qadi, Angeli Gamez

and Jaime A. Hernandez

Abstract Mechanistic-based pavement design guides gained popularity in the last

decade. Accurate computation of stress, strain, and the deformation field within
pavement, which are used in pavement design guides, is important to realistically
predict pavement performance over the design life. The layered elastic theory and
finite element method (FEM) are commonly used to compute the critical responses
of pavement structure. Although the layered elastic theory analysis is relatively
faster and easier to implement, it hinders real loading and pavement material
characterization. Therefore, FEM has become more attractive to pavement
researchers for investigating pavement behavior under tire loading since the last two
decades. Although several studies have been conducted for developing FE model
for pavement, limited work has been done to validate the finite element (FE) models
using in-situ pavement responses. This study presents an advanced
three-dimensional (3-D) pavement FE model validated using four pavement sec-
tions. Good agreement was noted between the FE analysis results and pavement
field instrument responses to loading, especially for vertical pressures and hori-
zontal tensile strains in the transverse direction. When using proper material
characterization parameters and accurate tire loading, the FE model is capable of
realistically simulating tire-pavement interaction in the field.

Keywords Finite element analysis Validation Flexible pavement Mechanistic

pavement design approach

O.E. Gungor (&)
I.L. Al-Qadi
A. Gamez
J.A. Hernandez

Department of Civil and Environmental Engineering,
University of Illinois at Urbana-Champaign, 205 North Mathews,
Urbana, IL 61801, USA
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 145

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_10
146 O.E. Gungor et al.

1 Introduction

Mechanistic-empirical based pavement design guides that link mechanistic pave-

ment responses to pavement distress prediction gained popularity in the last decade.
Accurate computation of stress, strain, and deformation fields within pavement
plays a crucial role in realistically predicting pavement performance over the design
life. The layered elastic theory (LET), originally developed by Burmister et al.
(1944), is the traditional approach to compute pavement response to vehicular
loading. Although LET is relatively faster and easier to implement, it hinders real
loading conditions and material characterization in pavement analysis.
Finite element method (FEM), on the other hand, considers variables that are
omitted in LET simulations, such as linear viscoelastic characterization of asphalt
concrete (AC), anisotropic stress-dependent material characterization for granular
layers, realistic model for simulating interaction between pavement layers, 3-D
non-uniform contact stresses, in addition to simulating continuously moving tire
loads. Considering those variables in pavement analysis not only results in accurate
computation of pavement responses but also helps researchers get better under-
standing of complex pavement behavior under tire loading. Therefore, FEM
became more attractive to pavement researchers in the past two decades.
A significant number of studies using FEM for analyzing pavement structure
have been conducted. FEM is applied in several ways in pavements, including the
following: Comparing the effect of using wide-base tire and dual-tire assembly
(DTA) on pavement performance (Al-Qadi et al. 2005; Gungor et al. 2014),
understanding the importance of nonlinear characterization of granular material in
pavement response (Kim et al. 2009), characterization of rutting within pavement
structure (Fang et al. 2004), understanding crack initiation and propagation with AC
(Ameri et al. 2011; Underwood et al. 2009) are examples of the main applications
of FEM on pavement analysis.
Although several studies have been conducted to develop pavement FE model
for a variety of purposes as listed above, limited research has been initiated to
validate FE models using in-situ pavement responses. This study presents and
validates an advanced 3-D pavement FE model using four pavement sections. The
following section introduces the developed pavement FE model and explains its
important features. Then, the validation results for each section are provided and
discussed. Finally, the last section presents the conclusion of the paper.

2 Finite Element Model

Realistic simulation of flexible pavement structure under tire load is quite a chal-
lenging task. Other than its geometry, every component of the simulation, such as
loading conditions and material characterization, is complicated. The tire applies
non-uniform, 3-D contact stress on pavement. AC exhibits viscoelastic behavior,
In-Situ Validation of Three-Dimensional Pavement … 147

meaning that stiffness is time, temperature, and frequency of loading dependent and
should be properly modelled in the simulation. Moreover, granular material shows
anisotropic stress-dependent nonlinear behavior. Not only the material’s stiffness
depends on the stress level it is exposed to, but also it behaves differently in each
principal direction (anisotropy). Additionally, modelling interaction between dif-
ferent pavement layers further complicates the behavior of pavement structure. The
literature clarifies the significant effect of these conditions on pavement responses.
Therefore, it is important to capture them while simulating pavement behavior
under tire loading to compute pavement responses accurately.
LET is the current analysis approach used in mechanistic-empirical guides to
obtain pavement responses mostly because of its computational efficiency.
However, unrealistic simplifications and assumptions are associated therewith.
Linear elastic characterization of AC and base materials, spring model assumption
for layer interface, two-dimensional (2-D) uniform tire pressure, and circular con-
tact area are only some examples of the simplifications and assumptions of LET.
FEM, on the other hand, is a promising numerical method that can overcome the
difficulties and challenges in flexible pavement simulation. It has the ability to
consider non-uniformity in loading conditions and complex nonlinear material
characterizations. Since there are a number inputs required for 3-D pavement FEM
that cannot be considered in LET (e.g., temperature, speed) these two pavement
analysis approaches are not comparable. However, comparison of FEM with
Mechanistic-Empirical Pavement Design Guide (AASHTO 2008) that incorporates
LET within its framework is studied as part of another study (Al-Qadi et al. In
Review).
The pavement FE model presented in this paper is the ultimate version of over
ten years of on-going research by Al-Qadi and his coworkers (Wang and Al-Qadi
2010; Elseifi et al. 2006; Yoo et al. 2007; Al-Qadi and Yoo 2007). The key features
of the developed FE model can be categorized into five different groups: geometry
and boundary conditions, loading conditions, material characterization, analysis
method, and interface interaction model. A brief explanation for each key feature is
given in the following sections.

2.1 Finite Element Model and Boundary Conditions

The FE model of flexible pavement structure that was developed in commercial FE

software, ABAQUS, is given in Fig. 1. Three main issues that need to be addressed
when developing pavement FE model: element size, i.e., mesh size (i), depth of the
model (ii) and lateral size of the model (iii). In traditional pavement analysis
approaches, such as LET, the subgrade is assumed to have infinite depth. However,
a finite depth needs to be assigned to subgrade in the FE model. In addition to
finding subgrade depth, the element size should be optimized. FE elements generate
more accurate results as the size of the element gets smaller. On the other hand,
computational time increases as the number of elements in the model increases.
148 O.E. Gungor et al.

Fig. 1 3-D finite element model

Therefore, optimum element size and subgrade thickness are determined by mesh
sensitivity analysis. In order to perform mesh sensitivity analysis, an elastic FE
model was compared to an LET software, known as Bisar, for five different critical
pavement responses: maximum transverse tensile strain at the bottom of AC;
maximum compressive strain within subgrade; and maximum vertical shear strain
within AC, base, and subgrade. The mesh was refined until the difference between
the FE model and Bisar was equal to or less than 5 %.
The pavement response yields haversine like pulse duration as demonstrated in
Fig. 2. As a result of trying different wheel path lengths (1000 and 2000 mm), it
was observed that a path length longer than 2000 mm was required to capture the
entire pulse duration. On the other hand, it was observed that there is no significant
difference in the maximum magnitude of pavement responses obtained from 1000
and 2000 mm of wheel path. Therefore, the wheel path was decided to be kept
around 1000 mm, which is long enough to capture maximum pavement response.

2.2 Loading Conditions

LET assumes 2-D uniform static vertical pressures with circular load. On the other
hand, the tire applies 3-D non-uniform contact stress, measured in the experiment
along with the realistic contact area. FEM considers these true measured contact
stresses and contact areas in the simulations. Details about tire contact measure-
ments can be found elsewhere (Hernandez et al. 2013). In addition to the
In-Situ Validation of Three-Dimensional Pavement … 149

Fig. 2 Pulse duration results for the two wheel paths

non-uniform contact stress, simulating the tire as a continuous moving load, rather
than a static steady load, is another important realistic consideration in the devel-
oped model.

2.3 Material Characterization

It is clearly stated in the literature that AC should be characterized as viscoelastic

material whose stiffness depends on the frequency of loading, time, and tempera-
ture. Time, temperature, and loading frequency dependencies can be captured by
various experiments that produce a plot called “master curve”. ABAQUS considers
shear and bulk moduli for viscoelastic young modulus through Prony coefficients
that are obtained from fitting to the master curve (Eqs. 1, 2). Afterwards, bulk and
shear modulus are calculated by assuming a constant Poisson’s ratio. Wlliams–
Landell–Ferry function was used to model time-temperature superposition of AC,
as shown in Eq. 3:
!
X
n  
t
GðtÞ ¼ G0 1  G i 1  e si ð1Þ
i¼1
!
X
n  
t
KðtÞ ¼ K0 1  K i 1  e si ð2Þ
i¼1
150 O.E. Gungor et al.

where
G = shear modulus,
K = bulk modulus
t = reduced relaxation time
G0 and K0 = instantenous shear and volumetric modulus
Gi, Ki and τi = Prony series parameters

C1 ðT  Tr Þ
logðat Þ ¼ ð3Þ
C2 þ ðT  Tr Þ

where
αt = shift factor
C1, C2 = regression coefficients
T = analysis temperature
Tr = reference temperature
In conventional pavement analysis approaches, both base and subgrade materials
are characterized as linear elastic material. However, it has been shown that stiffness
of the granular materials changes depending on the stress level they are exposed to.
Moreover, they behave differently in each principal direction. Therefore, they are
modelled as anisotropic stress-dependent nonlinear material using MEPDG model
(NCHRP 2004), as shown in Eqs. 4–6. It should be noted that effect of considering
non-linearity in material characterization is negligible for thick pavement, where
load is mostly carried by AC. Therefore, only granular material in thin pavement is
modelled as stress-dependent anisotropic material to decrease computational cost.
 k2  k3
h rd
Mrv ¼ k1 ð4Þ
po po
 k5  k6
h rd
Mrh ¼ k4 ð5Þ
po po
 k8  k9
h rd
Mrs ¼ k7 ð6Þ
po po

where
Mrv, Mrh, Mrs = vertical, horizontal and shear resilient modulus
h ¼ r1 þ r2 þ r3 = bulk stresses
po = unit reference pressure
k1, k2, k3, k4, k5, k6 = regression coefficients
In-Situ Validation of Three-Dimensional Pavement … 151

2.4 Analysis Method

There are three commonly used methods for pavement analysis: static, quasi-static,
and dynamic analysis. Static analysis assumes that the tire is not moving.
Quasi-static analysis models the tire as a moving load, but fails to capture the
inertial/damping effect and frequency-dependent material properties. Therefore,
dynamic analysis was used in this study to properly simulate moving tire loads with
nonlinear material characterization. The dynamic equation solved in ABAQUS is
given in Eq. 7. This equation can be solved using the implicit or explicit direct
integration method. In this study, the implicit direct integration method was selected
because it is more efficient for the level of frequencies observed in pavement
simulations.
   
€ þ ½C U_ þ ½KfUg ¼ fPg
½M U ð7Þ

where
[M] = mass matrix
[C] = damping matrix
[K] = stiffness matrix
{P} = external force vector
{Ü} = acceleration vector
{U̇} = displacement vector.

2.5 Interface Model

The model used for defining how two pavement layers interact with each other is
another key parameter for pavement simulation. All AC layers are assumed to be
fully bonded to each other in the developed model. On the other hand, AC-base and
base-subgrade interaction were simulated using a Coulomb model. In this model,
resistance of the movement is assumed to be proportional to normal stress at the
interface. In addition, a tolerance limit was set for shear strength above which two
layers start sliding relative to each other. In case of relative sliding, frictional stress
would be assumed constant.

3 Results and Discussion

Validation of FE model of flexible pavement was conducted for four different

pavement sections: Illinois Center of Transportation, Florida, Smart Road, and UC
Davis sections. The validation results and discussions are given below for each
section.
152 O.E. Gungor et al.

3.1 Illinois Center for Transportation Section

The first attempt to validate pavement FE model was performed on a pavement

section built at the Illinois Center of Transportation testing facility. The pavement
section, called “Section C” (Al-Qadi et al. 2008), was selected for validation. The
layer properties of this section are given in Table 1.
This section with 3 in of AC thickness is considered as thin section. The Prony
coefficients used for viscoelastic characterization of AC are given in Table 2. The
base material should be modelled as anisotropic stress-dependent material to
achieve more realistic pavement simulations. The nonlinear model parameters
(Eqs. 4–6) are given in Table 3.
The selected loading conditions for validation were 35.6 kN axle load and
690 kPa tire inflation pressure for DTA. A comparison between FEM and field data
was performed by using two pavement responses: vertical pressure at the bottom of
base and transverse strain at the bottom of AC (Table 4). The relative difference
between FEM and field data are computed as 9.7 and 24.8 %, respectively. The
variation of transverse strain at the bottom of AC could be attributed to bending of
the strain gauge in the field due to weak supporting platform for AC. In this case,
strain gauges report higher values since they possibly report summation of axial
tensile strain and bending tensile strain.

Table 1 Illinois Center for Transportation section input parameters

Layer Thickness Instantaneous/linear Poisson’s Density Rayleigh Rayleigh
(mm) modulus (MPa) ratio (×10−6) damping damping
parameters parameters
(α) (β)
AC 127 9840 0.35 2.30 N/A N/A
Base 305 Nonlinear 0.35 2.10 3.1416 0.000795
Subgrade – 45 0.4 1.56 3.1416 0.000795

Table 2 Prony coefficients i Gi (MPa) Ki (MPa) τi

of AC layer
1 6209.04 6209.04 0.0206
2 2469.84 2469.84 0.173
3 833.448 833.448 1.29
4 261.744 261.744 5.35
5 64.944 64.944 106

Table 3 Nonlinear model k1 1010 k4 227 k7 321

parameters for base layer (kPa) (kPa) (kPa)
k2 0.791 k5 1.071 k8 0.857
k3 −0.478 k6 −1.332 k9 −6.681
In-Situ Validation of Three-Dimensional Pavement … 153

3.2 Florida Section

Results of the Accelerated Pavement Test conducted at Florida Department of

Transportation facility were used to validate FEM pavement model. The test lanes
measured 92 m long and 3.7 m wide. The “Test pit” section was selected for FEM
validation. The pavement structure properties of this section and the other inputs
parameters for FEM are given in Table 5.
The AC layer is characterized as linear viscoelastic material by Prony series,
shown in Table 6. The pavement is considered as “thin pavement”, hence, the base

Table 4 Validation results for Illinois Center for Transportation section

Depth Location Pavement Simulation Field Difference
(mm) definition response results measurement (%)
432 Top of Vertical 16.8 15.2 10.9
subgrade pressure (kPa)
188 Bottom of AC Transversal 53.3 70.9 24.8
(BM-25) strain (με)

Table 5 Florida section input parameters

Layer Thickness Instantaneous/linear Poisson’s Density Rayleigh Rayleigh
(mm) modulus (MPa) ratio (×10−6) damping damping
parameters parameters
(α) (β)
AC 76.2 15,420 0.35 2.3 N/A N/A
Base 266.7 Nonlinear 0.4 2.1 3.1416 0.000795
Subbase 304.8 140 0.4 1.5 3.1416 0.000795
Subgrade 131 0.4 1.5 3.1416 0.000795

Table 6 Prony coefficients i Gi (MPa) Ki (MPa) τi

of AC layer
1 6859.929 6859.929 8.06E+00
2 5938.446 5938.446 7.89E+01
3 7098.832 7098.832 6.75E+02
4 5085.221 5085.221 6.49E+03
5 3890.706 3890.706 5.72E+04
6 1419.7664 1419.7664 8.16E+05
7 2969.223 2969.223 2.07E+06
8 219.44947 219.44947 7.61E+07
9 385.6577 385.6577 5.00E+08
10 246.07009 246.07009 2.88E+09
11 22.52514 22.52514 1.31E+10
154 O.E. Gungor et al.

Table 7 Nonlinear model parameters for base layer

k1 (kPa) 1038.7 k3 (kPa) 1038.7 k7 (kPa) 1038.7
k2 0.67 k5 0.67 k8 0.67
k3 −0.302 k6 −0.302 k9 −0.302
Because of lack of data, base layer is assumed to be isotropic material

was modelled as stress-dependent anisotropic material to have more realistic sim-

ulations. The Uzan parameters for the base is given in Table 7.
The loading conditions were defined with a half-axle load of 26.7 kN and tire
pressure of 552 kPa for DTA. Four critical responses, including vertical pressures at
bottom of AC and bottom of base and transverse and longitudinal strains at bottom
AC, were used for validation, as shown in Table 8. Four pressure cells are installed
in this section, two of them are located at the bottom of AC and the other two at the
bottom of the base layer. The relative difference between FEM and field are
reported as 1.2 and 16.1 %. On the other hand, a total of six strain gauges were
installed at the bottom of the AC layer: three in traffic direction and three in the
transverse direction. The relative differences between FEM and field data for these
responses were 11.9 and 52 % for traffic and transverse directions, respectively
(Table 9).

Table 8 Validation results for Florida section

Depth Location Pavement Simulation Field Difference
(mm) definition response results measurement (%)
76.2 Top of base Vertical pressure 166.9 168.9 1.18
(kPa)
342.9 Top of Vertical pressure 26.89 31 13.26
subgrade (kPa)
76.2 Bottom of AC Transversal strain 18.5 21.1 12.32
(με)
76.2 Bottom of AC Longitudinal 43 98 56.12
strain (με)

Table 9 Smart Road section input parameters

Layer Thickness Instantaneous/linear Poisson’s Density Rayleigh Rayleigh
(mm) modulus (MPa) ratio (×10−6) damping damping
parameters parameters
(α) (β)
AC 38 4230.9 0.35 2.30 N/A N/A
AC 150 4750.49 0.4 2.30 3.1416 0.000795
OGDL 75 2415 0.4 2.10 3.1416 0.000795
CTB 150 10342.14 0.4 2.10 3.1416 0.000795
Subbase 175 310.26 0.4 1.5 3.1416 0.000795
Subgrade – 262.00 0.4 1.5 3.1416 0.000795
OGDL open-graded drainage layer, CTB cement-treated base
In-Situ Validation of Three-Dimensional Pavement … 155

FEM results are in agreement for the vertical pressures and tensile strains in the
transverse direction (around 10 %). However, the prediction of the strain in traffic
direction is off by a factor of 2. The reason for this discrepancy can be explained as
follows: the base material is characterized as stress-dependent anisotropic nonlinear
material to simulate the behavior of the base layer. However, since the resilient
modulus test was conducted only using vertical deviatoric load, it was not possible
to obtain nonlinear material characterization parameters for all three dimensions.
Therefore, the base material was assumed isotropic; hence, FEM may have resulted
in underestimated tensile strains in the traffic direction.

3.3 Smart Road Section

Section B, which is considered a “thick” section, of the Smart Road project was
used to validate the FE model in this study. The section has the following input
parameters (Al-Qadi et al. 2004).
AC layers are characterized as viscoelastic and were defined using the Prony
series. Al-Qadi et al. (2008) shifted the Prony series to 25 °C and, therefore, the
William–Landel–Ferry coefficients are omitted (Table 10). In this scenario, the two
layers are assumed to be subjected to uniform temperature condition of 25 °C.
Additionally, since the pavement is considered as thick pavement, the base layer is
assumed to be linear elastic material to decrease computational cost.
Moreover, the loading condition was defined with a half-axle load of 35 kN and
720 kPa for the DTA. Given all presented input parameters, three critical responses
were determined from the FE model and compared with field data (Table 11).
Based on the resulting differences, the vertical pressure on top of the subgrade and
transversal and longitudinal strain underneath the AC layers varied by 2.4, 12.1,
and 2.1 %, respectively. For this section, FEM delivered quite accurate approxi-
mation for field pavement responses.

Table 10 Prony coefficients of AC layer

i Gi (MPa) Ki (MPa) τi Gi (MPa) Ki (MPa) τi
1 3277.4667 3277.4667 1.00 × 10−2 3085.7378 3085.7378 1.00 × 10−2
2 698.01388 698.01388 1.00 × 10−1 1069.5918 1069.5918 1.00 × 10−1
3 168.3602 168.3602 1.00 405.53936 405.53936 1.00
4 58.509116 58.509116 1.00 × 101 117.09055 117.09055 1.00 × 101
5 11.577435 11.577435 1.00 × 102 50.973233 50.973233 1.00 × 102
6 7.2373775 7.2373775 1.00 × 103 70927.191 70927.191 1.00 × 103
7 8.6331515 8.6331515 1.00 × 104 3.4551549 3.4551549 1.00 × 104
8 0.0699114 0.0699114 1.00 × 105 – – –
9 0.0288996 0.0288996 1.00 × 106 – – –
156 O.E. Gungor et al.

Table 11 Validation results for Smart Road section

Depth Location Pavement Simulation Field Difference
(mm) definition response results measurement (%)
588 Top of Vertical 40 40.9 2.44
subgrade pressure (MPa)
188 Bottom of Transversal 103.79 92.62 12.06
AC strain (με)
188 Bottom of Longitudinal 116.11 118.65 2.14
AC strain (με)

3.4 UC Davis Section

The last section used for FEM validation was built at the University of California
Pavement Research Center facility in Davis, California. The section, called
“671HC”, was selected for FEM validation. This section has two AC layers, each
60 mm thick. The top layer is warm mixed asphalt and the bottom is AC.
Underneath the two AC layers, a recycled base and a subbase layers exit having 250
and 270 mm layer thicknesses, respectively (Table 12).
The Prony coefficient for viscoelastic characterization of AC is given in
Table 13. For this section, it was important to assume a linear elastic behavior for
the base layer because resilient modulus data for this section was not available. The
base layer was assumed to have 60 MPa, which was obtained from the cone
penetration test.
The load case was selected as 26 kN axle load and 552 kPa tire pressure for
DTA. FEM was compared with field data for four pavement responses: vertical
pressures at the top of base and subgrade and tensile strain at the bottom of each lift
of AC layer. Comparison results are given in Table 14.
FEM prediction for vertical pressure at the top and bottom of the base layer was
much lower than field responses. The base material was characterized as linear
elastic material due to the lack of resilient modulus data. As reported in the liter-
ature (Kim et al. 2009), linear elastic characterization of base results is accepted in
stiffer behavior in pavement simulation. The effect gets even more significant for
pavement with thinner AC layer as in this section.

Table 12 UC Davis section input parameters

Layer Thickness Instantaneous/linear Poisson’s Density Rayleigh Rayleigh
(mm) modulus (MPa) ratio (×10−6) damping damping
parameters (α) parameters (β)
AC 60 25,070.5 0.35 2.30 × 10−6 N/A N/A
AC 60 20,200.2 0.4 2.30 × 10−5 3.1416 0.000795
Base 250 300 0.4 2.10 × 10−6 3.1416 0.000795
Subbase 270 300 0.4 2.10 × 10−7 3.1416 0.000795
Subgrade – 60 0.4 1.5 × 10−6 3.1416 0.000795
In-Situ Validation of Three-Dimensional Pavement … 157

Table 13 Prony coefficients of AC layer

i Gi (MPa) Ki (MPa) τi Gi (MPa) Ki (MPa) τi
1 8991.5622 8991.5622 6.43E+00 4303.1373 4303.1373 7.97E+00
2 4362.6974 4362.6974 7.44E+02 5422.6967 5422.6967 6.55E+01
3 4693.0468 4693.0468 5.99E+03 1115.6559 1115.6559 8.94E+02
4 1761.065 1761.065 7.70E+04 3151.1575 3151.1575 6.61E+03
5 3019.8724 3019.8724 4.21E+05 240.59757 240.59757 9.67E+04
6 728.78796 728.78796 6.57E+06 3450.7857 3450.7857 4.44E+05
7 142.43435 142.43435 9.09E+07 1169.9032 1169.9032 5.19E+06
8 586.31793 586.31793 6.19E+08 453.44754 453.44754 6.50E+07
9 673.50101 673.50101 4.87E+09 56.296935 56.296935 9.50E+08
10 65.073848 65.073848 1.09E+10 574.7684 574.7684 3.05E+09

Table 14 Validation results for UC Davis section

Depth Location Pavement Simulation Field Difference
(mm) definition response results measurement (%)
120 Top of base Vertical pressure 8.6 12.3 30.08
(kPa)
370 Top of Vertical pressure 4.5 5.8 22.41
subgrade (kPa)
120 Bottom of AC Transversal strain 11.57 16.9 31.54
(με)

4 Conclusion

This paper presents a study about in-situ validation of flexible pavement FE

modelling. The validation was conducted on four different sections. Three sections
are classified as “thin pavement”, where anisotropic stress-dependent base char-
acterization plays a critical role in the realistic simulation of pavement behavior
under tire loading. In all cases, AC was modelled as viscoelastic material to capture
its dependency on the frequency of loading and temperature. Moreover, the tire was
simulated as moving load to capture the dynamic effect on pavement, and realistic
temperature gradient was imposed on pavement.
Validation was performed for pavement responses of vertical pressure at layer
interfaces and transverse strains at the bottom of AC. In general, good agreement
was noted between FEM calculated and pavement field instrument responses. The
relative differences were computed around 10–15 % which is acceptable, consid-
ering the high variability in field data. The importance of proper characterization of
base layer for accurate prediction of pavement responses is also highlighted in this
paper. In conclusion, this study proves that FEM is a promising technique for
simulating tire-pavement interaction in the field when proper material characteri-
zation parameters and accurate tire loading are considered.
158 O.E. Gungor et al.

Acknowledgments This paper is based on the results of project DTFH61-11-C-00025, The

Impact of Wide-Base Tires on Pavement—A National Study. Project DTFH61-11-C-00025 is
conducted in cooperation with the Illinois Center for Transportation; the U.S. Department of
Transportation, Federal Highway Administration; Rubber Manufacturers Association, and the
following state departments of transportation: Florida, Illinois, Minnesota, Montana, New York,
Ohio, Oklahoma, South Dakota, Texas, and Virginia. The authors would like to acknowledge the
assistance provided by many individuals including Hasan Ozer, Mojtaba Ziyadi, Bouzid
Choubane, Jamie Green of FLDO, John Harvey of UC Davis, Erol Tutumluer of UIUC.

References

AASHTO. (2008). Mechanistic-empirical pavement design guide: A manual of practice.

Washington, DC: American Association of State Highway and Transportation Officials.
Al-Qadi, I., Dessouky, S., Kwon, J., & Tutumluer, E. (2008). Geogrid in flexible pavements:
Validated mechanism. Transportation Research Record: Journal of the Transportation
Research Board, 2045, 102–109.
Al-Qadi, I., Hernandez, J., Gamez, A., Ziyadi M., Gungor E., Kang S., et al. (In Review). The
impact of wide-base tires on pavement—A national study.
Al-Qadi, I. L., Loulizi, A., Elseifi, M., & Lahouar, S. (2004). The Virginia SMART ROAD: The
impact of pavement instrumentation on understanding pavement performance (with discus-
sion). Journal of the Association of Asphalt Paving Technologists, 73.
Al-Qadi, I. L., & Yoo, P. J. (2007). Effect of surface tangential contact stresses on flexible
pavement response (with discussion). Journal of the Association of Asphalt Paving
Technologists, 76, 663–692.
Al-Qadi, I. L., Yoo, P. J., Elseifi, M. A., & Janajreh, I. (2005). Effects of tire configurations on
pavement damage (with discussion). Journal of the Association of Asphalt Paving
Technologists, 74(1), 921–961.
Ameri, M., Mansourian, A., Khavas, M. H., Aliha, M. R. M., & Ayatollahi, M. R. (2011). Cracked
asphalt pavement under traffic loading—A 3D finite element analysis. Engineering Fracture
Mechanics, 78(8), 1817–1826.
Burmister, D. M., Palmer, L. A., Barber, E. S., & Middlebrooks, T. A. (1944). The theory of stress
and displacements in layered systems and applications to the design of airport runways.
Highway Research Board proceedings (Vol. 23).
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.
Fang, H., Haddock, J. E., White, T. D., & Hand, A. J. (2004). On the characterization of flexible
pavement rutting using creep model-based finite element analysis. Finite Elements in Analysis
and Design, 41(1), 49–73.
Gungor, O. E., Hernandez, J. A., Gamez A., & Al-Qadi, I. L. (2014). Quantitative assessment of
effect of wide base tire on pavement response using finite element analysis. Transportation
Research Record: Journal of the Transportation Research Board, 1990.
Hernandez, J. A., Al-Qadi, I., & De Beer, M. (2013). Impact of tire loading and tire pressure on
measured 3-D contact stresses. In Airfield and highway pavement 2013: Sustainable and
efficient pavements.
Kim, M., Tutumluer, E., & Kwon, J. (2009). Nonlinear pavement foundation modeling for
three-dimensional finite-element analysis of flexible pavements. International Journal of
Geomechanics, 9, 195–208.
National Cooperative Highway Research Program (NCHRP). (2004). Guide for the
mechanistic-empirical design for new and rehabilitated pavement structure, 1-37A.
Washington, DC: Transportation Research Board.
In-Situ Validation of Three-Dimensional Pavement … 159

Siddharthan, R. V., Yao, J., & Sebaaly, P. E. (1998). Pavement strain from moving dynamic 3D
load distribution. Journal of Transportation Engineering, 124(6), 557–566.
Underwood, B. S., Kim, Y. R., Savadatti, S., Thirunavukkarasu, S., & Guddati, M. N. (2009).
Response and fatigue performance modeling of ALF pavements using 3-D finite element
analysis and a simplified viscoelastic continuum damage model. Journal of the Association of
Asphalt Paving Technologists, 78.
Wang, H., & Al-Qadi, I. L. (2010). Impact quantification of wide-base tire loading on secondary
road flexible pavements. Journal of Transportation Engineering, 137(9), 630–639.
Yoo, P., & Al-Qadi, I. (2007). Effect of transient dynamic loading on flexible pavements.
Transportation Research Record: Journal of the Transportation Research Board, 1990, 129–
140.
Investigation of 3D-Move Responses
Under Traffic Speed Deflection
Devices (TSDDs)

Mahdi Nasimifar, Raj S. Siddharthan, Elie Y. Hajj

and Ramin Motamed

Abstract Traffic Speed Deflectometer (TSD) and Rolling Wheel Deflectometer

(RWD) that measure surface deflections at posted traffic speeds (up to 80–96 kph)
are being used in several countries to evaluate structural condition at network level.
Incorporating these devices to Pavement Management System (PMS) application
requires an analysis methodology for enabling the use of device measurements.
3D-Move Analysis evaluates pavement response using a continuum-based
finite-layer approach. Since 3D-Move has the capability of modeling moving
load and the resulting dynamic pavement responses, it is ideally-suited to simulate
pavement responses generated by these devices that move at high-speeds.
A recently completed FHWA study conducted field trials that used TSDDs where
in situ pavement response measurements at the MnROAD test site were made and
calibration of the 3D-Move model for application with TSDD loading was carried
out. Many calibration runs were performed and 3D-Move model provided a good
match with a variety of independent pavement responses that included surface
deflection bowls (measured using embedded geophone sensors) as well as hori-
zontal strains at the bottom of the Asphalt Concrete (AC) layers (measured using
embedded sensors). Using the calibrated 3D-Move software, an analytical inves-
tigation was then undertaken to explore relationships between load-induced pave-
ment structural-related responses and the corresponding surface deflection basin
related indices from TSDDs.

Keywords Traffic speed deflection devices
3D-move
Pavement responses

Continuous deflection basin Field data

M. Nasimifar (&)
R.S. Siddharthan
E.Y. Hajj
R. Motamed

University of Nevada, Reno, Reno, NV, USA
e-mail: [email protected]
R.S. Siddharthan
e-mail: [email protected]
E.Y. Hajj
e-mail: [email protected]
R. Motamed
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 161

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_11
162 M. Nasimifar et al.

1 Introduction

State Highway Agencies (SHA) routinely estimate structural adequacy as part of

their PMS activities based on surface deflection measurements. It has been rec-
ommended that advancing from stationary measuring deflection using Falling
Weight Deflectometers (FWDs) to deflections measured at traffic speeds is a more
viable alternative for network level pavement management applications. Recent
developments in moving deflection-testing devices enable such initiative. The
modern versions of the moving deflection testing devices that are actively used
today include the Greenwood Traffic Speed Deflectometer (TSD) and the Applied
Research Associates, Inc. (ARA) Rolling Wheel Deflectometer (RWD). These
devices are generally known as Traffic-Speed Deflection Devices (TSDDs).
A recently completed FHWA study (Rada et al. 2015) conducted field trials that
used TSDDs where in situ pavement response measurements were made and cal-
ibration of 3D-Move model for application with TSDD loading was carried out.
The MnROAD facility near Maplewood, Minnesota was selected as the primary
site for the field trials since it provided a multitude of test sections in one location as
well as a wealth of readily available information, including pavement structure,
pavement condition, and environmental and vehicle-induced pavement response
data. In addition to the existing MnROAD sensors (strain gauges, pressure cells,
etc.), four geophones were installed near the pavement surface to measure dis-
placements at three cells called accuracy cells.
The 3D-Move analytical model is capable of determining pavement responses
induced by vehicle travelling at traffic speeds (Siddharthan et al. 1998, 2002). This
model was calibrated as an initial undertaking before use in the development of a
methodology for incorporating TSDD measurements into network-level PMS
applications. Many calibration runs were performed and the results were compared
with measured deflection time histories (peak and basin) from the project geo-
phones and strain measurements taken by MnROAD strain gauges at various
locations within the pavement. Since load-induced strains are critical inputs to
pavement performance predictions, this effort was critical relative to the applica-
bility of the 3D-Move model. Field calibration brings together many other
important issues that include pavement layer thicknesses, material properties and
TSDD loading configuration, which are essential components of the pavement
response estimation.
Using the calibrated 3D-Move software, an analytical investigation was then
undertaken to explore relationships between load-induced pavement structural-
related response and the corresponding surface deflection basin related indices.
A key element was the simulation of pavement deflections using the 3D-Move
model with a focus on understanding the parameters that affect the TSDD mea-
surements that included vehicle speed and loadings, and pavement layer properties
(thickness and stiffness).
Investigation of 3D-Move Responses … 163

2 Overview of 3D-Move Program

The moving nature of the load is routinely overlooked in the conventional pave-
ment analysis in which the loads are considered as static and stationary. It has been
clearly shown from pavement responses measured in the field that the pavement
strain responses are affected by the speed of the vehicle (Sebaaly and Tabatabaee
1993). In general, there are two important factors that should be considered in any
dynamic pavement analysis: moving nature of the load and the dependency of the
material properties on the loading frequency (or vehicle speed).
Three-dimensional finite element based models have also been proposed for
pavement analysis (Huhtala and Pihlajamaki 1992; Sousa and Monosmith 1987).
Limitations associated with such methods are well-known, and they include sub-
stantial computational effort and the errors resulting from the need to incorporate
artificial lateral boundaries.
The 3D-Move model considers the vehicle loading as moving with all compo-
nents of contact stress distributions (normal and shear) of any shape and vis-
coelastic material characterization for the pavement layers. It takes advantage of the
horizontally-layered nature of the pavement structure in the formulation which
makes it significantly more computer efficient than the three-dimensional finite
element based models.
The analytical model (3D-Move) evaluates pavement response using a
continuum-based finite-layer approach. The finite-layer approach treats each
pavement layer as a continuum and uses the Fourier transform technique; therefore,
it can handle complex surface loadings such as multiple loads, non-uniform tire
pavement contact stress distributions, and any shaped tire imprints, including those
generated by wide-base tires (Siddharthan et al. 1998, 2000, 2002).
The following assumptions are used in the development of the model:
1. The domain is composed of horizontal layers of uniform thickness, which can
be of different materials;
2. Each layer can be either linear elastic or linear viscoelastic with a set of uniform
material properties (e.g., elastic modulus, Poisson’s ratio, unit weight), which
are time and space invariants;
3. Layers are modeled as single-phase;
4. Layers are finite, horizontally layered and rest on a rigid bottom layer;
5. The surface loads are assumed to move with constant speed (i.e., no accelera-
tion) along the vehicle direction (x-axis).
The 3D-Move model uses Fourier series expansion to decompose the loads into
harmonic components in space (x and y directions). Therefore, the applied load is
modeled as a two-dimensional periodic function in x- and y-directions, as shown in
Fig. 1.
Total response at a given location is then calculated by adding the individual
responses from each harmonic component. It can handle any number of layers with
the complex loading (non-uniform and any shape) at the surface and any number of
164 M. Nasimifar et al.

Fig. 1 Idealization of surface loads as a two-dimensional periodic function in 3D-move model

response evaluation points. Since the contact area can be of any shape, this approach
is suitable to analyze any tire imprints, including those generated by wide-base tires.
Since the 3D-Move has the capability of modeling moving loads and the
resulting dynamic pavement responses, it is ideally-suited to evaluate and compare
pavement responses measured using load-response devices that move at
high-speeds (e.g., TSD and RWD devices). Since rate-dependent material proper-
ties (viscoelastic) can be accommodated by the approach, it is an ideal tool to model
the behavior of the AC (asphalt concrete) layer and also to study pavement response
as a function of vehicle speed. In addition, because Fourier transform technique
along with frequency-domain solutions are adopted, the approach allows for the
direct use of the frequency sweep test data of the AC mixture (viscoelastic prop-
erties) in the analysis. The AC modulus |E*| at all of the sampling frequencies of
FFT are used in the computation.

3 TSDD Descriptions

The ARA RWD uses six spot lasers mounted on a horizontal aluminum beam to
measure the deflected pavement surface (longitudinally along the mid-point between
the dual tires). Two sensors (Sensor D located at 7.25 in. (184 mm) behind and
Sensor F located at 7.75 in. (197 mm) in front of the axle in Fig. 2) are within the
deflection bowl, while the other four sensors represent locations within the unde-
flected pavement surface. The A, B, C and E sensor readings are used to obtain the
load-induced surface deflection at the location of Sensors D and F.
The TSD utilizes Doppler lasers to estimate the so-called deflection velocity of
the road profile that is the velocity the pavement deflects due to the moving load. The
Investigation of 3D-Move Responses … 165

Fig. 2 RWD sensor locations

Fig. 3 TSD rear axle configuration and location of Doppler sensors (DV deflection velocity)

Greenwood TSD provides deflection velocities at between three and nine points,
with the model that evaluated in this study measuring six (at 100, 200, 300, 600, 900
and 1500 mm) as shown in Fig. 3. A theoretical algorithm is used to compute the
deflection basin that matches with the TSD measurements (Pedersen 2012).

4 MnROAD Field Trials

Cells 3 and 19 in the MnROAD Mainline and Cell 34 in the MnROAD Low
Volume Road (see http://www.dot.state.mn.us/mnroad/ for detail) were selected for
purposes of evaluating the accuracy of the TSDDs, and hence they are referred to as
the accuracy cells. All three cells consisted of flexible pavements. Each cell was
instrumented with four geophones during the project. The configuration of
embedded geophones at each site is shown in Fig. 4. These cells were also pre-
viously instrumented during construction by MnROAD with sensors such as strain
166 M. Nasimifar et al.

Fig. 4 Configuration of embedded project sensors and spacing

gauges, pressure cells, LVDTs, moisture gauges, thermocouples and tipping

buckets. The TSDDs field trials were conducted at various vehicle speeds. In the
calibration process, load-induced measured response values of strains, stresses and
displacements from the referenced sensors at the three accuracy cells were com-
pared with those computed with 3D-Move.

5 Material Characterization of MnROAD Accuracy Cells

The characterization of the material properties at the three accuracy cells is an

important step in the simulation process. The required 3D-Move inputs are:
• For surface layer (AC-considered viscoelastic):
1. Thickness;
2. Unit weight;
3. Dynamic modulus and damping as a function of temperature and frequency;
and
4. Poisson’s ratio.
• For unbound layers (Base and Subgrade—considered linear elastic):
1. Thickness;
2. Unit weight;
Investigation of 3D-Move Responses … 167

3. Resilient modulus and

4. Poisson’s ratio.
MnROAD maintains a database containing laboratory and field-testing results on
soils, aggregates, asphalt mixtures, asphalt binders, concrete mixtures, and other
materials. The MnROAD database also contains cell-specific information, including
layer thickness, type of layers and cross section of cells at time of construction and
subsequent treatments.
A critical input to 3D-Move is the material properties that are appropriate for the
conditions (e.g., temperature) that existed at the time of TSDD trials that were
conducted at different times of the day (e.g., early morning and late afternoon).
FWD testing data conducted prior to the TSDDs field trials were available. Figure 5
provides a sketch of the pavement sections and FWD backcalculated layer moduli
of Cells 3, 19 and 34. These pavement structures represent pavements with AC
thickness varying from 3 to 5 in. and base thickness varying from 12 to 43 in.
Standard deviation of FWD backcalculated layer moduli is shown in the figure with
σ value.
The AC modulus is highly affected by the temperature. Therefore, accurate
estimates of the AC layer temperatures at the time of testing were necessary.
Table 1 summarizes the average AC layer temperatures at the time of FWD testing
and TSDDs testing.

Fig. 5 Sketch of pavement structures used for Cells 3, 19, and 34

168 M. Nasimifar et al.

Table 1 Average temperatures within AC layer

Cell Temperature at time of Temperature at time of Temperature at time of
FWD testing, °F (°C) TSD testing, °F (°C) RWD testing, °F (°C)
3 99 (37) 91 (33) 99 (37)
19 81 (27) 68 (20) 63 (17)
34 108 (42) 91 (33) 90 (32)

The procedure used to develop the appropriate AC master curves took into
account the following considerations (Nasimifar et al. 2015):
(1) Undamaged AC modulus as determined from Witczak’s equation (Kim et al.
2011);
(2) FWD backcalculated modulus (in situ or existing modulus); and
(3) In situ modulus correction approach (Fatigue Damage Factor, dAC) used in the
MEPDG overlay design (Loulizi et al. 2007).
Existing AC moduli as a function of frequency for the three MnROAD accuracy
cells were developed using the steps described above. Figure 6 shows AC modulus
versus frequency for the temperatures associated with the TSDDs field tests for Cell
19. The curve appears realistic giving smooth variation at both the low and high
frequencies, and hence it was used as input to the 3D-Move model.

Fig. 6 Existing moduli for Cell 19 in field trials

Investigation of 3D-Move Responses … 169

Table 2 TSDDs rear axle RWD rear tires TSD rear tires
configuration and loads
Tire pressure 100 psi 116 psi
Rear axle load 4750 lbs/tire 5575 lbs/tire
Dual tire spacing 14.5 in. 13.5 in.

6 Loading Characteristics of TSDDs

One of the important sets of input for analyzing pavement structures is the char-
acterization of the applied load. During the TSDD field trials, the loads carried by
the axles were determined using a static scale owned and operated by MnROAD.
Table 2 shows the characteristics and loads of the rear axles for the TSDDs. Rabe
(2013) showed that the dynamic load can vary by as much as 33 % of the static
load. Therefore, the variation of axle load, which has a direct influence on the
computed deflection response, should be addressed.
Since the deflection measuring sensors for the two TSDDs are mounted close to
the rear trailer axles, the 3D-Move comparisons in this calibration effort focused on
the responses generated by those axles. Since data on the contact stress distribution
at the tire-pavement interface were not available, the rear axles were modeled in
3D-Move as dual circular loads with uniform contact pressure.

7 Case Scenarios for 3D-Move Analysis

Care was taken to select appropriate inputs for 3D-Move program that represent the
realistic condition of material properties and TSDDs loading at the time of testing.
However, many possible factors might have contributed to discrepancy of the
selected input parameters. As explained earlier, FWD backcalculation was used to
find existing AC modulus of accuracy cells. The backcalculated layer moduli were
estimated using the program MODULUS (Liu and Scullion 2001) from FWD data.
Assuming linear elastic theory ignores the effect of stress level on estimated
existing modulus. Also, the estimation of bedrock or stiff layer location is another
concern in the backcalculation process. Geology in the area suggests that the
subgrade thickness at the accuracy cell locations are substatially greater than those
predicted from the backcalculation effort. As stated earlier, the effect of wheel load
variation during operation at high speeds should also be properly addressed.
Accordingly, the following three 3D-Move case scenarios were considered for
analyses:
• Case 1: Three layer pavement structure with same thicknesses as used in the
FWD backcalculation and corresponding mean layer moduli derived from the
FWD backcalculation results;
• Case X: Three layer pavement with: (a) thicknesses used in the FWD back-
calculation except decreasing the AC layer thickness by 1 in. (25.4 mm),
170 M. Nasimifar et al.

(b) (mean − σ) of FWD backcalculated layer moduli for AC and base layers,
(c) (mean + σ) of FWD backcalculated layer moduli for subgrade, and
(d) +25 % of nominal tire load; and
• Case X1: Same as Case X, but with no reduction in AC layer thickness. This
case was used for Cells 3 and 34 which have the thinner AC thickness.
Case X was used only for Cell 19 to bracket the measured values.

8 3D-Move Results for Accuracy Runs

The results from the analyses with 3D-Move using the case scenarios described in
the previous section were initially compared with the measured values from the
project sensors. As a representative plot, Fig. 7 shows the comparison of 3D-Move
results and measured deflections for Cell 34 based on TSD trial at a vehicle velocity
of 30 mph (48 kph). Since the main focus is the comparison of the deflection bowls
(shape and maximum value), the plots have been shifted to align so that the
maximum displacement locations coincide. Since the GEO1 and GEO3 are located
along a plane parallel to the vehicle direction, they are expected to give very similar
deflection bowls. Therefore, the responses from these two sensors are shown in the
figure. The variation between the deflection bowls given by these two sensors may
be viewed as a measure of the overall variability in the measurements made by the

Fig. 7 TSD trials—3D-move predictions and measured deflections [Cell 34; V = 30 mph (48
kph)]
Investigation of 3D-Move Responses … 171

project sensors and possibly any spatial variability between 1 ft (0.3 m) of distance
between geophones (Fig. 4). The lower and upper bound of the project sensor data
were considered by treating GEO1 and GEO3 data as independent datasets.
3D-Move adequately captured the maximum and shape of measured displacements.
Case X1 provides the closest deflection basin to that measured.
Figure 8 shows the comparison of maximum displacements computed by
3D-Move and those measured by the project sensors for all accuracy runs made
with the TSD and RWD. In this comparison, Case X1 was used for Cells 3 and 34
and Case X was used for Cell 19. The figure shows a good match between com-
puted and measured maximum displacements.
In addition to surface sensors, the measured strains at the MnROAD facility
were compared with the results of analytical simulations. This comparison is
important since a validated 3D-Move program can be subsequently used to obtain
reliable deflection basin indices that relate well with the pavement response. The
computed horizontal tensile strains at the bottom of AC layers were compared with
measured strains from longitudinal strain gauges in the MnROAD facility. The
comparison of measured and computed maximum response in all cells agreed well
as shown in Fig. 9.

Fig. 8 3D-move analysis computed maximum displacement versus measured (all cells, vehicle
speeds during RWD and TSD trials)
172 M. Nasimifar et al.

Fig. 9 Computed versus measured maximum longitudinal strain at the bottom of AC layers (all
cells and vehicle velocities during RWD and TSD trials)

9 Deflection Basin Indices and the Relationships

with Pavement Responses

This section presents details concerning the analytical investigation undertaken to

explore relationships between one of the load-induced structural-related responses
of a pavement system and the corresponding surface deflection-basin related indi-
ces. This investigation and associated results were considered an important effort
toward the development of methodologies for incorporating TSDD measurements
into network-level PMS applications.
Pavement failure (or structural capacity), in terms of fatigue or rutting, can be
estimated from performance prediction equations (e.g., NCHRP 1-37A 2004, etc.).
These equations relate load-induced pavement responses to the following two
important pavement distresses: certain level of AC fatigue cracking and rutting
failure in the subgrade. The critical load-induced pavement responses that relate to
the two referenced distresses are the maximum tensile horizontal strain (also known
as fatigue strain) at the bottom of the AC layer and the vertical compressive strain
on top of the subgrade, respectively. In addition, there are studies that postulate the
existence of reliable correlations between deflection-basin related indices and
load-induced pavement structural responses (Horak 1987 and Thyagarajan et al.
2011), which suggest it is possible to correlate deflection basin indices to structural
capacity (or distress). Nasimifar et al. (2015, 2016a, b) evaluated many deflection
basin indices to identify the best indicators of the critical pavement performance.
A number of deflection-basin related indices have been proposed by researchers
that are perceived as strong predictors of the critical structural-related responses and
structural conditions (or capacity) of the pavements. Experimental and analytical
studies have reported strong correlations between some deflection basin indices and
pavement responses (Horak 1987; Thyagarajan et al. 2011). A total of 42 individual
Investigation of 3D-Move Responses … 173

Table 3 Deflection basin indices used in the evaluationa

Parameter and number of indices Indices for Parameter and number of Indices
evaluation indices for
evaluation
Radius of curvature (8)a R200 Deflection velocity (6), DV100
R = r2/(2Do(Do/Dr − 1)), (mils) R300 (mils/s) DV200
R400 DV300
R500 DV600
R600 DV900
R700 DV1500
R800
R900
Area (1) A = 6[1 + 2 (D300/Do) + 2 A Deflection Slope Index DSI100–200
(D600/Do) + D900/Do)] (5) DSI100-r = D100 − Dr,
Shape factors (2) F1 = (Do − D600)/D300 F1 (mils) DSI100–300
F2 = (D300 − D900)/D600 F2 DSI100–400
DSI100–500
DSI100–600
Surface Curvature Index TSD Deflecton Slope Index DSI200–300
(7) SCI = Do − Dr, (mils) SCI200 (4) DSI200-r = D200 − Dr, DSI200–400
(mils)
SCI300 DSI200–500
SCI400 DSI200–600
SCI500
SCI600
SCI900
RWD
D−184 − D197
Base Curvature Index BCI Slope of deflection SD200
(1) BCI = D600 − D900, (mils) (6) SD = tan−1 (Do − Dr)/r
Base Damage Index BDI SD300
(1) BDI = D300 − D600, (mils)
Area Under Pavement Profile (AUPP) AUPP SD400
(1) (5Do − 2D300 − 2D600 − D900)/2, SD500
(mils)
SD600
SD900
a
Numbers within parentheses indicate the number of indices for each deflection basin parameter
174 M. Nasimifar et al.

Table 4 Most appropriate Index Relation with tensile strain, μstrain

indices and correlations
related to tensile strain at the SCI300, (mils) 76.89 SCI0.7828
300 R2 = 0.93
MnROAD cells SCI400, (mils) 70.287 SCI0.6954
400 R2 = 0.93
SD300 118479 SD0.7828
300 R2 = 0.94
SD400 58278 SD0.6954
400 R2 = 0.93
DSI100–200, (mils) 150.62 DSI0.7917
100–200 R2 = 0.90
DSI100–300, (mils) 94.638 DSI0.7075
100–300 R2 = 0.91
DSI100–400, (mils) 82.799 DSI0.6406
100–400 R2 = 0.90
DSI200–300, (mils) 160.04 DSI0.6301
200–300 R2 = 0.90
DSI200–400, (mils) 119.34 DSI0.5692
200–400 R2 = 0.90
AUPP, (mils) 40.845 AUPP0.6266 R2 = 0.90

indices that were evaluated in this study are listed in Table 3. The indices were
correlated with tensile strains at the bottom of AC layer. The Surface Curvature
Index (SCI = Do − Dr) used deflection at the midpoint between the dual tires, Do,
as reference displacement. Dr in Table 3 is deflection at distance “r” from midpoint
between the dual tires. A new index called Deflection Slope Index (DSI) was also
considered with D100 and D200 as reference displacements.
Based on the pavement layer characteristics and loading configuration, the
maximum tensile strain response can occur at any of the transverse locations under
the dual tire. In order to capture the maximum response predicted by 3D-Move
program, the tensile strains are computed at several transverse locations at 1 in.
intervals. The maximum computed strain was used subsequently in the analysis.
The correlation between 3D-Move computed deflection indices and tensile strains
were evaluated and the best indices were selected. Table 4 summarizes the most
appropriate indices for MnROAD cells.
The 3D-Move results show that indices closer to the center of the load (e.g.,
SCI300, DSI200-300 and SD300) have a good relation to fatigue strains. The new
indices, DSI, were found also to be appropriate. Thus the best indices related to
fatigue strains are not necessarily depended on Do as found in literature, conse-
quently if D100 or D200 from the TSD is more reliably measured than Do, they can
be used as the reference deflections in the DSI calculations and subsequent relation
with fatigue strains.

10 Conclusion

The 3D-Move can take into account the modeling moving loads and the resulting
dynamic pavement responses. Therefore, it is ideally-suited to evaluate and com-
pare pavement responses measured using load-response devices that move at
high-speeds (e.g., TSD and RWD devices). The capability of the 3D-Move program
to simulate TSDDs was confirmed using the comparison between predicted pave-
ment responses from 3D-Move and those were measured by embedded sensors
Investigation of 3D-Move Responses … 175

during field trials at the MnROAD facility. Pavement responses considered inclu-
ded surface deflection bowls as well as horizontal strains at the bottom of the AC
layers. Then, using validated program, the best deflection basin indices related to
maximum horizontal strain at bottom of AC layer were selected and appropriate
relationships were developed. It should be noted the results are based on the sim-
ulation of TSDDs at the MnROAD facility. A comprehensive simulation of
pavement responses using dynamic simulation, but covering a larger pavement
database that considers the variability in pavement condition (layer configuration
and properties) is underway.

Acknowledgments The field measured data in this paper was originated from FHWA sponsored
study “Pavement structural evaluation at the network level” DTFH61-12-C-00031. We would like
to thank Dr. Nadarajah Sivaneswaran of the FHWA and Dr. Senthil Thyagarajan of ESC for their
kind support. We acknowledge Dr. Gonzalo Rada and Dr. Soheil Nazarian for their leadership and
encouragement. The conclusions presented in this paper are solely those of these authors.

References

Horak, E. (1987). The use of surface deflection basin measurements in the mechanistic analysis of
flexible pavements. In Proceedings of sixth international conference structural design of
asphalt pavements, University of Michigan, Ann Arbor, Michigan, USA (Vol. 1).
Huhtala, M., & Pihlajamaki, K. (1992). New concepts on load equivalency measurements. In
Proceedings of 7th international conference asphalt pavements, Nottingham, U.K.
(pp. 194–208).
Kim, Y. R., Underwood, B., Sakhaei Far, M., Jackson, N., & Puccinelli, J. (2011). LTPP computed
parameter: Dynamic modulus. Publication FHWA-HRT-10-035, FHWA, U.S. Department of
Transportation.
Liu, W., & Scullion, T. (2001, October). MODULUS 6.0 for windows: User’s manual. Texas
Transportation Institute.
Loulizi, A., Flintsch, G., & McGhee, K. (2007). Determination of in-place hot-mix asphalt layer
modulus for rehabilitation projects by a mechanistic-empirical procedure. In Transportation
Research Record: Journal of the Transportation Research Board, No. 2037. Transportation
Research Board of the National Academies, Washington. D.C., pp. 53–62.
Nasimifar, M., Siddharthan, R., Rada, G., & Nazarian, S. (2015). Dynamic analyses of traffic
speed deflection devices. International Journal of Pavement Engineering. doi:10.1080/
10298436.2015.1088152.
Nasimifar, M., Siddharthan, R. V., Rada, G. R., & Nazarian, S. (2016). Validation of dynamic
simulation of slow moving surface deflection measurements. In Transportation research board
95th annual meeting (no. 16-2492).
Nasimifar, M., Thyagarajan, S., Siddharthan, R., & Sivaneswaran, S. (2016). Robust deflection
indices from traffic speed deflectometer measurements to predict critical pavement responses
for network level PMS application. Journal of Transportation Engineering, ASCE, 04016004.
NCHRP. (2004). Guide for mechanistic-empirical design of new and rehabilitated structures. Final
Report for Project 1-37A, National Cooperative Highway Research Program, Transportation
Research Board, National Research Council, Washington, D.C.
Pedersen, L. (2012). Viscoelastic modeling of road deflections for use with the traffic speed
deflectometer. PhD study in collaboration with Greenwood Engineering, Technical University
of Denmark and the Ministry of Science and Innovation.
176 M. Nasimifar et al.

Rabe, R. (2013). Structural pavement monitoring with non-destructive measuring devices—

Experience from a pilot project in Germany. In Presentation at the ninth international
conference on bearing capacity of roads, railways and airfields, BCRRA 2013, Trondheim,
Norway.
Rada, G., Nazarian, S., Visintine, B., Siddharthan, R. V., & Thyagarajan, S. (2015). Pavement
structural evaluation at the network level, FHWA. Final report.
Sebaaly, P. E., & Tabatabaee, N. (1993). Influence of vehicle speed on dynamic loads and
pavement response. Transp. Res. Rec. 1410, Transportation Research Board (TRB),
Washington, D.C., pp. 107–114.
Siddharthan, R. V., Yao, J., & Sebaaly, P. E. (1998). Pavement strain from moving dynamic 3-D
load distribution. Journal of Transportation Engineering, 124(6), 557–566.
Siddharthan, R. V., Krishnamenon, N., & Sebaaly, P. E. (2000). Pavement response evaluation
using finite-layer approach. Transportation Research Record No. 1709, TRB, pp. 43–49.
Siddharthan, R. V., Krishnamenon, N., El-Mously, M., & Sebaaly, P. E. (2002). Investigation of
tire contact stress distributions on pavement response. Journal of Transportation Engineering,
128(2), 136–144.
Sousa, J. B., & Monosmith, C. L. (1987). Dynamic response of paving materials. Transportation
Research Record No. 1136, TRB, pp. 57–68.
Thyagarajan, S., Sivaneswaran, N., Petros, K., & Muhunthan, B. (2011). Development of a
simplified method for interpreting surface deflections for in-service flexible pavement
evaluation. In 8th international conference on managing pavement assets, Santiago, Chile.
Long-Term Pavement Performance
Monitoring and the Revision
of Performance Criteria for High Modulus
Asphalt in South Africa

Julius Komba, Joseph Anochie-Boateng, Johan O’Connell

and Benoit Verhaeghe

Abstract Enrobé à Module Élevé (EME) technology, a High Modulus Asphalt

(HiMA), was originally developed in France. The technology is primarily suitable
for construction of heavily trafficked routes, airports and container terminals. The
key performance characteristics of EME are high stiffness, high resistance to per-
manent deformation and fatigue cracking. EME is also designed to offer good
moisture resistance and good workability. The EME technology has been suc-
cessfully introduced to South Africa. The development of EME design guidelines in
South Africa started in 2006. A major outcome was the publication of Sabita
Manual 33: “Interim design procedure for high modulus asphalt in South Africa.”
The performance criteria/specifications stipulated in Manual 33 were based on
limited data. Based on further work against French mix design and analysis of data
collected in South Africa, a revised fatigue and stiffness specifications were adopted
in July 2015. Implementation of EME technology in South Africa started in 2011,
when a trial section consisting of an EME base layer was constructed on the heavily
trafficked South Coast Road in Durban. The section is a major entry route for heavy
vehicles travelling to the Durban harbour. Several attempts to rehabilitate the

J. Komba (&)
J. Anochie-Boateng
J. O’Connell
B. Verhaeghe

Council for Scientific and Industrial Research (CSIR),
P.O Box 395, Pretoria 0001, South Africa
e-mail: [email protected]
J. Anochie-Boateng
e-mail: [email protected]
J. O’Connell
e-mail: [email protected]
B. Verhaeghe
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 177

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_12
178 J. Komba et al.

section using conventional asphalt mixes had failed as a result of premature rutting
due to the heavy traffic volumes entering the Durban harbour. The heavy traffic
volumes at the section offered an ideal setting for an experiment in Accelerated
Pavement Testing (APT) without the use of a Heavy Vehicle Simulator (HVS),
which enabled the accelerated validation of the South African EME design pro-
cedure. The objective of this paper is to present the outcomes of the Long-Term
Pavement Performance (LTPP) monitoring programme that was undertaken to
assess the field performance of EME, and discuss the development of the newly
adopted South African EME performance specifications.

Keywords Enrobé à module élevé (EME)
Workability
Durability
Stiffness

Rutting Fatigue cracking

1 Introduction

Long-life and maintenance-free road design solutions are becoming increasingly

important to road owners, so as to curb congestion and minimize road user costs. In
the past few years, South African roads have experienced a significant increase in
volumes of heavy vehicles, resulting in the accelerated deterioration of the road
networks. The conventional asphalt pavement materials are no longer able to cope
with the significant increases in traffic volumes and loads. To ensure long-term
serviceability of the road networks, pavement design technologies and innovations
that address issues on heavily trafficked roads are becoming important to road
owners.
In 2006, the Southern African Bitumen Association (SABITA) identified the
need to implement flexible pavement solutions that would meet the needs of road
owners. One of the products identified was Enrobés à Module Elevé (EME) or High
Modulus Asphalt (HiMA), which was originally developed in France. The tech-
nology is primarily suitable for construction of heavily trafficked routes, airports
and container terminals. EME is hot-mix asphalt manufactured using hard,
unmodified bitumen (typically 10/20 or 15/25 penetration grades) as well as good
quality, fully crushed aggregate. The key performance characteristics of EME are
high stiffness, high resistance to permanent deformation and fatigue cracking (due
to high bitumen content). EME is also designed to offer good durability and
workability. The superior performance of EME allows for the construction of
pavements with longer life. International experience indicates that significant
pavement thickness reductions can also be achieved using EME resulting in saving
on the usage of asphalt materials. In France the use of EME has enabled the
reduction of asphalt layer thickness by between 25 and 40 % (Delorme et al. 2007).
Preliminary structural analyses performed on typical South African pavements have
Long-Term Pavement Performance Monitoring … 179

shown that asphalt base thickness reductions of approximately 30 % can be

achieved using EME (Denneman et al. 2011a). However, if the purpose is to
produce a longer-life pavement, it is recommended that layer thicknesses should not
be reduced.
The Council for Scientific and Industrial Research (CSIR) was tasked to conduct
a study to transfer EME technology to South Africa. A major outcome was the
publication of Sabita Manual 33: Interim design procedure for high modulus asphalt
in South Africa (Sabita 2013). The guideline provides procedures for the design and
evaluation of the structural performance of EME. Following the successful transfer
of the technology to South Africa, EME was used in the rehabilitation of a section
of the heavily trafficked South Coast Road owned by eThekwini municipality in the
Kwa-Zulu Natal province. The section is a major entry route for heavy vehicles to
the Durban harbor. Several attempts to rehabilitate the section using conventional
asphalt mixes had failed as a result of premature rutting due to the heavy traffic
volumes entering the Durban harbor. The heavy traffic volumes at the section
offered an ideal setting for an experiment in Accelerated Pavement Testing (APT)
without the use of a Heavy Vehicle Simulator (HVS), which enabled the accelerated
validation of the South African EME design procedure. In the absence of HVS, the
selection the extremely heavy trafficked section of the South Coast Road used in
this study ensured that heavy traffic loading can be applied over a short period time.
The extremely heavy traffic loading (i.e. 20 million Equivalent Standard Axle
Loads (ESAL’s) over a period of 4 years) provided an indication of the ability of
EME to withstand heavy traffic loading.
A Long-Term Pavement Performance (LTPP) monitoring programme was
undertaken to assess the field performance of the South Coast Road EME section
and the results indicate that EME can withstand heavy traffic loading. Subsequently,
EME has also been successfully used in the rehabilitation of a section N3 Highway
in Kwa Zulu Natal province between Candela Road and Paradise valley, and is
planned to be implemented on Buss Rapid Transit (BRT) routes and other heavily
trafficked routes in South Africa. In South Africa it became evident since the first
introduction of EME that mixes developed did not meet the interim fatigue criteria
contained in Sabita Manual 33 (Sabita 2013), whilst complying with the French
requirements when tested in France. Based on further benchmarking work against
French mix design and analysis of data collected in South Africa, revised fatigue
and stiffness criteria have been adopted (Sabita 2015).
The objective of this paper is to present the outcomes of the LTPP monitoring
programme that was undertaken to assess the field performance of the South Coast
Road EME section, and discuss the development of the newly adopted EME per-
formance specifications.
180 J. Komba et al.

2 An Overview of Process of Transferring EME

Technology to South Africa

In South Africa, EME technology has successfully been implemented following

research efforts to transfer the technology from France. The process followed to
transfer the EME technology to South Africa is well documented in Denneman
et al. (2011a). The process is briefly described in this section. The EME technology
transfer study consisted of four separate phases, namely:
• Phase one: Preliminary assessment of viability;
• Phase two: Development of preliminary mix and structurally design guidelines;
• Phase three: Validation of EME technology through APT, LTPP monitoring and
laboratory studies, and
• Phase four: Finalization of EME design guidelines and specifications.
The purpose of the phase one of the study was to get familiar with the com-
position, design methodologies, typical properties and use of EME in order to be
able to assess the feasibility and merits of importing the technology from France to
South Africa. A team from South Africa visited three French Companies namely
Colas, Total and Shell, as well as Transport Research Laboratory (TRL) in United
Kingdom (UK) to obtain information and gain insight on the design, application
and performance of EME from a European perspective. The visit took place in June
2008. The outcomes of the preliminary viability assessment formed the bases for
the subsequent phases of the EME technology transfer study.
Phase two focused on the development of performance related method for the
design of EME mixes, as well as the development of methodology for the structural
design of pavement structures containing EME layers. The developed South
African EME mix design method is based on the French methodology for the
design of bituminous mixtures as described by Delorme et al. (2007). According to
French mix design procedure, the performance of EME mix is evaluated with
respect to workability, durability, stiffness, resistance to permanent deformation
(rutting), and fatigue cracking. An extensive laboratory study was conducted to
translate French EME performance specifications to South African equivalents.
This was an important step due to the fact that the methods used to evaluate the
performance of asphalt mixes in France are different from those used in South
Africa, hence the need to develop EME specifications that are based on the local
South African test methods. Table 1 presents performance test methods used in the
French EME mix design procedure and the selected local South African
equivalents.
Detailed procedure for designing EME mixes is contained in Sabita Manual 33
(Sabita 2015). Brief descriptions of the main steps are as follow:
• The design process starts with selection of appropriate mix components in terms
of aggregate and bitumen. Aggregates should be fully crushed, with both
angularity and surface texture meeting requirements recommended in Sabita
Long-Term Pavement Performance Monitoring … 181

Table 1 French EME performance tests and selected South African equivalent
Property French South Africa
Test Method Test Method
Workability Gyratory compactor, EN Gyratory compactor, air ASTM
air voids after 100 12697-31 voids after 45 gyrations D6926
gyrations
Durability Duriez EN Modified Lottman ASTM
12697-12 D4867
Rutting Wheel tracking (large EN Repeated Simple Shear Test AASHTO
resistance device) at 60 °C and 12697-22 at Constant Height at 55 °C, T 320
30,000 cycles 5000 repetitions
Stiffness Two point bending EN Dynamic modulus test at AASHTO
flexural modulus 12697-26 10 Hz, 15 °C TP 79
10 Hz, 15 °C
Fatigue Two point bending EN Four point bending at 10 Hz, AASHTO
resistance 10 °C, 25 Hz to 50 % 12697-24 10 °C, to 50 % stiffness T 321
stiffness reduction reduction

Manual 33 (Sabita 2013). Filler (material passing 0.075 mm sieve) is also

required to meet the grading requirements. Hard bitumen, typically a 10/20 or
15/25 penetration-grades conforming to SANS 4001-BT1:2014 standard, is
required. There are two classes of EME, namely EME Class 1 and EME Class 2.
EME Class 2 has significantly higher binder content and superior performance
characteristics.
• After selection of mix components, a suitable grading is developed from the
different aggregate fractions. The grading is specified, in terms of the Nominal
Maximum Particle Size (NMPS), which is defined, as one sieve size larger than
the first sieve to retain at least 10 % of the aggregate (Asphalt Institute 1996).
Three grading types are specified in Sabita Manual 33, namely NMPS 10 mm,
NMPS 14 mm and NMPS 20 mm. These grading types should be used as a
point of departure, following which the selected grading needs to be optimized
using Bailey method principles described by Vavrik et al. (2002).
• Bitumen content is determined based on a minimum richness modulus, which is
similar to the binder film thickness used in the past for the design of conven-
tional asphalt mixes in South Africa.
• Using a trial mix design, a gyratory compactor is used to prepare specimens.
The specimens should meet the maximum voids requirement at 45 gyrations,
which is aimed at creating a workable mix. Once the workability requirement
has been met, specimens are subject to a durability test.
• Following the satisfactory workability and durability criteria, the structural
performance of the mix is assessed with respect to minimum dynamic modulus
(stiffness), a minimum level of resistance against permanent deformation/rutting
and against fatigue cracking.
182 J. Komba et al.

Table 2 EME performance tests and interim specifications (Sabita 2013)

Property Test Method Requirements
Class 1 EME Class 2 EME
Workability Gyratory compactor, air ASTM ≤10 % ≤6 %
voids after 45 gyrations D6926
Durability Modified Lottman ASTM ≥0.8 ≥0.8
D4867
Rutting Repeated Simple Shear Test AASHTO ≤1.1 % ≤1.1 %
resistance at Constant Height at 55 °C, T 320 strain strain
5000 repetitions
Stiffness Dynamic modulus test at AASHTO ≥14 GPa ≥14 GPa
10 Hz, 15 °C TP 79
Fatigue Four point bending at AASHTO ≥310 με @ ≥390 με
resistance 10 Hz, 10 °C, to 50 % T 321 106 @ 106
stiffness reduction repetitions repetitions

As part of phase two, an EME mix was designed in France using South African
mix components and tested in accordance with the French specifications. The EME
mix was subjected to extensive testing at the CSIR’s pavement material testing
laboratory using South African test methods (Anochie-Boateng et al. 2010a). The
South African test results were analyzed together with the French results to develop
interim South African specifications (Denneman et al. 2011a). Subsequently, Sabita
Manual 33: Interim design procedure for high modulus asphalt was published in
January 2013 (Sabita 2013). Table 1 provides a summary of the South African
EME performance tests and interim specifications (Sabita 2013).
Phases three and four focused on the validation of EME technology through an
LTPP monitoring study and finalization of EME guidelines/specifications respec-
tively. The intention was to revisit the EME interim performance specifications
based on the outcomes of the LTPP monitoring study and further testing of EME
mixes in South Africa (Table 2).
The outcomes of phases one and two of the EME technology transfer study have
been widely covered in the previous publications (Denneman et al. 2011a, b;
Nkgapele et al. 2012). This paper focuses on the outcomes of phases three and four.
Section three of the paper provides the outcomes of the LTPP monitoring study
(phase three) that was undertaken to validate the field performance of EME;
whereas section four covered the revision of EME performance specifications
(phase four).

3 South Coast Road LTPP Section

As mentioned earlier, phase three of the EME technology transfer study was to
validate the field performance of EME technology though LTPP monitoring pro-
gramme. A section of heavily trafficked South Coast Road owned by eThekwini
Long-Term Pavement Performance Monitoring … 183

municipality in the Kwa-Zulu Natal province was identified for the field validation
of EME. The project comprised the rehabilitation of the north bound lanes of South
Coast Road towards the intersection with Bayhead road, near the entrance of the
Durban harbor. The existing asphalt layer was severely rutted due to the heavy
traffic volumes entering the Durban harbor. The project was aimed to demonstrate
the ability to design, manufacture and construct EME pavements successfully
through long-term monitoring of its field performance. The developed interim
specifications/guidelines were used for the design of the EME mix, as well as for
the structural design of the pavement.

3.1 Materials, EME Mix Design and Performance

Evaluation

3.1.1 Materials and EME Mix Design

A combination of quartzite and tillite aggregates and mineral filler was used to
manufacture the EME mix using unmodified 10/20 penetration grade bitumen. A
grading with an NMPS of 20 mm was selected. Table 3 presents the design

Table 3 EME mix design grading, grading requirements and mix properties
Sieve size (mm) 25 20 14 10 7.1 5 2 1 0.600 0.300 0.150 0.075
% Passing 100 97 77 67 57 48 32 23 17 13 9 6.7
Recommended Min – – – – 47 42 25 – – – – 5.5
grading Max – – – – 66 62 38 – – – – 7.9
Mix property Design value
Bitumen content (%) 5.2
Design air voids (%) 4.6
Richness modulus (K) 3.4

Table 4 Summary of test results and interim performance requirements

Property Results Requirements
Class 1 EME Class 2 EME
Workability 2.9 % ≤10 % ≤6 %
Durability 0.8 ≥0.8 ≥0.8
Rutting resistance 0.24 % ≤1.1 % strain ≤1.1 % strain
Stiffness 24.4 GPa ≥14 GPa ≥14 GPa
Fatigue resistance 280 με @ 106 ≥310 με @ 106 ≥390 με @ 106
repetitions repetitions repetitions
184 J. Komba et al.

Fig. 1 Pavement structure

grading, along with the grading envelopes recommended in Sabita Manual 33. A
summary of the volumetric properties of the EME mix are also presented in
Table 3.

3.1.2 EME Mix Performance Evaluation

The performance of the EME mix was evaluated in the laboratory against the EME
specifications described in Sabita Manual 33. Table 4 shows a summary of the
performance tests results, along with the interim requirements contained in Sabita
Manual 33 (SABITA 2013). The mix satisfied the workability, durability and
stiffness requirements. The EME mix did not meet the fatigue requirement.
However, a decision was made to use the mix for the construction of the base layers
of the South Coast Road LTPP monitoring section (Fig. 1).

3.2 Construction of EME LTPP Section

Pavement analysis was performed using the CSIR’s Mechanistic Empirical

Pavement Analysis and Design Software (mePADS). Analysis was performed using
a dual wheel load of 20 kN per wheel, at 350 mm spacing, with a tyre inflation
pressure of 800 kPa. Based on the latest stress in motion studies this tyre inflation
pressure translates to an average 650 kPa contact stress on the pavement. Details on
the pavement structural design can be found in Denneman et al. (2011c). The EME
layer’s bearing capacity was estimated to be in excess of 100 million equivalent
standard axle loads (E80s), which is greater than the bearing capacity required for
20 years design life. Figure 2 shows the pavement structure.
The section (approximately 300 m long) was constructed in September 2011.
The existing severely rutted asphalt was milled off to a depth of approximately
Long-Term Pavement Performance Monitoring … 185

Severe rutting on the EME section prior to Condition of the rutting on the EME section four
rehabilitation years after rehabilitation

Fig. 2 Condition of the section prior and four years after rehabilitation

190 mm. The milled out asphalt was replaced by 160 mm thick EME 20 base layer,
which was constructed as two separate layers (80 mm thick each). After con-
struction of the EME base layer, a 30 mm SMA wearing course (EVA-modified
bitumen) was paved to provide fuel resistance in order to minimize potential
deformation as a result of fuel spillages from trucks.

3.3 Long-Term Pavement Performance Monitoring

The assessment of the field performance of the LTPP monitoring section consisted
of visual condition surveys, rut depth measurements, macro-texture measurements,
roughness measurements, and deflections measurements. Initial assessments were
conducted at 6-month intervals (i.e. 1, 6, 12, 18 and 24 months after construction).
A final assessment was performed 48 months (4 years) after construction at the end
of 2015. After 4 years (from September 2011 to October 2015), the traffic carried
by the section was estimated to be 20 million Equivalent Standard Axle Loads
(ESAL’s).

3.3.1 Visual Condition Survey

The visual survey was performed in accordance with the procedures stipulated in
the South African Technical Method for Highways 9 (TMH 9): Pavement man-
agement systems: Standard visual assessment manual for flexible pavements
(1992). The visual condition survey focused on surfacing assessments (texture,
voids, mechanical failures, bleeding/flushing, surface cracks, bitumen condition and
aggregate loss), structural assessments (cracks, pumping, rutting, undulation/set-
tlement, edge break, potholes, delamination and patching), and functional assess-
ment (riding quality, skid resistance, surface drainage and side drainage). No major
186 J. Komba et al.

Fig. 3 Maximum deflections measurements

distresses were observed during the visual condition assessments. Figure 2 shows
photos of the section prior to the rehabilitation and four years after rehabilitation. It
can be seen that the section is generally in good condition. There is no indication of
a significant structural damage to the section, despite having been heavily trafficked
for a period of four years.

3.3.2 Deflection Measurements

A Falling Weight Deflectometer (FWD) was used for deflection measurements. The
deflection measurements were performed using a standard load of 40 kN. The
pavement surface temperatures were also captured. Figure 3 shows maximum
deflection measurements, along with the measured pavement surface temperatures.
Overall, very low deflection values were measured during the entire monitoring
period, indicating that the pavement structure is still in sound/good condition.
Typically, a pavement structure with maximum deflections of less than 0.3 mm is
considered to be stiff (Horak 2008). It is also noted that higher deflection values
were measured after 6 months as compared to other surveys measurements (1, 12,
6, 18, 24 and 48 months after construction). It is also observed that higher tem-
peratures were measured during the 6 month survey. It is difficult to explain the
differences in the deflections, although temperature could have played a role.

3.3.3 Rut Depth Measurements

A digital high-speed laser profiler was used for rut depth measurements. Figures 4
and 5 show rut depth measurements performed on the right and left wheel paths,
respectively. The rut measurements performed 48 months after construction show
an increasing trend on some isolated sections. However, the magnitudes of the rut
measurements are within an acceptable range considering that the section has been
heavily trafficked over a period of 4 years.
Long-Term Pavement Performance Monitoring … 187

Fig. 4 Rut depth measurements on right wheel path

Fig. 5 Rut depth measurements on left wheel path

3.3.4 Macro-Texture Measurements

Macro-texture measurements were also performed using the digital high-speed laser
profiler. Figures 6 and 7 show Mean Profile Depth (MPD) measurements performed
on the right and left wheel paths, respectively. The MPD measurements performed
48 months after construction show an increasing trend over almost the entire sec-
tion. It should however be noted that the section was overlaid with an SMA. The
observed changes on macro-texture could be limited to the SMA wearing course.

3.3.5 Roughness Measurements

Figures 8 and 9 show roughness measurements in terms of International Roughness

Index (IRI). Overall, the trends of roughness measurements during the monitoring
period are virtually identical, indicating that the overall condition of the section has
not deteriorated significantly over the 4 year monitoring period. Relatively higher
IRI values were measured at the beginning of the section. This could be due to the
construction joint between the new and existing asphalt layers.
188 J. Komba et al.

Fig. 6 Macro-texture measurements on right wheel path

Fig. 7 Macro-texture measurements on left wheel path

Fig. 8 Roughness measurements on right wheel path

Long-Term Pavement Performance Monitoring … 189

Fig. 9 Roughness measurements on left wheel path

4 Revision of EME Performance Criteria

Phase four of the EME technology transfer study focused on the revision of the
EME performance criteria. Following a successful implementation of EME tech-
nology in South Africa, the South African asphalt industry started specifying EME
for the rehabilitation of heavily trafficked road sections. Several EME mixes were
evaluated in accordance with interim specifications contained in Sabita Manual 33
(Sabita 2013). The evaluated EME mixes did not satisfy the interim fatigue criteria.
During the early stage of implementation of EME in South Africa, it also became
apparent that South African EME mixes had stiffnesses significantly higher than the
interim specifications. Following several discussion with the industry it was decided
that both the interim fatigue and stiffness criteria be revised. A study aimed at the
revision of the South African EME performance specifications was initiated in April
2014. As part of the study, the performance of an EME mix designed in South
Africa using locally available materials (aggregate and binder) was evaluated at the
CSIR’s pavement material testing laboratory in South Africa and at the Shell
European Solution Centre in France. The results from the study were analyzed
together with a database of results of other EME mixes tested previously in South
Africa to revise the South African EME fatigue and stiffness performance specifi-
cations. More details of the EME mix design and performance testing can be found
in Lawrence and Jungmann (2014), Komba and Verhaeghe (2015) and Lawrence
et al. (2015). The remainder of this section will focus on the development of the
revised fatigue and stiffness performance specifications.
As mentioned earlier, the methods used to evaluate performance of asphalt
material in South Africa differ from those used in France. Therefore, the French
EME specifications cannot be directly translated to South Africa. At the CSIR’s
pavement material testing laboratory in South Africa, four-point beam fatigue tests
were conducted on prismatic beam specimens (400 × 63 × 50 mm). The Fatigue
tests were conducted under controlled-strain loading conditions at three strain
190 J. Komba et al.

amplitude levels of 200, 390 and 600 με at a frequency of 10 Hz and a temperature

of 10 °C. The tests were performed according to the protocol developed by
Anochie-Boateng et al. (2010b), which follows procedures in AASHTO T 321
(2009), with some modifications. The failure criterion was defined as the number of
load cycles at which the initial stiffness is reduced by 50 %. In France, two point
beam fatigue tests were conducted on the same EME mix at Shell European
Solution Centre in France in accordance with EN 12697-24 standard test method.
The French fatigue tests were performed at strain levels ranging from approxi-
mately 100–175 με which is in line with their standard practice. Figure 10 shows a
plot of the strain versus number of load cycles to failure for both South African and
French results. It can be seen that the EME mix just made the French fatigue
requirement for Class 2 EME which is 130 με at one million repetitions. For the
South Africa fatigue tests results, the EME mix had one million repetitions at
approximately 270 με as illustrated in Fig. 10.
In Fig. 11, the South African results of the EME mix used in the study are
plotted together with two other mixes with have been paved in South Africa to date
and demonstrated a satisfactory field performance. The two mixes are the EME

1000
Strain (Microstrain)

µ
µ
100

South Africa test results

French test results
10
1000 10000 100000 1000000 10000000
Number of load cycles

Fig. 10 Plots of South Africa and French fatigue results

1000
Strain (Microstrain)

280 µ

100

Study EME mix

South Coast Road EME mix
N3 Calendella EME mix
10
1000 10000 100000 1000000 10000000
Number of load cycles

Fig. 11 Revised South African EME fatigue requirement

Long-Term Pavement Performance Monitoring … 191

paved on South Coast Road in Durban and the EME mix paved on the N3 Highway
in Kwa-Zulu Natal province. On average, the EME mixes reached one million
repetitions at approximately 280 με as illustrated in Fig. 11. Considering the scatter
of the EME mixes fatigue test results, it was proposed that the South African fatigue
requirement for class 2 EME should be set ≥260 με at one million repetitions. This
slightly lower fatigue criterion was proposed in order to avoid the risk of reducing
the permanent deformation performance of EME. For Class 1 EME, the criterion
was estimated based on the ratio of the French criterion for class 1 EME and class 2
EME. The South African class 1 EME criterion was conservatively set ≥210 με at
one million repetitions. It should however be noted that class 2 EME is the most
preferred in South Africa.
Table 5 provides a summary of the stiffness results of the EME study mix and
two other mixes which have demonstrated a satisfactory field performance in South
Africa. It can be seen that the stiffnesses of all the mixes are generally higher than
the interim requirement of 14 GPa. It should be mentioned that the South African
interim requirement was set to be the same as the French. The French stiffness
requirement is based two point flexural bending in accordance with EN 12697-26
standard method, whereas the South Africa stiffness requirement is based on a
triaxial compression configuration in accordance with AASHTO TP 79. It was
realized later that the flexural bending stiffness is generally lower than the stiffness
based on triaxial compression configuration set up. Considering the differences in
the French and the South African tests, as well as the results obtained from South
African EME mixes to date, a minimum stiffness requirement of 16 GPa for both
Class 1 EME and Class 2 EME was proposed for South Africa. This is considered
conservative in order to avoid the risk of reduction of permanent deformation
performance of EME due to lower mix stiffness.
Table 6 shows the revised South African EME performance requirements. The
requirements have been incorporated on the revised guidelines for the design of
EME in South Africa (Sabita Manual 33) which was published in July 2015. In
addition, structural design guidelines have been included in the newly published
Sabita Manual 33, based on the outcomes of the research work by Steyn et al.
(2015).

Table 5 Summary of stiffness results and interim performance requirements

EME mix Stiffness (GPa) Interim requirements
Class 1 EME (GPa) Class 2 EME (GPa)
Study mix 29.2 ≥14 ≥14
South Coast Road mix 24.4
N3 Calendella mix 18.5
192 J. Komba et al.

Table 6 Revised EME performance requirements

Property Test Method Requirements
Class 1 EME Class 2 EME
Workability Gyratory compactor, air ASTM ≤10 % ≤6 %
voids after 45 gyrations D6926
Durability Modified Lottman ASTM ≥0.8 0.8
D4867
Rutting Repeated Simple Shear Test AASHTO ≤1.1 % ≤1.1 %
resistance at Constant Height at 55 °C, T 320 strain strain
5000 repetitions
Stiffness Dynamic modulus test at AASHTO ≥16 GPa ≥16 GPa
10 Hz, 15 °C TP 79
Fatigue Four point bending at AASHTO ≥210 με @ ≥260 με @
resistance 10 Hz, 10 °C, to 50 % T 321 106 106
stiffness reduction repetitions repetitions

5 Conclusions and Recommendation

The paper presented the outcomes of the LTPP monitoring programme and the
development of the revised EME performance criteria, which form part of the
transfer of EME technology to South Africa. In the absence of HVS, the extremely
heavy trafficked section of the South Coast Road was selected to allow for the
accelerated validation of the South African EME design procedure. Based on the
results and discussions contained in this paper, the following conclusions and
recommendations can be drawn:
• EME could provide an optimum solution for the design of heavily trafficked
roads. It was estimated that during the 4-year monitoring period, the section
carried approximately 20 million Equivalent Standard Axle Loads (ESAL’s) and
had no significant structural damage, implying that EME can easily withstand
heavy traffic loading.
• Revised fatigue and stiffness requirements have been developed, and incorpo-
rated in the newly published Sabita Manual 33. Further monitoring of the field
performance of the EME is expected to continue in South Africa, along with
building a database of laboratory test results for South African EME mixes. The
data will inform the need for future revisions of the EME performance
requirements.
• It is recommended that EME should be encouraged to be used on heavily
trafficked roads in South Africa and elsewhere.
Long-Term Pavement Performance Monitoring … 193

Acknowledgments The construction of the EME trial section on South Coast Road was funded
by eThekwini Municipality. The monitoring of the field performance of the trial section was
funded by the Southern African Bitumen Association (SABITA). Their financial contribution is
acknowledged and gratefully appreciated.

References

AASHTO T 320. (2007). Standard method of test for determining the permanent shear strain and
stiffness of asphalt mixtures using the superpave shear tester (SST). Washington, DC:
American Association of State Highway and Transportation Officials.
AASHTO T 321. (2009). Standard method of test for determining the fatigue life of compacted
hot-mix asphalt (HMA) subjected to repeated flexural bending. Washington, DC: American
Association of State Highway and Transportation Officials.
AASHTO TP 79. (2009). Standard method of test for determining the dynamic modulus and flow
number for hot-mix asphalt (HMA) using asphalt mixture performance tester (AMPT).
Washington, DC: American Association of State Highway and Transportation Officials.
Anochie-Boateng, J., Denneman, E., O’Connell, J., & Ventura, D. (2010a). High modulus asphalt
mix technology transfer: Level 1 report of high modulus asphalt reference mix. Report
prepared for Sabita, Report number CSIR/BE/IE/ER/2010/0080/C, CSIR, Pretoria.
Anochie-Boateng, J., Denneman, E., Mturi, G., O’Connell, J., & Ventura, D. (2010b). Hot-mix
asphalt testing for the South African pavement design method. In 29th Southern Africa
Transportation Conference, Pretoria.
Asphalt Institute. (1996). Mix design methods for asphalt concreate and other hot-mix asphalt.
Superpave Series No.2. USA.
Denneman, E., Nkgapele, M., Maina, J. W., Komba, J., Verhaeghe, B. M. J. W., & Duplessis, L.
(2011a). High modulus asphalt technology transfer: Final report phase 2. Report:
CSIR/BE/IE/ER/2011/0066/C Prepared for Sabita, CSIR, Pretoria.
Denneman, E., Nkgapele, M., Anochie-Boateng, J., & Maina, J. W. (2011b). Transfer of high
modulus asphalt mix technology to South Africa. In: Proceedings of the 10th conference on
asphalt pavements for Southern Africa, Champagne Sports Resort.
Denneman, E., Nkgapele, M., & Komba, J. (2011c). Implementation of high modulus asphalt
technology at South Coast Road in eThekwini. Technical Report No:
CSIR/BE/IE/ER/2011/0048/C. CSIR, Pretoria.
Delorme, J., De La Roche, C., & Wendling, L. (2007). LPC bituminous mixtures design guide.
Director. Paris: Laboratoire Central des Ponts et Chaussees.
Horak, E. (2008). Benchmarking the structural condition of flexible pavements with deflection
bowl parameters. Journal of the South African Institution of Civil Engineering, 50(2), 2–9.
Komba, J., & Vehaeghe. (2015). Refining Enrobé à Module Élevé (EME) fatigue criteria.
Technical Report No: CSIR/BE/TIE/IR/2015/0036/C. CSIR, Pretoria.
Lawrence, D., Fremont, S., & Hornsey, B. (2015). The development of a South African high
modulus asphalt mixture using French EME Methodology. In: Proceedings of the 11th
Conference on Asphalt Pavements for Southern Africa. Sun City.
Lawrence, D., & Jungmann, L. (2014). Technical Service request report No. TS EUSC-2014-147.
Shell European Solution Centre, France.
Nkgapele, M., Denneman, E., & Anochie-Boateng. (2012). Construction of a high modulus
asphalt (HiMA) trial section eThekwini: South Africa’s first practical experience with design,
manufacturing and pavement of HiMA. In Proceedings of the 31st Annual Southern Africa
Transport Conference, Pretoria.
Sabita. (2013). Interim design procedure for high modulus asphalt: Sabita Manual 33. ISBN
978-1-874968-62-7. Howard Place, South Africa.
194 J. Komba et al.

Sabita. (2015). Design procedure for high modulus asphalt (EME): Sabita Manual 33. ISBN
978-1-874968-62-7. Howard Place, South Africa.
Steyn, W. J. M., Maina, J. M., Myburgh, P., & Solomons, S. (2015). Evaluation of EME transfer
functions and critical layer thicknesses. In Proceedings of the 11th Conference on Asphalt
Pavements for Southern Africa. Sun City.
TMH 9. (1992). Pavement management systems: Standard visual assessment manual for flexible
pavements. CSRA. Pretoria, South Africa.
Vavrik, W. R., Huber, G., Pine, W. J., Carpenter, S. H., & Bailey, R. (2002). Bailey method for
gradation selection in hot-mix asphalt mixture design. Transportation Research E Circular,
Number E-C044.
Optimum Properties of Geocell
Reinforcement for Sustainable
Low-Volume Paved Roads

B. Bortz and M. Hossain

Abstract Geocells are three-dimensional honeycomb-like geosynthetic structures

filled with available geomaterials that vastly improve shear strength of those
materials. Geocells provide a possible viable solution for thinly paved roads that use
marginal infill geomaterials. The objective of this study was to find an optimum
geocell design that utilizes various infill materials and a thin hot-mix asphalt
(HMA) overlay. To achieve this study objective, four pavement test sections were
constructed at the accelerated pavement testing (APT) facility of the Civil
Infrastructure System Laboratory at Kansas State University. Three out of the four
lanes contained geocell-reinforced bases with three individual infill geomaterials:
crushed limestone, quarry by-products, and recycled asphalt pavement. The fourth
test lane, the control section, consisted of a crushed stone base. All sections were
heavily instrumented. Repeated loads (80 kN, single axle) were applied using an
APT load assembly. Sections with an HMA layer of 50 mm reached the failure
criteria of a 12.5 mm rut depth after 10,000 passes due to excessive stress in the
subgrade. Redesigned sections with HMA overlay of 100 mm carried more than
1,000,000 passes. Numerical simulation of the APT tests was done using the finite
element model. The optimum infill geomaterial property, geocell height, and
overlay thickness were obtained from this simulation.

Keywords Gerocell
RAP
Low-volume roads
Numerical simulation
APT

B. Bortz (&)
M. Hossain

Department of Civil Engineering, Kansas State University, Manhattan,
KS 66506, USA
e-mail: [email protected]
M. Hossain
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 195

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_13
196 B. Bortz and M. Hossain

1 Introduction

Geocells are three-dimensional (3-D) honeycomb-like geosynthetic structures filled

with an infill of granular materials, as shown in Fig. 1. This type of geosynthetic
has recently garnered increased interest as reinforcement for base courses of
pavement structures for low-volume roads. Geocells are made of strips of polymer
sheet or geotextile connected at staggered points so that a large honeycombed mat is
formed when the strips are pulled apart. Geocells mainly provide physical con-
tainment of an infill geomaterial. Geocells can be shipped to the job site in a
collapsed configuration, thereby increasing shipping efficiency. At the job site,
geocells are placed directly on the surface of the subsoil and propped open in an
accordion fashion with an external stretcher assembly. They are then filled with
geomaterials and compacted (Koerner 1994).
Because of their inherent (3-D) structure, geocells currently experience more
widespread use for confinement applications than any other planar geosynthetic
reinforcement (Yuu et al. 2008). However, research studies on geocells have pri-
marily demonstrated use of geocells for increasing bearing capacity and reducing
settlement of soft soil foundations (Sitharam et al. 2005; Dash et al. 2001, 2003,
2004). Although a few studies have demonstrated that geocells can provide good
soil confinement and enhance the performance of base course on weak subgrade,
acceptance of geocells for unpaved and paved roads is still limited due to a lack of
research and accepted design methods (Yuu et al. 2008).

Fig. 1 Geocellular confinement

Optimum Properties of Geocell Reinforcement … 197

In 2009, the University of Kansas and Kansas State University (KSU) conducted
accelerated pavement testing (APT) research on unpaved geocell-reinforced bases
over weak subgrade. Three base materials were used in the study: crushed stone,
reclaimed asphalt pavement (RAP), and quarry waste (QW) (Pokharel et al. 2011;
Han et al. 2011). The study resulted in the following conclusions for
geocell-reinforced unpaved roads:
1. A 170 mm geocell-reinforced base can outperform a 300 mm crushed stone
base.
2. RAP is the best performing infill material when compared to crushed stone and
quarry waste.
3. Geocells increase the stress distribution angle.
4. A thick (50–75 mm) cover is needed to minimize traffic damage to geocells.
Geocell performance in paved roads has not previously been studied.

2 Study Objective

The objective of this study was to determine optimum infill geomaterial property,
geocell height, and hot-mix asphalt (HMA) overlay thickness using numerical
simulation of APT.

3 Study Approach

In order to achieve the study objective, four lanes of pavement test sections were
constructed at the Civil Infrastructure System Laboratory (CISL) at KSU. Three of
the four lanes were geocell-reinforced and contained unique infill materials. The
fourth test lane, the control section, consisted of a crushed stone, AB-3 base. Three
types of infill geomaterials (AB-3, QW and RAP) were used in this study. In the
first test, the control and QW lanes were loaded with 70,000 repetitions and the
AB-3 and RAP lanes were loaded with 50,000 repetitions of an 80 kN single-axle
load. Due to excessive rutting, a thin overlay was placed over all sections. For the
second test, a cross section with thicker HMA was designed and constructed using
infill materials similar to those in the first experiment. These sections were loaded
with more than 1,000,000 repetitions, but no failure happened. ABAQUS, com-
mercially available finite element (FE) software, was used to numerically simulate
the geocell-reinforced sections under APT. Material properties obtained from
testing and actual geometry and boundary conditions were used in the models.
Compared to the high costs of APT testing, validation by numerical modeling
resulted in optimum geocell design at a low cost.
198 B. Bortz and M. Hossain

4 Facility and Equipment

The CISL houses an APT machine and three pits for constructing test sections. The
reaction frame of the APT machine covers a distance of 12.8 m and applies a
single-axle load of 80 kN with air bag suspension on dual tires. The wheel assembly
is belt driven by a 20 HP electric motor, and the load is controlled by hydraulic
pressure. Tire pressure can be variable, but 552 kPa was used in this study. The pits
have approximately identical dimensions of 6.1 m long, 4.9 m wide, and 1.8 m
deep. The moving wheel has a frequency of 0.167 Hz (i.e., 6 s/pass) at a speed of
11.3 km/h (Lewis 2008).

5 Material Properties

5.1 Geocells and Geotextile

Geocells used in this study were NEOLOY™ polymeric alloy (nano-composite

alloy of polyester/polyamide nano fibers, dispersed in a polyethylene matrix) (Han
et al. 2011). The geotextile used as a separator between the subgrade and the base
was a 3.5 oz nonwoven geotextile. Details of subgrade, infill geomaterials (AB-3,
QW, and RAP), and HMA properties can be found in Bortz et al. (2012).

5.2 Accelerated Pavement Testing Results

5.2.1 Test Lanes

In this study, two pits were divided into two lanes each (6.1 m long by 2.45 m
wide) for a total of four lanes. For the first test, the subgrade material for each lane
was clay (A-7-6) compacted to a California Bearing Ratio (CBR) of approximately
6 %. In the second test, the subgrade for the test sections was compacted to a CBR
of 12 %, which was based on analysis of in situ Dynamic Cone Penetrometer
(DCP) data from the Kansas Department of Transportation (KDOT) projects.
A nonwoven geotextile was used to separate the base and the subgrade. Based on
recommendations from Pokharel et al. (2011) and Han et al. (2011), the geocells
were laid out in a near-circular pattern with a dimension 250 mm in the wheel
direction (also the seam direction) and 210 mm in the transverse direction. In
addition, the Neoloy optimum geocell height found in that study was 75 mm, so the
height of the geocells in the first test was 75 mm. The geocells were filled and
compacted and then covered by 25 mm of infill geomaterial. However, this cover
was considered too thin because geocells were often exposed. The reinforced base
layer was paved with 50 mm of HMA of a Superpave mixture with 12.5 mm
Optimum Properties of Geocell Reinforcement … 199

nominal maximum aggregate size (NMAS). The second test consisted of a thicker
cross section of the reinforced and HMA layers. The geocell heights were increased
to 100 mm with a 50 mm infill cover. The thicker infill cover improved compaction
of the infill materials. HMA layer thickness was increased to 100 mm in the second
experiment, as shown in Fig. 2.
The APT machine is able to wander laterally while applying passes, so a wander
of ±152 mm was programmed into the machine. The wander was applied in a
truncated normal distribution. A full wander of +152 to −152 mm took 676 passes
to complete.

5.2.2 Instrumentation

All four lanes were instrumented with pressure cells on top of the subgrade and two
strain gauges at the bottom of the HMA layer. Thermocouples were also placed
below the HMA layer. Neoloy geocells were instrumented with five strain gauges
per lane.
The pressure cells were Geokon Model 3500 cells, and the H-Bar strain gauges
were Texas Measurements TML-60-2L gauges epoxied to two pieces of aluminum
as suggested by Lewis (2008). Type T thermocouples were used. Strain gauges
embedded on the body of the Neoloy geocells were Vishay C2A-06-250LW-120.
During placement of the HMA layer, some H-Bar gages were damaged. Details of
the instrumentation, including data acquisition, can be found in Bortz et al. (2012).

Fig. 2 Second test lane cross sections

200 B. Bortz and M. Hossain

6 Numerical Simulation

Numerical simulation evaluated mechanistic responses in the base and subgrade

during APT tests. Simulation was done by finite element (FE) modeling using
ABAQUS.

6.1 Material Properties

Material properties were determined from the laboratory and in situ tests. During
simulation, the base material was modeled with Mohr-Coulomb plasticity. HMA
layers were considered to be linear elastic, and layer moduli were backcalculated
from Falling Weight Deflectometer (FWD) data. Geocells were modeled as elastic
materials because no damage to the geocells was observed during testing and the
measured strains did not indicate nonlinearity or permanent deformation. Material
properties used in this study are tabulated in Table 1. The geocells were modeled as
diamonds to simplify meshing while maintaining basic reinforcing functionality.
Yang (2010) and Leshchinsky and Ling (2013) successfully modeled multiple
geocells as diamonds.

Table 1 Material properties for numerical simulation

Material Lane CTL QW RAP AB-3
Test# 2 2 2 2
Subgrade Density (tonne/mm3) 1.56E−06 1.56E−06 1.56E−06 1.56E−06
E (MPa) 68 52 58 59
v 0.45 0.45 0.45 0.45
Friction angle 0 0 0 0
Dilation angle 0 0 0 0
Cohesion (MPa) 0.104 0.104 0.104 0.104
Base Density (tonne/mm3) 2.03E−06 1.95E−06 1.81E−06 1.97E−06
E (MPa) 308 328 140 193
v 0.4 0.4 0.4 0.4
Friction angle 47.2 47.2 37 47.2
Dilation angle 17.2 17.2 7 17.2
Cohesion (MPa) 0.0047 0.005 0.05 0.0047
Geocell E (MPa) 550 550 550 550
v 0.45 0.45 0.45 0.45
HMA Density (tonne/mm3) 2.26E−06 2.26E−06 2.26E−06 2.26E−06
E (MPa) 6468 4867 4947 3848
v 0.3 0.3 0.3 0.3
Optimum Properties of Geocell Reinforcement … 201

Fig. 3 Boundary conditions for APT models

6.2 Model Boundary Condition

Only a quarter of the APT test pit was modeled in order to decrease required
computational time (Fig. 3). Because the pit is surrounded by concrete, displace-
ment was set for the bottom and two sides (back and left) of the model. The left side
was restrained from moving in the x-direction and the bottom was restrained from
moving in the z-direction. The front and right sides used symmetrical boundary
conditions; the front used symmetry in the y-direction and the right side used
symmetry in the x-direction.

6.3 Element/Mesh

Solid materials (HMA, base material, and subgrade) were meshed using an 8-noded
linear reduced integration hexahedral brick element (C3D8R). However, a trade-off
happens between accuracy and computational size: coarse meshes tend to be
inaccurate but fine meshes increase computational time. For the geocell-reinforced
section simulation, a balance was found between the HMA layer (shown in
Fig. 4a), base layer (shown in Fig. 4b), and subgrade (shown in Fig. 4c) with 1716,
1449, and 6279 elements, respectively. Tie constraints were used at the interfaces
between the HMA and the base material and between the subgrade and the base
material to speed up computation. Geocells were meshed using 10,080 S4R 4-node
doubly-curved thin or thick shell reduced integration elements, as shown in Fig. 4d.
Shell elements can be used when the ratio of one dimension is higher than the other
dimensions (ABAQUS 2004).
202 B. Bortz and M. Hossain

(a) HMA

(b) Base

(c) Subgrade

(d) Geocell

Fig. 4 Finite element mesh

Optimum Properties of Geocell Reinforcement … 203

Fig. 5 Embedded geocells in base layer

6.4 Geocell Embedment

An embedded region was used to place the geocells in the base layer. Embedded
regions are a group of elements within a “host” region. The embedded region
allows shell elements to be embedded into solid elements, as shown in Fig. 5. The
infill geomaterial was sandwiched between the HMA layer and the subgrade,
thereby preventing any movement of these materials. Embedded elements were
constrained by the response of the host elements; therefore, no contact friction was
attributed to the geocell wall. However, because the infill geomaterial had no room
to move, the assumption was made that friction between the geocell wall and the
infill provided little or no help to the reinforcement. This assumption was found to
be true during forensic studies after testing was complete (Bortz 2015). Geocells
were completely embedded in a slab-like structure (base layer). This embedment
differs in unpaved roads where the infill is able to move. Infill geomaterials in
unpaved roads can be pushed out of the cells; therefore, friction is an important
consideration. Embedded region elements show slab-like behavior. Since the
control section had no geocells, the model used the same model configuration but
without the embedded geocell.

6.5 Loading

Loading of the model occurred over the area of the tire print at a given point in the
load cycle. Symmetry was used for half of the tire print, as shown in Fig. 6. This
type of loading represented results obtained from the sensors. A static load was
applied using Abaqus/Explicit. Abaqus/Explicit analysis models highly nonlinear
behavior of materials better than Abaqus/Standard (Implicit) analysis. Abaqus/
Standard has convergence issues and uses very small time increments in soil
204 B. Bortz and M. Hossain

Fig. 6 Loading of model

analysis due to yielding of the soil (ABAQUS 2004). The base material in this
study had cohesion less than 10 % of the applied load; however, the creep material
model was not available in Abaqus/Explicit, so the HMA layer was modeled as a
linear elastic material. The load was applied as pressure over an area equal to the
tire imprints.
The total force of 80 kN (18 kip) was applied to the pavement through two sets
of dual tires. A tire pressure of 552 kPa (80 psi) was maintained during testing.
Rectangular tire imprints were assumed in this study, resulting in a total tire imprint
of 208 mm (8.2 in.) wide (measured tire imprint width) × 174.5 mm (6.9 in.) long
(calculated tire imprint length). Tire imprint length was divided into two due to
symmetry. One load cycle on a unit tire imprint was calculated to take 0.05 s to
pass. The step in the numerical analysis placed a pressure of 552 kPa (80 psi) on the
tire print for 0.05 s.

6.6 Pavement Responses from Numerical Analysis

Results for various simulation runs are shown in Table 2. During analysis, strain on
the geocell was analyzed at two locations. A path was set along five geocells where
strain gauges were located, as shown in Fig. 7. Maximum and minimum strain
Optimum Properties of Geocell Reinforcement … 205

Table 2 Numerical response analysis results

Lane HMA Geocell path Geocell path Geocell Geocell Subgrade
strain strain max strain min max strain min strain stress (kPa)
Microstrain
Control 148 28
QW 159 313 −57 393 −140 43
RAP 162 220 −63 328 −180 40
AB-3 186 334 −82 407 −194 41

Fig. 7 Geocell path

Table 3 HMA strain comparison

Control QW RAP AB-3
Microstrain Microstrain Microstrain Microstrain
(below HMA) (below HMA) (below HMA) (below HMA)
Numerical 148 159 162 486
analysis
Measured 140 146 209 220
% Difference 5.41 8.18 −29.0 54.7

values along the path were first obtained and then maximum and minimum strain
values for the entire geocell layer were obtained.
Results affirmed the intuition that thicker sections are stronger. When less stress
is transmitted to the subgrade, lower strains would also result in the geocell and
HMA layers. Table 3 compares numerical analysis results and measured results.
The control lane model yielded results that are close to the measured ones; however,
numerical analysis tended to overpredict HMA strain.
206 B. Bortz and M. Hossain

Table 4 Subgrade stress comparison

Control QW RAP AB-3
Pressure on Pressure on Pressure on Pressure on
subgrade subgrade subgrade subgrade
Numerical analysis (kPa) 28.0 43.0 40.0 41.0
Sensor 1 2 3 4 1 2 3 4
Measured (kPa) 32.2 22.9 52.8 54.4 46.4 81.3 35.2 48.2
% Difference −14.9 18.1 −22.7 −26.4 −16.0 −103.4 14.2 −17.5

A comparison of subgrade stress results from the numerical analysis and the
measured values is shown in Table 4. Subgrade stress results varied depending on
location due to base layer construction and initial layer compaction due to loading.
AB-3 had unique results, with one sensor measuring a response that was below the
predicted value and one measured response that was above the predicted value.
Numerical analysis underpredicted subgrade pressure on all RAP sections.

Table 5 Parametric study material properties

Parameter Base modulus Geocell height
Subgrade
Density (tonne/mm3) 1.52E−06 1.52E−06
E (MPa) 45 45
V 0.45 0.45
Friction angle 0 0
Dilation angle 0 0
h (mm) 1274 1274
Cohesion (MPa) 0.104 0.104
Base
Density (tonne/mm3) 2.03E−06 2.03E−06
E (MPa) Variable 150
V 0.4 0.4
Friction angle 47.2 47.2
Dilation angle 17.2 17.2
h (mm) 150 Geocell height + 50 mm
Cohesion (Mpa) 0.0047 0.0047
Geocell
E (MPa) 550 550
V 0.45 0.45
h (mm) 100 Variable
HMA
Density (tonne/mm3) 2.26E−06 2.26E−06
E (MPa) 4000 4000
V 0.3 0.3
h (mm) 100 100 or 150
Optimum Properties of Geocell Reinforcement … 207

6.7 Parametric Studies of Numerical Simulation

In order to determine the effect of infill/base material modulus and geocell height, a
parametric study was performed using the FE model. The variables were limited to
the base layer. Modulus of the base material provided insight into the base material
quality and compaction level in the geocells. Analysis used a range of base material
elastic modulus from 25 MPa (3626 psi) to 500 MPa (72,519 psi).
Increasing base modulus initially showed less stress on the subgrade layer and
then leveled off. Strain on the geocells increased as the base material modulus
increased up to 200 MPa (29,008 psi), followed by a decrease. Infill geomaterial in
the geocell significantly affected the effectiveness of geocell; however, the change
due to increase in modulus of base material became less prominent, indicating that
marginal and high-quality geomaterials can serve equally well as infill geomaterials.
In the geocell height study, ratios of geocell height, hGC, and two HMA thick-
nesses [150 mm (6 in.) and 100 mm (4 in.)], hHMA, were investigated. In the APT
test sections, a 50 mm (2 in.) cover was determined to be the minimum required
cover for constructability. During this study, the height of the base layer was
maintained at 50 mm (2 in.) over the height of the geoocell. Other parameters in the
model remained constant, as shown in Table 5. Study results showed that vertical
stress on the subgrade decreased or remained the same since the ratio of geocell
height to HMA thickness increased from less than 1 to 2. For both thickness ratios,
vertical stress on the subgrade increased since the ratio was greater than 2, and then
decreased significantly, as shown in Fig. 8. Strain in the geocells decreased as the
ratio increased, but a decrease in the benefit of reduced strain was observed as the
ratio exceeded 2 (Bortz 2015).

Fig. 8 Subgrade stress with varying geocell height

208 B. Bortz and M. Hossain

7 Conclusions

In this study, three polymeric alloy geocell reinforced bases were tested under APT,
with QW, RAP, and AB-3 as infill geomaterials and one control section with AB-3
base course. Optimum infill geomaterial property, geocell height, and HMA overlay
thickness were obtained by numerical simulation of the APT. The following con-
clusions were drawn from this study:
1. Higher moduli of infill geomaterial showed a decrease in subgrade stress up to a
point and then leveled off.
2. Strain on the geocells increased as the base material modulus increased up to
200 MPa (29,008 psi), followed by a decrease. Infill geomaterial in the geocell
significantly affects the effectiveness of geocell; however, increased base layer
modulus becomes less influential in determining geocell height.
3. Vertical stress on the subgrade layer decreased or remained constant because the
ratio of geocell height to HMA layer thickness increased from less than 1 to 2.
In both layer thickness cases, vertical stress on the subgrade increased because
the ratio was greater than 2, followed by a significant decrease.
4. Strain on the geocells decreased as the ratio of geocell height to the HMA layer
thickness increased. The benefit of reduced strain decreased as the ratio
exceeded 2.

Acknowledgments This project was funded by the Midwest States Accelerated Pavement
Testing Pooled Funds Program. The participants include KDOT, IADOT, MODOT, and
NYSDOT.

References

ABAQUS. (2004). User’s manual: Version 6.5. Abaqus. Inc., Pawtucket, RI.
Bortz, B. (2015). Geocellular confinement systems in low-volume roads. Ph.D. Dissertation,
Department of Civil Engineering, Kansas State University, Manhattan, Kan.
Bortz, B., Hossain, M., Halami, I., & Gisi, A. (2012). Accelerated pavement testing of low-volume
paved roads with geocell reinforcement. In Presented at the 91st Annual Meeting of the
Transportation Research Board, National Research Council, January 22–26, 2012,
Washington, D.C. and published in the compendium DVD.
Dash, S. K., Rajagopal, K., & Krishnaswamy, N. R. (2004). Performance of different geosynthetic
reinforcement materials in sand foundations. Geosynthetics International, 11(1), 35–42.
Dash, S. K., Rajagopal, K., & Krishnaswamy, N. R. (2001). Strip footing on geocell reinforced
sand beds with additional planar reinforcement. Geotextiles and Geomembranes, 19(8), 529–
538.
Dash, S. K., Sireesh, S., & Sitharam, T. G. (2003). Model studies on circular footing supported on
geocell reinforced sand underlain by soft clay. Geotextiles and Geomembranes, 21(4), 197–
219.
Han, J., Pokharel, S. K., Yang, X., Manandhar, C., Leshchinsky, D., Halahmi, I., et al. (2011).
Performance of geocell-reinforced RAP bases over weak subgrade under full-scale moving
wheel loads. Journal of Materials in Civil Engineering ASCE, 23(11), 1525–1534.
Optimum Properties of Geocell Reinforcement … 209

Koerner, R. M. (1994). Designing with Geosynthetics (3rd ed.). New Jersey: Prentice Hall.
Leshchinsky, B., & Ling, H. (2013). Effects of geocell confinement on strength and deformation
behavior of gravel. Journal of Geotechincal and Geoenviromental Engineering, 139(2),
340–352.
Lewis, P. (2008). Lessons learned from the operations management of an accelerated pavement
testing facility. In Proceedings of the 3rd International Conference on Accelerated Pavement
Testing, Madrid, Spain.
Pokharel, S. K., Han, J., Manandhar, C., Yang, X. M., Leshchinsky, D., Halahmi, I., & Parsons, R.
L. (2011). Accelerated pavement testing of geocell-reinforced bases over weak subgrade. In
Paper presented at the 10th International Conference on Low-Volume Roads, July 24–27,
2011, Lake Buena Vista, Florida.
Sitharam, T. G., Sireesh, S., & Dash, S. K. (2005). Model studies of a circular footing supported
on geocell-reinforced clay. Canadian Geotechnical Journal, 42(2), 693–703.
Yang, X. (2010). Numerical analyses of geocell-reinforced granular soils under static repeated
loads. Ph.D. Dissertation, University of Kansas, Lawrence.
Yuu, J., Han, J., Rosen, A., Parsons, R. L., & Leshchinsky, D. (2008). Technical review of
geocell-reinforced base courses over weak subgrade. In Proceedings of the First Pan American
Geosynthetics Conference and Exhibition, 2–5 March 2008, Cancun, Mexico.
Perspectives on Trends in International
APT Research

W. JvdM Steyn and F. Hugo

Abstract The National Academies sponsored a series of three APT Syntheses

since 1996. The last synthesis was conducted in 2011. This was followed by the
fourth, most recent international APT conference that was held in Davis, California
in 2012. It is thus now an opportune time to look back at the information identified
in these syntheses, the trends predicted and the actual development since 2011. The
paper starts with a brief evaluation of the trends identified in the 2011 synthesis,
identifying those that were deemed to be new and receiving attention in the
immediate future. This is compared to the APT work published in the last 5 years to
compare the generation of new knowledge in the area of APT, and a discussion and
analysis of the trends. At the same time, the synthesis of 2004 is scrutinized to
capture aspects that may still have been left without being addressed. Finally, an
opportunity is provided for updating the database of international APT facilities
with respect to the new international APT programs that were initiated since the
2011 synthesis.

1 Introduction

Accelerated Pavement Testing (APT) covers the controlled application of wheel

loads to layered, structural pavement system to determine pavement response and
performance under controlled, accelerated, accumulation of damage loading in a
compressed time period, while issues around environmental effects on the pavement
are controlled and measured. APT systems can be divided into full-scale systems
(f-sAPT) and small-scale systems. F-sAPT systems are those where a standard truck
tire (or combination of tires) is used for applying the loads to the pavement while

W. JvdM Steyn (&)

University of Pretoria, Pretoria, South Africa
e-mail: [email protected]
F. Hugo
Stellenbosch University, Stellenbosch, South Africa
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 211

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_14
212 W. JvdM Steyn and F. Hugo

small-scale systems are those where a scaled-down version of a truck tire and tire
load are applied to a pavement system.
The objective of APT is to apply traffic loads and sometimes also environmental
effects to a pavement at an accelerated rate compared with normal loading, and to
determine the reaction of the pavement and its constitutive layers to this loading in a
shorter time than would normally occur on a pavement under typical traffic and
environmental conditions. Trafficking is attained through the repeated loading by a
set of truck tires over a short section of pavement, often at increased loads. These
increased load levels causes the number of traffic loads applied to the pavement to
be multiplied by a factor of typically between 1 and 40 (care is required to consider
the effect of this action on the failure mechanisms of a pavement). Through this
process the response of the pavement can be quantified during a much shorter time
than would be the case if a normal road were monitored under standard load
applications.
APT forms a vital link between the laboratory evaluation of materials used in
pavement layers and the field behavior of these materials when combined into
pavement structures. For many years APT provided pavement engineers with
knowledge that improved their understanding of pavement materials and structures,
as well as their behavior under typical traffic and environmental loading. It formed
the basis for developing various theories about pavement behavior and supports
most of the current pavement design methods. Hugo et al. (1991) developed Fig. 1 to
indicate the relative context of APT within the broad basis of pavement engineering.
Although APT on its own can provide some insight into the performance of a

Fig. 1 Interrelationship between pavement engineering facets that collectively and individually
contribute to knowledge (Hugo et al. 1991)
Perspectives on Trends in International APT Research 213

pavement, it is acknowledged that APT programs should be supplemented with

laboratory testing and Long Term Pavement Performance (LTPP) programs to gain
the most benefit. This has gained momentum over the past decade.
The National Cooperative Highway Research Program (NCHRP) develops
regular synthesis documents on various highway-related issues. In terms of APT,
the first of these was published in 1996 (Metcalf 1996), followed by the second
synthesis in 2004 (Hugo and Martin 2004a, b) and the third in 2012 (Steyn 2012).
These syntheses contributed to the body of knowledge by evaluating developments
and advances around APT. The objective is to synthesize current international
developments in APT and identify gaps in knowledge and future research needs for
APT. This paper focuses on developments since 2011 (last year of data incorpo-
rated into NCHRP 433) based on available conference and journal publications,
with the focus on analysis of trends in APT.

2 NCHRP 433

2.1 Findings

NCHRP 433 (Steyn 2012) was based on an international questionnaire covering

relevant aspects of f-sAPT and the published conference and journal papers over the
period 2000–2011. Thirty-eight US State DOTs, twenty-nine f-sAPT programs and
five specialists responded to the questionnaire for a total of seventy-two unique
responses.
Overall, the synthesis found that the judicious use of f-sAPT contributes to and
supports the body of knowledge regarding the way that pavement materials and
structures react to controlled traffic and environmental loads. Through well-planned
studies the f-sAPT work highlighted strategic findings that provide important
information to the pavement engineering community to ensure the sustainable and
efficient supply of cost effective pavement-related infrastructure. The importance of
f-sAPT was perceived as high with a major role to be played in pavement structure
and basic materials research. The future of f-sAPT was mainly perceived as
growing and being a normal part of pavement research operations, benefiting
improved structural and material design methods, performance modeling and
evaluation of novel materials and structures. It was evident that many programs
share their facilities and data in order to expand the database that their specific
research is founded upon which. In this regard the formation of associations of
f-sAPT users with the general objective of improving the cost effectiveness of the
overall programs through cooperative efforts of program planning, data analysis and
device improvements is evident.
A wide scope of topics was addressed in the research conducted by the various
programs, with traditional focus areas such as Hot Mix Asphalt (HMA) response to
applied loads still quite evident. However, the data also showed that topics related
214 W. JvdM Steyn and F. Hugo

to environmental issues, such as the increased use of Warm Mix Asphalt

(WMA) and Recycled Asphalt Products (RAP) were receiving growing attention.
Another focus was on the extension of pavement life through application of HMA
overlays and Ultra-Thin Whitetopping (UTW). Evaluation of materials models for
newer pavement design methods (such as the Mechanistic Empirical Pavement
Design Guide (MEPDG) and California Mechanistic Empirical design method
(CalME)) was also quite popular.
F-sAPT aims to evaluate pavement sections under a range of loading and
environmental conditions in order to improve the knowledge of the potential per-
formance of the pavement layers and structure under a full range of operational
conditions. The majority of respondents relate their f-sAPT data to pavement
temperature and ambient air temperature, while also actively managing these
parameters during tests. This trend may be related to the majority of tests being
conducted on temperature sensitive HMA materials. New focus was placed by a
number of researchers on the major effects that tire contact stresses and loading
conditions have on pavement response, and the active management of this
parameter is visible in the published literature originating from many programs.
The improved characterization of loading conditions was mirrored by the use of
more complicated materials models that can react to these input conditions and
provide improved models of the materials’ load responses. Many programs actively
focused on the validation of the models incorporated in the MEPDG and CalME
pavement design processes, thereby reducing the risk involved in pavement design
as more appropriate parameters are incorporated into the design, and the effect of
each of the parameters are better understood.
The increased use of Finite Element Modeling (FEM) for analysis of moving
loads (as opposed to static load analysis) where factors such as mass inertia and
stress rotation are incorporated into the model, the increased use of materials
models that are not simply linear elastic, but which incorporate the effects on
non-linearity, viscosity and environmental sensitivity (i.e. moisture and tempera-
ture), the increased use of detailed definition of the applied loads in terms of both
load history and contact stress patterns and the increased cognizance given to the
effects of the environment on pavement response are prominent in terms of f-sAPT
modeling.
It appears that a virtual process is driven on several fronts where improved
computing technology allows the complexity of calculations to increase without
becoming too time- and resource consuming. Concurrently, the understanding of
materials properties are improving with the parallel development of appropriate
laboratory and field instruments and tests to obtain these parameters for different
materials, and the subsequent modeling is improved through the combination of
these factors. All facets that had been identified in NCHRP 325 and presented in the
APT 2004 Overview (Hugo 2004) have been receiving attention.
A major part of the modeling was still focused only on the surfacing layers and
that the effect and contribution of lower materials layers are generalized and
Perspectives on Trends in International APT Research 215

simplified, although these effects may sometimes affect the surfacing and other upper
layers significantly. It is specifically the strength-balance of the pavement that often
appears to be ignored in test planning and modeling.
Respondents viewed improved structural and material design methods, evalua-
tion of novel materials, improved performance modeling and the development of
performance related specifications as the major benefits of f-sAPT. Analogous to
this, the changed pavement engineering world focused on proving new techniques
and materials and development of a fundamental understanding of pavement
structures.
The most significant strategic level findings from f-sAPT in the last decade focus
on issues around materials characterization, pavement modeling, pavement
behavior and performance, pavement design method development and calibration,
benefits of specific materials and technologies, economic impacts of f-sAPT pro-
grams, calibration of pavement design methods, development of databases of
information on pavement performance that are shared between different pavement
research programs, cost savings through implementing f-sAPT and the development
of improved instrumentation and analysis methods.
In terms of more practical examples, issues such as an improved understanding
of failure mechanisms of top-down cracking, critical strain limits in HMA, the effect
of adequate layer compaction, variability of materials and layer properties,
improved understanding of the links between various materials’ laboratory and field
behavior and the effect of various real environmental conditions and traffic on
pavement behavior and performance are seen as major international findings.
The evaluation of economic benefits of f-sAPT has come to the forefront during
the past decade with more programs reporting attempts at performing Benefit Cost
Ratio (BCR) type evaluations of their research programs. It appears that the general
international economic conditions forces researchers to prove the benefit of their
research much more and identify, analyze and quantify the direct and indirect
benefits obtained from f-sAPT. The majority of programs are still only conducting
BCR analyses after the research has been completed (43.5 %), while 17.4 % of
respondents indicated that they perform BCRs as an input in the research planning.
Estimates of BCR from respondents ranged broadly between 1.4 and 11.6, although
some respondents to the questionnaire indicate that their perception of the BCR for
their programs is over 30 [in NCHRP 325 values were reported as being between 10
and 15 to one for three major APT entities (Hugo 2004)].
It is evident that the f-sAPT community is moving towards the future with a
focus on calibration of f-sAPT outputs with in situ pavement data, specifically with
the view of incorporating environmental and real traffic issues that cannot be
modeled using f-sAPT. Many of the trends identified in the synthesis are old issues
that were known to the pavement engineering community for a long time, but which
did not receive the required attention in the research and testing environment
(Hugo and Martin 2004a, b).
Questionnaire respondents indicated that issues such as a more detailed focus on
vehicle-pavement interaction (including improved load and contact stress models),
environment-pavement interaction (including climate change issues), development
216 W. JvdM Steyn and F. Hugo

of and improvements in performance related specifications, improved MEPDG

validation, evaluation of sustainable pavement solutions (energy efficient tech-
nologies and re-use of available infrastructure) and improved reliability in pave-
ment design are important future focus areas.

2.2 Trends

APT follows an evolutionary approach where new developments in technology

(machines and sensors) is combined with new insights into pavement behavior and
performance to expand knowledge and understanding of pavement science. In this
regard it is prudent to keep monitoring existing and developing trends in the field.
NCHRP 325 (Hugo and Martin 2004a, b) identified the following four items for
future research:
• in situ field performance of pavements tested in f-sAPT for future comparative
studies;
• closer association between f-sAPT and in service pavement evaluations formal
SHRP Long Term Pavement Performance (LTPP) studies and pavement man-
agement systems to validate and evaluate f-sAPT results;
• exploration of vehicle-pavement-environment interaction to enhance the ability
to do quantitative performance prediction of different pavement structures under
specific conditions, and
• prudent use of available information and collaborative research efforts to
improve the reliability of findings and establish confidence limits.
NCHRP 433 (Steyn 2012) found that these four items were adequately covered
by the international f-sAPT community. Incorporation of LTPP sections into pro-
grams were more evident than before (although it was not the standard as yet for all
programs and efforts comparing data between APT and LTPP were just starting),
vehicle-pavement-environment interaction received ample attention with the
majority of reports at least indicating the conditions under which specific tests were
conducted (including tire types and conditions, temperature and moisture content at
which a test was conducted and the speed at which the test was conducted).
Collaborative research has increased tremendously during the last decade with the
development of associations leading to improved communications around common
interests and facilities, and different facilities and programs using data collectively
to evaluate models developed on specific facilities and validate the applicability of
specific models and design methods. The regular webinars offered by the TRB
AFD40 committee is evident that this trend continues positively.
NCHRP 433 (Steyn 2012) stated that many of the trends identified in Synthesis
433 were known issues, which did not receive the required attention in the research
and testing environment as would be expected. These included stress rotation within
Perspectives on Trends in International APT Research 217

pavement layers, and the use of non-linear, elasto-plastic, elasto-visco-plastic and

other complex models in research and analysis procedures. Further, economic
evaluation of f-sAPT programs was receiving more attention, although most pro-
grams appeared to be satisfied that they do get good data from the programs, and that
more detailed economic analyses were not required. The continuous improvements
in computer technology allowed the use of more complex models that are providing
data much closer to real pavement’s response to loads. These more complex models
also incorporate the details of tire-pavement contact stress area and magnitudes.
Future significant evaluations from f-sAPT was identified as:
• a more detailed focus on vehicle-pavement interaction (including improved load
and contact stress models);
• environment-pavement interaction (including climate change issues);
• development of and improvements in performance related specifications;
• improved MEPDG validation;
• evaluation of sustainable pavement solutions (energy efficient technologies and
re-use of available infrastructure);
• improved reliability in pavement design, and
• increased use of dedicated airfield APT facilities with special load magnitudes
and bogies used by this traffic.

3 Current Developments (Since NCHRP 433)

3.1 New Programs

A number of f-sAPT facilities have been developed since 2011. These include
facilities in the US (mainly airfield), China, Indonesia, India, Mexico, and Saudi
Arabia. These facilities are using various mobile and stationary loading devices.
Most of these programs are currently in the process of initiating research, with
limited outputs to date (HVSIA 2015; MLS 2015). Current estimates of f-sAPT
facilities is around 40, based on a survey of current and new facilities (Africa 2,
Asia 6, Australasia 3, Canada 1, Central America 1, Europe 11, Middle East 1,
South America 1, USA 14).
The development of cooperative programs was one of the main trends identified
in NCHRP 433. One of the major developments in this area has been the part-
nership between NCAT and MnROAD, whereby NCAT partnered with MnROAD
starting with their 2015 research cycle. The current objective is to quantify the
benefits of pavement preservation for both hot and cold climates and both flexible
and rigid pavements. This collaborative research effort allows for sharing of both
resources and expertise. It also allows a facility like NCAT to incorporate cold
weather climate information into their research, and ultimately the effort starts
218 W. JvdM Steyn and F. Hugo

addressing national rather than state-focused research. This example should serve as
a prompter to other programs to interface their field studies with in-service
performance evaluation with detailed monitoring of climatic variation.

3.2 New Technologies

All the major APT device suppliers have upgraded their devices over the last
5 years through the addition of capabilities or improvements in existing capabili-
ties. In this paper, no detailed analysis of the changes in the various devices are
discussed, as detailed information on new technical machine-related developments
are available from the various suppliers. Suffice to indicate that issues such as
improved loading speed and improved control of loading levels have been some of
the aspects that were addressed by suppliers in the recent past.
Most of the new facilities/programs are centralized around a dedicated facility,
incorporating quality laboratories, instrumentation development units and often
local universities. This is almost a shift from the traditional mobile APT in the field
to much better controlled construction and environmental conditions being con-
trolled during the testing. There are, however, still field studies conducted on
existing pavements by some programs, and it is hoped that this practice will not be
ignored entirely, due to the benefits of testing on real-in-service pavements,
specifically those that have undergone a number of years of real trafficking and
environmental changes.
The concept of big data analysis is also starting to become prevalent in APT
research, with techniques developed in alternative sciences being applied to eval-
uate the full data sets obtained from APT research, rather than to focus on average
measurements conducted at set intervals during the tests. Although no specific
papers or publications on this specific aspect have been published as yet, the
techniques are being investigated by selected programs.
The use of tire-pavement contact stress measurements has increased, using
systems such as the Stress-In-Motion (SIM) De Beer et al. (2012) device as well as
flexible film devices (Steyn and Ilse 2015). Although the technologies are different,
a mutual aim of improved understanding of the surface stress conditions and thus a
more detailed and correct analysis, modeling and understanding of specifically
surfacing conditions of pavements exists.
The use of instrumented pavements in support of pavement research and in
conjunction with APT programs is also receiving attention in some programs. In
South Africa a recent program focused on the installation of detailed sensors in ten
different pavement structures as part of the improvement in the SA Roads Design
System (SARDS). Again, no papers have been published on the findings as yet, but
the intention is to conduct this as a combined LTPP, instrumented pavement/APT
facility over the next 5 years (Steyn 2015).
Perspectives on Trends in International APT Research 219

3.3 New Research

NCHRP 433 (Steyn 2012) indicated that the needs and trends in terms of research
are:
• Hot Mix Asphalt (HMA) as well as studies with environmentally focused topics
such as Warm Mix Asphalt (WMA) and recycling, and
• Evaluation of materials models for the Mechanistic Empirical Pavement Design
Guide (MEPDG).
Evaluating recent (2012–2015) publications and conferences indicate recent
research in the following major areas (only topics and titles are provided in this
summary paper and not detailed discussion of the findings. Also, the publication
details are shown at the end of each of the titles with enough details to find the
respective papers using a standard web search, as the content of the papers are not
cited and discussed). The topics indicated are deemed the major current topics, and
may not be exhaustive:
• Asphalt Concrete and Warm Mix Asphalt
– Warm mix asphalt for airfields (TRR 2456, 2014);
– Permeability of Porous Friction Course Pavements (TRR 2456, 2014);
– Evaluation of Strain-Relieving Interlayer to Retard Thermally Induced
Reflective Cracking on AC overlaid rigid pavements for airfields (NAPTF)
(TRB 2015, Paper 15-3236);
– Evaluation and Implementation of PG 76-22 Asphalt Rubber Binder in
Florida (TRB 2015, Paper 15-3261);
– Field Performance and Structural Characterization of Full-Scale Cold Central
Plant Recycled Pavements (TRB 2015, Paper 15-0781);
– Evaluation of Warm-Mix and Rubber-Modified Open-Graded Friction
Course Test Sections Made Without Fibers in South Carolina (TRB 2015,
Paper 14-4205);
– Enhanced Gradation Guidelines to Improve Asphalt Mixture Performance
(TRB 2014, Paper 14-1153);
– Porous Asphalt Performance in Cold Climate: Research at MnROAD (TRB
2013, Paper 13-4381);
– Evaluation of a Heavy Polymer-Modified Binder Through Accelerated
Pavement Testing (TRB 2013, Paper 13-2150);
– Evaluation of Long-Lasting Perpetual Asphalt Pavement with Life-Cycle
Cost Analysis (TRB 2013, Paper 13-0330);
– Rutting of Rubberized Gap-Graded and Polymer-Modified Dense-Graded
Asphalt Overlays in Composite Pavements (TRB 2012, Paper 12-3494);
– Laboratory and Full-Scale Evaluation of 4.75-mm Nominal Maximum
Aggregate Size Superpave Overlay (TRB 2012, Paper 12-3725);
– Long Term Performance of a Thin Asphalt Overlay on the NCAT Pavement
Test Track (TRB 2012, Paper 12-3842)
220 W. JvdM Steyn and F. Hugo

– Performance Evaluation of Thin Wearing Courses Through Scaled

Accelerated Trafficking (Virginia Center for Transportation Innovation and
Research, Final Report VCTIR 14-R7);
– Performance Testing for Superpave and Structural Validation (FHWA
Publication No. FHWA-HRT-11-045, November 2012) Comprehensive
findings from TPF-5(019): Full-Scale Accelerated Performance Testing for
Superpave and Structural Validation and SPR-2(174): Accelerated Pavement
Testing of Crumb Rubber Modified Asphalt Pavements;
– Implementation of New Quieter Pavement Research: Accelerated Pavement
Testing and Laboratory Evaluation of Different Open-Graded Hot-Mix
Asphalt Materials (UCPRC, UCPRC-RR-2013-04).
• Other aspects and materials
– Concrete Block Paving evaluation and design tables (11th International
Conference on Concrete Block Pavement, 2015);
– Accelerated pavement testing on slab and block pavements using the MLS10
Mobile Load Simulator (Hugo et al. 2012);
– Relationship between Tire-Pavement Interaction Noise and Surface
Characteristics in 2009 Pavement Test Track Cycle (TRB 2013, Paper
13-0328);
– Impact of Wide-Base Tires on Pavements: Results from Instrumentation
Measurements and Modeling Analysis (TRB 2012, Paper 12-0727);
– Performance Evaluation of Uncrushed Aggregates in Unsurfaced Road
Applications Through Accelerated Pavement Testing (TRB 2012, Paper
12-2251);
– Lessoned Learned During the First Application of MSP for Extracting
Asphalt Slabs in Comparative Testing of Fatigue Performance of Warm
Mix RAP Asphalt MMLS3 Trafficking (ISAP APE 2015);
– Accelerated Traffic Load Testing of Expansion Joints for the Self-anchored
Suspension Section of the New San Francisco-Oakland Bay Bridge East
Span (UCPRC, UCPRC-RR-2011-06);
– Full-Depth Recycling (UCPRC, UCPRC-RR-2014-03);
– A Synthesis of Applications of the MLS as an Innovative System for
Evaluating Performance of Asphalt Materials in Pavement Engineering
(CAPSA 2015);
– Evaluation of Tire/Surfacing/Base Contact Stresses and Texture Depth
(International Journal of Transportation Science and Technology 2015);
– Strain response of a semi-rigid base asphalt pavement based on heavy-load
f-Scale APT with fibre Bragg grating sensors RMPD, 16:2);
– International case studies in support of successful applications of APT in
Pavement Engineering (Hugo et al. 2012);
– Analysis of the long-term structural performance of flexible pavements using
full-scale accelerated pavement tests [Transport Research Arena
(TRA) 2012];
Perspectives on Trends in International APT Research 221

– Testing of bridge deck surfacing for the new Liaohe Bridge constructed in
Liaoning Province in China (Hugo et al. 2012).
• Low Volume Roads
– Comparison of Low-Volume Road Pavement Performance with Results of
Accelerated Pavement Testing (Johnston et al. 2015, Volume 2).
• Portland Concrete
– The design, construction and Heavy Vehicle Simulator testing results on
roller compacted concrete test sections at the CSIR innovation site and on a
full scale test road at Rayton (Du Plessis et al. 2014);
– Ultra-thin reinforced concrete pavements UTCRCP: addressing the design
issues (Du Plessis et al. 2014).
• APT 2012 sessions—Cost benefit, Airports, Sustainable materials, AC,
Instrumentation, MEPDG calibration and design, Functional performance,
Impact on practice, PCC, Links to laboratory results, future roles and
sustainability.
Analysis of the general focus of these research indicates that surfacing materials
and environmental aspects of pavements are still a major research field in inter-
national APT.
TRB has been organizing regular webinars through the AFD40 committee on a
range of APT-related topics. Although these are directly related to current and
completed research, and therefore mostly reported in journals and conferences, a
selection of typical topics is included as support for some of the new developments
in international APT (Pavetrack 2015):
• Accelerated Pavement Testing Program with the Mobile Load Simulator
MLS10—Temperature Analysis (BASt);
• Follow up on HVS testing of Roller Compacted Concrete and Ultra-Thin
Reinforced Concrete Test sections (CSIR);
• FDOT’s APT program, expanding to include a concrete test road in addition to
their longstanding HVS program (FDoT);
• Using Accelerated Pavement Testing to Evaluate Permeable Interlocking
Concrete Pavement Performance (UCPRC), and
• Cold Central Plant Recycling—A Proven Paving Approach Using 100 %
Recycled Materials.
Although this paper focuses on trends in the period 2011–2015, it is interesting
to review the Research needs as defined through the TRB system as an indication of
future focus for APT research. In this regard, the following research need state-
ments have been developed and published:
• Effect of Accelerated Testing on the Rutting and Cracking Performance of
Flexible Pavements (posted since 2009), and
222 W. JvdM Steyn and F. Hugo

• Development of a Permanent Deformation Test Procedure for Evaluating

Rutting Potentials of Pavement Granular Base/Subbase Layers (posted since
2015).
The following are currently seen as priorities by the AFD40 committee and
being developed into final proposals. Although these are shown as preliminary
research need statements, it indicates the current thinking around new research and
issues that need to be addressed:
• Cost Benefit Determination for Pavement Research;
• The effect of speed, lateral wander and HMA temperature on rutting and
cracking in HMA, and
• The relationship between APT test results and the performance of in-service
pavements (an on-going issue since the early days of APT, and highlighted in all
APT synthesis reports).
Again the focus on surfacing materials is evident, but also the vehicle-pavement
interaction issues in terms of speed effects and linkages to real life pavements and
economics.
A synthesis of applications of the MLS in evaluating performance of asphalt
materials in pavement engineering (Hugo and Steyn 2015) supported most of the
NCHRP 433 areas to focus on, such as the need to incorporate higher loading speeds
in APT, the importance of accurate environmental testing conditions, and the
importance of accurate tire-pavement contact stress application and analysis. As
Al-Qadi indicated in his keynote lecture at APT 2012, the variability of pavement
material and in construction quality can influence the response and performance of
pavement materials and structure and thus need to be incorporated into the analytical
evaluation of the pavement, rather than attempting to discard it (Hugo et al. 2012).

4 Challenges

Major challenges for APT are currently viewed as access to funding, continued
institutional support of active programs and economic evaluation of APT programs.
There is international consensus that funding of infrastructure in general, and
infrastructure research specifically, is under pressure. Figures in the US indicates
that between 1994 and 2014 total federal, state, and local investment in trans-
portation has fallen as a share of GDP, while population, traffic and congestion, and
maintenance backlogs have increased. The US is perceived to lag behind many of
its overseas competitors in transportation infrastructure investment, with 65 % of
America’s major roads rated in less than good condition (NEC 2014).
Although the potential benefits of investing in infrastructure, and by definition,
infrastructure research, are clear and well understood (Cambridge Systematics 2012;
Perspectives on Trends in International APT Research 223

NEC 2014; OECD 2011), the general international economic situation focuses on a
more austere environment where expenditures on issues for which the benefits are
not clear-cut and short-term are difficult to motivate (Utt 2011; Whitfield 2012).
In this economic environment there are a number of current and traditional APT
programs that are experiencing significant constraints in terms of continued funding
and support for their programs, even where technical merits of the programs are
well-understood (due to sensitivities around the issue, no specific names of pro-
grams are mentioned in this paper). From an institutional viewpoint, it has become
extremely important to motivate the continued need for accelerated APT in general
at executive level given the technical and economic benefits that have accrued
nationally and internationally from its application.
It is especially important to motivate this need in the light of improvements in
modeling techniques and computer simulations that may appear to the lay persons,
to enable all that is being done using APT, to be done at a fraction of the cost and
time solely in that way. The importance of communication regarding the need and
requirement to develop accurate pavement materials and structural models, to
enable the related computer simulations to provide accurate outputs, cannot be
overstated.
The economic benefit of APT programs can be calculated objectively, and most
facilities either calculate or assume BCRs of between 1 and 10. It appears as if BCR
evaluations are being performed by more APT owners in the last number of years,
although it still often happens after the testing and not as part of the planning
process.

5 Conclusions and Recommendations

5.1 Conclusions

It is apparent that there has been significant progress over the full range of topics
that form the basis of pavement engineering during the past decade. This has been
stimulated by the growth in the extent of APT in the national and international
arena. Unfortunately this is now experiencing pressure of economic constraints at a
time when there is an urgent need for reconstruction of highways that are carrying
increased traffic volumes and loads. These issues have not been discussed in any
detail, in the paper albeit that they have also been the focus of APT. A spinoff of the
research findings has been the progress of producing more durable pavements that
should help to reduce the long-term effect of the current situation.
224 W. JvdM Steyn and F. Hugo

5.2 Recommendations

The issues regarding the need for improved communication between the stake-
holders that have been raised should be addressed as a topic at the conference and
the related associated entities such as the AFD40 committee and other TRB
standing committees. In a connected world, improved communication and collab-
oration are required to ensure that short-term economic issues are not stifling
innovative research that will support the required infrastructure investment to allow
for renewed international economic growth.

Acknowledgments The authors would like to acknowledge the support and information obtained
from most APT role players in providing information or access to information to generate this
paper. The opinions in the paper remains those of the authors.

References

Cambridge Systematics. (2012). Assessing the economic benefit of transportation infrastructure

investment in a mature surface transportation system. The National Cooperative Highway
Research Program, NCHRP Project 20-24, Task 80, National Cooperative Highway Research
Program, Washington, D.C.
De Beer, M., Maina, J. W., van Rensburg, Y., & Greben, J. M. (2012). Toward using tire-road
contact stresses in pavement design and analysis. Tire Science and Technology, 40(4), 226–271.
Du Plessis, L., Kilian, A. & Mngaza, K. (2014). Ultra-thin reinforced concrete pavements
(UTRCP): Addressing the design issues. In Proceedings of 33rd Southern African transport
Conference, CSIR international convention center, Pretoria, South Africa.
Hugo, F. (2004). Accelerated pavement testing—Comfort, concerns, constraints and challenges. In
Opening overview, 2nd international conference on accelerated pavement testing, CD.
University of Minnesota, Minneapolis-St Paul, USA, September 2004.
Hugo, F., & Martin, A. E. (2004a). Significant findings from full-scale accelerated pavement
testing. NCHRP Synthesis 325. National Cooperative Highway Research Program (NCHRP),
Transportation Research Board, (TRB), National Research Council, Washington, D.C.
Hugo, F., & Martin, A. E. (2004b). Synthesis of highway practice 325: Significant findings from
full-scale accelerated pavement testing. Transportation Research Board, National Research
Council, Washington, D.C.
Hugo, F., McCullough, B. F., & Van der Walt, B. (1991). Full-scale accelerated pavement testing
for the texas state department of highways and public transportation. Transportation Research
Record 1293, Transportation Research Board, National Research Council, Washington D.C.
pp. 52–60.
Hugo, F., & Steyn, W.J.vdM. (2015). A synthesis of applications of the MLS as an innovative
system for evaluating performance of asphalt materials in pavement engineering. In 11th
conference on asphalt pavements for Southern Africa (CAPSA), 16 to 19 August 2015, Sun
City, South Africa.
Hugo, F., Zefeng, T., Arraigada, M., Kim, Y. R., & Li, S. M. (2012). International case studies in
support of successful applications of APT in pavement engineering. Paper presented and
published in the proceedings of 4th international conference on accelerated pavement testing,
Sacramento, CA, September 2012.
HVSIA. (2015). HVS international alliance website. www.hvsia.co.za. Accessed October 22,
2015.
Perspectives on Trends in International APT Research 225

Johnston, M. G., Núñez, W. P., & Cerrati, J. A. (2015). Comparing a low volume road pavement
performance to accelerated pavement testing results. In 11th international conference on low
volume roads, Pittsburgh, Pennsylvania, July 13–15, 2015.
Metcalf, J. B. (1996). Synthesis of highway practice 235: Application of full-scale accelerated
pavement testing. Transportation Research Board, National Research Council, Washington, D.
C., 1996, 117 pp.
MLS Test Systems. (2015). http://academic.sun.ac.za/mls/. Accessed October 22, 2015.
NEC (National Economic Council). (2014). An economic analysis of transportation infrastructure
investment. National Economic Council and the President’s Council of Economic Advisers,
The White House, Washington, D.C. https://www.whitehouse.gov/sites/default/files/docs/
economic_analysis_of_transportation_investments.pdf. Accessed October 22, 2015.
OECD. (2011). Strategic transport infrastructure needs to 2030. OECD international futures
program. Paris: OECD Publications.
Pavetrack. (2015). NCAT pavement test track. www.pavetrack.com. Accessed October 22, 2015.
Steyn, W.J.vdM. (2012). Full-scale accelerated pavement testing, 2000 to 2011. NCHRP Synthesis
433. National Cooperative Highway Research Program (NCHRP), Transportation Research
Board (TRB), National Research Council, Washington D.C.
Steyn, W.J.vdM. (2015). Mechanistic-empirical pavement design enhancements in South
Africa—Field validation of models. In 94th annual transportation research board meeting,
Washington D.C.
Steyn, W.J.vdM., & Ilse, M. (2015). Evaluation of tire/surfacing/base contact stresses and texture
depth. International Journal of Transportation Science and Technology, 4(1), 107–118. ISSN
1475-472-X.
Utt, R. D. (2011). How to create an effective transportation program in an age of fiscal austerity.
Washington D.C., USA: The Heritage Foundation.
Whitfield, D. (2012). In place of austerity—Reconstructing the economic, state and public
services. Nottingham: Spokesman.
Potential Benefits of APTF for Evaluation
of Flexible Pavement for Its Permanent
Deformation Behaviour

M.N. Nagabhushana, Shahbaz Khan, Abhishek Mittal

and Devesh Tiwari

Abstract More than 95 % of road network in India is made up of flexible pave-

ment, especially due to the relative lesser initial cost, ease of construction and
maintenance. The design of flexible pavement in India is done with criteria of load
associated fatigue cracking and permanent deformation (rutting). While design
methods try to consider these failure modes, more rational inputs based on intrinsic
evaluations are required for better and improved designs. The Accelerated
Pavement Testing Facility (APTF) is regarded as a versatile tool all over the world
in pavement evaluation and performance monitoring due to its advance evaluation
capabilities within a short time. This paper presents the rutting performance char-
acteristics of a typical flexible pavement built as per Indian design practice, eval-
uated using APTF. APTF is capable of evaluation with regard to total structural
ability of a pavement based on number of standard axle load passes which repre-
sents time period and load sustainability of a pavement with respect to defined
failure criteria. The pavement was instrumented with interfacial response moni-
toring, wherein deflections were effectively measured using state of art instru-
mentation technique. The study result and possible rutting mechanism under the
operational capabilities of APTF and test condition have been discussed in the
paper.

Keywords Accelerated pavement testing facility
Permanent deformation

Flexible pavement Benefits

M.N. Nagabhushana (&)

APTF Group, CSIR-CRRI, New Delhi, India
e-mail: [email protected]
S. Khan
A. Mittal
CSIR-CRRI, New Delhi, India
D. Tiwari
Pavement Engineering Area, CSIR-CRRI, New Delhi, India

© Springer International Publishing Switzerland 2016 227

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_15
228 M.N. Nagabhushana et al.

1 Introduction

1.1 Accelerated Pavement Testing as a Tool

for Rutting Evaluation

Accelerated Pavement Testing (APT) is a link between the Pavement Engineering

(P-E) tools of laboratory evaluation of pavement materials and also understanding
of full-scale field behaviour of materials within a pavement structure. It has been
used widely to support most of the current pavement design methods and also forms
the basis for developing various theories about pavement behavior. Therefore, APT
has been used as a tool which provides information to pavement engineers to
understand behavior of pavement materials and structures under typical traffic
loading in known environmental condition. Accelerated pavement testing facility is
defined as “controlled application of wheel loading to pavement structures for the
purpose of simulating the effects of long-term in service loading conditions in a
compressed period of time” (Hugo and Martin 2004).
Several authors used APTF for rutting evaluation such as comparison for overlay
evaluation (Harvey and Popescu 2000), evaluation of coarse and fine grained
superpave mixture (Choubane et al. 2006), development of pavement rutting pre-
diction model (Archilla and Madanat 2000; Ji et al. 2015), comparison of rutting
resistance of unmodified and SBS-modified Superpave mixtures (Sirin et al. 2008),
use of geocell-reinforced sand base (Yang et al. 2012), evaluating permanent
deformation in asphalt rubber mixtures (Fontes et al. 2010), evaluation of the
structural contribution of full-depth reclamation material stabilized with foamed
asphalt (Romanoschi et al. 2004), effect of Tire Pressure on Flexible Pavement
Response and Performance (Bonaquist et al. 1989). Long term pavement perfor-
mance can be done through field surveys but it is time consuming and often needs
lane closures for inspection of actual site conditions. Therefore, APT facility
bridges the gap between design, laboratory and field performance of pavement
(Hugo et al. 1991). The application of accelerated pavement testing facility and its
relation with long term pavement performance has been well documented (Hugo
2004). Also, the 1.1 HVS passes of 10 kmph speed and NCAT track speed of
70 kmph are equivalent (Powell 2008). Many researchers as per NCHRP synthesis
433 (Steyn 2012) have the concern of the performance of HVS tested in the real
field owing to the reduced speed of the HVS as the contact time at low speed is
more as compared to the high speed, so the accumulation of damage will be more at
low speed in HVS. Furthermore, the under increased load levels the use of
sub-grade strain is the poor indicator of the pavement life (Alabaster et al. 2004).
Potential Benefits of APTF for Evaluation of Flexible Pavement … 229

1.2 Indian APT with Instrumentation

and Material Used

The current Accelerated Pavement Testing Facility (APTF) in India was procured
from South Africa and is known as Heavy Vehicle Simulator (HVS) Mark IV+ with
capabilities as given in NCHRP synthesis 433.
The measurement zone of the test section is 8 m × 1 m (length × width) and is
divided into 16 points with 1 m on both end side (length wise) of the test strip as
acceleration and deceleration zone making 6 m effective length for uniform speed
with loading. The test section was instrumented with Multi Depth Deflectometer
(MDD) at the centre of the test section. The Multi-Depth Deflectometer (MDD) is
an linear variable displacement transducers (LVDT) deflection measuring device
which is retrofitted into pavement layers. A maximum of six MDD modules may be
installed in a single 1.5-inch diameter hole. The modules are clamped against the
sides of the hole at the required depths and the center core is attached to an anchor
located approximately 7 feet below the pavement surface (Scullion et al. 1988).
The MDD measure the in situ elastic and permanent deformation of each layer. For
the present case four module MDD was installed at the interface of each layer as
shown in Fig. 1. The laser profilometer was used to measure the transverse profile
of the test section.
APT facility was experimented on a pre-designed and constructed flexible
pavement test section based on design specification of Indian Roads Congress code
(IRC:37 2001) for California Bearing Ratio (CBR) of 5 % of soil sub-grade and
design life of 30 million standard axles (MSA). The total crust composition of
flexible pavement test section is shown in Fig. 1.
The soil sub-grade as per soil classification is well graded silty soil found in Delhi
having granular sub-base which is moderately coarse graded. The unbound base
layers are typically dense graded Wet Mix Macadam (WMM) covered with binder

Fig. 1 Crust composition and MDD position within the pavement ( )

230 M.N. Nagabhushana et al.

Table 1 Material properties of bituminous layer

Mix Optimum binder Marshall Retained Air Voids filled with
specification content (%) stability (kN) Marshall voids Bitumen
stability % (%) (VFB in %)
DBC 5.1 10.22 83 4.2 71.7
DBM 5.0 10.15 79 4.9 69.3

course of Dense Bituminous Macadam (DBM) and wearing course of Dense

Bituminous Concrete (DBC). The maximum dry density of 1.87 gm/cc of WMM was
found at optimum moisture content of 12 % which are above the required designed
specification. The granular sub-base and base layers were meeting the gradation
specification of the Ministry of Roads Transport and Highway. The entire test section
construction was done as per Ministry of Road Transport and Highway (MORTH
2001). The material properties of bituminous layer are presented in Table 1.

2 Permanent Deformation

Rutting is described as a longitudinal deformation developing in flexible pavements

under the action of repeated channelized loadings thereby reducing serviceability
and safety. There are three mechanisms that might be involved in the development
of rutting, that is—plastic movement, mechanical deformation and consolidation
(Kandhal et al. 1998). Rutting can be the result of permanent reduction in volume
(consolidation/traffic densification), permanent movement of the material at con-
stant volume (plastic deformation/shear) or a combination of the both (Roque et al.
2004). Bituminous mixes are time, temperature and stress dependent materials and
exhibits elastic/plastic/visco-elastic response when subjected to repeated loading.
The inherent elastic properties of bituminous mixes are represented by stiffness
modulus and Poisson’s ratio because they do not contribute to permanent defor-
mation, whereas plastic properties contribute to permanent deformation under
cumulative repeated loading. Factors such as vehicle speed, contact time and
contact pressure influences rutting and represented directly in the creep rate model,
whereas temperature, characteristics of bituminous mixture and construction quality
are represented in the constant values. Shear resistance properties of bituminous
mixes need to be properly addressed for limiting rutting. Most of the pavement
structural design practices mainly focus on protecting the subgrade by excessive
vertical strain and shear resistance of the bituminous layers are left to mix designers.
Adequacy of shear resistance in bituminous mixture is commonly assessed with
empirical methods, such as Marshall Stability. There is also a continuous scientific
effort to shift towards the use of fundamental performance related test methods.
Generally two procedures are being vastly used to limit rutting; one to limit the
vertical compressive strain on top of the subgrade and the other is to limit the total
accumulated permanent deformation on the pavement surface based on the prop-
erties of each individual layer (Huang 2004). The procedure based on former
Potential Benefits of APTF for Evaluation of Flexible Pavement … 231

approach is to limit rutting through number of load repetitions (Nr) by relating

vertical compressive strain εz coming on top of the subgrade as presented in Eq. 1.
 f 2
1
Nr ¼ f 1  ð1Þ
ez

It has been found that different values for constants f1 and f2 have been reported
by several agencies. In India, the rutting model has been developed for rutting
failure of flexible pavements from large number of data obtained from various
national level research studies at 80 % reliability. Limiting the allowable rut depth
as 20 mm and beyond it as failure, the rutting equation has been in use for Indian
Design from Indian Road Congress (IRC-37) as Eq. 2.
 4:5337
1
Nr ¼ 4:1656  108  ð2Þ
ez

Nr Number of load repetitions in rutting failure

εz Compressive strain on the top of subgrade
Thus, to find out the number of load repetitions for failure due to rutting, the
compressive strain at the top of the subgrade is calculated for pavement structure.

3 Objective of the Present Study

The objective of this study is to summarize the potential benefits of the use of Heavy
Vehicle Simulator (HVS) in India for evaluation the rutting phenomenon in a flexible
pavement constructed as per Indian design concept i.e. Indian Road Congress
(IRC):37-2001 based on test conducted in CSIR-Central Road Research Institute
(CRRI). Two different test programs within the objectives were formulated i.e.
• To review the present IRC model for pavement design for permanent defor-
mation using instrumentations and
• To find out effect of Uni-directional and bidirectional trafficking using HVS.

4 Rutting Evaluation with APT

4.1 Performance of Test Section with Respect to Design

The loading sequence applied on dual wheels to the APT test section started from a
standard load of 40 kN with a tyre pressure of 650 kPa. Single wheel path,
bi-directional traffic was used for load application. The load for first 2,50,000
232 M.N. Nagabhushana et al.

repetition has been fixed to 40 kN, thereafter 60 kN for further 2,20,000 repetitions
and 80 kN for about 5,00,000 repetition adding to about 9 MSA using the fourth
power law. The channelized trafficking was done with no wandering. The average
speed was 10 kmph and test period was from June 2011 to Feb. 2012, during day
time only. Temperature sensors embedded at bituminous layers were used for the
measurement of pavement temperature. The bituminous mix properties were as per
Table 1 shown above, while no temperature control measures were adopted in order
to get data nearer to true life situation. The deflection data of pavement different
layers were captured from MDD using Data Acquisition system. The weather
station was mounted on the top of HVS (roof top) to measure the variation in
climatic temperature throughout the testing period. The MDD gives the deflection
bowl after every passes. The average of peak deflection of the consecutive three
passes were used to find the elastic deflection with respect to depth and number of
passes as shown in Figs. 2 and 3 respectively. The deflection at 40 mm depth is the
sum of the deflections of all the lower layers.
Here MDD-40, MDD-160, MDD-410, MDD-710 shows the deflection at 40,
160, 410, 710 mm respectively.
The above Fig. 3 shows elastic deflection (X-axis) with depth (Y-axis) and
increasing number of passes (X-axis). The nomenclature in graph deflection-911,
deflection-6317, deflection-52557 etc. shows the deflection value of the pavement
after 911, 6317, 52557 number of passes of HVS respectively. The permanent
deformation of a specific layer is the amount of the permanent shift of the respective
MDD as measured with reference point which is its original position at the initial
installation. The plastic deformation of each layer was found out from MDD data
and has been plotted below as shown in Fig. 4. The permanent deformation is the
deformation of each layer with respect to anchor position of MDD. The permanent
deformation at 40 mm depth is the total permanent deformation of all the lower
layers after the specified number of passes for a given pavement structure. It helps

Fig. 2 Typical deflection bowl obtained of each layer

Potential Benefits of APTF for Evaluation of Flexible Pavement … 233

Fig. 3 Elastic deflection with depth and increasing passes

Fig. 4 Rutting of each layer with number of passes

to arrive at the deformation of any individual layer by subtracting the total defor-
mation of other layers from the effective deformation measured at the top surface.
The permanent deformation of different layers were calculated using MDD
whereas the permanent deformation of bituminous layers were calculated by sub-
tracting the total surface deformation found through comparing the initial profile
(zero passes) using laser profilometer (Fig. 5) with respect to total deformation at
40 mm depth.
234 M.N. Nagabhushana et al.

Fig. 5 Total surface deformation

Table 2 Total rutting after Layer Deformation (mm) Percentage (%)

approximately 9,30,000
passes Subgrade 2.51 12.85
Sub-base 1.61 8.25
Base 4.45 22.80
DBM 5.31 27.20
DBC 5.64 28.90
Total 19.52

The total contribution of each pavement layers to rutting found through APTF on
experiment test section is given in Table 2. Indian Roads Congress Code (IRC-37)
has prescribed the failure criterion of more than 20 mm depth.

4.2 Rutting Evaluation for Uni- Versus Bi-directional

Loading

Uni- versus Bi-Directional Loading of trafficking was used in accelerated pavement

testing facility to check which mode of trafficking has greater potential for distress
in terms of permanent deformation for the given pavement with the design speci-
fications as given earlier. The HVS has two different mode of operation which are
with and without wandering and for the current study wandering was not consid-
ered. Under bidirectional mode of operation, 900 load repetition was achieved per
hour with an operational speed of 10 kmph whereas about 500 passes per hour
could be applied in the Uni-directional mode of trafficking. Both Bi-directional and
unidirectional mode of trafficking were made with following conditions:
Potential Benefits of APTF for Evaluation of Flexible Pavement … 235

Tyre pressure 700 kPa

Traffic load 80 KN (half axle)
Tyre type Dual wheel (11R/22.5)
Traffic speed 10 kmph
Traffic length 6 m with 1 m on each side acceleration and deceleration zone
All these test parameter were kept constant through the entire test period and
environmental parameters was measured through weather station and pavement
temperature was monitored throughout all the test period.
The section was divided into 16 points with the spacing of 0.5 m uniformly
making it to a total 8 m long., as shown in Fig. 6, but for analysis purpose, only 13
points were considered, leaving the initial acceleration zone and deceleration zone
of 1 m each on either side of the section. Initial profile of the test section was
recorded using a Laser Profilometer for all the 16 points and is taken as reference
profile for rutting evaluation. Subsequent profiles were taken using laser
profile-meter and were plotted using Microsoft excel. The difference between the
initial profile of the section under the wheel load and subsequent profile on the
section due to the moving load depicts the rut depth in the wheel path at different
points, using which the average rut depth has been calculated.
The average rut depth with increasing number of passes has been plotted as
shown in Fig. 7 for bi-directional mode of traffic. The best fit line was plotted and
R2 value of 0.9498 from above figure was found out for the polynomial equation
and the average rut depth was found out to be around 4 mm. Similarly for
Uni-directional mode, the rut depth is plotted with number of passes and the rut
depth after same number of passes was found to be about 9 mm as shown in Fig. 8.

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Fig. 6 Test section

236 M.N. Nagabhushana et al.

Fig. 7 Rut depth with number of passes in bidirectional mode

Fig. 8 Progression of rut with number of passes in unidirectional mode

4.2.1 Air Void

The air voids of the test section were also determined before and after trafficking of
the HVS in the uni-directional and bi-directional mode respectively. The cores were
taken from test section from wheel path as well as away the wheel path. The results are
shown in Table 3. The air void was determined by automatic vacuum sealing method.
Potential Benefits of APTF for Evaluation of Flexible Pavement … 237

Table 3 Air voids of the test section

Bituminous layer Air voids (%)
Unidirectional Bidirectional Reference (non-rutted)
BC 5.21 5.26 5.66
DBM 5.18 6.3 6.81

From the above Table 3, it has been found that the non-rutted portion (away
from the wheel path) has the maximum air voids whereas it is decreasing from
Bi-directional (from wheel path) to Unidirectional (from wheel path), accordingly.
This indicates Uni-directional rutting for both the layers are having more reduction
in air voids than others.

5 Conclusions

Rutting is developed either due to overloading of designed pavement, instability of

bituminous mix used in the pavement and poor quality control during the con-
struction of the pavement structure. The difference of pavement design life between
theoretical design aspect and achieved through constructed pavement can be well
understood by using APTF. This also indicates the sustainability of a pavement
structure for which it has been designed.
The conclusions drawn from the present study have been summarized below:
• With the trafficking of 2,50,000 passes for the 40 kN (standard axle) with a tyre
pressure of 650 kPa most of the rutting has taken place in the upper layer with a
major contribution of bituminous layer and a negligible effect on the sub-grade.
This indicates that controlled axle loads do not damage the subgrade whereas it
affects the upper layers. Thus, sustainability of a pavement depends more on
controlled axle loads.
• The total rutting/deformation is 19.52 mm due to load applications of about 9
MSA whereas the design was for 30 MSA which clearly suggest that there is
need to switch to more sophisticated model as MEPDG, CALME etc. which
account for each layer rutting.
• The difference in design and actual life of pavement from HVS can be attributed
to design concept for rutting as Indian design consider vertical sub-grade strain
as rutting criteria but as can be seen sub-grade rutting is only 13 % as compared
to total rutting.
• The air voids of the bituminous decreased with indicates significant densifica-
tion of the bituminous layer, since the density of DBM was slightly less than the
DBC, so it is more compacted under traffic, thereby more reduction of air void
in DBM layer. So, there is a need for more improved mix design methods.
238 M.N. Nagabhushana et al.

• Rutting is affected by the directional flow of traffic with uni-directional rutting

doubling as compared to that of bi-directional rutting, with the other conditions
of the testing being almost identical.
• The reduction in air void is more pronounced in Uni-directional rutting as
compared with bi-directional rutting, that too in the DBM layer which is
comparable to the result obtained by others (Byron et al. 2004)
• The use of HVS led to the better understanding of pavement permanent
deformation behaviour and failure mechanism which can be incorporated in
design standards

References

Alabaster, D., Arnold, G., & Steven, B. (2004). The equivalent standard axle approach and flexible
thin surfaced pavements. In International Conference on Accelerated Pavement Testing, 2nd,
2004. Minneapolis, MN, USA.
Archilla, A. R., & Madanat, S. (2000). Development of a pavement rutting model from
experimental data. Journal of Transportation Engineering, 126(4), 291–299.
Bonaquist, R., Surdahl, R., & Mogawer, W. (1989). Effect of tire pressure on flexible pavement
response and performance (No. 1227).
Byron, T., Choubane, B., & Tia, M. (2004). Assessing appropriate loading configuration in
accelerated pavement testing. In Proceedings of 2nd International Conference on Accelerated
Pavement Testing.
Choubane, B., Gokhale, S., Sholar, G., & Moseley, H. (2006). Evaluation of coarse-and
fine-graded superpave mixtures under accelerated pavement testing. Transportation Research
Record: Journal of the Transportation Research Board, 1974, 120–127.
Fontes, L. P., Triches, G., Pais, J. C., & Pereira, P. A. (2010). Evaluating permanent deformation
in asphalt rubber mixtures. Construction and Building Materials, 24(7), 1193–1200.
Harvey, J., & Popescu, L. (2000). Accelerated pavement testing of rutting performance of two
caltrans overlay strategies. Transportation Research Record: Journal of the Transportation
Research Board, 1716, 116–125.
Huang, Y. H. (2004). Pavement analysis and design. Upper Saddle River: Pearson Printice Hall.
Hugo, F. (2004). Accelerated pavement testing overview-comfort; concerns; constraints and,
challenges. In Proceedings of the Second International Conference on Accelerated Pavement
Testing. The University of Minnesota, Minneapolis, MN, September. (pp. 1–35).
Hugo, F., & Martin, A. E. (2004). Significant findings from full-scale accelerated pavement testing
(Vol. 325). Transportation Research Board.
Hugo, F., McCullough, B. F., & van der Walt, B. (1991). Full-scale accelerated pavement testing
for the Texas State Department of Highways and Public Transportation. Transportation
Research Record (1293).
IRC:37-2001. Guidelines for the design of flexible pavements, 2nd revision. New Delhi: Indian
Roads Congress.
Ji, X., Zheng, N., Niu, S., Meng, S., & Xu, Q. (2015). Development of a rutting prediction model
for asphalt pavements with the use of an accelerated loading facility. Road Materials and
Pavement Design, 1–17.
Kandhal, P. S., Mallick, R. B., & Brown, E. R. (1998). Hot mix asphalt for intersections in hot
climates. National Center for Asphalt Technology, NATC Report (98-6).
Ministry of Road Transport and Highways. (2001). (4th ed.). New Delhi.
Potential Benefits of APTF for Evaluation of Flexible Pavement … 239

Powell, R. L. (2008). Modeling rutting performance on the NCAT pavement test track. In
Transportation Research Board 87th Annual Meeting (No. 08-2596).
Romanoschi, S., Hossain, M., Gisi, A., & Heitzman, M. (2004). Accelerated pavement testing
evaluation of the structural contribution of full-depth reclamation material stabilized with
foamed asphalt. Transportation Research Record: Journal of the Transportation Research
Board, 1896, 199–207.
Roque, R., Birgisson, B., Darku, D., & Drakos, C. (2004). Evaluation of laboratory testing
systems for asphalt mixture design and evaluation: volume 1 (no. Uf 4910-4504-659-12).
Scullion, T., Uzan, J., Yazdani, J. I., & Chan, P. (1988). Field evaluation of the multi-depth
deflectometers. Interim Research Report, September 1986–September 1988 College Station:
Texas A&M University.
Sirin, O., Kim, H. J., Tia, M., & Choubane, B. (2008). Comparison of rutting resistance of
unmodified and SBS-modified Superpave mixtures by accelerated pavement testing.
Construction and Building Materials, 22(3), 286–294.
Steyn, W. J. (2012). NCHRP synthesis of highway practice 433: Significant findings from
full-scale accelerated pavement testing.
Yang, X., Han, J., Pokharel, S. K., Manandhar, C., Parsons, R. L., Leshchinsky, D., et al. (2012).
Accelerated pavement testing of unpaved roads with geocell-reinforced sand bases. Geotextiles
and Geomembranes, 32, 95–103.
Study of the Bearing Capacity of Swiss
Standard Pavements Under MLS10
Loading

Martin Arraigada, Andreas Treuholz and Manfred N. Partl

Abstract Accelerated Pavement Testing (APT) is an efficient testing procedure to

evaluate the performance of different pavement concepts. With APT, it is possible
to determine and measure the structural response and pavement performance under
a controlled, accelerated accumulation of load damage that simulate the long-term
in-service condition, but in a compressed period of time. The repeatability of the
loads applied permits making accurate predictions about the bearing capacity and
lifespan of the pavement under study. It also allows making comparative assess-
ments of potential long term performance of any type of pavements. This paper
summarizes the results of a series of initial tests carried out on pavements with
similar structure but different stiffnesses designed following Swiss standards. Each
pavement was loaded with the Mobile Load Simulator MLS10 until reaching its
bearing capacity limit. The aim of these tests was to use the results as a benchmark
to compare it to other non-standard structures with similar stiffness in terms of
dimensioning and/or materials. It was found that the pavements’ life-spans were
basically longer to the ones anticipated by the standards. This difference was larger
for the pavements with lower stiffness. Further, as expected, the distress mecha-
nisms were similar for all the pavements considered.

Keywords MLS10 Lifespan of pavements
Bearing capacity
Pavement dis-


tress Pavement design

1 Background

One of the purposes of accelerated pavement testing (APT) is to provide experi-

mental, realistic data to develop and validate pavement design methods. APT is also
an important tool to evaluate innovative pavement structures or materials that have

M. Arraigada (&)
A. Treuholz
M.N. Partl

Road Engineering/Sealing Components Laboratory, EMPA,
Ueberlandstrasse 129, 8600 Duebendorf, Switzerland
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 241

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_16
242 M. Arraigada et al.

been tested at laboratory scale and have shown favourable results. APT offers the
possibility to validate these laboratory results under simulated traffic loads instead
of constructing a real road directly. This avoids the risk of costly failures if the real
road pavement doesn’t perform as expected. APT is the necessary tool to bridge the
gap between testing materials in the laboratory and in the field, to perform evalu-
ations that consider the interaction among material, traffic load and weather.
The loads to simulate traffic used in APT are rolling tires and, in consequence,
one step closer to reality than laboratory loads. However, APT loads and real traffic
loading differ considerably in speed, distribution, rate of application and resting
periods, among an enormous amount of other variables. All these factors influence
the distress development of the studied pavements differently. Further, in APT it is
not possible to simulate the influence of the environment on the aging of the
materials of the structure.
Therefore, each APT method or device can lead to different results in terms of
number of applied loads until reaching the bearing capacity limit of a structure
(lifespan). This depends on the characteristics of the loads, the construction con-
ditions and the environmental control capabilities. Consequently, after establishing
a new APT system, it is important to study what is the effect of the individual APT
characteristics on the distress and bearing capacity of existing traditional pavement
technologies. This information can be then used as a reference or benchmark when
testing other new types of structures.
In this sense, one of the uncertainties to clarify after the Mobile Load Simulator
(MLS10) prototype APT device was established in Switzerland was the necessary
amount of time or number of load cycles required to reach the bearing capacity of
pavements with different stiffnesses. This question had a practical reason, because
when organizing a test of an innovative type of pavement, it is important to know
the approximate time required to reach its failure. Testing a structure dimensioned
with the design standards would provide a validation of the design under a con-
trolled load application.
This work presents a summary of results from a series of APT experiments
carried out on pavements built using traditional materials and construction tech-
nologies, but with different stiffnesses. These pavements were loaded with the
MLS10 in order to establish a framework of different pavement tests results and to
have a benchmark for comparison purposes. Moreover, these experiments can be
considered as an evaluation of the Swiss design standards and their ability to make
an accurate prognosis of the bearing capacity limit of these structures.

1.1 The MLS10

The MLS10 at Empa is a prototype APT device developed in 2006 in South Africa
and since 2008 active in Switzerland. Details of its working principle and technical
characteristics can be found elsewhere (Arraigada et al. 2009; Arraigada et al.
2012). In order to set it to an acceptable working level and to avoid the occurrence
Study of the Bearing Capacity of Swiss Standard Pavements … 243

of extensive breakdowns periods during testing phases, the device underwent

important refurbishing processes in the first and second year of activity after pur-
chase. Therefore, the machine was deployed initially in different test sections with
the purpose of understanding its operation mechanism, identifying and improving
possible prototype problems, defining the testing procedures and implementing
measurement and data analysis procedures. Afterwards, the MLS10 was used for
loading a selection of pavements presented in this work, with the aim of calibrating
the machine for different pavement stiffnesses and local climatic conditions. Other
applications of the prototype MLS10 are reported in Hugo et al. (2012).

1.2 Pavement Design in Switzerland

There are many pavement design methods, formulas, software and catalogues for
calculating layers and overall thickness for different types of constructions and
loads. Most of them are still of empirical background. This is also the case in
Switzerland; where the design of pavements is based on the AASHO tests con-
ducted between 1958 and 1960 in Illinois, USA (VSS 2011a). According to this
study, it is possible to define a coefficient named Structural Number (SN) of each
structure. The SN is calculated based on the addition of each layer’s structural
contribution considering a coefficient “a” depending on the material properties and
thicknesses of each layer. It is important to mention that the Swiss dimensioning
norms have developed local values to calculate the SN and therefore, they are
different than those of the AASHTO Guide (AASHTO 1993). This SN is then
compared to the required SN value for a certain traffic class (from T1 to T6),
defined by the number of accumulated Equivalent Single Axle Loads (ESAL) and
an assessment of the bearing capacity of the subgrade. Pavements are also
dimensioned in a second stage against frost. Therefore the final structure thickness
is established according to the values that resist the expected traffic and the frost,
whatever is the most critical condition.
The standards also define a range of different capacities of the subgrade, cor-
responding to the plate loading tests. All pavements are designed to survive a period
of 20 years, were a certain amount of accumulated ESALs are expected. The
conversion of any load to ESALs can be carried out following the well-known
conversion formula obtained in the AASHO tests and defined with more precision
for Switzerland in the standards (VSS 2011b). Due to the uncertainties and the lack
of more precise structural design methods in Switzerland, the tests presented here
are significant to evaluate these standard pavement technologies under the load of
MLS10.
244 M. Arraigada et al.

2 Methodology

Three pavements for low to middle traffic volume constructed ad-hoc are consid-
ered in this study. The selected structures, the loading regime as well as the data
analysis methodology employed are explained in the next sections.

2.1 Pavement Types

The pavements presented in this paper were constructed using local, typical road
construction materials available in Switzerland. The three pavements are shown
schematically in Fig. 1.
As previously described, the types of structures considered were chosen
according to what is mostly used in the country for road construction: a flexible
pavement consisting of different asphalt layers on top of a granular base layer.
Pavement T2 was constructed following the structural design catalogue, taking into
account the required stiffness of a pavement for low volume traffic. The bearing
capacity of the subgrade after compaction was measured with the plate loading test
(VSS 1998), having as an objective to reach a medium class bearing capacity S2,
according the design standard (VSS 2011a).
The other two pavements (T3 and T4) were built in another testing area with
different soil and climatic conditions than the first pavement. The subgrade was
compacted and conditioned to reach similar bearing capacity, as required in the
standards. However, due to the danger of frost, it was decided to build a thicker
granular layer (650 mm). The thickness of this layer is higher than the values
suggested for low and medium traffic volume in the standards, if the pavements are

Fig. 1 Description of the pavement structures considered in this work, including thickness of each
layer, material coefficient “a” and calculation of the SN. The different asphalt concrete layers are
defined following to the standards (VSS 2008)
Study of the Bearing Capacity of Swiss Standard Pavements … 245

Table 1 SN, ESALs and expected number of MLS10 load cycles for each pavement
Name T2 T3* T4*
SN 75 92 108
Max. Nr of ESAL 730,000 2,190,000 7,300,000
Max. expected MLS10 load cycles 102,900 308,700 1,028,900
*Thickness asphalt layers below what recommended by the standards

dimensioned to traffic. Therefore, in order to reach similar stiffness according to the

Structural Number (SN) in the standards, a slight reduction in the asphalt layers
thicknesses was proposed.
Table 1 summarizes the information about the theoretical loading capacity of
each pavement. Each structure is designed to survive a certain amount of ESALs
depending on the stiffness. The calculation of the theoretical number of MLS10 load
cycles, according to the values described in next section is also included in the table.

2.2 MLS10 Loading

For all of the tests, super single tires with a load of 65 kN were used. This
corresponds to a full-scale axle load of 130 kN. The loading speed for class T2
pavement was 21.6 km/h, meaning a rate of 6000 load cycles an hour. In order to
protect the moving parts of MLS10 against fatigue damage, pavements T3 and T4
were loaded with a 5000 load applications per hour. This means a rolling tire speed
of 18 km/h. Pavement T4 was loaded in two locations. In the first location, the
loading was carried out in a season with mild to high temperatures (spring to
autumn). In the second location, the loading phase was carried out in winter, with
relative low temperatures. The reason to do this was to examine the effect of the
temperature on the rutting performance, as is explained later. In all cases, unidi-
rectional loading without lateral wandering was applied. All loading phases were
carried out outside, without temperature control. The total amount of load cycles in
each of the pavements is summarized below:
T2 800,000 load cycles
T3 900,000 load cycles
T4 (mild/high temp.) 1,000,000 load cycles
T4 (low temp.) 380,000 load cycles

2.3 Evaluation of Pavement Response and Distress

The evaluation of the pavements’ bearing capacity consists of the assessment of the
conditions that lead to failure of the structure. The exact definition of failure of a
pavement is however a subjective judgement. In this work, several measurements
246 M. Arraigada et al.

were carried out before, during and after the end of the loading phase to establish a
comparable distress condition. The methods selected are similar to the techniques
used in other APT programs and include not only measurements of pavement
responses but also measurements of the environmental conditions that affect the
pavement. For the sake of brevity, in this work only a selection of measurement
results are presented.
As the temperature has an important influence on the response and distress
development of asphalt pavements, thermocouples were installed in the pavement
structure and measurements were carried out continuously. Permanent deformation
of the structure was evaluated by periodic measurements with a profilometer in
three transverse locations along the wheel path. Visual inspections of the pavement
surface were carried out simultaneously with the profile measurements. Strain
gauges were installed below and on top of the asphalt layers during construction, in
order to monitor the structural deformation under MLS10 loading. Other mea-
surements, like Falling Weight Deflectometer measurements or forensic investi-
gations carried out after the tests are not presented in this paper.

3 Data Analysis and Interpretation

3.1 Temperature

Table 2 reviews the average, minimum and maximum temperatures measured

inside for the duration of the tests. The measurements take into account the values
obtained from a set of sensors installed inside the pavement structure, approxi-
mately 5 cm below the surface, and consider only the temperature measured when
the MLS10 was in service. The average temperatures are relatively similar for all
the pavements, except for the second loading phase of pavement T4. For the first
loading phase of T4 and T3, the temperature ranges are the highest, reaching a
difference of more than 40 °C between the lowest and highest temperatures. The
highest recorded temperature in average was on pavement T2.

Table 2 Recorded temperatures

Pavement Min. temperature Average temperature Max. temperature
(°C) (°C) (°C)
T2 10.8 24.8 37.8
T3 −1.1 18.6 43.1
T4 (mild/high temp.) 0.9 20.9 40.5
T4 (low temp.) −0.8 7.8 19.8
Study of the Bearing Capacity of Swiss Standard Pavements … 247

3.2 Transverse Pavement Profile

As an example, Fig. 2 presents the deformation of the surface of all the pavements
after the accumulated application of 300,000 MLS10 load cycles. This figure shows
the shape of the channelized loading of the super single tires in all the pavements
simultaneously. The fact that none of these shapes present any elevation at the
edges of the rut indicates that there is no sideward shoving of asphalt material from
underneath the tire-path to the sides. In fact, the weaker pavements (T2 and T3)
present a pronounced downwards deformation at each sides of the rut. This brings
into evidence that the deformation in the weaker pavements was mostly produced
by the post-compaction of the granular layer, due to the loading of the MLS10. In
pavement T4 however, the deformation is mostly allocated in the asphalt layers.
The influence of the temperature is quite evident comparing the profiles of the T4
pavement loaded at mild/high and low temperatures. In the same pavement, the
temperatures registered between 400,000 and 500,000 load applications reached
40.5 °C and had a strong influence on the permanent deformation. This can be
observed in Fig. 3 which displays the final condition for each pavement. The shape
of the profiles indicate that in pavement T4 there is strong shoving of the asphalt
material to the sides of the rut.
The rutting of each measured profile, calculated as the difference between the
highest upward peak in the profile and the averaged rut depth under the tire, is
presented in Fig. 4. The curves show the progression of the rutting in terms of
accumulated load cycles. As expected, increase of rutting is higher at the beginning
of the loading phase and decreases with the accumulation of load cycles, presenting
a curve that can be modelled with an exponential function. The effect of the high
temperature period is evident for pavement T4.

Fig. 2 Transverse pavement profiles at 300,000 load cycles

248 M. Arraigada et al.

Fig. 3 Transverse pavement profiles at the end of the loading phase. Missing data is due to
problems in the profile measurements

Fig. 4 Comparison of the progression of the rut in each test

The failure criterion for flexible pavements of surface rutting used in this work is
20 mm. The number of load cycles required to reach this rutting value was
approximately:
T2 137,000 load cycles
T3 530,000 load cycles
T4 (mild/high temp.) 1,000,000 load cycles
These values are higher than the theoretical maximum expected number of
MLS10 load cycles except for the pavement T4, loaded at mild/high temperatures.
Study of the Bearing Capacity of Swiss Standard Pavements … 249

In fact, the strong influence of the high temperatures in the development of the
rutting can be clearly observed in Fig. 4; particularly in pavement T4 which has
noticeable increase the rut depth as a consequence of a period of high temperatures.
In order to account for the influence of temperature and accumulated load
applications in the progression of rutting, the rut rate was defined as the rut depth
produced by one single loading pass. It was calculated as the increment of rut
divided by the number of load cycles between the profile measurements. These
values, together with the maxima, minima and average temperatures recorded in all
pavements are presented in Fig. 5. This figure confirms that the prediction of rutting
for a flexible pavement structure without temperature control is a complex task and
results cannot be extrapolated with confidence to the same structural types. The rut
rate correlates with the temperature fluctuations but also is defined by the previous
compaction due to the loading history and, of course, by the asphalt layers materials
and thicknesses.

3.3 Cracking

The pavements were regularly inspected by eye to asses crack development on the
surface. All sections showed visible cracking at the end of the loading phase. The kind
of cracks observed can be classified in longitudinal cracks, at the edge of the rut due to
the large deformation of the pavement, and in cracks in the wheel path, due to fatigue.
Figure 6 shows the final conditions of the pavements after loading. The cracks are
highlighted with a line marker. It can be observed that the pavement with the highest
amount of cracks at the end of the loading phase was T3. In this section, the cracks
observed were mostly transversal to the driving direction and in many locations they
showed pumping of fine aggregates of the base layers. This is a consequence of the
low temperatures registered during the first half of the loading phase.
Table 3 contains the number of load cycles when the first cracks appeared in the
pavement surface and is an indication of the rate of deterioration of the structures.

3.4 Strain Measurements in the Pavements

Strain gauges were embedded in the pavements to obtain the deformation response
under MLS10 loading in order to estimate the stiffness of the structure. Any change
in the response that is not related to temperature changes or different loading
conditions is considered as a modification of the structural stiffness.
Measurements of strains were performed at regular intervals. The analysis of the
datasets was carried out by calculating and plotting the strain amplitudes of single
MLS10 passing vs. the amount of accumulated load cycles. Any variation in the
trend of the strain amplitudes could be an indication of a change in the structural
stiffness. In order to assess the influence of the temperature, only strains recorded at
250 M. Arraigada et al.

Fig. 5 Average, max. and

min. temperatures (dotted
lines), and rut rate in μm per
load cycle (straight lines)
Study of the Bearing Capacity of Swiss Standard Pavements … 251

Long.shear
cracks

Fatigue
T2 cracks T3 T4

Fig. 6 Cracking patterns on the surface of each pavement at the end of the tests

Table 3 Cracking Pavement T2 T3 T4

progression
Edge cracks 120,000 350,000 450,000
Fatigue cracks 300,000 500,000 750,000

a certain temperature level are considered. Further the shape of the strains were
visually controlled because only taking into account the pure strain amplitudes
might lead to a wrong interpretation of the results. In this way, an alteration in the
strain field can be detected and considered as a possible effect of the stiffness
alteration.
In this work only a selected amount of strain gauges results are presented.
Figure 7 presents the strains measured by a sensor located between the granular
layer and the asphalt layers in the direction perpendicular to traffic. The upper part
of the figure corresponds to the results obtained in pavements T2. Below are the
results of pavement T3. The curves show typical tension strains that don’t change
significantly with the accumulation of load cycles. The tension values are consid-
erably smaller in T2 than in T3 and in the latter, they increase until 400,000 load
cycles. Afterwards, they start to decrease. This can be an indication of a consid-
erable change in pavement stiffness due to the initiation and propagation of cracks.
Figure 8 shows the strains for a sensor positioned under the asphalt layers and
coincident, e.g. in line to the traffic direction. The top part of the figure corresponds
252 M. Arraigada et al.

Fig. 7 Strain curves and amplitudes for pavement T2 (top) and T3 (bottom). They correspond to a
sensor located below the asphalt layers and transversal to the traffic direction

to pavement T3, while the lower part presents the results for pavement T4. In
general, the curves show typical compression-tension-compression peaks although
they present slight changes throughout the tests. The strain amplitudes are, as
expected, higher in pavement T3 and present in this case a turning point near
350,000 load applications. In pavement T4, the 700,000 load cycles mark seems to
be a turning point and might show a significant change in the stiffness of the
pavement.

3.5 Failure Mechanism

After taking trenches from all the pavements, the failure mechanism can be
observed. All the pavements have a similar distress characteristic: they present a
Study of the Bearing Capacity of Swiss Standard Pavements … 253

Fig. 8 Strain curves and

amplitudes for pavement T3
(top) and T4 (bottom). They
correspond to a sensor located
below the asphalt layers and
coincident to the traffic
direction
254 M. Arraigada et al.

Interlayer debonding Cracks

Fig. 9 Punching of the asphalt layers into the granular base layer with initiation of fatigue cracks
in the bottom of the bound material

punching of the asphalt layers into the granular base layer. They also show
bottom-up cracks and interlayer debonding close to the rut edges due to the large
rutting deformations. The trench for pavement T4 can be observed in Fig. 9.

4 Summary and Conclusions

This paper summarizes the results of a series of tests carried out to evaluate the
performance of different pavements constructed using conventional structures and
materials used in Switzerland. The results should provide a benchmark for esti-
mating the bearing capacity limit and life-span of different structures under the load
of the prototype MLS10. Furthermore, the findings would serve as a verification of
the Swiss dimensioning standards.
Typical pavement performance measures such as rutting progression and crack
growth by means of visual inspection as well as strain development in the pavement
are presented in this work. In general, it can be stated that designing the pavements
according to the Swiss design method is safe. Considering rutting performance,
general pavements survived considerably more than the maximum expected amount
of loads. Temperature is however a factor that needs to be controlled in future tests.
The same can be concluded when analysing crack and strain data: cracks developed
later than expected. Temperature however, has a big influence to these results.
Therefore it is recommended that in future experiments, the object of study should
be tested simultaneously with a reference, standard pavement for comparison
purposes.
Study of the Bearing Capacity of Swiss Standard Pavements … 255

References

AASHTO. (1993). AASHTO guide for design of pavement structures. Washington, D.C.:
American Association of State Highway and Transportation Officials.
Arraigada, M., Kalogeropoulos, A., Hugenschmidt, J., Partl, M. N., Caprez, M., & Rabaiotti, C.
(2009). Pilotstudie zur Evaluation einer mobilen Grossversuchsanlage für beschleunigte
Verkehrslastsimulation auf Strassenbelägen. Eidgenössisches Departement für Umwelt,
Verkehr, Energie und Kommunikation, Bundesamt für Strassen. ASTRA 2004/018, Report
Nr 1261.
Arraigada, M., Partl, M. N. & Pugliessi, A. (2012) Initial tests results from the MLS10 mobile load
simulator in Switzerland. In Proceedings of the 4th international conference on accelerated
pavement testing, Davis, CA, USA (pp. 277–285). September 19–21, 2012.
Hugo, F., Arraigada, M., Shuming, L., Zefeng, T., & Kim, Y. R. (2012). International case studies
in support of successful applications of accelerated pavement testing in pavement engineering.
In Proceedings of the 4th international conference on accelerated pavement testing, Davis, CA,
USA (pp. 93–104). September 19–21, 2012.
Schweizerischer Verband der Strassen- und Verkehrsfachleute VSS. (2008). Asphaltmischgut –
Mischgutanforderungen– Teil 1: Asphalt-beton (Asphalt mixes – requirements – Part 1 Asphalt
concrete). Swiss Standard SN 640431-1 (in German).
Schweizerischer Verband der Strassen- und Verkehrsfachleute VSS. (1998). Böden -
Plattendruckversuch EV und ME (Subgrade – Plate-loading test EV and ME). SN 670317b
(in German).
Schweizerischer Verband der Strassen- und Verkehrsfachleute VSS. (2011a). Dimensionierung
Strassenaufbaus, Unterbau und Oberbau (Dimensioning of road constructions, subgrade and
pavement). Swiss Standard SN 640324 (in German).
Schweizerischer Verband der Strassen- und Verkehrsfachleute VSS. (2011b). Dimensionierung
des Strassenaufbaus Äquivalente Verkehrlast (Dimensioning of equivalent loads of road
constructions). Swiss Standard SN 640320 (in German).
The Development of APT Methodology
in the Application and Derivation
of Geosynthetic Benefits in Roadway
Design

Jonathan Cook, Michael Dobie and David Blackman

Abstract Geosynthetic materials are used in roadway pavement construction to

improve performance. In order to quantify performance it is necessary to carry out
trafficking trials, the results of which may be used empirically to establish or modify
pavement design methods. Several such trials have been carried out over the last
30 years, in various configurations. Eight trials carried out by TRL over the period
2000–2013 are described in detail, providing information on both the pavement
construction and the performance testing. These trials were all carried out in a
consistent manner on a single layer of sub-base quality material over a clay sub-
grade, with relatively small test areas constructed indoors in a pit, and then traf-
ficked with an automated wheel load. Important trends and findings from examining
the results of all 93 test panels are summarised in order to establish whether the
geosynthetic improves pavement performance by confinement or by tensioned
membrane. Of these two mechanisms only confinement is relevant to permanent
pavements, and performance limits are established which should be achieved for
confinement to be the dominating mechanism.

1 Introduction

Engineers have been benefiting from the inclusion of geosynthetic materials in

roadway pavement construction for over 30 years. The ability to increase the traffic
carrying capability of a roadway using geosynthetics has now become accepted

J. Cook (&)
Tensar International, Blackburn, UK
e-mail: [email protected]
M. Dobie
Tensar International, Jakarta, Indonesia
D. Blackman
Transport Research Laboratory, Wokingham, UK

© Springer International Publishing Switzerland 2016 257

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_17
258 J. Cook et al.

technology. This in turn provides opportunities to rationalise the pavement sub-base

construction thickness allowing projects to benefit from environmental and eco-
nomic cost reduction. Existing guidance from several public authorities around the
world such as AASHTO, FHWA, UK Highways England and Norwegian Roads
has been generated after several years of positive experience using geosynthetics.
Potential users are guided that the benefits of geosynthetics can be used where
ground conditions are poor (IAN 73, UK) or that they can reduce layer thickness
and/or increase trafficking life (AASHTO R50/09). However these benefits cannot
be derived theoretically, and in common with most pavement design methods must
be established empirically, by carrying out pavement trafficking tests and trials.
Table 1 lists a number of trafficking trials carried out over the last 30 years,
where the main aim was to establish the benefit of including geosynthetics in the
granular layer. In some cases the pavement was a single granular layer, and in other
cases an asphalt surfacing was included (see column AC). Also of relevance is the
type and size of the pavement under test (see column Type), which varies from
relatively small pits, to longer straight tracks and in some cases oval tracks. Some
trials were outdoors and others were covered.
Accelerated Pavement Testing (APT) technology and methodology has been
developed and used by the Transport Research Laboratory (TRL) in UK in eight
trials since 2000 on behalf of Tensar International, first reported by Jenner et al.
(2002). An important aim of these developments has been to reduce test variables
and external influences to a minimum, in order to generate ever more consistent,

Table 1 Trafficking trials carried out with the aim of quantifying the performance of
geosynthetics
Trial AC Area (m2) Type Covered Traffic Max Reference
ESAL
TRL No 85 Pit Yes Truck 3,100 Chaddock (1998)
USCoE Yes 670 Track Yes Truck 10,000a Webster (1991)
Uni Yes 448 Oval No Truck 80,000 Cancelli and
Milan Montanelli (1999)
Montana No 780 Track No Truck 120 Cuelho and Perkins
Ph1 (2009)
USCoE Yes 111 Track Yes APT 100,000 Jersey et al. (2012)
Ph1
TRI Yes 555 Oval No Truck 10,000 Wrigley and Valero
(2013)
Montana No 1260 Track No Truck 900 Cuelho et al. (2014)
Ph2
USCoE Yes 93 Track Yes APT 200,000 Norwood and Tingle
Ph2 (2014)
TRL No 24 per Pit Yes APT 15,000 This paper
T1-T8 section
a
Trafficking applied by 13T single wheel
The Development of APT Methodology in the Application … 259

Table 2 Summary of the eight trials carried out by TRL since 2000
Trial No Sections Panels Controls Non-standarda Panels with geosynthetics
T1 4 12 2 0 10
T2 3 9 2 0 7
T3 2 6 2 0 4
T4 3 9 3 0 6
T5 3 9 0 0 9
T6 4 12 1 5 9
T7 6 18 3 0 15
T8 6 18 1 0 17
Totals 31 93 14 5 77
a
Standard is CBR = 2 % and thickness = 300 mm approximately

Rut depth

Deformation

Fig. 1 Definition of surface deformation and rut depth

reliable and repeatable results. Therefore the tests are carried out in a 10 m long pit,
on a single layer of sub-base quality material supported by a clay subgrade.
Trafficking is carried out indoors by an automated trafficking machine. The main
features of the eight trials carried out are summarised in Table 2.
This paper describes in detail the method of construction and quality control of
the test panels required to create consistent pavement sections, as well as the
technique used for trafficking. This extensive and unique programme of work has
generated a significant amount of trafficking data allowing the mechanism of
interaction between aggregate and geosynthetic to be examined in detail, as dis-
cussed in the final section. In reading this paper, it is important to be aware of the
terms used to define surface performance, namely rut depth and deformation, as
shown in Fig. 1.

2 Accelerated Pavement Testing Methodology

2.1 Pavement Test Facility

The Pavement Test Facility (PTF) at the Transport Research Laboratory,

Wokingham, UK is an indoor testing facility and the largest within the UK. It
260 J. Cook et al.

comprises a concrete pit measuring 10 m wide by 25 m long by 3 m deep with a

gantry spanning the width of the pit. The main features of the PTF are given in
Table 3 and a general view of the PTF is shown in Fig. 2. A carriage fitted with a
test wheel, shown in Fig. 3, is suspended from the underside of the gantry, and
trafficking is achieved by moving the carriage backwards and forwards across the
full width of the test pit. A typical test pavement is 10 m long by 2.4 m wide
although the size can be increased or decreased as required. For the trials covered in
this paper, all the sections were 2.4 m wide but concrete blocks were placed along
the inside edge of the test pit to support the wheel at the start and end of each pass,
which reduced the length of the test section to 8.8 m. Each section was sub-divided
into three sub-sections, the two outer sub-sections were 3.15 m long and the central
sub-section was 2.5 m long.

Table 3 Main test features and specification of test parameters for the PTF at TRL
Test feature Specification
Wheel load and accuracy Variable from 23 to 100 kN ±2 % of set load
Wheel speed and accuracy Variable 1–20 km/h ±0.25 km/h
Wheel configuration Dual pair or wide-base single
Lateral positioning of wheel Centreline ±450 mm in 1 mm steps (NB: fixed position
used for these trials)
Direction of loading Uni-directional or bi-directional
Maximum rate of trafficking 1000 passes per hour

Fig. 2 General view of the pavement test facility (PTF)

The Development of APT Methodology in the Application … 261

Fig. 3 View of PTF carriage fitted with dual wheel assembly

The vertical loading, direction of travel, speed and lateral position of the wheel, as
well as the wheel type configuration and tyre inflation pressure, may be selected as
appropriate for different test scenarios. Pavements in the PTF can be trafficked at
either ambient temperature or at temperatures above ambient to a maximum of 45 °C
by use of a stand-alone infra-red heating system controlled by thermocouples
installed in the test pavement.

2.2 Subgrade Preparation

Working with natural clay materials at full scale presents challenges in terms of
being able to achieve a uniform level of stiffness across a large test area. In each test
since 2000, the clay sub-grade was a grey silty London clay described by
Casagrande’s extended soil classification as Type CH, a clay of high plasticity.
For each of the eight trials conducted in this test programme the target strength
for the subgrade was 2 % CBR. As other experiments had been undertaken between
the different trials the strength, depth and area of available existing subgrade was
often not suitable. In order to provide a uniform and comparable subgrade layer for
each of the trials the existing clay within the pit over the area required for the trial
was excavated to a depth of about 1.5 m and reconditioned. To generate enough
suitably conditioned material often required the addition of extra clay which was
sourced from the original quarry in Colnbrook, Berkshire, UK and/or existing
stockpiles of the material held at TRL. Using the same source for the supply of the
262 J. Cook et al.

clay for each test provides a suitable degree of consistency. To ensure uniformity
within the subgrade layers and across the trial area, the different sources of clay
were thoroughly mixed together during the conditioning process to minimise
possible differences in the characteristics and subsequent performance of the
material during trafficking. Conditioning of the clay was achieved by spreading the
material out on a concrete hard standing, spraying it with water and mixing it with a
tractor mounted rotovator, as shown in Fig. 4. It was not possible to rotovate the
full depth of the material due to the underlying concrete hard standing so it was
turned at intervals during the conditioning process using a tracked excavator. This
also further ensured a thorough mix of the different sources of clay and reduced the
likelihood of localised wet or dry spots.
During this conditioning process the CBR of the clay was monitored through
regular sampling for MCV (moisture condition value) testing and moisture content
determination. One advantage of repeatedly using the same source of clay was that
a very good understanding of the relationship between strength and moisture
content was established and how the material gained strength with time after mixing
and compacting.
After conditioning, the clay was placed and compacted in layers in accordance
with the Series 600 of the UK Specification for Highway Works (Manual of
Contract Documents for Highway Works or MCHW 2016, Volume 1). Each
compacted layer was nominally 110 mm thick and compaction was achieved with
12 passes of a Rammax RW2900 pad foot roller having a mass per unit width of
roll of 1250 kg/m. To ensure the finished surface level was at the correct depth in
relation to the top of the pit slightly more clay was placed and compacted than

Fig. 4 Conditioning the subgrade

The Development of APT Methodology in the Application … 263

Fig. 5 Finished subgrade surface

actually required. The excess clay was then graded off and the surface was ‘sealed’
by 4 passes of a non-vibrating smooth roller. This approach also produced a flat
even surface finish to all the sections which helped eliminate variations between
sections and successive trials thereby aiding direct comparative performance of the
different geosynthetics. A view of the finished subgrade surface, marked out for the
different sections and sub-sections is shown in Fig. 5.

2.3 Sub-grade Strength Measurement

Following construction of the subgrade layer a series of tests were undertaken to

determine a relationship between the CBR value of the clay and the Cone Index
(CI) reading from a soil assessment cone penetrometer developed by the British
Military Engineering Experimental Establishment (MEXE). This was achieved by
measuring the CI readings at a number of locations within the test pit and
comparing them with the measured CBR values determined from laboratory tests
performed on samples taken adjacent to these locations. Typically the relation
between the CBR value and the CI value for the MEXE penetrometer fitted with a
20 mm cone over the eight trials was:

CBR ¼ 0:035  CI

where
CI = the average value for the top 150 mm of subgrade.
264 J. Cook et al.

CBR (%) CBR (%)

0 1 2 3 4 5 0 1 2 3 4 5
0 0
Nine MEXE Nine MEXE
Depth below subgrade surface (mm)

Depth below subgrade surface (mm)

100 profiles are 100 profiles are
executed for executed for
200 each panel 200 each panel
300 300

400 400
Before After
500
trafficking 500
trafficking
T7/5A T7/5A
600 600
T7/5B T7/5B
700 700
T7/5C T7/5C
800 800

Fig. 6 Typical CBR profiles based on MEXE probe readings for three panels in one section

These measurements were taken at least twenty four hours after the final com-
paction as historically it had been noted that the measured CBR increases with time
and this is often associated, in part, with dissipation of the pore pressures within the
clay. Penetrometer profiles were measured at nine locations per panel, and examples
of typical profiles for one test section (consisting of three panels) are shown in
Fig. 6, which includes all measurements taken (27 profiles with 12 readings per
profile) both before and after trafficking.

2.4 Granular Sub-base Grading and Placement

For all the trials a crushed granite sub-base conforming to the requirements of the
800 Series of the Specification for Highway Works (SHW) (MCHW 2016 Volume
1) for a Type 1 material was used. Granite was chosen to reduce the likelihood of
any cementing action that could occur if a limestone material was used. To ensure
there was minimal variation over the different sections within each trial in terms of
the grading and moisture content of the sub-base material all the required lorry
loads were delivered to site in advance. They were then mixed together to reduce
segregation and moisture content variations which can occur during stock piling
and transportation. Although there is now no specified optimum moisture content
requirement for Type 1 material, tests were taken to ascertain the optimum moisture
content value and the actual moisture content of the material prior to laying. If it
was found to be too wet or dry it was either allowed to dry out or additional water
was added to bring the moisture content as close as possible to optimum.
Considerable care was taken when placing and spreading the sub-base such that
the geosynthetics were not disturbed or damaged and no granular material got
underneath them. To achieve this, they were pre-cut to size, placed on the subgrade
surface and weighted down with concrete blocks. In Trials T1–T7 the products
were abutted to each other on all sides with no overlap, however during trafficking
The Development of APT Methodology in the Application … 265

local deformation occurred along the wheel path at the interface of the sub-sections,
where the grid type changed. To reduce this effect during Trial T8 the products
were overlapped longitudinally by 150 mm; there was no overlap of the products
between adjacent sections. The sub-base was then placed and spread in two layers,
each having a target compacted thickness of 150 mm, using a tracked excavator.
Once the first layer had been loosely spread the concrete blocks weighing down the
geosynthetics were removed. Compaction was achieved with eight passes of a twin
drum tandem vibrating roller falling into the over 1300 up to 1800 kg mass per
metre width of vibrating roll as given in Table 8/4 of the 800 Series of the SHW.

2.5 Pavement Surface Stiffness

Whilst considerable care had been taken to ensure that the subgrade and sub-base
were comparable over all the sections, in order to ascertain if there were any large
differences which could explain subsequent performance variations between the
different sections the pavement stiffness was measured using a Falling Weight
Deflectometer (FWD) described by Sorensen and Hayven (1982). The FWD was
fitted with a 300 mm diameter segmented loading plate and the weight fell from a
predetermined height to produce a stress of 150 kPa on the surface of the sub-base.
The pavement stiffness (E) was calculated from the equation:

E ¼ 2qað1  m2 Þ=d

where:
q = stress under the plate ν = Poisson's ratio of the sub-base (taken to be 0.45)
a = radius of the plate δ = maximum deflection at the centre of the plate

3 Trafficking

For each of the trials in these series of testing, trafficking was applied
bi-directionally with no lateral deviation at a speed of 15 km/h. Loading was
applied using a dual wheel assembly, tyre type 295/80R22.5, with a wheel load of
40 kN and a tyre inflation pressure of 700 kPa.
During the construction process optical levels were taken on a 0.5 m grid on both
the subgrade and sub-base surface in order to determine the thickness of the sub-base
layer. In addition, two transverse profile lines were established on the subgrade and
sub-base surface within each sub-section and optical levels were taken every
100 mm along these lines. The measurements on the sub-base surface were repeated
at intervals during the trafficking so the development of surface deformation could be
tracked as shown in Fig. 7. This figure also shows that the wheel path was nominally
266 J. Cook et al.

Distance across section (m)

0 0.4 0.8 1.2 1.6 2 2.4
40 N=0
Sub-base N = 100
20 surface N = 200
Surface level (mm)

0 N = 500
N = 1000
-20 N = 2000
N = 3000
-40
N = 5000
-60 N = 7500
N = 10000
-80

Fig. 7 Development of transverse surface profile with trafficking

600 mm wide and centred along the middle of each section with the geosynthetics
extending 900 mm beyond the edges of the wheel path.
At the end of trafficking the sub-base was removed in order to retrieve the
geosynthetic products. This also enabled the surface profile of the subgrade to be
re-measured to allow the surface profiles of the two layers to be compared, as
shown in Fig. 11 for six panels as discussed in the following section of this paper. It
also enabled the strength of the subgrade to be re-assessed using the MEXE
penetrometer and Fig. 8 shows the change between the mean initial subgrade CBR
and the CBR 4 months after trafficking, for part of Trial T7.
Optical levels were also taken along the centre of each wheel of the dual pair to
record surface deformation with trafficking. As optical levelling was a time con-
suming exercise the development of the surface profile was assessed during traf-
ficking using a 2 m straight edge and wedge, as shown in Fig. 9 (left). This is a
relatively quick exercise and allowed an informed decision, based on the increase in
rut depth, to be made as to when the next set of optical level measurements should
be taken. The change in rut depth with increasing trafficking for six panels
discussed in the following section of this paper are shown in Fig. 13.
As previously mentioned, after trafficking the sub-base and geosynthetics were
removed which also allowed the rut at the subgrade surface to be measured as
shown in Fig. 9 (right). One benefit of taking both optical level measurements and
rut depths is that further comparison of performance can be made as optical levels
give the relative change in surface profile with respect to a fixed datum whereas rut
depth gives a measurement for the total deformation and heave that has occurred.

4
S1 Initial S1 Final S2 Initial S2 Final S3 Initial S3 Final
3
CBR (%)

0
S1A S1A S1B S1B S1C S1C S2A S2A S2B S2B S2C S2C S3A S3A S3B S3B S3C S3C

Fig. 8 Subgrade CBR at the start and end of the trial

The Development of APT Methodology in the Application … 267

Fig. 9 Surface rutting under the straightedge (left) and rut on subgrade surface (right)

4 Discussion of Trends Observed in the TRL Trafficking

Trials

The eight TRL trials outlined in the introduction and described in detail in the
previous sections of this paper provide an incomparable set of data concerning the
performance of a single layer of sub-base quality material under trafficking. Most
panels had the same target arrangement, namely thickness = 300 mm and
CBR = 2 %. Examination of the data indicates that there are variations in beha-
viour, with occasional panels appearing to be anomalous compared to others.
However with a total of 93 panels tested, it is possible to establish general trends of
behaviour, to take averages when trends are similar and to reject outliers. These
benefits only arise because such a large body of data exists.
A specific issue with regards to defining the performance of granular pavements
that incorporate a geosynthetic layer is the mechanism of interaction with the
aggregate particles which develops as trafficking takes place. Of the 93 panels
tested, 77 included a geosynthetic material placed between the sub-base and the
subgrade, as summarised in Table 2. A wide range of geosynthetics have been
tested, all of which are intended to provide an improved performance compared to a
control:
• Geogrids made by punching and stretching a polymer sheet (both biaxial and
multi-axial)
• Geogrids made by extrusion and stretching
• Welded geogrids
• Woven geogrids
• Woven geotextiles
• Geocomposites
• NB: non-woven geotextiles intended to be used as separators or filter layers
have not been tested
268 J. Cook et al.

The main aim of the discussion presented in this section is not to associate
performance with the specific geosynthetics used, but rather to investigate the
principal mechanism developed in each test panel in order to establish trends and
limitations. Generally all the geosynthetics tested in the TRL trials had a beneficial
effect on surface performance, but to varying degrees depending on the principal
mechanism developed. Giroud (2006) discusses mechanisms which may develop
when geosynthetic reinforcement or stabilisation materials are used in pavements,
in particular the mechanisms depicted in Fig. 10: confinement and tensioned
membrane.
Important points related to these mechanisms are summarised as follows:

Confinement Applies to geogrids which can interlock effectively with the aggregate
particles of the layer immediately above them. The benefit comes from
improved load distribution due to lateral restraint (here termed
confinement) as well as reducing aggregate deterioration. In a simple way
effective confinement makes it very difficult for the aggregate to dilate, but
dilation must take place if deformations are to become large (i.e. for
development of the tensioned membrane, discussed below). Confinement
works with small deformation, and is therefore applicable to permanent
surfaced pavements. The geogrid operates at very low strain, and the
general term used to describe this mechanism is mechanical stabilisation.
Tensioned Applies to both geotextiles and geogrids which are unable to develop
membrane confinement. A large deformation of the geosynthetic is developed due to
the formation of a deep rut in the subgrade surface. This results in a
significant tensile load being developed in the geosynthetic and the vertical
component of that load helps support the wheel. This further requires that
the geosynthetic immediately beyond the rut is anchored by the weight of
the pavement material above the geosynthetic combined with its pull-out
resistance, and therefore acts as a reinforcement. Giroud (2006) confirmed
that the tensioned membrane effect was relatively small and limited to
channelised traffic on unsurfaced pavements. It is not applicable to
trafficked areas because the traffic is not channelised, nor to surfaced roads,
because the deformations required to create the tensioned membrane would
not be acceptable at the base of a permanent surfaced pavement.

Confinement Tensioned
membrane

Fig. 10 Two mechanisms of support: confinement on left and tensioned membrane on right
The Development of APT Methodology in the Application … 269

In the TRL trials all 77 panels with a geosynthetic have been examined to
establish how their final condition (i.e. as determined after excavation of the
sub-base material, see Fig. 9) relates to the two mechanisms depicted in Fig. 10,
namely confinement or tensioned membrane. Not surprisingly there is a wide
variety of performance, ranging from perfect confinement to perfect tensioned
membrane. Therefore a simple classification has been established to rate the final
pavement shape, as described in Table 4.
In order to illustrate the classifications given in Table 4, the initial and final
sections for six panels, all with geogrids at the subgrade/sub-base interface, are
shown in Fig. 11. Table 5 summarises all of the main performance data for these
panels, which include total passes reached, as well as passes to reach 50 mm
surface deformation (defined as the deformation measured from the original surface
level, see Fig. 1). The table also gives final surface deformation and final heave
(taken as measured rut depth minus surface deformation). An important perfor-
mance parameter is the rate of deformation increase at the end of trafficking, given
as deformation in mm per 1000 passes.
The middle part of Table 5 gives the initial section parameters, and it can be seen
that all sections have a mean thickness close to 300 mm and CBR close to 2 %. The
final column (headed MEM*) is the only exception, being a non-standard sec-
tion 250 mm thick. The relevance and performance of this section is discussed
later. Initial surface stiffness is included in Table 5, as well as a parameter termed
the Road Index (or RI). Road Index is defined as follows:

RI ¼ 10  h  CBR0:5

where
h = thickness of sub-base quality material (in m), CBR = of subgrade in percent.
This expression is similar to the design equation for unsurfaced pavements
published by Giroud and Noiray (1981), but with the CBR raised to the power of
0.5 rather than 0.63 (using 0.5 was found to give better agreement with the design
method for granular pavements published in Austroads 2012). The reason for

Table 4 Definition of terminal pavement shape and condition

Pavement shape Description
classification
CON Perfect confinement, subgrade deformation negligible
CON- Confinement, with some deformation of the subgrade surface but
negligible development of the membrane shape
50/50 Some subgrade deformation with membrane shape beginning to
develop
MEM- Clearly defined membrane behaviour, but subgrade rut significantly
smaller than surface rut
MEM Perfect membrane, with surface rut and subgrade rut of much the same
size and geometry
270 J. Cook et al.

Distance across section (m) Distance across section (m)

0 0.4 0.8 1.2 1.6 2 2.4 0 0.4 0.8 1.2 1.6 2 2.4
200 200
Behaviour = CON Behaviour = CON-
100 100
Sub-base surface
Surface level (mm)

Surface level (mm)

0 0
Sub-base
-100 -100 surface

-200 -200

-300 -300
Subgrade surface Subgrade surface
-400 -400

Distance across section (m) Distance across section (m)

0 0.4 0.8 1.2 1.6 2 2.4 0 0.4 0.8 1.2 1.6 2 2.4
200 200
Behaviour = 50/50 Behaviour = MEM-
100 100
Sub-base surface
Surface level (mm)

Surface level (mm)

0 0
Sub-base
surface
-100 -100

-200 -200

-300 -300 Subgrade

Subgrade surface surface
-400 -400

Distance across section (m) Distance across section (m)

0 0.4 0.8 1.2 1.6 2 2.4 0 0.4 0.8 1.2 1.6 2 2.4
200 200
Behaviour = MEM Behaviour = MEM*
100 100
After 4600 passes
Surface level (mm)

Surface level (mm)

0 Sub-base 0
Sub-base
surface
-100 -100 surface

After 10,000
-200 -200
passes

-300 Subgrade -300 Subgrade

surface surface
-400 -400

Fig. 11 Transverse surface profiles before and after trafficking

including the factor 10 is to give numbers to one decimal place for the resulting
index values. Making reference to Austroads (2012), for low volume roads RI is in
the range 5–7, whereas for full pavements RI is from 7 to 14. The purpose of the
Road Index is to give a simple method of comparing pavements from trial to trial,
as well as assessing the relevance of trials to real pavements. As stated above, RI is
only relevant to single layer pavements of sub-base quality material, but different
types of pavement material can be taken into account by using equivalence factors.
The Development of APT Methodology in the Application … 271

Table 5 Specific section and performance data for the six pavements shown in Fig. 11
CON CON- 50/50 MEM- MEM MEMb
Trafficking parameter
Total passes of ESAL (N) 11,500 10,000 10,000 10,000 4,600 1,000
ESAL for 50 mm deformation (N) NR NR 7,000 3,300 1,750 330
Final surface deformation (mm) 38 45 56 69 103 85
Final surface heave (mm) 0 13 16 48 96 76
Final deformation per 1000 passes 1.3 1.2 2.0 2.4 30 36
(mm)
T4 T2 T2 T7 T7 T6
Initial section parameters/trial
Subgrade CBR (%) 1.84 2.20 2.19 2.22 2.31 2.06
Section mean thickness (mm) 314 286 292 298 302 255
Surface stiffness from FWD (MPa) 40.1 34.7 33.5 32.4 42.0 22.4
Road index (RI) 4.3 4.2 4.3 4.4 4.6 3.7
Initial section parameters of controla
Subgrade CBR (%) 1.85 2.08 2.09 2.14b 2.14b NA
Section mean thickness (mm) 299 280 292 301b 301b NA
Surface stiffness from FWD (MPa) 40.2 29.4 30.8 39.6b 39.6b NA
Road index (RI) 4.1 4.0 4.2 4.4b 4.4b NA
NA not available, NR not reached
a
Nearest available control
b
Mean of three controls

Although not a parameter intended for use in design, the RI scale is logarithmic in
nature, with RI = 7 having 10 times the life in terms of ESAL compared to RI = 6.
The lower part of Table 5 includes the section parameters for the nearest control
(i.e. panel without a geosynthetic) to each section shown in Fig. 11. Again it can be
seen that all control sections have a mean thickness close to 300 mm and CBR
close to 2 %. The Road Index values are much the same for all sections, and on
average, the TRL trials have RI = 4.3, with slight variation as seen in Table 5 (with
the exception of the 250 mm thick panel). Also of interest is the initial surface
stiffness for each panel, as measured using the FWD. Inspection of the values in
Table 5 indicates that there is more variation in surface stiffness than in RI, which
probably indicates slight variations in the sub-base material itself from trial-to-trial.
Inspection of the six sections in Fig. 11 indicates major differences in perfor-
mance. Assignment of the pavement shape classification is subjective to a certain
extent. Both CON and CON- appear to be similar, except that CON- has a slight rut
development in the subgrade surface, especially on the right side in the case
included in Fig. 11. The section designated 50/50 clearly shows the tensioned
membrane shape appearing at the subgrade surface, but it is relatively small
compared to MEM-, the next designation. The MEM designation indicates major
disruption and distortion at the subgrade level, to a similar extent as the sub-base
272 J. Cook et al.

surface (it should be noted that the subgrade profile in MEM was measured on
completion after 10,000 passes).
The other performance parameters listed in the upper part of Table 5 are all
measured values taken from the detailed trafficking records as described in Sect. 3
of this paper, and are not subjective. By associating these parameters with the
pavement shape classification, is has been possible to establish limits which define
whether or not confinement is the relevant mechanism in any particular panel at the
termination of trafficking. The resulting limits are defined in Table 6. By com-
parison with the actual performance parameters from Table 5, it can be seen that
CON and CON- both fall close to or inside the limits, whereas 50/50 falls outside
all limits except that it does reach full term.
It is also helpful to view the performance of these same sections in terms of
deformation versus the number of passes (1 pass = 1 ESAL for TRL trials). All six
sections are shown combined in Fig. 12, and based on this, it might be guessed that
CON, CON-, 50/50 and MEM- all have a similar performance, as they go full term,
and appear to flatten off. However comparing CON and MEM- in Fig. 11 confirms
that there are substantial differences, and that the geosynthetic used for MEM- is
behaving as a tensioned membrane. By way of providing a baseline to the beha-
viour seen in Fig. 12, an average trafficking result for all associated control sections

Table 6 Terminal trafficking performance limits required for confinement mechanism

Trafficking parameter Requirement for section to be considered as confinement
Total passes of ESAL Should reach full term (normally 10,000–11,500 passes)
Pavement shape classification Should be CON or CON-
Final surface deformation <50 mm
Final surface heave <10 mm
Final deformation per 1000 passes Around 1 mm/1000 passes or less

Passes of ESAL (no)

0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 11000 12000
0

20
Surface deformation (mm)

40

60

80 CON CON-
*250mm 50/50 MEM-
100 thick MEM MEM*
section
Av control
120

Fig. 12 Development of surface deformation versus passes of the wheel

The Development of APT Methodology in the Application … 273

is included. There is a slight variation between controls, but at the ESAL scale used
for Fig. 12, they would appear very close together, with the panels having highest
initial surface stiffness giving the best performance. Adding this average control to
Fig. 12 confirms that all geosynthetics are resulting in some improvement in the
deformation behaviour, however the MEM deformation is only slightly less than
the control, confirming the statement given above, namely that the tensioned
membrane effect is relatively small (Giroud 2006).
It is also instructive to examine the trafficking data of the six panels in terms of
rut depth versus number of passes, as shown in Fig. 13. Although the trends are
similar to Fig. 12, this does confirm the significantly better performance of the
section designated as CON, and it is now clear that MEM- involves much larger
surface disruption (i.e. deformation combined with heave) than the 50/50 section.
The final item of discussion is related to the panel designated MEM*, with the
250 mm thick non-standard section. In terms of the types of geosynthetic used in
the six panels in Fig. 11, all are geogrids, and all are made by punching and
stretching a polymer sheet, with the exception of MEM which is a woven geogrid.
Almost all woven geogrids tested in the eight TRL trials reached MEM well before
full term, similar to the performance shown in Figs. 11, 12 and 13. From the other
sections, CON and MEM* are multi-axial geogrids, whereas the others are biaxial
geogrids of various aperture shapes and sizes.
The difference between the panels CON and MEM* is their thickness, with
MEM* being 50 mm thinner, although with the same type of geosynthetic, a
punched and stretched multi-axial geogrid, developed as a stabilisation geogrid
rather than a reinforcement. As can be seen in Table 5, this results in a significantly
reduced RI and initial surface stiffness of almost half, such that the geosynthetic
behaviour goes almost immediately into tensioned membrane. Therefore a trial of

Passes of ESAL (no)

0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 11000 12000
0

20

40

60
Rut depth (mm)

80

100

120

140 CON CON-

160 *250mm 50/50 MEM-

thick MEM MEM*
180 section
Av control
200

Fig. 13 Development of rut depth versus passes of the wheel

274 J. Cook et al.

this arrangement provides no information relevant to permanent surfaced pave-

ments, partly because of the rapid development of tensioned membrane behaviour,
and partly because the geosynthetic used is not intended to behave in this way. This
comparison confirms that the TRL standard section of 300 mm sub-base over a
subgrade with CBR = 2 % (and RI *4.2) is on the lower limit of creating traf-
ficking performance relevant to the design of permanent pavements, in cases where
confinement is developed.

5 Conclusions

An important part of the investment in the full scale trafficking trials described in
this paper was to establish and maintain consistency through the testing programme
which has proved beneficial when the whole body of results is gathered together.
Important aims of the test programmes have been to minimise test variables and
external influences, which is vital if the resulting data is to be meaningful and
applicable to developing design methods. The PTF at TRL provides accurate wheel
load application along a well-defined path (always channelised in the case of these
trials) inside a building, so protected from rain and extreme drying conditions. This
is considered to have considerable advantage over trials conducted outdoors, using
truck loading around oval circuits or long straight tracks which can cover very large
areas, making consistent preparation of the subgrade very difficult. Although pos-
sibly considered to add realism to a trial, such test conditions also add variability
and uncertainly both to the data and to the conclusions which can be derived.
Testing methodologies have been employed and developed across the 14 years
of testing which relate to several key aspects: subgrade soils and their preparation,
sub-base and its placement, quality control of these features as well as the extensive
deformation measurements recorded. Benefits have been realised in the derivation
and quantification of the performance advantages available from using the
geosynthetic products assessed, which in turn has allowed the derivation of product
specific factors for use in the design of geosynthetic stabilised roadways and traf-
ficked areas.
The discussion presented in this paper is based on examining the terminal
trafficking performance from all eight trials in order to examine the mechanism
which develops, in particular does the geosynthetic function by way of confine-
ment, or by way of tensioned membrane. By developing a simple pavement shape
classification scale, it has been possible to identify the terminal pavement perfor-
mance which defines true confinement, such that a significant improvement in
surface deformation is seen with very little deformation at the base of the pavement.
The conditions are that: the trial should go full term; surface deformation is less
than 50 mm; heave is less than 10 mm; the rate of deformation increase should be
around 1 mm per 1000 passes or less at the end of the trial. For a clear confinement
mechanism to be operating, all these criteria should be met.
The Development of APT Methodology in the Application … 275

References

AASHTO. (2009). Standard practice for geosynthetic reinforcement of the aggregate base course
of flexible pavement structures. Washington, DC: American Association of State Highway &
Transportation Officials, R 50-09.
Austroads. (2012). Guide to pavement technology: Part 2: Pavement structural design. Austroads
Publication AGPT02-12, Sydney: Austroads Ltd.
Cancelli, A., & Montanelli, F. (1999). In-ground test for geosynthetic reinforced flexible paved
roads. In Proceedings geosynthetics’99 (pp. 863–878). St Paul, Minnesota.
Chaddock, B. C. J. (1998). Deformation of road foundations with geogrid reinforcement. TRL
Report RR140, Transport Research Laboratory, Wokingham.
Cuelho, E., & Perkins, S. (2009). Field investigation of geosynthetics used for subgrade
stabilisation. Report FHWA/MT-09-003/8193, Western Transportation Institute, Montana
State University, Bozeman, Montana.
Cuelho, E., Perkins, S., & Morris, Z. (2014). Relative operational performance of geosynthetics
used as subgrade stabilisation. Report FHWA/MT-14-002/7712-251, Western Transportation
Institute, Montana State University - Bozeman, Montana.
Giroud, J. P., & Noiray, L. (1981). Geotextile reinforced unpaved road design. Proceedings
American Society of Civil Engineers, 107(GT9), 1233–1254.
Giroud, J. P. (2006). Functions of geosynthetics in road engineering. Keynote Lecture at Seminar
on Road Pavements in Indonesia, Indonesian Chapter of IGS, Jakarta, 6 April 2006.
IAN 73. (2009). Design guidance for road pavement foundations. Interim Advice Note 73/06,
2009.
Jenner, C. J., Watts, G. R. A., & Blackman, D. I. (2002). Trafficking of reinforced, unpaved
subbases over a controlled subgrade. In Proceedings of 7th international conference on
geosynthetics (pp. 931–934). Nice.
Jersey, S. R., Tingle, J. S., Norwood, G. J., Kwon, J., & Wayne, M. (2012). Full-scale evaluation
of geogrid reinforced thin flexible pavements. Transportation Research Board 91st Annual, 22
to 26 January 2012.
Manual of Contract Documents for Highway Works (MCHW). (2016) Specification for highway
works volume 1: Series 800 road pavements—unbound, cement and other hydraulically bound
mixtures. London: The Stationery Office.
Manual of Contract Documents for Highway Works (MCHW). Specification for highway works
volume 1: Series 600 earthworks. London: The Stationery Office.
Norwood, G. J., & Tingle, J. S. (2014). Performance of geogrid-stabilised flexible pavements.
Report ERDC/GSL TR-14-28, Geotechnical and Structures Laboratory, US Army Corps of
Engineers, Engineer Research and Development Center, Vicksburg, Mississippi.
Sorensen, A., & Hayven, M. (1982). The dynatest 8000 falling weight deflectometer test system.
In Proceedings of the international symposium on bearing capacity of roads and airfields
(pp. 464–470). Trondheim, 23–25 June 1982.
Webster, S. L. (1991). Geogrid reinforced base courses for flexible pavements for light aircraft:
Test section construction, behaviour under traffic, laboratory tests and design criteria. Report
DOT/FAA/RD-92/25, Geotechnical Laboratory, Department of the Army, Waterways
Experimental Station, Corps of the Engineers, Mississippi.
Wrigley, N. E., & Valero, S. N. (2013). Roads design and associated full-scale geogrid testing. In
Proceedings of 6th international conference on geosynthetics (pp. 47–57). Middle East 2013,
UAE.
Tire Type Effect on Pavement Responses;
Accelerated Pavement Testing Results

Mojtaba Ziyadi and Imad L. Al-Qadi

Abstract The proposed paper summarizes the accelerated pavement testing

database obtained from various locations, including a database of recent studies.
Test sections built as part of a study investigating the effect of new generation
wide-base tire (NG-WBT 445/50R22.5) and dual-tire assembly (DTA
275/80R22.5) on flexible pavement response were used. These test sections were
established at three different test locations throughout the U.S.—Florida,
Davis-California, and Ohio State. Various instrumentations were installed at several
locations within the pavement system to capture the most critical pavement
responses. Instrumentations installed include strain gauges, pressure cells, ther-
mocouples, multi-depth defloctometers, surface strain gauges, and strain gauge
rosettes. Critical pavement responses, including strain at bottom of AC and stress
on top of subgrade, were analyzed. Moreover, the effect of temperature, differential
tire pressure, and near surface responses were investigated for NG-WBT versus
DTA.

Keywords APT
Instrumentation
New generation wide-base tire
Pavement

response

1 Introduction

The effect of tire type on pavement response and performance was thoroughly
investigated in the past decades. As a result, the tire industry has been trying to
modify the traditional tire design. Dual-tire assemblies (DTA) and wide-base tires
(WBT) are the two major types of tires used in the trucking industry. WBT have
been continuously improving in terms of cost, environmental effects, and safety

M. Ziyadi (&)
I.L. Al-Qadi

University of Illinois at Urbana-Champaign, Champaign, USA
e-mail: [email protected]
I.L. Al-Qadi
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 277

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_18
278 M. Ziyadi and I.L. Al-Qadi

(Al-Qadi and Wang 2009; Al-Qadi and Elseifi 2007; Markstaller et al. 2000). Super
single (425/65R22.5) tires were introduced as the first generation of WBT. They are
wider, but cheaper, than typical dual tires. However, several studies have showed
the adverse effect of these tires on pavement (Greene et al. 2010). Since then, the
tire industry has been upgrading the tire design characteristics to improve its per-
formance. Since the introduction of the new-generation of WBT (NG-WBT)
(445/50R22.5 and 455/55R22.5), industry had been trying to encourage the use of
NG-WBT to increased safety as well as cost and environmental benefits. Several
studies have focused on the effect of WBT on pavement. However, results vary
depending on the type of test, pavement structure, and environmental condition.
A study by environmental protection agency showed significant fuel savings due to
the use of WBT (Routhier 2007).
A recent study showed that super single tires tend to generate more damage to
pavement than NG-WBT (Greene et al. 2010). Some studies conducted at Virginia
Tech indicate that the pavement damage caused by WBT-445 is less than that
resulting from the use of DTA-275 (Al-Qadi and Elseifi 2004; Al-Qadi et al. 2005;
Elseifi et al. 2005).
Green et al. (2009) showed that WBT-425 had the worst rutting performance and
generated the highest transverse strain at the surface. Also, they concluded that
WBT-455 resulted in lowest shear strain at edge of the tire; WBT-445 and DTA
generated similar values (Greene et al. 2009). In line with previous studies, Xue and
Weaver (2011) observed that WBT-425 produced higher response (shear strain)
than other tire types (Xue and Weaver 2011). Several studies also modeled the tire
type effect using finite element (FE) modeling techniques (Hernandez et al. 2015;
Al-Qadi and Wang 2012; Greene et al. 2010; Al-Qadi and Elseifi 2004).
Instrumentation results are mainly used for calibration of and validation purposes in
FE modeling studies.
There has been no conclusive evidence with regard to the effect of NG-WBT on
pavements since the introduction of this type of tires. This is partially due to that
fact that conducting full-scale studies, considering all aspects of tire design and
pavement structure, is expensive and would require full-scale testing. Thus, a
comprehensive approach is needed to compare NG-WBT with DTA.
A national study was conducted to assess the effect of NG-WBT on pavement
using accelerated pavement testing facility and controlled truck testing. Three test
sections were built at three locations: Florida, UC Davis and Ohio. While controlled
truck testing was used for the Ohio sections, full-scale accelerated testing was
performed on the other two sections. This paper presents details of the instru-
mentation and a summary of main results.
The instrumentations installed include strain gauges, pressure cells, thermo-
couples, multi-depth defloctometers, surface strain gauges, and strain gauge
rosettes. Critical pavement responses, including strain at the bottom of AC and
stress on top of the subgrade, were analyzed. Moreover, the effect of temperature,
differential tire pressure, and near surface responses was investigated for the
NG-WBT versus DTA.
Tire Type Effect on Pavement Responses; Accelerated … 279

2 Experimental Program

A unique experimental program was designed to be conducted at three new test

sections to study the effect of NG-WBT on pavements. Contact stresses/forces of
the tires were also measured by the Council for Scientific and Industrial Research in
South Africa for validation purposes. Table 1 shows the experimental test matrix.
Two types of NG-WBT (445/50R22.5) and DTA (conventional 2 × 275/
80R22.5) were used. Three candidate sections were identified at three different
climate regions, including Florida, UC Davis, and Ohio.

2.1 Florida Test Sections

Two sections were built in Florida: test pit and test track. The test pit pavement
consists of two similar 38.1 mm Superpave (SP-12.5) layers with a PG 67-22
asphalt binder. The test track consists of an existing 38.1 mm SP-12.5 layer with a
PG 67-22 asphalt binder, with two new layers of 38.1 mm SP-12.5 layer with PG
76-22 and 25.4 mm layer of 4.75 mm mixture with PG 76-22.
Figure 1 shows the instrumentation plan for the test pit section. Instrumentation
included surface and embedded strain gauges as well as pressure cells. Six sets of
foil gauges, three in each traffic direction, were installed on the surface of both test
sections. Each set has four foil gauges at 76.2 mm offset from the tire edge. Three
H-type embedded strain gauges in both directions and two pressure cells were
installed at the bottom of the new AC layers (76.2- and 63.5-mm below the surface
for the test pit and test track sections, respectively). Additionally, two pressure cells
were placed on top of the subgrade at the test pit section. Accelerated loading was
performed using a heavy vehicle simulator (HVS). The speed during testing was

Table 1 Experimental program conducted at test locations

Tire type Inflation pressure Tire loading (kN) (kips)
(kPa) (psi)
WBTa and DTAb 552 (80) 26.7 35.6 44.5 62.3 80
WBT and DTA 689 (100) (6) (8) (10) (14) (18)
WBT and DTA 758 (110)
WBT and DTA 861 (125)
DTA Only 413/758c
DTA Only 552/758c
a
WBT New generation of wide-base tire (445/50R22.5)
b
DTA Dual-tire assembly (conventional 2 × 275/80R22.5 tire)
c
Differential tire inflation pressure in dual tires
280 M. Ziyadi and I.L. Al-Qadi

PLAN VIEW

24 24 24 24 24 24 24

FGT-10 FGL-10 FGT-11 FGL-11 FGT-12 FGL-12

144
FGT-7 FGL-7 FGT-8 FGL-8 FGT-9 FGL-9 4@3 in

FGT-4 FGL-4 FGT-5 FGL-5 FGT-6 FGL-6

FGT-1 FGL-1 FGT-2 FGL-2 FGT-3 FGL-3 3 in

Edge of
Wheel Path

Wheel
Path
HGT-1 HGL-1 HGT-2 HGL-2 HGT-3 FGL-3 PC 1 PC 2
PC 3 PC 4

Shoulder

PROFILE VIEW
FGT-1 FGL-1 FGT-2 FGL-2 FGT-3 FGL-3
FGT-4 FGL-4 FGT-5 FGL-5 FGT-6 FGL-6
1.5 in, SP-12.5 PG76-22 3
AC FGT-7 FGL-7 FGT-8 FGL-8 FGT-9 FGL-9 1.5 in, SP-12.5 PG76-22
FGT-10 FGL-10 FGT-11 FGL-11 FGT-12 FGL-12
HGT-1 HGL-1 HGT-2 HGL-2 HGT-3 FGL-3
PC 1 PC 2

Limerock
10.5
Base

PC 3 PC 4

Subgrade Florida DOT Proposed

+ Test Pit Section 12
Limerock
Transverse strain foil gauge
Longitudinal strain foil gauge
Transverse strain H-gauge
Subgrade
Longitudinal strain H-gauge

Pressure cell

Fig. 1 Florida test pit section instrumentation plan (unit in figure is inch, 1 in. = 2.54 mm)

8.0 km/h and each loading combination was conducted at 25, 40, and 55 °C. The
load and tire inflation pressure used during the accelerated pavement testing fol-
lowed the experimental program shown in Table 1.

2.2 UC Davis Test Sections

APT at the University of California Pavement Research Center facility in Davis was
conducted on two test sections; thin and thick sections. One of the test sections had
a single 60 mm lift of HMA and the second had two 60 mm lifts, both on top of
Tire Type Effect on Pavement Responses; Accelerated … 281

T T

0 4T 8 12 16 0 4T 8 12 16

HMA Lift 2
HMA Lift 1 HMA Lift 1

Recycled Recycled Failed during

Base Base construction
Failed
DNSTduring Failed during
testing testing

Old Base Old Base

T Thermocouple Tree Pressure Cell T Thermocouple Tree Pressure Cell

MDD Module MDD Module
Longitudinal Strain Gauge Transverse Strain Gauge Longitudinal Strain Gauge Transverse Strain Gauge

(a) Thin section (b) Thick section

Fig. 2 Davis instrumentation plan: a thin section and b thick section

full-depth reclaimed aggregate base with no stabilization. Recycle depth was set at
250 mm on top of a 320 mm aggregate base course.
The layout of the embedded instruments is shown in Fig. 2a, b. Instrumentation
included strain gauges at the bottom of each lift, pressure cells at the bottom AC
layer and on top of subgrade for only one section, thermocouples at different depths,
and multi-depth defloctometer (MDD). MDD essentially consists of a stack of
linear variable differential transformer (LVDT) modules. The LVDT modules have
non-spring loaded core slugs linked together into one long stick fixed at the bottom
of the borehole. The APT was conducted using a HVS following the experimental
program provided in Table 1. The tests were run at 20, 35, and 50 °C. Moreover,
three lateral positions of the centerline of the tire assembly (CLTA) of 0, 177.8 and
304.8 mm were run at a fixed temperature of 50 °C, half-axle load of 44.5 kN, and
two tire pressures of 552 and 861 kPa. A total of 100 load repetitions were applied
for each loading combination, and loading was applied in a channelized mode (i.e.,
no wander) in both directions, unless otherwise specified.

2.3 Ohio Test Sections

Three test sections were built in Ohio for the purpose of the study. The total
thicknesses of the AC layer for the sections are 330 mm for Sections A and B and
381 in for Section C. The thickness of the asphalt-treated base (ATB) is 152 mm for
Sections A and B and 203 mm for Section C. Also, a 102 mm fatigue resistance
layer (FRL) was constructed on top a 152 mm aggregate base layer for all sections.
Instrumentation of the sections included H-type strain gauges at three different
depths (bottom of FRL, bottom of ATB, and bottom of upper lift of the surface
layer). Six longitudinal sensors were placed at the bottom of the FRL; six at the
bottom of the ATB (three longitudinal and three transverse); and four close to the
surface (two longitudinal and two transverse). Instrumentation of these sections also
282 M. Ziyadi and I.L. Al-Qadi

PLAN VIEW
Traffic

CL

Wheel Path

Shoulder

PROFILE VIEW

AC

ATB
FRL

DGAB

Subgrade

SECTIONS A - B
Pressure Cell
Linear Variable Displacement Transducer
Longitudinal Strain Gage
Transverse Strain Gage

Strain Gage Rosette

Fig. 3 Ohio sections A and B instrumentation plan (1 in. = 25.4 mm)

included LVDT, pressure cells, and strain gauge rosettes (SGR), as shown in Fig. 3.
In addition to the pressure cells on top of the subgrade, two additional pressure cells
were installed at the bottom of the FRL. A total of 16 SGR were installed in
Sections A and B, with a total of two holes and eight rosettes in each hole, at four
different depths. One of the two holes was circular and the other rectangular. One of
the two rosettes at each depth was installed in the direction of the traffic, while the
other was perpendicular to traffic. The SGR was placed in the middle of each lift to
calculate shear strain at mid-depth of each layer.
Truck loading was used to apply each load case according to the experimental
program provided in Table 1. However, the full experimental program was not
followed, and only some load cases were run at specific axle configuration.
Different axle configurations (single-axle dual, tandem-axle dual, single-axle
wide-base, tandem-axle wide-base) were added to the test. Trucks were loaded and
axles were weighed before testing to ensure conformity with the test plan. Detailed
truck load distribution on each tire was provided as part of the test. A straight path
was set over the sensor lines to drive the trucks in a straight line. However, sand
was used at the entrance and exit of the truck path to measure wandering. The
trucks were set to drive at 8, 40.2, 72.4, and 88.5 km/h.

3 Data Analysis

The effect of different factors influencing pavement responses was studied for both
tire types (DTA vs. WBT). The following paragraphs summarize the results.
Tire Type Effect on Pavement Responses; Accelerated … 283

3.1 Load Effect

According to the data obtained from the Florida test pit, increasing the load from
26.7 to 80 kN increases the longitudinal strain at the bottom of AC, 50 % under
DTA and 40 % under WBT. The effect worsens as the thickness decreases. It was
also observed that under higher temperatures and increased loading, the strain does
not necessarily increase. This is a complex phenomenon which involves interaction
of temperature-dependent pavement material and load-dependent stress distribution.
According to a study by Hernandez et al., 3-D tire contact stress distribution varies
significantly at higher loads. Maximum stress under loading moves from center to
the edge of the tire; as expected this is more pronounced for wide-base tire
(Hernandez et al. 2013). This can significantly affect the sensor readings, especially
in thin pavements.
A comparison between the longitudinal strain at the bottom of AC layer at UC
Davis thin section and Florida test track, under different loading, 758 kPa tire
inflation pressure, and 20 °C temperature, is shown in Fig. 4a, b.
Generally, WBT resulted in higher strain than DTA, as shown in Fig. 4a. At
higher loads, however, the difference between the two tire types decreases. In one
case at the Florida test track section (Fig. 4b), DTA even resulted in higher strain
than WBT. Figure 5 shows this clearly in terms of percent higher strain resulting
from WBT compared with DTA at the bottom AC layer.
It was noticed that the increase in the load resulted in diminishing the difference
between WBT and DTA; in some cases, DTA resulted in higher responses than
WBT. It must be noted that all measurements are compared at 25 °C.

450 35
400 DTA WBT DTA WBT
30
Strain (MicroStrain)

350
Strain (Microstrain)

300 25
250 20
200 15
150
10
100
50 5
0 0
6kips 8kips 10kips 14kips 18kips 6kips 8kips 10kips 14kips 18kips
Half Axle Load Half Axle Load

(a) (b)

Fig. 4 Strain at bottom of AC for a UC Davis thin section and b Florida test track
section (1 kip = 4.45 kN)
284 M. Ziyadi and I.L. Al-Qadi

Fig. 5 Load effect on 120%

FL_Test Pit

% Strain Difference (WBT vs.

longitudinal strain at the 100% FL_Test Track
bottom of AC 80% Davis_Thin
(1 kip = 4.45 kN) Davis_Thick
60%

DTA)
40%
20%
0%
6kips 8kips 10kips 14kips 18kips
-20%
-40%
Half Axle Load

Fig. 6 Temperature effect on 450 DTA

longitudinal strain at bottom 400 WBT
of AC for Florida test pit Strain (Microstrain) 350
300
section under 44.5 kN loading 250
and 110 psi tire inflation 200
pressure (1 kip = 4.45 kN) 150
100
50
0
25 40 55
Temperature (C)

3.2 Temperature Effect

Figure 6 shows the temperature effect on longitudinal strain at the bottom of AC for
the Florida test pit section under 44.5 kN loading, and 758 kPa tire inflation
pressure. In general, a 5–10 % increase in strain was observed for one unit change
in temperature for all sections.

3.3 Tire Inflation Pressure Effect

Based on the data analysis, an increase in the tire inflation pressure results in an
increase in the strain at the bottom of AC. However, this effect is considered very
minimal compared with the effect of other parameters such as load and temperature.
At the Florida and Davis sections, an increase in tire inflation pressure from 552 to
861 kPa resulted in <10 % increase in strain for DTA and 20 % for WBT. The
effect of tire inflation pressure was more pronounced for WBT.

3.4 Near-Surface Responses

Analyzing surface strain gauge data for the Florida test pit track section showed that
an increase in tire pressure from 552 to 861 kPa resulted in 30 and 40 % increase in
Tire Type Effect on Pavement Responses; Accelerated … 285

Fig. 7 Tire type effect on DTA_3in_Offset DTA_12in_Offset

WBT_3in_Offset WBT_12in_Offset
surface strain under 10 kips
160
loading and 25 °C

Surface Strain (Microstrain)

140
(1 psi = 6.89 kPa)
120
100
80
60
40
20
0
80 100 110 125
Tire Inflation Pressure (psi)

surface transverse strain for DTA and WBT, respectively, with 76.2 mm offset from
the wheel path. The effect was higher on transverse strain than longitudinal strain.
Also, with the increase in offset from 76.2 to 304.8 mm, the effect was reduced to <10
and −5 % for DTA and WBT, respectively. This means that strain under WBT is
more sensitive to change under inflation pressure compared with DTA. Figure 7
shows the surface strain versus tire inflation pressure under 44.5 kN loading at 25 °C.
Figure 8 shows the effect of sensor offset on the surface strain. As can be noted,
the transverse strain near loading is smaller than the longitudinal strain. Further
apart from loading, the transverse strain becomes greater.
Figures 9 and 10 show the effect of axle load, temperature, and tire inflation
pressure on strain and stress at the bottom of AC layer, respectively. The data are
based on UC Davis’ thin and thick sections.
Therefore, WBT resulted in higher responses than DTA. The difference in strain
responses decreased with the increase in loading and temperature. As previously
mentioned, this could partly be attributed to the shear development in higher
temperatures, which in turn affects the longitudinal strain. The temperature effect on
WBT was more pronounced.

Long_80psi Long_125psi Long_80psi Long_125psi

Trans_80psi Trans_125psi Trans_80psi Trans_125psi
180 180
Surface Strain (Microstrain)
Surface Strain (Microstrain)

160 160
140 140
120 120
100 100
80 80
60 60
40 40
20 20
0 0
3 6 9 12 3 6 9 12
Offset (inches) Offset (inches)

(a) DTA (b) WBT

Fig. 8 The effect of sensor offset on the surface strain under a DTA and b WBT
(1 in. = 25.4 mm)
286 M. Ziyadi and I.L. Al-Qadi

1500
DTA_80psi DTA_100psi DTA_110psi
DTA_125psi WBT_80psi WBT_100psi
1300

1100
Strain (Microstrain)

900

700

500

300

100
6kips 8kips 10kips 14kips 18kips 6kips 8kips 10kips 14kips 18kips 6kips 8kips 10kips 14kips 18kips
20C 35C 50C

Fig. 9 Effect of axle load, temperature, and tire inflation pressure on strain at bottom of AC at UC
Davis thin section (1 kip = 4.45 kN)

35
DTA_80psi DTA_100psi DTA_110psi DTA_125psi
WBT_80psi WBT_100psi WBT_110psi WBT_125psi
30

25

20
Stress (psi)

15

10

0
6kips 8kips 10kips 14kips 18kips 6kips 8kips 10kips 14kips 18kips 6kips 8kips 10kips 14kips 18kips
20C 35C 50C

Fig. 10 Effect of axle load, temperature, and tire inflation pressure on stress on top of subgrade at
UC Davis thick section (1 kip = 4.45 kN)

4 Conclusion

A comprehensive experimental program was designed and conducted at three test

locations to study the effect of NG-WBT (445/50R22.5) on pavements. The sec-
tions were built in different climate regions: Florida, Davis, and Ohio. A unique
experimental program was followed to conduct the accelerated pavement testing
and controlled truck test. Extensive instrumentation was installed in each section
including longitudinal and transverse strain gauges, surface strain gauges, SGR,
pressure cells, MDD, and thermocouples.
Tire Type Effect on Pavement Responses; Accelerated … 287

In summary, the following results were obtained:

1. NG-WBT resulted in approximately 30 % higher pavement responses in terms
of strain at the bottom of AC. This can result in higher pavement bottom-up
fatigue cracking.
2. Increasing load from 6 to 18 kips increased the longitudinal strain at the bottom
of AC, 50 % under DTA and 40 % under WBT.
3. At higher temperatures as load increased, longitudinal strain at bottom of AC
decreased. This is partially due to the shear development process within the
AC. Moreover, at high loading values, previous studies showed that contact
stress pattern changes dramatically which in turn will adversely affect the
sensor readings.
4. In general, 5–10 % increase in strain was observed for one unit change in
temperature for all sections.
5. The effect of tire inflation pressure decreased with depth.
6. An increase in tire inflation pressure from 80 to 120 psi resulted in <10 %
increase in strain at the bottom of AC for DTA and 20 % for WBT.
7. The effect of tire inflation pressure was more pronounced for WBT.
8. An increase in tire pressure from 80 to 120 psi resulted in 30–40 % increase in
surface transverse strain for DTA and WBT, respectively. The effect was higher
on transverse strain than longitudinal strain.
9. The effect of tire inflation pressure on longitudinal strain diminished further
apart from loading. However, the effect on transverse strain increased up to 6 in
apart from loading and then decreased afterwards.
10. The effect of tire inflation pressure on stress and strain was more distinct at
higher loads, especially for NG-WBT.

References

Al-Qadi, I. L., Elseifi, M. A., & Yoo P. J. (2004) Pavement damage due to different tires and
vehicle configurations. Final report submitted to Michelin Americas Research and
Development Corporation, 515 MichelinRoad, SC.
Al-Qadi, I. L., & Elseifi, M. A. (2007). New generation of wide-base tire: Impact on trucking
operations, environment, and pavements. Transportation Research Record: Journal of the
Transportation Research Board, 2008, 100–109.
Al-Qadi, I. L., Elseifi, M. A., & Yoo, P. J. (2005). Characterization of pavement damage due to
different tire configurations. Journal of the Association of Asphalt Paving Technologists, 84,
921–962.
Al-Qadi, I. L., & Wang, H. (2009). Full-depth pavement responses under various tire
configurations: Accelerated pavement testing and finite element modeling. Journal of the
Association of Asphalt Paving Technologists, 78.
Al-Qadi, I. L., & Wang, H. (2012). Impact of wide-base tires on pavements—Results from
instrumentation measurements and modeling analysis. Transportation Research Record:
Journal of the Transportation Research Board, 2304, 169–176.
288 M. Ziyadi and I.L. Al-Qadi

Elseifi, M. A., Al-Qadi, I. L., Yoo, P. J., & Janajreh, I. (2005). Quantification of pavement damage
caused by dual and wide-base tires. Transportation Research Record: Journal of the
Transportation Research Board, 1940, 125–135.
Greene, J., Toros, U., Kim, S., Byron, T., & Choubane, B. (2009). Impact of wide-base single tires
on pavement damage. Research report, FL/DOT/SMO/09–528, Florida Department of
Transportation.
Greene, J., Toros, U., Kim, S., Byron, T., & Choubane, B. (2010). Impact of wide-base single tires
on pavement damage. Transportation Research Record: Journal of the Transportation
Research Board, 2155, 82–90.
Hernandez, J., Al-Qadi, I., & De Beer, M. (2013). Impact of tire loading and tire pressure on
measured 3D contact stresses. Airfield and Highway Pavement. doi:10.1061/9780784413005.
044.
Hernandez, J. A., Gamez, A. M., & Al-Qadi, I. L. (2015). Effect of wide-base tires on nationwide
flexible pavement systems—Numerical modeling. Transportation Research Record: Journal of
the Transportation Research Board (submitted for publication).
Markstaller, M., Pearson, A., Janajreh, I. (2000). On vehicle testing of michelin new wide base
tire. In Proceedings of the 2000 SAE international conference, no. 2000–01-3432. Detroit, MI:
Society of Automotive Engineers.
Routhier, B. (2007). Wide-base tires fleet experiences. Presented at the international workshop on
the use of wide-base tires. McLean, VA: FHWA Turner Fairbank, Highway Research Center.
Xue, W. J., & Weaver, E. (2011). Pavement shear strain response to dual and wide-base tires.
Transportation Research Record: Journal of the Transportation Research Board, 2225, 155–
164.
 Part IV
Accelerated Pavement Testing Focused on
Mechanistic—Empirical Pavement Design
Procedures and Models
Analysis of Dynamic Response of Asphalt
Pavement in Heavy Vehicle Simulator
Tests

Aimin Sha, Jie Wang, Liqun Hu and Xiaolong Zou

Abstract In order to analyze the dynamic response of asphalt pavement, heavy

vehicle simulator (HVS) was used to test with variables of axle load, speed, and
temperature on five full-scale test sections with different structures. Pre-embedded
strain sensors and pressure cells were used to measure the strain response at the
bottom of the surface layer and the vertical stress at the top of the subgrade under
wheel load. The results of tests show that the tensile strain at the bottom of surface
layer of semi-rigid base asphalt pavement and rigid base asphalt pavement is small.
The vertical stress at the top of the subgrade has a good positive exponential
relationship with axle load. The maximum tensile strain at the bottom of surface
layer gradually decreases with speed and increases with temperature. The tensile
strain at the bottom of surface layer of flexible base asphalt pavement is affected by
the speed most significantly. The increase rate of strain increases with temperature,
which demonstrates the temperature has significant effects on the mechanical
properties of the asphalt pavement. Based on the results of HVS tests, a prediction
model of tensile strain at the bottom of the surface layer was established. Therefore,
this paper could contribute to deepening the understanding of the dynamic response
characteristics of asphalt pavement and improving the design of pavement structure.

Keywords Asphalt pavement Accelerated pavement testing
Heavy vehicle


simulator Dynamic response, strain prediction

A. Sha (&)
J. Wang
L. Hu
X. Zou

Key Laboratory for Special Region Highway Engineering of Ministry of Education,
Chang’an University, Xi’an 710064, Shaanxi, China
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 291

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_19
292 A. Sha et al.

1 Induction

The current design systems of asphalt pavement in all countries set vehicle load as
static load, or simplify it approximately as static load (Yoder and Witczak 1975;
Claessen et al. 1977; Huang 1993; Alagappan et al. 2011). The pavement design
methods using static load have been well developed and commonly applied (Bari
and Witczak 2015). With the rapid development of highway industry in China,
speed and load of vehicles gradually increases. Although increasing number of
axes, tire size, and number of tires can make static pressure of a heavy vehicle on
pavement close or equal to that of a light vehicle, the impact load and inertial load
of vehicle at high speed caused by vibration will be greatly increased with the
increasing of vehicle speed and axle load. The increased impact load and inertial
load could enlarge the difference between equivalent static load and the actual load
of moving vehicle on the pavement. It is not suitable to describe the dynamic
characteristics of the pavement structure using static load methods. Therefore,
dynamic response characteristics and changing rule of the pavement internal
structure have become a hot research topic nowadays.
The theory of elastic layered system was used to correct the static analysis results
and took the influence of vehicle speed and load frequency into consideration
(Monismith 1992; Chatti et al. 1995), but this theory used an elastomer constitutive
relation model without considering the visco-plastic of asphalt mixture. Dong
Zhong-hong et al. assumed pavement materials as viscoelastic materials and used
the linear theory and the dynamics of continuum system to simplify the axle load of
a moving vehicle based on Fourier transform method (Zhonghong and Fengying
2014). They developed a mathematical model of dynamic response of asphalt
pavement, but this mathematical model ignores the effects of axle load and tem-
perature on dynamic response of the pavement. The above-mentioned studies are
applicable under certain assumptions; however, the dynamic response of asphalt
pavement under the actual traffic load is not only affected by vehicle axle load,
accumulative loading time, speed, and tire parameters, but also affected by envi-
ronmental factors such as temperature and humidity. Besides, it is closely related
with pavement structure and material property. Although many dynamic models of
pavement structure have been established considering some of the above-mentioned
factors (Deng and Sun 2000; Goktepe et al. 2006; Jing-song et al. 2007; Perez et al.
2007; Beainy et al. 2013; Chen et al. 2015), they still cannot effectively describe the
dynamic characteristics of actual pavement structure.
In order to understand thoroughly the dynamic response characteristics of
asphalt pavement under the actual traffic load, five full-scale test sections with
different structures were built. A large accelerated pavement testing (APT) facility
with Heavy Vehicle Simulator (HVS) was used to analyze the influence of vehicle
Analysis of Dynamic Response of Asphalt Pavement in Heavy … 293

axle load, speed, and temperature upon the dynamic response of asphalt pavement.
A prediction model of the tensile strain at the bottom of asphalt layer was estab-
lished based on speed, axle load, and temperature in the middle depth of asphalt
layer. Therefore, this paper could contribute to deepening the understanding of the
dynamic response characteristics of asphalt pavement and improving the design of
pavement structure.

2 Full-Scale Accelerated Pavement Testing

2.1 The Characteristics of APT Facility

A HVS was used as loading facility in the test. In this APT facility, a test load can
be simulated by the half-axle and double-tire load with the load range of 60–200 kN
(applied by the half axle), which can be applied bidirectionally or unidirectionally.
The maximum test speed of HVS is 12 km/h. The maximum number of repetition
of HVS traveling per day is 28,000 bidirectionally or 14,000 unidirectionally. The
effective loading length of HVS is 6 m. In addition, the wheel track distribution of
the actual traffic can be simulated in this APT facility by automatically controlling
the transverse displacement of HVS.

2.2 Full-Scale Test Sections

Five full-scale test sections with different types of base course were built in the
cement concrete test slot in the lab, which were free from impacts by rainfall,
sunshine, and other weather conditions. Considering the tensile strain at the bottom
of thicker asphalt layer is relatively small, thin asphalt layer was adopted in each
test section, with the length of 10 m and the width of 3.5 m. The pavement
structures of the full-scale test sections are shown in Table 1.

Table 1 Pavement structures of the full-scale test sections

Pavement Section A Section B Section C Section D Section E
structures (structure 1) (structure 2) (structure 3) (structure 4) (structure 5)
Surface 5 cm 5 cm 5 cm AC-16 5 cm AC-16 5 cm AC-16
layer AC-16 AC-16
Base 20 cm 20 cm 20 cm skeleton 20 cm suspension 20 cm asphalt
course cement graded cement-stabilized cement-stabilized stabilized
concrete gravel gravel gravel gravel
(ATB-25)
Subgrade 95 cm silt 95 cm silt 95 cm silt 95 cm silt 95 cm silt
294 A. Sha et al.

2.3 Materials

The five test sections were paved with the same surface layer and different base
courses. The surface layer was paved with dense-graded (AC-16) mixture, in
which limestone aggregate and a binder asphalt with penetration grade 90 (PG
64-28) were used. A commercial cement concrete C25 was used in base course of
section A. A typical dense-graded gravel was used in base course of section B. The
gradations of AC-16 and gravel used in base courses of section C, D, and E were
shown in Table 2.

2.4 Sensor Embedment

Strain sensors and pressure cells were embedded in the central point of test sections,
which was located in the gap area between the two wheel paths. The strain sensors
embedded at the bottom of the asphalt layer and the pressure cells embedded at the
top of the subgrade were used to measure strain and vertical compress stress,
respectively. Due to the rigid base of section A, the compress strain at the top of the
subgrade could be small, there is no pressure cell embedded in section A. The strain
sensors were embedded along the traveling direction of HVS in the test sections to
monitor the longitudinal strain at the bottom of the asphalt layer.

2.5 Test Conditions

In order to analyze the influence of vehicle axle load, speed, and pavement surface
temperature upon the dynamic response of the asphalt pavement, HVS tests were
conducted with variables of axle load, speed, and temperature on five full-scale test
sections. Four axle load levels of 100, 120, 140 and 160 kN were applied in tests.
Four speed levels of 3, 6, 9, and 12 km/h were included. Pavement surface tem-
perature was used as a control index. Each test section was tested at four temper-
ature levels. The pavement surface temperatures in HVS tests are shown in Table 3.
HVS tests were conducted with variables of axle load, speed, and temperature at
four levels, respectively. Therefore, each test section was tested under 64 different
conditions totally. The test was repeated three times under each condition. The
dynamic response data of five test sections under different conditions were
collected.
Table 2 Gradations of asphalt mixture and materials in base courses
Gradation Passing percentage (%)
31.5 26.5 19.0 16.0 13.2 9.5 4.75 2.36 1.18 0.6 0.3 0.15 0.075
AC-16 – – 100 96.5 86.5 71.1 43.7 33.0 22.7 14.9 11.4 7.5 5.5
Skeleton cement-stabilized gravel 100 – 77 – – 48 27 22 – 11.5 – – 1.5
Suspension cement-stabilized gravel 100 86 63 53 41 26 11 7.0 6.0 4.0 3.0 1.0 0.0
Asphalt stabilized gravel ATB-25 100 92 70 58 52 42 30 23 17 13 9.0 6.0 4.0
Analysis of Dynamic Response of Asphalt Pavement in Heavy …
295
296 A. Sha et al.

Table 3 Pavement surface Test sections Pavement surface temperature/°C

temperatures in HVS tests
Section A 27.7 31.5 32.8 36.0
Section B 28.1 31.7 34.9 35.7
Section C 24.9 27.3 29.6 31.2
Section D 15.4 17.9 19.1 21.4
Section E 12.6 14.9 16.1 17.8

3 Results and Discussion

3.1 Influence of Axle Load on Dynamic Response

of the Asphalt Pavement

Figure 1 shows the time-course curve of longitudinal strain at the bottom of the
asphalt layer in the test section B when axle load is 100 kN and pavement surface
temperature is 28.1 °C.
Figure 1 shows that, with the pass of moving wheels, the strain at the bottom of
the asphalt layer alternately presents compressive strain and tensile strain. When the
wheel approaches and leaves the test point, the strain presents compressive strain.
The compressive strain when the wheel leaves is less than the compressive strain
when the wheel approaches. The tensile strain is generated when the wheel reaches
the test point. The maximum tensile strain is about 391 % of the maximum com-
pressive strain. After the wheel leaves, the strain can usually restore to the original
state. Because the strain at the bottom of the asphalt layer presents both tensile
strain and compressive strain, a strain ratio (the ratio of maximum compressive
strain and the maximum tensile strain) or a strain amplitude (the sum of maximum
compressive strain and the maximum tensile strain) could be taken as a control
index in the analysis of fatigue life of the asphalt pavement.
Figure 2 shows the time-course curve of vertical compressive stress at the top of
the subgrade in section C. The curve is asymmetric which indicate that the time of
stress concentration is slightly less than the time of stress relaxation.

Fig. 1 The time-course curve of strain at the bottom of surface layer

Analysis of Dynamic Response of Asphalt Pavement in Heavy … 297

Fig. 2 The time-course curve of vertical compressive stress at the top of the subgrade

According to Figs. 1 and 2, the residual strain at the bottom of the asphalt layer
and the residual stress at the top of the subgrade are zero when the wheel leaves.
The results indicates that the viscosity of the asphalt pavement has an insignificant
influence upon the strain at the bottom of the asphalt layer and the compress stress
at the top of the subgrade at a moderate temperature.
Figures 3, 4, 5, 6 and 7 show the curves of strain at the bottom of the asphalt
layer and vertical compress stress at the top of the subgrade changing with axle load
in the five test sections, respectively, at the speed of 12 km/h.
As shown in Figs. 3, 4, 5, 6 and 7:
(1) With the increasing of axle load, the maximum compressive strain, the
maximum tensile strain, and strain amplitude in five test sections increase
constantly, which could indicate the influence of axle load on the strain
response at the bottom of the asphalt layer.

Fig. 3 Dynamic response of

structure 1 changing with axle
load at 27.7 °C
298 A. Sha et al.

Fig. 4 Dynamic response of

structure 2 changing with axle
load at 28.1 °C

Fig. 5 Dynamic response of

structure 3 changing with axle
load at 24.9 °C

Fig. 6 Dynamic response of

structure 4 changing with axle
load at 15.4 °C
Analysis of Dynamic Response of Asphalt Pavement in Heavy … 299

Fig. 7 Dynamic response of

structure 5 changing with axle
load for at 12.6 °C

(2) Under the condition of each test temperature, the strain response at the bottom
of the asphalt layer and the axle load is linearly correlated, which indicates that
dynamic characteristics and fatigue life of flexible pavement can be studied
according to linear elastic theory.
(3) When the axle load of the test pavement is 160 kN, the maximum tensile strain
at the bottom of the asphalt layer of structure 1 is 27.13 με and the value of
structure 3 is 25.56 με, which is relatively small. The reason is that structure 1
is rigid base and structure 3 is semi-rigid base. The greater overall rigidity of
the two structures leads to smaller strain at the bottom of the asphalt layer,
which indicates that while designing rigid base pavement and semi-rigid
pavement, the tensile strain at the bottom of the asphalt layer is not suitable to
be used as a main control index.
(4) The compressive stress at the top of the subgrade in five test sections increase
rapidly with the increasing of axle load, indicating that the dynamic response
is strongly affected by axle load. The compressive stress of the structure 2 and
5 is significantly larger than that of the structure 3 and 4, indicating that the
compressive stress at the top of the subgrade of flexible base asphalt pavement
is larger than that of the semi-rigid base asphalt pavement. The compressive
stress (σ) at the top of the subgrade and the axle load (L) have good expo-
nential relationship by regression analysis, as shown in Table 4.

Table 4 Exponential relationship model between compressive stress at the top of the subgrade
and axle load
Test pavement structure Relationship formula Correlation coefficient (R2)
Structure 2 σ = 52.006e0.0051L
0.9919
Structure 3 σ = 36.716e0.0061L 0.9977
Structure 4 σ = 47.411e0.0041L 0.9976
Structure 5 σ = 54.925e0.0041L 0.9984
300 A. Sha et al.

3.2 Influence of Speed on Dynamic Response of the Asphalt

Pavement

Figures 8, 9, 10, 11 and 12 show the curves of dynamic response of different

structures changing with speed with axle load as 100 kN.
As shown in Fig. 8, 9, 10, 11 and Fig. 12:

Fig. 8 Dynamic response of

structure 1 changing with
speed at 27.7 °C

Fig. 9 Dynamic response of

structure 2 changing with
speed at 28.1 °C

Fig. 10 Dynamic response

of structure 3 changing with
speed at 24.9 °C
Analysis of Dynamic Response of Asphalt Pavement in Heavy … 301

Fig. 11 Dynamic response

of structure 4 changing with
speed at 15.4 °C

Fig. 12 Dynamic response

of structure 5 changing with
speed at 12.6 °C

(1) With the increasing of speed, the maximum tensile strain, the maximum
compressive strain, and strain amplitude of the five structures decrease grad-
ually. The decreasing rate of maximum tensile strain lowers gradually mainly
due to the damping effect and hysteresis effect of asphalt pavement. For
structure 1, the maximum compressive strain at the bottom of the asphalt layer
is always larger than the maximum tensile strain, while for each of the other
four structures, the maximum compressive strain is obviously smaller than the
maximum tensile strain, indicating that the compressive strain at the bottom of
the asphalt layer on the rigid base is relatively large.
(2) The compressive stress of the five structures decreases gradually with the
speed increasing from 3 to 12 km/h. The compressive stress decreases by 7.4,
9.6, 7.4, and 7.3 % for structure 2, 3, 4, and 5, respectively. The decreasing
rate gradually eases, which indicates that the influence of speed upon vertical
compressive stress at the top of the subgrade is insignificant.
(3) Compared with the maximum compressive strain, the maximum tensile strain
is significantly affected by speed. Compared with the maximum tensile strain
at speed of 12 km/h, the maximum tensile strain of the five structures at speed
of 3 km/h decreases by 29.1, 25.6, 24.2, 29.5, and 25.3 %, respectively.
302 A. Sha et al.

Table 5 Exponential relationship model between speed and tensile strain at the bottom of the
asphalt surface layer
Test pavement structure Relationship formula Correlation coefficient (R2)
−0.039V
Structure 1 ε = 35.217e 0.9993
Structure 2 ε = 44.188e−0.033V 0.9964
Structure 3 ε = 34.592e−0.029V 0.9596
Structure 4 ε = 27.947e−0.029V 0.9337
Structure 5 ε = 26.861e−0.032V 0.9827

(4) Through regression analysis, the maximum tensile strain (ε) at the bottom of
the asphalt layer and speed (V) presents a good exponential relationship, as
shown in Table 5.

3.3 Influence of Temperature on Dynamic Response

of the Asphalt Pavement

Figures 13, 14, 15, 16 and 17 show the curves of strain at the bottom of the asphalt
layer changing with temperature for five structures, respectively, with the axle load
of 100 kN at the speed of 12 km/h.
As shown in Figs. 13, 14, 15, 16 and 17:
1. The maximum tensile strain, the maximum compressive strain, and strain
amplitude of the five structures increase with temperature. Under the condition
of different temperature levels, the maximum compressive strain is always
greater than the maximum tensile strain for structure 1.
2. The maximum tensile strain at the bottom of the asphalt layer of structure 1 at
36 °C increases by 32.7 % compared with that at 27.6 °C. Similarly, the rates of
structure 2, 3, 4, and 5 are 24.8, 19.4, 15.0, and 16.8 %, respectively. The

Fig. 13 Strain at the bottom

of the asphalt layer of
structure 1 changing with
temperature
Analysis of Dynamic Response of Asphalt Pavement in Heavy … 303

Fig. 14 Strain at the bottom

of the asphalt layer of
structure 2 changing with
temperature

Fig. 15 Strain at the bottom

of the asphalt layer of
structure 3 changing with
temperature

Fig. 16 Strain at the bottom

of the asphalt layer of
structure 4 changing with
temperature

results of tests indicate the increase rate of strain increases with temperature,
which demonstrates temperature has significant effects on the mechanical
properties of the asphalt pavement.
304 A. Sha et al.

Fig. 17 Strain at the bottom

of the asphalt layer of
structure 5 changing with
temperature

Table 6 Exponential relationship model between tensile strain at the bottom of the asphalt
surface layer and temperature
Test pavement structure Relationship formula Correlation coefficient (R2)
Structure 1 ε = 8.5937e0.0345T
0.9743
Structure 2 ε = 14.806e0.0275T 0.9932
Structure 3 ε = 11.924e0.0282T 0.9998
Structure 4 ε = 14.023e0.02388T 0.9808
Structure 5 ε = 12.705e0.0299T 0.9998

3. Through the nonlinear regression analysis, it’s found that the maximum tensile
strain (ε) at the bottom of the asphalt layer and pavement surface temperature
(T) satisfy a good exponential relationship model, as shown in Table 6.

4 Prediction of Tensile Strain at the Bottom of the Asphalt

Layer

In order to reduce or eliminate repeated costly tests in the future design, a prediction
model of the tensile strain at the bottom of different base asphalt pavement was
established based on temperature in the middle depth of asphalt layer, speed, and
axle load through nonlinear regression analysis. This model could contribute to
deepening the understanding of the dynamic response characteristics of asphalt
pavement with different base courses and provide reference for the future design of
different pavement structures.
Analysis of Dynamic Response of Asphalt Pavement in Heavy … 305

4.1 Establishment of Strain Prediction Model

According to the above-mentioned analysis of test results, the strain at the bottom of
the asphalt layer has positive exponential relationship with temperature and axle
load and has negative exponential relationship with speed. Therefore, the following
prediction equation can be established:

e ¼ a  V b  cT  d N ð1Þ

where ε is tensile strain at the bottom of the asphalt layer (με); V is the vehicle
speed (km/h); T is the temperature in the middle depth of the asphalt layer; N is
vehicle axle load; a, b, c, and d are regression coefficients of the equation.
The literature (Willis et al. 2009) shows that the temperature in the middle depth
of the asphalt layer has closest correlation with the tensile strain at the bottom of the
asphalt layer. According to research findings of SHRP, pavement surface temper-
ature and pavement internal temperature are calculated with the Eq. (2) (Huber
1994; Kennedy et al. 1994; Robertson 1997):

Td ¼ ðTsurf þ 17:78Þð1  2:48  103 d þ 1:085  105 d 2  2:441  108 d 3 Þ

 17:78
ð2Þ

where Tsurf is the temperature on pavement surface; Td is the temperature at the

depth of d from the pavement surface. In this paper, the thickness of the asphalt
layer of the test sections is 5 cm. Therefore, according to Eq. (2), the temperature
(T2.5) in the middle depth of 2.5 cm can be calculated, which approximates Tsurf.

4.2 Determination of the Coefficients of the Prediction

Model

Based on the results of HVS tests, the strain prediction model for each test section
was established by a regression analysis software, as follows:
8
>
> 16:414  V0:215  1:028T  1:002N ; R2 ¼ 0:963 structure 1
>
>
< 20:769  V0:157  1:026T  1:001N ; R2 ¼ 0:921 structure 2
e ¼ 18:711  V0:190  1:027T  1:001N ; R2 ¼ 0:979 structure 3 ð3Þ
>
>
>
> 19:184  V0:195  1:028T  1:001N ; R2 ¼ 0:988 structure 4
:
17:266  V0:169  1:030T  1:001N ; R2 ¼ 0:966 structure 5

The correlation coefficients are larger than 0.92, which indicates that the equa-
tion are highly reliable. Based on this model, the strain response at the bottom of the
asphalt layer can be predicted effectively. Through brief comparison of the
306 A. Sha et al.

prediction models for tensile strain of five structures, following conclusions can be
reached:
(1) b is the speed coefficient which indicates the change of tensile strain at the
bottom of the asphalt layer along with vehicle speed. The order of coefficient b
is: structure 2 > structure 5 > structure 3 > structure 4 > structure 1, which
indicates that structure 2 is most significantly affected by speed.
(2) c is the temperature coefficient which indicates the changes of tensile strain at
the bottom of the asphalt layer along with temperature. When the temperature
increases with 5 °C interval from 20 to 40 °C, the increasing rate coefficient of
tensile strain for five pavement structures are Δ1(1.148,1.148,1.148,1.148),
Δ2(1.139,1.137,1.137,1.137), Δ3(1.142,1.143,1.143,1.142), Δ4(1.148,1.148,
1.148,1.148), Δ5(1.159,1.159,1.159,1.158), respectively. Structure 2 is affec-
ted least by temperature. The increasing rate coefficient of tensile strain of
structure 5 is most significant due to the material of base course is
asphalt-stabilized gravel, which is a temperature-sensitive material.
(3) d is the axle load coefficient. The axle load coefficient of five pavement
structures is extremely close to 1, indicating that the effect of axle load
magnitude is considerably less than the effects of velocity and temperature on
the tensile strain at the bottom of the asphalt layer.

5 Conclusions

This study applied the APT facility with HVS to analyze the influence of vehicle
axle load, speed, and temperature on the dynamic response of five full-scale test
sections with different structures. Four axle load levels, four speed levels, and four
temperature levels were included in this study. Therefore, each test section was
tested under 64 different conditions totally. The dynamic response data of five test
sections under different conditions were collected. The prediction model of the
tensile strain at the bottom of asphalt layer was developed based on speed, axle
load, and temperature in the middle depth of asphalt layer. The following con-
clusions can be drawn:
(1) With the pass of moving wheels, the strain at the bottom of the asphalt layer
alternately presents compressive strain and tensile strain. The effect of tensile
strain and compressive strain should be taken into account in the analysis of
fatigue life of the asphalt pavement. At the axle load level of 160 kN, the
maximum tensile strain at the bottom of the rigid base asphalt pavement and
the semi-rigid base asphalt pavement is much smaller than the suggested limit.
The tensile strain at the bottom of the asphalt layer is not suitable to be used as
a main control index in the design of rigid base asphalt pavement and
semi-rigid base asphalt pavement. The vertical compressive stress at the top of
subgrade is influenced significantly by the axle load.
Analysis of Dynamic Response of Asphalt Pavement in Heavy … 307

(2) The maximum tensile strain, the maximum compressive strain, and the strain
amplitude of the five structures decrease gradually with the increasing of speed
from 3 to 12 km/h. The decreasing rate of maximum tensile strain lowers
gradually mainly due to the damping effect and hysteresis effect of asphalt
pavement. Besides, influence of speed upon vertical compressive stress at the
top of the subgrade is insignificant.
(3) The maximum tensile strain, the maximum compressive strain, and the strain
amplitude of the five structures increase with temperature. The results of tests
indicate the increase rate of strain increases with the pavement surface tem-
perature, which demonstrates the temperature has significant effects on the
mechanical properties of the asphalt pavement.
(4) The prediction model for tensile strain based on temperature in the middle
depth of the asphalt layer, axle load and vehicle speed is highly reliable.
According to the prediction model, axle load has little influence on the tensile
strain. The tensile strain at the bottom of the asphalt layer increases roughly by
115 % with every 5 °C increase of temperature. The rates of changing of
tensile strain of structure 5 at different temperatures are most significant due to
the material of base course is asphalt-stabilized gravel, which is a
temperature-sensitive material.

References

Alagappan, P., Krishnan, J. M., & Veeraragavan, A. (2011). Mechanical response of modified
asphalt pavements. In: Transportation and development institute congress 2011@: Integrated
transportation and development for a better tomorrow (pp. 438–448). ASCE.
Bari, J., & Witczak, M. (2015). New predictive models for viscosity and complex shear modulus
of asphalt binders: For use with mechanistic-empirical pavement design guide. Transportation
Research Record: Journal of the Transportation Research Board.
Beainy, F., Commuri, S., & Zaman, M. (2013). Dynamical response of vibratory rollers during the
compaction of asphalt pavements. Journal of Engineering Mechanics.
Chatti, K., Mahoney, J. P., Monismith, C. L., & Moran, T. (1995). Field response and dynamic
modeling of an asphalt concrete pavement section under moving heavy trucks. In 4th
international symposium on heavy vehicle weights and dimension., Ann Arbor, Michigan,
USA.
Chen, Y., Zhang, H., Zhu, X. Q., & Liu, D. W. (2015). The response of pavement to the multi-axle
vehicle dynamic load. In 2015 international conference on electrical, automation and
mechanical engineering. Atlantis Press.
Claessen, A. I. M., Edwards, J. M., Sommer, P., & Uge, P. (1977). Asphalt pavement design—the
shell method. In: Volume I of proceedings of 4th international conference on structural design
of asphalt pavements, Ann Arbor, Michigan, August 22–26, 1977.
Deng, X. J., & Sun, L. (2000). Study on dynamics of vehicle-ground pavement structure system.
Beijing: China Communications Press.
Goktepe, A. B., Agar, E., & Lav, A. H. (2006). Advances in backcalculating the mechanical
properties of flexible pavements. Advances in Engineering Software, 37(7), 421–431.
Huang, Y. H. (1993). Pavement analysis and design.
308 A. Sha et al.

Huber, G. A. (1994). Weather database for the SUPERPAVE (TM) mix design system (Revised
edition).
Jing-song, S., Xiao-ming, H., & Gong-yun, L. (2007). Dynamic response analysis of pavement
structure under moving load. Journal of Highway and Transportation Research and
Development, 1 (130).
Kennedy, T. W., Huber, G. A., Harrigan, E. T., Cominsky, R. J., Hughes, C. S., et al. (1994).
Superior performing asphalt pavements (Superpave): The product of the SHRP asphalt
research program. Strategic Highway Research Program, National Research Council.
Monismith, C. L. (1992). Analytically based asphalt pavement design and rehabilitation: Theory
to practice, 1962–1992.
Perez, S. A., Balay, J. M., Tamagny, P., & Petit, C. (2007). Accelerated pavement testing and
modeling of reflective cracking in pavements. Engineering Failure Analysis, 14(8), 1526–
1537.
Robertson, W. D. (1997). Determining the winter design temperature for asphalt pavements.
Journal of the Association of Asphalt Paving Technologists, 66.
Willis, R., Timm, D., West, R., Powell, B., Robbins, M., et al. (2009). Phase III NCAT test track
findings. NCAT report, 8–9.
Yoder, E. J., & Witczak, M. W. (1975). Principles of pavement design. Hoboken: Wiley.
Zhonghong, D., & Fengying, N. (2014). Dynamic model and criteria indices of semi-rigid base
asphalt pavement. International Journal of Pavement Engineering, 15(9), 854–866.
Automating Mechanistic-Empirical
Pavement Design Calibration Studies

Xiaolong Guo and David H. Timm

Abstract Accelerated pavement testing (APT) applies intensive traffic loading for
the purpose of observing long-term performance of pavements in a short time. The
testing is beneficial for researchers as a timely and valuable data resource. As
Mechanistic-Empirical (M-E) pavement design approaches emerge as a favorable
alternative to traditional empirical approaches, there is a need for conducting cal-
ibration of the performance models in the M-E design framework. Data from APT
testing may be considered advantageous for calibration due to research-grade
accuracy conducted over shorter time intervals, but even so, M-E calibration is still
time-consuming because many combinations of the respective calibration coeffi-
cients need to be evaluated against observed performance from pavement sections.
The current version of the AASHTOWare™ Pavement ME Design software,
though serving as a powerful pavement design and analysis tool, is not necessarily
conducive to calibration studies as many interactions are needed by the user to
populate the software with trial calibration coefficients and extract performance
predictions after simulations have been completed. This study proposes using
automation to facilitate calibration studies whereby the repetitive manual operations
required by the software are recorded and built into a macro that greatly reduces the
required human interactions. The software automation is demonstrated through a
calibration study conducted at the National Center for Asphalt Technology (NCAT)
Test Track that examined fatigue cracking calibration. The efficiency of the auto-
mated approach is compared against conducting calibration requiring human
interaction with the software.

Keywords Data quality
M-E calibration
Automation
Required human

interactions

X. Guo (&)
D.H. Timm

Department of Civil Engineering, 238 Harbert Engineering Center,
Auburn 36849, USA
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 309

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_20
310 X. Guo and D.H. Timm

1 Background

The Mechanistic-Empirical (M-E) approach is considered more advanced than

traditional approaches because it can better account for changing paving technology
and diverse design conditions which should provide more cost-effective designs.
Noticing years of wear and tear, unrelenting traffic, an explosion of heavy trucks,
weather conditions, and delayed maintenance because of tight budgets and soaring
construction costs (AASHTO 2009), the cost-effectiveness of M-E-derived pave-
ment design is obviously timely and valuable for the highway community. Many
state highway agencies are undertaking implementation of the M-E design approach
into routine design practice. A questionnaire (NCHRP 2014) responded to by 48 U.
S. (92 %) and 9 Canadian (69 %) agencies suggested that three have already
implemented the MEPDG and 43 planned to implement the Mechanistic-Empirical
Pavement Design Guide (MEPDG) within 5 years. However, a number of chal-
lenges confront the implementation of the new approach. State agencies are
required to carry out a series of activities: conducting input sensitivity analyses,
developing input libraries (e.g., materials, climate, and load spectra), local cali-
bration of transfer functions, and delivering training courses. Among these activi-
ties, local calibration of the MEPDG (or its new version AASHTOWare™
Pavement ME Design) performance prediction models (i.e., transfer functions) was
indicated by the majority of agencies to be one of two activities that require
assistance, along with training in the use of AASHTOWare™ Pavement ME
Design software (Pierce and McGovern 2014).
Local calibration of the AASHTOWare™ Pavement ME Design models is
meant to reduce bias and standard error between the model-predicted and
field-measured performance, if they exist locally. The model prediction is very
likely to be adjusted to reflect the local trend, thereby yielding viable and
cost-effective pavement designs. Studies pertaining to local calibration have been
conducted by many, yet previous efforts (e.g., Kang and Adams 2007; VonQuintus
and Moulthrop 2007; Darter et al. 2013; Velasquez et al. 2009; ARA 2004; Guo
and Timm 2015) indicated two problematic issues encountered in conducting these
studies: research data quality and time and labor investment. Research data captures
pavement behaviors subject to loading and weathering so its quality has a signifi-
cant effect on gained model accuracy. On the other hand, time and labor investment,
though required by any investigation, is indeed a prominent concern for local
calibration studies due to the time-consuming iterations of software execution. One
calibration study (ARA 2004) provides a sense of considerable time and labor
investment: 4896 MEPDG runs were required and each run took about 30 min.
As for research data quality, several aspects of data were considered by previous
studies. Overall, data of good quality are of high value for local calibration. It was
believed that local calibration requires a vigilant selection of in-service pavement
sections that represent local pavement design, construction practices, and perfor-
mance (Haider et al. 2015). Particularly, the detail of input data may affect per-
formance prediction significantly. One study found a relevant clue that the rutting of
Automating Mechanistic-Empirical Pavement Design Calibration … 311

asphalt concrete was more accurately predicted at input level 1, whereas it was
over-predicted at input level 3 (Zhou et al. 2013). The determination of cracking
type sometimes requires coring or trenching in addition to visual surface inspection.
As noted by many, cracking observed on the surface can be caused by layer
slippage or top-down cracking, rather than traditional bottom-up cracking (Guo and
Timm 2015). Additionally, the data regarding pavement materials, environmental
conditions, and traffic loading should be accurately collected on an as-is basis, or at
least reasonably reflects characteristics of local design scenarios. Furthermore, it is
imperative that a consistent definition and measurement of the surface distresses
and other data be used and maintained throughout the calibration and validation
process (ARA 2004). It can be seen that the aspects of data quality in terms of
representativeness, the level of detail, availability and accuracy, compatibility and
consistency need to be considered when selecting data for local calibration.
Data from three sources were typically used in previous local calibration studies:
APT (or Test Track), Pavement Management System (PMS), and Long Term
Pavement Performance Program (LTPP). Of course, some studies may adopt a
combination of PMS and LTPP data or APT and LTPP data. Previous studies, listed
in Table 1, indicated data from these three sources has, if noteworthy enough, some
different patterns appearing in the aforementioned aspects of research data quality.
The comparison of these patterns may shed light on advantages and disadvantages
of adopting a source so that agencies may well have a better strategy to select data
for local calibration. In brief, data from APT has general advantages in level of
detail and availability and accuracy, whereas data from LTPP or PMS tend to have
a better representation of local traffic and climatic characteristics.
Beyond concerns of research data, the time and labor investment is another
obstacle to deal with for conducting local calibration. A great many iterations of
software execution were needed by previous studies (as shown in Table 2), and
thus, considerably large time and labor investment were required. In fact, the
iteration could include a series of human interactions: initialization of design pro-
ject, entry of calibration coefficients for trial, saving the design and running project
analysis, compiling of predicted and measured performance data, and performing
necessary calculation of statistics.
As a whole, the quality of research data depends on routine collection practice,
along with project budget and management strategy. It is also seen that APT is
probably not available for many due to expensive cost. Therefore, some only have
access to PMS or LTPP data. Yet the time and labor investment can be addressed by
the method of managing computer interaction to facilitate local calibration studies.
The limitation of time and labor investment can be resolved, to some degree, by
adopting an automation method or other approaches (e.g., using multiple computers
or hiring more labor only if the budget allows).
As for automation methods, only a few studies have mentioned the convenience
or adoption of M-E software execution. One study suggested that macro software
may be used for reading data from a spreadsheet and entering the data into the
correct MEPDG menu (Schram and Abdelrahman 2010). Excel macros can be
easily adapted to automatically retrieve the performance predictions for each
312 X. Guo and D.H. Timm

Table 1 Calibration data from typical sources

APT (or test track) PMS LTPP
Representativeness 10-Million ESALs, 19 Interstate and one Interstate or major
local design, single state highway highways widespread
climate area (Guo and sections, local design, over North America,
Timm 2015) multiple climate areas local design, diverse
Interstate sections, (Zhou et al. 2013) climate area (ARA
local design, single 38 state highway and 2004; FHWA 2014)
climate area (Hoegh Interstate sections,
et al. 2010) local design, multiple
climate area (Rahman
et al. 2013)
Level of detail Traffic: level 1 Traffic: mostly level 3; Traffic: levels 1, 2, 3
Materials: mostly Materials: mostly Materials: levels 1, 2,
level 1 levels 2 and 3; 3
Environment: on-site Environment: Environment: weather
weather collection estimated from nearby station (ARA 2004)
record (Hoegh et al. weather station
2010; Tompkins and records (Zhou et al.
Khazanovich 2007; 2013; Rahman et al.
Guo and Timm 2015) 2013)
Availability and Traffic: precisely Traffic: Traffic:
accuracy monitored, even WIM-monitored or not WIM-monitored
controlled available Environment: weather
Environment: on-site Environment: limited station monitored
monitored, even availability Materials: laboratory
controlled Materials: only a few and FWD testing
Materialslaboratory laboratory and FWD Performance: yearly
and FWD testing testing equipment surveying
Performance: routine Performance: (e.g., dipstick profiler,
visual identification equipment surveying laser equipment)
coupled with crack (e.g., dipstick profiler, (FHWA 2012, 2014;
mapping, coring, and laser equipment) Perera and Kohn
trenching (Ullidtz (Haider et al. 2015; 2001)
et al. 2008; Hoegh Mallela and
et al. 2010; Tompkins Titus-Glover 2014;
and Khazanovich Zhou et al. 2013;
2007; Guo and Timm Rahman et al. 2013;
2015) Mamlouk and Zapata
2010; Li et al. 2009;
Muthadi and Kim
2008)
Compatibility and Compatible and Compatibility and Compatible and
consistency consistent in most consistency not consistent in most
cases (Hoegh et al. guaranteed (Mallela cases (ARA 2004)
2010; Tompkins and and Titus-Glover
Khazanovich 2007; 2014)
Guo and Timm 2015)
Automating Mechanistic-Empirical Pavement Design Calibration … 313

Table 2 Time investment for sensitivity and calibration studies

Research purpose Investigators Number Estimated Automation Lines of
of runs execution time tool code
Local calibration Velasquez >202,664 N/A N/A N/A
and sensitivity et al. (2009)
National calibration ARA (2004) 4896 30 min/run N/A N/A
Local calibration Guo and >172 3 min/run Macro 200
Timm (2015) recorder

distress and store them in a single spreadsheet (Schram and Abdelrahman 2010).
Another study mentioned the use of macro recorder software and MATLAB to
avoid repetitive manual maneuver of running MEPDG analysis (Jadoun and Kim
2012). Some other also discussed an automation method applied in local calibra-
tion, which utilized macros to automatically execute required repetitive computer
actions. The adoption of automation reduced a great amount of labor for local
calibration (Guo 2013). Actually, the automation methods have been successfully
implemented in some MEPDG studies (Schram and Abdelrahman 2010; Jadoun
and Kim 2012; Guo 2013) to facilitate the software execution. Based on these
studies, there is sufficient reason to believe that automation methods can feasibly
benefit users in terms of automating software execution. Thus, this study was
conducted to demonstrate one automation method in the AASHTOWare™
Pavement ME Design and evaluate the efficiency of the automation method.

2 Objectives

The first objective was to apply automation within the M-E software to facilitate
local calibration studies. The second objective was to demonstrate automation using
a case study and quantify the benefits in time and human interaction savings.

3 Scope of Work

To accomplish the two objectives, an on-going calibration study by NCAT was

implemented with an automation tool to demonstrate how to facilitate the repetitive
execution of AASHTOWare™ Pavement ME Design software. The calibration data
were collected from the NCAT Test Track and laboratory equipment. A suitable
commercially-available software, Automation Anywhere, was used to record and
build macros that greatly reduce required time and human interaction investment.
Additionally, Microsoft Excel was an integral part of the automation schemes.
314 X. Guo and D.H. Timm

4 Automation Software Selection

A number of automation programs are available for use, such as AutoIt,

AutoHotKey, Automation Anywhere®, Keyboard Maestro, Livrot Mic Command©,
Macro Express®, Workspace Macro Pro, UBot Studio©, and GNU Xnee. Three
major criteria were used to select the suitable program based on the interaction
needs of the ME Design software. First, the automation software must be capable of
recording mouse movements and keyboard operations as these are integral to
operating the ME Design software. All the programs listed above but AutoHotKey
and UBot Studio have this capacity. Second, the capability of compilation to EXE
files is a necessary function since the developed macro can be turned into an
Executable program for other users. This function can be used to further develop
self-contained programs for distribution. Only three programs, AutoIt,
AutoHotKey, and Automation Anywhere can compile to EXE files, as well as
satisfying the first criterion. Third, “smart” mouse recording (record the absolute
and relative coordinate) is a required function for this study since the mouse macro
crashes easily if only the absolute coordinate is used to locate the mouse pointer. Of
the programs listed above, Automation Anywhere is the only suitable software that
has the function of smart mouse recording and the other features. Therefore, the
software was adopted in this study to interact with the ME Design software.

5 Case Study

The case study exemplifies the application of automation within a local calibration
study currently being performed at NCAT. The study is to evaluate the nationally-
calibrated distress prediction models in the AASHTOWare™ Pavement ME Design
software to assess their local applicability and conduct local calibration to improve
model accuracy, if warranted. This paper focuses on the fatigue cracking calibra-
tion, which is the only attempted calibration so far. In this study, the main function
of automation is to initialize the ME software, enter trial calibration coefficients,
launch ME software analysis, extract performance predictions, and calculate the
sum of squared error. The sum of squared error was computed for each trial
combination of calibration coefficients so that the least sum of squared error among
all trials identified the best-fit local calibration coefficients. Since the current version
of AASHTOWare™ Pavement ME Design software does not have built-in local
calibration features, the application of automation to local calibration study is of
significant value for designers to facilitate the local calibration of the
AASHTOWare™ Pavement ME Design software.
The automation flow chart in this study is presented in Fig. 1. The automation
started with defining the variables required for loop control and data transfer and set
some initial values for a loop counter. Then, a loop containing repetitive actions
was executed for each trial of calibration coefficients. In the loop, the variables for
Automating Mechanistic-Empirical Pavement Design Calibration … 315

Fig. 1 Flowchart of automation applied in a local calibration study

data transfer were assigned with user-defined information. The user-defined infor-
mation was arranged in a formatted “Input Summary” Excel spreadsheet: the col-
umns classified variables and the rows specified trials. Next, the project file was
opened followed by automatically entering the trial calibration coefficients, and
software analysis was executed. When the ME Design software finished execution,
output data were extracted and transferred to a formatted “Output Summary” Excel
316 X. Guo and D.H. Timm

spreadsheet: the predicted time-series distress data were placed in columns along-
side the measured time-series distress data, and thus comparisons between predicted
and measured distress data could be performed in rows. The “Output Summary”
Excel spreadsheet was used for computing the sum of squared error. The number of
loops corresponded to the number of trials. The logic of automation applied to local
calibration study is similar to that applied to sensitivity analysis, which indicates
fair portability of this automation logic. Similar studies in the future may be able to
adopt this logic as well.
With the automation logic illustrated in Fig. 1, the actions that needed to be
automated were recorded stepwise by the macro recorder software. However, to
guarantee the fluency of automation, some issues during macro recording were
worthy of special attention. First, once the automation was run, the focus of each
computer action needed to be specifically designated in case of running errors (i.e.,
interruption). For instance, when the software analysis was completed, it was
necessary to set the computer focus to the specific software output Excel spread-
sheet which contained performance prediction results, before moving forward to
compile information to the “Output Summary” Excel spreadsheet. Thus, the focus
was first set on the specific software output spreadsheet and then shifted to the
“Output Summary” spreadsheet. The function of “activating sheet” in Automation
Anywhere may be considered for use to ensure an effective focus shift. Secondly,
the Excel spreadsheet running in the background is a problem for AASHTOWare™
Pavement ME Design software to generate Excel output after executing analysis.
This may happen when Excel is opened and closed many times. Since many trials
were executed, Excel spreadsheets were opened and closed as trials were run
one-by-one. Therefore, a precautionary measure was needed for a check on whether
any Excel spreadsheet was still running in the background before the ME Design
software itself started to generate Excel output. The measure for this check was
built into macros by using the “Window Exists” function. If an Excel spreadsheet
was detected as running in the background, then an action was needed to close it.
To resolve this, a self-contained program was developed for terminating the running
of Excel in the background. Once these macro issues were resolved, the automation
scheme ran smoothly without error.
The metric of quantifying benefits is primarily the time savings achieved through
automation. The labor savings can be quantified by the actual time requirement of
manual maneuver excluding the part of executing software analysis. For fatigue
cracking calibration, three coefficients (i.e., bf1, bf2, bf3) were iterated eleven times
with 25 test sections within the study. Therefore, 33,275 (11  11  11  25)
total trials were required. Table 3 lists the time required for all the trials. By using
automation, the total time required was reduced by 44 %. Again, this includes ME
Design software running time. If only actions requiring human interaction are
considered, the time savings become 62.5 %. The labor savings for this project is
99,826 min.
Automating Mechanistic-Empirical Pavement Design Calibration … 317

Table 3 Time required for software execution by all trials

Stepwise work Time (min)
Manual maneuver Automation Savings
Initialize software 16,638 8319 8319
Adjust calibration coefficients 16,638 8319 8319
Execute software analysis 49,913 49,913 0
Extract output 33,275 8319 24,956
Calculate sum of squared error 33,275 8319 24,956
Total 149,739 83,189 66,550

6 Discussion

It has been discussed that research data are crucial to local calibration studies. Data
of different quality impacts the results of local calibration. However, it is still a
question in which way different data quality affects the calibration results. The
question, if answered, may well help agencies to select the most appropriate data for
local calibration. It has been commonly noted that the goal of local calibration is to
improve model accuracy thereby optimizing the design options that highway
agencies have. Nevertheless, the effectiveness of local calibration can be evaluated
by model validation through trials of specific design scenarios. Therefore, it might
be worthy to quantify the effects of research data quality on the effectiveness of
local calibration. So far very few studies, if any, have suggested a significant
relationship, for instance, data obtained from measurement equipment that has
better accuracy or data satisfying a higher input level has a correlation with levels of
accuracy improvement in validation (i.e., increase in R2 or percent decrease in sum
of squared error).
As demonstrated by the case study, the time committed to software execution can
be greatly reduced by using automation. Total time saving of 44 % was achieved,
while the time savings in regards to human interaction were 62.5 %. Additionally,
since the required human interaction is significantly minimized during software
execution, automation has the potential to greatly reduce human error. If automation
is adopted, the software execution becomes merely a “mouse click” or “button
press,” instead of cumbersome manual software maneuvers. Speaking of ease-of-
use, the programming of automation is straightforward and easy to understand. The
macro recorder software is object-oriented so that user can record the actions that
need to be repeated. The lines of code generated from recording are less than that in
previous studies (Schwartz et al. 2011; Guo and Timm 2015) listed in Table 1.
Automation applied to local calibration is beneficial since it saves a considerable
amount of time and effort that is otherwise placed on researchers. The automation
converts the time-consuming factorial experiment into a batch run with compiled
results, providing researchers with a concise and clear relationship between trial
calibration coefficients and prediction accuracy indicator. Given any factorial, the
best-fit local calibration coefficients can be determined after running automation.
318 X. Guo and D.H. Timm

7 Recommendations

The M-E based approach advances many aspects of pavement design, especially in
characterizing design factors (e.g., traffic, climate, and materials) and predicting
specific modes of distress. The cost of such advances, however, lies in the com-
plexity of the design approach, along with time and labor required for implemen-
tation. Specifically, the AASHTOWare™ Pavement ME Design software requires
calibration data of good quality and tremendously large amounts of human inter-
action, which complicates the implementation of the M-E based approach. The
impact of data quality on calibration results is recommended to be investigated so as
to provide clues on the strategy of calibration data selection. Given the fact that the
time of software execution and demands of human interaction can be greatly
reduced, it is recommended that an automation tool be applied to local calibration
studies, if not yet implemented. Furthermore, if automation is warranted by many
researchers, it is recommended to develop a self-contained and easy-to-use program
that requires no programming of automation macros.

References

AASHTO. (2009). Rough roads ahead. Publication code: RRA-1, Washington, DC.
ARA. (2004). Guide for mechanistic-empirical pavement design of new and rehabilitated
pavement structures. ARA Inc. Eres Division.
Darter, M. I., Titus-Glover, L., & Wolf, D. J. (2013). Development of a traffic data input system in
Arizona for the MEPDG. Final report 672, Arizona Department of Transportation Research
Center.
FHWA. (2012). LTPP beyond FY 2009: What needs to be done? Publication no.
FHWA-HRT-09-052.
FHWA. (2014). LTPP 2014 and beyond—What is needed and what is needed and what can be
done? Publication no. FHWA-HRT-15-017.
Guo, X. (2013). Local calibration of the MEPDG using test track data. Master’s thesis, Auburn
University.
Guo, X., & Timm, D. H. (2015). Local calibration of MEPDG using national center for asphalt
technology test track data. In Proceedings of the 94th annual transportation research board,
Washington, DC.
Haider, S. W., Brink, W. C., Buch, N., & Chatti, K. (2015). Process and data needs for local
calibration of performance models in the pavement-ME. In Proceedings of the 94th annual
transportation research board, Washington, DC.
Hoegh, K., Khazanovich, L., & Jensen, M. (2010). Local calibration of Mechanistic-Empirical
pavement design guide rutting model. Transportation Research Record: Journal of the
Transportation Research Board, 2180, 130–141.
Jadoun, F. M., & Kim, Y. R. (2012). Calibrating Mechanistic-Empirical pavement design guide for
North Carolina. Transportation Research Record: Journal of the Transportation Research
Board, 2305, 131–140.
Kang, M., & Adams, T. M. (2007) Local calibration for fatigue cracking models used in the
Mechanistic-Empirical pavement design guide. In Proceeding of the 2007 mid-continent
transportation research symposium. Ames, Iowa.
Automating Mechanistic-Empirical Pavement Design Calibration … 319

Li, J., Pierce, L. M., & Uhlmeyer, J. (2009). Calibration of flexible pavement in
Mechanistic-Empirical pavement design guide for Washington State. Transportation
Research Record: Journal of the Transportation Research Board, 2095, 73–83.
Mallela, J., & Titus-Glover, L. (2014). Role of pavement management data in the implementation
of AASHTO’s pavement ME design methodology. Paper prepared for presentation at the
experience with pavement ME design session of the 2014 conference of the transportation
association of Canada Montreal, Quebec.
Mamlouk, M., & Zapata, C. E. (2010). Necessary assessment of use of state pavement
management system data in Mechanistic-Empirical Pavement design guide calibration process.
Transportation Research Record: Journal of the Transportation Research Board, 2153, 58–66.
Muthadi, N. R., & Kim, Y. R. (2008). Local calibration of Mechanistic-Empirical pavement design
guide for flexible pavement design. Transportation Research Record: Journal of the
Transportation Research Board, 2087, 131–141.
Perera, R. W., & Kohn, S. D. (2001). LTPP data analysis: Factors affecting pavement smoothness.
Prepared for National Cooperative Highway Research Program, Transportation Research
Board, National Research Council, Plymouth, Michigan.
Pierce, L. M., & McGovern, G. (2014). Implementation of the AASHTO Mechanistic-Empirical
pavement design guide and software. NCHRP Synthesis 457, Transportation Research Board
of the National Academies, Washington, DC.
Rahman, M. S., Williams, R. C., & Scholz, T. (2013). Local calibration of the fatigue prediction
models in the MEPDG for pavement rehabilitation in Oregon. In Proceedings of the 92th
annual transportation research board, Washington, DC.
Schram, S. A., & Abdelrahman, M. (2010). Integration of Mechanistic-Empirical pavement design
guide distresses with local performance indices. Transportation Research Record: Journal of
the Transportation Research Board, 2153, 13–23.
Tompkins, D., & Khazanovich, L. (2007). MnRoad lessons learned. Publication
No. MN/RC-2007-06.
Ullidtz, P., Harvey, J., Tsai, B., & Monismith, C. L. (2008). Calibration of Mechanistic-Empirical
models for flexible pavements using California Heavy Vehicle Simulators. Transportation
Research Record: Journal of the Transportation Research Board, 2087, 20–28.
Velasquez, R., Hoegh, K., Yut, L., Funk, N., Cochran, G., Marasteanu, M., & Khazanovich, L.
(2009). Implementation of the MEPDG for new and rehabilitated pavement structures for
design of concrete and asphalt pavements in Minnesota. Publication no. NM/RC 2009-06,
Minnesota Department of Transportation.
VonQuintus, H. L., & Moulthrop, J. S. (2007). Mechanistic-Empirical pavement design guide
flexible pavement performance prediction models for Montana: Volume I executive research
summary. Publication no. FHWA/MT-07-008/8158-1, Federal Highway Administration.
Zhou, C., Huang, B., Shu, X., & Dong, Q. (2013). Validating MEPDG with Tennessee pavement
performance Data. American Society of Civil Engineers: Journal of Transportation
Engineering, 139(3).
Calibration of ME Design Using
a Combination of APT and PMS Data

Jeremy D. Lea, Rongzong Wu and John T. Harvey

Abstract The ultimate goal of ME design, and much of the pavement engineering
behind it, such as APT, is to provide true estimates of how pavements will perform
in the field. In the past, most design methods have been calibrated with APT data,
and there has been some difficultly in calibrating to field data, including LTPP data.
While this is partly due to a lack of understanding of the performance of pavements,
much of the difficultly seems to arise because there is a disconnect between how
performance is explained in design methods and how it is measured in the field.
This paper details a framework for understanding APT data and PMS data, and for
jointly modelling both sets while calibrating an ME design method.

Keywords Mechanistic-empirical pavement design
Calibration
Pavement

management

1 Introduction

Pavement engineers need to be able to predict how a pavement will perform in the
field, if we are to move past “catalog” designs of what has worked in the past, to
effective pavement design. A major step in this process is the development of
Mechanistic-Empirical (ME) pavement design methods that provide a more realistic
prediction of pavement behavior based on fundamental mechanical processes rather
than direct empirical relationships to past performance. Much of the work in
Accelerated Pavement Testing (APT) has focused on experiments to understand
these mechanical processes and develop the calibration data to explain how they
affect the behavior of full-scale pavement systems. However, a gap will always
remain between APT and field performance, because we will never be able to
model a full range of traffic and environmental loading in APT and the field

J.D. Lea (&)
R. Wu
J.T. Harvey

University of California Pavement Research Center, UC Davis,
One Shields Ave, Davis, CA, USA
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 321

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_21
322 J.D. Lea et al.

construction of projects will always differ from the final design, because materials
and construction processes cannot be fully controlled. This is where the empirical
component of the design method must be calibrated to match field performance.
While some work has been done on using Long-Term Pavement Performance
(LTTP) and Pavement Management System (PMS) data for calibration, the results
of various calibrations show very low correlations between observed and predicted
performance (NCHRP 2004).
This paper lays out a framework that explains why previous attempts at
LTPP/PMS calibration have failed, and provides a way forward for calibration of
ME design methods with PMS data. The framework relies on a number of prin-
ciples that may be controversial since they run counter too much of the language
used in relation to ME design, even if they are already ingrained in the mathematics.
These principles are outlined as the framework is explained.
The calibration procedures here assume an incremental-recursive ME design
method like CalME (Ullidtz et al. 2010) that is able to provide predictions of
performance in time, not just a final prediction of design life. The calibration also
assumes that extensive detailed PMS data, with small measurement intervals, is
available. Agencies that have been collecting data at 0.1 mile or smaller intervals,
and storing the raw historical information, should have sufficient data to perform
calibration.

2 Background

First generation ME design methods that used direct “transfer functions”, such as
the SAMDM (Theyse et al. 1996), have typically been calibrated using data with a
single prediction of N (the number of repeats to failure, for this failure mode), based
on a single determination of some stress or strain. Within these methods, some have
been calibrated using APT data, and some using PMS data, where it was possible to
predict stress or strain for in-service pavements. In these methods, N is the
dependent variable in the empirical component and the failure distress level is a
dependent variable.
The initial calibration of the MEPDG (NCHRP 2004) used a similar approach,
based on LTPP data, and local calibration processes are being pursued using
additional LTPP or PMS data, depending on what is available. The only real
difference is that the MEPDG does not use a direct prediction of N, but rather an
incremental prediction as time/loading progresses. In other words, the loading (n) is
a dependent variable and the distress is now the independent variable. As a result,
this allows multiple calibration points through time from a single LTPP section and
decouples the selection of failure criteria from the empirical model.
The calibration of CalME to date has taken a different approach, where the
method has been incrementally developed using calibration from HVS and
full-scale tests. Rather than adding models because they were available, models
have only been added to CalME when they were needed to explain the individual
Calibration of ME Design Using a Combination of APT and PMS Data 323

behavior of a HVS test. Starting with relatively simple structures, and a handful of
models known to be required (bottom-up fatigue cracking and rutting), models were
added for observed behaviors such as reflection cracking, slip, crushing, and others
—as documented in the various calibration reports (Ullidtz et al. 2006). CalME uses
the same approach as the MEPDG, in that distress is modelled as a function of
loading (n), and distress is a dependent variable.

2.1 Variability Within ME Design Methods

In all engineering, a factor of safety needs to be used to account for unexplained

behavior or defects in the materials and construction. In pavement design, this has
traditionally been expressed as “design method reliability.” In methods that predict
N this takes the form of a probability distribution around this prediction. It is then a
matter of reducing the predicted value so that, for example, 90 % of the predictions
of N fall above the expected loading n. This distribution is a function of the
observed variability in the calibration data, so the selected confidence level reflects
a probability that any particular section within the calibration data will fail after the
predicted loading. As such, it is not a prediction of within-section variability, but
rather assumes that within-section variability is included in the definition of failure
(and is often 50 % within the type of calibration data used, since it is the mean
distress for the section). In addition, it only reflects the between-section variability
within the calibration data, not the network wide observed between-section
variability.
On the other hand, in pavement management there is often an implicit
within-segment reliability. For example, many PMSs will trigger maintenance
actions when more than 10 % of a segment has past some distress threshold.
A pavement design should not be designed so that the expected (50 %)
within-project performance matches this threshold, but rather so that the combi-
nation of within- and between-section reliability of the design is 90 % (i.e. 10 %
failed).

3 Framework

By definition, calibration requires that we have a standard to calibrate against. In

pavement design, this standard is often vague. It is clear that we would like to
determine how some theoretical pavement structure would perform at some real
location with future climate and traffic. In practice, we cannot know the future
climate or traffic, and we cannot know many of the properties of the actual pave-
ment that might be constructed in response to the theoretical design presented. As a
result, we need a design method that can be calibrated against actual in-service
pavements, with historical climate and traffic information, but be used in such a way
324 J.D. Lea et al.

Fig. 1 Within-project damage and distress process from vehicle i

that it can predict performance over a wide range of possible outcomes from the
construction process and future traffic and climate. This is broader than the tradi-
tional split of “between” and “within” project variability.
In is paper, the following language is used (as shown in Fig. 1): a pavement that
consists of a number of layers is constructed along a project. Each layer has initial
properties determined through the unique combination of materials and construc-
tion, which vary along the length of the project (spatial variability). This is shown
on the top panel of the figure. At some time t, each layer has a set of current
properties that are determined by the initial properties, the recent
climate/environmental conditions, and the current damage (the second panel). At
time t a vehicle i traverses the pavement, causing a dynamic response within the
pavement (both in space and time). This dynamic response causes incremental
damage within the structure that accumulates with the current damage to become
accumulated damage. This accumulated damage within the structure might be
partially observable within the layers (typically via a surrogate variable, using
Calibration of ME Design Using a Combination of APT and PMS Data 325

non-destructive testing). Moreover, this damage results in internal distress (e.g.

cracking, deformation, or crushing), that can be observed directly (typically via
destructive testing, such as coring). The internal distress, in turn, manifests as
surface distress that can be directly observed. The occupants of vehicle i perceive
this surface distress as affecting the serviceability of the pavement. The perceived
serviceability of the pavement for vehicle i is also reflected as an incremental cost
for that vehicle to traverse the project. The perceived serviceability, as measured by
the average road user, is captured as the functional condition of the pavement. An
engineer, looking at the internal/surface distress can also rate the structural con-
dition of the pavement, typically through some scale that transforms the raw surface
distress into a severity. The functional condition of the pavement is also typically
translated into a scale that reflects the severity of the loss of condition from some
ideal/design level (how fit-for-purpose is a particular location at the current time).
The full history of functional and structural condition over time is the functional or
structural performance of the pavement. The history of a particular distress is the
deterioration for that distress.
Based on a combination of the current severity of functional and structural
condition, and the extent over which severity levels manifest over the project,
various maintenance, rehabilitation or reconstruction (MR&R) actions may be
appropriate, but this is beyond the scope of this paper. Note, however, that the
framework does not define or use the terms failure or design life. The assumption is
that decisions within the design process will be based on a decision-making
framework such as LCCA, LCA, or PMS that determine future treatment needs, and
define an analysis timeframe. It is also expected that this decision-making frame-
work will trigger MR&R actions based on the severity of individual components of
distress rather than an aggregate condition.
In addition to these definitions, we also need definitions for variability, as shown
on Fig. 2 (for construction of a new pavement). Within a project, there are different
types of variability as the project progresses. At time t the project will have some
expected condition, with some error. During pavement design, the expected con-
dition of the project will be the same at all locations along the project (because there
is no prior knowledge of any spatial differences along the project—for a mainte-
nance or rehabilitation project we may have some prior knowledge of spatial
variability), and will be based on the network-wide expected performance distri-
bution. It is this distribution that the framework within this paper is addresses. Once
the bid process is complete and the contractor selects materials, this expected
performance changes to match the performance of the selected materials, but is still
uniform spatially. Then once construction is complete the spatial variability within
the layers is built in and the expected condition is further refined. After the actual
passage of time, to time t, the traffic and past environmental conditions are now
known, and further limit the expected performance.
However, we would still have some error on the expected condition, because we
would not know exactly how the materials and structure performed until we
actually observe the condition of the pavement. Even then there is some difference
between the measured condition of the pavement and the actual perceived
326 J.D. Lea et al.

Fig. 2 Differences in within-project variability

serviceability, because we cannot truly measure serviceability, we can only

approximate it through measures such as rutting, cracking and roughness. There
will also be real errors in these measurements, because of technological limitations.
Finally, within a PMS we could use the measured condition to update a per-project
prediction equation for the project (e.g. using Bayesian updating of parameters in a
performance model). All of these are measures of the distribution of within-project
variability for the project at time t, and we must be careful when we use this term
about exactly which is intended.
Between-project variability is also a difficult concept to pin down. The definition
would be the length-weighted distribution of the mean performance of different
projects with identical designs, environmental conditions, and traffic at time t after
construction. However, it is not actually possible to find such a set of projects,
because even if we had several identical designs, we would not find pavements with
the same traffic or environmental loading. This issue will be addressed below. In
design, we are also interested in the combined within- and between-project vari-
ability, because the ultimate question is not the expected performance of a project,
Calibration of ME Design Using a Combination of APT and PMS Data 327

but the variability in performance of a new project. In other words, if we design this
pavement, not knowing the exact materials, construction, future loading, or climate,
what is the probability that, for example, more than 10 % of the pavement will have
fatigue cracking 20 years after construction?
Finally, the vertical bar on Fig. 2 represents what would be seen on a typical
APT section. While we will still encounter all of the different types of
within-project variability, these will typically be over a small enough distance that
the actual observed within-section variability is small, meaning that the condition
on the APT section can be averaged and treated as a single sample. With LTPP data
one typically has the worst type of data: the section is long enough that it exhibits
significant within-section variability (although not all of the within-project vari-
ability), but the condition data is averaged over the section, so this critical data is
lost, masking the true variability in performance.

3.1 Guiding Principles

The first guiding principle of the framework is that pavement analysis will never be
deterministic, but will always have a probabilistic component. This does not mean
that we will never be able to develop a fully mechanistic model for some particular
pavement under a set of defined loading and environmental conditions, only that we
will never be able to specify these defined loading and environmental conditions
deterministically in analysis for design. As a result, pavement analysis will always
have a probabilistic component, and this is exploited in the framework to express all
of the performance parameters in probabilistic terms: the system only provides the
probability that a certain severity threshold for some distress will be exceeded after
some time.
The second principle for the framework is that the mechanistic models are used
to explain the severity of the distress, while the extent is explained through the
variability in distress from both variability in the inputs and uncertainty in the
empirical component of the method. This can be seen as dividing the pavement into
a series of independent transverse “slices” that each model the damage → dis-
tress → condition process in a deterministic fashion. A random sample of such
slices then is used to capture the behavior along the project. In the calibration
process, short APT sections (such as the traditional 8 m HVS section) are assumed
to constitute one of these slices.
The third is that the current layer properties are only a function of damage within
the layer, not of the corresponding distress. This limits the framework to not
allowing second order impacts of distress to influence performance—for example,
the framework would not allow cracking to alter the stiffness of the material, or for
rutting to change the layer thickness. This principle could be removed, but this
would make the calibration significantly more difficult, because it would require a
simultaneous calibration of the damage and distress models. This assumption does
impact calibration to APT and other data where the condition is often run to
328 J.D. Lea et al.

extremes, allowing these second order factors to influence behavior. For in-service
pavements, it is unusual to allow these second order factors to come into play and,
once they have, it is often after critical MR&R decisions have been made (so, for
example, once one needs to account for dynamic loading because of roughness, an
economic analysis will usually indicate the maintenance should already have been
performed).
It should not need to be stated, but the models and formulations used in the
calibration framework are dependent on the earlier components, and these com-
ponents cannot be replaced without a full recalibration of the later components.
Thus, if a layered linear-elastic solution is used to derive the pavement response to
loading, it cannot be replaced with a non-linear visco-elastic solution—this does not
lead to more accurate results, only to incorrect/biased results. This is the fourth and
final principle of the framework: that all of the outputs from the various components
are not true values that need to be “shifted” to match reality. They are random
variables that reflect some underlying but unobserved process, and are used in later
components as endogenous latent variables that are better correlated to the observed
dependent variables than the observed independent variables.

4 Calibration Procedure

The calibration procedure proceeds in three or four phases. The first two phases use
APT data, and the third and fourth use PMS data. The first phase is to calibrate the
recursive components of design method against dynamic data, which in CalME
involves calibration of the damage models. The second is to calibrate the distress
models, to obtain the correct shapes for the distress curves. The final phases are to
calibrate the means and standard deviation of the predicted distress to observations
from PMS data, for individual projects in the third phase and for the entire network
in the fourth phase. The first two phases will not be covered in detail, since they do
not differ from existing published work on the calibration of CalME (Ullidtz et al.
2010).

4.1 Phase 1: Calibration of Recursive Components

This phase requires at least one APT section that exhibits behaviors of the model
that one wishes to calibrate. A detailed study of the section should have been
undertaken to establish material properties, layer thicknesses, dynamic responses,
etc. This implies a properly managed APT program, focused on “value-added”
testing rather than “proof testing” of new structures. The actual process of col-
lecting this data has been covered elsewhere (Lea 2008). The APT section should
be split into as many subsections as possible that exhibit unique behavior, if the
source of these differences can be established. An initial model is built based on the
Calibration of ME Design Using a Combination of APT and PMS Data 329

layer properties and material properties. In CalME, the coefficients from material
tests are used directly where possible. The coefficients of the damage models are
then adjusted to match the dynamic response data from the section. Typically, it is
easier to match the peak deflection information than strain or stress data, because
these rely on more complex mechanistic models.
Each model in the ME analysis has a shift factor (traditionally A), that defines the
scale and distribution of the model. The calibration process here involves adjusting
the shift factor (or rather the mean of the shift factor distribution), so that the
predicted dynamic response matches the observed response. The residual error in
the fit is then attributed to the shift factor as the standard deviation.

4.2 Phase 2: Calibration of Distress Models

Once the damage models are calibrated then the distress models can be calibrated.
The distress models in pavements are normally cracking, rutting (permanent
deformation), and crushing models. The cracking and crushing models can often be
expressed as a time to initiation based on the latent damage variables and then a
progression model. Permanent deformation is a more complex process, where the
effects of wander and other inputs must be accounted for. CalME currently uses a
simple process of predicting vertical permanent deformation within each layer,
which works relatively well. This area of ME design is the most in need of active
current research.
The result of the calibration is a small set of deterministic performance models
for the APT section. This model will still have some residual error, but should be as
close to well specified and unbiased as is possible for such a complex system. If
multiple sections with the same materials are available, then a process similar to
mixed-effects modelling needs to be followed to reconcile any differences in fitted
coefficients between the same materials, moving the discrepancies into a section (or
sub-section) specific error parameter (Lea 2012).
Once again, this process involves a shift factor calibration. However, the primary
output of this calibration is to set parameters other than the shift factor that cannot
be set from materials testing, so that the shape of the deterioration matches the
observed values. The shift factors here are specific to the section, and the number of
sections is too small to form a valid statistical set.

4.3 Phase 3: Calibration to Section Specific PMS Data

If there are any PMS/LTPP sections with sufficient information to pursue a full
calibration then these should be considered as an additional calibration dataset.
However, even with data at the level of the US national LTPP project it is unlikely
that there will be sufficient information to divide these full-scale sections into
330 J.D. Lea et al.

transverse segments that can be handled independently. Typically, there is no

detailed spatial information available for layer properties (e.g. thickness and
air-voids), so sub-section models on the one to ten meter scale are not possible. As a
result, the calibration process for these sections needs to progress differently. On the
assumption that there is material test data available, these sub-section models need
to be built using a Monte Carlo simulation approach to handle the spatial distri-
bution in the sub-sections. The LTPP performance data typically includes either
severity/extent ratings for various surface distress types or means and standard
deviations. These sections need to be calibrated using the procedure outlined below
in phase 4, except that they provide true within-project variability for a particular
project. The value of the results from this phase is not in the obtained calibration
(since this is for a section that will never be again), but in validation of the models
from phases 1 and 2, to provide confidence that the models can predict performance
on an in-service project. As such, this phase and phase 4 could be switched, or run
contemporaneously.

4.4 Phase 4: Calibration to Network Wide PMS Data

The available network wide PMS data needs to be cleaned and processed to be
ready for the calibration process. The resulting dataset is then filtered to distinct
project/treatment types, based on the layers and material types. There will be
projects that are unique or overly complex, that should be discarded, as they will
not be easy to model. For each surface distress type that was recorded, various
severity levels are chosen (the data will often constrain this selection) and then a
statistical model is fitted to this data. These models always take the form of
percentage-of-extent-above-threshold-severity, and so are a type of failure proba-
bility model. As such, they are best modelled using a generalized linear regression
model (such as a probit or logit), and because the calibration data will consist of a
time series for each section (forming a panel data set), a mixed-effects model is
appropriate. Using appropriate statistical methods, a regression model is fitted to the
available PMS data that controls for factors such as design layer thickness, climate
zones, and other independent variables. This model can make use of
Empirical-Mechanistic principles (selecting model variables and functional forms
based on the mechanistic model rather than statistical significance) (Madanat et al.
2008). At this point, the goal is to obtain a model that allows one to control for
various differences in these independent variables, and for as many input variables
in the ME analysis method as possible. This model (or models) would also be
appropriate for direct use in PMS software. Because the model is fitted using
in-service pavements, the variance in the model accounts for the true variance of the
process: it takes into account variance both within- and between-projects, and
variance caused by different materials, traffic, and all of the other unobserved
variables in pavement behavior.
Calibration of ME Design Using a Combination of APT and PMS Data 331

This statistical model of the PMS data allows for the prediction of the expected
performance of a particular design pavement, controlling for layer thickness, traffic,
climate zone, and any of the other observed independent variables. As such, a
conditional model can be developed by fixing these variables and allowing them to
be entered into the ME analysis as deterministic values, and providing an expected
distribution of performance which is conditional on these values. This conditional
model addresses the issue of not being able to determine between-project variability
because we cannot find projects with the same climate, traffic, etc. In the ME
analysis the remaining parameters should either be fixed at their expected values or
have an input distribution into the Monte Carlo simulation. Running a Monte Carlo
simulation, with the ME analysis and values sampled from the input distributions,
provides a very rich data set of expected performance. Recall that each run in the
Monte Carlo simulation is treated as an independent transverse cross-section of the
pavement, and the outputs from the ME analysis are expressed as severity values
for various distresses. This data can be processed to extract the same
percentage-of-extent-above-threshold-severity information as is used for the
empirical modelling of the PMS data, and is predicted by the conditional model.
Since the conditional model includes all of the observed variance, what remains is
to shift and scale the distribution obtained in the Monte Carlo simulation to match
the conditional PMS distribution. This is done through the shift factors on the
distress models. The shift factors should be set to a constant during the Monte Carlo
simulation (this is the reason for the third principle: that the distress models cannot
impact the recursive components of the models), and then the distributional
parameters for shift factor can be established by a process of distribution fitting.
The Monte-Carlo simulation provides a sample of distress extents, that each
represent the expected value for a single simulation. This sample can be used to
derive an empirical distribution for within-project variability. The shift factor is first
scaled so that the expected distress matches the “observed” distress (which is
actually also an expected distress from the PMS model, conditioned on the various
input parameters), and then the parameters of the shift-factor distribution are set so
that the convolution of the shift-factor distribution with the empirical distribution of
distress from the Monte-Carlo simulation matches that of the conditional distri-
bution from the PMS model. This process is repeated for different layer thicknesses,
traffic, climate, and thresholds, and the shift factors set based on a fit to the entire
set, rather than just one distribution.
In this ME analysis a set of standard materials should be used. Either these can
be based on results from real materials with “average” performance (based on
expert judgement) or hand fitted coefficients to create materials that have reasonable
performance compared to the historical PMS data. The distributional parameters for
the shift factors are linked to these standard materials, and so they must be used for
all design involving the calibrated shift factors. This does not preclude the use of
other/new materials in design, only that the designer needs to be aware that the shift
factors associated with these materials may not account for between-project vari-
ability, especially if they have been calibrated with limited data. However, this
cannot be avoided; since we cannot know how much variability a new material will
332 J.D. Lea et al.

have in-service until it is actually used—as engineers we must make educated

assumptions about how these materials might perform. What this framework does
provide is a way to split the variability in the PMS data into within- and
between-project components, so that various rational assumptions about these can
be made concerning new materials.

5 Current Progress and Future Work

The UCPRC has investigated many of the approaches outlined here, and has
completed Phase 1 and 2 calibration of CalME for California. The historical PMS
data has also been extracted, cleaned, and merged with the detailed condition
information obtained in recent APCS surveys. There has been some delay in
continued APCS surveys, which have limited the ability to obtain a proper time
series for many distress values, preventing the development of the required
empirical models for the PMS data. Caltrans has made great strides in the digiti-
zation of their construction history information, providing details for thousands of
historical projects involving a full range of treatments. Together the construction
history and historical PMS data are being used to generate panel datasets for the
modelling of various types of distress and treatment, although there is still signif-
icant data cleaning work involved in the development of a final network-wide PMS
dataset.
Some work on empirical modeling has been done, but much remains, and is a
current research focus area. CalME is also being rewritten, and one of the goals of
that process is to make it easier to run Monte Carlo simulations with appropriate
input distributions and to capture the outputs. The final components needed for
Phase 3/4 are thus not fully in place. UCPRC will likely perform some Phase 3
calibration on individual PMS projects, just to validate the CalME methodology,
before proceeding to a Phase 4 calibration for new pavement designs and then
various MR&R treatments.
No doubt, significant hurdles remain in pursuing a network wide calibration of
ME design methods. The framework presented here attempts to tie together ME
design and PMS data in a fashion that respects the nature and meaning of the
models and data from each, and defines performance and variability in a consistent
and rigorous way. It is likely that as full calibration is attempted many sources of
unexplained error will be found, requiring additional ME development work to
address these previously unaccounted for behaviors. However, such calibration
must be attempted for ME design methods to reach their full potential.
Calibration of ME Design Using a Combination of APT and PMS Data 333

References

Lea, J. D. (2008). Experience gained in APT Database Design through the development of the
caltrans/UCPRC HVS database. In 3rd International Conference on Accelerated Pavement
Testing. Madrid, Spain.
Lea, J. D. (2012). Using point level accelerated pavement testing data for calibration of
performance models. In D. J. Jones, J. T. Harvey, A. Mateos, & I. Al-Qadi (Eds.), 4th
International conference on accelerated pavement testing: Advances in pavement design
through full-scale accelerated pavement testing (pp. 453–459). Davis, CA: CRC Press.
Madanat, S., Nakat, Z., Farshidi, F., Sathaye, N., & Harvey, J. T. (2008). Development of
empirical-mechanistic pavement performance models using data from the Washington
State PMS database. Davis & Berkeley, CA: University of California Pavement Research
Center.
NCHRP. (2004). Guide for mechanistic-empirical design of new and rehabilitated pavement
structures. Washington, DC: National Cooperative Highway Research Program (NCHRP).
Theyse, H. L., De Beer, M., & Rust, F. (1996). Overview of South African mechanistic pavement
design method. Transportation Research Record: Journal of the Transportation Research
Board, 1539, 6–17.
Ullidtz, P., Harvey, J. T., Basheer, I., Jones, D., Wu, R., Lea, J. D., et al. (2010). CalME, a
mechanistic-empirical program to analyze and design flexible pavement rehabilitation.
Transportation Research Record: Journal of the Transportation Research Board of the
National Academies, 2153, 143–152.
Ullidtz, P., Harvey, J. T., Tsai, B.-W., & Monismith, C. L. (2006). Calibration of
incremental-recursive flexible damage models in CalME using HVS experiments. Research
Report, University of California Pavement Research Center, Davis and Berkeley.
Key Concepts in Dynamic Signal
Processing from Instrumented
Pavement Sections

David H. Timm

Abstract Accelerated pavement test facilities have often featured dynamic pave-
ment response instrumentation, such as strain gauges, pressure plates and
multi-depth deflectometers since the 1960s. While the instrumentation and data
acquisition systems have rapidly evolved since then, researchers still face the
fundamental problem of converting raw dynamic signals gathered at high sampling
rates into meaningful information. These signals are inevitably infused with elec-
tronic noise that complicates extraction of useful information. Furthermore, the
resources to collect and store raw data often far outpaces the capabilities to process
and analyze the signals efficiently. The National Center for Asphalt Technology
(NCAT) Pavement Test Track has utilized dynamic strain gauges and earth pressure
cells since 2003 within structural pavement research studies and NCAT has also
been involved in live-traffic instrumented test sites in China, Oklahoma and New
Mexico. Each test site experienced the problems described above and also faced
unique challenges related to site-specific conditions. This paper provides
lessons-learned from the past 14 years of dynamic signal processing at each of
these test sites. A common set of practices is described and applied to two of the
sites that demonstrates the effectiveness and efficiency of the developed processing
schemes. Emphasis is placed on speed of data processing and visual inspection for
quality control.

Keywords Instrumentation Signal processing Pavement response measurement

1 Introduction

Instrumenting experimental sections at accelerated pavement test facilities has been

a common practice since the 1960s. Virtually every road test or accelerated
pavement test facility (APT), with the exception of the WesTrack experiment

D.H. Timm (&)

Auburn University, 238 Harbert Engineering Center, Auburn, AL 36849, USA
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 335

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_22
336 D.H. Timm

(Epps et al. 2002), has utilized embedded instrumentation to gain valuable

mechanistic-empirical (M-E) information. Sites such as the Minnesota Road
Research Project (Mn/ROAD), the Virginia Smart Road and the National Center for
Asphalt Technology (NCAT) Test Track are good examples of U.S. facilities
devoted to pavement research that have utilized extensive sets of instrumentation to
monitor in situ pavement responses under a variety of experimental conditions.
These responses often include strain and stress measurements under moving traffic
loads that require high frequency data acquisition systems. It is not uncommon to
sample the embedded sensors at frequencies ranging from 1000 to 2000 Hz.
Modern data acquisition systems are certainly capable of sampling at these high
rates, and many could easily exceed 2000 Hz. However, the problem does not lie
within data acquisition itself but rather in processing the raw data (voltage versus
time) into meaningful information that achieves the objectives of the particular
study. The challenges in data processing fall into three major categories that include
handling of large data sets, cleaning signals of electronic noise and extracting and
storing meaningful information from the raw signals. Each is briefly discussed in
the following paragraphs.
From the perspective of gauge sampling rates, the volume of data sampled can
easily overwhelm tools typically used for data analysis such as Excel. As an
example, consider a typical gauge array used at the NCAT Test Track of fourteen
gauges (12 strain and two pressure plates) embedded in an asphalt pavement sec-
tion. Sampling data from these fourteen gauges at 1000 Hz over a three second
interval to capture the passing of a single truck yields 42,000 discrete data points.
Considering that multiple truck passes are obtained during a data collection cycle
(typically 15 passes), the number of data points becomes 630,000 from a single test
section. Obviously, this volume of data requires data reduction tools other than
Excel which are better equipped to handle such large data sets. Many research
facilities have developed customized tools to handle these large data sets.
Mn/ROAD, for example, worked with the Department of Electrical and Computer
Engineering at the University of Minnesota to develop a stand-alone peak picking
program that could select baseline values and peak responses from dynamic
pavement response signals (Burnham et al. 2007). Similarly, researchers involved
with the Virginia Smart Road developed a customized program, SmartWave, to
view and process the dynamic strain traces (Al-Qadi et al. 2004).
In addition to the sheer volume of data, researchers must consider the inherent
noise in the electronic signals that are collected. As an example, Fig. 1a illustrates
the strain response from a typical truck pass at the NCAT Test Track. Though the
eight axle events are clearly identifiable by their clear vertical spikes, there is noise
evident throughout the signal that must be addressed. Figure 1b shows a zoomed-in
portion of the truck pass in Fig. 1a focused on just the steering axle where the noise
is much more evident. Strategies for “cleaning” this noise from the signal must be
considered and developed as part of an overall data processing scheme.
Key Concepts in Dynamic Signal Processing from Instrumented … 337

a) Entire truck pass.

b) Steer axle only.

Fig. 1 Raw strain gauge data (voltage vs. time)

After having devised a plan for handling large data sets and signal cleaning, one
must finally consider what points to extract from the signal and store for further
investigations. Figure 1a clearly shows a truck event captured by a strain gauge, but
deciding what to extract and tabulate is critical to achieving specific project
objectives. For example, simply the deviation from the baseline to a peak response
may be important for one investigation while the rate of strain increase may be
critical to another. Regardless of the study, decisions must be made regarding
extraction and storage of critical information.

2 Objectives and Scope of Work

The issues described above are not unique to any particular accelerated pavement
testing facility and are in fact shared by any pavement research study conducted
with embedded instrumentation. The objective of this paper is to provide practical
338 D.H. Timm

lessons learned from experience at the NCAT Test Track and other instrumented
projects pertaining to dynamic signal processing and analysis. To meet the this
objective, a general overview is first presented that discusses key concepts related to
data processing. Examples are then provided from the NCAT Test Track and a field
projects in Oklahoma.

3 General Principles

There are some common issues that must be considered when working with
dynamic pavement response data. Discussed in the following sub-sections, these
include:
• Developing a data collection scheme to meet the experimental objectives.
• Separating actual pavement response from electronic noise in the signal.
• Identifying important features of dynamic response.
• Developing an efficient user-friendly system for processing and compiling data.

3.1 Developing a Data Collection Scheme

As noted above, it is critical to match the data collection and processing scheme to
the experimental objectives. Thoughtful consideration should be given to planned
uses for the data before the experiment begins since the data needs are directly
related to the experimental objectives. Another important concept is that data col-
lection does not necessarily mandate data processing and analysis. It would be
possible, for example, to collect data continuously throughout an HVS experiment,
but only process and analyze axle passes at key points during the experiment. When
developing a scheme it is important to answer the question, “How rapidly will
conditions change during the experiment?” Under relatively stable conditions, the
amount of data needed will generally be less than when experimental factors are
more transient. Estimating the amount of data required can be based on experience,
or by examining some actual data and determining the stability of the experimental
conditions. Figure 2 illustrates two different strain data sets. Figure 2a comes from
a heavy vehicle simulator where the strain traces are nearly identical between the
five passes while Fig. 2b comes from two passes of different trucks NCAT Test
Track where wheel wander effects are much more evident. Clearly, the Test Track
data collection scheme will need more passes of each truck to accurately represent
strain since the data are much more variable.
Key Concepts in Dynamic Signal Processing from Instrumented … 339

a) HVS strain data.

b) NCAT Test Track strain data.

Fig. 2 Strain data comparisons between HVS and NCAT Test Track

3.2 Removing Noise from Signal

Electronic noise is a common problem that must be dealt with when examining
dynamic pavement response data. To minimize the amount of noise, efforts should
be made to minimize it at the power supply source, but it still can be a problem.
Electronic noise can be effectively eliminated by simply computing a moving
average of data points. Most data processing programs contain this as a standard
feature. It can also be done manually in Microsoft Excel, though it can be overly
time consuming for large data sets. Figure 3 illustrates the effect of a moving
average on a noisy strain signal. As seen in the figure, care must be taken so that the
moving average is not so large that the true engineering response is lost. In this
example, the raw signal represents 2000 Hz with a 10 point moving average pro-
viding a sufficiently clean signal. Wider moving averages would significantly alter
the true response measurement. If higher ratios are needed, then more complex
340 D.H. Timm

Fig. 3 Effect of moving average on raw strain signal

Raw Signal

Filtered Signal

Fig. 4 Effect of single pole analog low pass filter

methods of signal cleaning may need to be employed. For example, Fig. 4 illus-
trates the effect of a single pole analog low pass filter at 40 Hz applied to another
strain trace. Most signal processing programs will have some filtering algorithms
built into the program that may be used.
Key Concepts in Dynamic Signal Processing from Instrumented … 341

3.3 Identifying Important Features in the Dynamic

Response

There are many approaches to defining the “pavement response” from a dynamic
signal. Some researchers define strain as the peak response relative to a baseline
reading. Others define it by subtracting the minimum from the maximum response.
Regardless of the approach, from a data processing perspective, it is important to
define strain for a particular experiment so that the data processing scheme can be
tailored to acquire the necessary information from the response signal.
Consider the first part of the strain trace shown in Fig. 5 from a gauge at the
NCAT Test Track. From this relatively simple trace, strain could be defined in a
variety of ways including subtracting point 1 from point 3 or subtracting point 2
from point 3. Alternatively, there may be useful information that can be obtained
from the unloading portion of the curve (points 3–4 or 4–5). Since these would
generate drastically different strain values, it is critical for the researchers under the
guidance of experience and the experimental objectives to clearly define the strain
response. From a data processing perspective, the issue is to select and store the
appropriate points from which strain can be computed.
The importance of visually reviewing data cannot be overemphasized with
respect to identifying the key components of the dynamic response. Depending on
the placement and orientation of gauges, drastically different responses can be
observed. Therefore, careful review of the raw dynamic traces should be accom-
plished before beginning to develop a data processing scheme.

3.4 Developing a Data Processing Scheme

Once decisions have been made regarding the volume of data to collect, signal
filtering and the definition of pavement response, the data processing scheme can be

1 5

4
2

Fig. 5 Identifying important parts of the strain trace

342 D.H. Timm

developed. For the sake of efficiency, it is important to have a scheme that is as

automated as possible. For the sake of quality, it is important to have some visual
human interaction with the data. Effectively balancing these two components will
ensure a scheme that provides useful information in a timely manner.
To maximize automation, consistency must be maintained in the data collection
process. In a given experiment, maintaining consistent file naming and file for-
matting are critically important. Seemingly minor decisions such as whether a file is
tab- or comma-delimited can wreak havoc on the ability to automatically upload
data files into a processing template. It does not necessarily matter which format is
chosen, so long as that format is maintained throughout the experiment.
It is important, when developing a processing scheme, to carefully consider how
the data will be organized and compiled into a database. Identifying the important
factors in the experiment, as they relate to the experimental objectives, will help
define the database. These factors essentially become column headings within the
database and the processing scheme should be designed to automatically create data
for the database.
There are many proprietary software packages to choose from when building a
processing scheme. Programs such as EXCEL, LabView, MATLAB and DaDISP
are just a few examples of the numerous software packages available. Each package
has its own advantages and disadvantages in terms of handling large data files,
user-friendliness, built-in computational capabilities and cost. For example, EXCEL
is readily available and most engineers are very comfortable working in this
environment. However, there are file size limitations, there are not many built-in
signal processing features and computations can be somewhat slow compared to
other programs available. Deciding upon a particular package is a decision every
researcher must make and should be based on the factors mentioned above. It is
important to point out, however, that a particular processing scheme can be
implemented in most software packages.

4 Data Processing Schemes

Using the concepts presented above, data processing schemes were developed for a
variety of full-scale instrumented pavement sections. This paper presents two as
case studies (NCAT Test Track and I-35 in Oklahoma). For a facility such as the
Test Track, where the axle loads are applied by a small fleet of known vehicles, the
data processing schemes are relatively straightforward since they do not need to
count axles or classify vehicles. This information is entered as part of the data
processing scheme. This advantage exists for any APT facility where traffic is
known. Open access facilities, such as I-35 in Oklahoma, are more difficult since
the processing scheme needs to first determine axle numbers and spacing before
processing the data.
Key Concepts in Dynamic Signal Processing from Instrumented … 343

4.1 NCAT Test Track

Instrumented pavement sections have been a part of the NCAT Test Track since
2003. Over the past 12 years, the data processing schemes to handle the strain and
pressure data have evolved from plotting data in Excel to more sophisticated
automated schemes developed in the commercial software DaDISP. DaDISP is a
visual spreadsheet program that efficiently handles very large data sets in nearly
instantaneous fashion. It also utilizes a “series processing language” that enables
development of sophisticated algorithms for data processing purposes. This soft-
ware has been used since 2003 to develop a series of data processing templates
tailored to specific objectives of each Test Track research cycle. The following
description pertains to the most recently developed templates.
Figure 6 illustrates the typical gauge arrangement at the Test Track where there
are two earth pressure cells (EPC) and 12 asphalt strain gauges (ASG). Figure 7
shows the corresponding data processing template for a series of 15 truck passes at
the Test Track on a single section. It is organized into individual windows that
display one of the following: raw data (W1 through W14), user-entered data (W16),
processed data (W17 through W34) or tabulated data to export (W38) to a database.
The raw data windows are important for the individual processing the data (a.k.a.,
data processor) to quickly inspect the data to verify adequate signals were obtained
during data collection or identify potential problems with particular signals.
Decisions can then be made about which signals or truck events to process.
After inspecting the data for quality, the data processor enters some basic vehicle
information that informs the processing algorithms about axle counts and spacing
for each truck pass. Again, this is a luxury pertaining to closed-access facilities
where the axle or truck load events are always known. Once this information has

Fig. 6 NCAT Test Track typical gauge arrangement

344 D.H. Timm

Fig. 7 NCAT Test Track data processing template

Use steer axle peak

Clean all signals with low Find all steer axle peaks in
timestamps to compute
pass filter at 70 Hz EPC signals
vehicle speed (v=d/t)

Use computed vehicle speed Use computed vehicle speed

and known axle spacings to Use computed time windows and known spacing between
target time window where to find maximum responses first EPC and ASG to target
other axle peaks occur in in EPC signals time window where peak
EPC signals (t=v/d) strain responses occur (t=v/d)

Use computed time windows Find minimum responses

to find maximum responses between strain peaks to serve as Tabulate Data
in ASG signals inflection points

Fig. 8 NCAT Test Track data processing scheme

been entered, the data processor clicks a button to begin the automated data pro-
cessing algorithms depicted schematically in Fig. 8.
The first step in the data processing scheme is to clean each signal. After some
trial-and-error, it was found that single pole analog low pass filter with 70 Hz as a
cutoff frequency worked well for these gauges. Therefore, the filter is first applied and
all subsequent operations are performed on the cleaned signals. After signal cleaning,
the processing algorithms rely heavily on the known distances between gauges
(shown in Fig. 6), the known axle spacings (entered by the data processor) on each
vehicle and the ability to compute velocity from distance divided by time (v = d/t).
Key Concepts in Dynamic Signal Processing from Instrumented … 345

After signal cleaning, as shown in Fig. 8, the processing algorithms look for the
peak responses from the steer axles on the EPC gauges. The peak responses are
found by looking for a local maximum value above a pre-defined minimum voltage
threshold. Once these peaks, and their corresponding timestamps are found, the
speed of each vehicle is determined by dividing the physical distance between the
two EPC’s (12 ft) by the difference in time between the peaks.
The next step in Fig. 8 is to use the computed vehicle speeds to find the other
EPC responses from the non-steer axles using a targeted time window. The center
of this time window is found by dividing the known distance between axles by the
calculated speed of the vehicle which establishes the idealized time when the peak
should occur assuming no vehicle speed change. Because vehicle speed may
change during the duration of the truck event, a window of time is placed around
the idealized time to allow for some discrepancy between speed calculated from the
steer axle and the speed of the remainder of the vehicle during the entire truck
event. The maximum response is then found within that time window for each axle.
A similar approach is taken with determining peak strain responses. However, the
known distance between the first EPC and each ASG is used in the time compu-
tation (t = v/d) to identify the center of the time window. Again, the maximum is
found during the time window for each axle on each ASG. An additional step is
taken at this point to identify minimum values between peak strains in the ASG
signals which help to establish the maximum range of strain responses. The trucks,
axles, speeds, strain and pressure events are then tabulated in a spreadsheet that can
be exported to Excel or Access for storage and analysis.
Figures 9 and 10 illustrate the graphical output of the processing scheme
depicted in Fig. 8 for the EPC’s and one strain gauge, respectively, for a single

Fig. 9 Processed EPCs from NCAT scheme

346 D.H. Timm

Fig. 10 Processed ASG from NCAT scheme

truck pass. These are magnified plots corresponding to W17 and W22 from Fig. 7.
This visual representation allows the data processor to quickly inspect the data for
quality before exporting to a database. When problems are encountered, the data
processor may adjust a few settings to improve the capture of the pavement
responses. A well-trained data processor may successfully process a test section in
5–10 min. Given that it takes approximately 15–20 min to collect the data, the
processing time is considered efficient.

4.2 Oklahoma I-35 Project

The second case study was from a full-scale experiment conducted by the
Oklahoma Department of Transportation (ODOT) on I-35. This effort was led by
the University of Oklahoma with NCAT in a supporting role with respect to
pavement instrumentation (Solanki et al. 2009). The instrumentation scheme for
this site is depicted in Fig. 11 which is very similar to the gauge arrangement
presented in Fig. 6 with the addition of an EPC in the center of the array and axle
sensing strips upstream of the main gauge array. These sensing strips were added to
the array to enable axle counting, axle spacing computation, speed determination,
wheel offset computation and facilitate data processing.
The data processing scheme developed for the I-35 site included two sets of
DaDISP templates. The first, pictured in Fig. 12, allowed the data processor to
upload the data file and handled the main portion of the signal processing. The
second template, pictured in Fig. 13, allowed the processor to quickly inspect the
Key Concepts in Dynamic Signal Processing from Instrumented … 347

Fig. 11 ODOT I-35 instrumentation scheme

Fig. 12 I-35 main processing template

processed data for quality control purposes. These could have existed in the same
DaDISP template, but it was decided to separate them due to screen size limitations.
Figure 12 contains 5 visible windows, with many more hidden in the background.
The top left window (W1) provides the upload location of all the raw data, though
only the first strain gauge is shown. The bottom left window (W19) contains event
markers uploaded by the data processor. The I-35 system operates in a triggered
mode which is a common feature of many data acquisition systems. Every time the
system is triggered for a truck event, a time stamp is automatically stored and
tabulated which makes for easy entry into the template. These event markers are
used in the processing algorithms to subdivide the continuous data file, containing
many trucks, into individual truck events for signal processing. Once the event
348 D.H. Timm

Fig. 13 I-35 visual checking template

markers have been loaded, the data processor clicks a single button to begin the
automated processing subroutines which follow the basic logic layed out in Fig. 8.
The top right window (W20) contains the cleaned axle sensing strip signals and
automatically-obtained time stamps that are used for computing vehicle speed,
wheel offset, etc. The bottom two windows on the right hand side (W31 and W32)
contain the tabulated processed data. W31 may be copied into the visual check
template (Fig. 13) which rapidly displays all the data for a quality control check.
W32 contains the summary data which may be uploaded into an Excel spreadsheet
or Access database for storage and further analysis. Again the entire process takes
just a few minutes per data file and includes a visual check to ensure data quality.

5 Summary

Embedded instrumentation is often an important part of accelerated pavement

testing. Though results are often presented, very little has been documented about
processing of the dynamic pavement response signals themselves. This paper aimed
to present some basic principles and present schemes from two different instru-
mented test sites. Regardless of the facility or experimental objectives, the fol-
lowing important issues should be considered when developing a scheme. First, the
data collection and processing scheme should target the experimental objectives.
Second, signal noise must be considered and dealt with during the signal processing
phase. Third, important features of the dynamic response should be identified for
storage during signal processing. Fourth, an efficient and user-friendly system
should be developed for processing and compiling the data. There are many options
Key Concepts in Dynamic Signal Processing from Instrumented … 349

for successfully addressing these issues that range from development of stand-alone
software to utilizing commercially-available software customized to the particular
facility. The examples from the NCAT Test Track and the I-35 site in Oklahoma
used the DaDISP software to meet the needs of their respective experiments.

References

Al-Qadi, I. L., Loulizi, A., Elseifi, M., & Lahouar, S. (2004). The Virginia Smart Road: The
impact of pavement instrumentation on understanding pavement performance. The Journal of
Association of Asphalt Paving Technologists, 73, 427–465.
Burnham, T. R., Tewfik, A., & Srirangarajan, S. (2007). Development of a computer program for
selecting peak dynamic sensor responses from pavement testing. MN/RD-2007-49, Minnesota
Department of Transportation. http://www.lrrb.org/PDF/200749.pdf.
Epps, J. A., Hand, A., Seeds, S., Schulz, T., Alavi, S., Ashmore, C., Monismith, C. L., et al.
(2002). Recommended performance-related specification for hot-mix asphalt construction:
Results of the Westrack project. NCHRP Report 455, Transportation Research Board, National
Research Council.
Solanki, P., Zaman, M., & Muraleetharan, K. K. (2009, August). Field performance monitoring
and modeling of instrumented pavement on I-35 in McClain County—Construction and
Instrumentation Report. ODOT Item No. 2200.
Protocols for Accelerated Pavement
Testing of Fully Permeable Pavements

D. Jones, R. Wu and H. Li

Abstract Fully permeable pavements are used as an alternative means of

managing stormwater on roads, parking lots, and other paved areas. In the past, they
have primarily been used only in areas with very light traffic, and designs have been
empirical. With road agencies coming under increasingly stricter stormwater
management requirements, there is growing interest in using these types of pave-
ments under heavier traffic loads. A study was therefore initiated by the Interlocking
Concrete Pavement Institute (ICPI) and the UCPRC to better understand the
behavior of fully permeable pavements and to formalize the design process using a
more mechanistic type approach. The study was undertaken in phases and included
laboratory testing, modeling, accelerated load testing, development of design pro-
cedures, life-cycle cost assessment, and environmental life cycle assessment. Most
fully permeable pavements include a subbase/reservoir layer of large (75 mm)
railway ballast aggregate, to support the load under wet conditions, and a base of
smaller (25 mm) open graded aggregate as an intermediate layer between the large
subbase aggregate and the surfacing. The use of this large open-graded aggregate
limit the use of the testing protocols, instrumentation (e.g., multi-depth deflec-
tometers), and data analysis procedures traditionally used in accelerated loading
tests. Alternative testing protocols ad analysis procedures were therefore developed
for the UCPRC study. This paper summarizes the design procedures and protocols
developed for accelerated pavement testing of fully permeable pavements and
discusses how the results were used to develop a mechanistic empirical design
procedure to optimize layer thicknesses.

D. Jones (&)
R. Wu
H. Li
Department of Civil and Environmental Engineering, University of California
Pavement Research Center, UC Davis, Davis, CA 95616, USA
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 351

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_23
352 D. Jones et al.

1 Introduction

Fully permeable pavements (i.e., where the water flows through the pavement
structure and into the underlying subgrade) is being increasingly used on parking
lots, alleys and low-volume streets, with demand emerging from national, state and
municipal laws to decrease stormwater runoff and pollution. There is also growing
interest in their use on higher traffic volume pavements. To address this interest,
The Interlocking Concrete Pavement Institute (ICPI) developed a structural design
procedure for permeable interlocking concrete pavement (PICP) (Smith 2011),
based on the flexible pavement design method from the American Association of
State Highway and Transportation Officials 1993 Guide for Design of Pavement
Structures (AASHTO 1993). While conservative input values are used, the appli-
cation of the AASHTO-based design method is theoretical with no genuine
empirical observations from PICP to develop equations that calculate layer thick-
nesses. Like other permeable pavement design methods, PICP requires full-scale
load testing to validate design methods and to provide designers and road agencies
with increased confidence in the use of PICP performance. A study was therefore
initiated by the ICPI and the University of California Pavement Research Center
(UCPRC) to better understand the behavior of fully permeable pavements and to
formalize the design process using a more mechanistic type approach. The study
was undertaken in phases and included laboratory testing, modeling, accelerated
load testing, development of design procedures, life-cycle cost assessment, and
environmental life cycle assessment (Li et al. 2014).
Most fully permeable pavements include a subbase/reservoir layer of large
(75 mm) railway-type ballast aggregate to support the load under wet conditions,
and a base of smaller (25 mm) open graded aggregate as an intermediate layer
between the large subbase aggregate and the surfacing. The use of these large
open-graded aggregates requires different design procedures to those used for
designing conventional pavements, and limits the use of the testing protocols and
instrumentation (e.g., multi-depth deflectometers) traditionally used for measuring
permanent deformation in individual pavement layers during accelerated loading
tests. Alternative testing protocols were therefore developed for this UCPRC study.

2 Research Objectives and Tasks

The objective of this research was to produce design tables for permeable inter-
locking concrete pavement (PICP) based on mechanistic analysis and partially
validated with accelerated pavement testing (APT). The tasks to complete this
objective included a literature review, field deflection testing of existing projects
and test sections, estimation of the effective stiffness of each layer in PICP struc-
tures, mechanistic analysis and structural design of a test track incorporating three
different subbase thicknesses (low, medium, and higher risk), tests on the track with
Protocols for Accelerated Pavement Testing of Fully Permeable … 353

a heavy vehicle simulator (HVS) to collect performance data to validate the design
approach using accelerated loading, refinement and calibration of the design pro-
cedure using the test track data, development of a spreadsheet based mechanistic
design tool, and development of revised design tables using the design tool. This
paper summarizes the design procedures and protocols developed for accelerated
pavement testing of fully permeable pavements and discusses how the results were
used to develop a mechanistic empirical design procedure to optimize layer
thicknesses.

3 Literature Survey

A literature review (Jones et al. 2013a) found that only a few organizations
worldwide have undertaken detailed research on the design of permeable pave-
ments, with many studies focusing on infiltration on low volume traffic roads, rather
than structural design of roads carrying truck traffic. Limited published record was
found on mechanistic design approaches and on controlled load testing on per-
meable pavements in general and PICP in particular. No published information was
found on instrumentation or the measurement of permanent deformation in indi-
vidual layers in a permeable pavement.

4 Test Track Design

4.1 Design Criteria and Variables

The test track for accelerated load testing on PICP was located at the UCPRC in
Davis California. The design was derived from a sensitivity analysis that considered
a range of mechanistic values from worst-case to best-case scenarios. Selection of
the input values was based on previous work by the authors (Jones et al. 2013a, b;
Harvey et al. 2010), work by others on the topic identified during literature reviews
(Tutumluer and Seyhan 1999; Thompson et al. 2002; Tutumluer et al. 2004; Kim
and Tutumluer 2006; Li et al. 2010; Chow and Tutumluer 2014; Das 2007; Nova
2012; Carter and Bentley 1991; Wnek et al. 2013; Huang et al. 2009), and the
results of the deflection testing study (Jones et al. 2013b).
The most likely failure mode of PICP is permanent deformation in the base,
subbase, and/or subgrade layers, which will manifest as rutting and/or paver dis-
placement on the surface. The design criteria for the test track were therefore
focused on this type of distress.
Rut development rate as a function of the shear stress to shear strength to ratios
at the top of the subbase and the top of the subgrade was used as the basis for the
design approach. This approach was selected based on a review of the literature,
354 D. Jones et al.

past research on permeable pavements by the authors, and the results of deflection
testing on in-service permeable interlocking concrete pavements. The shear
stress/strength ratio was originally developed for airfield pavements where the shear
stresses from aircraft loads and tire pressures are high relative to the strengths of the
subgrade materials (Thompson et al. 2002; Tutumluer et al. 2004). On permeable
road pavements, subgrade materials are often uncompacted or only lightly com-
pacted and wet or saturated for much of the service life, resulting in relatively low
shear strengths compared with the high shear stresses from trucks. Deeper ruts are
usually also tolerated on permeable pavements due to the absence of ponding on the
surface during rainfall. The alternative approach of using a vertical strain criterion
was considered inappropriate for permeable pavements, given that this is typically
used where the shear stresses relative to the shear strains are relatively low, which
typically results in low overall rutting.
Shear stress/strength ratio is defined as the ratio between the applied shear stress
(τf) and the material shear strength [τmax (τmax = c + σf tan ϕ in a triaxial strength
test, where c is the cohesion of the material)] on the failure plane at a specific
applied normal and confining stress state (Thompson et al. 2002). The normal and
shear stresses (σf and τmax) acting on a failure plane (oriented at an angle of 45° + ϕ/
2, where ϕ is the internal friction angle of the material) can be calculated according
to the Mohr-Coulomb failure theory for specific confining (σ3) and deviator (σd)
stresses applied to a laboratory specimen during triaxial testing.
Materials with lower shear stress/strength ratios are less likely to fail due to shear
(i.e., rutting and permanent deformation) than materials with higher shear
stress/strength ratios. Research studies (Kim and Tutumluer 2006; Li et al. 2010)
have shown that materials subjected to shear stress/strength ratios higher than 0.7
are likely to accumulate high permanent deformation and present a higher rutting
risk, leading to rapid shear failure in the pavement. Materials with shear
stress/strength ratios between 0.3 and 0.7 represent a medium risk with a steady but
reasonable rate of rutting, while those with shear stress/strength ratios less than
about 0.3 are expected to have little or no rutting after an initial small “bedding-in”
rut. Based on these findings, the following three shear stress ratio design variable
categories aligned to the level of rutting risk were defined for permeable inter-
locking concrete pavements (Jones et al. 2013a; Tutumluer et al. 2004):
SSR < 0.3, low risk of rutting;
0.3 ≤ SSR ≤ 0.7, medium risk of rutting;
SSR > 0.7, high risk of rutting.
The equations used to calculate the SSR corresponding to the stress state applied
during triaxial testing or other conditions are listed below:

Shear Stress=Strength Ratio ðSSRÞ ¼ sf =smax ð1Þ

sf ¼ ððr1  r3 Þ=2Þ cos / ¼ rd =2 cos / ð2Þ

Protocols for Accelerated Pavement Testing of Fully Permeable … 355

T max ¼ c þ rf tan / ð3Þ

rf ¼ ððr1 þ r3 Þ=2Þ  ððr1  r3 Þ=2Þ sin /

ð4Þ
¼ ððrd þ 2r3 Þ=2Þ  ðrd =2Þ sin /

where
τmax is applied shear stress acting on the failure plane oriented at an angle of
45° + ϕ/2;
σf is applied normal stress acting on the failure plane oriented at an angle of
45° + ϕ/2;
τf is the shear strength of the material under a certain stress state;
σ1 and σ3 are the major and minor principal stresses, respectively;
σd is the deviator stress, σd = σ1 − σ3;
c is the cohesion of the material;
ϕ is the internal friction angle of the material (ϕ = 0 for stress-independent
materials).
In mechanistic analyses, the major and minor principal stresses (σ1 and σ3) on
top of the base and subgrade layers are the critical responses required for calculating
the shear stress/strength ratio for designing permeable interlocking concrete pave-
ments. These stresses can be calculated using multilayer linear elastic theory. In this
study, the OpenPave software program (Lea undated) was used for these analyses.

4.2 Pavement Structure

A standard permeable interlocking concrete pavement structure with the following

layers was used in the mechanistic analysis:
• Surface (interlocking concrete paver, 80 mm thick).
• Bedding layer (ASTM #8 aggregate, 50 mm thick).
• Base layer (ASTM #57 aggregate, 100 mm thick).
• Subbase layer (ASTM #2 aggregate, with varying thickness).
• Subgrade soil.
The bedding and base layers provide intermediate levelling layers between the
coarse subbase aggregate and the concrete pavers. Rather than applying marginally
different shear strengths to each layer, all of the aggregate layers (bedding, base,
and subbase) were integrated into one nominal aggregate base (AB) layer and
assumed to have similar strength properties (a range of strength properties was
used). Nine different thicknesses of this nominal aggregate base (AB) layer, ranging
from 300 to 1500 mm, were used in the mechanistic analysis (Li et al. 2014).
356 D. Jones et al.

4.3 Materials Properties

The material properties used in the mechanistic analysis included stiffness and
Poisson’s ratio for each layer in the pavement structure, and cohesion and internal
friction angle of the composite base aggregate and subgrade soil materials. These
properties were selected from the deflection testing analyses and from the results of
other studies documented in the literature (Li et al. 2014). No laboratory testing to
measure actual material properties was undertaken in this study.
Four different stiffnesses were selected for each layer (surface, base and sub-
grade) based on the backcalculated effective stiffnesses from the deflection testing
analyses (Li et al. 2014):
• Surface (pavers): 200, 500, 1000 and 2000 MPa.
• Base (combined base, and subbase layers): 60, 90, 120 and 180 MPa.
• Subgrade: 20, 50, 100 and 150 MPa.
The Poisson’s ratio for each layer was assumed to be 0.35 based on measure-
ments documented in other studies (Kim and Tutumluer 2006; Chow and
Tutumluer 2014).
The cohesion and internal friction angle (c, ϕ) of the aggregate base material was
assumed to be 0 kPa and 45°, respectively, based on a review of the literature (Kim
and Tutumluer 2006; Chow and Tutumluer 2014). For the subgrade material, both
non-zero (ϕ ≠ 0) and zero (ϕ = 0) internal friction angles were used in the analysis
for all stiffness levels to simulate drained and soaked, undrained soil conditions,
respectively. Based on a review of the literature (Das 2007; Nova 2012; Carter and
Bentley 1991; Wnek et al. 2013; Huang et al. 2009), the subgrade cohesion and
internal friction angles (c, ϕ) were set at the following levels for each of the four
subgrade stiffnesses (Li et al. 2014):
• 20 MPa: 10 kPa and 20° and 0°.
• 50 MPa: 15 kPa and 25° and 0°.
• 100 MPa: 20 kPa and 30° and 0°.
• 150 MPa: 25 kPa and 35° and 0°.

4.4 Traffic Load

A single rear axle with dual wheels was used in the analysis. The axle load was set
at 89 kN and the tire pressure was set at 700 kPa (tire pressure used in HVS tests).
The distance between the two tire centers was set at 340 mm. The stress under the
wheel and the stress between the wheels were both calculated to identify the most
critical stress (Li et al. 2014).
Protocols for Accelerated Pavement Testing of Fully Permeable … 357

4.5 Mechanistic Analysis Results

The results of the mechanistic analysis for the different base layer (combined
bedding, base, and subbase layers) stiffness values and thicknesses include the
major and minor principal stresses, normal stress at the failure plane, shear strength
at the selected stress state, shear stress at the failure plane, and the shear
stress/strength ratio at the failure plane at the top of the combined base layer. The
results indicated that, according to the multilayer linear elastic design theory, an
increase in the thickness of the combined base and subbase layer would not nec-
essarily reduce the stresses at the top of that layer, as expected (Li et al. 2014).
The results of the mechanistic analysis for the subgrade included the same
parameters used in the combined base and subbase layer analysis, except that the
shear stress/strength ratio at the top of the subgrade was calculated. The results
indicated that increasing the thickness of the subbase layer reduces the stresses
(absolute values) at the top of the subgrade soil layer, as expected (Li et al. 2014).
During dry conditions, when the subgrade is relatively dry (or at equilibrium
moisture content) and has a nonzero internal friction angle (ϕ ≠ 0), the shear
strength of the subgrade soil changes with the thickness of the combined base and
subbase layer. Interestingly, the effective shear strength of the subgrade soil
decreases slightly as the thickness of the base/subbase layer increases. This is
attributed to the effective shear strength of subgrade soils being positively corre-
lated with the normal stress at the failure plane under dry conditions (ϕ ≠ 0), which
provides confinement. An increase in the thickness of the base/subbase layer sig-
nificantly reduces the normal stress at the failure plane at the top of the subgrade
soil layer, and consequently, the effective shear strength of the subgrade soil
decreases slightly as the thickness of the base/subbase layer increases.
Under wet conditions [i.e., when the subgrade is soaked and has a zero internal
friction angle (ϕ = 0)], the effective shear strength of subgrade soils does not
change with the thickness of the base/subbase layer. This is because the shear
strength of materials with zero internal friction angle is independent of the normal
stress applied and is determined only by the cohesion of the material. Therefore,
soaked subgrade soils will have constant effective shear strength regardless of an
increase in the thickness of the base/subbase layer. The effective shear strength will
be equal to the cohesion of the material which is slightly lower than the effective
shear strength of the subgrade soil under dry conditions.
The normal stress and the shear stress are both higher under wet conditions than
those under dry conditions for an identical structure and identical material prop-
erties). The shear stress/strength ratio under wet conditions is also higher than the
shear stress/strength ratio under dry conditions for an identical structure and
identical material properties, as expected. This confirms that wet conditions are the
most critical condition influencing rutting and permanent deformation in the sub-
grade in permeable pavements.
Based on the mechanistic analysis results, the base/subbase layer thicknesses
with shear stress/strength ratios of 0.8 (i.e., >0.7), 0.5 [i.e., intermediate between 0.3
358 D. Jones et al.

and 0.7 (0.3 ≤ SSR ≤ 0.7)], and 0.2 (i.e., <0.3), representing different rutting risk
levels, were estimated using interpolation of the different material properties and
subgrade moisture conditions. The main observations from the analysis with regard
to required base/subbase layer thicknesses for PICP include the following:
• Higher shear stress/strength ratios, which equate to a higher risk of rutting,
require thicker base/subbase layers, as expected.
• For the same shear stress/strength ratio, an increase in the effective stiffness of
the base/subbase layer reduces the required thickness of that layer, especially
when the subgrade has a low stiffness.
• An increase in the stiffness of the surface layer reduces the required
base/subbase layer thickness to achieve the same shear stress/strength ratio.
However, the effect of the surface layer stiffness is not significant due to the
relatively low thickness of the pavers (80 mm) and the reduced interlock
between them compared to pavers with sand joints.
• For the same shear stress/strength ratio, wet conditions require thicker
base/subbase layers compared to the dry condition, confirming that undrained
wet conditions are the most critical condition for design.
• The theoretical optimal design base thicknesses (combined bedding, base, and
subbase layers) for low, intermediate, and higher risk levels (subgrade shear
stress/strength ratios of 0.2, 0.5 and 0.8, respectively) under dry subgrade
moisture conditions are approximately 1300, 800 and 500 mm, respectively. In
wet conditions, the theoretical optimal design thicknesses increase to 1400,
1000 and 600 mm, respectively.

4.6 Test Track Layer Thicknesses

Based on the results of the mechanistic analysis, three subbase thicknesses of 450,
650 and 950 mm, were selected for the HVS test track design to provide high,
intermediate, and low risk scenarios (Fig. 1). The bedding layer and base layer
thicknesses were fixed at 50 and 100 mm, respectively, equating to total structure
thicknesses of 600, 800 and 1100 mm for the three subsections. These subbase
layer thicknesses are mostly thinner than the theoretical optimal design thicknesses
discussed above and were selected to ensure that the performance and behavior of
the test track structure could be fully understood within the time and budgetary
constraints of the project.
The design approach described in the ICPI PICP guide (Smith 2011) was fol-
lowed to determine a benchmark design for the test track that could be used to
compare with the results from the mechanistic design. Using the pavement structure
detailed above, and designing for a subgrade soaked California Bearing Ratio
(CBR) of 4 % (determined from DCP tests) and lifetime equivalent standard axle
loads (ESALs) of 1,000,000 (expected traffic loading with the HVS), a subbase
thickness of 675 mm under bedding and base layers of 50 and 100 mm,
Protocols for Accelerated Pavement Testing of Fully Permeable … 359

y Top View
N
15 m

0.2 m
1m
0.2 m

x
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
Station Number (unit: 0.5m)
Cross Section
Thermocouples (0 and 25mm) Thermocouple 80mm Paver
50mm Bedding Layer
100mm Base
450mm 650mm 950mm

Subbase

Subgrade
Legend
Permanent deformation Road surface deflectometer Pressure cell Thermocouple

Surface profile measured at every station between Station 3 and Station 27 NOT TO SCALE

Fig. 1 Pavement structure and layout for PICP test track (not to scale)

respectively, would be required. This was close to the calculated intermediate risk
thickness of 650 mm.
The track dimensions were 30 m long by 8 m wide, long enough to accom-
modate the HVS and wide enough for three side-by-side test sections to allow
testing under dry, wet, and drained conditions.

5 HVS Testing Protocols

The test focused on assessing permanent deformation under dry and wet conditions.
Water levels were measured in perforated pipes (75 mm diameter) installed on the
east and west sides of the test track on each subsection. The bottom of the pipe
rested on the subgrade and the top of the pipe extended above the surface.
Submersible pressure transmitters were used to measure hydraulic pressure at
30 min intervals, with water level calculated from the hydraulic pressure. Dipstick
measurements were also taken in the pipes to verify the pressure transmitter
measurements and to check consistency of the water level across the test track.
All testing was carried out at ambient temperatures and the HVS was not
enclosed in an environmental chamber. Ambient and pavement temperatures were
measured with Type-K thermocouples.
360 D. Jones et al.

A laser profilometer was used to the following surface profile parameters. The
analysis software was modified to account for the gaps between the pavers.
• Maximum total rut depth at each station (see Fig. 1).
• Average maximum total rut depth for all stations.
• Average deformation for all stations.
• Location and magnitude of the maximum rut depth for the section.
• Rate of rut development over the duration of the test.
Measuring permanent deformation in the underlying layers was more of a
challenge. In standard pavement structures, multi-depth deflectometers (MDD) or
stand-alone linear variable differential transformers (LVDT) are commonly used for
this activity. However, these are not appropriate for large dimension, open-graded
aggregates since the instruments cannot be appropriately anchored. Consequently, a
simple dipstick-type system was designed to measure vertical permanent defor-
mation between the pavement surface and the top of the subbase, and the pavement
surface and the top of the subgrade. The gauges were custom fabricated for the
experiment. The gauge consists of a 100 mm square stainless steel target plate that
is positioned on a thin layer of bedding sand on the top of the selected layer
(Fig. 2). A 25 mm sleeve is welded to the plate. A stainless steel pipe, cut to a
height equivalent to 25 mm below the top of the paver is inserted into the sleeve.
The top of this pipe slots into a hole drilled into the paver. A stainless steel rod is
inserted into the pipe to measure permanent deformation on a simple measuring jig
(Fig. 3). Gauges were installed in each subsection of the dry and wet test sections.
Surface deflection was measured with a road surface deflectometer (RSD).
Measurements were taken under a creep-speed 40 kN half-axle load at regular
intervals.
Pressure cells were installed in each subsection of the dry and wet test sections to
measure vertical pressure (stress) under the moving wheel. One cell was installed
level with the surface of the subgrade in each subsection after excavation and
compaction. Another cell was installed on the top of the subbase. Bedding sand for
the cells was placed on top of a square of geotextile and used as a leveling course for
the instrument to limit any movement on top of the uneven coarse aggregate subbase.

Fig. 2 Permanent deformation gauge

Protocols for Accelerated Pavement Testing of Fully Permeable … 361

Fig. 3 Permanent deformation measurements

Jointing stone depth relative to the surface was measured with a depth gauge.
The stone was replenished if the measured depth was below 25 mm (stone
replenishment was only required on the wet test).

6 HVS Testing Program

A summary of the HVS loading program and loads applied on the three test sections
is shown in Table 1. All trafficking was carried out with a dual-wheel configuration,
using radial truck tires (Goodyear G159—11R22.5—steel belt radial) inflated to a
pressure of 700 kPa, in a bidirectional loading mode with a 1.0 m wide wander
pattern (i.e., trafficking in both directions in line with standard procedures for
testing base layer performance). Load was checked with a portable weigh-in-motion
pad at the beginning of each test.

Table 1 Summary of HVS Test Half-axle wheel Repetitions ESALsa

loading program load (kN)
Wet 25 100,000 15,259
40 100,000 100,000
60 140,000 708,750
Section total 340,000 824,009
Dry 25 100,000 15,259
40 100,000 100,000
60 140,000 708,750
80 40,000 640,000
Section total 380,000 1,464,009
Drained 25 100,000 15,259
40 25,000 25,000
Section total 125,000 40,259
Project total 845,000 2,328,277
a
ESAL Equivalent single axle load = ((axle load/40)4.0) × load
repetitions
362 D. Jones et al.

7 Key Observations from HVS Testing

Key observations from the HVS testing include the following (Li et al. 2014):
• There was a significant difference in rutting performance and rutting behavior
between the wet and dry tests, as expected.
• A large proportion of the rutting on all three sections occurred as initial
embedment in the first 2000–5000 load repetitions of the test and again after
each of the load changes, implying that much of the rutting in the base and
subbase layers was attributed to bedding in, densification, and/or reorientation of
the large aggregate particles in the subbase. Although, this behavior is consistent
with rutting behavior on other types of structures, better compaction of the base
and subbase in the test track may have limited the extent.
• During testing under dry conditions, limited permanent deformation (<4 mm)
was recorded in the bedding and base layers on all three subsections, and most
occurred very early in the test. On the subsection with the 450 mm subbase,
rutting occurred in both the subbase (10 mm rut) and subgrade (13 mm rut). On
the 650 and 950 mm subbase subsections, rutting occurred mostly in the sub-
base. Total permanent deformation on the 450, 650 and 950 mm subbase
subsections was 27, 23 and 17 mm, respectively, implying a generally linear
trend of increasing permanent deformation with decreasing subbase thickness.
• During testing under wet conditions, rutting in the bedding and base layers was
dependent on the thickness of the subbase (9, 5 and 2 mm on the 450, 650 and
950 mm subbase subsections, respectively). Rutting occurred in both the sub-
base and the subgrade on all subsections, with rutting in the subbase consistent
across all three sections (*25 mm). Rutting in the subgrade differed between
sections relative to subbase thickness, with 15, 6 and 4 mm of rut recorded on
the 450, 650 and 950 mm subbase subsections, respectively. The number of
load repetitions and equivalent standard axles required to reach the terminal rut
depth (25 mm) set for the project is summarized in Table 2. The sensitivity of
the pavement structure to water (i.e., standing water in the subbase) and to load
is clearly evident.
• Although only limited testing was undertaken under drained conditions (i.e., no
water in the subbase), rutting behavior appeared to show similar trends and
behavior to the test under dry conditions.

Table 2 Repetitions and ESALs required to reach terminal rut

Test 450 mm 650 mm 950 mm
Load repetitions at terminal rut (25 mm)
Dry 340,000 Rut < 25 mm Rut < 25 mm
Wet 95,259 180,000 210,000
ESALs at terminal rut (25 mm)
Dry 824,009 Rut < 25 mm Rut < 25 mm
Wet 165,884 220,000 216,519
Protocols for Accelerated Pavement Testing of Fully Permeable … 363

• The thickness of the subbase influenced rut depth in the subgrade, as expected,
but did not influence the rutting behavior in the subbase itself. Rutting in this
layer therefore appears to be governed by the aggregate properties, and con-
struction methods and quality.
• The rate of rut depth increase escalated with increasing load, indicating that the
pavement structure was load sensitive, especially at load levels close to and
above the legal design load.
• Deflection during dry testing was dependent on subbase thickness and it
increased with increasing load. Deflections were relatively high compared to
more traditional pavements with dense graded layers. Deflection during wet
testing was higher compared to that recorded during dry testing, with deflection
on the 450 mm subbase subsection significantly higher compared to that
recorded on the 650 and 950 mm subbase subsections, indicating a
load-sensitive, weaker overall structure as a result of the wet subgrade.
• No distress was noted on any individual pavers and no pavers were dislodged
from the pavement during testing.
• The measured infiltration rate of water through the joints between the pavers
reduced marginally over the course of HVS testing; however, visually it was still
considered to be both rapid and effective.

8 Data Analysis

8.1 Rutting Models

The design criteria, design variables, and critical responses used in the test track
design and discussed above were used in the mechanistic analysis. However, in the
initial design, the bedding, base, and subbase layers were combined into a single
aggregate base layer for purposes of the design. The results of HVS testing indi-
cated that rutting behavior in the bedding and base layers differed from that in the
subbase layer during the wet test, and consequently, the bedding and base layers
were combined into a single layer and analyzed separately from the subbase layer in
this stage of the analysis. Based on the observations made and measurements taken
during HVS testing, it was also concluded that no deformation or distress occurred
in the pavers, only in the underlying layers, which resulted in measurable defor-
mation on the surface. Consequently, although surface rutting is the primary cri-
terion designed for and predicted in the analysis, it is assumed that the pavers
themselves would not deform. It should also be noted that, although individual
pavers have a relatively high stiffness, a permeable surface layer constructed of
pavers does not, due to the wider spacing and reduced interlock between the pavers
compared to pavers with sand joints.
During testing under dry conditions, a rapid embedment of about 4 mm was
recorded in the combined bedding and base layer on the three subsections in the
364 D. Jones et al.

Table 3 Summary of rut models

Layer Rut modela Moisture Model parameters
condition a b c
Combined bedding RDBB = a × h_SB + b Dry 0 4.0 –
and base Wet −0.012 13.1 –
Subbase RDSB = (a × SSRb) × Nc Dry 3.10E−06 2.70 1
Wet 3.10E−06 2.70 1
Subgrade (silty RDSG = (a × SSR + b) × Nc Dry 0.03 −0.01 0.5
clay) Wet 0.03 −0.01 0.5
a
RDxx rut depth of xx layer [BB surface (paver, bedding and base); SB subbase; SG subgrade]
(mm); h_SB thickness of subbase (mm); SSR shear stress/strength ratio at the top of the layer;
N load repetition; a, b, c, model constants

first 2000 load repetitions. Thereafter, no further significant rutting was recorded for
the remainder of the test on any of the subsections. The rut depth model of this
combined layer under dry conditions was therefore set as a constant (4 mm). Under
wet conditions, the rut depth in the combined bedding and base layers varied
depending on the subgrade thickness. On the 650 and 950 mm subbase subsections,
most of the rutting occurred during early embedment after which it remained rel-
atively constant. On the 450 mm subbase subsection, rutting was influenced by
changes in the load. The rut depth model of this combined layer for wet conditions
was therefore set as a linear function of the subbase thickness (see Table 3).
A typical mechanistic approach using the general formulas in Eq. (5) was used
to develop the rutting model for the subbase layer (Li et al. 2014).

RDSB ¼ aN b and RDSG ¼ aN b ð5Þ

where
RDSB and RDSG are the rut depths in the subbase and subgrade layers, respectively
N is the number of load repetitions,
a and b are constants and are a function of the shear stress/strength ratio (SSRSB and
SSRSG) at the top of the subbase and subgrade layers.
A two-step model development process was followed (parameters for the sub-
base model are provided):
• Step 1. Fit RDSB = a(dN + N0)b for each testing case with different subbase
thickness (h_SB), test load (L), and test moisture condition (Dry and Wet),
considering the effect of early embedment in the initial stages of trafficking.
RDSB is the total rut depth in the subbase for a load level i; dN is the incremental
repetition under that load level i; N0 is a model constant for considering the
effect of earlier loading.
• Step 2. Fit a * f(SSRSB) and b * f(SSRSB) for all testing cases.
Protocols for Accelerated Pavement Testing of Fully Permeable … 365

Using the rut test data from the HVS testing, it was found that the subbase rut
depth showed an approximately linear relationship with load repetitions after early
initial embedment for all testing cases. Consequently, the power constant b was set
as 1. The constant a is a function of SSRSB and SSRSG, calculated for each case
using Eqs. (1)–(4). Rut depth in the subgrade had a power relationship of
approximately 0.5 with load repetitions after early embedment for all testing cases.
Consequently, the power constant c was set as 0.5. Models are summarized in
Table 3.
Given that rutting in the subbase cannot be prevented simply by increasing the
thickness of this layer, only rutting in the subgrade was used in the development of
the design tables. Changing the rutting behavior of the subbase would require
tighter specifications for the properties of the materials used in this layer, and/or the
construction methods (e.g., higher relative compaction/lower air void content).
Consideration for these alternatives would require laboratory and additional
accelerated pavement testing to quantify the effects of the changes, which was
beyond the scope of this study.

8.2 Input Parameters for M-E Design

The default input parameters for mechanistic-empirical design of PICP were revised
from those used in the test track design phase discussed earlier based on the HVS
test results. Three separate layers were modelled (bedding plus base, subbase, and
subgrade). The same material properties used in the earlier analysis were used;
however, the values were adjusted to match the materials and layer thicknesses used
in the test track. The following default stiffnesses were selected for each layer under
both wet and dry conditions based on the backcalculated effective stiffnesses from
the deflection data collected during HVS testing:
• Combined surface layer: Dry = 110 MPa, Wet = 87 MPa.
• Subbase: Dry = 122 MPa, Wet = 73 MPa.
• Subgrade: Dry = 60 MPa, Wet = 37 MPa.
The same Poisson’s ratio used in the original analysis (0.35) was used for this
phase of the analysis. The default cohesion (c) of the subbase material remained the
same at 0 kPa under both dry and wet conditions. However, in this phase of the
analysis, different default internal friction angles (ϕ) were assumed for dry and wet
conditions. Selected values were 45° under dry conditions and 35° under wet
conditions, based on differences in rutting performance in this layer in the dry and
wet tests and a review of the literature (Kim and Tutumluer 2006; Chow and
Tutumluer 2014).
The default cohesion and internal friction angles of the subgrade were set at
15 kPa and 25° for dry conditions, and 9 kPa and 15° for wet conditions.
366 D. Jones et al.

Based on the HVS test results, the worse-case design condition for PICP is when
the subbase contains standing water (i.e., the subbase is serving as a reservoir for
collected rainwater while that water infiltrates the subgrade or drains through a
subsurface drainage system). The number of wet days when the subbase contains
standing water is required to distribute the traffic between dry and wet periods. The
default number of wet days was set at 50 (Li et al. 2014).
Traffic input was expanded from that used in the original analysis (Li et al.
2014). An axle load spectrum for single and tandem axles with dual wheels was
used to characterize traffic. Twelve single axle loads ranging between 10 and
160 kN and ten tandem axle loads ranging between 20 and 200 kN were consid-
ered. Tire pressure was set at 700 kPa. The tandem axle load was treated as two
independent single axle loads. The distance between the two tire centers was set at
340 mm. The stress under the wheel and the stress between the wheels were both
calculated to identify the most critical stress.
The total truck traffic volume was calculated from two-way annual average daily
traffic (AADT), percentage of trucks, direction distribution factor, lane distribution
factor, annual growth rate, design life, and traffic safety factor (Li et al. 2014). The
total truck traffic volume was divided into different axle loads according to the
axle-load distribution factor. The equivalent single axle load (ESAL) was then
calculated using the 4th-power law.

8.3 Design Tool and Validation

A mechanistic-empirical design tool for PICP was created in an Excel® spreadsheet

and used to run the analyses and prepare example design tables. The critical
responses, including the major and minor principal stresses (σ1 and σ3) on top of the
subbase and subgrade layers, required for calculating the shear stress/strength ratio
(SSR), were calculated with the inputs listed above using multilayer linear elastic
theory. The OpenPave software program (Lea undated) was used for these analyses.
An incremental recursive analysis method (Ullidtz et al. 2010) was then used to
calculate the rut depth in the M-E design tool. The tool was used to provide the
following:
• The expected total rut depth (derived from the predicted rut depth in the sub-
grade) based on the input values of pavement structure, material properties,
climate, and traffic.
• The required minimum thickness of the subbase layer for a given allowable rut
depth.
The design tool was validated and the results calibrated with the data collected
during the three HVS tests. No independent validation of the calibration was
undertaken since only one set of HVS test data, and no comprehensive long-term
performance measurements with traffic and subbase water content data from
Protocols for Accelerated Pavement Testing of Fully Permeable … 367

in-service pavements are available. The maximum error between the measured and
calculated rut depths ranged from −4 to +5 %, with rut depth shift factors of 1.10
for dry conditions and 1.23 for wet conditions.
The main observations from the earlier mechanistic analysis discussed above
were refined as follows after this analysis:
• Higher shear stress/strength ratios at the top of the subgrade, which equate to a
higher risk of rutting in the subgrade, require thicker subbase layers, as
expected.
• For the same shear stress/strength ratio at the top of the subbase, an increase in
the stiffness of the subbase layer reduces the required thickness of that subbase
layer, especially when the subgrade has a low stiffness.
• For the same shear stress/strength ratio at the top of the subgrade, wet conditions
require thicker subbase layers compared to the dry condition, confirming that
wet conditions are the most critical condition for design.

9 Conclusions

This paper summarizes the research undertaken to develop criteria for revising
design tables for permeable pavements in general and specifically permeable
interlocking concrete pavement using a mechanistic-empirical design approach and
validated with results from an accelerated pavement test on a custom built test track.
When compared to the values in the current ICPI design tables, the new recom-
mended minimum subbase thicknesses required to prevent subgrade rutting do not
differ significantly; with new thicknesses slightly thinner or slightly thicker
depending on the design traffic, resilient modulus of the subgrade, and number of
days that the subbase contains water. In scenarios where lower numbers of days
with standing water in the subbase are selected, the new recommended minimum
subbase thicknesses are mostly less conservative that the thicknesses proposed in
the current ICPI tables.
The use of the shear stress to shear strength ratios at the top of the subbase and
top of the subgrade as inputs for modelling the rut development rate at the top of
these layers is considered to be an appropriate design approach for permeable
pavements.

Acknowledgments The authors acknowledge the financial support for the project provided by the
ICPI Foundation for Education and Research, the Concrete Masonry Association of California and
Nevada, and the California Nevada Cement Association.
368 D. Jones et al.

References

American Association of State Highway and Transportation Officials. (1993). Guide for design of
pavement structures. Washington, DC: American Association of State Highway and
Transportation Officials.
Carter, M., & Bentley, S. P. (1991). Correlations of soil properties. London, UK: Pentech Press.
Chow, L. C., & Tutumluer, E. (2014). Framework for improved unbound aggregate base rutting
model development for mechanistic-empirical pavement design. In Proceedings transportation
research board 93rd annual meeting. Washington, DC: Transportation Research Board.
Das, B. M. (2007). Advanced soil mechanics. New York, NY: Taylor and Francis.
Huang, H., Tutumluer, E., & Dombrow, W. (2009). Laboratory characterization of fouled railroad
ballast behavior. Transportation Research Record: Journal of the Transportation Research
Board, 2117, 93–101.
Jones, D., Harvey, J., Li, H., Wang, T., Wu, R., & Campbell, B. (2010). Laboratory testing and
modeling for structural performance of fully permeable pavements under heavy traffic: Final
report. Davis, CA: University of California Pavement Research Center. UCPRC-RR-2010-05/
CTSW-RR-09-249.04.
Jones, D., Li, H., & Harvey, J. T. (2013a). Development and HVS validation of design tables for
permeable interlocking concrete pavement: Literature review. Davis, CA: University of
California Pavement Research Center. UCPRC-TM-2013-03.
Jones, D., Li, H., Harvey, J. T. (2013b). Development and HVS validation of design tables for
permeable interlocking concrete pavement: Field testing and test section structural design.
Davis, CA: University of California Pavement Research Center. UCPRC-TM-2013-09.
Kim, I., & Tutumluer, E. (2006). Field validation of airport pavement granular layer rutting
predictions. Transportation Research Record: Journal of the Transportation Research Board,
1952, 48–57.
Lea, J. (undated). OpenPave: An N-layer, N-load, N-point multi-layer elastic half space
calculation. www.OpenPave.org. Accessed September 10, 2013.
Li, H., Harvey, J. T., & Jones, D. (2010). Summary of a computer modeling study to understand
the performance properties of fully permeable pavements. Davis, CA: University of California
Pavement Research Center. UCPRC-TM-2010-04.
Li, H., Jones, D., Wu, R., & Harvey, J. (2014). Development and HVS validation of design tables
for permeable interlocking concrete pavement: Final report. Davis, CA: University of
California Pavement Research Center. UCPRC-RR-2014-04.
Nova, R. (2012). Soil mechanics. Hoboken, NJ: Wiley
Smith, D. R. (2011). Permeable interlocking concrete pavements (4th ed.). Chantilly, VI:
Interlocking Concrete Pavement Institute.
Thompson, M., Gomez-Ramirez, F., & Bejarano, M. (2002). Illi-Pave based flexible pavement
design concepts for multiple wheel heavy gear load aircraft. In Proceedings 9th international
conference on asphalt pavements. Copenhagen, Denmark: International Society of Asphalt
Pavements.
Tutumluer, E., Kim, I., & Santoni, R. (2004). Modulus anisotropy and shear stability of
geofiber-stabilized sands. Transportation Research Record: Journal of the Transportation
Research Board, 1874, 125–135.
Tutumluer, E., & Seyhan, U. (1999). Laboratory determination of anisotropic aggregate resilient
moduli using an innovative test device. Transportation Research Record: Journal of the
Transportation Research Board, 1687, 13–21.
Protocols for Accelerated Pavement Testing of Fully Permeable … 369

Ullidtz, P., Harvey, J. T., Basheer, I., Jones, D., Wu, R., Lea, J., et al. (2010). CalME, a
mechanistic-empirical program to analyze and design flexible pavement rehabilitation.
Transportation Research Record: Journal of the Transportation Research Board, 2153,
143–152.
Wnek, M., Tutumluer, E., Moaveni, M., & Gehringer, E. (2013). Investigation of aggregate
properties influencing railroad ballast performance. Transportation Research Record: Journal
of the Transportation Research Board, 2374, 180–189.
 Part V
Accelerated Pavement Testing on Asphalt
Concrete Pavements
Behavior Evolution on Performance
of UV-Irradiation Aged Asphalt Mixtures
Under Reduced-Scale Accelerated
Trafficking

Jinting Wu, Fen Ye, Frederick Hugo, Yinting Wu, Feng Wang
and Xinhui Ding

Abstract In order to investigate the actual behavior evolution of asphalt pavement

after the Ultra Violet irradiation (UV-irradiation) aging, reduced-scale accelerated
pavement tests were conducted on two separate asphalt mixtures with a stone
matrix asphalt (SMA-13) structure, respectively, crumb-rubber modified asphalt
mixture and its matrix asphalt mixture, at ambient air temperature. In each group,
UV-irradiation aged asphalt mixture after ultraviolet irradiation of 300 h and its
corresponding non-aged asphalt mixture were trafficked at the same time by the 1/3
model mobile load simulator (MMLS3) till the cumulative loading applications up
to 500,000. The test results show that the performance attenuation degree of the
crumb-rubber modified asphalt mixture was lower than that of the matrix asphalt
mixture, so the rubber additive helps to reduce the effect of the UV-irradiation
aging. Rutting deformation, three-dimensional dynamic strain and seismic modulus
were measured during the accelerated loading test. The rutting results (including the
rutting depth, rutting area and its development rates), the surface dynamic strains,
and the decay rate of the seismic modulus from the UV-irradiation aged asphalt
mixture were greater than those of non-aged sample, and the difference narrowed
after the cumulative loading applications reached more than 100,000. In addition
the analysis of the modification rutting factor from the dynamic shear rheometer
(DSR) test, the UV-irradiation aging weakened the high temperature stability and

J. Wu (&)
F. Ye
F. Wang
X. Ding

Tongji University, 1239 Siping Road, Shanghai 200092, People’s Republic of China
e-mail: [email protected]
F. Hugo
Stellenbosch University, Stellenbosch 7600, South Africa
e-mail: [email protected]
Y. Wu
School of Geology Engineering and Geomatics, Chang’an University, Xi’an,
People’s Republic of China
e-mail: [email protected]
Y. Wu
National Engineering Laboratory for Offshore Oil Exploration, Xi’an Jiaotong University,
126 Yan Ta Road, Xi’an 710054, People’s Republic of China

© Springer International Publishing Switzerland 2016 373

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_24
374 J. Wu et al.

mechanical strength. Therefore, the UV-irradiation aging aggregated the perfor-

mance decay of asphalt mixture and will inevitably shorten its service life.

Keywords Reduced-scale accelerated pavement testing Crumb-rubber asphalt
mixture

Matrix asphalt mixture

Rutting deformation

Three-dimensional


strain Seismic modulus

1 Introductory Historical Background Notes

Research has shown that under the ultraviolet irradiation, the internal composition
and chemical structure of asphalt binder could change with the oxygen involved in
the process. When the asphalt pavement is directly exposed to the sunlight, the
ultraviolet light, accounting for about 5 % of the sunlight, will affect the part in the
depth range of 0.1 mm from the surface; and then the irradiation influence will
expand to 1 mm depth because of the diffusion of asphalt molecular; moreover, air
void or crack damage will cause the irradiation influence to continue to expand into
a depth of 10 mm from the pavement surface (Yu et al. 2012). Because the asphalt
film thickness is generally in the range of 5–15 μm in the asphalt mixture, the
ultraviolet irradiation affects the asphalt pavement inevitably (Wu et al. 2007).
Kemp and Predoehl (1981) exposed asphalt mixtures to the light-oxidative
condition for 18 h at the temperature of 35 °C. The test shows that aging hardening
occurred only in the thickness of 5 μm at different mixtures’ surface layer. Under
the same conditions of ultraviolet irradiation, the aging degree of asphalt mixtures
was much lower than that of asphalt binder (Hugo and Kennedy 1985). Tia et al.
(1988) carried out asphalt mixture aging tests at different time by forced air con-
vection and ultraviolet irradiation at 60 °C, and they thought that the combination
of both conditions can simulate asphalt mixture’s long-term aging; Ultraviolet
irradiation was the main reason for the asphalt mixture’s aging; the degree of Ultra
Violet (UV) aging and PAV aging to the same mixture was similar.
In terms of splitting strength, Wu and Liang (2007) pointed out that the styrene
butadiene rubber (SBR) modified asphalt mixture had better resistance to UV aging
than the matrix asphalt mixture. Qian et al. (2005) compared the styrene butadiene
styrene (SBS) and styrene ethylene butadiene styrene (SEBS) modified asphalt and
their mixtures, and thought that the aging resistance performance of the latter was
more than double the former, and the SEBS ones were fit for the South China and
the areas with strong ultraviolet irradiation. Compared with the SBR modified
asphalt mixture, the UV aging resistance performance of SBS was better (Durrieu
et al. 2007; Liu et al. 2009). Lins et al. (2008) exposed hot asphalt mixtures to the
ultraviolet irradiation and humidity cycling conditions, and the results showed that
the UV aging decreased the thermal stability and molecular weight, along with the
oxidation process. From the UV aging results conducted by Chen et al. (2010) and
Guo and Zhang (2012), the low temperature performance of asphalt mixtures were
reduced and that of modified asphalt mixture was better than the matrix asphalt
Behavior Evolution on Performance of UV-Irradiation … 375

mixture’s. From the variation values of Marshall stability and flow value before and
after UV aging, He et al. (2012) showed that the UV aging degree of dense-graded
asphalt mixture was smallest, followed by the semi-open-graded and open-graded
in sequence. In the test conducted by Wang (2013), the asphalt mixture’s dynamic
stability, low temperature crack resistance and resilience modulus all decreased
after UV aging. Li (2013) put two kinds of SBS modified asphalt mixtures into UV
aging chamber at fixed radiation intensity and different irradiation time, and then
got the decline law of their road performance, and finally put forward the optimal
gradation and asphalt-aggregate ratio in Inner Mongolia under strong ultraviolet
irradiation. Ye and Xu (2014) compared the performance of SBS, SBR and matrix
asphalt mixtures with the same gradation before and after UV aging, and con-
cluded: the splitting strength overall increased after UV aging; the resistance ability
of high temperature deformation improved; and the aged performance of SBS
modified asphalt mixture was better than the SBR modified asphalt mixture. Li et al.
(2014) conducted rutting test on the UV aging asphalt mixtures under different
aging time, gradation types and asphalt-aggregate ratios. The results show that, the
dynamic stability decreased along with the ultraviolet irradiation time increasing
and then it tended to be stable after a certain accumulation of ultraviolet aging; with
the increase of medium particle size aggregates, the high temperature stability of
aged asphalt mixture significantly improved; the dynamic stability of asphalt
mixture increased with a lower asphalt-aggregate ratios than the optimal one, on the
contrary, the dynamic stability decreased. Zhang (2015) conducted ultraviolet
irradiation on the Marshall specimens and rutting plate specimens, and obtained the
results that the Marshall stability, splitting strength and freeze-thaw splitting
strength ratio reduced to a certain extent after UV irradiation aging.
Based on the above, reported studies on the UV-irradiation aging of asphalt
mixtures, there are no unified test procedures as reference at present. Distinct
differences exist in the test equipment, test method and test condition, such as the
ultraviolet irradiation intensity, irradiation time and asphalt film thickness, among
units and researchers. Therefore, the results are not same, even on the contrary, and
poor comparability are among them. Moreover, the existing research is generally
focus on the matrix asphalt and SBS modified asphalt, but few on the crumb-rubber
asphalt mixture. In addition, at present, the performance decay of materials and
structures caused by the UV-irradiation aging is not considered in asphalt material
and structure design. However, in the region with high intensity ultraviolet,
UV-irradiation is an inevitable and overlooked factor.
Since the late 1990s, MMLS3 has been used to study the relationship and
differences of fatigue, high temperature and deformation properties obtained from
accelerated loading tests and traditional indoor tests among different materials or
structures. Hugo et al. (1999) compared the results found by MMLS3 and
Texas MLS (TxMLS), and concluded that the permanent deformation and fatigue
damage were similar in both tests; the results from MMLS3 loading test were
similar to the results of core sample laboratory test. Rutting deformation was pre-
dicted and evaluated in the MMLS3 loading test (Epps et al. 2002; Kruger et al.
2004). In China, Peng (2007) adopted MMLS3 to study the effect of temperature on
376 J. Wu et al.

the permanent deformation of asphalt mixture and compared the results with the
wheel rutting test and Hamburg rutting test. In Tongji University, asphalt mixture
samples with different types and materials were tested by MMLS3 (Lu et al. 2011;
Wu et al. 2012). Shen et al. (2012) conducted accelerated loading test on the in-site
pavement, and analyzed the influence of air void and loading applications on the
high temperature performance of the upper and middle layers. An overview of case
studies of the comparison between performances under full-scale and scaled traf-
ficking, is reviewed in the MLS synthesis by Hugo and Steyn (2015).
It was concluded that MMLS3 could be used to distinguish the performance
difference of different materials and structures. Based on this finding, this paper
reports on reduced-scale accelerated loading test by MMLS3 on the UV-irradiation
aged asphalt mixture sample and non-aged one, in order to investigate the effect of
ultraviolet irradiation on the asphalt pavement.

2 Test Overview

2.1 Asphalt Mixture

Considering the SMA-13, with the nominal maximum aggregate size of 13.2 mm,
as a typical upper layer used in the asphalt pavement structure, SMA-13 was
selected for the ultraviolet aging test and MMLS3 accelerated loading test.
Four kinds of aggregates were used in the test: the aggregate of 0–3 mm size was
limestone, and the aggregates of 3–5, 5–10 and 10–15 mm sizes were all basalt.
Limestone powder was chosen in the test. The sand equivalent of limestone
aggregate was 65, which met the requirement of more than 60. The basic properties
of aggregates met requirements. In order to enhance the effect of oil absorption,
lignin fiber stabilizer was added into SMA-13, and its consumption reached about
0.3 % of the aggregate quality. The asphalt-aggregate ratio was 6.3 %. The mea-
sured Marshall volume parameters and performance results of SMA-13 met
requirements.
A multi dimension vibratory roller named Slabcompactor was used to form the
samples of crumb-rubber asphalt mixture and matrix asphalt mixture at the com-
paction temperature of 175 °C. The sample dimension was 50 cm long, 30 cm wide
and 10 cm high. Each sample was divided into two longitudinally by a cutting
machine. The 1/2 sample was placed under the ultraviolet rays to receive the
UV-irradiation aging, and the other 1/2 sample kept away from light under ambient
air temperature, as the non-aged comparison sample.
The ultraviolet irradiation intensity in the longitudinal center of the aged sample
can reach 400 W/m2, and the intensity on both sides decreased along with the
deviation distance from the longitudinal center, in accordance with the quadratic
polynomial model. The UV-irradiation time of the matrix and crumb-rubber asphalt
Behavior Evolution on Performance of UV-Irradiation … 377

mixtures was set as 300 h. The highest temperature of the aged sample surface was
not more than 60 °C.
The MMLS3 loading test divides into two groups—the crumb-rubber asphalt
mixture and matrix asphalt mixture. In each group, the UV-irradiation aged 1/2
specimen and the non-aged 1/2 specimen were placed in the loading steel cell by
splicing together longitudinally. So they can be trafficked at the same time, which
can meet the performance comparison to analyze the UV-irradiation effect on the
mixtures.

2.2 Loading Parameters

The MMLS3 loading system comprises the load simulator, environment control
chamber, heating, refrigeration and drying system, loading cell, and so on. It can
simulate the designed loads, temperature and humidity. The overall size of the
loading device MMLS3 is 2.40 m long, 0.60 m wide, and 1.20 m high. It com-
prises a rigid outer frame with 4 individual wheel tires of Φ300 and 80 mm wide,
and the tire pressure was approximately 0.87 MPa. The load per wheel was
2.70 kN. The average ground contact area of a single tire was measured to be
approximately 81.85 cm2, and the actual tire pressure area was approximately
30.80 cm2. In order to ensure the loading stability, the initial loading speed was
3600 applications per hour, and then the loading speed of 6000 applications per
hour was used in the subsequent loading. For the crumb-rubber and matrix asphalt
mixtures, the cumulative loading applications limits to the loading speed change
were 60,000 and 10,000, respectively. Lateral wander was not considered in the
reduced-scale loading test. The cumulative loading applications was up to 500,000
at ambient air temperature, and the sample surface temperature was about 33 °C
during trafficking. When the cumulative loading applications reached a certain
specified value, trafficking was suspended for the measure of performance
parameters.

2.3 Measure Elements

In the loading process, rutting deformation, dynamic strain and seismic modulus
were measured. Considering the factors such as sample size, measurement accuracy
and loads disturbance, the platinum resistance strain gauge, BX120-30AA, was
used to measure the longitudinal and transverse strains in the top and bottom
surface and the longitudinal and vertical strains in the sample side. The dynamic
strain data acquisition was carried out by the DHDAS (5920-1394) system all
around the loading process and the acquisition frequency was 100 Hz. MLS
Profilometer Driver-P900 and PSPA (Portable Seismic Property Analyzer) were
378 J. Wu et al.

used to collect rutting deformation and seismic modulus data, respectively, when
the cumulative loading applications achieved a certain set value.

2.3.1 Strain Measuring Point Arrangement

Grid lines were drew in the top surface, bottom surface and side surface of the
prepared asphalt mixture specimen. There the longitudinal line was the center line
in each measure surface, and the transverse line was set every 5 cm. In each surface,
two measuring points were arranged, 10 and 25 cm apart from the stitching loca-
tion, respectively. In the top and bottom surface, a longitudinal and transverse strain
gauges were arranged at every point, and what’s more, the gauges in the top surface
were one to one corresponding with theses in the bottom surface. Likewise, in the
side surface, longitudinal and vertical strain gauges were arranged at every point.
The strain gauge layout is shown in Fig. 1. In order to obtain the more accurate
dynamic strain, another two mixture samples were prepared with the same material,
structure and strain gauge layout as the loaded sample. The duplicate samples were
placed beside the test cell to do the temperature compensation one-to-one for the
measured strain.
The strain gauges were bonded to the measuring points with epoxy resin, and the
every stain gauge in the top surface was covered and protected by a small piece of
silicon sheet (its thickness was less than 1 mm), which was also glued with a small
amount of epoxy resin. The above gauges preparation was completed at least 48 h
before the trafficking. The sample layout with strain gauges is shown in Fig. 2.

2.3.2 Rutting Deformation Section Layout

Three transverse sections on the non-aged asphalt mixture specimen were selected
as the rutting measure sections, and its distances apart from the stitching location
were 5, 20 and 35 cm, respectively. Similarly, on the UV-irradiation aged asphalt
mixture specimen, five transverse sections were selected and its distances apart
from the stitching location were 5, 15, 26, 35 and 45 cm, respectively. The sketch
of the rutting sections are shown in Fig. 1.

10cm 5cm 10cm 5cm 5cm 5cm×3 10cm×3

10cm

non-aged sample UV-irradiation aged sample

50cm 50cm
Longitudinal Transverse Vertical Seismicmodulus Rutting

Fig. 1 Measure elements layout

Behavior Evolution on Performance of UV-Irradiation … 379

Strain gauges covered

with silicon sheets

Non-aged sample

UV-irradiation aged sample

Duplicate samples for
temperature compensation

Rutting deformation

Steel cell

Fig. 2 Samples layout

2.3.3 Seismic Modulus Measuring Point Arrangement

The intersection of rutting section and the surface longitudinal center line, shown as
black dots in Fig. 1, was regarded as the measuring points for seismic modulus.

3 Performance Results Analysis

3.1 High Temperature Stability

3.1.1 Rutting Deformation

Each rutting profile section was 130 mm in length and rutting data was collected at
2 mm interval. The rutting curves are shown in Fig. 3. The rutting curve of matrix
asphalt mixture specimen was in a typical “upheaval-depression-upheaval” model.
However, depression was the main deformation in the rutting of crumb-rubber
asphalt mixture specimen, while the upheaval deformation was not obvious.
The original morphology of each rutting section without trafficking was as the
baseline for rutting deformation calculation. The part higher than the baseline
means upheaval deformation, and the relative deformation is expressed as a positive
value. Correspondingly, the part lower than the baseline indicates depression
deformation, and the relative deformation is expressed as a negative value. The
algebraic difference between the highest point and the lowest point in the rutting
curve is defined as the rutting depth. Rutting area is defined as the area surrounded
by the rutting deformation curve and the baseline. Likewise, the upheaval area
380 J. Wu et al.

Fig. 3 Rutting deformation 2

curve
1

Rutting deforrmation /mm

Transverse length /mm
0
0 50 100
-1
2
-2
-3
-4
-5
-6 Matrix-non-aged Matrix-UV-aged

above the baseline is expressed as a positive value, and the depression area below
the baseline is expressed as a negative value.
Figure 4 shows the results of mean value of rutting depth along with the loading
applications. Under the MMLS3 trafficking, the mean rutting depth of
UV-irradiation aged asphalt mixture was greater than that of the corresponding
non-aged asphalt mixture, and the gap was gradually reduced after 100,000 loading
applications. The mean rutting depth of crumb-rubber asphalt mixture was less than
that of matrix one, and the gap between both increased with the increase of loading
applications, and the difference between the UV-irradiation aged and non-aged
crumb-rubber asphalt mixture was less than that of the matrix asphalt mixture. The
relationship between the mean rutting depth and loading applications was found to
be a power function (R2 > 0.97). The rutting results show that the high temperature
stability of asphalt mixture was reduced by the ultraviolet irradiation, and the effect
of ultraviolet aging on the crumb-rubber asphalt mixture was much less than that of
the matrix asphalt mixture.

Fig. 4 Rutting depth

Behavior Evolution on Performance of UV-Irradiation … 381

400 UV-aged-upheaval y = 67.789x0.48 365.83

UV-aged-depression R² = 0.97 330.27
350
Non-aged-upheaval 291.59
300 Non-aged-depression 263.25
Rutting area /mm2

250 222.49 228.27

205.34 208.20
188.69
200 175.60
161.62
139.25
150 105.30 112.15
98.90
81.73 y = 35.548x0.52
100 59.88 R² = 0.89
45.94
50

0
1 2 4 6 10 20 30 40 50
Loading applications /104
(a) Matrix asphalt mixture (b) Rubber asphalt mixture

Fig. 5 Rutting area of asphalt mixture

Figure 5 shows the results of mean value of rutting area along with the loading
applications. There was a power function relationship between the loading appli-
cations and the rutting area of asphalt mixtures. The rutting area of the
UV-irradiation aged matrix asphalt mixture sample was greater than that non-aged,
and the former was about twice of the latter, but the multiple difference of the both
reduced with the increase of loading applications. For the crumb-rubber asphalt
mixture, the rutting area after UV-irradiation aging was slightly greater than that
non-aged, and the multiple ratio, approximately in the range of 1.1–1.2, was stable
during the loading test. The upheaval area of the non-aged and UV-irradiation aged
matrix asphalt mixture was about 18 and 3 %, respectively. So, although the
UV-irradiation reduced the high temperature stability of matrix asphalt mixture, but
to a certain extent, it also reduced the lateral flow deformation.
The node and phase deformation rates were calculated to study the rutting
deformation development in the loading test. The node deformation rate indicated
that the cumulative rutting deformation’s growth with the increase of the loading
applications, which was defined as the ratio of the total cumulative rutting depth (or
area) and the loading applications from the beginning to the calculation node. The
phase deformation rate was used to characterize the growth rule of rutting defor-
mation in a certain period of trafficking, as the ratio of the added rutting depth (or
area) and the added loading applications in a loading phase. The detailed calcula-
tion methods can be defined according to the following Eqs. 1 and 2. The calcu-
lation results of rutting depth and area development rate are shown in Fig. 6.

Rn
Vnode ¼ ð1Þ
Nn
Rn  Rn1
Vphase ¼ ð2Þ
Nn  Nn1

where Vnode is the node rate of the rutting deformation (rutting depth or area); Vphase
is the phase rate of the rutting deformation (rutting depth or area) with given
382 J. Wu et al.

2.5 Matrix-non-aged-node 150 Matrix-non-aging-node

Matrix-non-aged-phase
2.0 Matrix-UV-aged-node Matrix-non-aging-phase
Matrix-UV-aged-phase Matrix-UV-aging-node
100 Matrix-UV-aging-phase
1.5
Rutting depth rate /mm·10-4

Rutting area rate /mm2·10-4

1.0 y = 2.17x-0.715
50 R² = 0.99
0.5
0
0.0
0 10 20 30 40 50
0 10 20 30 40 50
1.5
-0.5
-50
-1.0
1.0
-100
-1.5
0.5

-2.0
0.0 -150
Loading applications /104
Loading applications /104
(a) Rutting depth development rate (b) Rutting area development rate

Fig. 6 Development rate of rutting deformation

loading applications; Rn and Rn−1 are the deformation (rutting depth or area) cor-
responding to the loading applications of Nn and Nn−1, respectively.
It can be seen that the development rate of rutting depth (or area) of asphalt
mixture decreased with the loading applications, in the form of power function. The
rutting development rate of UV-irradiation aged matrix (or crumb-rubber) asphalt
mixture was about 1.5 times (or 1.2 times) that of the corresponding non-aged one.

3.1.2 Advanced Rutting Factor

In order to further study the development of high temperature performance under

UV-irradiation condition, DSR time sweep test was conducted on the crumb-rubber
asphalt binder and its matrix asphalt binder. The asphalt binders were subjected
UV-irradiation for different time, including 100, 200, 300, 400, 500 and 600 h.
Moreover, at each irradiation time, there were three kinds of irradiation intensity,
that is, high intensity (400 W/m2), medium intensity (200 W/m2) and low intensity
(50 W/m2). The DSR parameters are as follows: the sample diameter was 25 mm,
the sample thickness was 1 mm and the temperature was set from 40 to 82 °C and
the time sweep test was conducted every 6 °C.
Since the Superpave specification has been put forward, the researchers found
that G*/sinδ, as the rutting factor, has a great limitation to evaluate the high tem-
perature stability of asphalt binders, especially for the modified asphalt binders.
Therefore, some researchers have questioned the factor’s applicability and proposed
different evaluation indexes or standards for high temperature performance, one of
which proposed by Shenoy (2004) has the largest impact, namely the factor of
G*/(1 − 1/(sinδ
tgδ)). Although it has taken the delayed elasticity of asphalt binder
into consideration, directly uses the unrecoverable deformation to characterize the
high temperature performance of asphalt binder and is more sensitive to the change
Behavior Evolution on Performance of UV-Irradiation … 383

(a) Matrix asphalt binder (b) Crumbe-rubber asphalt binder

Fig. 7 G*/(sinδ)9 of asphalt binder

of δ, the unrecoverable strain cannot be less than zero, namely G*/(1 − 1/

(sinδ
tgδ)) > 0, that is δ should be more than 51.8°. Therefore, when δ is less than
51.8°, the index is no longer applicable. According to this situation, Shenoy found
that G*/(sinδ)9 and G*/(1 − 1/(sinδ
tgδ)) had excellent correlation and G*/(sinδ)9
applied to all values of δ. Therefore, G*/(sinδ)9 can replace G*/(1 − 1/(sinδ
tgδ)) as
the optimization advanced rutting factor.
Taking the time sweep results at 64 °C in the DSR test for example, analysis of
the development of G*/(sinδ)9 with the UV-irradiation time of crumb-rubber
asphalt binder in the low intensity, medium intensity and high intensity was done.
The advanced rutting factor decreased with the increase of UV-irradiation time and
intensity. Under the same UV-irradiation time, the G*/(sinδ)9 and irradiation
intensity have power function relationship. The relationship between G*/(sinδ)9 and
UV-irradiation energy (obtained from the product of UV-irradiation intensity and
UV-irradiation time) can be expressed by power function or logarithmic function,
as shown in Fig. 7.
The advanced rutting factor G*/(sinδ)9 of the matrix asphalt binder has little
change with the UV-irradiation time increasing, but it increased at first and then
decreased with the UV-irradiation intensity increasing. Therefore, under the high
UV-irradiation intensity and high temperature conditions, G*/(sinδ)9 of the matrix
asphalt binder will decay rapidly and the resistance to permanent deformation will
be weakened.

3.2 Dynamic Strain Response

In the trafficking process, the dynamic strain data acquisition is carried out to obtain
the development law of spatial distribution and time history of three-dimensional
strains on the top surface, bottom surface and side surface of the asphalt mixtures.
384 J. Wu et al.

3.2.1 Instantaneous Dynamic Strain

From the measured strain analysis, it can be known that the instantaneous dynamic
strain curves obtained under different loading rates and samples are similar. The
instantaneous longitudinal dynamic strain was subjected to compression-tension-
compression cycles, which can accelerate the fatigue failure of asphalt mixture:
when the wheel load was close to the measure point, it was compressed in longi-
tudinal direction; then when the wheel load just moved on the measure point, the
longitudinal was tension and up to its peak value; with the wheel load away, the
measure point was recovered from the tension state. The instantaneous transverse
dynamic strain generally showed a one-way form: the measure point at the top
surface suffered transverse compressive strain; while the measure point at the
bottom surface suffered transverse tensile strain. The instantaneous vertical
dynamic strain was in a one-way compression state.
Compared with the instantaneous strains at the top surface, due to the diffusion
of trafficking loads in the 10 cm thick asphalt mixture, the measure points at the
bottom surface were almost in the load influence zone in the whole loading process;
however, just when the trafficking load was very close to the measure point, strain
response at the top surface began appear. That is, the measure points at the top
surface presented strain recovery when the wheel load was not on them, however,
the strain recovery at the bottom surface was not obvious.
For the matrix asphalt mixture, the instantaneous dynamic longitudinal and
transverse strains on the surface were about 300 με, and the longitudinal and
vertical strains on the sides were about 100 με. It can be seen that on the top
surface, the transverse strain of the matrix asphalt mixture was greater than the
longitudinal stain, and the multiple difference decreased along with the loading
applications; the former was 1.03–1.22 times the latter on the UV-irradiation aged
matrix asphalt mixture, and the corresponding times was in the range of 1.17–1.53
on the non-aged matrix asphalt mixture. To the bottom surface, the instantaneous
transverse strain was larger than the longitudinal one on the matrix asphalt mixture,
while the instantaneous transverse strain was less than the longitudinal one on the
crumb-rubber asphalt mixture.
At different loading phases, the strain ratio between the UV-irradiation and
non-aged asphalt mixtures is shown in Table 1. Concrete results are as follows. In
the loading process, to the matrix asphalt mixture, the ratio of the longitudinal
surface strain between the UV-irradiation aged and non-aged samples was in the
range of 1.00–1.20; similarly, the ratio of the transverse surface strain was in the
range of 0.89–1.17; the ratio of the vertical side strain was in the range of 1.29–3.71
and the ratio of the longitudinal strain was in the range of 0.60–0.78. The longi-
tudinal strain at the bottom surface of the crumb-rubber asphalt mixture was less
than 250 με, and the transverse strain at the bottom surface and the longitudinal
strain at the side surface were both less than 100 με. The strains at the bottom
surface of the UV-irradiation aged and non-aged crumb-rubber asphalt mixtures had
little difference (that is, the ratio was approximately 1.00), as well as the longitu-
dinal strain at the side surface. The vertical strain at the side surface of the
Table 1 Strain ratio of UV-irradiation aged and non-aged asphalt mixtures
Type Loading Longitudinal-top Transverse-top Longitudinal-bottom Transverse-bottom Longitudinal-side Vertical-side
phase/104
Matrix asphalt 0–1 1.20 1.07 1.10 0.96 0.78 1.29
mixture 1–2 1.10 1.17 1.08 0.94 0.78 3.71
2–4 1.13 1.08 1.05 1.03 0.81 3.57
4–6 1.15 1.03 1.20 1.03 0.63 2.22
6–10 1.12 1.00 1.11 1.03 0.65 2.98
10–20 1.06 0.91 1.09 1.01 0.57 1.44
30–40 1.00 0.91 0.99 0.95 0.60 2.72
40–50 1.04 0.89 0.97 1.02 0.61 2.59
Behavior Evolution on Performance of UV-Irradiation …

Crumb-rubber 0–1 1.25 1.37 1.15 1.26 1.05 2.35

asphalt mixture 1–4 1.19 1.16 1.09 1.17 1.06 2.65
4–10 1.30 1.01 1.13 1.25 1.01 2.94
10–20 1.09 1.06 1.07 1.18 0.96 3.09
20–30 0.96 1.22 1.01 1.11 0.92 3.10
30–50 1.07 1.25 1.02 0.97 0.90 2.48
385
386 J. Wu et al.

0-1 1-2
400 2-4 4-6
6-10 10-20
350 30-40 40-50

Transverse strrain /µε

300

250

200

150
3600 6000
100
Loading phase

(a) Longitudinal cumulative strain (b) Transverse cumulative strain

Fig. 8 Cumulative strain of matrix asphalt mixture

UV-irradiation aged crumb-rubber asphalt mixture was approximately twice to

triple that of the non-aged sample; the longitudinal strain at the top surface of the
former was 1.07–1.25 times of the latter; and the transverse strain at the top surface
of the former was 1.01–1.37 times of the latter.
It can be seen that, the UV-irradiation aging had little effect on the strains of
bottom surface, but the top surface strains of the UV-irradiation asphalt mixture
were larger than the corresponding values of the non-aged asphalt mixture; with the
increase of the loading applications, the surface strain gaps between UV-irradiation
aged and non-aged samples decreased, that is, the UV-irradiation effects on the
mechanical response of asphalt mixture decreased.
Based on the above, the UV-irradiation aging has greater impact on the top
surface strains and the side vertical strain. The highest increase ratio of the surface
strains caused by UV-irradiation aging was up to 30 %, and the ratio was up to
200 %. The strain results add credence to the view that the UV-irradiation induces
greater rutting deformation of the asphalt mixtures.

3.2.2 Cumulative Dynamic Strain

Considering the UV-irradiation has greater impact on the top surface of asphalt
mixture specimens, here the top surface strains of the matrix asphalt mixture are
taken as an example, shown in Fig. 7. The peak-valley difference in the curve of
instantaneous dynamic strain to loading applications is taken as the strain ampli-
tude. In each loading phase, the mean amplitude within one hour was calculated one
by one, in order to analyze the cumulative variation of dynamic strain along with
the loading applications. As shown in Fig. 8, the cumulative strains decreased when
the loading speed of 3600 applications per hour changed into 6000 applications per
hour. In the suspense period of 3 h, when the rutting deformation and seismic
modulus were measured after a certain loading applications, the strains recovered
Behavior Evolution on Performance of UV-Irradiation … 387

till the next loading phase. In each loading phase, the surface cumulative strains of
the matrix asphalt mixture increased with the loading applications, while in the first
two loading phase, namely from the beginning to the time when the cumulative
loading applications were up to 40,000, the surface longitudinal and transverse
strains of the crumb-rubber asphalt mixture decreased, and then in the subsequent
loading phases, the cumulative strains increased with the increasing loading
applications.
For the matrix asphalt mixture, after 500,000 loading applications, the cumu-
lative longitudinal strains at surface of the UV-irradiation aged asphalt mixture
were up to in the range of 300–360 με, the transverse strains were up to 310 με
approximately, and the vertical strain at the side surface was about 85 με. For the
crumb-rubber asphalt mixture, the strains corresponding to the above values were
up to in the range of 290–300, 400 and 100 με, respectively. For the crumb-rubber
asphalt mixture, the cumulative strains were less than the corresponding values of
the matrix sample. For the UV-irradiation aged crumb-rubber asphalt mixture, the
cumulative longitudinal and transverse strains at top surface and the vertical strains
at the side were approximately up to 200, 290 and 100 με, respectively. For the
non-aged crumb-rubber asphalt mixture, the values corresponding to the above
were approximately 260, 345 and 55 με, respectively.

3.3 Seismic Modulus

The decay of the material modulus reflects the weakening process of the material’s
mechanical properties, and the modulus of the asphalt surface layer is an important
parameter affecting the permanent deformation. After the rutting measurement,
PSPA was used to measure the seismic modulus of asphalt mixture. PSPA is
composed of 1 seismic wave generator and 2 receivers. The average seismic
modulus of the pavement surface beneath the receivers can be calculated from the
Rayleigh wave time difference propagating between the two receivers.
The initial parameters in the seismic modulus measurement are set as follows:
the Poisson ratio is 0.25, the material density is 2.5 g/cm3, the R wave speed is
640 MPs, and the P wave speed is 1200 MPs. The surface temperature of measure
points was approximately in the range of 30–32 °C measured by an infrared
detection gun. Seismic modulus data of each measure point was collected at least 10
times. And the average of them were taken as the seismic modulus of the given
measure point. The variation coefficient of seismic modulus of each measure point
was less than 20 %. The results of seismic modulus are shown in Fig. 9.
The seismic modulus of the UV-irradiation aged asphalt mixture rapid decayed
with the trafficking and became less than that of the non-aged sample. After
500,000 loading applications, the seismic modulus of the UV-irradiation aged and
non-aged matrix asphalt mixtures were approximately 65 and 80 % of its corre-
sponding initial moduli, respectively. Similarly, the above two ratios of the
crumb-rubber asphalt mixture were 70 and 75 %, respectively. Compared to the
388 J. Wu et al.

(a) Matrix asphalt mixture (b) Crumb-rubber asphalt mixture

Fig. 9 Seismic modulus of asphalt mixture

matrix asphalt mixture, the seismic modulus of crumb-rubber asphalt mixture was
smaller. The seismic modulus of the crumb-rubber asphalt mixture attenuated with
the increase of loading applications. In addition, at the later phase of trafficking, the
seismic modulus increased slightly due to the decrease of temperature. Through the
fitting analysis, the regression formula between the seismic modulus, the loading
applications and the temperature can be shown in Eq. 3, with a larger variance.

M ¼ aN 2 þ bN þ cT þ d ð3Þ

where M is the seismic modulus; N is the loading applications; T is the temperature;

a, b, c and d are fitting parameters.
For the crumb-rubber asphalt mixture, the fitting parameters a, b, c and d are
0.00076, −0.02870, 0.17466 and −3.98010, respectively, with the corresponding
variance R2 of 0.66. Moreover, what should note is that the amount of the measured
data in this test is finite, further experiment or research is needed.

4 Conclusions

The comparison of the performance (high temperature stability, dynamic strain

response and seismic modulus) of the UV-irradiation aged and non-aged asphalt
mixtures, including the crumb-rubber sample and matrix sample, is analyzed in this
paper, and the following conclusions are drawn:
1. The average rutting depth, rutting area and the rutting development rate of the
asphalt mixture samples were related to the loading applications by power
function. The average rutting depth and rutting area of the crumb-rubber asphalt
mixture were less than that of the matrix asphalt mixture, and the gap between
both increased with the loading applications. For the crumb-rubber and its
matrix asphalt mixtures, the average rutting deformation of the UV-irradiation
Behavior Evolution on Performance of UV-Irradiation … 389

samples were larger than that of non-aged samples, and the difference of the
rutting deformation between the UV-irradiation aged and non-aged
crumb-rubber asphalt mixture was less than that of matrix one. Ultraviolet
aging weakens the high temperature stability of asphalt mixture and the influ-
ence of ultraviolet aging on the crumb-rubber asphalt mixture is less than that of
the matrix asphalt mixture.
2. The longitudinal dynamic strain of asphalt mixture was subjected to
compression-tension-compression cycles, while the transverse and vertical
strains were one-way response to the trafficking. With the loading applications
increasing, multiple difference between the transverse and longitudinal strains
decreased. UV-irradiation aging has little effect on the strains at the bottom
surface of asphalt mixture. The strains at the top surface and the side strains of
the UV-irradiation aged asphalt mixture were larger than the corresponding
values of the non-aged samples. For example, the surface strains increased by
30 % approximately, and the maximum increase ratio of the vertical strain was
up to 200 % approximately. With the increase of the loading applications, the
difference between the UV-irradiation aged and non-aged asphalt mixtures
reduced, that is the effect of UV-irradiation aging on the mechanical response of
asphalt mixture gradually reduced.
3. The decay rate of the seismic modulus of the UV-irradiation aged asphalt
mixture was larger than that of the non-aged one. With the loading applications
increasing, the seismic modulus of asphalt mixture attenuated. Moreover, the
seismic modulus increased slightly due to the decrease of temperature.
A regression formula between the seismic modulus, the loading applications and
the temperature is put forward as a reference.
It is apparent that the performance of the crumb-rubber asphalt mixture was
better than that of the matrix asphalt mixture, and the performance degradation
induced by UV-irradiation of the former was less than that of the latter. So, to a
certain extent, the modification of rubber powder reduces the effect of
UV-irradiation aging. As in the macro level, the performance change caused by the
UV-irradiation presents larger rutting deformation, smaller rutting factor, larger
instantaneous and cumulative dynamic strains and faster decay rate of the seismic
modulus. Therefore, the UV-irradiation aging contributes to the performance
degradation and inevitably shortens the service life of asphalt pavement.

Acknowledgments The research was supported by the National Natural Science Foundation of
China (Grant Nos. 51168044, 41204076), the Postdoctoral Science Foundation of China
(2013M540756) and the Fundamental Research Funds for the Central Universities
(0009-2014G2260010). The authors wish to acknowledge the support and contributions of the
project sponsors, other participating institutions and their personnel. The authors also thank the
anonymous referees who reviewed an earlier version of this manuscript for their constructive,
helpful comments.
390 J. Wu et al.

References

Chen, H. X., Jiang, Y., Li, S., & Zhang, H. L. (2010). Low-temperature properties of aged asphalt
mxitures. Journal of Chang’an University (Natural Science Edition), 30(01), 1–5.
Durrieu, F., Farcas, F., & Mouillet, V. (2007). The influence of UV aging of a
styrene/butadiene/styrene modified bitumen: Comparison between laboratory and on site
aging. Fuel, 86(10), 1446–1451.
Epps, A. L., Ahmed, T., Little, D. C., & Hugo, F. (2002). Performance prediction with the
MMLS3 at WesTrack. CD-Rom. In Proceedings of the Ninth International Conference on
Asphalt Pavements Copenhagen. pp. 17–22.
Guo, W. W., & Zhang, Z. T. (2012). Influence of ray aging on low temperature performance of
asphalt mixture in plateau-cold region. Journal of Chongqing Jiaotong University (Nature
Science), 31(1), 51–53.
He, W. H., Guo, W. W., & Yang, Q. (2012). Effect of xenon lamp accelerated aging on the
performance of asphalt concrete. Journal of East China Jiaotong University, 29(2), 15–20.
Hugo, F., & Kennedy, T. W. (1985). Surface cracking of asphalt mixtures in Southern Africa.
Proceedings of Association of Asphalt Paving Technologists, 54, 454–501.
Hugo, F., Fults, K., Chen, D. H., Smit, A. D. F., & Bilyeu, J. (1999). An overview of the TxMLS
program and lessons learned. CD-Rom. In Proceedings of the International Conference on
Accelerated Pavement Testing, Reno, Nevada.
Hugo, F., & Steyn, W. J. M. (2015). A synthesis of applications of the MLS as an innovative
system for evaluating performance of asphalt materials in pavement engineering. In
Proceedings of the 11th Conference on Asphalt Pavements for Southern Africa (CAPSA), 16
to 19 August 2015, Sun City, South Africa.
Kemp, G. R., & Predoehl, N. H. (1981). A comparison of field and laboratory environments on
asphalt durability. In Proceedings of International Conference on Association of Asphalt
Paving Technologists, p. 50.
Kruger, J., Hartman, A. M., & Loots, H. (2004). Towards developing a test protocol for field
permanent deformation performance evaluation using the MMLS3. In Proceedings of the 8th
Conference on Asphalt Pavements for Southern Africa (CAPSA’04).
Li, L. (2013). Experimental study on the anti-ultraviolet aging performance of asphalt mixture in
Northwest Inner Mongolia. Master Dissertation. Xi’an: Xi’an University of Architecture and
Technology.
Li, H. X., Guo, J. M., & Tong, S. J. (2014). Research on influence factors of high temperature
stability of ultraviolet aging asphalt mixture. Highway Engineering, 39(02), 72–75.
Lins, V. F. C., Araújo, M. F. A. S., Yoshida, M. I., Ferraz, V. P., Andrada, D. M., & Lameiras, F.
S. (2008). Photodegradation of hot-mix asphalt. Fuel, 87(15), 3254–3261.
Liu, L. P., Dong, W. L., Sun, L. J., & Jiang, T. (2009). Ultraviolet radiation aging performance of
SBS and SBR modified asphalt. Journal of Building Materials, 12(06), 676–678.
Lu, H. W., Zhang, H. C., Wang, J., & Guo, Y. N. (2011). Research on asphalt mixtures’ stability at
high temperatures applying MMLS3. Journal of Building Materials, 14(05), 624–629.
Peng, Y. Z. (2007). Temperature effect on the high temperature stability of asphalt mixture.
Journal of Highway and Transportation Research and Development, 7, 51–53.
Qian, C. X., Xie, J. G., & Wang, H. B. (2005). Anti-aging capability of SBS and SEBS modified
asphalt and mixture. Journal of the Southeast University (Natural Science Edition), 35(06),
945–949.
Shen, A. Q., Guo, Y. C., Chen, F., Yin, W., & Li, Y. L. (2012). Influence of asphalt mixture
segregation on long-term high temperature performance of asphalt pavement based on
MMLS3. Journal of Highway and Transport, 25(3), 80–86.
Shenoy, A. (2004). High temperature performance grading of asphalts through a specification
criterion that could capture field performance. Journal of Transportation Engineering, 130(1),
132–137.
Behavior Evolution on Performance of UV-Irradiation … 391

Tia, M., Ruth, B. E., Charai, C. T., Shiau, J. M., Richardson, D., & Willianms, J. (1988).
Investigation of original and in-service asphalt properties for the development of improved
specifications-final phase of testing and analysis. Final Report, Engineering and Industrial
Experiment Station, University of Florida, Gainesville, FL.
Wang, C. 2013. Research on asphalt mixture of strong ultraviolet region for weather ability and
aggregate gradation optimum. Master Dissertation. Xi’an: Xi’an University of Architecture
and Technology.
Wu, H., & Liang, N. X. (2007). Analysis of the ultraviolet aging of SBR modified asphalt mixture.
Journal of Chongqing Jiaotong University, 26(02), 72–74.
Wu, J. T., Ye, F., & Zhao, Q. Q. (2012). High temperature stability of different modified asphalt
mixtures based on MMLS3. Journal of Building Materials, 15(05), 654–659.
Wu, S. P., Pang, L., Yu, J. Y., Qiu, J., & Ma, L. X. (2007). Research progress on photooxidation
aging of asphalt. Petroleum Asphalt, 21(02), 1–6.
Ye, C. Y., & Xu, D. J. (2014). UV aging resistance capability studies of SBS and SBR modified
asphalt mixture. Journal of Wuhan University of Technology (Transportation Science and
Engineering), 38(04), 883–886.
Yu, J. Y., Pang, L., & Wu, S. P. (2012). Aging and anti-aging of asphalt material. Wuhan: Wuhan
University of Technology Press.
Zhang, P. (2015). Research on characteristics and detection methods of asphalt mixture damage
induced by water-temperature-radiation. Ph. D. Dissertation. Jilin: Jilin University.
Effects of Binder and Mix Properties
on the Mechanistic Responses of Fatigue
Cracking APT Sections

Daba S. Gedafa, Mustaque Hossain, Stefan A. Romanoschi

and Mbakisya A. Onyango

Abstract One of the major benefits of Accelerated Pavement Testing (APT) as a

research tool is that the performance of pavement materials and structures can be
evaluated at a reduced cost and in a short period of time. The Civil Infrastructure
Systems Laboratory (CISL) for APT at Kansas State University was established in
1997. Six fatigue pavement sections that had 100 mm thick hot mix asphalt layer
were constructed for APT in CISL. The sections had the same base and subgrade
materials. The sections were instrumented using strain gauges and pressure cells to
measure pavement responses under APT loading. They were loaded with a 100 kN
single axle load at 20 °C. The main objective of this paper is to investigate the
effects of binder content, binder grade, and mixture nominal maximum aggregate
size (NMAS) on the mechanistic responses from the experimental sections. Stress,
strains, deflections, and profile data were collected periodically. Normalized falling
weight deflectometer (FWD) deflection data were used to back-calculate moduli
using the EVERCALC software. Measured pavement temperatures were used to
adjust back-calculated moduli to a temperature of 20 °C. The stiffer the binder, the
lower the longitudinal strain, permanent deformation, and roughness, but the higher
the transverse strain, stress on the top of subgrade in general, and back-calculated
moduli. Higher binder content results in a higher longitudinal strain, transverse

D.S. Gedafa (&)

Department of Civil Engineering, University of North Dakota,
243 Centennial Drive Stop 8115, Grand Forks, ND 58202-8115, USA
e-mail: [email protected]
M. Hossain
Department of Civil Engineering, Kansas State University, 2118 Fieldler Hall,
Manhattan, KS 66506, USA
e-mail: [email protected]
S.A. Romanoschi
University of Texas-Arlington, UTA Box 19308, Arlington, TX 76019-0308, USA
e-mail: [email protected]
M.A. Onyango
Department 2502 - Civil Engineering, University of Tennessee, 615 McCallie Avenue,
Chattanooga, TN 37403-2598, USA
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 393

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_25
394 D.S. Gedafa et al.

strain, stress on top of subgrade, permanent deformation, rut depth, and roughness.
The larger the NMAS, the higher the transverse strain, back-calculated moduli,
permanent deformation, and roughness, but the lower the longitudinal strain, stress
on the top of subgrade, and rut depth.

Keywords Fatigue cracking
Mechanistic responses
Binder grade
Binder

content

1 Introduction

The Kansas State University, in cooperation with Kansas Department of

Transportation (KDOT), developed the Civil Infrastructure Systems Laboratory
(CISL) in 1997. This indoor facility allows the performance of full-scale accelerated
tests on pavement structures using the CISL machine as the loading device.
A full-size truck axle passes over the pavement at about every five seconds,
applying a total axle load between 80 and 130 kN. In the bi-directional loading
mode, approximately 650 passes of the bogie are applied in 1 h of operation, and
about 100,000 passes in 1 week. Typically, two test pavement sections are con-
structed in each pit and loaded simultaneously with one wheel of the axle passing
over each pavement section. Single and dual tires as well as single and tandem axles
can be accommodated in this system. The operation is typically stopped for several
hours weekly for the maintenance of the loading machine and the measurement of
pavement response and performance. The facility allows the control and monitoring
of the temperature at the surface and in the pavement layers (Romanoschi et al.
2014).

1.1 Objective of the Study

This paper is based on results from CISL Experiment No. 14, which was conducted
with the aim to verify mechanistic-empirical design models for flexible pavements
through APT. The main objective of this paper is to investigate the effects of binder
content, binder grade, and nominal maximum aggregate size (NMAS) on the
evolution of mechanistic responses from ‘fatigue cracking’ sections. The responses
that were considered are the longitudinal and transverse strains at the bottom of the
hot-mix asphalt (HMA) layer, the stress on the top of the subgrade, back-calculated
moduli, permanent deformation, rut depth, and roughness of the experimental
sections.
Effects of Binder and Mix Properties … 395

2 Experimental Sections

Three Midwestern states, Iowa, Kansas, and Missouri, sponsored the construction
and instrumentation of twelve experimental pavement sections at CISL. The sec-
tions were constructed in six pairs, in pits that are 4.5 m wide, 6 m long, and 1.8 m
deep. Three pairs were ‘fatigue cracking’ sections and aimed to study the ‘fatigue
cracking’ behavior of flexible pavements. The remaining three pairs were ‘rutting’
sections and aimed to study the rutting behavior of asphalt concrete pavements. In
total, six HMA mixtures were used, two for each of the three states. One ‘fatigue
cracking’ and one ‘rutting’ pavement were built for each mix. Only fatigue cracking
sections are included in this paper for brevity. Fatigue cracking sections were
constructed in three layers: 100 mm of HMA surface layer, 150 mm of crushed
unbound granular base course (commonly known as AB-3 in Kansas), and 1.55 m
of A-7-6 clayey soil subgrade. The sections were loaded at a pavement surface
temperature of 20 °C. It should be noted that all the sections had the same base and
subgrade materials.

2.1 Asphalt Mix Designs

Material selection is a key aspect to better performing asphalt mixes. The

Departments of Transportation (DOTs) of Iowa, Kansas, and Missouri provided the
asphalt mix designs of the six asphalt mixes (two per state). A local contractor in
Manhattan, Kansas, constructed the experimental sections at CISL with materials
transported from the three states. The six asphalt mixes comprised: a Kansas coarse
mix (KS-3) with 19 mm NMAS; a Kansas fine mix (KS-4) with 12.5 mm NMAS;
two 12.5 mm NMAS Missouri mixes with different binder grades; and two
12.5 mm NMAS Iowa mixes with the same binder but slightly different fine
aggregate content and different design Equivalent Single Axle Loads (ESALs). The
Iowa sections, IA-3 and IA-4, were designed for 30 and 3 million ESALS,
respectively (Romanoschi et al. 2014). Table 1 shows mixture specifications for the
experimental sections.
The constructed sections had some mixtures that had properties that deviated
from the asphalt mix design parameters provided. The in situ binder content for the
IA-3 and IA-4 mixes was higher than the design binder content. IA-3 had higher
as-constructed binder than IA-4. However, these variations provided valuable
insights into the performance of the mixtures as a result of construction variability
(Onyang 2009; Romanoschi et al. 2014).
396 D.S. Gedafa et al.

Table 1 Mixture specification and notations

State Binder mix NMAS Binder Binder Design ESAL Notation
type (mm) grade content (%) (millions)
Iowa HMA 30M L-4 12.5 PG 64-22 7.5 30 IA-3
HMA 3M L-4 12.5 PG 64-22 7.0 3 IA-4
Kansas SM 19A 19A PG 64-22 5.6 6.4 KS-3
SM 12.5A 12.5A PG 64-28 5.2 2.6 KS-4
Missouri SP 125C 12.5 PG 70-22 5.3 3.3 MO-3
SP 125C 12.5 PG 64-22 5.4 3.3 MO-4

3 Data Collection

The experimental sections were instrumented to measure different mechanistic

responses periodically. Strain gauges were used to measure longitudinal and
transverse strains at the bottom of the HMA layers. Thermocouples were used to
measure pavement temperature at the top, at mid-depth, and at the bottom of the
HMA layers. Pressure cells were used to measure stress on the top of the subgrade.
The sections were loaded with up to 2,100,000 load repetitions of a 100 kN
single axle load. The APT machine is equipped with a temperature chamber that
allows testing at a controlled temperature within ±2.0 °C; the test temperature was
20 °C for this study. Lateral movement was provided by a lateral wandering device
that moved the entire frame of the APT machine in the lateral direction with a
maximum lateral wander of ±61 cm (Romanoschi et al. 2014).
Permanent deformation in each layer, longitudinal and transverse strains at the
bottom of the HMA layers, vertical stress on the top of the subgrade layer, and the
temperature in the HMA layers were measured periodically during the experiment.
For all sections, the APT loading started at least 2 months after their construction
(Romanoschi et al. 2008). In addition to the responses mentioned, deflection data
were collected periodically using falling weight deflectometer (FWD). Longitudinal
and transverse profile data were also collected on the same day as the strain and
stress responses.

3.1 Horizontal Strains at the Bottom of the HMA Layer

The strains were measured when the APT wheel assembly was in the middle of the
wheel path (corresponding to the “0” position of the wandering device), at a
sampling rate of 100 readings per second for four complete cycles (8 passes) of the
Effects of Binder and Mix Properties … 397

APT bogie. Data was recorded and stored in spreadsheet format together with the
data for the axle magnitude and position (Romanoschi et al. 2009).

3.2 Mid-Depth Pavement Temperature

Four thermocouples were placed in each pavement structure at the center of each
lane. Two sensors were placed at the bottom of the asphalt concrete layer and two at
the bottom of the base layer.

3.3 Transverse Profiles

Transverse profile measurements were performed periodically at the same time

intervals as the longitudinal profile and strain/stress. On each pair of pavements,
transverse profiles were measured at three different spatial locations. Each profile
consists of elevation data at 210 points spaced at 12.5 mm intervals.

3.4 Longitudinal Profiles

The longitudinal profiles of the experimental sections were recorded by measuring

the elevations at 19 points spaced at 30.5 cm intervals longitudinally on the outside
wheel path with surveying equipment.

4 Data Analysis

4.1 Back-Calculated Moduli

Back-calculation was done with a multilayered linear-elastic program. All pave-

ment sections were modeled as three-layer systems. The back-calculation program
EVERCALC was used.
EVERCALC is a linear elastic analysis back-calculation program developed by
the Washington State Department of Transportation. It uses WESLEA as the
response analysis program, a complex integration algorithm based on Wendle’s rule
and a nonlinear least squares optimization technique (Chen et al. 1999). It is capable
of evaluating HMA pavement structures containing up to five layers. The program
uses an iterative approach to find a set of moduli that would provide a calculated
398 D.S. Gedafa et al.

deflection basin closest to the measured deflection basin as characterized by the

root-mean-square (RMS) error (Zhou et al. 1997).
The program requires a set of seed modulus values to start and adopts Newton’s
method to search for a set of deflections that best match the measured deflections. It
uses regression equations to determine seed moduli for HMA pavements with up to
three layers. The seed moduli must be user defined when the number of layers
exceeds three.
Chen et al. (2000) developed Eq. (1) based on deflections from intact locations.
Equation has been used to correct back-calculated moduli to 20 °C based on
measured mid-depth pavement temperature.

E TC
ETW ¼ h i ð1Þ
ð1:8TW þ 32Þ 2:4462
ð1:8TC þ 32Þ2:4462

where ETW = adjusted modulus of elasticity at TW (MPa); ETC = measured modulus

of elasticity at TC (MPa); TW = temperature to which the modulus of elasticity is
adjusted (°C); and TC = mid-depth temperature at the time of FWD data collection
(°C).

4.2 Transverse Profiles

Permanent deformation and rut depth were derived from the transverse profiles.
Permanent deformation and rutting are sometimes used interchangeably since they
are both depressions that occur in the wheel path after traffic loading; the difference
is due to the measurement method. Permanent deformation is measured as the depth
of depression with reference to the original profile, whereas rutting is measured
using a straight edge as the depth between the highest and the lowest points on a
profile.
Permanent deformation at the pavement surface was calculated first in each of
the 105 (210/2) points of the profile by subtracting measured elevation after a given
number of passes from the initial elevation. The permanent deformation was pos-
itive when the current elevation of the point was lower than the initial elevation.
Then, for each pavement, and for a particular transverse profile, the permanent
deformation was computed as the maximum value obtained from the 105 points.
Factors contributing to the permanent deformation in asphalt mixes include binder,
aggregates, and HMA (Monismith et al. 1994). The effects of binder content, binder
grade, and NMAS on mechanistic responses from the experimental sections have
been investigated in this paper.
Rut depth (RD) for each pavement, for a particular transverse profile, was
computed as the difference between the elevation of the highest and lowest points of
that profile.
Effects of Binder and Mix Properties … 399

4.3 Longitudinal Profiles

The roughness of the longitudinal profile was estimated from the elevation data
using the Slope Variance (SV) as the roughness statistic. The SV was selected
because of its simplicity. Other indices that are computed based on elevation data
require a minimum length of pavement section; for example, to compute the
International Roughness Index (IRI), the road section must be at least 10 m long.
The SV was computed using Eq. (2). It is important to note that SV was computed
only to compare its evolution for the experimental sections under study.
 2
SUM Si  Savg 100ðhi þ 1  hi Þ
SV ¼ and Si ¼ ð2Þ
N 1 d

where N = number of segments where the slope is computed (N = 18 for the CISL
sections); Si = slope in point i, in percent; h = elevation (in.); and d = spacing
between points (d = 12 in.).

5 Results and Discussions

5.1 Horizontal Strains at the Bottom of the HMA Layer

5.1.1 Longitudinal Strain

Figure 1 shows the longitudinal strain for the experimental sections. The higher the
binder content, the higher the longitudinal strain for the Iowa sections. The mea-
sured longitudinal strain for the Kansas section, KS-3, with a larger NMAS is lower
than the one with a smaller NMAS, KS-4. The stiffer the binder, the lower the
magnitude of the longitudinal strain for the Missouri sections except the first few
number of passes. The effect of time or number of passes on the evolution of
longitudinal strain is not clear.

5.1.2 Transverse Strain

The higher the binder content, the higher the transverse strain at the bottom of
HMA layer based on the Iowa sections as shown in Fig. 2. The measured transverse
strain on the SM-19A section (KS-3) is higher than the one on the SM-12.5A
(KS-4). The softer the binder, the lower the magnitude of the transverse strain in
general, as can be seen from the results for the Missouri sections. This confirms that
softer binders, when used for layers with the same pavement thickness, can result in
pavements with longer fatigue life. The effect of time or number of passes on the
evolution of transverse strain is not clear except for Iowa sections where the strains
increased.
400 D.S. Gedafa et al.

IA-3 IA-4 KS-3 KS-4 MO-3 MO-4

500
Longitudinal Strain (microstrain)

450
400
350
300
250
200
150
100
50
0
0 5 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 200
Load Repetitions (x10,000)

Fig. 1 Evolution of longitudinal strains

IA-3 IA-4 KS-3 KS-4 MO-3 MO-4

200
Transverse Strain (microstrain)

175

150

125

100

75

50

25

0
0 5 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200
Load Repetitions (x10,000)

Fig. 2 Evolution of transverse strains

5.2 Vertical Stresses on Top of the Subgrade

Vertical compressive stresses on top of the subgrade were measured in each

pavement structure at two locations. The average of these locations corresponding
to the “0” position has been presented in Fig. 3. In general, there is an increase in
Effects of Binder and Mix Properties … 401

IA-3 IA-4 KS-3 KS-4 MO-3 MO-4

180
Stress on the Top of Subgrade (kPa)

160

140

120

100

80

60

40

20

0
0 5 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200
Load Repetitions (x10,000)

Fig. 3 Evolution of stress on top of the subgrade

the stress on top of the subgrade as the number of passes increases. The higher the
binder content, the higher the stress on the top of subgrade based on Iowa sections.
The higher the NMAS, the lower the stress on the top of subgrade based on Kansas
sections. The stiffer the binder, the higher the stress on the top of subgrade based on
Missouri sections in general.

5.3 Back-Calculated Moduli

The effect of binder content on back-calculated moduli is not clear since the trend
changed over time as shown in Fig. 4. The larger the NMAS in a mix, the higher the
back-calculated moduli (SM-19A in KS-3 vs. SM-12.5A in KS-4) at all numbers of
passes. The stiffer the binder, the higher the back-calculated moduli (PG 70-22 in
MO-3 vs. PG 64-22 in MO-4). There is no clear trend whether there is an increase or
a decrease in back-calculated moduli as the number of passes increases.

5.4 Transverse Profiles

5.4.1 Permanent Deformation

Figure 5 shows the evolution of permanent deformation for the experimental sec-
tions. As can be expected, there is an increase in permanent deformation as the
402 D.S. Gedafa et al.

IA-3 IA-4 KS-3 KS-4 MO-3 MO-4

5000
4500
Back-calculated Moduli (MPa)

4000
3500
3000
2500
2000
1500
1000
500
0
0 2.5 5 10 20 200
Load Repetitions (x10,000)

Fig. 4 Evolution of back-calculated HMA layer moduli

IA-3 IA-4 KS-3 KS-4 MO-3 MO-4

10

9
Permanent Deformation (mm)

0
5 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200
Load Repetitions (x10,000)

Fig. 5 Evolution of permanent deformation

number of passes increases in general. The highest permanent deformation was

measured on IA-1, which was the section with the highest as-constructed binder
content. Kansas section with SM-12.5A shows a lower permanent deformation than
a section with SM-19A at all pass intervals except one. Also the Missouri section
Effects of Binder and Mix Properties … 403

IA-3 IA-4 KS-3 KS-4 MO-3 MO-4

35

30

25
Rut Depth (mm)

20

15

10

0
0 5 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200
Load Repetitions (x10,000)

Fig. 6 Evolution of rut depth

with a stiffer binder (MO-3) performed better than the section with a softer binder
(MO-4) except at one point.

5.4.2 Rut Depth

The highest rut depth was measured in the section with the highest as-constructed
binder content (IA-3), as shown in Fig. 6. This confirms that increasing binder
content has a negative impact on the rut resistance of pavements. The larger the
NMAS, the lesser the rut depth as the number of passes increased. There is no
significant increase in the rut depth as the number of passes increases except with
Iowa sections and Kansas section with lower NMAS. Iowa sections failed in rutting
and the tests for other sections was stopped after 2,000,000 passes for the sake of
time since the sections did not fail.

5.5 Longitudinal Profiles

There is no clear trend as to whether there is an increase in roughness with an

increase in the number of passes, as shown in Fig. 7. The higher the as-constructed
binder content, the higher the roughness of the Iowa section. This was due to faster
evolution of permanent deformation on this section. The Kansas section with a
smaller NMAS is smoother than the one with larger NMAS. The stiffer the binder,
the smoother the section as can be seen from Missouri sections.
404 D.S. Gedafa et al.

IA-3 IA-4 KS-3 KS-4 MO-3 MO-4

3.5

3.0

2.5
Slope Variance

2.0

1.5

1.0

0.5

0.0
0 5 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200
Load Repetitions (x10,000)

Fig. 7 Evolution of roughness

6 Conclusions

The effects of binder grade, binder content, and mixture NMAS on the evolution of
mechanistic responses on a number of APT test sections have been investigated.
The responses considered were the strains (longitudinal and transverse) at the
bottom of the HMA layers, stress on the top of the subgrade, back-calculated
moduli, transverse profiles (permanent deformation and rut depth), and longitudinal
profile (roughness). Based on this study, the following conclusions can be drawn.
• The stiffer the binder,
– the higher the transverse strain, stress on the top of subgrade in general, and
back-calculated moduli.
– the lower the longitudinal strain, permanent deformation, and roughness.
• The higher the binder content, the higher the longitudinal strain, transverse
strain, stress on top of subgrade, permanent deformation, rut depth, and
roughness.
• The larger the NMAS,
– the higher the transverse strain, back-calculated moduli, permanent defor-
mation, and roughness.
– the lower the longitudinal strain, stress on the top of subgrade, and rut depth.
• As the number of passes increases, there is an increase in the stress on the top of
subgrade and permanent deformation in general.
Effects of Binder and Mix Properties … 405

Acknowledgments This study was funded by the Midwest States Accelerated Pavement Testing
Pooled Funds Program. The participants included the state Departments of Transportation of Iowa,
Kansas, and Missouri.

References

Chen, D. H., Bilyeu, J., Lin, H. H., & Murphy, M. (2000). Temperature correction on
falling-weight deflectometer measurements. Transportation Research Record: Journal of the
Transportation Research Board, No. 1716, Transportation Research Board of the National
Academies, Washington, DC (pp. 30–39).
Chen, J., Hossain, M., & Latorella, T. M. (1999). Use of falling weight deflectometer and dynamic
cone penetrometer in pavement evaluation. Transportation Research Record: Journal of the
Transportation Research Board, No. 1655, Transportation Research Board of the National
Academies, Washington, DC (pp. 145–151).
Monismith, C. L., Hicks, R. G., Finn, F. N., Sousa, J., Harvey, J., Weissman, S. et al. (1994).
Permanent deformation response of binder aggregate mixes. SHRP-A-415, Strategic Highway
Research Program of the National Research Council, Washington, DC.
Onyang, M. A. (2009). Verification of mechanistic prediction models for permanent deformation
in binder mixes using accelerated pavement testing. Ph.D. Dissertation, Kansas State
University, Manhattan, KS.
Romanoschi, S. A., Dumitru, C., Lewis, P., & Hossain, M. (2009). Accelerated testing for studying
pavement design and performance (FY 2004): Thin bonded rigid overlay on PCCP and HMA
(CISL Experiment No. 13). FHWA-KS-08-8, Kansas Department of Transportation, Topeka, KS.
Romanoschi, S. A., Lewis, P., Dumitru, O., & Banda, S. (2008). Accelerated testing for studying
pavement design and performance (FY 2003): Evaluation of the chemical stabilized subgrade
soil (CISL Experiment No. 12). FHWA-KS-07-8, Kansas Department of Transportation,
Topeka, KS.
Romanoschi, S., Lewis, P., Gedafa, D., & Hossain, M. (2014). Verification of
mechanistic-empirical design models for flexible pavements through accelerated pavement
testing. FHWA-KS-14-02 Kansas Department of Transportation, Topeka, KS.
Zhou, H., Rada, G. R., & Elkins, G. E. (1997). Investigation of back-calculated moduli using
deflections obtained at various locations in a pavement structure. Transportation Research
Record: Journal of the Transportation Research Board, No. 1570, Transportation Research
Board of the National Academies, Washington, DC (pp. 96–107).
Evaluating Resistance of Hot Mix Asphalt
Overlays to Reflective Cracking Using
Geocomposites and Accelerated Loading

Mansour Solaimanian, Ghassan Chehab and Marcelo Medeiros

Abstract Placing a structural asphalt overlay atop existing pavements is one of the
conventional methods used in pavement rehabilitation. However, reflective crack-
ing in the new overlay continues to be a challenge associated with pavement
rehabilitation. Traffic loading and environmental effects are the primary external
causes of reflective cracking. A properly designed geocomposite or geogrid may
help delay or reduce reflective cracking by providing reinforcement and strain
relief. This study was conducted to evaluate the effectiveness of using such a
geocomposite interlayer to improve pavement performance, specifically in terms of
resistance to reflective cracking. Asphalt concrete was placed on the top of concrete
slabs. The underlying concrete slab was placed in two different thicknesses, with a
space between slabs to simulate the joint. The thinner slab was placed on a rubber
mat to deliver uniform thickness and flat surface for placement of asphalt. Two
slabs were made, one as the control system without a geocomposite interlayer, and
the other as the experimental system with a geocomposite interlayer. Accelerated
loading of the asphalt overlay was conducted using a 3rd Scale Model Mobile Load
Simulator (MMLS3) system. The bottom-up reflective cracking could not be
developed during this test in neither the control nor the experimental slabs.
However, it was clearly observed that, the geocomposite significantly improved the
top-down cracking resistance of the asphalt overlay.

Keywords Accelerated testing
Asphalt concrete
Reflective cracking
Load

simulator

M. Solaimanian (&)
Penn State University, 221 Transportation Research Building, University Park,
PA 16802, USA
e-mail: [email protected]
G. Chehab
American University of Beirut, Beirut, Lebanon
e-mail: [email protected]
M. Medeiros
University of New Hampshire, Durham, USA
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 407

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_26
408 M. Solaimanian et al.

1 Introduction

Structural and functional rehabilitation of roads using asphalt overlays is very

common. In many cases, such overlays are placed on the top of an old jointed
concrete or an old asphalt surface which has already developed cracks of various
types. A major problem with construction of such overlays is reflective cracking,
which develops in the newly paved asphalt within the first 2 years after placement.
Reflective cracking involves the development of cracks in a new overlay that mirror
the cracks and/or joints in the old, existing pavement. Intensity of traffic loading,
harshness of the environment, temperature variations, and pavement structural
conditions such as stiffness and thickness determine the rate of crack development.
Many geogrid and geotextile manufacturers offer products as pavement interlayers
to delay reflective cracking. Some of these geotextiles are laid in a tacking agent
prior to application of the overlay. Some manufacturers offer self-adhesive rein-
forcing mesh or grid containing fiberglass strands coated with polymer modified
bitumen. Evidence exists that a geocomposite interlayer combining a high-modulus
geogrid with a lightweight, non-woven geotextile can be effective in reduction or
delaying development of reflective cracks. It is believed that such layer of geo-
composite, once properly installed, serves its purpose by providing reinforcement
and strain relief.
Most of the available research indicates that utilizing any type of stress
absorbing membrane (SAMI) improves the crack resistance but to various degrees.
Study by Schultz and Peters (1987) indicated that 39 % of original cracks reflected
to the surface for the asphalt rubber SAMI test sections compared with 80 % for the
control section without SAMI. Using model scale testing, Ogundipe et al. (2012)
demonstrated that SAMI made with glass fiberglass sandwiched between emulsion
layers and covered with aggregate significantly delayed reflection of cracks in
asphalt overlay.
Several researchers have investigated the benefit of geosynthetic and geocom-
posite layers in delaying reflective cracks. From review of literature it appears that
most of the past investigation in this regard has been through field studies. Some
researchers have reported a combination of field and laboratory work. For example,
Pasquini et al. (2005) report of a lab and field experiment using geocomposites
where they found significant improvement in resistance to reflective cracking. The
researchers laboratory work included 95-mm cylindrical specimens and rectangular
slabs with dimensions of 285 × 90 × 80 mm. The work by Hughes and Somers
(2000) using three reinforcing systems (geogrid, paving fabric, and fiber-glass grid)
indicated a similar performance of reinforced sections in comparison with the
control section, but most field studies are indicative of reduction or delay in
reflective cracking as a result of using paving fabrics. One of the most compre-
hensive reviews was conducted by Carmicheal and Marienfeld (1999) who eval-
uated 30 sites and their general conclusion was that reduction of reflective cracking
was observed in 26 out of these 30 sites. Cleveland et al. (2003) conducted a
theoretical/experimental analysis of the benefits of geocomposites in reducing the
Evaluating Resistance of Hot Mix Asphalt Overlays … 409

reflective cracking. The authors cite existence of many complex interactions

between fracture properties of the overlay and the geosynthetic material, making
such analysis difficult. However, they do mention that based on their analysis and
experiment using Texas overlay tester grids and composites perform better than
fabrics in delaying reflection of cracks. While considerable information can be
found in the literature about field performance of geogrids and various types of
paving fabrics, laboratory scale information with accelerated loading is very lim-
ited, and hence the reason for conducting this research.

2 Objective

The present study was conducted to evaluate the effectiveness of using a geo-
composite interlayer to improve pavement performance, specifically in terms of
resistance to reflective cracking. Performance of asphalt concrete overlaying
hydraulic cement concrete was evaluated, with and without the inclusion of a
geocomposite interlayer.

3 Experimental Program

The experiment included building two slabs of asphalt concrete over hydraulic
cement, in a specific configuration, and loading them with a model scale accelerated
testing equipment. The control slab did not include any reinforcement. The
experimental slab included a layer of geocomposite in the middle of asphalt layer.
Through a laboratory experiment using bending beams, Khodaii et al. (2009)
concluded that if reinforcement is placed closer to the middle of the asphalt overlay,
compared with placement at the bottom of overlay, less permanent deformation will
be collected in the overlay.
Testing temperature was targeted at 20 °C. This temperature was selected to
expedite crack initiation. Testing was conducted under dry conditions (i.e., no water
was induced into the system). Furthermore, no wandering was allowed and traf-
ficking was channelized to create more aggressive loading.

3.1 Configuration of Slabs

Figures 1 and 2 show the cross-sectional and dimensional plots of the two slabs
made for this experiment. The completed slab was 16 in. wide and 58.5 in. long.
The picture in Fig. 3 shows one of the prepared slabs made with this configuration.
The underlying concrete slab consisted of two adjacent sections (slabs). One section
was 16 in. wide and 3 in. thick, placed directly on the firm ground. The other
410 M. Solaimanian et al.

Fig. 1 Setup of the slab with geocomposite placed in between asphalt overlay

Fig. 2 Setup of the control slab

Fig. 3 One of the slabs prepared for this study. Clockwise from top left Neoprene rubber on the
left and the 3” thick concrete slab to the right; 2” in. and 3” thick concrete slabs next to each other;
Tack coated slabs; completed compacted and asphalt mix
Evaluating Resistance of Hot Mix Asphalt Overlays … 411

Fig. 4 Side view of the slab showing the asphalt concrete over hydraulic concrete and rubber mat

section was 16 in. wide and 2 in. thick, with a 1-in. thick neoprene rubber pad
underneath. The idea supporting this set-up was to aggravate vertical movement at
the joint expediting reflective cracking. Such condition could occur in real pave-
ment where load transfer efficiency of the slabs at the joint is reduced, resulting in
vertical differential movement between adjacent slabs. Differential movement at the
joint could also induce higher shear stresses at the asphalt section right on the joint.
The joint under the asphalt concrete layer is noticeable in the picture of Fig. 4.
Compaction of asphalt mix over the concrete slab was achieved with the use of a
vibratory plate compactor.

4 Materials

Materials used can be classified into two major groups: those for the asphalt
overlays and those for constructing the concrete slabs serving as the base of the
asphalt overlay. In addition, for each testing configuration, rapid setting tack coat
was applied to the surface of each concrete slab at least 2 h before placement of the
asphalt layer over the concrete slab. Tacking was applied to achieve better bonding
between the concrete slabs and the asphalt overlay.
The materials for the 2-in. thick asphalt overlay were hot-mix asphalts
(HMA) procured from local asphalt mixing plants where local aggregate batches
were mixed with a PG64-22 asphalt binder. The mix was a Superpave 9.5 mm mix
412 M. Solaimanian et al.

Table 1 Mix design for the concrete slab

Dimensions of the concrete slab 29” × 16” × 2” and 29” × 16” × 3”
Volume of 2” thick slab 928 in3 (0.01989 yd3)
Volume of 3” thick slab 1392 in3 (0.02984 yd3)
3
For 1 Yd of concrete 2-in. thick slab 3-in. thick slab
Water (lbs) 344.8 6.9 10.3
Cement (lbs) 666.7 13.3 19.9
Coarse aggregate (lbs) 1645.2 32.7 49.1
Fine aggregate (lbs) 1180.2 23.5 35.2

with 5.7 % asphalt content designed for traffic level between 0.3 to 3 million
equivalent single axle loads (ESALs). The concrete slabs in this study were con-
structed with four ingredients: water, cement, coarse aggregates, and fine aggre-
gates. Proportion of ingredients for making concrete slabs are presented in Table 1.
All slabs were cured for at least 28 days in order to achieve maximum strength
before asphalt was placed and compacted on the top of the slab. The geocomposite
used in this experiment was bituminous coated high modulus polyester combined
with ultra-lightweight nonwoven geotextile on one side to provide a strong bond
with asphalt. The rubber material under one of the concrete slabs was neoprene with
Shore hardness of 60.

4.1 Testing Equipment

The One-Third Scale Model Mobile Load Simulator (MMLS3) testing equipment
used for the project was comprised of three parts: the MMLS3 loading unit, the air
conditioner, and the electronic profilometer. The MMLS3, shown in Fig. 5, is an
accelerated pavement testing device that applies unidirectional trafficking to the
pavement in a controlled laboratory environment or on full-scale pavements in the
field. The MMLS3 has four tires, each with diameter of 12.0 in. and width of
3.1 in.. The actual wheel path generated by the MMLS3 is about 54 in. long. The
load exerted by each wheel of the MMLS3 is 225 lbf with a corresponding tire
pressure of approximately 90 psi. The traffic speed was set to 7200 axles (wheels)
per hour, or two axles (wheels) per second, delivering a moving speed of
approximately 5 mph. Total of 400,000 cycles were applied to each slab.
The air conditioner and environmental chamber in Fig. 6 were used to maintain
the temperature inside the chamber while testing. The air conditioner can send cool
or hot air at a specified temperature into the insulated environmental chamber
through two air hoses. A thermocouple connected to the air conditioner on one side
can measure the air temperature inside the chamber and send feedback to the air
conditioner to adjust the air temperature.
Evaluating Resistance of Hot Mix Asphalt Overlays … 413

Fig. 5 MMLS3 loading of one of the slabs for this study

Fig. 6 The air conditioner and environmental chamber for temperature control
414 M. Solaimanian et al.

5 Results and Discussion

5.1 Temperature Challenge

This experiment was intended to be conducted in a controlled temperature envi-

ronment. The target temperature was selected as 20 °C. Maintenance of this precise
temperature at all times was not achieved, but the temperature was kept within 18–
22 °C most of the time.

5.2 Asphalt Concrete Air Void and Strength

Compaction of asphalt concrete over the concrete slab was achieved with the aid of
a vibratory plate compactor. Three cores were taken after completion of the loading
the slabs to determine air void and strength. The air void varied in the range of 10–
12 % implying the difficulty that existed with compaction of laboratory slabs using
a plate compactor. The average indirect tensile strength of the cores was 85 psi with
average 3.4 % vertical strain at failure.

5.3 Crack Observation

The asphalt overlay was investigated for signs of distress and cracking at specific
cycle intervals. The control setup without using geocomposite exhibited top-down
cracking at the vicinity of the joint after experiencing 20,000 trafficking cycles,
while the experimental setup with geocomposite in the overlay showed top-down
cracking at similar position after 150,000 cycles (Fig. 7). The crack width for both
setups was almost the same, and approximately 0.3–0.4 mm. Figures 8 and 9
exhibit the surface condition of the slabs after 400,000 cycles of trafficking. This
observation of top down cracking is consistent with field observations by Nunn
(1989) who indicated an overwhelming evidence of reflection cracks initiated at the
surface of the asphalt layer rather than reflecting from the bottom. These cracks start
from the top of the pavement but are aligned with the crack in the underlying layer
(Fig. 10).
The bottom up cracking could be developed neither in the control slab, nor in the
experimental slab within the number of cycles conducted in this study. Probably, a
significantly higher number of load repetitions would have been needed to see
bottom-up reflective cracking in the asphalt layer.
Evaluating Resistance of Hot Mix Asphalt Overlays … 415

Fig. 7 Comparison of the number of cycles which initiated top-down reflective cracking in the
slabs

6 Summary and Conclusions

A study was undertaken to investigate the impact of using geocomposite to reduce

reflective cracking in asphalt concrete overlays. Slabs were made with asphalt
concrete over hydraulic cement concrete and rubber mat. The underlying concrete
slabs were used at two different thickness levels, with the thinner side sitting on a
neoprene foundation. In the control slab, the 2-in. asphalt overlay was placed
without the geocomposite interlayer. In the experimental slab, the 2-in. asphalt
overlay included a geocomposite interlayer. A 0.5-in. spacing was created between
the underlying concrete slabs. The slabs for this testing resided on the laboratory
concrete floor. Loading the asphalt overlay was conducted using an MMLS3 sys-
tem. Each slab experienced 400,000 cycles of MMLS 3 tire traffic load. Neither slab
exhibited bottom-up reflective cracking up to the 400,000 load cycles applied.
However, it was clearly observed that under the first configuration, the geocom-
posite significantly improved the top-down cracking resistance of the asphalt
overlay, as the system with no geocomposite manifested initiation of such cracking
at 20,000 cycles and the system with geocomposite showed signs of top-down
cracking at 150,000 cycles.
416 M. Solaimanian et al.

Fig. 8 Crack at the middle joint position for the control setup after 20,000 trafficking cycles

Fig. 9 First appearance of the crack in the experimental setup (with geocomposite) after 150,000
cycles
Evaluating Resistance of Hot Mix Asphalt Overlays … 417

Fig. 10 Crack at the middle joint position for the experimental setup (with geocomposite) after
400,000 trafficking cycles

Acknowledgments The financial support of Mid-Atlantic University Transportation Center and

the Huesker, Inc. is greatly appreciated. Special thanks go to Mr. Scott Milander, NECEPT
laboratory coordinator, who led the laboratory efforts in conducting the accelerated testing.

References

Carmichael, R. F., & Marienfeld, M. L. (1999). Synthesis and literature review of nonwoven
paving fabrics performance in interlayers. Transportation Research Record,
No. 1687 (pp. 112–124).
Cleveland, G. S., Lytton, R L., & Button J. W. (2003). Reinforcing benefits of geosynthetic
materials in asphalt concrete overlays using pseudo strain damage theory. Transportation
Research Record No. 1849, Transportation Research Board, National Research Council,
Washington, DC (pp. 201–211).
Hughes, J. J., & Somers, E. (2000). Geogrid mesh for reflective crack control in bituminous
overlays. Final Report, Report No. PA 200-013-86-001, Pennsylvania Department of
Transportation.
Khodaii, A., Fallah, S., & Moghadas Nejad, F. (2009). Effect of geosynthetics on reduction of
reflection cracking in asphalt overlays. Geotextiles and Geomembrances, 27, 187–195.
Nunn, M. E. (1989). An investigation into reflection cracking in composite pavements. In
Proceedings of 1st Conference on Reflection Cracking in Pavement (pp. 146–153), Liege,
Belgium.
418 M. Solaimanian et al.

Ogundipe, O. M., Thom, N. H., Collop, A. C., & Richardson, J. (2012). Performance of ‘SAMI’S
in simulative testing. In: 7th RILEM International Conference on Cracking in Pavements (Vol.
4, pp. 93–102).
Pasquini E., Pasetto, M., & Canestrari, F. (2005). Geocomposites against reflective cracking in
asphalt pavements: Laboratory simulation of a field application. Road Materials and Pavement
Design. Online publication, May.
Schultz, R. L., & Peters, A. J. (1987). Asphalt_rubber binder stress absorbing membrane
interlayer. Final Report, Report Number WA-RD 128.1, Washington State Department of
Transportation, July.
Optimum Tack Rate for Hot-Mix
Asphalt Bonding

A. Sufian, M. Hossain and G. Schieber

Abstract Construction of hot-mix asphalt (HMA) pavement involves spraying

tack coat on an existing surface in order to promote an acceptable interlayer bond
and to ensure that multiple pavement layers behave monolithically. Insufficient tack
coat application has been linked to premature cracking failure, a major factor in
reduced HMA overlay life in Kansas. The Kansas Department of Transportation
(KDOT) currently specifies that a slow-setting polymer-modified emulsion
(SS-1hP) tack coat be applied at a rate of 0.23 L/m2 (0.05 gal/yd2) for overlays.
This study utilized accelerated pavement testing (APT) to determine the optimum
rate of SS-1hP tack material. At the APT facility at Kansas State University, an
existing pavement pit (20 ft long × 14 ft wide) was milled and cleaned with a
broom and compressed air. The pit was then divided into four test sections with
varying rates of SS-1hP (50, 100, 160, and 240 % of 0.05 gal/yd2). Inlay paving
was done with a 12.5 mm (0.5 in.) Nominal Maximum Aggregate Size Superpave
mixture. An 89 kN (20 kip) single-axle load was bi-directionally applied over a
6.1 m (20 ft) long lane at approximately 11.3 km/h (7 mph). Rut depth and strain
measurements at the overlay interface were done at periodic intervals during
application of 550,000 load repetitions. Results showed that the section with 50 %
(of 0.05 gal/yd2) SS-1hP tack coat had higher rutting when compared to other
sections.

Keywords HMA
Bond
Tack coat
Emulsion
APT

A. Sufian (&)
M. Hossain

Department of Civil Engineering, Kansas State University, Manhattan, KS 66506, USA
e-mail: sufi[email protected]
M. Hossain
e-mail: [email protected]
G. Schieber
Kansas Department of Transportation, 700 SW Harrison, ESOB 7th Floor, Topeka,
KS 66603, USA
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 419

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_27
420 A. Sufian et al.

1 Introduction

Hot-mix asphalt (HMA) pavement is typically laid in lifts that vary in thickness
according to maximum aggregate size of the mixture, layer thickness, etc. These
lifts may be compacted over spans of days or months. Upon completion of con-
struction, all these lifts are expected to be monolithic and collaboratively carry
applied traffic loads. Therefore, a tack coat is typically placed between pavement
lifts in order to bond these layers together. The most commonly used tack coat is
asphalt emulsion. Asphalt emulsion is produced in a colloid mill where hard asphalt
cement is sheared into smaller particles and an emulsifying agent is used to suspend
naturally neutral asphalt particles with a charge that prevents these particles from
coalescing. When an emulsion is sprayed over an existing surface, water in the
emulsion evaporates in a process known as “breaking,” leaving a thin film of
asphalt layer on the surface that acts as a tack coat (Brown et al. 2009).

2 Study Objective and Approach

This study focused on accelerated pavement testing (APT) of HMA overlay

interface bond on an existing HMA pavement for varying application rates of
SS-1hP. The objective was to find the optimum emulsion application rate using
APT. An existing asphalt pavement was milled, and four overlay test sections were
paved with four tack rates for SS-1hP. Repetitive loads were applied using an
89 kN (20 kip) single-axle load. The sections were monitored for rutting and
cracking at various loading intervals.

3 Problem Statement

Inadequate bond between HMA pavements lifts or between an HMA pavement

overlay on an existing HMA pavement results in decreased structural capacity. The
absence of a specified tack coat may lead to premature failure in the form of
debonding, mat slippage, and potential fatigue cracking, thereby reducing pavement
life (Weston 2003). The Kansas Department of Transportation (KDOT) primarily
specifies a slow-setting, polymer-modified emulsion (SS-1hP) for tack coat, with a
recommended application rate of 0.23 L/m2 (0.05 gal/yd2). KDOT has recently
identified premature cracking on overlay projects and is investigating if this
cracking may be attributed to lack of bond between the overlay and existing
pavement. Therefore, an optimal tack rate must be determined in order to ensure
adequate bond.
Optimum Tack Rate for Hot-Mix Asphalt Bonding 421

4 Accelerated Pavement Testing of HMA Interface Bond

The HMA interface bond has been extensively studied since the early 2000s.
A majority of the studies have been dedicated to developing a test procedure that
can evaluate interface bond in a laboratory setting. Only three studies have been
conducted with APT. In the late 1990s, Caltrans studied the performance of asphalt
layers with and without a tack coat in a series of experiments with a Heavy Vehicle
Simulator (HVS) (Harvey et al. 2000). Results showed a 10- to 45-fold increase in
load-carrying capacity of bonded sections (with a tack coat) compared to unbonded
(no tack coat) sections. The Illinois Center for Transportation studied interface
bonding between HMA overlays and Portland cement concrete (PCC) pavement
under APT (Leng et al. 2010; Al-Qadi et al. 2009). Results showed that SS-1hP and
PG 64-22 binder as tack coats offered higher rut resistance compared to cutback
asphalt. The milled surface showed lower rutting compared to a transverse-tined
and smooth PCC surface. Cleanliness of the PCC surface and a uniform tack coat
application rate were shown to significantly affect the interface bond strength.
A research project at Kansas State University studied tack coat for rehabilitation
of pavements using various rates of SS-1hP and Emulsion Bonding Liquid
(EBL) on six test sections under APT (Mealiff 2014). Sections with EBL showed
lower rut depths than the SS-1hP sections though cracking was similar on these test
sections after 400,000 repetitions of a 80 kN (20 kip) load. Cracks observed on the
test sections were confirmed to be top-down cracking.

5 APT Facility at Kansas State University

Kansas State University (KSU) has a 650 m2 (7000 ft2) APT facility at the Civil
Infrastructure Systems Laboratory (CISL). The CISL contains two rectangular pits
(south and middle) measuring 4.88 m (16 ft) × 6.10 m (20 ft) and one rectangular
pit (north) measuring 4.27 m (14 ft) × 6.10 m (20 ft). The pits are 1.83 m (6 ft)
deep and underlaid by a reinforced concrete layer to prevent intrusion of ground-
water or soils into the subsurface layer. The north pit contained a milled, crack-free
asphalt pavement that was overlaid with a 50 mm (2 in.) HMA overlay with
varying rates of SS-1hP as the tack coat for this study.
The APT load assembly hydraulically applies loading to a single axle with dual
tires as the axle moved across the test pit (6.1 m or 20 ft), traveling at approxi-
mately 11.3 km/h (7 mph) (Fig. 1a). A protective housing insulates this load
assembly and a built-in air conditioner maintains any given temperature. This
custom-built machine also applies wander up to ±150 mm (±6 in.) with a truncated
normal distribution in order to simulate realistic load application (Fig. 1b). When
operating without interruption, 100,000 passes of this machine are possible per
week (Lewis 2008).
422 A. Sufian et al.

Fig. 1 a CISL APT load assembly, b CISL APT wheel wander distribution

6 Description of the APT Study

6.1 Test Section Construction

This APT project, part of experiment CISL #17, was conducted on the north pit. The
pit was paved in a previous experiment, CISL #14, in order to verify
mechanistic-empirical design models for flexible pavements. The existing pavement
consisted of a 100 mm (4 in.) layer of 12.5 mm Nominal Maximum Aggregate Size
(NMAS) Superpave mixture. This layer was constructed on top of a 150 mm (6 in.)
AB-3 crushed stone base. The AB-3 base was placed on top of 1.52 m (5 ft) of A-7-6
clay subgrade. The existing HMA mix contained primarily APAC Sugar Creek rock
from Kansas City, Missouri, and 25 % Humble flint chat. In the CISL #14 experi-
ment, the pit was subdivided into two 2.13 m (7 ft) wide lanes, as illustrated in
Fig. 2. The north lane contained 5.3 % asphalt content (AC) with PG 70-22, and the
south lane had 5.4 % AC with PG 64-22. After more than 2.2 million Equivalent
Single-Axle Loads (ESALs) of loading during the CISL #14 experiment, the pave-
ment section in this pit did not show any signs of cracking (Romanoschi et al. 2014).
Therefore, this existing pavement provided a crack-free base for the current study.
A 50 mm layer was milled off the existing pavement to remove the overlay placed
during the first part of the CISL #17 experiment with SS-1hP and EBL (Mealiff
2014). The surface was then thoroughly cleaned with a power broom, and then
further cleaned with compressed air. The pit was divided into four test sections, as
shown in Fig. 2. The pit measured 6.1 mm × 4.26 m (20 ft × 14 ft), and each test
section was 2.13 m (7 ft) wide and 3.05 m (10 ft) long. Tack material was applied
using a hand-pumped pressure sprayer. Spray rate was verified by measuring the
weight of tack used with even coverage. Figure 3 shows the tack application process.
Optimum Tack Rate for Hot-Mix Asphalt Bonding 423

Fig. 2 Tack application rate test section layout

Fig. 3 Tack application

6.2 Instrumentation

Eight H-Bar strain gauges from Tokyo Sokki Kenkyujo Co. Ltd. were placed to
measure strain at the interface between the existing pavement and the HMA overlay
layer. Based on previous experience of stock lead wires degrading in hot HMA,
these wires were replaced with high-quality shielded wiring. The soldered con-
nection at the base of the strain gauge was covered by heat shielding. Each strain
gauge was epoxied to two notched aluminum bars, and the gauge and bars were
attached to the milled surface using metal staples.
424 A. Sufian et al.

Fig. 4 Installed strain gauges

Three strain gauges were installed per section, as illustrated in Fig. 4. The
gauges were covered with loose HMA mix and hand-compacted in order to protect
them from the paver.
The HMA mixture for the overlay was a 12.5 mm NMAS Superpave mixture
with 25 % Recycled Asphalt Pavement (RAP) (SR-12.5A) and PG 58-28 binder.
The mixture was placed into a 50 mm (2 in.) thick lift. The mat was compacted
with a steel-drum vibratory roller in order to reach the targeted 7 % in situ air voids.
A nuclear density gauge was used to check the density of the mat after rolling.
The test was conducted at room temperature, and no water was added to the
pavement during APT. Because the pavements were constructed in pits and the
asphalt concrete surface layer was paved wall-to-wall, the moisture content in
the subgrade soil remained relatively constant during APT. This was confirmed by
the Time Domain Reflectometry (TDR) gauges installed in the tested pavement
structure subgrade that indicated no change in volumetric moisture content in the
CISL #14 experiment (Romanoschi et al. 2014).
APT load was applied on the test section bi-directionally. Transverse profiles
were taken intermittently using a digital Linear Variable Differential Transformer
(LVDT) with a roller attachment at the end that could be mounted on a leveled
beam, as illustrated in Fig. 5. Two transverse profiles were taken for each test
section with respect to benchmarks that were epoxied to the concrete floor to ensure
a consistent location unaffected by loading. Elevation at every 12.5 mm (0.5 in.)
was measured across the entire section. The APT was run at 89 kN (20 kip) loading
in increasing increments with periodic strain measurements. The sections were also
inspected at fixed intervals for any signs of cracking. A total of 550,000 repetitions
of APT load were applied.
Optimum Tack Rate for Hot-Mix Asphalt Bonding 425

Fig. 5 APT transverse profiler

7 Results and Analysis

7.1 Rutting

Permanent deformation, the difference between identical elevation points at dif-

ferent times (or load repetitions), is illustrated in Fig. 6. The largest difference
between these identical elevation points was the reported value of permanent
deformation, which was taken because of the lack of noticeable heaving on the
sections. Figure 7 shows the progression of permanent deformations on the test
sections with loading. The SS-1hP section with 50 % of tack rate as required by
KDOT (0.05 gal/yd2) experienced the most deformation over time. The sections
with 100 and 160 % of required tack rate (0.05 gal/yd2) performed similarly with

Fig. 6 Definition of permanent deformation

426 A. Sufian et al.

Fig. 7 Permanent deformation on each test section

respect to rutting. The section with 240 % of required tack rate (0.05 gal/yd2)
performed slightly better in rutting, indicating that 0.05 gal/yd2 should be an
acceptable tack rate only if none of it is removed by construction traffic. No added
benefit of heavy tack rate was observed as it was for EBL in the first part of
CISL #17 experiment (Mealiff 2014).

7.2 Strain

In this study, strain gauges demonstrated a high failure rate. Although 12 gauges
were installed (three per section), only eight gauges were operational after con-
struction, potentially as a result of soldering issues and stress from construction
equipment. No gauges on the section with 240 % tack rate survived, probably due
to the heavy tack application shorting those gauges. Typical output from the
remaining gauges is shown in Fig. 8. Results showed that the median strain value
for a 90-s interval aligned with the baseline for that unit of time, as illustrated in
Fig. 8. Distance from the baseline to the peak was the maximum strain taken from
the pulse because, after the wheel load pass, the strain pulse somewhat returned to
the baseline. This strain was measured for an entire wander cycle, or 676 load
repetitions. The top-50 strains for a wander cycle were averaged and plotted, as
tabulated in Table 1 and illustrated in Fig. 9.
The results clearly demonstrated that the section with the KDOT-specified tack
rate (0.05 gal/yd2 of SS-1hP) showed the most consistent low-strain behavior. The
section with 160 % tack rate showed a large increase in strain after loading began
and another jump in strain after 400,000 repetitions, likely due to slippage. A large
Optimum Tack Rate for Hot-Mix Asphalt Bonding 427

Fig. 8 90-s strain response

Table 1 Microstrain values during loading

Load repetitions (×1000) Microstrain values during loading
50 % 100 % 160 % 240 %
SS-1hP SS-1hP SS-1hP SS-1hp
0 479 96 50 No strain gauges survived
10 142 82 184
25 79 70 187
50 63 60 227
100 51 66 226
200 41 49 222
425 1 61 184
450 151 71 301
550 163 80 308

Fig. 9 Strain development with load repetitions

428 A. Sufian et al.

increase in strain after 400,000 repetitions was also observed for the section with
50 % tack rate, indicating deteriorated interface on this section as well. This unu-
sual increase may indicate that these two sections will crack soon.

8 Conclusions

The following conclusions can be drawn based on results of this study:

1. APT testing indicated that SS-1hP at a recommended rate of 0.05 gal/yd2
showed good performance as a tack coat material based on the rutting on the
overlay. Strain at the interface of the overlay and the existing HMA pavement
was lowest for this rate.
2. Although EBL showed good performance as a tack coat material as far as rutting
is concerned in a previous APT experiment, similar results could not be obtained
by increasing the amount of SS-1hP.

Acknowledgments The authors wish to acknowledge the financial support for this study pro-
vided by the Kansas Department of Transportation under the Pooled Funds Program.

References

Al-Qadi, I. L., Carpenter, S. H., Leng, Z., Ozer, H., & Trepanier, J. S. (February 2009) Tack coat
optimization for HMA overlays: Accelerated pavement test. Report. No. FHWA-ICT-09-035,
University of Illinois at Urbana-Champaign.
Brown, E., Kandhal, P., Roberts, F., Lee, D. Y., & Kennedy, T. W. (2009). Hot mix asphalt
materials, mixture design, and construction. NAPA Research and Education Foundation,
Lanham, Maryland.
Harvey, J. T., Roesler, J., Coetzee, N. F., & Monismith, C. L. (2000). Caltrans’s accelerated
pavement test (CAL/APT) program-summary report six year period: 1994–2000. Report
No. FHWA/CA/RM-2000/15, Pavement Research Center, University of California, Berkeley.
Leng, Z., Al-Qadi, I. L., Carpenter, S. H., & Ozer, H. (2010). Interface bonding between hot-mix
asphalt and various portland cement concrete surfaces. Transportation Research Record:
Journal of Transportation Research Board, 2127, 20–28.
Lewis, P. (2008). Lessons learned from the operations management of an accelerated pavement
testing facility. In Proceedings of the 3rd International Conference on Accelerated Pavement
Testing, Madrid, Spain, October 2008.
Mealiff, D. (2014). Evaluation of Interface Bond of Hot-Mix Asphalt Overlays. Master’s Thesis.
Department of Civil Engineering, Kansas State University.
Romanoschi, S., Lewis, P., Gedafa, D., & Hossain, M. (2014). Verification of
mechanistic-empirical design models for flexible pavements through accelerated pavement
testing. Final Report No. FHWA-KS-14-02, Kansas Department of Transportation, Topeka.
Weston, J. (2003). Tech notes—Materials laboratory. Washington State Department of
Transportation.
Rutting Resistance of Asphalt Mixes
Containing Highly Modified Asphalt
(HiMA) Binders at the Accelerated
Pavement Load Facility in Ohio

Issam Khoury, Shad Sargand, Roger Green, Benjamin Jordan

and Paul Cichocki

Abstract Highly modified asphalt (HiMA) mixes have been proposed as a stronger
alternative to standard asphalt concrete mixes that will enable thinner perpetual
pavements in Ohio. HiMA uses a polymer-enhanced binder has a higher unit cost,
but has the potential to reduce the thickness of perpetual pavements, resulting in
savings compared to conventional HMA perpetual pavement projects. Test sections
containing HiMA in the surface, intermediate, and base layers were constructed
with total thicknesses of 20, 23, and 25 cm, created by varying the depth of the AC
treated base, in the Accelerated Pavement Load Facility (APLF) at Ohio University
in Lancaster, Ohio. An additional 28 cm section was installed as a control with a
non-HiMA base layer and HiMA surface and intermediate layers to compare with
the experimental sections. Each pavement was subjected to 10,000 passes with a
single axle load of 40 kN while the pavement was held at each of two pavement
temperatures of 21.1 and 37.8 °C. Rutting on the surface of the pavement was
measured using a rolling wheel profilometer after 100, 300, 1000, 3000, and 10,000
wheel passes at each temperature. The four sections showed no significant rutting
damage after being subjected to a total of over 20,000 passes, and remained well
below ODOT’s “low rutting” threshold of 3.2 mm. Extrapolating the rutting data
indicated the worst-performing section would reach the low rutting threshold after
more than a million passes of the load. The rutting data were compared to those
obtained in a previous perpetual pavement study in the APLF that had surface
courses including Aspha-min warm mix asphalt and a conventional hot mix asphalt,
which crossed the low rutting threshold well before 10,000 passes at the high
temperature.

I. Khoury (&)
S. Sargand
R. Green
B. Jordan
P. Cichocki

Civil Engineering Department, Ohio University, Athens, OH 45701-2979, USA
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 429

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_28
430 I. Khoury et al.

1 Introduction

1.1 Rutting in AC Pavements

Structural rutting is a permanent deformation that develops when the strength of the
pavement structure is inadequate (Newcomb et al. 2010). Rutting deformation can
manifest in one or multiple layers. Structural rutting occurs in the aggregate base or
subgrade while surface rutting occurs in the surface layer of the asphalt layer.
Rutting potential can be minimized using different techniques. Studies have
shown a threshold effect where rutting potential due to compressive strains on the
subgrade significantly decreases after a certain thickness is reached (Nunn and
Ferne 2001). Structural rutting can be prevented not only by designing a thicker
pavement structure but also by increasing the stiffness in portions of the structure as
well (Newcomb et al. 2010).

1.2 Highly Modified Asphalt (HiMA)

Highly Modified Asphalt (HiMA) utilizes a new modified binder created by Kraton
Polymers that contains a 7.5 % styrene butadiene styrene (SBS) polymer to
improve stiffness at high temperatures. HiMA mixture has also been investigated at
the National Center for Asphalt Technology (NCAT) test track near Auburn,
Alabama. Based on the promise shown in the NCAT evaluations, an experiment
was proposed at the Accelerated Pavement Load Facility (APLF) to determine if
HiMA’s improved stiffness would allow for a thinner perpetual pavement design.
The four test sections at the APLF range in thicknesses between 20 and 28 cm in
2.54 cm increments to determine where the pavement transitions from standard to
perpetual response, with the 28 cm pavement built as a control to match the depth
of the thinnest test section already built on the SHRP Test Road on US Route 23 in
Delaware County, Ohio, henceforth referred to as “DEL-23”.
More details on the pavement buildups, composition, and mechanical properties
are available in the project report (Sargand et al. 2015) and in another article
presented at this conference (Khoury et al. 2016).

2 Objectives

The objectives of the study was to document the performance of the test pavements
under controlled conditions in an accelerated pavement test facility including
10,000 passes of a controlled load, at each of two temperatures. In particular, the
surface profile was measured at certain intervals during the load sequence to
determine the amount of rutting the pavement endured.
Rutting Resistance of Asphalt Mixes … 431

3 Methods

3.1 Accelerated Pavement Load Facility (APLF)

The Accelerated Pavement Load Facility (APLF), located on the Ohio University
campus at Lancaster, Ohio, consists of a concrete lined rectangular pit 13.7 m long
by 11.6 m wide by 2.5 m deep. This indoor facility has two large sliding doors on
the north and south end of the facility which allow large construction vehicles to
enter and exit freely, a feature which enables the construction of full-size pavements
for testing. The facility also houses industrial grade heating and cooling units to
maintain pavements at desired temperatures in the range of −12.2 to 54.4 °C during
testing. Humidity and subgrade moisture levels can also be controlled. Dynamic
and static wheel loads are applied with either a dual or single wide-based tires, with
the wheel mounted on a track system supported by two I-beams. The mounted
wheel support feature allows for the ability to maintain a desired wheel load and
pressure on a precise location for repeated loads on a test pavement. The APLF is
capable of a wheel load of up to 133 kN with the option of wheel offsetting to
measure wander effects. A dynamic loading test inside the APLF can be seen in
Fig. 1. Previous research (Sargand and Edwards 2004) found the relationship
between the log of the number of passes and log of rut depth was a linear function
to 100,000 passes and can be characterized after 10,000–20,000 passes. Therefore,
10,000 passes of a wheel loaded to 40 kN at 21.1 °C and an additional set of 10,000
wheel passes conducted at 37.8 °C have been used to evaluate the performance of
the mixes for this research.

Fig. 1 Load wheel passing over test pavement in the APLF

432 I. Khoury et al.

3.2 Measuring Rutting with the Profilometer in the APLF

A rolling wheel profilometer was used to measure surface rutting across the four test
sections. Rutting, next to cracking, is one of the most common distresses an asphalt
pavement will endure over its entire lifespan. This profilometer, developed by
ORITE, consists of a 3 m track that allows a wheel to measure elevations to
130 µm or 1 % accuracy at 13 mm intervals along a linear path across the pave-
ment surface (Sargand et al. 2009). The rolling wheel profilometer can be seen in
Fig. 2.
From previous research performed in the APLF with the profilometer, it was
suggested the profilometer be rotated to minimize the effects of rutting in adjacent
test lanes on the elevation of the profilometer supports as well as to allow more data
points to be collected. The profilometer takes measurements over a path 2.84 m in
length; to fit the profilometer’s full path inside the 2.43 m width of each test strip,
the device was oriented at an angle of 31° away from the perpendicular. To ensure
rut measurements were taken from the same locations, metal washers were epoxied
to the surface of each test pavement at the location where each support of the
profilometer contacted the asphalt. The total distance between each leg was mea-
sured and an optimum layout was created using AutoCAD such that four profiles
were taken at equally spaced intervals along each test section. In order to com-
pensate for the starting location of the profilometer wheel, which is in front of the
legs, profilometer positions on Lane C and Lane D were a mirror image of those on
Lane A and Lane B. The profilometer layout on the pavement can be seen in Fig. 3.

Fig. 2 Rolling wheel profilometer in use at the APLF

Rutting Resistance of Asphalt Mixes … 433

Fig. 3 Overhead view of test pavements in APLF showing placement of profilometer

4 Results and Discussion

The asphalt mixes tested in the APLF appear to be very resistant to rutting.
Amongst all test sections the 20 cm test section, which exhibited the highest peak
strains for both temperatures, was analyzed as an example to show the minimal
434 I. Khoury et al.

amounts of rutting measured in the test pavements. Rutting was measured to be

0.30 mm in the 20 cm test section (Lane A) after 10,000 passes of a wheel loaded
to 40 kN at 21.1 °C. Tire pressure was 785 kPa. The accumulated rutting was
measured after an additional set of 10,000 wheel passes conducted at 37.8 °C and
was found to be 1.168 mm. Figure 4 shows the profile history in the 20 cm section
from profiles taken after 100, 300, 1000, 3000, and 10,000 wheel passes at 21.1 °C,
and Fig. 5 shows the profile history with testing conducted at 37.8 °C. Both figures
show the initial profiles of the 20 cm section as a bold line and indicated by the
number “0” and the 10,000 passes indicated by the dotted line.
Further evaluation was performed at locations that showed the largest rutting in
each test section. Maximum rut depths were calculated by taking the original profile
elevation and subtracting successive profile section elevations. The maximum
difference in elevations in each test section occurred along the wheel path when the
sections were kept at a temperature of 37.8 °C. Table 1 shows the maximum rut
depths for the APLF test sections.
The maximum values from the previous table were plotted against the number of
passes at 40 kN load and can be seen in Fig. 6.
Using the trend lines in Fig. 6, logarithmic equations of the form
y = A × lnx + B were used to model the rutting behavior of each test section;
Table 2 shows the trend line parameters and correlation coefficients.

Fig. 4 Profile history of 20 cm section (Lane A) in APLF at 21.1 °C. The numbers above each
line indicate the number of passes of the loaded wheel before the profile was measured
Rutting Resistance of Asphalt Mixes … 435

Fig. 5 Profile history of 20 cm section (Lane A) in APLF at 37.8 °C. The numbers near each line
indicate the number of passes of the loaded wheel before the profile was measured

Table 1 Maximum rut depth for APLF lanes at 37.8 °C

Number of passes Lane A (cm) Lane B (cm) Lane C (cm) Lane D (cm)
100 0.032 0.023 0.034 0.03
300 0.051 0.041 0.055 0.065
1000 0.074 0.066 0.093 0.096
3000 0.091 0.091 0.095 0.137
10,000 0.117 0.128 0.138 0.174
Pavement thickness 20 23 25 28

ODOT uses four classifications of rutting: high, medium, low, and none. High
rutting consists of any rut depth that exceeds 1.91 cm, medium rutting falls in the
range 1.91–0.95 cm, low rutting falls in the range 0.95–0.32 cm, and any rutting
below 0.32 cm is considered “No Rutting” (Sargand et al. 2009). For this project no
section exceeded the “Low Rut” threshold of 0.32 cm, as can be seen in Table 3,
even at the high temperature.
Since only one type of surface mix was used in the HiMA project, rutting data
from the warm mix asphalt (WMA) project in 2009 was used to compare against
the HiMA surface mix, which had been subjected to a similar experimental protocol
(Sargand et al. 2009). The WMA project included four 2.43 m wide test perpetual
pavements, three topped with different formulations of warm mix asphalt, and the
436 I. Khoury et al.

Fig. 6 Rut depth versus number of passes for APLF sections at 37.8 °C. The boxes around each
line fit equation are colored to match the line of actual data being fitted

Table 2 Trend line parameters for rutting in HiMA lanes

Pavement section Trend line parameters at 37.8 °C
A (cm) B R2
Lane A 0.0183 −0.0208 0.9982
Lane B 0.0226 −0.0339 0.9868
Lane C 0.0218 −0.0262 0.9598
Lane D 0.0312 −0.0454 0.9979

Table 3 Average maximum rut depth for APLF and WMA surface mixes under high temperature
Number of HiMA Lane D WMA 3N Aspha-min WMA 4N Control
passes (cm) (cm) (HMA) (cm)
100 0.03 0.116 0.064
300 0.065 0.165 0.102
1000 0.096 0.244 0.169
3000 0.137 0.348 0.267
10,000 0.174 0.513 0.441
WMA data are from Sargand et al. (2009)
Rutting Resistance of Asphalt Mixes … 437

fourth with a surface course of traditional hot mix asphalt (HMA) as a control. The
testing protocol, including the temperatures, load sequence, and measurement
intervals, was the same as for this project. The Aspha-min and Control surface
mixes were chosen to compare against the HiMA surface mix. Table 4 compares
the maximum rut depths found in the HiMA and WMA projects while the test
sections were maintained at high temperatures of approximately 37.8 °C; the
average rut depths are also plotted in Fig. 7.
As shown in Fig. 7, the HiMA surface produced much lower rutting over the
same load repetitions at the high temperature. To determine how much more rut
resistant the HiMA was compared to the WMA mixes, the logarithmic trend lines
were extrapolated until rutting reached the ODOT threshold for “low rutting” (rut
depth equal to 0.32 cm) and the number of wheel passes computed. Similar trend
lines were established for the WMA project. Table 4 shows the results. The HiMA

Table 4 Number of passes of 40 kN load to reach low rutting classification (0.32 cm)
Mix Number of passes to reach low rutting threshold
HiMA 1,038,885
Aspha-Min 2260
Control 4551

Fig. 7 Comparison of average rut depths between HiMA and WMA mixes at high temperature in
APLF. The horizontal line marks ODOT’s threshold for “low rutting”
438 I. Khoury et al.

surface would need over 1 million passes at 40 kN to reach the “low rutting”
threshold. This value is significantly greater than those for Aspha-Min. (2260 pas-
ses) and the control (4551 passes), which occurred during the WMA test in the
APLF, as can be seen in Fig. 7.
In 2013, NCAT published Report 13-03 (Timm et al. 2013), which included a
comparison of the rut resistance of a 14.6 cm HiMA Kraton mix section to that of a
control section consisting of 17.8 cm of HMA. There was a significantly lower rate
of rutting in the HiMA compared to the control mix (Timm et al. 2013). Based on
the findings between the APLF and the WMA projects, this report agrees with that
claim. The HiMA binder appears to resist rutting in the surface and intermediate
layers better than standard HMA and WMA mixes.

5 Conclusions

Four test lanes of perpetual pavement were constructed in the indoor Accelerated
Pavement Load Facility (APLF) which included highly modified asphalt (HiMA),
which contains a high-polymer binder designed to increase the strength of the
pavement. The pavement was heated and tested at temperatures of 21.1 and 37.8 °C
with humidity and subgrade moisture controlled, then each lane was subjected to
10,000 passes of a wide-base single tire load of 40 kN at each temperature. At the
beginning of each run, and after 100, 300, 1000, 3000, and 10,000 passes, rutting
was measured with a profilometer that measured the surface height to 130 µm or
1 % accuracy at 1.3 cm intervals along the device’s 2.84 m length; which was
rotated 31° to fit in the 2.43 m lane width.
The rutting profile data were published and the maximum depth of rutting noted
at each stage of the experiment. Very little rutting was seen in any of the lanes; the
maximum amount seen at the end of the experiment was 0.174 cm, which well
below the threshold of “low rutting”, 0.32 cm used by the Ohio Department of
Transportation Office of Pavement Engineering. An extrapolation of the data using
the logarithmic curve fit suggested that the “low rutting” threshold would be
reached after 1,038,885 passes of a 40 kN load at 37.8 °C.
By comparison, rutting data from an earlier experiment with similar perpetual
pavements in the APLF (Sargand et al. 2009) with surfaces using Aspha-min WMA
and traditional HMA showed sufficient rutting to cross the ODOT “low rutting”
threshold during the experiment.

References

Khoury, I., Sargand, S., Jordan, B., Cichocki, P., & Sheer, M. (2016). Structural study of perpetual
pavement performance in Ohio. In 5th International Conference on Accelerated Pavement
Testing, San Jose, Costa Rica, September 19–21, 2016.
Rutting Resistance of Asphalt Mixes … 439

Newcomb, D. E., Willis, R., & Timm, D. H. (2010). Perpetual asphalt pavements a synthesis. (IM
40) Lanham, MD: Asphalt Pavement Alliance.
Nunn, M., & Ferne, B. W. (2001). Design and assessment of long-life flexible pavements.
Transportation Research Circular, 503, 32–49. ISSN: 0097-8515.
Sargand, S., & Edwards, W. (2004). Accelerated testing of Ohio SHRP sections 390101, 390102,
390105, and 390107. Report FHWA/OH-2004/012 for the Ohio Department of Transportation,
Athens, Ohio: Ohio Research Institute for Transportation and Environment (ORITE), Ohio
University, December, 2004.
Sargand, S., Figueroa, J. L., Edwards, W., & Al-Rawashdeh, A. S. (2009). Performance
assessment of warm mix asphalt (WMA) pavements. Report No. FHWA/OH-2009/08 for the
Ohio Dept. of Transportation, Athens, OH: Ohio Research Institute for Transportation and the
Environment, Ohio University, September, 2009. http://cdm16007.contentdm.oclc.org/cdm/
ref/collection/p267401ccp2/id/4561. Accessed April 29, 2016.
Sargand, S., Khoury, I., Jordan, B., Scheer, M., & Cichocki, P. (2015). Implementation and
thickness optimization of perpetual pavements in Ohio, Report FHWA/OH-2015/17 for Ohio
Dept. of Transportation, Athens, OH: Ohio Research Institute for Transportation and the
Environment, Ohio University, June, 2015. http://cdm16007.contentdm.oclc.org/cdm/ref/
collection/p267401ccp2/id/12760. Accessed April 29, 2016.
Timm, D. H., Robbins, M. M., Willis, J. R., Tran, N., & Taylor, A. J. (2013). Field and laboratory
study of high-polymer mixtures at the NCAT Test Track. Final report (NCAT Report 13-03).
Auburn, AL: National Center for Asphalt Technology.
Structural Study of Perpetual Pavement
Performance in Ohio

Issam Khoury, Shad Sargand, Benjamin Jordan, Paul Cichocki

and Matthew Sheer

Abstract Three perpetual pavement test sections were constructed on U.S. Route
23 in Delaware, Ohio (DEL-23) with AC thicknesses 28, 33, 38 cm and instru-
mented to detect strains in Fatigue Resistant Layer (FRL) and base layers. The 28
and 33 cm sections were constructed on lime stabilized subgrade, while the 38 cm
section was constructed on compacted subgrade. Four additional test pavements
were built in the Accelerated Pavement Load Facility (APLF) and instrumented
similarly to DEL-23. The sections were thinner, but included Highly Modified
Asphalt (HiMA) with Kraton polymer binder in sections of depth 20, 23, and
25 cm. An 28 cm section used conventional asphalt in the base as a control with
HiMA in surface and intermediate layers. There was no FRL. All sections were
placed on 46 cm of cement stabilized subgrade. Strains at bottom of FRL during
Controlled Vehicle Load (CVL) testing on DEL-23 in summer indicated the 33 cm
section on stabilized subgrade and 38 cm section on compacted subgrade met the
perpetual pavement criteria of NCHRP Project 9-44A, while 28 cm section on
stabilized subgrade did not. Testing at the APLF included thoroughly heating the
pavement to 37.8 °C and subjecting each section to 10,000 passes of a 40 kN wheel
load. Pavement strains at the bottom of the base and intermediate layers in the
longitudinal and transverse directions were measured after 1000, 3000, and 10,000
wheel passes using test loads of 27, 40, and 53 kN. The serviceability of the
pavements was determined by comparing the longitudinal strains within the base
layer of each pavement to fatigue endurance limits (FEL) calculated by using
flexural stiffness standards from NCHRP 9-44A and a method recommended by
Kansas researchers. At the APLF, the thinnest section produced maximum average
strains higher than the calculated FEL at 37.8 °C using the NCHRP 9-44A equa-
tion, while the 25 and 28 cm sections met the perpetual pavement criteria. Using the
Kansas researchers approach, all four test sections were found to have lower lon-
gitudinal strains than the calculated FEL.

I. Khoury (&)
S. Sargand
B. Jordan
P. Cichocki
M. Sheer

Civil Engineering Department, Ohio University, Athens, OH 45701-2979, USA
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 441

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_29
442 I. Khoury et al.

Keywords Perpetual pavement Highly modified asphalt Accelerated pavement


testing CVL testing

1 Introduction

Perpetual pavements are asphalt concrete (AC) pavement structures designed to

carry the heaviest vehicle loads without the need for major structural rehabilitation
or reconstruction for at least 50 years (Asphalt Pavement Alliance 2002). These
pavement structures obtain their longevity from resisting the failure modes of
bottom-up fatigue cracking and structural rutting. They are multiple layer systems
where thickness and stiffness of each layer is designed to resist specific types of
distress. Properly designed perpetual pavement structures can provide superior
performance that minimizes user delays and enhances economic efficiency.
Fatigue endurance limits (FELs) are used to evaluate these perpetual pavement
systems and predict remaining life before the pavement completes service. It was
discovered that if certain strain responses remain below specific thresholds, the
pavement structure, in theory, can withstand an infinite amount of load cycles
without development of damage. During loading, reactions at critical locations can
be monitored and if the reactions remain below the FEL, bottom-up fatigue
cracking and structural rutting and the resulting damage will be diminished (Timm
and Newcomb 2006).
Battaglia et al. (2010), report on two test sections of perpetual pavement con-
structed on a weigh station entrance ramp on I-94 near Kenosha, Wisconsin.
Annual performance surveys conducted after 4, 5, and 6 years of service found
many instances of alligator cracking and longitudinal cracking in the wheel paths.
The cracking was top-down, and had not yet reached the middle layer of HMA.
Thus it is believed the observed premature cracking was due to segregation and
over-compaction of the upper layers of pavement. Rutting was less than 2.5 mm
and ride quality was good. The authors also noted it is difficult to estimate an
expected service life before the first surface layer replacement. The recommended
FEL (εallowable at the bottom of the pavement) was set to 70 µε, and the constructed
pavement generally had low strains during testing.
Park and Kim (2003) proposed a method for predicting the remaining life of
flexible pavements from FWD measurements. The performance of pavement
undergoing fatigue cracking could be predicted by their method except for sections
in wet freeze regions that had experienced high levels of cracking that appear to
have been caused by environmental distresses, such as low-temperature cracking.
The synthesis by Newcomb et al. (2010) recommends several areas of continued
investigation, such as the development of high-modulus asphalt pavements and mix
designs. They also recommend identifying factors that influence the fatigue limit
and otherwise refining the determination of the FEL, beyond the findings of
NCHRP Study 9-44A.
Structural Study of Perpetual Pavement Performance in Ohio 443

Willis (2009) and Willis and Timm (2009a, b) compared laboratory measure-
ments of fatigue thresholds to cumulative strain distributions via a fatigue ratio.
Correlating the laboratory fatigue threshold to one point on the cumulative strain
distribution was difficult. Instead, a ratio of the strain to the 95th percentile lower
bound of the fatigue confidence interval determined in the laboratory was used to
quantify the strain distribution A table of upper bound fatigue limits is presented by
Willis (2009), with the recommendation that with similar mix designs and traffic
patterns to those analyzed, sections whose measured cumulative strain distributions
remain below the upper bound in the table should withstand fatigue cracking. The
authors recommend further research to develop the links between laboratory fatigue
thresholds, measured field strains, and pavement performance.
The above studies indicated there was a great need for developing a compre-
hensive understanding of the design of perpetual pavements. The main deficiency
with the current perpetual pavement design method is it does not ensure an opti-
mum structure and/or layers that will last for a 50-year design life.

2 DEL-23 Project Background and Instrumentation

2.1 Project Site Description

As part of the ODOT’s Partnered Research Exploration Program, ODOT and the
Ohio Research Institute for Transportation and the Environment (ORITE) devel-
opment of optimized perpetual pavement designs for Ohio through the instru-
mentation of four pavement test sections with different thicknesses on the Strategic
Highway Research Program (SHRP) Test Road on U.S. Route 23 in Delaware
County, Ohio, near the town of Waldo, henceforth referred to as “DEL-23”. The
project was focused on analyzing the impact of varying total pavement thicknesses
by monitoring data obtained during Controlled Vehicle Load (CVL) testing. The
optimal design thickness would minimize the life cycle cost of the pavement.
Two test sections were constructed on North Waldo Road, section 39BN803 in
the northbound lane and section 39BS803 in the adjacent southbound lane. The two
remaining test sections were constructed on the northbound lanes of the U.S. Route
23 mainline, with section 39D168 residing in the driving lane and section 39P168
in the adjacent passing lane. More details on the construction and instrumentation
of the test sections on DEL-23 are in the project report (Sargand et al. 2015).

2.2 Design of Test Sections

The goal in designing a perpetual pavement is eliminating bottom-up fatigue

cracking. The perpetual pavement design includes a top layer designed to resist
rutting and tire wear, an intermediate layer that also resists rutting, and a rich
444 I. Khoury et al.

bottom layer designed to resist fatigue cracking, known as the fatigue resistant layer
(FRL). The thicknesses of the layers are increased so that the strain at the bottom of
the FRL, where it contacts the dense graded aggregate base (DGAB) never exceeds
a specified value of the longitudinal tensile strain, called the “fatigue endurance
limit” (FEL), in regular use.
The pavement thicknesses were also selected to validate results previously
obtained from existing sections on WAY-30 (Sargand et al. 2008; Liao and Sargand
2010; Sargand and Figueroa 2010) and the Accelerated Pavement Load Facility
(APLF) (Sargand et al. 2009). Sections with a total asphalt thickness of 38 and
33 cm were similar to sections constructed in the APLF and were expected to have
strains of 70 µε or less at the bottom of the FRL. The section with a total asphalt
thickness of 28 cm was expected to have a strain of 70 µε or more. Table 1 lists
specifications for each layer used.
Table 2 provides the specified layer thicknesses for each section. Each section
was constructed on a 15 cm layer of Dense Graded Aggregate Base (DGAB). The
subgrade for all four sections had a soil classification of A-6. In addition, the
subgrade for sections 39BS803 and 39BN803 had been chemically stabilized
through lime treatment during a previous construction project. Testing revealed the
subgrade beneath sections 39D168 and 39P168 had an average resilient modulus of
138 MPa while the stabilized subgrade for sections 39BS803 and 39BN803 had an
average resilient modulus of 276 MPa. After construction of the test sections, core
samples obtained revealed section 39P168 had been constructed with a total
pavement thickness of 41 cm, a 7.6 cm increase from what had been specified. Due

Table 1 Layer specifications for DEL-23 perpetual pavement sections

Layer Description PG binder Binder content
grade (%)
Surface Fine graded polymer asphalt PG 76-22 M 7.6
concrete
Intermediate Asphalt concrete, 19 mm PG 64-28 4.9
Base Asphalt concrete PG 64-22 4.2
Fatigue Fatigue resistant base layer PG 64-22 4.6
resistant

Table 2 Specified layer thicknesses of DEL-23 pavement sections

Section 39D168 (cm) 39P186 (cm) 39BS803 (cm) 39BN803 (cm)
Layer
Surface layer 2.5 2.5 2.5 2.5
Intermediate layer 5.1 5.1 5.1 5.1
Base layer 20.3 15.2 15.2 10.2
Fatigue resistant layer 10.2 10.2 10.2 10.2
Total pavement thickness 38.1 33 33 27.9
Structural Study of Perpetual Pavement Performance in Ohio 445

to this discrepancy, data obtained from this section are not presented here.
Henceforth, sections 39BN803, 39BS803, and 39D168 will be identified as the 28,
33, and 38 cm sections respectively.

2.3 Instrumentation of Test Sections

In order to measure pavement responses while being exposed to dynamic loading

and various environmental conditions, numerous sensors were installed throughout
the pavement structure following a comprehensive instrumentation plan.
Thermocouples were installed at various depths in the pavement to measure tem-
perature throughout the structure; air temperature was also measured with a ther-
mometer. Linear Variable Differential Transformers (LVDTs) were installed to
measure deflections at the pavement surface and at the top of the subgrade. Pressure
cells recorded pressures in the subgrade and the Fatigue Resistant Layer (FRL).
Horizontal strain gauges were installed in every asphalt layer in the longitudinal,
transverse, and 45° positions. A detailed layout of the entire instrumentation plan is
presented in Fig. 1.

Fig. 1 Instrumentation layout for 33 and 38 cm sections (1 ft = 0.305 m)

446 I. Khoury et al.

3 Controlled Vehicle Load Testing on DEL-23

Controlled vehicle load (CVL) testing was performed by driving ODOT dump
trucks loaded to a known weight at a controlled speed over the pavement and
recording the sensor responses. The axle configuration and tire pressure were
recorded as was the lateral offset of the tires where they passed over the sensors,
which was recorded by measuring imprints in sand placed on the surface. For every
variation of load, axle type, tire pressure, and speed, at least five tests were per-
formed and the three tests with the least amount of lateral offset measured were used
for analysis. CVL tests were conducted both in the fall and summer, with both
Tandem axle and Single axle trucks. The most critical data is for single axle trucks
in the hot temperature which will be discussed further.

3.1 CVL Testing Pavement Response in Summer

CVL testing was conducted on the perpetual pavement test sections during July 1,
10, and 11, 2013 on the 38, 28, and 33 cm sections, respectively, to determine
pavement response under hot weather conditions. Each section was tested during
the warmest period of the day, between 10:00 AM and 3:00 PM, with temperatures
measured every 10 min. Although pavement temperatures remained fairly consis-
tent during testing some variations were seen throughout the course of testing for
each section and between testing days. The air temperature sensor was placed
directly over the pavement at a low elevation, so that the temperature recorded
includes ambient air temperature and heat radiated from the pavement surface, thus
peak values for air temperature are higher than commonly reported ambient tem-
peratures, and in this case at times exceeded 37.8 °C. The test procedure was
identical to the cold weather test procedure, using the same axle configurations,
loads, speeds, and tire pressures.

3.1.1 Longitudinal Strain in the FRL

The 28 cm section received the highest longitudinal strains in the FRL in com-
parison with the 33 and 38 cm sections. Table 3 shows the average maximum
longitudinal strain and maximum longitudinal strain measured in the FRL for
testing conducted on the 28 cm section. It compares strains found at various speeds
and tire pressures for both the single axle wide-based tire and tandem axle dual tire
trucks. The maximum longitudinal strain recorded in the FRL for the 28 cm section
was 143.25 µε, captured during single axle truck testing at a truck speed of 8 km/h
and a tire pressure of 758 kPa. Average longitudinal strains in the FRL ranged
between 56 and 107 µε for the single axle truck and between 48 and 78 µε for the
tandem axle truck. The highest average and maximum values for the longitudinal
Structural Study of Perpetual Pavement Performance in Ohio 447

Table 3 Maximum longitudinal strain in the FRL for the 28 cm section (µε)
Tire pressure Speed [mph (km/h)]
5 (8) 30 (48) 55 (89)
(psi) (kPa) Average Max Average Max Average Max
Single axle wide-base tire (93 kN axle load)
80 552 95.95 137.69 73.09 102.83 72.65 92.12
110 758 106.50 143.25 73.86 100.34 64.23 89.41
125 862 101.45 138.56 73.11 99.57 56.41 71.43

strain in the FRL for the 28 cm section were consistently seen during 8 km/h testing
for each tire pressure.
The 33 cm section received significantly lower longitudinal strains in the FRL
than the 28 cm section. Table 4 presents a summary of longitudinal strains mea-
sured in the FRL during testing on the 33 cm section. The maximum FRL longi-
tudinal strain obtained in the 33 cm section was 85.04 µε, when testing at 8 km/h
with a tire pressure of 80 psi (552 kPa) and the single axle truck, which was
considerably less than the maximum strain for the 28 cm section. Average longi-
tudinal strain in the FRL for the 33 cm section ranged between 35 and 76 µε for the
single axle truck and between 26 and 53 µε for the tandem axle truck. Similar to the
28 cm section, strains produced by the single axle truck were consistently higher
than strains produced by the tandem axle truck. Longitudinal strains in the FRL
were also, once again, highest when testing at 8 km/h for each tire pressure,
although there was no clear dependence on tire pressure.
The 38 cm section had longitudinal strain in the FRL greater than those of the
33 cm section, but still substantially lower than those of the 28 cm section. As
mentioned for the autumn test section, the discrepancy may be due to the higher
modulus of the stabilized subgrade under the 13 in (33 cm) section.
Table 5 displays the maximum and average maximum longitudinal strains
recorded in the FRL for the 38 cm section. The maximum strain was 102.18 µε,
measured using the single axle truck at 8 km/h and a tire pressure of 80 psi
(552 kPa). Average longitudinal strain in the FRL ranged between 42 and 78 µε for
testing utilizing the single axle truck and 37 and 53 µε with the tandem axle truck.
Once more, the highest strains were obtained during 8 km/h for each tire pressure.

Table 4 Maximum longitudinal strain in the FRL for the 33 cm section (µε) for single axle
wide-base tire (93 kN axle load)
Tire pressure (kPa) Speed (km/h)
8 48 89
Average Max Average Max Average Max
552 75.04 85.04 44.2 50.92 39.88 42.15
758 68.51 77.14 45.59 49.62 37.62 41.26
862 66 73.99 40.13 43.63 35.98 40.23
448 I. Khoury et al.

Table 5 Maximum longitudinal strain in the FRL for 38 cm section (µε) single axle wide-base
tire (93 kN axle load)
Tire pressure (kPa) Speed (km/h)
8 48 89
Average Max Average Max Average Max
552 77.92 102.18 57.62 66.32 42.11 47.77
758 74.97 96.44 54.42 62.28 47.79 50.97
862 70.46 93.2 55.2 65.19 44.23 46.38

4 Evaluation of Highly Modified Asphalt

Highly Modified Asphalt (HiMA) utilizes a new modified binder created by Kraton
Polymers that contains a 7.5 % styrene butadiene styrene (SBS) polymer to
improve stiffness at high temperatures. HiMA mixture has also been investigated at
the National Center for Asphalt Technology (NCAT) test track near Auburn,
Alabama. Based on the promise shown in the NCAT evaluations, an experiment
was proposed at the Accelerated Pavement Load Facility (APLF) to determine if
HiMA’s improved stiffness would allow for a thinner perpetual pavement design.
The four test sections at the APLF range in thicknesses between 20 and 28 cm in
2.5 cm increments to determine where the pavement transitions from standard to
perpetual response, with the 28 cm pavement built as a control to match the depth
and composition of the thinnest DEL-23 section.

4.1 The Accelerated Pavement Load Facility

The Accelerated Pavement Loading Facility (APLF), located in Lancaster Ohio,

includes a concrete lined rectangular pit 13.7 m long by 11.6 m wide by 2.5 m deep
in which the test pavements were constructed. The facility has two large sliding
doors on the north and south end of the facility which allow large construction
vehicles to enter and exit freely, a feature which makes the APLF unique. As an
indoor facility, tests may be performed at any time in the year. The facility also
houses industrial grade heating and cooling units to test and maintain pavements at
desired temperatures in the range of −12.2 to 54.4 °C. Dynamic and static wheel
testing can be conducted using dual or single wide-based tires. The wheel is
mounted onto a track system that is supported by two I-beams which apply a
controlled load up to 133 kN through the wheel, with variable offsetting possible to
measure wander effects. More details on the APLF can be found in the project
report (Sargand et al. 2015), and in another article contributed to this conference
(Green et al. 2016).
Structural Study of Perpetual Pavement Performance in Ohio 449

4.2 Instrumentation of Pavements in the APLF

In an agreement with the Ohio Department of Transportation (ODOT), it was

decided to use an instrumentation plan similar to those employed in prior perpetual
pavement studies in the APLF (e.g. Sargand et al. 2009); therefore, strain gauges,
thermocouples, and LVDTs were chosen to measure the dynamic load responses of
the pavement and the subgrade. Strain gauges were installed to measure the hori-
zontal strains at the base of the fatigue and the intermediate layers. Thermocouples
were installed throughout the entire depth of the pavement in each test sections to
monitor the internal temperatures during testing. The LVDTs were used to observe
the vertical deflections of the subgrade-pavement interface and the asphalt surface.
Upon completing gauge selection, an AutoCAD document was created for use as a
reference during construction. An example of the instrumentation for test lane A
can be seen below in Fig. 2.
A 15 cm layer of DGAB was placed, leveled, and compacted on top of the
existing subgrade, which was previously stabilized with cement to a depth of
46 cm. Following compaction, the gauges were installed following the grid layout
in Fig. 2 designed using AutoCAD was implemented inside the APLF.
Four test lanes were measured to size and the centerlines were marked with spray
paint. From previous research studies performed in the APLF, it was decided the first
strain gauge in the bottom of the fatigue layer would be placed along the centerline of
each test section 1.8 m from the concrete platform on the south end of the facility. At
this distance the initial balancing of the loads prior to the motion of the wheel would
have dissipated enough to eliminate the possibility of false readings.

Fig. 2 Gauge layout in APLF test lane A, which is similar to the other lanes
450 I. Khoury et al.

5 APLF Pavement Responses

Dynamic load testing was conducted in the APLF during May through September
2014. Upon completion of 10,000 passes of a 40 kN load at 8 km/h on all lanes at
70 °F (21.1 °C), the temperature was increased to 100 °F (37.8 °C), the tempera-
ture allowed to stabilize through the pavement thickness, and 10,000 passes of the
load were then applied to all lanes. Profiles were taken to measure rutting of each
test section prior to loading and upon completion of 100, 300, 1000, 3000, and
10,000 passes of a 40 kN load. After completion of 100, 3000, and 10,000 passes,
the lead wires were attached to the LVDTs and twelve passes of three wheel loads
(27, 40 and 53 kN), at 8 km/h with various offsets to the centerline of the lane, were
applied to analyze the test section’s pavement response. Longitudinal strains in the
base layer of each test section were compared to calculated endurance limits to
determine which sections met the perpetual pavement design criteria. LVDT
deflections were measured with reference to the bottom of the DGAB and 91 cm
into the subgrade.

5.1 Strain Responses in the HiMA Pavement

During testing, strain responses were measured in the same direction as traffic (the
rolling load wheel), called the longitudinal direction, and in the perpendicular or
“transverse” direction. Strains in the longitudinal direction were compressive while
the wheel approached and left the gauge as, but were tensile while the wheel was
over the gauge. Transverse strains measured at the bottom of the AC base were
tensile, while transverse strains measured at the bottom of the intermediate layer
were compressive.

5.2 Longitudinal Strain in HiMA Base Caused by Adjusted

Load

Table 6 shows the average and maximum longitudinal strains found in Lane A
(20 cm), Lane B (23 cm), Lane C (25 cm), and control Lane D (28 cm) produced
during testing in the 21.1 °C and the 37.8 °C temperatures under wheel loads of 27,
40, and 53 kN. The highest maximum strains of 79 and 113 με were produced
under the 53 kN wheel loads for both temperatures on the thinnest (20 cm) test
section. The 25 cm section (Lane C) using the HiMA, produced strains lower than
the 28 cm control (Lane D) for testing at 21.1 °C and yielded similar longitudinal
strains at 37.8 °C.
Structural Study of Perpetual Pavement Performance in Ohio 451

Table 6 Average and maximum longitudinal strains in APLF base layer (µε)
Lane AC thickness Load [lb (kN)]
6000 (27) 9000 (40) 12,000 (53)
(in) (cm) Avg Max Avg Max Avg Max
70 °F (21.1 °C)
A 8 20 35 43 54 61 70 79
B 9 23 31 36 48 54 62 69
C 10 25 21 24 35 39 46 51
D 11 28 27 43 40 55 52 67
100 °F (37.8 °C)
A 8 20 62 66 89 93 106 113
B 9 23 41 46 63 73 79 83
C 10 25 34 44 50 56 61 67
D 11 28 27 34 43 56 56 73

Figure 3 is a plot of average maximum longitudinal strains versus the three

wheel loads applied during testing at 21.1 °C, and the corresponding results at
37.8 °C can be seen in Fig. 4. Typically the 28 cm control lane (Lane D) produced
the lowest average maximum longitudinal strains in the AC base layer at 37.8 °C
under all three loads, while at 21.1 °C, the 25 cm HiMA lane (Lane C) produced

Fig. 3 Average max longitudinal strains for APLF test sections at 21.1 °C
452 I. Khoury et al.

Fig. 4 Average max longitudinal strains for APLF test sections at 37.8 °C

the lowest average maximum longitudinal strains. Longitudinal strains in the 25 cm

section (Lane C) were less than those in the 28 cm control section (Lane D) during
testing conducted at 21.1 °C. However, the 20 cm and 23 cm sections both pro-
duced average maximum longitudinal strains higher than strains in the AC base of
the 28 cm control lane. At 37.8 °C, once again the two thinner sections produced
higher longitudinal strains than the other two sections. As seen in Fig. 4, the effects
of temperature on the 20 cm section caused much higher strains to be produced
compared to those in thicker sections.

6 Estimating the Fatigue Endurance Limit

The FEL of a perpetual pavement design can be estimated using the results of the
dynamic modulus test to estimate the initial flexural stiffness (E0). The NCHRP
project 9-44A included a vigorous laboratory testing program and development of a
model which estimates the fatigue endurance limit based on results of the beam
fatigue test (Witczak et al. 2013). This model calculated the Stiffness Ratio (SR)
based on applied tensile strain, rest period, number of loading cycles, and the initial
flexural stiffness for the beam fatigue test. The endurance limit of an asphalt mix
can be determined by setting the Stiffness Ratio equal to one and solving for the
tensile strain using Eq. 1.
Structural Study of Perpetual Pavement Performance in Ohio 453

SR ¼ 2:0844  0:1386  logð0:145E0 Þ  0:4846  logðet Þ  0:2012  logðN Þ

þ 1:4103  tanhð0:8471  RPÞ þ 0:0320  logð0:145E0 Þ  logðet Þ
 0:0954  logð0:145E0 Þ  tanhð0:7154  RPÞ  0:4746  logðet Þ  tanhð0:6574  RPÞ
þ 0:0041  logðN Þ  logð0:145E0 Þ þ 0:0557  logðN Þ  logðet Þ
þ 0:0689  logðN Þ  tanhð0:259  RPÞ
ð1Þ

where
SR = stiffness ratio = stiffness measured at any load cycle during beam fatigue
testing divided by the initial stiffness of the specimen
E0 = initial flexural stiffness (MPa)
εt = applied tensile strain (µε)
RP = rest period (s)
N = number of load cycles
The NCHRP project 9-44A report included a sensitivity study which concluded
N has little to no effect on the Stiffness Ratio, hence N is set to the recommended
value of 200,000 cycles for this analysis, per Witczak et al. (2013). The Rest Period
(RP) had a major effect on the endurance limit, up to 5 s, above which the
endurance limits became similar. Thus for the laboratory analysis of the DEL-23
pavements and the HiMA pavements inside the APLF, a common rest period of 5 s
was used.
Values for E0 were determined from laboratory measurements conducted on
specimens collected at the asphalt plant during construction of each test pavement
and compacted to 7 % air void on site using a Superpave Gyratory Compactor and
following Test Specification AASHTO T312. Each compacted specimen was fur-
ther prepared and tested using an asphalt mixture performance tester following
AASHTO TP 62 to determine dynamic modulus at temperatures of 4.4, 21.1, 37.8,
and 54.4 °C and load frequencies of 25, 10, 5, 1, 0.5, and 0.1 Hz. Master curves of
dynamic modulus as a function of load frequency and log shift factor as a function
of temperature were prepared. Results of the laboratory measurements of the
dynamic modulus at 10 Hz are in Table 7 for the DEL-23 FRL, the HiMA AC
base, and the APLF control section AC base.

Table 7 Dynamic modulus results at 10 Hz as measured in the laboratory

Temperature (° Dynamic modulus (E*) at 10 Hz
C) DEL-23 FRL HiMA-AC base-Kraton APLF-AC base-control
(GPa) (GPa) (GPa)
4.4 17.2 19.09 18.82
21.1 10.58 10.7 10.45
37.8 4.52 5.05 4.67
54.4 1.8 2 1.45
454 I. Khoury et al.

6.1 Comparison of Endurance Limits and Strain Results

from Field Testing

A pavement structure is perpetual if the stiffness ratio is greater than or equal to one
when the maximum strain is entered into the model. By inserting the maximum
strain values measured in the field into Eq. 1 and solving for Stiffness Ratio,
perpetual pavement status can be determined. The dynamic modulus used is the
value measured in the laboratory adjusted to the temperature of the field testing,
using a quadratic best fit.
It is important to note the strains measured for the DEL-23 testing is based on a
single axle, single wide based tire load of 14 kip (62 kN) whereas the APLF
average maximum strains recorded were based on a tandem axle, dual tire load of
12 kip (53 kN). The loads are similar but the single tire configuration is expected to
result in higher strain readings. The initial flexural stiffness is estimated using both
E0 = E* and E0 = E*/2. Tables 8 and 9 show the FEL of the DEL-23 test tem-
peratures and the results of the stiffness ratio when the average maximum strains
measured are put into the endurance limit prediction model.

Table 8 DEL-23 FEL at test temperatures with E0 = E*, RP = 5 s, f = 10 Hz, N = 200,000,

SR = 1
Date Lane Pavement Average Dynamic Initial flexural FEL,
depth Temp., modulus, E* stiffness, E0 εt
(cm) T (°C) (GPa) (GPa) (µε)
11/29/2012 39D168 38 5 16.96 16.96 70
12/18/2012 39BN803 28 6.7 16.3 16.3 71
12/19/2012 39BS803 33 6.7 16.22 16.22 71
7/1/2013 39D168 38 28.9 7.41 7.41 88
7/10/2013 39BN803 28 26.7 8.22 8.22 85
7/11/2013 39BS803 33 27.2 7.94 7.94 86

Table 9 DEL-23 stiffness ratio based on the average peak strain measured during the controlled
vehicle load tests (E0 = E*), RP = 5 s, f = 10 Hz, N = 200,000
Date Lane Pavement Avg. Initial Stiffness Average FEL,
depth Temp., flexural ratio, SR peak εt
(cm) T (°C) stiffness, E0 strain (µε)
(GPa) (µε)
11/29/2012 39D168 38 5 16.96 1.15 38 70
12/18/2012 39BN803 28 6.7 16.3 1.1 47 71
12/19/2012 39BS803 33 6.7 16.22 1.2 31 71
7/1/2013 39D168 38 28.9 7.41 1.04 74 88
7/10/2013 39BN803 28 26.7 8.22 0.96 101 85
7/11/2013 39BS803 33 27.2 7.94 1.05 70 86
Structural Study of Perpetual Pavement Performance in Ohio 455

Based on the testing results from DEL-23, the summer testing showed the 28 cm
section (39BN803) had a stiffness ratio less than one when the initial flexural
stiffness was set equal to the dynamic modulus at the respective test temperature.
All other test sections for summer and winter showed the stiffness ratio greater than
or equal to one; they could be classified as perpetual pavement sections based on
the model prediction. Tables 10 and 11 show the results of the model prediction
when the initial flexural stiffness is equal to one half of the dynamic modulus.
When using the assumption the initial flexural stiffness is half of the dynamic
modulus, all of the pavement sections on DEL-23 have a stiffness ratio greater than
or equal to one and can be classified as perpetual based on the beam fatigue model.
The temperatures inside the APLF were controlled for testing purposes. Two
temperatures were selected for testing 70 °F (21.1 °C) and 100 °F (37.8 °C). Also,
the strains used for perpetual pavement analysis were based on a tandem axle, dual
tire configuration with a 12,000 lb (53 kN) load applied to the pavement. Tables 12
and 13 show the FEL of the HiMA test temperatures and the results of the stiffness

Table 10 DEL-23 fatigue endurance limits at test temperatures (E0 = E*/2), RP = 5 s, f = 10 Hz,
N = 200,000, SR = 1
Date Lane Pavement Average Dynamic Initial FEL,
depth Temp., modulus, flexural εt
(cm) T (°C) E* (GPa) stiffness, E0 (µε)
(GPa)
11/29/2012 39D168 38 5 16.96 8.48 85
12/18/2012 39BN803 28 6.7 16.3 8.15 85
12/19/2012 39BS803 33 6.7 16.22 8.11 86
7/1/2013 39D168 38 28.9 7.41 3.7 105
7/10/2013 39BN803 28 26.7 8.22 4.11 102
7/11/2013 39BS803 33 27.2 7.94 3.96 103

Table 11 DEL-23 stiffness ratio based on the average peak strain measured during the controlled
vehicle load tests (E0 = E*/2), RP = 5 s, f = 10 Hz, N = 200,000
Date Lane Pavement Average Initial Stiffness Average FEL,
depth Temp., flexural ratio, SR peak εt
(cm) T (°C) stiffness, strain (µε)
E0 (GPa) (µε)
11/29/2012 39D168 38 5 8.48 1.2 38 85
12/18/2012 39BN803 28 6.7 8.15 1.15 47 86
12/19/2012 39BS803 33 6.7 8.11 1.25 31 86
7/1/2013 39D168 38 28.9 3.7 1.09 74 105
7/10/2013 39BN803 28 26.7 4.11 1 101 102
7/11/2013 39BS803 33 27.2 3.96 1.1 70 103
456 I. Khoury et al.

Table 12 APLF fatigue endurance limits at test temperatures (E0 = E*), RP = 5 s, f = 10 Hz,
N = 200,000, SR = 1
Mix Test Temp., Dynamic modulus, Initial flexural stiffness, FEL, εt
T (°C) E* (GPa) E0 (GPa) (µε)
AC base: 21 10.45 10.45 80
control 38 4.67 4.67 99
AC base: 21 10.7 10.7 79
Kraton 38 5.05 5.05 97

Table 13 APLF stiffness ratio based on the average peak strain measured during the controlled
load tests (E0 = E*), RP = 5 s, f = 10 Hz, N = 200,000
Lane Mix Pavement Test Dynamic Initial Stiffness Average FEL,
thickness Temp., modulus, flexural ratio, SR peak εt
(cm) T (°C) E* (GPa) stiffness, E0 strain (µε)
(GPa) (µε)
D AC 28 21 10.45 10.45 1.1 52 80
base: 38 4.67 4.67 1.14 56 99
control
C AC 25 21 10.7 10.7 1.13 46 79
base: 38 5.05 5.05 1.11 61 97
Kraton
B AC 23 21 10.7 10.7 1.06 62 79
base: 38 5.05 5.05 1.05 79 97
Kraton
A AC 20 21 10.7 10.7 1.03 70 79
base: 38 5.05 5.05 0.98 106 97
Kraton

Table 14 APLF fatigue endurance limit at test temperatures (E0 = E*/2), RP = 5 s, f = 10 Hz,
N = 200,000, SR = 1
Mix Test Temp., Dynamic modulus, Initial flexural stiffness, FEL, εt
T (°C) E* (GPa) E0 (GPa) (µε)
AC base: 21 10.45 5.22 96
control 38 4.67 2.34 118
AC base: 21 10.7 5.35 95
Kraton 38 5.05 2.53 116

ratio when the average maximum strains measured are inputted into the endurance
limit prediction model. Only Lane A, the thinnest section, did not meet the criteria
from perpetual pavement when the initial flexural stiffness was set equal to the
dynamic modulus.
Structural Study of Perpetual Pavement Performance in Ohio 457

Table 15 APLF stiffness ratio based on the average peak strain measured during the controlled
load tests (E0 = E*/2), RP = 5 s, f = 10 Hz, N = 200,000
Lane Mix Pavement Test Dynamic Initial Stiffness Average FEL,
thickness Temp., modulus, flexural ratio, SR peak εt
(cm) T (°C) E* (GPa) stiffness, E0 strain (µε)
(GPa) (µε)
D AC 28 21 10.45 5.22 1.15 52 96
base: 38 4.67 2.34 1.19 56 118
control
C AC 25 21 10.7 5.35 1.18 46 95
base: 38 5.05 2.53 1.16 61 116
Kraton
B AC 23 21 10.7 5.35 1.11 62 95
base: 38 5.05 2.53 1.1 79 116
Kraton
A AC 20 21 10.7 5.35 1.03 70 95
base: 38 5.05 2.53 1.02 106 116
Kraton

Tables 14 and 15 show the results of the endurance limit estimation when the
initial flexural stiffness is half of the dynamic modulus. Unlike the previous anal-
ysis, when E0 = E*/2, Lane A met the criteria for perpetual pavement status, along
with the others.

7 Conclusions

The endurance limit is the key concept in the perpetual pavement design. By
building the pavement thick enough and using the appropriate materials to ensure
the tensile stress at the bottom of the asphalt fatigue resistance layer never exceeds
the endurance limit, one can prevent the bottom-up damage that typically leads to
the need for a full-depth reconstruction. Instead the perpetual pavement will per-
form for 50 years or more with periodic surface course milling and replacement.
Early studies, such as those discussed in Romanello (2007) and Sargand et al.
(2008), used a fixed endurance limit criterion of 70 με. Since then the endurance
limit has been adjusted, usually upward, depending on the material properties of the
asphalt and/or aggregate being used. For example, it is possible to derive an
endurance limit from the NCHRP 9-44A equation (Eq. 1) using the approach with
E0 = E* or E0 = E*/2.
The three pavements constructed and studied on DEL-23, all met the original
conservative 70 με criterion for perpetual pavement except the thinnest pavement
with 28 cm thickness. Thus using a similar design and mix, a thickness of 33 cm
will be sufficient to create a perpetual pavement, when constructed on a stabilized
subgrade. The only section included in the study on compacted subgrade consisted
458 I. Khoury et al.

of 38 cm of asphalt. Therefore, using a similar design and mix, a thickness of

38 cm will be sufficient to create a perpetual pavement on compacted subgrade.
In the APLF study, using HiMA in the pavement, the thickness of a perpetual
pavement may be further reduced to 25 cm if the 70 με criterion is used, or to
23 cm if the NCHRP 9-44A FEL is used.

References

Green, R., Khoury, I., Sargand, S., & Cichocki, P. (2016). Rutting resistance of asphalt mixes
containing highly modified asphalt (HiMA) binders at the accelerated pavement load facility in
Ohio. In: Fifth International Conference on Accelerated Pavement Testing, San Jose, Costa
Rica, September 19–21, 2016.
Newcomb, D. E., Willis, R., & Timm, D. H. (2010). Perpetual asphalt pavements a synthesis. (IM
40) Lanham, MD: Asphalt Pavement Alliance.
Nunn, M., & Ferne, B. W. (2001). Design and assessment of long-life flexible pavements.
Transportation Research Circular, 503, 32–49. ISSN: 0097-8515.
Robbins, M. M., & Timm, D. H. (2008). Temperature and velocity effects on a flexible perpetual
pavement. Paper presented at the 3rd International Conference on Accelerated Pavement
Testing, Madrid, Spain.
Robbins, M. M., & Timm, D. H. (2009). Effects of strain pulse duration on tensile strain in a
perpetual pavement. In Proceedings from International Conference on Perpetual Pavement
2009, Columbus, OH.
Romanoschi, S. A., Gisi, A. J., & Dumitru, C. (2006). The dynamic response of Kansas perpetual
pavements under vehicle loading. Paper presented at the Proceedings for the International
Conference on Perpetual Pavements, Columbus, OH.
Romanoschi, S. A., Gisi, A. J., Portillo, M. M., & Dumitru, C. (2008). First findings from the
Kansas perpetual pavements experiment. Transportation Research Record, 2068, 41–48.
Sargand, S., Figueroa, J. L., Edwards, W., & Al-Rawashdeh, A. S. (2009). Performance
assessment of warm mix asphalt (WMA) pavements. Report No. FHWA/OH-2009/08 for the
Ohio Dept. of Transportation, Athens, OH: Ohio Research Institute for Transportation and the
Environment, Ohio University, September, 2009. http://cdm16007.contentdm.oclc.org/cdm/
ref/collection/p267401ccp2/id/4561. Accessed April 29, 2016.
Sargand, S., Figueroa, J. L., & Romanello, M. (2008). Instrumentation of the WAY-30 test
pavements. (Report No. FHWA/OH-2008/7). Athens, OH: Ohio Research Institute for
Transportation and the Environment, Ohio Department of Transportation.
Sargand, S., Khoury, I., Jordan, B., Scheer, M., & Cichocki, P. (2015). Implementation and
thickness optimization of perpetual pavements in Ohio. Report FHWA/OH-2015/17 for Ohio
Dept. of Transportation, Athens, OH: Ohio Research Institute for Transportation and the
Environment, Ohio University, June, 2015. http://cdm16007.contentdm.oclc.org/cdm/ref/
collection/p267401ccp2/id/12760. Accessed April 29, 2016.
Sargand, S. M., Khoury, I. S., Romanello, M. T., & Figueroa, J. L. (2006). Seasonal and load
response instrumentation of the WAY-30 perpetual pavements. Paper presented at the
Proceedings for the International Conference on Perpetual Pavements, Columbus, OH.
Timm, D. H., Robbins, M. M., Willis, J. R., Tran, N., & Taylor, A. J. (2013). Field and laboratory
study of high-polymer mixtures at the NCAT test track. Final Report (NCAT Report 13-03).
Auburn, AL: National Center for Asphalt Technology.
Willis, J. R. (2009). Field-based strain thresholds for flexible perpetual pavement design, Ph.D.
Dissertation, Auburn University, Auburn AL, May 9, 2009. http://etd.auburn.edu/etd/bitstream/
handle/10415/1580/Willis_James_17.pdf?sequence=1. Viewed September 9, 2011.
Structural Study of Perpetual Pavement Performance in Ohio 459

Willis, J. R., & Timm, D. H. (2009a). Field-based strain thresholds for flexible perpetual
pavement design. (NCAT Report 09-09). Auburn, AL: National Center for Asphalt
Technology.
Willis, J. R. & Timm, D. H. (2009b). A comparison of laboratory fatigue thresholds to measured
strains in full-scale pavements. In: Proceedings of the International Conference on Perpetual
Pavement 2009, Columbus, Ohio, September 30–October 2, 2009.
Willis, R., Timm, D., West, R., Powell, B., Robbins, M., Taylor, A., Smit, A., et al. (2009). Phase
III NCAT test track findings. (NCAT Report 09-08). Auburn, AL: National Center for Asphalt
Technology.
Witczak, M., Mamlouk, M, Souliman, M., & Zeiada, W. (2013), Laboratory validation of an
endurance limit for asphalt pavements. NCHRP Report 762. Washington, D.C: Transportation
Research Board.
Study of In-Service Asphalt Pavement
High-Temperature Deformation Based
on Accelerated Pavement Test

Liping Liu, Yu Yuan and Lijun Sun

Abstract This paper explores high temperature deformation regularity of asphalt

pavement/asphalt mixture test specimen under different conditions: in-service asphalt
pavement, 10 cm milled and overlaid asphalt course with and without vibration
rolling, normal overlaid asphalt pavement section, based on full-scale accelerated
pavement testing equipment MLS66 and lab repeat wheel rolling test. The research
findings reveal that asphalt pavements under different conditions show similar high
temperature deformation regularity apparently in three-stage deformation. The degree
of compaction and the age have significant impact on high temperature deformation
performance of asphalt pavement. Based on the number of accumulation loading
applied to the pavement and rutting depth, preliminary conversion relationship
between accelerated loading actions and actual standard loading actions is drawn from
the perspective of resistance to deformation. The concept of deformation stability is
introduced to indicate the high-temperature deformation resistance of asphalt pave-
ment and the high-temperature APT deformation stability results could be converted
to the lab test deformation stability results with a conversion coefficient of 1.2–2.8.

Keywords Full-scale accelerated pavement testing In-service asphalt pavement



High temperature deformation Degree of compaction

1 Introduction

Accelerated pavement test (APT) refers to a test in which loadings by prototype

wheel loads were repeated on a prototype or practical multi-layer pavement
structure in a short duration and under controlled conditions to investigate the

L. Liu
Y. Yuan
L. Sun (&)

Key Laboratory of Road and Traffic Engineering of the Ministry of Education,
Tongji University, 4800 Cao’an Road, Shanghai 201804, China
e-mail: [email protected]
L. Liu
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 461

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_30
462 L. Liu et al.

performance of the structure. This method has been widely applied for the per-
formance evaluation of materials and the pavement structure for years (Galal and
White 1999; Harvey and Bejarano 2000; Hugo and Martin 2004; Lee et al. 2006).
Nevertheless, it was not given much attention in the early days. It was until the end
of the Strategic Highway Research Program (SHRP) when pavement damages
could not be explained by traditional theories for the drastically increasing traffic,
and also in order to verify the research achievements of SUPERPAVE that APT
attracted significant attentions. Current applications of APT focus on comparisons
of different structures, materials and loading patterns, modelling and material per-
formance analysis, and were generally indoor or specific field tests (Cho et al. 2010;
NCAT 2011; Ullidtz et al. 2006, 2010a, b). Few studies of in-service asphalt
pavement by APT have been reported.
Currently, studies related to APT are focusing on loading acceleration, envi-
ronment control and holographic detections on condition that the structure and loads
are maintained (Sun et al. 2013). The main functions of APT are as follows: (1) to
monitor and record relevant information and analyze the interactions of different
factors; (2) to investigate the quantitative effects of loadings and environment on
fatigue and deformation of the pavements; (3) to establish a reliable database to
facilitate the pavement design (Sun et al. 2013). Therefore, it can be concluded that
APT is a reliable method for pavement studies. In this article, high-temperature
deformations of in-service asphalt pavement under different conditions were
investigated using the Mobile Load Simulator 66 (MLS 66). High-temperature
loading tests were applied on pavements before and after milling-planing overlay
coating. Additionally, the conversion from accelerated loading to practical loading in
terms of deformation resistance was explored based on the conditions (total practical
loadings experienced and rutting depth generated) of the pavement tested.

2 Accelerated Loading Equipment and Test Proposals

2.1 MLS 66

Figure 1 shows the MLS 66 used in this study. With dimensions of

15 × 2.87 × 3.5/3.8 (m), this equipment can apply practical wheel loads to study

Fig. 1 Mobile load simulator 66 (MLS 66)

Study of In-Service Asphalt Pavement High-Temperature Deformation … 463

Fig. 2 MLS profilometer driver-P2003

the effects of the axle load proposed on the pavement. MLS 66 consists of a rigid
frame and six wheel carriers equipped with twin tires (305/70 R22.5). The wheel
carriers can move in vertical circular guideways and the wheels would apply a
pressure (loading) on the pavement once the guideway is at the bottom. Consisting
of two hydraulic drives and a nitrogen container, the wheels can apply a load of
75 kN. The maximum speed of the carrier is 6 m/s and 1000 loadings/h can be
applied by each wheel. Additionally, MLS 66 is equipped with a heating system (to
heat up the pavement, if necessary) and the Profilometer Driver-P2003 (to measure
the pavement deformation) (Fig. 2). Wheel loading, rutting deformation and other
parameters are monitored in real-time and shown PCs that are wirelessly connected
to the equipment so that timely adjustments can be made.

2.2 Test Pavement

The target asphalt pavement in this study was an improvement section (pile
No. K20 +259*+389) of the Baoqian Road, Shanghai. As a typical secondary
road, the original pavement consisted of a 3 cm layer of fine grained type asphalt
concrete, a 9 cm layer of coarse grained type asphalt concrete, a 45 cm triple slag
layer of fly ash, and a 15 cm sandy gravel layer. Since the last reconstruction in
2002, the traffic has been rapidly increasing, resulting in severe pavement damages
such as rutting, transverse cracks, pitted surface and pot holes. The PCI testing
result of the target pavement was 70 and no significant damages other than slight
rutting were observed. The target section was within a truck lane and the whole
section was closed during the entire testing process for safety purpose. Besides
essential treatment, a double overlay structure consisting of 5 cm AC-20C and
4 cm SBS modified SMA13 was applied and elevation transition sections were
levelinged using AC-20C.
464 L. Liu et al.

2.3 Accelerated Test Proposal

As the wheel tracks of MLS 66 were 6.6 m, seven sections (inter-distance = 1 m)

were involved to maximize the data points obtained. The testing zone was the
intersection of wheel tracks and the sections, as shown in Fig. 3. A 14 h testing
(equipment was turned off during nights) was executed per day and loading cycles
were not limited. Instead, the loading was terminated once a 2 cm rutting or sig-
nificant fatigue cracks were observed. The pavement was heated up (via the heating
plates) to and maintained at 60 °C. However, the spontaneous cooling by the
atmosphere resulted in a temperature variation of 30–60 °C.
At the end of each loading session, deformations of the pavement were measured
using MLS Profilometer Driver-P2003. This profilometer consists of
wheels-equipped guideways, a charging box and a PC. In the measurement process,
a wheel was rolling on the section to be tested and the deformation was recorded
regularly. With a maximum 1500 mm range and precision of 0.1 mm, this pro-
filometer could precisely monitor the deformations of all sections.
Four cases were included in this study:
(1) High-temperature loading on in-service pavement till deformation exceeded
2 cm;
(2) High-temperature loading on milling-planing overlay pavement (compaction
without vibration, 60 m) till deformation exceeded 2 cm;

Section 1
1m
Section 2
1m
Section 3
1m
Section 4
1m
Section 5
1m
Section 6

Driving direction 1m
Section 7

Fig. 3 Intersection of wheel tracks and the sections tested

Study of In-Service Asphalt Pavement High-Temperature Deformation … 465

(3) High-temperature loading on milling-planing overlay pavement (compaction

with vibration, 60 m) till deformation exceeded 2 cm;
(4) High-temperature loading on pavement overlayed 3 months ago (normal
construction) till deformation exceeded 2 cm.

2.4 Lab Repeat Wheel Rolling Test

Based on the four cases mentioned above, 30 cm × 30 cm × 10 cm samples were

taken in un-loading area for lab wheel rutting tests. The method was as described in
T07190-2011 of JTG E20-2011 (Standard test methods 2011). Deformation data
were recorded at the end of each loading session, the test temperature was 60 °C
and the loading rate was 42 times/min. The test was terminated once sample failure
was observed or the accumulated rutting depth exceeded 10 mm (10 % of the
sample thickness).

3 Results and Discussion

3.1 High-Temperature Deformation of Different Cases

in APT

As shown in Fig. 4, as the loading cycles increased, sinkages were observed on the
area in contact with the tire, while uplift due to shear flow deformation of the
asphalt mixture were observed on the sides.
Figure 5 shows the deformation curves of the 7th section of the existing pavement
at different loading cycles (measurement width = 1500 mm, step distance = 4 mm),

Fig. 4 Pavement after APT

466 L. Liu et al.

N
pavement elevation/mm

Lateral Position, mm

Fig. 5 Deformation curves of the 7th section at different loading cycles

which is generated automatically by the system. As can be seen, sinkages were

observed on the area in contact with the tire, while uplift due to shear flow defor-
mation of the asphalt mixture were observed on the sides. The uplift deformation is
defined as the maximum uplift height of each cross-section and the sinkage defor-
mation is defined as the maximum sinkage deformation of each cross-section. The
sum of the maximum uplift height and the maximum sinkage deformation is defined
as the rutting depth (Li et al. 2009). The rutting depth increased continuously from
7.8 mm before testing to 10 mm after 300,000 loading cycles.
Figure 6 shows the rutting depth versus loading cycles in all cases. Temperatures
of different loading sessions were also indicated. As can be seen, rutting on different
sections in the same case exhibited similar trends, while the variations of rutting on
different sections in different cases differed from each other (variation in Case 4 was
the lowest). This can be attributed to the non-uniformity of materials, structures and
construction. With 20 mm as the critical rutting level, maximum loading cycles in
Case 1, 2, 3 and 4 were 400,000, 100,000, 130,000 and 200,000, respectively,
indicating that the difference in the maximum loading cycles in different cases.
Hence, it is concluded that the in-service pavements exhibited better
high-temperature deformation resistance because of asphalt aging and structural
stabilization. Fully compaction also contributed to the deformation resistance, and
normal construction and the age was preferred for reduced deformations.
Figure 7 shows the average rutting depth of all sections of case 1 tested. As can
be seen, the rutting curve in Case 1 can be divided into the creep stage and the
accelerated deformation stage (in the three-stage model for permanent
Study of In-Service Asphalt Pavement High-Temperature Deformation … 467

Temperature at depth of 20mm from surface/ :

59 30 72 49 54

(a) Case 1: high-temperature loading on existing pavement

Temperature at depth of 20mm from surface/ :

56 31 36

(b) Case 2: high-temperature loading on milling-planing overlay pavement (compaction without

vibration)

Fig. 6 Rutting depth versus loading cycles in four cases. Note The criterion to stop the test is the
appearance of rutting depth more than 2 cm, while the real-time deformation couldn’t be
monitored during the loading process, which could merely be judged empirically. And, the
condition of the equipments and testing time can also make affections to the loading cycles.
Therefore, the times of load cycles and times gaps of the 4 cases are different

deformations). In the former stage, the rutting growth rate was stabilized (slightly
decreasing); in the latter stage, the rutting growth rate was significantly higher.
Figure 8 shows the average rutting depth as a function of loading cycles in all
cases. Herein, the existing deformation of in-service pavements (7.8 mm) was taken
into consideration. As can be seen, the rutting depth of new pavements of case 4
468 L. Liu et al.

Temperature at depth of 20mm from surface/ :

33 49 46 37

(c) Case 3: high-temperature loading on milling-planing overlay pavement (compaction with

vibration)

Temperature at depth of 20mm from surface/ :

51 52 48 56 56 57

(d) Case 4: High-temperature loading on pavement overlayed three months ago (normal
construction)

Fig. 6 (continued)

reached 7.8 mm after 40,000 loading cycles in APT. According to statistics, the
total axle loading applied on the pavement before testing was 1.49 × 107
(3.43 × 106 if the transverse distribution coefficient (Liu and Sun 2005) was con-
sidered). Therefore, in the case of Baoqian Road, it was concluded that 40,000 APT
loadings (0.9 MPa) may be converted to 3.43 million standard axle loadings
(0.7 MPa). Additionally, the permanent deformation growth could be described by
Study of In-Service Asphalt Pavement High-Temperature Deformation … 469

Fig. 7 Average rutting depth in Case 1

Fig. 8 Average rutting depths versus loading cycles in all cases

the AASHTO 2002 three-stage model (first stage = rapid deformation growth for
compaction, second stage = creep growth, third stage = accelerated deformation
growth).
470 L. Liu et al.

3.2 High-Temperature Deformation in Lab Repeat Wheel

Rolling Test

Figure 9 shows the high-temperature deformations as a function of loading cycles

in lab tests for cutting samples. As can be seen, lab repeat wheel rolling tests were
capable of precisely describing the deformation of asphalt mixture specimen, as
well as the deformation resistances of pavements cutting samples with and without
milling-planing overlay. The initial deformation growth rate of in-service pavement
was extremely low, which may be attributed to the hard shell (Wei et al. 2012)
formed on top of the aged asphalt layer. After considerable loading cycles, the hard
shell was ruptured and the deformation was accelerated in the second and third
stage. For samples of overlay pavements, the deformations in first and second
stages were improved, which is consistent with the APT results.
According to Miner’s Rule, the rutting depth of asphalt mixtures was accumu-
lated in a linear superposition pattern. At a certain temperature (60 °C in this study),
a rutting depth of Ri was obtained at loading cycles of Ni. Ni/Ri, which is the
number of loading cycles for 1 mm deformation, is then defined as the deformation
stability. Critical points of curves in Fig. 9 were obtained after curve smoothing and
loading cycles corresponding to different stages could be defined. In this study,
points corresponding to 1260 and 27,720 cycles were defined as the critical points
of the second stage and the third stage. As shown in Table 1, samples in Case 1
exhibited significantly higher deformation stability compared with samples in other
cases. This can be explained by the fact that the compaction of samples in Case 1
had been achieved during the service of the pavement. On the other hand, the third
stage, in which deformation growth was significantly accelerated and the defor-
mation stability was drastically reduced, of samples in Case 1 started earlier than

Fig. 9 Sample deformation versus loading cycles in lab tests

Study of In-Service Asphalt Pavement High-Temperature Deformation … 471

Table 1 Deformation stabilities of samples in all cases of lab tests

Case Case 1 Case 2 Case 3 Case 4
Deformation in the first stage (mm) 0.194 2.022 1.58 1.141
Deformation in the second stage (mm) 7.509 6.952 7.054 6.317
Deformation stability (cycle/mm) 3692 3987 3930 4388
Deformation stability in the first stage 6494 623 797 1104
(cycles/mm)
Deformation stability in the second stage 3617 5348 5553 5775
(cycles/mm)

Table 2 Conversion coefficients of site APT results and lab test results
Case Line Case 1 Case 2 Case 3 Case 4
number
APT loading cycles (cycles)/20 mm ① 165,000 100,000 142,000 248,000
APT deformation stability obtained in site ② 8250 5000 7100 12,400
measurements (cycle/mm)
deformation stability obtained in lab ③ 3692 3987 3930 4388
measurements (cycle/mm)
Conversion coefficient ④ 2.23 1.25 1.81 2.83

that of other samples did. Additionally, the deformation stability serves as an

indicator of the deformation resistance as the deformation resistance was positively
related to the deformation stability, which was demonstrated in all cases.
The primary conversion of APT site measurements and lab tests could be
obtained by analysing APT results in terms of the deformation stability. As shown
in Table 2, the high-temperature APT deformation stability results could be con-
verted to the lab test deformation stability results with a conversion coefficient of
1.2–2.8. Explains as follows: Line ① gives the loading cycles obtained from the
APT results under the 4 cases (Fig. 6) when the rutting depth reached 20 mm; Line
② gives the APT deformation stability which is defined as loading cycles divided
by 20 mm; Line ③ gives the deformation stability of laboratory test, which is from
Table 1; Line ④ gives the conversion coefficient of site APT results and lab test
results which is defined as the figure in Line ② divided by the corresponding one in
Line ③.

4 Conclusions

High-temperature deformations of in-service asphalt pavement/asphalt mixture test

specimen under different conditions were investigated using the Mobile Load
Simulator 66 (MLS 66) and lab repeat wheel rolling test. The cases investigated
472 L. Liu et al.

include existing pavements, milling-planing overlay pavement (compaction with

vibration), milling-planing overlay pavement (compaction without vibration), and
overlay pavement after standard construction. The following conclusions can be
drawn:
• Three deformation stages, including the compaction stage, the creep stage and
the rupture stage, have been demonstrated in both APT and lab tests;
• Good compaction and properly service durations can reduce the deformation
rate, thus increasing pavement service life;
• The high-temperature deformation resistance can be indicated by the deforma-
tion stability, which is defined as the number of loading cycles for 1 mm
deformation;
• Based on the existing rutting depth and the cumulative number of standard loads
experienced by the pavements, the loading cycles in APT can be converted into
practical standard loading cycles;
• The high-temperature APT deformation stability results could be converted to
the lab test deformation stability results with a conversion coefficient of 1.2–2.8.

Acknowledgments This research was sponsored by Shanghai urban construction and manage-
ment committee, China (project number: 2013-004-004). The authors gratefully acknowledge its
financial support for allowing them the opportunity to perform the present study.

References

Cho, N. H., Suh, Y. C., Mun, S., & Cho, Y. H. (2010). Accelerated pavement testing used for the
rutting model calibration of Korean mechanistic-empirical pavement design guide. ISAP 2010.
Galal, K. A., & White, T. D. (1999). INDOT-APT test facility experience paper: CS8-4.
International Conference on Accelerated Pavement Testing. Reno, Nevada. October 18–20,
1999.
Harvey, J., & Bejarano, M. (2000). Performance of Caltrans asphalt concrete and asphalt-rubber
hot mix overlays at moderate temperatures—Accelerated pavement testing evaluation.
California Department of Transportation.
Hugo, F., & Martin, A. E. (2004). Significant findings from full-scale accelerated pavement
testing. Transportation Research Board.
Lee, S. J., Seo, Y., Kim, Y. R. (2006). Performance prediction of asphalt pavements loaded by
MMLS3 using material-level laboratory test methods and performance models. In 10th
International conference on the structural design of asphalt pavements.
Li, L., Chen, J., & Su, Z. (2009). Analysis of uplift deformation and loose phenomena of asphalt
layer. Journal of Tongji University (Natural Science), 37(1), 52–56.
Liu, L., & Sun, L. (2005). Research on wheelpath lateral distribution for freeway asphalt
pavements. Journal of Tongji University (Natural science), 11, 31–34+50.
NCAT. (2011). Pavement test track research findings (2000–2010).
Standard test methods of bitumen and bituminous mixtures for highway engineering
(T07190-2011). (2011). The ministry of transport of the People’s Republic of China.
Beijing: The People’s Communications Press.
Study of In-Service Asphalt Pavement High-Temperature Deformation … 473

Sun, L., et al. (2013). Structural behavior study for asphalt pavements. Shanghai: Tongji
University Press.
Ullidtz, P., Harvey, J., Basheer, I., Wu, R., & Lea, J. (2010). Process of developing a
mechanistic-empirical asphalt pavement design system for California. In10th International
Conference on the Structural Design of Asphalt Pavements.
Ullidtz, P., Harvey, J., Kanekanti, V., Tsai, B. W., & Monismith, C. (2006). Calibration of
mechanistic-empirical models using the California heavy vehicle simulators. In 10th
International Conference on the Structural Design of Asphalt Pavements.
Ullidtz, P., Mateos, A., Ayuso, J., Harvey, J., & Basheer, I. (2010). Simulation of full scale
accelerated pavement test from CEDEX using the Californian predictive pavement design
system (CalME). In 10th International conference on the structural design of asphalt
pavements.
Wei, X., Zheng, J., & Wu, L. (2012). Failure mode of soft clay hard shell layer foundation. Journal
of Highway and Transportation Research and Development, 8, 31–35+43.
Thickness and Binder Type Evaluation
of a 4.75-mm Asphalt Mixture Using
Accelerated Pavement Testing

James Howard Greene and Bouzid Choubane

Abstract An increasing number of transportation agencies are investing in thin

overlays consisting of small aggregate size asphalt mixtures to address routine
maintenance needs and pavement preservation initiatives. Reported benefits of thin
overlays include a smooth riding surface, reduced tire-pavement noise, and the
ability to more easily maintain and control grade and cross-slope. The smallest
aggregate size mixture the Florida Department of Transportation (FDOT) currently
specifies is a 9.5-mm nominal maximum aggregate size (NMAS) mixture. When
used as a dense graded friction course, the 9.5-mm NMAS mixture is placed in
1.0 in. lifts, which allows more flexibility and better optimization of milling depths,
as well as structural or overbuild lift thickness to achieve a structurally appropriate
and economical pavement. However, the ability to place an even thinner asphalt
layer may potentially provide designers with more cost-effective alternatives.
FDOT recently investigated the layer thickness of a 4.75-mm NMAS mixture using
accelerated pavement testing (APT). During the APT study, overlay thicknesses of
0.5, 0.75, and 1.0 in. were placed with PG 67-22 (unmodified) and PG 76-22
(polymer modified) asphalt binders. Pavement instrumentation, laboratory data, and
results from a field test section were also considered. This report documents the
findings of the APT study and provides recommendations for the use and appro-
priate thickness of a 4.75-mm NMAS overlay.

1 Background

1.1 Introduction

The use of a 4.75 mm nominal maximum aggregate size (NMAS) overlay has
gained interest among many transportation agencies as a pavement maintenance or
preservation technique. The Florida Department of Transportation (FDOT) recently

J.H. Greene
B. Choubane (&)

Florida Department of Transportation, Tallahassee, FL, USA
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 475

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_31
476 J.H. Greene and B. Choubane

placed several short sections of a 4.75 NMAS Superpave designed friction course
(FC-4.75) as maintenance overlays and also constructed two sections as an
experimental pavement preservation technique. Local agencies in Florida are also
interested in the potential benefits of a thin FC-4.75 overlay due to the reported
benefits. Benefits of thin 4.75 mm NMAS overlays found in literature include a
smooth riding surface, reduced tire-pavement noise, and the ability to more easily
maintain and control grade and cross-slope (Watson and Heitzman 2014). Other
benefits of mixtures with smaller aggregate sizes include good workability and
decreased permeability (West et al. 2011). Perhaps one of the most important
advantages to FDOT’s pavement designers is that a thin lift or overlay allows
flexibility and better optimization of milling depths, structural layers, and overbuild
lifts to achieve structurally adequate and economical pavement sections. FDOT
currently allows 1 in. lifts of Superpave designed 9.5-mm NMAS mixtures
(SP-9.5). However, the ability to place an even thinner asphalt layer may potentially
provide designers with more cost-effective alternatives. According to NCHRP
Report 531, lift thicknesses should be three to four times the NMAS to achieve
adequate compaction (Brown et al. 2004). Considering this criteria, a 4.75-mm
NMAS overlay could be placed thinner than 1 in. assuming it meets other
requirements of a surface course such as resistance to rutting and cracking and
adequate friction properties.
The National Center for Asphalt Technology (NCAT) conducted a survey in
2004 that found 11 states used a 4.75-mm mixture primarily as surface mixtures and
leveling courses (West et al. 2011). Despite the extensive use by several states,
documented performance of 4.75-mm NMAS mixtures is limited. The lack of
performance studies is likely due to the predominant use of the mixture as surface
courses for light volume roadways and as leveling courses. One of the first studies
to examine the performance of 4.75-mm NMAS mixtures found that these mixtures
had excessive laboratory rutting measured with the Asphalt Pavement Analyzer
(APA) due to high asphalt contents (James et al. 2003). Kansas researchers
observed that rutting performance is aggregate specific and higher binder grade may
not improve performance using the Hamburg Wheel Tracking device (Rahman
et al. 2010). A recent NCAT study revealed that rutting resistance of 4.75-mm
mixtures assessed with the Mixture Verification Tester can be improved by
increasing dust content, using high angularity aggregate, and using stiffer binders
(West et al. 2011). However, it should be noted that laboratory testing of asphalt
mixtures is typically conducted on specimens much thicker than what will typically
be placed in the field. Stress conditions that a thin asphalt overlay is subjected to
will likely be much different than what laboratory methods can measure.
Another concern that some agencies have is the potential for hydroplaning and
friction loss (Williams 2008; Li et al. 2011). A recent FDOT study that analyzed
crash statistics, geometrical data, pavement condition, and pavement surface types
found that conditions for hydroplaning are more likely on dense-graded pavement
surfaces (Gunaratne et al. 2012). Likewise, an NCHRP study linked surface texture
to water film thickness and hydroplaning potential through the following rela-
tionship (Anderson et al. 1998).
Thickness and Binder Type Evaluation of a 4.75-mm Asphalt Mixture … 477

 0:5
nLi
WFT ¼ MTD
36:1S0:5

where
WFT Water film thickness (in.)
n Manning’s roughness coefficient
L Drainage path length (in.)
i Rainfall rate (in./h)
S Slope of drainage path (mm/mm)
MTD Mean texture depth (in.)
Considering the limitations of laboratory testing of 4.75 mm NMAS mixtures,
full-scale field tests and accelerated pavement testing (APT) are perhaps the best
methods to assess the performance of these thin overlays. One of the only APT
studies of a 4.75-mm NMAS mixture was conducted at the Turner Fairbank
Highway Research Center (Li et al. 2011). The 4.75-mm mixture inlay was
approximately 1 in. thick and was evaluated for rutting and fatigue resistance.
Full-scale rutting indicated satisfactory performance and it was concluded that 4.75
NMAS overlays have the ability to significantly delay top-down cracking as long as
the binder does not age excessively. Lane slices and cores revealed air void contents
that ranged from 10.4–16.3 %, which was significantly higher than the target of
10 %. The authors also suggested that performance may have been better had field
densities not been so low.

1.2 Objective and Scope

The primary objective of this study was to determine the optimum thickness of a
Superpave designed 4.75 mm NMAS friction course (FC-4.75) through full-scale
accelerated pavement testing (APT). Test sections were constructed with an
FC-4.75 thickness of 0.5, 0.75, and 1.0 in. In addition, the FC-4.75 included a PG
67-22 unmodified asphalt binder for one set of test sections while a second set of
sections used a polymer modified PG 76-22 asphalt binder. The APT performance
of the FC-4.75 overlays were compared to results of APT sections surfaced with a
Superpave designed fine-graded 12.5 mm NMAS structural course (SP-12.5) that
was found to have good performance in a previous APT study. Laboratory testing
supplemented the APT results and provided an indication of fracture resistance.
Finally, field measurements from a pavement preservation test section constructed
in 2012 validated findings of the APT study.
478 J.H. Greene and B. Choubane

2 Accelerated Pavement Testing

2.1 Test Lane Construction

Six full-scale pavement test lanes measuring 12 ft wide each were constructed for the
APT study. Two of the lanes served as a control and consisted of two 1.5 in. thick
fine-graded SP-12.5 mixtures that were paved over a milled portion of a SP-12.5
mixture from a previous study. The only difference in these lanes were the type of
asphalt binder. One lane included a PG 67-22 unmodified asphalt binder in both
SP-12.5 lifts while the second lane used a polymer modified PG 76-22 asphalt binder
in both lifts. The other four lanes included one 1.5 in. lift of a SP-12.5 with a PG
67-22 unmodified asphalt binder followed by a 1.5 in. lift with a polymer modified
PG 76-22 asphalt binder. All four lanes were surfaced with a FC-4.75 mixture with a
staggered thickness along the length of the lane of 0.5, 0.75, and 1.0 in. The FC-4.75
overlay included a PG 67-22 unmodified asphalt binder in two of the lanes while the
other two included a polymer modified PG 76-22 asphalt binder. The FC-4.75 had a
target asphalt content of 6.5 % and target air voids of 5.9 %. Both mixture types
were Traffic Level C (3–10 million Equivalent Single Axle Loads) and used 75
gyrations. The supporting layers of all of the test lanes included a 10.5 in. limerock

Table 1 Job mix formula and as-built properties

Sieve size FC-4.75 SP-12.5
Job As-built As-built Job As-built As-built
mix with PG with PG mix with PG with PG
formula 67-22 76-22 formula 67-22 76-22
3/4 in. 100 100.0 100.0 100 100.0 100.0
1/2 in. 100 100.0 100.0 100 97.3 98.4
3/8 in. 100 100.0 100.0 87 86.3 87.5
#4 99 98.7 98.7 62 57.8 61.9
#8 77 79.1 78.2 41 39.2 42.4
#16 56 55.7 55.0 29 28.7 31.1
#30 39 40.3 40.0 22 22.1 24.0
#50 26 26.7 26.0 12 13.2 13.8
#100 15 15.1 15.5 4 4.9 5.2
#200 9 8.9 9.5 2 2.8 2.9
% AC 6.5 6.0 100.0 5.1 4.9 4.9
% Air voids at 5.9 6.3 4.6 4.0 3.3 3.9
Ndes
% Density from NA NA NA 93.0 93.2 93.6
cores
Thickness and Binder Type Evaluation of a 4.75-mm Asphalt Mixture … 479

Fig. 1 Pavement structures evaluated

base and a 12 in. stabilized subgrade consisting of a mixture of limerock and native
soil. The pavement structure represented a typical Florida roadway and was con-
structed according to standard FDOT specifications. Table 1 shows the design and
as-built gradations, asphalt content, and air voids of each mixture. Figure 1 illus-
trates the pavement structures. The FC-4.75 mixture design was based on a FDOT
developmental specification, which was developed considering criteria proposed by
the National Center for Asphalt Technology (West et al. 2011). Granite aggregate
was used in all of the asphalt mixtures. The FC-4.75 mixture used aggregate from a
Georgia pit and local sand. The SP-12.5 mixture used granite from the same Georgia
pit, but also used granite from a Nova Scotia source.

2.2 Accelerated Loading

FDOT uses APT to allow for a faster and more practical assessment of pavements
under closely simulated in-service conditions. In Florida’s APT program, the
accelerated loading is performed using a Heavy Vehicle Simulator (HVS).
Accelerated loading is performed uni-directionally and the pavement temperature is
maintained at 120 °F. To normalize the effect of construction and differential
pavement aging prior to testing, at least three tests are performed for each pavement
section using a randomized test sequence. Rut depth measurements are conducted
periodically using a laser profiling system mounted on the either side of the HVS
carriage. Each test section is trafficked until a rut depth of 0.5 in. is accumulated or
100,000 passes is achieved and rutting has stabilized. In this study, accelerated
loading was performed using a wide-base Super Single tire (Goodyear G286 A SS,
480 J.H. Greene and B. Choubane

Fig. 2 FDOT’s heavy vehicle simulator

0.5 inch 0.75 inch 1.0 inch Control (PG 67-22)

0.5

0.4
Rut Depth (inch)

0.3

0.2

0.1

0.0
0 20,000 40,000 60,000 80,000 100,000
HVS Pass

Fig. 3 Rut depth measurements for FC-4.75 overlays and SP-12.5 control with PG 67-22 asphalt
binder
Thickness and Binder Type Evaluation of a 4.75-mm Asphalt Mixture … 481

0.5 inch 0.75 inch 1 inch

Control (PG 76-22) Control (PG 67-22)
0.5

0.4
Rut Depth (inch)

0.3

0.2

0.1

0.0
0 20,000 40,000 60,000 80,000 100,000
HVS Pass

Fig. 4 Rut depth measurements for FC-4.75 overlays and SP-12.5 control with PG 76-22 asphalt
binder

425/65R22.5) loaded to 9 kips and inflated to 100 psi. Test sections were trafficked
with a wheel wander of 4 in. The Super Single tire was selected because it
accelerates pavement damage. A previous HVS study found that the Super Single
tire generated a 0.5 in. rut depth approximately 10 times faster than a standard dual
tire when loading dense graded surfaces (Greene et al. 2010) Fig. 2 shows FDOT’s
HVS with insulated panels installed to maintain the test temperature. More detailed
information of FDOT’s APT facility is described elsewhere (Byron et al. 2004).
Figures 3 and 4 show the average rut progression of each test section. In general,
thicker FC-4.75 overlays performed better than thinner overlays. All FC-4.75
overlay sections, regardless of thickness or asphalt binder type, had similar or better
rut resistance than the SP-12.5 control without polymer modified binder. This
indicates that all overlay options would provide adequate rutting resistance on low
volume roadways. On roadways with higher volumes of truck traffic where a
polymer modified binder is required, the 0.75 and 1.0 in. FC-4.75 overlay had
similar or better rut resistance than the SP-12.5 control with PG 76-22 asphalt binder.
Cores and lane slices taken from each test section indicated that rutting was confined
to the upper FC-4.75 or SP-12.5 layer and did not extend to the granular layers.
As mentioned previously, it is an FDOT APT standard to heat the pavement to
120 °F when evaluating rut resistance. This elevated temperature simulates summer
asphalt conditions and accelerates rutting. It is uncommon for cracks to form during
HVS testing under these conditions. However, hairline longitudinal cracks were
observed on several 0.5 and 0.75 in. FC-4.75 test sections with PG 67-22 asphalt
binder. The cracks formed approximately 6–8 in. from the tire edge as shown in
482 J.H. Greene and B. Choubane

Fig. 5 Hairline longitudinal crack on 0.75 in. FC-4.75 overlay with PG 67-22 asphalt binder

0.5 inch (PG 67-22) 1 inch (PG 67-22)

160
140
Tensile Microstrain

120
100
80
60
40
20
0
3 5 7 9 11
Distance from Tire Edge (inch)

Fig. 6 Surface strain measurements on FC-4.75 overlays

Fig. 5. In response to the observations, surface strain gauges were installed at

various offsets from the tire edge on a 0.5 and 1 in. thick FC-4.75 section with PG
67-22 asphalt binder. Figure 6 shows that the maximum tensile strain occurred at a
similar location from the tire edge as the observed cracks. Cores revealed that the
cracks were extremely shallow and consisted of minor surface distortions.
Thickness and Binder Type Evaluation of a 4.75-mm Asphalt Mixture … 483

2.3 Laboratory Data

Asphalt material was sampled from delivery trucks during construction of the HVS
test lanes for laboratory performance testing. Samples were reheated and prepared
to determine dynamic modulus and fracture properties. Figure 7 shows the dynamic
modulus master curves of the FC-4.75 and SP-12.5 mixtures. Dynamic modulus
data at lower frequencies is often associated with rutting performance while low
temperature cracking performance is related to dynamic modulus data collected at
higher frequencies. In general, binder grade does not appear to have an effect on the
dynamic modulus of either mixture. The SP-12.5 mixtures have a slightly greater
dynamic modulus at higher frequencies which suggests better low temperature
crack resistance. The additional asphalt content and smaller maximum aggregate
size of the FC-4.75 may have resulted in the slightly lower dynamic modulus.
The Energy Ratio (ER) concept and the Laboratory Overlay Tester (OT) were
used to assess mixture fracture properties. The Energy Ratio model links asphalt
mixture damage to dissipated creep strain energy (DCSEf), which is a function of
the maximum tensile strength, resilient modulus of the mixture, and creep com-
pliance (Zhang et al. 2011). According to this model, two crack initiation thresholds
exist. Mixtures can fail under repeated loads with low magnitudes when the
accumulated damage induced exceeds the dissipated creep strain energy (DCSE)
threshold. Damage below this threshold is healable, while damage above this

SP-12.5 w/ PG 67-22 SP-12.5 w/ PG 76-22

FC-4.75 w/ PG 67-22 FC-4.75 w/ PG 76-22
3,500
Dynamic Modulus (ksi)

3,000
2,500
2,000
1,500
1,000
500
0
0.000001 0.0001 0.01 1 100 10000 1000000
Reduced Frequency (Hz)

Fig. 7 Dynamic modulus of FC-4.75 mixtures and reference SP-12.5 mixtures

484 J.H. Greene and B. Choubane

threshold will initiate a macro-crack or will propagate a macro-crack if one already

exists. The second threshold occurs when the fracture energy (FE) is exceeded due
to a single critical axle load. The ER is defined as the ratio of the dissipated creep
strain energy at failure (DCSEf) and the minimum dissipated creep strain energy
required to produce a 2 in. crack (DCSEmin). To determine the properties required
to assess ER, indirect tensile (IDT) tests were performed at 50 °F. This test tem-
perature has been shown to correlate well with field cracking performance of
Florida pavements. Mixtures with good cracking performance have been found to
have an ER that exceeds 1.0 (Roque et al. 2004).
The Laboratory OT was developed to assess mixture crack resistance due to the
opening and closing of joints and cracks. Tests were conducted at 77 °F with a
0.025 in. displacement. Testing was conducted until a 93 % reduction of the
maximum load was recorded. Florida has not yet developed OT performance cri-
teria, but Texas has proposed that mixtures should have greater than 300 cycles for
adequate reflective cracking resistance and rich-bottom mixtures should have
greater than 750 cycles for satisfactory reflective cracking performance (Zhou and
Scullion 2004).
Table 2 shows the results of the ER and OT tests. Overall, all of the mixtures had
satisfactory cracking resistance properties. However, it is interesting to note that the
FC-4.75 mixture with polymer modified PG 76-22 asphalt binder has ER and OT
results greater than double the other mixtures. This suggests that this mixture may
also be a good candidate as an interlayer to resist reflection cracks. It should be
noted that reclaimed asphalt pavement (RAP) was not included in either of the
mixtures. RAP has been shown to reduce mixture fracture resistance depending on
the age and quality of the RAP. Figure 8 shows that ER and OT results have a
positive correlation and rank the mixture performance similarly.

Table 2 Energy ratio and overlay tester results

Mixture Energy ratio properties Overlay tester cycles to
Fracture energy DCSEf Energy failure
(kJ/m3) (kJ/m3) ratio
SP-12.5 with PG 2.3 2.1 2.8 362
67-22
SP-12.5 with PG 2.8 2.6 4.2 469
76-22
FC-4.75 with PG 3.2 2.9 4.7 701
67-22
FC-4.75 with PG 4.4 4.1 10.9 922
76-22
Thickness and Binder Type Evaluation of a 4.75-mm Asphalt Mixture … 485

Overlay Tester Cycles to Failure 1000

900
800
700
y = 63.51x + 253.83
600
R² = 0.84
500
400
300
200
100
0
0 2 4 6 8 10 12
Energy Ratio

Fig. 8 Energy ratio and overlay tester relationship

2.4 Friction and Surface Texture Measurements

of the APT Test Sections

Friction measurements of the FC-4.75 overlay were made using the Dynamic
Friction Tester (DFT). In addition, the Circular Track Meter (CTM) was used to
determine the mean profile depth (MPD). Ten friction and texture measurements
were randomly made outside of the loaded wheel path area on two of the lanes with
both binder types. Six friction and texture measurements were made within the
wheel paths to determine the potential impact of traffic. DFT measurements made at
40 mph (DFT40) were converted to Friction Number values at 40 mph using a
ribbed tire (FN40R) based on a relationship developed in a previous FDOT study
that included SP-9.5 and SP-12.5 surfaces (Choubane et al. 2012). No difference
was found between asphalt binders used in the FC-4.75. Table 3 shows the average
friction and texture measurements. There was approximately 39 % difference in
friction due to loaded wheel traffic. However, friction within the wheel path was
still satisfactory according to an FDOT minimum required FN40R of 35. The high
friction measurements may be due to the aged asphalt surface. Both set of

Table 3 FC-4.75 friction Location DFT40 × 100 FN40R MPD

and texture measurements (in.)
Outside wheel 0.65 62 0.009
path
Within wheel path 0.44 43 0.016
486 J.H. Greene and B. Choubane

FC-9.5 FC-12.5 Open-Graded FC

450
400
350
300
Lane Miles

250
200
150
100
50
0
0.016

0.031
0.013

0.019
0.022
0.025
0.028

0.034
0.037
0.040
0.043
0.046
0.049
0.052
0.055
0.058
0.061
0.064
0.067
0.070
MPD (inch)

Fig. 9 MPD distribution of FDOT friction courses

measurements were made after completion of the APT study, which was more than
2 years after construction of the test sections.
MPD values measured outside of the wheel path shown in Table 3 were gen-
erally less than those of typical FDOT SP-9.5 and SP-12.5 mixtures. MPD values
measured within the wheel path where typical of FDOT SP-9.5 mixtures, sug-
gesting the texture increased as traffic was applied. Figure 9 shows the MPD dis-
tribution of newly placed FDOT friction courses measured over an 8 year period
using a 64 kHz laser mounted on a vehicle travelling at highway speeds. An FDOT
study found the MPD determined with the CTM and the 64 kHz laser were similar
(Choubane et al. 2012).

3 Pavement Preservation Test Section

A pavement preservation test section consisting of ten treatments was constructed in

2012 on both directions of US-98 in Gulf County. At the time of construction, it
was standard for the asphalt binder to include 5 % ground tire rubber in Superpave
designed friction courses and 12 % ground tire rubber in the open-graded friction
course. Unlike standard FDOT pavement rehabilitation methods, several sections
were not milled prior to overlay to assess the effectiveness of the treatment in
delaying reflection cracks. Each test section is approximately 1500 ft long and has
been monitored annually since construction. Pavement smoothness and rut depth
were determined using a high speed inertial profiler. Friction measurements were
collected with a locked-wheel friction tester. A 64 kHz laser mounted on the
Thickness and Binder Type Evaluation of a 4.75-mm Asphalt Mixture … 487

Table 4 Pavement preservation test sections

Test Treatment Predominant crack percentage
section at year 4
Wheel path Outside
wheel path
1, 9 Control (no resurfacing) 30 % Class III 20 % Class
III
2 Microsurfacing 30 % Class I 5 % Class I
3 0.5 in. FC-4.75 overlay 25 % Class I 5 % Class I
4 0.75 in. FC-4.75 overlay 35 % Class I 5 % Class I
5 1 in. FC-9.5 overlay 10 % Class I 3 % Class I
6 1 in. mill and 1 in. FC-9.5 overlay 10 % Class I 7 % Class I
7 1.5 in. mill and 1.5 in. FC-12.5 overlay 10 % Class I 3 % Class I
8 0.75 in. open-graded friction course with heavy 10 % Class I 2 % Class I
polymer tack coat (bonded friction course)
10 1.5 in. hot-in-place recycling (HIPR) 15 % Class I, 5 % Class I
Rippling
Class I—Cracks less than or equal to 1/8 in.
Class II—Cracks greater than 1/8 in. and less than 1/4 in.
Class II—Cracks greater than 1/4 in.

locked-wheel friction tester was used to assess the surface texture in the wheel path.
Cracking was assessed through a visual survey. The average annual daily traffic for
this site is approximately 7000 vehicles per day with 5.5 % trucks.
Table 4 describes the test sections and the most recent crack survey observa-
tions. Two sections were not treated in order to serve as controls. The predominant

Year 1 Year 2 Year 3 Year 4

0.30

0.25
Rut Depth (inch)

0.20

0.15

0.10

0.05

0.00
Control 2 3 4 5 6 7 8 10
Test Section

Fig. 10 Pavement preservation test section rut depth

488 J.H. Greene and B. Choubane

Year 1 Year 2 Year 3 Year 4

140
120
Mean IRI (inch/mile)

100
80
60
40
20
0
Control 2 3 4 5 6 7 8 10
Test Section

Fig. 11 Pavement preservation test section smoothness

crack types for both travel directions are shown in the table. While the FC-4.75
overlay has improved the extent of cracking compared to the control, fine cracks are
propagating throughout the wheel paths at a greater rate than for most other
treatments. However, the 1 in. FC-9.5 overlay appears to be more crack resistant
than the thinner FC-4.75 overlays. Milling also appears to delay the presence of
reflection cracking.
Figures 10 and 11 show the pavement rut depth and smoothness of each section.
Treatment types and both control section measurements were averaged. Overall, the
sections with an FC-4.75 are performing well. Both sections have rut depths less
than 0.15 in. and are among the smoothest sections with a mean IRI of approxi-
mately 50 in./mile. Upon construction, all of the treatment sections had friction
values (FN40R) that ranged from 50 to 60 and all except the bonded friction course
had a drop in friction by approximately 15 % after the first year. The bonded
friction course values dropped by approximately 8 %. The average MPD for the
FC-4.75 overlay sections is 0.017 in., which is in the typical range of standard
dense-graded friction courses shown previously in Fig. 9. All of the sections
continue to have acceptable friction properties after 4 years of service.

4 Summary and Conclusions

A full-scale APT study, supporting laboratory testing, and field measurements have
shown that an FC-4.75 overlay may be effective as a preservation treatment when
properly designed and constructed. Laboratory testing showed that the FC-4.75
mixtures had greater cracking resistance than the SP-12.5 mixtures, but visual
Thickness and Binder Type Evaluation of a 4.75-mm Asphalt Mixture … 489

observations of the pavement preservation test sections showed that reflection

cracks propagated through the FC-4.75 overlay at a greater rate than a similarly
prepared but thicker FC-9.5 overlay. Milling prior to an overlay also appeared to
reduce the appearance of cracks. In addition, minor surface distortions were found
on the 0.5 and 0.75 thick FC-4.75 APT test sections with unmodified PG 67-22
asphalt binder. Another potential concern is the hydroplaning potential of a new
FC-4.75 mixture due to minimal surface texture. However, measurements of the
pavement preservation test sections and the APT test sections showed that FC-4.75
surface texture and friction values were in the typical range of standard FDOT
friction courses. Findings and recommendations based on the results of this study
can be found below.
• For low volume roadways, all combinations of binder (PG 67-22 and PG 76-22)
and thickness (0.5, 0.75, and 1 in.) provided adequate rutting resistance.
However, overlays of 0.5 and 0.75 in. with a PG 67-22 binder were found to
have minor surface distortions in the APT study which may indicate less
resistance to near surface cracking.
• For higher volume roadways, a 0.75 in. or thicker FC-4.75 overlay with PG
76-22 asphalt binder will provide adequate rutting and cracking resistance.
• An FC-4.75 overlay should not be used on roadways with design speeds of 50
mph or greater until more research is conducted to better quantify the
hydroplaning potential of newly placed FC-4.75 mixtures.
• Prior to overlay, the existing pavement should be milled to minimize reflection
cracking. If milling is not possible or cracks cannot be completely removed,
thicker overlays and the use of a polymer modified binder will help delay
reflection cracking.
• An SP-4.75 mixture with polymer modified binder showed potential as a crack
relief interlayer. More research is needed to optimize the mixture and quantify
the potential benefit.

Acknowledgments The work presented is the result of a team effort. The authors would like to
acknowledge the HVS Research Group and the Bituminous Laboratory of the State Materials
Office for their diligent efforts and contributing knowledge.

References

Anderson, D., Huebner, R., Reed, J., Warner, J., & Henry, J. (1998). NCHRP web document 16,
improved surface drainage of pavements. Washing DC: Transportation Research Board,
National Research Council.
Brown, E., Hainin, M., Cooley, A., & Hurley, G. (2004). NCHRP report 531 relationship of air
voids, lift thickness, and permeability in hot mix asphalt pavements. Washington, DC:
Transportation Research Board, National Research Council.
Byron, T., Choubane, B., & Tia, M. (2004). Assessing appropriate loading configuration in
accelerated pavement testing. In Proceedings, 2nd International Conference on Accelerated
Pavement Testing, Minneapolis, MN.
490 J.H. Greene and B. Choubane

Choubane, B., Lee, H., Holzschuher, C., Upshaw, P., & Jackson, M. (2012). Harmonization of
texture and friction measurements on Florida’s open and dense graded pavements.
Transportation Research Record 2306, Transportation Research Board of the National
Academies, Washington, DC, pp. 122–130.
Greene, J., Toros, U., Kim, S., Byron, T., & Choubane, B. (2010). Impact of wide-base tires on
pavement damage. In Transportation Research Record, Journal of the Transportation
Research Board, No 2010, National Research Council, Washington DC, pp. 82–90.
Gunaratne, M., Lu, Q., Yang, J., Metz, J., Jayasooriay, W., Yassin, M., & Amarasiri, S. (2012).
Hydroplaning on multi lane facilities. FDOT report no. BDK84 977-14, Florida Department of
Transportation, Tallahassee, FL.
James, R., Cooley, A., & Buchanan, S. (2003). Development of mix design criteria for 4.75-mm
superpave mixes. Transportation Research Record 1819, Transportation Research Board of the
National Academies, Washington, DC, pp. 125–133.
Li, S., Noureldin, S., Jiang, Y., & Sun, Y. (2011). Evaluation of pavement surface friction
treatments. Indiana Department of Transportation. Report No. FHWA/IN/JTRP-2012/04.
Indianapolis, In.
Li, X. Gibson, N., Qi, X., Clark, T., & McGhee, K. (2012). Laboratory and full-scale evaluation of
4.75 mm NMAS superpave overlay. Transportation Research Record 2293, Transportation
Research Board of the National Academies, Washington, DC, pp. 29–38.
Rahman, F., Hossain, M., Romanoschi, S., & Hobson, C. (2010). Evaluation of 4.75-mm
superpave mixture. CD-ROM Transportation Research Board of the National Academies,
Washington, DC.
Roque, R., Bigission, B., Drakos, C., & Dietrich, B. (2004). Development and field evaluation of
energy-based criteria for top-down cracking performance of hot-mix asphalt. Journal of the
Association of Asphalt Paving Technologies, 73, 229–260.
Watson, D., & Heitzman, M. (2014). Thin asphalt concrete overlays. NCHRP Synthesis 464,
National Cooperative Highway Research Program, Transportation Research Board of the
National Academies, Washington, DC.
West, R., Hietzman, M., Rausch, D., & Grant, J. (2011). Laboratory refinement and field validation
of a 4.75 mm superpave designed asphalt mixtures. NCAT report 11-01, Auburn, AL.
Williams, S. (2008). Surface friction measurements of fine-graded asphalt mixtures. Fayetteville,
AR: University of Arkansas.
Zhang, Z., Roque, R., & Birission, B. (2011). Evaluation of laboratory measured crack growth
rate for asphalt mixtures. Transportation Research Board of the National Academies,
Washington, DC, pp. 67–75.
Zhou, F., & Scullion, T. (2004). Overlay tester: A rapid performance related crack resistance test.
Report No. FHWA/TX-05/0-4467-2, Texas Department of Transportation, Austin, TX.
Towards Understanding Tyre-Pavement
Contact in APT Research on Flexible
Pavements

Morris De Beer and Colin Fisher

Abstract An important element of Accelerated Pavement Testing (APT) is the

selection of appropriate test tyre(s) to be used for APT testing. By far the majority
of full scale (or scaled) APT devices uses pneumatic rubber tyres. Tyres differ not
only in size, but also in composition stiffness, tyre tread patters, etc. Together with
selected tyre loading scenarios, tyre inflation pressure plays a very important role
and which is often neglected during planning the APT research. The thinner the
flexible surfacing layers or the higher the pavement temperature within the asphalt
surfacing and base layers, the more important the tyre-pavement interaction
becomes. This is so because of the magnitude and non-uniformity of the
tyre-pavement contact within the tyre contact patch, amongst others. The general
(traditional) assumption is that the tyre-pavement contact is uniformly distributed
within the tyre contact patch. A multitude of studies indicated that it is not only the
tyre tread pattern and surface friction that influences these tyre contact stress pat-
terns, but mainly the tyre loading/inflation pressure combination. In order to address
these issues, the concept of “tyre fingerprinting” is proposed for APT users. Each
tyre load/inflation pressure combination produces a different contact pattern inside
the contact patch. For example, tyre loadings combined with high inflation pressure
results in peak stresses (higher than inflation pressure) to develop in the centre
portion of the contact patch, where relatively low inflation pressures combined with
relatively high tyre loadings as often used to accelerate testing during APT research,
may result in tyre edge loading/stresses, also much higher than the tyre inflation
pressures. The aim of this paper is to address and clarify above with respect to
selecting the appropriate tyre loading/inflation pressure levels for flexible pavement
research using APT devices.

Keywords Tyre

Flexible

Pavement Loading


Fingerprinting



Tyre-pavement contact stress (TPCS) Accelerated pavement testing (APT)

M. De Beer (&)
C. Fisher

CSIR Built Environment, PO Box 395, Brummeria, Pretoria 0001, South Africa
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 491

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_32
492 M. De Beer and C. Fisher

1 Introduction

The characteristics of pneumatic test tyres used by full-scale and scaled pavement
accelerated pavement testing devices are for the majority of testing ignored, as the
tyre is seen just as the load/stress transmissioning device between Accelerated
Pavement Testing (APT) device and the test surface. Tyres differ not only in size,
but also in rubber composition stiffness, tyre tread patterns, tyre inflation pressure
(TiP) associated with tyre load (L) magnitudes. An important element of APT is the
selection of appropriate test tyre(s) to be used for pavement testing. By far the
majority of full scale (or scaled) APT devices uses pneumatic rubber tyres. Together
with selected tyre loading scenarios, which most APT users ranked very important
(Jones 2012), TiP plays a very important role and which is often neglected during
planning the APT research. The thinner the flexible surfacing layers or the higher
the pavement temperature within the asphalt surfacing and/or base layers, the more
important the tyre-pavement interaction becomes. For a given TiP and tyre loading,
the non-uniform characteristic of the tyre-pavement contact within the tyre contact
patch, is defined and quantified in three dimensions (3D), i.e. Vertical (Z), Lateral
(Y) and Longitudinal (X) in this paper. The general (traditional) assumption is that
the tyre-pavement contact is uniformly distributed within the tyre contact patch.
A multitude of studies indicated that it is not only the tyre tread pattern and surface
friction that influences these tyre contact stress patterns, but mainly the tyre
loading/inflation pressure combination. In order to address these issues, the concept
of “tyre fingerprinting” is proposed for APT users. Some background reading can
be found in De Beer et al. (1997); De Beer and Maina (2011); De Beer and Sallie
(2012); De Beer and Van Rensburg (2015); Maina et al. (2013). Each tyre
load/inflation pressure combination produces a different contact pattern inside the
contact patch. For example, vertical free rolling tyre loadings combined with high
TiPs result in peak contact stresses (often higher than TiP) to develop in the centre
portion of the contact patch, where relatively low TiPs combined with relatively
high tyre loadings as often used to accelerate testing during APT research, may
result in tyre edge loading/stresses, also much higher than the TiP. The aim of this
paper is to address and clarify the foregoing with respect to selecting the appro-
priate tyre loading/inflation pressure levels combinations for flexible pavement
research using APT devices.

2 Background to Pneumatic Rubber Tyres

A nominal description of the main components of a pneumatic rubber tyre is given

in the cut section in Fig. 1, and is self-explanatory. For the purpose of this paper, it
is important to note that Tread Width is smaller than the Overall as well as the
Section Width. The Tread Width defines the cross section width of the tyre contact
patch on the surface of the road (or APT test section).
Towards Understanding Tyre-Pavement Contact in APT Research … 493

Fig. 1 Section through a pneumatic treaded modern rubber tyre (from: za.vc.8059.2001_004_01.
jpg, originally from: https://law.resource.org/pub/za/ibr/za.vc.8059.2001.html)

2.1 Tyre Sidewall Height Profiles

The evolution of tyre dimension changed over time. This is caused by both his-
torical circumstances and practical reasons in the past, if the sidewall’s height was
close to its width, its height didn’t have to be given (http://www.oponeo.co.uk/tyre-
article/tyre-sizes). As the tyre market was developing, the contact patch increased,
which forced the manufacturers to also include the cross-section (profile percent-
age) of the tyre, e.g. “315/80”, where the aspect ratio is therefore 80 %.

2.2 Effects of Tyre Inflation Pressure (TiP)

The effects of over- and under inflation relative to the correct tyre inflation pressure,
w.r.t road contact are schematically illustrated in Fig. 2. In essence, over inflation
(combined with relatively low tyre load) will result in road contact (and tyre wear)
at the centre of the tread width. On the other hand under inflation (combined with
over loading for that pressure) will cause the opposite, in-which the maximum road
contact will be at the two tyre edges. These concepts are critical for understanding
during APT research programs, as the observed phenomena (rutting, cracking) will
494 M. De Beer and C. Fisher

Fig. 2 Differences in road contact area as a result of different inflation pressure conditions—note
the darker contact areas within the tyre contact patch (from: http://www.ewinda.com/img/
information/tyre-pressure-image.jpg. Last visited on Oct 6 2015)

Fig. 3 Typical channelized

rutting (close up) and
Wandering (top right) on an
asphalt surfacing/base at bus
and lane stops (modified from
WS DOT, 2002)

be largely dictated by the tyre loading and tyre inflation pressure characteristics, and
therefore the tyre-road contact—which are the main loading/stress inputs onto the
APT test sections from the test tyre. This is in addition of the so-called “Wandering
Mode” and “Channelized Mode” of APT loading application. See Fig. 3 for a real
world example of rutting, both in channelised as well as wandering pattern. In the
next section the concept of Tyre Pavement Contact Stress (TPCS) is discussed.
Towards Understanding Tyre-Pavement Contact in APT Research … 495

3 Characterizing Tyres for Use in APT

3.1 Tyre-Pavement Contact Stress (TPCS)

with Stress-in-Motion (SIM) Device

In the current era of fast paced research and development towards implementation
of permanent infrastructure such as roads, more and more is required to know the
operational environment. A full understanding of the correct engineering input
parameters during the design stage is therefore required. One of the main inputs
towards design optimisation is the tyre-road interaction forces, highlighted by many
APT researchers as significant (Steyn 2012). The challenge here is the quantifica-
tion of these forces under rolling tyres up to the 3rd dimension, i.e. in 3D. For this
purpose the results of the Stress-in-Motion (SIM) system is used in this paper. Tyre
3D tyre-pavement contact stresses (TPCSs) were measured with the SIM device
described by De Beer and Fisher (2013). See the associated images in Figs. 4, 5, 6,
7, 8 and 9, which are rather self-explanatory.

3.2 Examples of TPCS on Rutting (Z-Only)

Research with the SIM technology with the Heavy Vehicle Simulator (HVS) over
the years, indicated amongst others, that tyre-road interaction is not that simple, but
constitutes a multitude of parameters such as: the very important relationship(s)
between tyre vertical tyre loading, and tyre inflation pressure. It was found that each
tyre presents a unique “fingerprint” for the vertical, lateral and horisontal forces,
which can easily be converted to contact stress units. It was shown that different
combinations of tyre load (L) and tyre inflation pressure (TiP) of the same tyre

Fig. 4 Heavy vehicle

simulator (HVS) MK VI+
(from Gautrans roads)
496 M. De Beer and C. Fisher

Fig. 5 HVS dual test wheel

configuration approaching the
stress-in-motion (SIM) device
(next to yellow beam)

Fig. 6 Close-up of HVS dual

wheel confiduration on SIM
device

Fig. 7 Overloaded and

underinflated HVS dual tyres
on the SIM device
Towards Understanding Tyre-Pavement Contact in APT Research … 497

Fig. 8 Steering tyre—

overloaded and underinflated.
Insert Ink printed contact
patch

Fig. 9 Effect of temperature

—HVS channelized testing—
HMA @ 60 °C on road R80
in Pretoria, South Africa

could potentially influence the rutting patterns on asphaltic surfaced road surfaces
and is described in the next section.

3.3 Tyre Fingerprinting: of 12R22.5 HVS Tyres in 3D

(i.e. Z, Y and X Directions)

Test matrices of the tyre fingerprinting results for HVS test tyre 12R22.5 are
illustrated in Figs. 10, 11, and 12. Figure 10 shows the captured Vertical Contact
Stress (VCS) by the SIM system, for a range of tyre loading from 15 kN to 50 kN
(on y-axis), and on the x-axis the range of TiP from 520 to 800 kPa. It is clear that
with an increase in tyre loading, the shape of the VCS changes accordingly,
showing the transition from what is referred to as a “n-Shape” pattern to a
“m-Shape” pattern, across the width of the actual measured lateral width of the tyre
498 M. De Beer and C. Fisher

Fig. 10 12R22.5 Tyre “fingerprint” for vertical (Z) contact stress distribution at different loads
and tyre inflation pressure. [Tyre SA06-TYRE 18 caravan side (CS)]

Fig. 11 12R22.5 Tyre “fingerprint” for Lateral (±Y) contact stress distribution at different loads
and tyre inflation pressure. [Tyre SA06-TYRE 18 caravan side (CS)]

contact patch. Further, with an increase in the TiP, there is a general increase in the
VCS near the centre portion of each of the data plots, showing the typical “balloon”
effect upon increasing inflation pressure. The traces alongside the tyre contact patch
shows the typical parabolic shapes of the measured VCS for each of the active
measuring pins from the SIM system. For this tyre the “ideal shape” is a relatively
flat shape of the VCS across the tyre width at a TiP of approximately 740 kPa, and
Towards Understanding Tyre-Pavement Contact in APT Research … 499

Fig. 12 12R22.5 Tyre “fingerprint” for Longitudinal (±X) contact stress distribution at different
loads and tyre inflation pressure. [Tyre SA06-TYRE 18 caravan side (CS)]

maximum tyre loading of 27.5 kN. This represents the approximate conditions for
maximum tyre life at maximum tyre loading. It is therefore postulated that in this
way, the life of the APT test tyres will be maximized. Any tyre loading exceeding
27.5 kN, will result in VCS concentrations at one or both the tyre edges, i.e.
“m-Shape” conditions, which is considered extreme for both tyre and APT test
sections. These concentrations may directly affect the way of say the lateral asphalt
deformations observable from the APT test sections, in both the channelized and
wandering mode of APT traffic, as well as for on normal traffic and pavements. See
4.2 later. The fingerprint measurements also included the other Contact Stresses
(CSs) in the lateral (±Y) and longitudinal (±X). See Figs. 11 and 12, respectively.
Understanding these effects, improved decision making could potentially follow for
APT test tyre maintenance as well as a better understanding of APT test results, in
particular the rutting (or permanent deformation) behaviour of “plastic” material
such as hot mix asphalt (HMA) at elevated temperatures.

4 Normalized Contact Pressures (NCP)

For a rather simplistic representation of the VCS results discussed in Sect. 3.3, the
concept of normalized contact stress (or normalized contact pressure) across the
tyre patch is used. The Normalized tyre Contact Pressure (NCPz) is defined as the
dimensionless ratio between the maximum vertical tyre contact stress (Z) and the
TiP. See Eq. 1 below:
500 M. De Beer and C. Fisher

Fig. 13 Normalized contact pressure (NCPz) for “n-Shape” (concave) and “m-Shape” (convex)
results

VCS
NCP ¼ ð1Þ
TiP

A summary schematic of the NCP results shown in Fig. 10 is given in Fig. 13. The
NCPs for the left hand side (LHS) and right hand side (RHS) of both the “n-Shape
(concave) and “m-Shape” (convex) across the tyre width are illustrated in Fig. 13.
Note that although the concept of NCP is used here, it does not effectively dis-
criminate between the different shapes, i.e. “n-Shape” and “m-Shape” profiles.
However, the relatively higher NCPs (by *55 % higher than TiP) are often
associated with the tyre edges of the typical “m-Shape” (convex) VCS patterns (i.e.
the Overloaded/Underinflated condition). The NCPs at tyre center is almost equal
for the two cases in this example, at *33 % higher than TiP. The NCPz is the factor
that the TiP can be multiplied with in order to get an idea of the actual vertical tyre
contact stress on the pavement. Therefore, the APT user might be in a better
position to clarify especially rutting patterns from APT devices, relative to labo-
ratory test and its associated stress regime(s) during the testing. It is postulated that
this approach might inform towards the minimization of the gap between APT
results and laboratory results, especially for rutting behaviour. It may also inform on
cracking behaviour, but needs further investigation.

4.1 Typical “ESAL” Footprint

As an example, for the HVS test tyre 12R22.5, finger printing was done at a total
load of 40 kN on dual tyres (see Fig. 6), at a TiP of 800 kPa for both tyres. This half
single axle with dual wheels on the HVS represents a standard truck axle load for
Towards Understanding Tyre-Pavement Contact in APT Research … 501

Fig. 14 Vertical (Z) contact stresses of dual tyre configuration on HVS at total tyre loading of
40 kN and tyre inflation pressure = 800 kPa (*ESAL). Maximum contact stress: 1.09 MPa

Fig. 15 Lateral (±Y) contact stresses of dual tyre configuration on HVS at total tyre loading of
40 kN and tyre inflation pressure = 800 kPa (*ESAL). Contact stress range: −0.164 kPa to
+0.172 kPa

dual tyres, and any other tyre (or axle load) is referenced back to this standard axle,
using the well-known 4th power law. The result of the conversion is known as the
“Equivalent Single Axle Load (i.e. ESAL). The Vertical (Z), Lateral (±Y) and
Longitudinal (±X) contact stress patterns are illustrated in Figs. 14, 15 and 16,
respectively. Note this is the result of a single measurement using the SIM system
(De Beer and Fisher 2013). Although the TiP of this tyre was 800 kPa, the max-
imum vertical stress (Z) was 1.09 MPa, hence the Normalized Contact Pressure
ratio is 1009/800 = 1.26 (i.e. 26 % higher than TiP). See also Fig. 13. Note: All
tests are with a “free-rolling” loaded tyre at creep speed (*5 k/h). Further, the
Lateral (±Y) contact stress range was from −0.164 to +0.172 kPa, as shown in
502 M. De Beer and C. Fisher

Fig. 16 Longitudinal (±X) contact stresses of dual tyre configuration on HVS at total tyre loading
of 40 kN and tyre inflation pressure = 800 kPa (*ESAL). Contact stress range: −0.068 to
+0.096 kPa

Fig. 15, whilst the Longitudinal (±X) contact stress range was from −0.068 to
+0.096 kPa, see Fig. 16.

4.2 Examples of Rutting and Temperature and Contact

Stress Shapes

Examples of rutting (and temperature effects) on a 40 mm HMA overlay, using an

“n-Shape” contact stress (see inserts) are illustrated in Figs. 17 and 18. The tyre
type was a 11R22.5 tyre (16 ply, 125 psi, HSR, Regional Traffic), with a tread
pattern incorporating 3 longitudinal grooves in center portion of tyre patch. with
tyre traffic loading at 40 kN, with a TiP of 620 kPa. A very important message from
Fig. 18 is the influential effect of temperature gradient across the width of the HVS
test section, which accidently occurred. The planned test average temperature was
60 °C, which was achieved on the Caravan Side (CS), but only 55 °C was achieved
on the Traffic Side (TS) of this particular HVS test section. Note also the relatively
low rutting levels found at a temperature of 40 °C. Figure 19 illustrates an example
of the effect of an “overloaded tyre”, i.e. “m-Shape” contact stress condition. The
“m-Shape” applied stress condition is clearly shown by the cross section profile of
the same shape. In this case the tyre was also of type 11R122.5, but with only two
grooves near the center portion. The footprint and rutting effects of this tyre is
clearly distinguishable by studying Figs. 20, 21, 22, 23 and 24, which are all
self-explanatory.
Towards Understanding Tyre-Pavement Contact in APT Research … 503

Fig. 17 Effect of temperarure differential between the two lanes of channelized uni-directional
trafficking with “n-Shape” contact stress pattern—HVS rutting study. Note CS caravan side, TS
traffic side (modified from: Steyn and Fisher 2008)

Fig. 18 Effect of the actual shape of the tyre contact stress on actual cross section observed during
HVS rutting tests. Note typical “m-Shape” tyre edge contact stress for test section 448A4, opposed
to sections 446A4 and 447A4, where “n-Shape” were used
504 M. De Beer and C. Fisher

Fig. 19 SIM measured

vertical contact stress: tyre at
relatively high tyre inflation
pressure (overinflation)
showing a typical “n-shape”
contact stress pattern
(concave)

Fig. 20 Channelised traffic

—tyre at relatively high tyre
inflation pressure causing
maximum rutting in the centre

Fig. 21 Example rutting

profile (from Fig. 20)
Towards Understanding Tyre-Pavement Contact in APT Research … 505

Fig. 22 SIM measured

vertical contact stress: tyre at
relatively low tyre inflation
pressure (underinflation)
showing a typical “m-shape”
contact stress pattern (convex)

Fig. 23 Channelised traffic

—tyre at relatively low tyre
inflation pressure causing
maximum rutting in both
outside edges of the tyre
width

Fig. 24 Example rutting

profile (from Fig. 23)
506 M. De Beer and C. Fisher

5 Aspects of Tyre-Pavement Modeling

The next challenge to overcome is to use the discussed non-uniform tyre-pavement

contact stresses towards modeling some way, or another. Some examples demon-
strating the foregoing using closed form solutions based on linear elasticity (LE) (as
opposed to approximate finite element (FE) solutions), are given by De Beer and
Maina (2011), De Beer and Sallie (2012), Maina et al. (2013), De Beer and Fisher
(2013) and De Beer and van Rensburg (2015). It is, however, accepted that much
more modeling work is needed in the visco-elastic as well as the visco-plastic
domains, especially dealing with the thicker (>100 mm) HMA layers, not common
in southern Africa.

6 Summary, Conclusions and Recommendations

In this paper an attempt was made towards the understanding tyre-pavement contact
in APT research on flexible pavements. Some basic tyre information as well as
examples of the concept of tyre fingerprinting is discussed. The aim of the paper
was to inform towards selecting appropriate tyre loading/inflation pressure level
combinations for flexible pavement research using APT devices. Amongst others,
the data in this paper indicates that the maximum vertical tyre-pavement contact
stress for the tyre investigated is much higher than the tyre inflation pressure, by a
factor ranging from 34 to 76 %, with associated effects in both the lateral and
longitudinal directions within the contact patch. It is concluded that enough
information is currently available for optimizing APT research in general, including
upscaling the tyre-pavement model techniques used by practitioners.
It is therefore recommended that APT practitioners take note of the tyre related
aspects discussed in this paper (amongst others), in order to address the important
issue of correctly quantifying the effect of pneumatic tyres on flexible road pave-
ments more precisely (i.e. tyre “fingerprinting”). Secondly, to understand the effects
of increasing the test tyre loading and associated tyre inflation pressure for
increased APT productivity, especially in the light of increased pavement tem-
peratures and associated rutting (or plastic flow) of HMA during APT testing.

References

De Beer, M., & Fisher, C. (2013). Stress-in-motion (SIM) system for capturing tri-axial tyre-road
interaction in the contact patch, measurement (Vol. 46, Issue 7, August 2013, pp. 2155–2173.
ISSN0263-2241). doi:10.1016/j.measurement.2013.03.012. http://www.sciencedirect.com/
science/article/pii/S0263224113000791.
De Beer, M., Fisher, C., & Jooste, F. J. (1997). Determination of pneumatic tyre/pavement
interface contact stresses under moving loads and some effects on pavements with thin asphalt
Towards Understanding Tyre-Pavement Contact in APT Research … 507

surfacing layers. In Eight (8th) international conference on asphalt pavements (8th ICAP ‘97)
(Vol. 1, pp 179–227. August 10–14, 1997), Seattle: Washington. ISBN 8790145356.
DP-97/011.
De Beer, M., & Maina, J. W. (2011). Using tire-road contact stresses in road pavement design and
analysis. In: Tire technology international annual review, UK. http://viewer.zmags.com/
publication/54f87b66#/54f87b66/76.
De Beer, M., & Sallie, I. (2012). An appraisal of mass differences between individual tyres, axles
and axle groups of a selection of heavy vehicles in South Africa. International conference on
weigh-in-motion: ICWIM6 (organized with NATMEC 2012), Dallas, Texas June 4–7, 2012.
In B. Jacob, A.-M. McDonnell, & F. Schmidt (Eds.), Proceedings of the 6th international
conference on weigh-in-motion (ICWIM 6), Wiley Cunagin, ISTE (UK), Wiley (USA). ISBN:
978-1-84821-415-6.
De Beer, M., & van Rensburg, Y. (2015). Pavement damaging effects from dual tyre
configurations of heavy vehicles with tyre inflation pressure differentials. In 11th
Conference on asphalt pavements for Southern Africa, 2015 (CAPSA 2015), focus area:
FA1—Innovative structural design procedures. Paper 10, 16 to 19 August 2015, Sun City:
South Africa.
Jones, D. (2012). In I. Al-Qadi (Ed.), Advances in pavement design through full-scale accelerated
pavement testing. CRC Press, 2012 Print. ISBN: 978-0-415-62138-0. eBook ISBN:
978-0-203-07301-8.
Maina, J.W., De Beer, M., & Van Rensburg, Y. (2013). Modelling tyre-road contact stresses in
pavement design and analysis. In 32nd Annual Southern African transport conference 2013
(SATC 2013), 8–11 July 2013 “Transport and Sustainable Infrastructure”. CSIR International
Convention Centre, Pretoria, South Africa.
Steyn W. J. V. D. M. (2012). A decade of full-scale accelerated pavement testing. Advances in
pavement design through full-scale accelerated pavement testing. Sep 2012, pp. 13–22.
Steyn, W. J. V. D. M, & Fisher C. (2008). Technical memorandum: Phase 1 of HVS testing: Road
P159/1–441A4, 442A4, 443A4, 444A4, 445A4, 445A4A, 446A4, 447A4, 448A4, 449A4 and
450A4. Unpublished report for the Gauteng Department of Public Transport, Roads and Works
(GDPTRW), Transport Infrastructure Engineering CSIR, March 2008.
Use of APT for Validating the Efficiency
of Reinforcement Grids in Asphalt
Pavements

Martin Arraigada, Federico Perrotta, Christiane Raab,

Gabriele Tebaldi and Manfred N. Partl

Abstract Reinforcement of flexible asphalt pavements with interlayer systems

containing grids is increasingly used to extend service life of roads. Typically,
reinforcements are placed between layers. They can be utilized to prevent the
propagation of reflective cracking, prolonging the service life of asphalt pavements.
Many research and practical experiences have shown the advantages of using
reinforcements to increase the life-span of a pavement. However, only few tests
were done using Accelerated Paving Testing (APT), where the application of loads
is carried out with controlled passing of truck tires. Further, there is still insufficient
research carried out to predict the actual effect on the extension of the pavement’s
life-span. This information is critical for pavement design purposes and to improve
the design standards. This research focuses on the effect of different reinforcement
systems on the life-span extension of two-layered slabs and an asphalt pavement,
using scaled and full scale APT with the Model Mobile Load Simulator MMLS3
and the Mobile Load Simulator MLS10 respectively. Laboratory and full-scale
results with MMLS3 and MLS10 show that the different grids used in the test will
extend the life-span of the tested pavements up to two times. The full-scale tests
show, that the effect of the construction of a reinforcement system combining a grid
with a stress absorbing membrane (SAMI) with the grid, increases the rutting of the
pavement.

Keywords MLS10
MMLS3
Reinforcement grids
Reflective cracking

M. Arraigada (&)
C. Raab
M.N. Partl

Road Engineering/Sealing Components Laboratory, EMPA, Ueberlandstrasse 129,
CH8600 Duebendorf, Switzerland
e-mail: [email protected]
F. Perrotta
G. Tebaldi
Dipartimento di Ingegneria Civile, dell’Ambiente, del Territorio e Architettura,
Università degli Studi di Parma, Parco Area Delle Scienze 181/a, Parma, Italy

© Springer International Publishing Switzerland 2016 509

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_33
510 M. Arraigada et al.

1 Introduction

The reinforcement of pavements using different interlayer systems is a method used

mainly to delay the initiation and propagation of reflective cracking in asphalt
overlays. Further, different reinforcement systems are offered as commercial
products claiming to have the following benefits:
• Preventing reflective cracking
• Improving the distribution of loads over a cracked layer
• Providing shear resistance against rutting
• Enhancing the fatigue resistance of the asphalt layers
• Providing additional bearing capacity
• Act as a barrier for water infiltration
Although the general efficiency of these systems seems to be out of discussion
(de Bondt 2012; Ling and Liu 2001), there are still many open questions about their
application. For example, there is no clear understanding about which materials,
geometries and installation procedures are the most effective. There is also little
evidence about the influence of the different types of reinforcements on the life-span
of the pavement. Moreover, there is a lack of information about the proper
dimensioning of a reinforced flexible pavement structure. Finally, the effect on this
type of system in the interlayer bonding combined with the influence of traffic
loading requires further research.
This work focuses on the results of two Accelerated Pavement Testing
(APT) experiments carried out to evaluate the efficiency of different reinforcement
systems to delay reflective cracking produced by a controlled application of
traffic-similar loading. One of the APTs was carried out in the laboratory using the
scaled load of the Model Mobile Load Simulator MMLS3 and slabs as testing
objects. The other APT was performed under real climatic conditions and on a
full-scale pavement construction using the Mobile Load Simulator MLS10.

2 Methodology

This paper presents the final part of a research project in which three different
commercial asphalt pavement reinforcement systems were studied. This part
includes the use of two accelerated pavement technologies with reduced and
full-scale loading. The aim of these experiments was to investigate the efficiency of
each of these systems to delay the propagation of cracks from the bottom layers to
the surface. The investigation included the study of a pavement without rein-
forcement. Within this paper, the three studied reinforcement systems were des-
ignated as described below:
Use of APT for Validating the Efficiency of Reinforcement Grids … 511

a) b) c)

Fig. 1 View of the reinforcement grids. Systems A, B and C in subfigures a–c respectively

• System A: polyester grid (Fig. 1a)

• System B: glass and carbon fiber grid (Fig. 1b) It is important to line up the
direction of the carbon fibers with loading direction
• System C: glass fiber grid (Fig. 1c). The system was installed in combination
with a stress absorbing membrane interlayers (SAMI)
In all cases, the tested specimens (slabs or pavements) were constructed using
the same materials, geometries and construction methods. However, the rein-
forcements were applied by the companies responsible of their commercialization.

2.1 Accelerated Pavement Testing in the Laboratory

2.1.1 The MMLS3

In the laboratory experiments, the MMLS3 (MLS Test Systems 2002) was used to
simulate the loading of passing vehicles. This apparatus is a scaled APT device
capable of applying 7200 load cycles per hour on a length of 1.2 m, using small
pneumatic wheels with a diameter of 30 cm. The load used for these tests was
2.7 kN and the speed of the wheels was 2.5 m/s. More information about the device
can be found in (Arraigada et al. 2014).

2.1.2 Experimental Setup

The setup employed in the laboratory analysis was designed to simulate the distress
mechanism that occurs in a real pavement after rehabilitation of a layer containing
cracks. More precisely, the goal was to load special slabs with the MMLS3 until
reaching the complete cracking or delamination of the specimen due to the prop-
agation of artificial cracks. To that end, two-layered rectangular specimens (slab)
were prepared using a steel roller compactor. The 3 cm thick bottom layers were
produced with AC 11 asphalt. The top layer was built with AC 8, also having 3 cm
thickness. Each slab had a length of 1.8 m and a width of 0.435 m. A total of eight
slabs were produced. Two of them were constructed using only tack coat between
the layers and were considered as reference slabs, i.e. slabs with no reinforcement.
512 M. Arraigada et al.

The remaining six slabs were reinforced with the three systems under study: for
each system two slabs were constructed (Slab 1 and Slab 2). Unfortunately, the
production of more slabs and the performance of additional tests were not possible
due to the lack of time and resources. A view of the construction process is
presented in Fig. 2.
In order to simulate the presence of cracks in the bottom layers, two transversal
cuts were planned each 2.5 cm deep as shown schematically in Fig. 3. These
artificial cracks were designed to concentrate tension strains produced by the
bending moment of the slabs under MMLS3 loading. Hence, the slabs were
positioned laying on two supports. The supports were situated close to the ends of
the slab, such that the entire assembly was simulating a simply supported beam.
Below the slab, a cellular rubber pad was placed in order to simulate the soft
foundation provided by the subgrade. The entire construction was supported by a
thick concrete plate to give stability. The system was positioned inside a container
with temperature control and the MMLS3 was placed on top of the slab.
Temperature sensors were embedded in the slabs to provide feedback to the tem-
perature control system. In order to monitor the vertical deformation of the system

a) b) c) d)

Fig. 2 a Image of the roller compactor, b, c application of a reinforcement grid, d slab on testing
position

Fig. 3 View of the MMLS3 on testing position and schema of the test set-up
Use of APT for Validating the Efficiency of Reinforcement Grids … 513

under loading, six displacement sensors (D1–D6 in the figure) were set on one of
the sides of the slab. These sensors were used to help detecting a change in the
stiffness of the slabs due to the progression of the cracks. It was considered that
under constant 20 °C temperature inside the slab and identical loading, the slab
should have the same bending deformation until the development of fatigue cracks,
delamination or any other kind of distress appears. Each slab was loaded up to
reaching the bearing capacity limit of the system, i.e. until cracks progressed
through the entire thickness of the slab. To detect that condition, visual inspections
were carried out at regular intervals of time.

2.1.3 Data Analysis and Interpretation

Three phases can be identified in the progression of the distress. In the first phase,
there is no visible crack and the slab shows typical bending deformation under
MMLS3 loading. It can however occur, that micro-cracks start to generate in the
bottom region of the slab, causing a moderate increase of the bending deformation.
Then, a crack initiates from one or both cuts. The crack(s) propagate(s) from the
bottom to the top surface of the slab, sometimes followed by a debonding between
layers. In this situation, the bending deformation starts to increase dramatically.
Finally, the crack(s) reaches the upper surface of the slab (Fig. 4). At this stage, the
broad cracking of the slab adjacent to the cuts generate a mechanical condition
similar to a hinge. Then, the bending deformation tends to reach a maximum value.
In this moment it is considered that the slab reaches the total failure.
Visual inspections of the tested slabs were important to help identify the different
stages described before. Still, detecting thin cracks visually is difficult and requires
expertise. Further, inspections can be carried out only at discrete periods of times,
during working hours. Therefore, the deformation sensors located close to the

Crack
AC 8
Reinforcement system

AC 11

Fig. 4 Outlook of a cracked slab

514 M. Arraigada et al.

middle part of the slab were used to detect the phases already mentioned and to
establish the number of MMLS3 load cycles required for the cracks, to propagate
through the interlayer. Figure 5 presents the temperature of the slabs throughout the
duration of tests. In the same figure, deflection amplitudes obtained by sensors D4
and D6 positioned on each side of the centre of the slabs are plotted against the
accumulated number of MMLS3 load cycles. Unfortunately, the results of the tests
on Slab 2 of the reinforcement System A had to be discarded because an inadequate
functioning of the MMLS3, which was detected only after finishing with the
loading phase. The deflection amplitudes were calculated based on 30 s long

Fig. 5 Deflection amplitudes of the slabs vesus number of accumulated MMLS3 load cycles for
two given displacement sensors D4 and D6. In addition, the temperature of the slab is depicted
Use of APT for Validating the Efficiency of Reinforcement Grids … 515

measurements carried out every 5 min. They stand for the average difference
between the maximum and minimum deflections for each loading cycle. The dis-
continuities in the curves correspond to interruptions in the loading phase, mostly
over the weekends and due to repair works. During these resting periods a tem-
perature reduction of the slab and a healing and a viscoelastic recovery of the
asphalt concrete were observed, as expected and reported in the literature (Lee et al.
2011; Little et al. 2001). It was noticed that the continuous loading of the slab
induces an increase of temperature due to internal friction. The temperature fluc-
tuations are responsible of fluctuation in the moment of bending of the slabs. The
bearing capacity limit of each specimen was the established from these curves
combined with visual inspections. It was observed that, in general, the deflection
amplitude curves show an exponential increase when the cracks reach the top
surface of the slab. These values are summarized in Table 1.
The variability in the number of loading cycles until failure of each pair of
similar specimens is produced due to the difficulty of constructing slabs with same
properties. Their size, the compaction procedure, their manipulation for cutting and
applying the reinforcements are just a few factors that contribute to the dispersion of
the results. In order to obtain more robust conclusions from these types of exper-
iments, future tests should consider a larger amount of slabs. Nevertheless, a trend
can be observed from these results. According to the average values, the System C
showed a poor performance compared to the reference slabs. On the other hand, the
slab with the System A slab was able to take almost twice as much load as the
reference. The best performance was obtained by the slabs with the reinforcement
System B, that had more than two times additional loading endurance than the
reference slabs.

Table 1 Summary of the MMLS3 testing results

Load cycles until crack propagation through interlayer

Reinforcement (from visual inspections and sensor’s results)
system
Single measurements Bar chart

Reference – Slab 1 500‘000 Average reference

Reference – Slab 2 240‘000
System A – Slab 1 680‘000
System A - Slab 2 Invalid measurement
System B – Slab 1 850‘000
System B – Slab 2 890‘000
System C – Slab 1 150‘000
System C – Slab 2 460‘000
516 M. Arraigada et al.

2.2 Full-Scale Accelerated Pavement Testing

2.2.1 The MLS10

The MLS10 at Empa is a prototype APT developed in 2006 in South Africa and
since 2008 active in Switzerland. This machine is a full-scaled APT device capable
of applying 6000 load cycles per hour over a length of 4.2 m, using truck
super-single or twin tires (Fig. 6). The MLS10 can apply a maximum load of
65 kN, which correspond to a 130 kN axle. The speed of the loading wheels can
reach 6 m/s. More information about the device can be found elsewhere (Arraigada
et al. 2009, 2012).

2.2.2 Experimental Setup

Full-scale tests were carried out using the MLS10. The pavements were built in a
special testing site with no vehicle traffic loading, but using standard construction
machines and real construction practices. The goal was to compare the performance
of a two-layered pavement structure containing artificial cracks and the reinforce-
ment systems described earlier as well as the same pavement without any rein-
forcement as reference.
The structure comprised two asphalt concrete layers on top of a 60 cm thick
granular base (Fig. 7). The surface course consisted of a 4 cm asphalt concrete
AC 11 S, whereas the binder course was made of 6.5 cm thick AC T 22 S.

Fig. 6 View of the MLS10 with and schema of the loading mechanism
Use of APT for Validating the Efficiency of Reinforcement Grids … 517

Reinforcement system

Scaled view
MLS10

Loading position 1 Loading position 2

5m
4.2m
Reference

9m

36m

Fig. 7 Schema and distribution of the pavements

The description of the material characteristics can be found in VSS (2008). The
total length of the paved area was 36 m and the width was 5 m. However, the area
was divided into four different sections of 9 m length. The reinforcement systems
analyzed in this study were applied in three pavement sections. The remaining
section was used as reference and only tack-coat was applied between layers. The
testing site was located in an open air space and did not allow any temperature.
Since the environmental conditions during loading have an extremely significant
effect on the performance of the materials, it was important that each section was
tested at approximately the same temperature. Therefore, and due to the fact that the
length of the MLS10 wheel path is only 4.2 m, two loading positions were planned.
Each loading position included two different pavement sections, i.e. two pavements
were loaded simultaneously. Further, artificial cracks were prepared in the binder
course by cutting the entire 6.5 cm thick layer perpendicular and in the same
direction of the loading. Similarly as in the slabs, these cuts were intended to
concentrate the tension strains produced by the bending moment of the asphalt
layers under MLS10 loading.
The Loading position 1 comprised the reference pavement and the pavement
with reinforcement system C. The Loading position 2 included the pavements
containing reinforcement systems A and B. For the tests, the MLS10 was equipped
with super-single tires, each applying 65 kN and operating without lateral wan-
dering at a speed of 18 km/h. The pavements were instrumented with thermo-
couples to measure the temperature of the materials in the testing phase. Moreover,
the structural responses were monitored through a set of strain gauges and
accelerometers with the intention of determining the influence of the different
reinforcements on the strain-deformation condition. The surface was visually
inspected at regular intervals of time, seeking for the presence of cracks. At the
same time, transversal profiles were taken to assess the rutting performance of each
pavement. The total amount of accumulated load cycles in each loading position
518 M. Arraigada et al.

was 300,000. The machine was moved from one loading position to the other every
100,000 load cycles to have, on average, the same temperature profile in both
positions. The loading phase was carried out within one and a half months in
autumn, with the purpose of having mild/cold temperature ranges, favourable for
the formation and propagation of cracks in asphalt concrete.

2.2.3 Data Analysis and Interpretation

In this paper, only a selected amount of results is presented. One of the factors that
play a fundamental role in the development of the distresses is temperature. Table 2
presents the average temperatures, organized in steps of 50,000 load cycles for both
loading positions. It can be observed, that the temperature is slightly higher for
Loading position 1. The difference, however, is below 3.4 °C. The temperature
ranges are also typical for a mild climate.
One of the most important factors to consider in the analysis of the performance
of the pavements was the capability of the different reinforcement systems to delay
the propagation of the cracks from the binder course to the surface. This was carried
out through visual inspections. In coincidence with the loading interruptions (every
50,000 load cycles), the pavement was carefully examined and any new crack
observed was defined with a white marker. The first transversal crack in the ref-
erence pavement was observed at the 150,000 load cycles interruption, whereas in
the other pavements the first crack appeared at the end of the tests, after 300,000
load cycles.
Figure 8 shows a picture of the pavement surface at the end of the loading phase.
It can be perceived, that the density of cracks is much higher for the reference
pavement and the pavement reinforced with system C, than for the other two
pavement sections. Further, the majority of the cracks are transversal to the loading
direction and coincident with the cut in the binder course, at least at an early
cracking stage. All detected cracks had a thickness of less than 1 mm.
A comparative summary of the characteristics of the cracking development of
each reinforcement system is presented in Fig. 9. From this figure it can be
observed that all systems succeeded in delaying the manifestation of cracks in more

Table 2 Average pavement temperatures

Loading cycles Temp. loading position 1: Temp. loading position 2: system
ref/system C (°C) B/system A (°C)
0–50,000 16.1 13.1
50,000–100,000 18.6 15.5
100,000–150,000 10.8 12.7
150,000–200,000 12.3 12.6
200,000–250,000 12.2 13.0
250,000–300,000 13.3 9.9
Use of APT for Validating the Efficiency of Reinforcement Grids … 519

Fig. 8 Overview of the surface cracks at the end of the tests after 300,000 load cycles

Fig. 9 Overview of the surface cracks at the end of the tests

than two times compared to the pavement without reinforcement. However, the
amount of cracks and their total length is the highest in the pavement with rein-
forcement C. From the amount and total length of the cracks, the system B was the
one that had the best performance.
The rutting performance of each pavement is summarized in Fig. 10. This figure
shows the progression of average rutting depths, and the average temperatures
obtained (see Table 2). Pavements with reinforcement systems A and B show a
better rutting performance than the other pavements. This can be partially attributed
to the temperatures during testing, which in average and in most part of the loading
phase were lower. Nevertheless, for the same loading temperature, system C shows
worse performance than the reference pavement. This effect can be due to the fact
520 M. Arraigada et al.

Fig. 10 Average rutting depths

that in this system, the glass fiber grid was installed together with a SAMI.
The SAMI might help reducing the amount of water that infiltrates to the bottom
layers but is definitely a negative influence for the permanent deformation.

3 Summary and Conclusions

In this work, two APT technologies were used to evaluate the performance of
different reinforcement systems including a first laboratory testing phase with the
reduced scale MMLS3 device and a field phase with the full-scale MLS10.
According to the test results with the MMLS3, the reinforcement System B was
able to delay the progression of the cracks in the slabs by more than two times,
compared to the slab with no reinforcements. The System A also showed a good
performance by extending the life-span of the slab by a factor below two. The
System C didn’t show any improvement compared to the reference slab. Overall it
was observed that the longest life-span was obtained by System B. It is recom-
mended however, to perform future tests with at least three or more slabs in order to
have robust average values.
Similarly, the second part of the works with the full scale MLS10 loading
showed comparable results to those obtained with the MMLS3. These indicate that
all the reinforcement systems were able to delay the appearance of cracks in the
surface of the pavement by a factor close to two. However, it was System C the one
that showed higher amount and length of cracks. At the same time, this rein-
forcement presented the highest rutting, which is attributed to the use of SAMI.
Reinforcement System B presented instead, the least amount of cracks and the
rutting was similar to the reinforcement System A.
Use of APT for Validating the Efficiency of Reinforcement Grids … 521

References

Arraigada, M., Kalogeropoulos, A., Hugenschmidt, J., Partl, M. N., Caprez, M., & Rabaiotti, C.
(2009). Pilotstudie zur evaluation einer mobilen grossversuchsanlage für beschleunigte
verkehrslastsimulation auf Strassenbelägen. Eidgenössisches Departement für Umwelt,
Verkehr, Energie und Kommunikation, Bundesamt für Strassen. ASTRA 2004/018, Report
Nr 1261.
Arraigada, M., Partl, M. N., & Pugliessi, A. (2012). Initial tests results from the MLS10 mobile
load simulator in Switzerland. In Proceedings of the 4th international conference on
accelerated pavement testing, Davis, CA, USA, Sep 19–21 2012, pp. 277–285.
Arraigada, M., Pugliessi, A., Partl, M. N., & Martínez, F. (2014). Effect of full-size and down scale
accelerated traffic loading on pavement behaviour. Materials and Structures, 47, 1409–1424.
de Bondt, A. H. (2012). 20 Years of research on asphalt reinforcement—Achievements and future
needs. In 7th RILEM conference on cracking in pavements: Mechanisms, modeling, testing,
detection and prevention case histories (vol. 4, pp. 327–335).
Lee, J., Lee, S., & Kim, Y. (2011). Evaluation of healing effect by rest periods on asphalt concrete
slab using MMLS3 and NDE techniques. KSCE Journal of Civil Engineering, 15(3), 553–560.
Ling, H., & Liu, Z. (2001). Performance of geosynthetic-reinforced asphalt pavements. Journal of
Geotechnical and Geoenvironmental Engineering, 127(2), 177–184.
Little, D. N., Lytton, R. L., Williams, A. D., & Chen, C. W. (2001). “Microdamage Healing in
Asphalt and Asphalt Concrete”, Volume 1: Microdamage and microdamage haling. Project
summary report. College Station: Texas Transportation Institution.
MLS Test Systems. (2002). MMLS3 traffic simulator operator’s manual. April 2002.
Schweizerischer Verband der Strassen- und Verkehrsfachleute VSS. (2008). “Asphaltmischgut—
Mischgutanforderungen—Teil 1: Asphalt-beton” (Asphalt mixes—requirements—Part 1
Asphalt concrete), Swiss Standard SN 640431–1 (in German).
 Part VI
Accelerated Pavement Testing on Airports
Estimation of In-Situ Shear Strength
Parameters for Subgrade Layer Using
Non-destructive Testing

Hadi Nabizadeh, Elie Y. Hajj, Raj V. Siddharthan, Sherif Elfass

and Peter E. Sebaaly

Abstract The Falling Weight Deflectometer (FWD) test has been widely used for
evaluating the structural condition and load-carrying capacity of asphalt pavement
systems as a non-destructive testing device. Conventionally, the in situ stiffness
properties of the various pavement layers are estimated from the analysis of the
surface deflection measurements using a backcalculation procedure. In this pilot
study, an innovative and novel approach was investigated for estimating the shear
strength parameters (C and ϕ) of the subgrade layer by means of FWD testing. Such
parameters become important and necessary when assessing the risk of instanta-
neous shear failure in asphalt pavement layers under non-standard heavy vehicles.
In order to assess the applicability of proposed approach, numerically simulated
FWD test as well as measured FWD field data conducted on the APT asphalt
pavement sections at the National Airport Pavement Test Facility (NAPTF) were
analyzed. The surface deflection measurements from the FWD testing at multiple
load levels were used in conjunction with backcalculation process to capture the
stress dependent behavior of unbound layers and subsequently used to determine
the shear strength parameters of the subgrade layer. It was found that the proposed
approach was capable of estimating the in situ shear strength parameters of the
subgrade material and the results were consistent with those obtained from con-
ventional laboratory testing. Based on the findings from this study, work is

H. Nabizadeh (&)
E.Y. Hajj
R.V. Siddharthan
S. Elfass
P.E. Sebaaly

University of Nevada, Reno, Reno, USA
e-mail: [email protected]
E.Y. Hajj
e-mail: [email protected]
R.V. Siddharthan
e-mail: [email protected]
S. Elfass
e-mail: [email protected]
P.E. Sebaaly
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 525

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_34
526 H. Nabizadeh et al.

currently undergoing to extend and validate the proposed approach for different
type of asphalt pavement structures and subgrade properties.

Keywords Shear strength parameters
Falling weight deflectometer
Stress

dependent behavior Asphalt pavement

1 Introduction

Non-destructive testing (NDT) techniques including lasers, Ground-Penetrating

Radar (GPR), Falling Weight Deflectometer (FWD), penetrometers, and seismic
technologies have been increasingly utilized in pavement industry for the purpose
of quality control and quality assurance (QC/QA) as well as evaluation of pavement
performance during its service life (Von Quintus et al. 2009). Among the various
NDT techniques, the Falling Weight Deflectometer has attracted substantial atten-
tion from highway agencies in the past decades. The FWD simulates pavement
responses under a moving truck wheel by applying a stationary dynamic load to the
pavement surface. The applied load in conjunction with the measured vertical
surface displacements at different radial distances from the load application (called
deflection basin) have been used to estimate the stiffness of the various pavement
layers using backcalculation techniques. Stiffness information can be used to
evaluate the structural capacity of asphalt pavements for rehabilitation and pave-
ment management purposes (Alavi and Tavares 2008; Schmalzer 2006).
Despite the fact that the FWD has been successfully employed to estimate
stiffness of pavement layers and subsequent structural capacity of the pavement
structure, the applicability of this device to determine the in situ shear strength
parameters (cohesion, C, and friction angle, ϕ) of pavement layers, especially for
unbound materials, has not been examined. Determination of shear strength
parameters for pavement layers is needed particularly when the risk of instanta-
neous shear failure in pavement layers under a non-conventional superheavy load
(SHL) move is of concern.
The need to move SHLs on US highways is a vital economic necessity for many
important industries such as chemical, oil, electrical and mining etc. Many of such
superheavy components are much larger in size and weight and they often require
specialized trailers and hauling units. The movement of such loads has become
more common over years (Chen et al. 2013). Previous studies reveal that
Mohr-Coulomb yield criterion can be employed in order to assess the possibility of
pavement damage under SHLs and the potential for edge shear failure under a SHL
moving close to the edge (Jooste and Fernando 1994; Chen et al. 1996; Fernando
1997). The 3D Mohr-Coulomb yield criterion was also implemented in a recent
study to evaluate the potential for pavement failure under SHLs moves (Chen et al.
2013). In past studies, the required shear strength parameters (C and ϕ) were
determined by means of laboratory tests or engineering experiences. The destruc-
tive and time consuming nature of laboratory measurements may sometimes lead to
Estimation of In-Situ Shear Strength Parameters … 527

an inaccurate estimation of shear strength parameters based on engineering judg-

ments. Consequently, the estimation of in situ shear strength parameters for
unbound materials in a non-destructive manner can be an invaluable technique
which can be used in pavement analysis.
In this paper a novel and innovative approach based on FWD is proposed in
order to estimate the in situ shear strength parameters (C and ϕ) of subgrade. This is
to be accomplished by focusing on FWD surface deflections at locations away from
the center of the loading plate at a number of load levels.

2 Proposed Methodology

The triaxial compression tests have been traditionally used as a laboratory method
to determine the shear strength of soil materials. A typical stress-strain relationship
in the triaxial compression test is represented by a hyperbola as presented in Eq. 1.
Subsequently, a widely accepted procedure is used to determine the deviator stress
at failure (σdf) by rewriting hyperbolic equation in a linear form (see Eq. 2). Hence,
the inverse of the slope of ε1/σd versus ε1 is equal to deviator stress at failure (σdf) as
illustrated in Fig. 1, (Duncan and Chang 1970; Byrne et al. 1987; Stark et al. 1994).

1
rd ¼
1 ð1Þ
1
þ
Ei rdf

1 1 1
¼
1 þ ð2Þ
rd rdf Ei

Fig. 1 Hyperbolic approach to calculate σdf

528 H. Nabizadeh et al.

In these equations, σd is the triaxial deviator stress, ε1 is the axial strain corre-
sponding to deviator stress, and Ei is the initial tangent modulus. In order to develop
the ε1/σd versus ε1 relationship, multiple datasets of σd and ε1 are required. In this
study, the σd and ε1 for a representative triaxial element in the subgrade layer will
be estimated using the FWD test results at multiple load levels.
It is commonly accepted that unbound materials such as crushed aggregate base
and subgrade soils exhibit nonlinear stress-dependent behaviors. This means that
the resilient modulus of an unbound layer is a function of the stress condition. Such
a characteristic of the unbound material is usually reflected in the FWD backcal-
culated moduli values when multiple load levels are applied. Figure 2 represents
typical backcalculated moduli values for two different types of subgrade materials
at different load levels.
The backcalculated layer moduli values at each load level are used in a layered
elastic program (LEP) to compute the stress tensor (σij) at any location in the
pavement, particularly at a representative element in the subgrade layer. It may be
mentioned that an element located at the depth of B/2 (B is the diameter of the
FWD plate) from the top of subgrade can be treated as the representative element to
determine the load-induced stresses. The representative element (at B/2 from top of
subgrade surface) is bounded by shearing zones in the subgrade and experiences the
largest vertical strain (Schmertmann et al. 1978). Moreover, the assumptions used
in the backcalculation procedures and the aforementioned layered elastic programs
are consistent (e.g., static and axisymmetric loading).
The induced stress tensor (σij) is then “transformed” to an equivalent laboratory
triaxial stress testing conditions by the use of stress invariants. The use of stress
invariants as the most appropriate method of comparing the stress states that affect
material behavior and characterization was suggested in previous studies (Brown
and Bell 1977; Hajj et al. 2011). Stress invariant values are the same regardless of
the orientation of the coordinate system chosen. The octahedral normal (σoct) and
shear (τoct) stresses are used to convert the stress tensor observed in the

Fig. 2 Backcalculated subgrade modulus at different load levels; a softening behavior,

b hardening behavior
Estimation of In-Situ Shear Strength Parameters … 529

representative subgrade element under the FWD loads to deviator (σd) and con-
fining (σc) stresses in a triaxial testing set-up using Eqs. 3–6. In these equations, σ1,
σ2, and σ3 are the major, intermediate, and minor principal stresses, respectively.

1
roct ¼ ðr1 þ r2 þ r3 Þ ð3Þ
3
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
jsoct j ¼ ðr1  r2 Þ2 þ ðr2  r3 Þ2 þ ðr3  r1 Þ2 ð4Þ
3
3
rd ¼ pffiffiffi jsoct j ð5Þ
2
rd
rc ¼ roct  ð6Þ
3

The calculated σd at each load level can be used to compute ε1 as expressed by

Eq. 7 where (MR)SG is the backcalculated resilient modulus at the corresponding
load level. Consequently, the inverse slope of ε1/σd versus ε1 is equal to deviator
stress at failure (σdf), as shown in Fig. 1. Hence, based on the determined σdf and
computed σc-avg which is the average of σc at each load level (Eq. 6) it is possible to
establish one single Mohr circle. Accordingly, the shear strength parameters can be
estimated using Eq. 8 by assuming an acceptable range for the friction angle when
the soil classification for the subgrade is known. It should be mentioned that there
exists solid documents in the literature for selecting an appropriate range of friction
angle based on the soil classification (Das 2014).
rd
e1 ¼ : ð7Þ
ðMR ÞSG

ðr3 þ rdf Þ  r3
tan2 ð45 þ /=2Þ

C¼ ð8Þ
2 tanð45 þ /=2Þ

3 Validation of Proposed Methodology

In order to assess the applicability and validity of the proposed approach to estimate
the shear strength parameters, numerical analysis that simulates FWD data as well
as measured FWD field data were utilized. The results of these analyses are pre-
sented in the following sections.
530 H. Nabizadeh et al.

3.1 Synthetic FWD Data

Synthetic pavement responses under FWD loading were generated using

ILLI-PAVE 2005 program (Thompson and Elliott 1985). Substantially lower
computational effort because of the use of axi-symmetric finite element formulation
and the consideration of stress dependency (non-linearity) along with failure con-
ditions of unbound materials (C and ϕ) are the main advantages of this software. It
may be noticed that ILLI-PAVE 2005 was also employed in previous studies to
obtain a FWD database (Ceylan et al. 2005a, b; Pekcan et al. 2008; Tutumluer and
Sarker 2015).
FWD tests on different flexible pavements were simulated by applying various
uniform loads (9–27 kips) on a circular plate with 5.91 in. radius. The corre-
sponding pavement surface displacements at 0 in. (D0), 8 in. (D8), 12 in. (D12), 18
in. (D18), 24 in. (D24), 36 in. (D36), 48 in. (D48), 60 in. (D60) and 72 in. (D72)
away from the center of the FWD plate were computed.
Pavement layers properties play a crucial role in the realistic simulation of the
FWD loading. Accordingly, appropriate material properties for the asphalt concrete
layer (AC), crushed aggregate base (CAB), and subgrade (SG) to characterize their
behavior is an important aspect. In the current analysis, asphalt concrete layer was
represented with constant linear elastic properties (EAC and ν). Stress-dependent
constitutive resilient modulus models along with shear strength parameters (C and
ϕ) were specified for CAB and SG. The Theta (Eq. 9) and Uzan (Eq. 10) models
were selected to model the stress-dependent behavior of CAB and SG materials,
respectively:

MR ¼ KðhÞn ð9Þ

MR ¼ K1 ðhÞK2 ðrd ÞK3 ð10Þ

where, MR is the resilient modulus; θ is the bulk stress; σd is the deviator stress; K,
n and ki are regression constants. Table 1 shows the material properties as well as
the regression constants used for the various pavement layers. The model constants
for unbound materials were adopted from the literatures (Garg and Thomson 1998;
Santha 1994). Three different pavement sections with different layer thicknesses
were simulated in ILLI-PAVE.
Figure 3 depicts the deflection basins under different FWD load levels for the
pavement structure II. Simulated FWD data obtained with ILLI-PAVE runs were
employed in the backcalculation analysis using BAKFAA (Federal Aviation
Administration 2015b) which is a widely used open source software utilized for
handling the backcalculation process. Since linear elastic behavior was assumed for
AC layer, a constant elastic modulus was specified for this layer at all load levels
during the backcalculation process. Figure 4 shows the backcalculated moduli
values for subgrade and base layers. This figure reveals that an increase in FWD
load level resulted in a reduction in backcalculated subgrade modulus, indicating a
Estimation of In-Situ Shear Strength Parameters … 531

Table 1 Material properties of pavement layers used in ILLI-PAVE

Properties AC Crushed aggregate base Subgrade (fat clay)
Density (pcf) 145 120 100
K0 0.67 0.36 0.80
Resilient modulus (psi) 300,000 Theta model Mr ¼ K ðhÞn Uzan model Mr ¼ K1 ðhÞK2 ðrD ÞK3
K = 5358 K1 = 1793
n = 0.32 K2 = 0.19
K3 = −0.36
Friction angle, ϕ – 39 12
Cohesion, C (psi) – 1 19
Poisson’s ratio 0.35 0.35 0.45
Layer thickness (in.)
Pavement Structure I 6.0 8.0 Semi-infinite
Pavement Structure II 9.0 10.0 Semi-infinite
Pavement Structure III 12.0 12.0 Semi-infinite

Fig. 3 Deflection basin under different load levels for pavement structure II

softening material behavior. On the other hand, an increase in the FWD load level
resulted in an increase in the base resilient modulus, revealing a hardening behavior
of the material (Yau and Von Quintus 2002).
Subsequently, based on the backcalculated moduli values for the base and
subgrade layers at each of the load levels, the stress tensor (σij) at 5.91 in. (i.e., B/2
where B is the plate diameter) below the subgrade surface was computed by
assuming static loading conditions using the 3D-Move Analysis software
(Siddharthan et al. 1998). Finally, the stress condition was converted to stress
invariants and corresponding triaxial testing conditions as outlined before.
532 H. Nabizadeh et al.

Fig. 4 Backcalculated modulus of crushed aggregate base (a) and subgrade (b)

As mentioned earlier, the inverse slope of ε1/σd versus ε1 is equal to deviator

stress at failure (σdf). As shown in Fig. 5, the inverse slope of ε1/σd versus ε1 for the
pavement structure II is equal to 50.96 psi. Accordingly, the shear strength
parameters can be estimated if the soil classification for the subgrade is known (see
Eq. 8). For instance, by assuming an acceptable range for the friction angle, the
corresponding cohesion can be calculated. In the current analysis, the subgrade
cohesion, CSG, is estimated to be 20.5 psi when the subgrade angle of internal
friction, ϕSG, was assumed to be 12°. It should be mentioned that a similar analysis
for pavement structures I and III resulted in an estimate for the CSG equal to 21.3

Fig. 5 Hyperbolic approach to estimate σdf for pavement structure II

Estimation of In-Situ Shear Strength Parameters … 533

and 18.1 psi, respectively. These set of estimated values when compared with the
input values for the subgrade material (CSG = 19 psi and ϕSG = 12°) implies that
the current approach is capable of estimating the shear strength parameters of the
subgrade layer from the FWD simulated measurements when a good estimate for
the angle of internal friction is made.

3.2 Real Field FWD Data

A series of tests known as Falling/Heavy Weight Deflectometer (FWD/HWD)

Round-Up tests were conducted at the National Airport Pavement Test Facility
(NAPTF) at the William J. Hughes Technical Center in Atlantic City, New Jersey
(Federal Aviation Administration 2015a). A series of FWD/HWD tests were con-
ducted at nine different stations on a flexible and rigid pavement sections when the
FWD was applied at three load levels of 9, 12, and 16 kips and the HWD target
loads were 12, 24, and 36 kips. The pavement sections and drop locations are
illustrated in Fig. 6. In order to evaluate the validity of the proposed approach, the
analysis focused on the drop location 17 on the flexible pavement (see Fig. 6).
Figure 7 depicts the measured deflection basins under the different FWD load
levels. Figure 8 presents the backcalculated moduli values for the subgrade and
base layers from the FWD data. Softening behavior in the subgrade layer and
hardening behavior in the base layer were observed. It should be noted that, since
AC modulus is expected to be constant and independent of the FWD load level, a
constant AC modulus was determined based on the average of the backcalculated
AC moduli values at different load levels from the first backcalculation process.
Subsequently, this constant AC modulus value was applied during the second
backcalculation process for all load levels to correctly capture the stress-dependent
behavior of the unbound layers.
Based on the backcalculated moduli values for base and subgrade layers, the
stress tensor (σij) was computed at 5.91 in. below the subgrade level using 3D-Move
static analysis for each of the FWD load levels. Subsequently, using the procedure
outlined above the stress tensor was used to compute the octahedral stress compo-
nents (invariants) and the equivalent stress conditions associated with the corre-
sponding triaxial tests. The next step is to fit a hyperbolic relationship to obtain the
deviator stress at failure (σdf) as shown in Fig. 9 which is equal to 23.7 psi.
As mentioned earlier, the shear strength parameters can be estimated if the soil
classification is known. In the FAA study, the subgrade material was classified as
Dupont Clay (see Fig. 6). Hence, assuming ϕSG is equal to 10°, corresponding
estimated CSG will be equal to 9.9 psi. Tutumluer et al. (2004) conducted a study to
determine the stress-strain relationship of Dupont clay. Figure 10 illustrates the
axial stress versus the axial strain for Dupont clay at different compaction levels in a
534 H. Nabizadeh et al.

Evaluated Drop Location

Fig. 6 Pavement sections and drop locations in FAA study

series of unconfined compression tests (Tutumluer et al. 2004). In the FAA study,
the average CBR of the subgrade was around 6 and therefore the corresponding
un-drained shear strength from Fig. 10 is equal to 12.5 psi. An estimate of the shear
Estimation of In-Situ Shear Strength Parameters … 535

Fig. 7 Deflection basin under different load levels in FAA study

Fig. 8 Backcalculated modulus of crushed aggregate base (a) and subgrade (b) in FAA study

strength of subgrade in saturated condition (Cu)SG can be made for this clay using
Eq. 8 by assuming ϕSG = 0°. This results in an estimate of 11.8 psi for CSG, which
is consistent with the reported measured data of 12.5 psi.
536 H. Nabizadeh et al.

Fig. 9 Hyperbolic approach to estimate σdf in FAA study

Fig. 10 Stress-strain curve for Dupont clay from Tutumluer et al. (2004)
Estimation of In-Situ Shear Strength Parameters … 537

4 Conclusion

The Falling Weight Deflectometer (FWD) test as a common non-destructive

pavement testing device has been widely employed to measure the deflection basin
under applied dynamic load and subsequently determine the in situ stiffness of
pavement layers using backcalculation techniques. Although stiffness parameters
can be applied to evaluate structural adequacy and load-carrying capacity of asphalt
pavement systems, shear strength parameters are needed when the risk of instan-
taneous shear failure in asphalt pavement layers under non-standard heavy vehicles
is particularly assessed. In this pilot study, an innovative and novel approach was
proposed and investigated to estimate the shear strength parameters (C and ϕ) of the
subgrade layer in a non-destructive manner by means of FWD testing.
This approach consisted of the following steps: (1) Backcalculate pavement
layer moduli values under all FWD load levels; (2) Obtain the stress tensor (σij) at a
representative location in the subgrade using 3D-Move static analysis for each of
the FWD load levels using backcalculated modulus values; (3) Use the determined
stress tensor to compute the octahedral stress components, which are invariants and
the equivalent stress conditions associated with triaxial tests (σd and σc) for each of
the FWD load levels; (4) Evaluate the corresponding triaxial axial strain is the
backcalculated subgrade modulus; (5) Fit a hyperbolic relationship to the triaxial
responses (σd and ε1) to obtain the ultimate deviator stress, σdf; and (6) Use Eq. 8 to
estimate the shear strength parameters of subgrade layer.
In order to assess the applicability of the proposed approach, numerical analyses
to simulate FWD data as well as measured FWD field data conducted on the APT
asphalt pavement sections at the National Airport Pavement Test Facility (NAPTF)
were independently analyzed. In summary, it was found that the proposed approach
was capable of estimating the in situ shear strength parameters of the subgrade
material. Accordingly, large-scale experiments are currently undergoing at the
University of Nevada, Reno to extend and validate the proposed approach for
different type of asphalt pavement structures and subgrade properties under the
controlled environment.

References

Alavi, S., & Tavares, M. P. (2008). NCHRP synthesis 381: falling weight deflectometer usage.
Washington, DC: Transportation Research Board, National Research Council.
Brown, S. F., & Bell, C. A. (1977). The validity of design procedures for the permanent
deformation of asphalt pavements. In Proceedings of the 4th international conference on the
structural design of asphalt pavements, Ann Arbor (Vol. 1).
Byrne, P. M., Cheung, H., & Yan, L. (1987). Soil parameters for deformation analysis of sand
masses. Canadian Geotechnical Journal, 24(3), 366–376.
Ceylan, H., Gopalakrishnan, K., & Guclu, A. (2005a). Advanced approaches to characterizing
nonlinear pavement system responses. Journal of the Transportation Research Board.
538 H. Nabizadeh et al.

Ceylan, H., Guclu, A., Tutumluer, E., & Thompson, M. R. (2005b). Backcalculation of full-depth
asphalt pavement layer moduli considering nonlinear stress-dependent subgrade behavior.
International Journal of Pavement Engineering, 6(3).
Chen, D., Fernando, E., & Murphy, M. (1996). Application of falling weight deflectometer data for
analysis of superheavy loads. Journal of the Transportation Research Board (1540).
Chen, X., Lambert, J. R., Tsai, C., & Zhang, Z. (2013). Evaluation of superheavy load movement
on flexible pavements. International Journal of Pavement Engineering, 14(5).
Das, B. M. (2014). Shallow foundations. Boca Raton, FL: CRC Press.
Duncan, J. M., & Chang, C. (1970). Nonlinear analysis of stress and strain in soils. Journal of Soil
Mechanics and Foundations Division, ASCE, 96(5).
Federal Aviation Administration. (2015a). Falling/heavy weight deflectometer (F/HWD) equip-
ment round-up. http://www.airporttech.tc.faa.gov/Databases/FWD-HWD. Last accessed on
October 2015.
Federal Aviation Administration. (2015b). BAKFAA version 2.0. http://www.airporttech.tc.faa.
gov/Download/Airport-Pavement-Papers-Publications-Detail/dt/Detail/ItemID/34/BAKFAA-
version-20. Last accessed on October 2015.
Fernando, E. G. (1997). Guidelines for evaluating superheavy load routes. College Station: Texas
Transportation Institute. (Report No. TX -98/3923-S)
Garg, N., & Thomson, M. R. (1998). Triaxial characterization of Mn/ROAD granular materials.
Journal of the Transportation Research Board (1557).
Hajj, E. Y., Ulloa, A., Siddharthan, R., & Sebaaly, P. E. (2011). Estimation of stress conditions for
the flow number simple performance test. Journal of the Transportation Research Board
(2281).
Jooste, F. J., & Fernando, E. G. (1994). Victoria superheavy load move: report on route
assessment and pavement modeling. College Station: Texas Transportation Institute. (Report
No. FHWA/TX -94/1335-1)
Pekcan, O., Tutumluer, E., & Thompson, M. R. (2008). Nondestructive pavement evaluation using
ILLI-PAVE based artificial neural network models. Illinois Center for Transportation. (Report
No. FHWA-ICT-08-022)
Santha, B. L. (1994). Resilient modulus of subgrade soils: Comparison of two constitutive
equations. Journal of the Transportation Research Board (1462).
Schmalzer, P. N. (2006). LTPP manual for falling weight deflectometer measurements, version
4.1. Beltsville: MACTEC Engineering and Consulting, Inc. (Report No. FHWA-HRT-06-132)
Schmertmann, J. H., Brown, P. R., & Hartman, J. P. (1978). Improved strain influence factor
diagrams. Journal of the Geotechnical Engineering Division, 104(8), 1131–1135.
Siddharthan, R. V., Yao, J., & Sebaaly, P. E. (1998). Pavement strain from moving dynamic 3D
load distribution. Journal of Transportation Engineering, 124(6), 557–566.
Stark, T. D., Ebeling, R. M., & Vettel, J. J. (1994). Hyperbolic stress–strain parameters for silts.
Journal of Geotechnical Engineering, 120(2), 420–441.
Thompson, M. R., & Elliott, R. P. (1985). ILLI-PAVE based response algorithms for design of
conventional flexible pavements. Journal of the Transportation Research Board (1043).
Tutumluer, E., & Sarker, P. (2015). Development of improved pavement rehabilitation procedures
based on FWD backcalculation. (NEXTRANS Project No. 094IY04)
Tutumluer, E., Thompson, M. R., & Brar. R. Subgrade soil support and stabilization. http://www.
ceat.uiuc.edu/PUBLICATIONS/presentations/2004-OCT%20Open%20House/
Oct2004Tutumluer.ppt. Last accessed on October 2015.
Von Quintus, H. L., Minchin, R. E., Nazarian, S., Maser, K. R., & Prowell, B. (2009). NCHRP
Report 626: NDT technology for quality assurance of HMA pavement construction.
Washington, DC: Transportation Research Board, National Research Council.
Yau, A., & Von Quintus, H. L. (2002). Study of LTPP laboratory resilient modulus test data and
response characteristics. Austin: Fugro-BRE, Inc. (Report No. FHWA-RD-02-051)
Understanding Airfield Pavement
Responses Under High Tire Pressure:
Full-Scale Testing and Numerical
Modeling

Hao Wang, Navneet Garg and Maoyun Li

Abstract This paper aims to investigate airfield flexible pavement responses under
heavy aircraft loading with high tire pressure through an integration of full-scale
testing and numerical modeling. The new generations of aircraft, such as Boeing
787 and Airbus 350/380, have tire pressure close to or exceeding 232 psi. Limited
information is available on the effect of high tire pressure on HMA pavement
responses. A new series of high tire pressure tests were conducted at the Federal
Aviation Administration’s (FAA) heavy vehicle simulator-airport version (HVS-A)
Test Strips. An advanced three-dimensional (3-D) finite element (FE) model was
developed that characterized the hot-mix asphalt (HMA) layer as a viscoelastic
material to predict time- and temperature-dependent pavement responses under
various loading conditions. The accelerated pavement testing results indicate that
there is an insignificant effect of tire pressure on pavement rutting. The effect of tire
pressure was more significant beyond failure (1 in. surface rut). This is consistent
with the previous findings from high tire pressure tests. The numerical modeling
results show that as the critical pavement responses in the asphalt layer increased
slightly as tire pressure increased from 210 to 254 psi. The cross-anisotropic
non-linear behavior of granular base affects the tensile strains at the bottom of
asphalt layer significantly. The comparison between predicted and measured strains
emphasizes the importance of considering the realistic tire-pavement interaction and
the appropriate constitutive model for each pavement layer. The numerical mod-
eling can support and supplement the full-scale testing results and provide valuable
suggestions for mechanistic-based airfield pavement design under heavy aircrafts
with high tire pressure.

Keywords High tire pressure

Accelerated pavement testing
Numerical




modeling Viscoelastic Cross-anisotropic Nonlinear modulus

H. Wang
M. Li
Rutgers, The State University of New Jersey, New Brunswick, USA
N. Garg (&)
FAA Airport Technology R&D, WJHTC, Egg Harbor Township, USA
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 539

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_35
540 H. Wang et al.

1 Introduction

Traditionally, aircraft tire inflation pressure ranges from 174 to 218 psi (1.2–
1.5 MPa) depending on aircraft gross weights and landing gear configurations. The
new generations of aircrafts, like Boeing 787 and Airbus 350/380, have tire pres-
sure close to or exceeding 232 psi (1.6 MPa). This creates a challenge for the
traditional Pavement Classification Number (PCN) rating that includes four pres-
sure categories: W (no pressure limitation), X (218 psi [1.5 MPa] limitation), Y
(145 psi [1.0 MPa] limitation), and Z (72.5 psi [0.5 MPa] limitation) (ICAO 1983).
At the same time, the aircraft wheel load has increased significantly with the
development of long-range wide-body aircrafts. Currently, the single wheel-load of
A380 and B777 are close to or exceed 60,000 lbs. Therefore, it is necessary to
investigate the impact of heavy aircraft wheel load with high tire pressure on airport
pavement performance.
Accelerated pavement testing (APT) provides an acceptable solution between
real field pavement loading and laboratory tests to evaluate the loading and design
parameters on pavement damage. Accelerated pavement testing compresses many
years of pavement load-related deterioration into just a few months or weeks of
testing. During the APT, pavement responses to loading can be measured using
pavement instrumentation. The parameters that can be measured include strains,
stresses, deflections, moisture, temperature, etc. In situ measurements of pavement
responses allow for understanding the key factor affecting pavement responses and
developing accurate performance models for mechanistic-empirical pavement
design approaches. Pavement instrumentation has recently become an important
tool to monitor in situ pavement condition in response to loading and the
environment.
On the other hand, numerical modeling has become a powerful tool to simulate
pavement responses under different loading scenarios. The appropriate utilization of
numerical modeling could reduce the significant efforts that are required for con-
struction of full-scale pavement sections and pavement instrumentation. After the
numerical model is calibrated and validated, the computational environment would
enable to consider different combinations of material properties, structure designs,
loading configurations, and environmental conditions. The numerical modeling
results can be also used to check the reliability and accuracy of measurements and
predict pavement responses that are difficult to measure in the field. Modeling
analysis of pavement can be either based on the multilayer elastic theory (MLE) or
finite element model (FEM). The FEM approach is more powerful for pavement
analysis because it could simulate realistic tire-pavement interaction and complex
material behavior of each pavement layer.
Understanding Airfield Pavement Responses … 541

2 Objective

This paper aims to investigate airfield flexible pavement responses under heavy
aircraft loading with high tire pressure through an integration of full-scale testing
and numerical modeling. A new series of high tire pressure tests were conducted at
the National Airport Pavement Test Facility (NAPTF) using the state-of-art heavy
vehicle simulator-airport version (HVS-A). An advanced three-dimensional (3-D)
finite element (FE) model was developed that characterized the hot-mix asphalt
(HMA) layer as a viscoelastic material and considered the cross-anisotropy and
stress-dependency of aggregate base layer. Preliminary results on critical pavement
responses under various loading conditions are presented and analyzed. The sim-
ulation results were compared to the measured pavement responses for model
calibration and validation.

3 Review of Previous Studies on High Tire Pressure

3.1 Boeing and FAA High Tire Pressure Test

A series of full-scale tests have been conducted at the National Airport Pavement
Test Facility (NAPTF) to evaluate pavement responses under aircraft loading with
high tire pressure. The first high tire pressure test was initiated in 2005 by Boeing
(Roginski 2007). Three sections with different asphalt surface thicknesses (2, 4, and
6 in. [25, 100, and 150 mm]) were loaded by a single wheel at 2.5 mph (3.2 km/h).
The loading cycles were applied by increasing the load from 40,000 to 50,000 lbs
(178–222 kN) and the tire pressure from 140 to 240 psi (0.96–1.65 MPa). The
results showed that higher tire pressure with maximum single wheel load of 55,000
lbs (245 kN) can produce increasing rutting depth or extensive cracks to pavement
failure. The testing was stopped when the rutting depth reached 0.5–0.75 in. (12.5–
19 mm) or extensive cracking was observed. The pavement temperature during
testing was in the range of 70–80 °F. It was found that the rutting in the asphalt
layer was the main failure mode under heavy aircraft loading, and high tire pressure
had no adverse effect on flexible pavements that have stable asphalt layers and meet
thickness requirement.
A second series of high tire pressure testing was conducted in 2009 on heated
pavement sections (FAA 2010). The purpose was to duplicate the worst-case
conditions for pavement rutting likely to be encountered in the field. Two different
heating methods, hydronic heating system with hot water pipes and electrically
heated wire mesh system, were used in the initial test sections. The test sections
were rebuilt with the strengthened pavement structure that included a 5-in. asphalt
surface layer and a 17-in. (432 mm) Econocrete base layer placed on the DuPont
542 H. Wang et al.

clay subgrade. Two asphalt binder grades (PG 64-22 and PG 76-22) were used. The
hydronic heating system was finally used since it was proven to be more reliable
than the electrical heating system. The pavement temperature was kept between 100
and 110 °F during the cyclic loading process. Dual tires with 54-in. (137.2-cm)
spacing were used by using different levels of inflation pressure (210 and 245 psi
[1.45 and 1.69 MPa]) for each tire. The applied wheel loads were 52,500 and
61,300 lbs (234 and 273 kN) with three different wandering locations. The loads
were applied in one direction only with a trafficking speed of 1 ft/s (0.3 m/s). The
primary, secondary, and tertiary phases of rutting development were observed as
the loading cycles increased. It was found that the observed differences in rutting
depth were in the range 0–4 % due to the tire pressure effect.

3.2 Airbus High Tire Pressure Test

A series of high tire pressure tests were carried out in Toulouse, France by Airbus
and French STAC (Civil Aviation Technical Service) in 2010 to evaluate pavement
performance under multi-wheel loading of heavy aircraft (Airbus 2010). The Airbus
Heavy Traffic Simulator (HTS) was used in the test, which has full-scale landing
gear with modular assembly up to five bogies and had the capacity of generating up
to 70,500 lbs (314 kN) for each single wheel with a maximum speed of 5 mph
(8 km/h). The simulator was equipped with A340 tires and the tire inflation pres-
sure could be adjusted to simulate other aircraft tires. Four dual-wheel assemblies
were used to apply different levels of wheel load and tire pressure that were suf-
ficiently spaced enough to prevent any interaction between multiple wheels.
Seven test sections were designed according to the French airport pavement
design method for 10,000 passes of B747-400 gear. The pavement sections had a
10-in. (254-mm) asphalt layers with different thickness combination of surface and
base asphalt layers and a 15-in. subbase layer of untreated graded aggregate sup-
ported by a foundation. The loading configurations used in the test included two
load levels (57,400 and 66,400 lbs [255 and 295 kN]) and two pressure levels (218
and 254 psi [1.5 and 1.75 MPa]). The loading was applied on pavement up to
10,000 passes to cause the rutting depth of 0.5–0.75 in. (12.5–19 mm), which is
considered a medium severity rutting for airport pavement remediation. The aver-
age temperatures of asphalt surface layer were 68–86 °F (20–30 °C) in the initial
configurations and then increased to 104–122 °F (40–50 °C). The full-scale test
results showed that rut depth differences ranged from 0.075 to 0.2 in. (1.9–5 mm)
that varied depending on the magnitude of wheel loading. This indicated that the
increase of tire pressure from 218 to 254 psi (1.5–1.75 MPa) would not have critical
impact on the rutting development in the asphalt layer. The rutting initiation
appeared more affected by the average temperature in the asphalt layer than the
traffic condition or loading level.
Understanding Airfield Pavement Responses … 543

4 Full-Scale Testing Using HVS-A

4.1 Test Program

The FAA accepted the Heavy Vehicle Simulator-Airport Version (HVS-A) at the
National Airport Pavement Test Facility (NAPTF) in November 2013. The HVS-A
was designed to apply single- and dual-wheel system that can load up to 100 kips
(445 kN) in a wandering range of 6 feet (0.15 m). The pavement surface temper-
ature in the chamber can be heated up to 150 °F (65 °C) and the test speed for the
simulated wheel loading ranges from 0.17 to 5 mph (0.27–8 km/h). It was designed
to accommodate the 52×21.0R22 radial tire for single-wheel and smaller tires (such
as B737-800) would be assigned for dual-wheel assembly.
In the acceptance test of HVS-A, two pavement sections were built to evaluate
the effect of aircraft high tire pressure on responses of airfield flexible pavement.
The pavement structure is composed of a 10-in. (254-mm) asphalt layer (P401) with
the PG76-22 binder and a 15-in. (381-mm) aggregate base layer (P209). The
pavement sections were constructed on sandy subgrade with a California Bearing
Ratio (CBR) of 20. During construction, strain gauges were embedded at the
bottom of the asphalt layer and temperature gages were placed in the asphalt layer
at different depths (Fig. 1).
The accelerated pavement testing consisted of response tests and traffic tests. The
tire pressure was set to 210 psi (1.45 MPa) for tests on test strip-1 and 254 psi
(1.75 MPa) for tests on test strip-2. The objective of response tests was to study
asphalt concrete strains at different load levels. The pavement surface temperature
was 120 °F (49 °C) measured at a depth of 1 in. below pavement surface. The test
speed was 2 mph. Three levels of wheel loads were applied—30,000, 40,000, and
50,000 lbs (133, 178, and 222 kN). The lateral wander pattern included five wander
positions with the maximum offset of 20 in. (0.5 m) away from centerline. Data
from sensors embedded in pavement were collected for every loading cycle. Tests
on test strip-2 were performed after the traffic tests on test strip-1 were completed.

Fig. 1 Heavy Vehicle Simulator-Airport Version (HVS-A)

544 H. Wang et al.

Test parameters for traffic tests were mostly the same as the ones implemented in
response tests except the wheel load increased to 61,300 lbs (272 kN). Pavement
surface profile measurements were made using HVS-A on-board profiler before the
start of traffic tests. After that, surface profiles were measured after 18, 72, 144, 576,
1152, 2304, 4608, 9216 passes. Straight-edge rut depth measurements were mea-
sured at three locations (middle three blue lines on the test strips) at the same
intervals. The loading was applied in a predetermined wander pattern that simulates
a normally distributed traffic with approximate standard deviation of 12-in. (0.3 m)
and mean of zero. Each wander pattern consisted of 18 passes.

4.2 Test Results

Figure 2 shows the measured tensile strains from the response test at different load
levels, respectively, for longitudinal and transverse tensile strains. The time histo-
ries of tensile strains under the moving tire loading were plotted. All of the strain
gauges were embedded at the bottom of asphalt layer. Both the LSG-1 and LSG-2
gauges were embedded longitudinally at the centerline of the test strip. The TSG-2
gauge was embedded transversely at the centerline of test strip, while the TSG-2
strain gauge was embedded transversely at an offset of 28 in. away from the
centerline. Only the measured tensile strains under tire loading with the inflation
pressure of 210 psi (1.45 MPa) were reported here since the strain gauges in the test
strip-2 for testing at tire pressure of 254 psi (1.75 MPa) were damaged during
construction.
The longitudinal strains are negative values (compression) as the tire is far away,
then the positive values (tension) as the tire approaches the strain gage, followed by
the negative values (compression) again as the tire is leaving. On the other hand, the
transverse strains are always in tension or compression as the tire loading is
approaching and leaving, depending on the transverse offsets of strain gauges to the
loading location. This is because the direction of transverse strains was always
perpendicular to the tire moving direction. An unsymmetrical pattern of strain
shapes were observed for both longitudinal and tensile strains due to the vis-
coelasticity of asphalt mixture. The results clearly show that increasing loads
produced the greater peak values for the tensile strains. It is noted that the peak
values of longitudinal tensile strains were observed at different timings due to the
offset locations of the embedded strain gages.
Figure 3 shows the rut depth measurements for the two different tire pressures
levels. It shows that the rut depth increases nonlinearly as the loading pass
increases. The difference of rut depth under two pressure levels was around 0.1 in.
(2.54 mm) when the rutting depth is smaller than 1.0 in. (25.4 mm), while the
difference increased to 0.3 in. (76 mm) when the rutting depth reached 1.5 in.
(38 mm). This indicated that the effect of high tire pressure on rutting became more
noticeable when the rutting development progressed. It is believed that the total rut
Understanding Airfield Pavement Responses … 545

(a)

(b)

Fig. 2 a Longitudinal and b transverse strains measured under APT

546 H. Wang et al.

Fig. 3 Rut depth under APT at different tire pressure

depth measured at the pavement surface is caused by the accumulation of perma-

nent deformation in the asphalt layer, and not in the P-209 aggregate base layer, or
subgrade due to thick asphalt layer and high temperature.

5 Finite Element Modeling

5.1 Finite Element Model

A 3-D FE model of flexible pavement was simulated using the general-purpose FE

software ABAQUS. In the model, eight-node, linear brick elements with reduced
integration were used in the finite domains, whereas infinite elements were used at
boundaries to reduce a large number of far-field elements without significant loss of
accuracy and to create a “silent” boundary for the dynamic analysis. The infinite
element has a special shape function for the geometry at the infinite boundary and
thus has zero displacement as the coordinate approaches infinity. In addition, the
infinite elements introduce additional normal and shear tractions (viscous damping
boundary) on the FE boundary that are proportional to the normal and shear
components of the velocity of the boundary (ABAQUS 2010). This could minimize
the reflection of dilatational and shear wave energy back into the FE mesh and it
usually provides acceptable results for most practical cases.
Understanding Airfield Pavement Responses … 547

Figure 4 illustrates the 3-D FE model that discretizes the pavement structure.
The FE mesh is refined around the loading area along the wheel path; a relatively
coarse mesh is used far away from the loading area. The element horizontal
dimensions along the aircraft wheel loading area were dictated by the tire rib and
groove geometries. Hence, the length of elements within the loading area was
selected at 0.8–1.2 in. (20–30 mm) in the transverse direction and 1.6 in. (40 mm)
in the longitudinal (traffic) direction to have good aspect ratios. Based on the mesh
convergence analysis, the element thicknesses were selected to be 0.5 in. (12.5 mm)
for the asphalt surface layer and 1.2–2 in. (30–50 mm) for the aggregate base layers
in order to have a smooth stress transition between elements. A sensitivity analysis
was performed to define the infinite boundaries at both sides, as well as the bottom
of FE mesh. After comparing the maximum tensile and shear strains in the asphalt
layer, the locations of the infinite boundary in three directions from the load center
needed to be greater than 8.2 feet (2.5 m) in order to obtain the stable solutions (less
than 5 % changes). To achieve the balance between computation cost and accuracy,
the final selected domain size (finite + infinite) has an in-plane dimension of
20 × 30 feet (6 × 9 m) and a vertical dimension of 13 feet (4 m).
The aircraft tire loading was characterized with the non-uniformly distributed
contact stresses at the tire-pavement interface (Wang et al. 2013). The non-uniform
vertical contact stress distributions were based on the contact stress measurements
under heavy aircraft tire loading reported by Rolland (2009). In the longitudinal
direction, a half-sinusoidal distribution of vertical contact stress was used along the
contact length of each rib. The peak contact stresses beneath two edge ribs were
assumed equal to around 2.0 times the tire inflation pressure; while the peak contact
stresses under all other ribs were assumed equal to around 1.1 times the tire
inflation pressure. The aspect ratio (the ratio of contact width to length) of the tire
print was controlled depending on the load level. As the tire load increases, the

Fig. 4 Illustration of FE model and pavement structure

548 H. Wang et al.

contact length increases more significantly than the contact width due to the rigidity
of tire sidewall. The aspect ratio was selected as 0.71 when the tire loading is 30
kips (133 kN) loading and 0.63 when the tire loading is 50 kips (222 kN).

5.2 Material Properties

The viscoelastic material properties of asphalt mixture were considered in the finite
element model. Figure 5 shows the measured dynamic modulus and the fitted
master curve using the sigmoid function at a reference temperature of 68 °F (20 °
C). As expected, under a constant loading frequency, the dynamic modulus
decreases as the temperature increases; while under a constant testing temperature,
the dynamic modulus increases as the frequency increases.
The relaxation modulus was inter-converted from the dynamic modulus using
Eqs. 1 and 2 assuming that the linear viscoelascity of HMA was represented by a
generalized Maxwell solid model (Park and Kim 1999). The relaxation modulus
and relaxation times were determined by minimizing the sum of squares of the
errors (Eq. 3). The bulk and shear relaxation moduli were calculated assuming a
constant Poisson’s ratio. The relationship between the shift factor and the tem-
perature can be approximated by the Williams-Landell-Ferry (WLF) function
(ABAQUS 2010).

Fig. 5 Master curve of dynamic modulus at 20 °C

Understanding Airfield Pavement Responses … 549

Xn
x2 s2i Ei
E 0 ðxÞ ¼ E1 þ ð1Þ
i¼1
1 þ x2 s2i

Xn
xsi Ei
E 00 ðxÞ ¼ ð2Þ
i¼1
1 þ x2 s2i
" 2  00 2 #
Xk
E0 ðxÞcalcualted E ðxÞcalculated
min 1 þ 1 ð3Þ
j¼1
E0 ðxÞmeasured E 00 ðxÞmeasured

0
where E ðxÞ is real part of the dynamic modulus; E00 ðxÞ is imaginary part of the
dynamic modulus; E1 is equilibrium relaxation modulus at infinite time; x is
angular frequency; Ei , and si are Prony series parameters for relaxation modulus; n
is number of Maxwell elements; and k is number of data points from the
measurements.
To investigate the effect of aggregate base nonlinearity on pavement responses,
both linear isotropic and cross-anisotropic nonlinear models were used to predict
pavement responses. For the linear isotropic analysis, the elastic modulus was set
equal to a typical value of 47,826 psi (330 MPa) for the P-209 base layer. In the
nonlinear cross-anisotropic model, the vertical modulus is described using the
generalized model adopted in the proposed Mechanistic-Empirical Pavement
Design Guide (MEPDG), Eq. 4 (ARA 2004). In this model, the first stress invariant
or bulk stress term considers the hardening effect, while the octahedral shear stress
term considers the softening effect. The nonlinear coefficients (k1 = 2.800,
k2 = 1.184, and k3 = −1.597) were obtained from the literature where the
stress-dependent modulus of P-209 base material was measured at different mois-
ture contents (Nazarian et al. 2014). The stress dependency of Poisson’s ratios was
not considered in this study and the in-plane and out-of-plane Poisson’s ratios are
assumed constant. A user-material subroutine (UMAT) was developed for solving
nonlinear problems in ABAQUS (Wang and Al-Qadi 2013).
 k2  k3
h soct
Mrv ¼ k1 pa þ1 ð4Þ
pa pa
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
with h ¼ r1 þ r2 þ r3 and soct ¼ 13 ðr1  r2 Þ2 þ ðr2  r3 Þ2 þ ðr1  r3 Þ2
where Mrv is vertical resilient modulus (kPa) (E3 in Eq. 2); Mrh is horizontal resilient
modulus (kPa) (E1 in Eq. 2); Gr is shear resilient modulus (kPa) (G13 in Eq. 2); h is
bulk stress (kPa); soct is octahedral shear stress (kPa); r1 , r2 , and r3 are maximum,
middle, and minimum principal stresses; k1 ; k2 ; k3 are exponent parameters; and pa
is atmospheric pressure (100 kPa).
For the cross-anisotropic modulus, the horizontal and shear modulus ratios (n
and m) were used, as shown in Eqs. 5 and 6. Previous research has found that
horizontal modulus ratios and shear modulus ratios had a relatively small range of
550 H. Wang et al.

variation (Tutumluer and Thompson 1997). The modulus ratios were assumed
constant as 0.35 for n and m in this study as typical values. The linear elastic
modulus of the subgrade in the thin asphalt pavement section was estimated from its
California Bearing Ratio (CBR) value.

n ¼ Mrh =Mrv ð5Þ

m ¼ Gr =Mrv ð6Þ

5.3 Modeling Results

Table 1 summarizes the critical pavement responses caused by different tire pres-
sure levels, respectively, for tensile strains and shear strains in the asphalt layer. The
maximum tensile strains were found at the bottom of asphalt layer under tire center;
while the maximum shear strain was found at the shallow depth of asphalt layer at
the outmost tire rib. All critical pavement responses were calculated using the
nonlinear cross-anisotropic model for base layer and linear elastic modulus for
subgrade. The results show that as the tire pressure increases from 210 to 254 psi
(1.45–1.75 MPa), the changes of strain are smaller than 5 % in general. This means
that the effect of high tire pressure on fatigue cracking potential is not significant for
the airfield pavement structure with a 10-in. asphalt layer. On the other hand, the
effect of tire load on strain responses was found much more significant compared to
tire pressure effect. As the load increases, the increase of shear strain is more
significant than the increase of tensile strain. This is probably because the tire load
increase mainly causes the concentration of tire contact stress at tire edge ribs due to
the rigidity of tire sidewalls.
Table 2 compares the strain responses calculated using different models for the
granular base layer under two loading levels (30 and 50 kips [133 and 222 kN])
with tire pressure of 210 psi (1.45 MPa). The results show that the cross-anisotropic
stress dependent model results in 27–58 % greater tensile strains but only 2–4 %

Table 1 Comparison of pavement responses under different tire loads and pressure
Parameters Load = 30 kips Load = 50 kips
using different using different
tire pressure tire pressure
levels levels
Tire inflation pressure (psi) 210 254 210 254
Longitudinal tensile strain 764 785 910 944
at the bottom of asphalt layer (micro)
Transverse tensile strain 996 1032 1081 1123
at the bottom of asphalt layer (micro)
Maximum shear strain in asphalt 2176 2204 3746 3926
layer (micro)
Understanding Airfield Pavement Responses … 551

Table 2 Comparison of pavement responses using different models for aggregate base layer
Pavement responses Load = 30 kips using Load = 50 kips using
different base models different based models
Linear Nonlinear Linear Nonlinear
isotropic cross-anisotropic isotropic cross-anisotropic
Longitudinal tensile strain 599 764 718 910
at the bottom of asphalt
layer (micro)
Transverse tensile strain at 661 996 690 1081
the bottom of asphalt layer
(micro)
Maximum shear strain in 2083 2176 3666 3746
asphalt layer (micro)

greater shear strains, compared to the results obtained using the traditional linear
isotropic model for granular base. It indicates that the stress-dependency and
cross-anisotropy of the aggregate base layer has more significant effect on the
tensile strain than on the critical shear strain of asphalt layer. This may be attributed
to the fact that the bending behavior of asphalt layer is more sensitive to the
underlying layer support.
Figure 6 plots the strain-time history predicted form the FE model, respectively,
for longitudinal and transverse tensile strains. The strain development trend is
consistent with the measured strain pulses. This emphasizes the importance of
considering moving load pattern and viscoelastic asphalt layer in the FE model for

Fig. 6 Caculated tensile strains from FE model

552 H. Wang et al.

Fig. 7 Comparision of tensile strains from accelerated pavement testing and numerical modeling

predicting pavement responses. Figure 7 compares the calculated and measured

tensile strains in the asphalt layer under two loading levels (30 and 50 kips [133 and
222 kN]) with tire pressure of 210 psi (1.45 MPa) when the tire loading has
different offsets away from the strain gauge. In general, the modeling results agree
well with measurement results. The discrepancy could be caused by the reason that
the plastic behavior of asphalt concrete was not considered in the model, which will
be considered in the future work.

6 Conclusions

The following conclusions were concluded from the analysis:

1. The accelerated pavement testing results indicate that the number of passes
required for 1-in. surface rut (failure criterion) were similar for the two pressure
levels (210 and 254 psi [1.45 and 1.75 MPa]). This observation was obtained
from pavement testing conducted at the high temperature and it is consistent
with the previous findings from high tire pressure tests.
2. Although the comparison of strain responses were not available from the
accelerated pavement testing, the numerical modeling results show as that the
critical pavement responses in the asphalt layer increased slightly as tire pressure
increased from 210 to 254 psi (1.45 to 1.75 MPa).
3. The cross-anisotropic non-linear behavior of granular base affects the tensile
strains at the bottom of asphalt layer significantly. The comparison between
Understanding Airfield Pavement Responses … 553

predicted and measured strains emphasizes the importance of considering the

realistic tire-pavement interaction and the appropriate constitutive model for
each pavement layer.
The numerical modeling can support and supplement the full-scale testing results
and provide valuable suggestions for mechanistic-based airfield pavement design
under heavy aircrafts with high tire pressure. It is noted that this paper presented the
preliminary analysis results from the FE model that focused on the strain responses
in the asphalt layer. Ongoing work is being conducted to predict rutting depth at
different pressure levels using mechanistic-empirical pavement analysis framework.

Acknowledgments The authors would like to acknowledge Federal Aviation Administration for
providing funding for the research. The contents of the paper reflect the views of the authors, who
are responsible for the facts and accuracy of the data presented within. The contents do not
necessarily reflect the official views and policies of the FAA. The paper does not constitute a
standard, specification, or regulation.

References

ABAQUS. (2010). ABAQUS/standard user’s manual, Version 6.10. Pawtucket, RI: Hibbitt,
Karlsson & Sorenson, Inc.
Airbus. (2010). High tire pressure test. Technical Report.
ARA, Inc., ERES Division. (2004). Guide for mechanistic-empirical design of new and
rehabilitated pavement structures. Washington, DC: TRB. (NCHRP 1-37A Final Report)
FAA. (2010). Full-scale high tire pressure tests on heated pavement. Final Report, Submitted to
ICAO Aerodrome Operations and Services Working Group.
ICAO. (1983). Aerodrome design manual, part 3, pavements (2nd ed.). Document 9157-AN/901,
International Civil Aviation Organization.
Kim, I. T., & Tutumluer, E. (2004). Predicting rutting performance of pavement granular layers at
the FAA’s National Airport Pavement Test Facility. In Proceeding of 2004 FAA worldwide
airport technology transfer conference.
Nazarian, S., et al. (2014). Modulus-based construction specification for compaction of earthwork
and unbound aggregate: appendices. Transportation Research Board. (NCHRP 10-84)
Park, S. W., & Kim, Y. R. (1999). Interconversion between relaxation modulus and creep
compliance for viscoelastic solids. Journal of Materials in Civil Engineering, 11(1), 76–82.
Rolland, E. (2009). Tire pressure test effect on pavement. Airbus high tire pressure workshop (in
CD). Toulouse, France.
Wang, H., & Al-Qadi, I. L. (2013). Importance of nonlinear anisotropic modeling of granular base
for predicting maximum viscoelastic pavement responses under moving vehicular loading.
Journal of Engineering Mechanics, 139(1), 29–38.
Wang, H., Portas, S., Coni, M., & Al-Qadi, I. L. (2013). Three-dimensional finite element
modeling of instrumented airport runway pavement responses. Transportation Research
Record: Journal of the Transportation Research Board, No. 2367, TRB, National Research
Council, pp. 76–83.
Use of Ground Penetrating Radar
at the FAA’s National Airport Pavement
Test Facility (NAPTF)

Injun Song and Albert Larkin

Abstract The Federal Aviation Administration (FAA) in the United States has
used a ground-coupled Ground Penetrating Radar (GPR) at the National Airport
Pavement Test Facility (NAPTF) since 2005. One of the primary objectives of the
Accelerated Pavement Testing (APT) at the facility is to provide full-scale pave-
ment response and failure information for use in airplane landing gear design and
configuration studies. During the traffic testing at the facility, a GSSI GPR system
was used to develop new procedures for monitoring Hot Mix Asphalt
(HMA) pavement density changes that is directly related to pavement failure.
A new methodology showing HMA density changes in terms of dielectric constant
variations, called dielectric sweep test, was developed and applied in full-scale
pavement test. The dielectric constant changes were successfully monitored with
increasing airplane traffic numbers. The changes were compared to pavement
performance data (permanent deformation). The measured dielectric constants
based on the known HMA thicknesses were compared with computed dielectric
constants using an equation from ASTM D4748-98 Standard Test Method for
Determining the Thickness of Bound Pavement Layers Using Short-Pulse Radar.
Six inches diameter cylindrical cores were taken after construction and traffic
testing for the HMA layer bulk specific gravity. The measured bulk specific gravity
was also compared to monitor HMA density changes caused by aircraft traffic
conditions.

Keywords Ground penetrating radar Accelerated pavement testing
Full-scale



pavement testing HMA density Dielectric constant

I. Song (&)
SRA International, Inc., 1201 New Road Suite#242, Linwood, NJ 08221, USA
e-mail: [email protected]
A. Larkin
FAA Airport Technology R&D Branch, ANG-E26 William J. Hughes Technical Center,
Atlantic City International Airport, Egg Harbor Township, NJ 08405, USA

© Springer International Publishing Switzerland 2016 555

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_36
556 I. Song and A. Larkin

1 Introduction

Electromagnetic (EM) wave as a geophysical test method was adopted successfully

for the application of Ground Penetrating Radar (GPR) in pavement area since early
1970s (Hironaka et al. 1976). The Federal Aviation Administration (FAA) has used
both ground and air couple GPR at the National Airport Pavement Test Facility
(NAPTF) and in-service runway airfield pavements since 2005. The application of
GPR for the airport pavement is introduced and the behaviors of the HMA surface
layer under the full-scale traffic loading are characterized in terms of the surface
layer dielectric constant changes and the progress of permanent deformation (rut-
ting); a dielectric constant sweep test was introduced for monitoring the constant
changes corresponding to the pavement performances.
Song et al. developed a methodology to monitor dielectric constant changes in
Hot Mix Asphalt (HMA) layer so called dielectric constant sweep test (Song et al.
2014). In this paper the methodology was validated using bulk specific gravities
from cylindrical cores taken from full-scale HMA pavement.

2 Ground Penetrating Radar Applications at the Federal

Aviation Administration

A full-scale HMA pavement was constructed at the NAPTF which is an indoor

facility excluding water infiltrations to the pavement layers under airplane loading
conditions. In addition to the 900 MHz ground couple GPR antenna, the FAA

Fig. 1 The FAA customized pavement evaluation vehicle mounted ground couple (400 MHz)
and air couple (2 GHz) GPR antennas
Use of Ground Penetrating Radar … 557

Fig. 2 GPR applications to a asphalt, b concrete, and c soil at the FAA’s NAPTF
558 I. Song and A. Larkin

customized a pavement evaluation vehicle called NDT Van mounted 400 MHz and
2 GHz GPR antennas, line laser imaging system, mobile inertial profiler, and as
shown in Fig. 1.
The FAA applies GPR to asphalt mixtures, concrete, and soils for measuring
pavement layer thickness, locating dowel bars, and identification of sewer pipes,
respectively, at multiple locations including in-service airfield pavements. Figure 2
shows the GPR applications to different materials and the results using manufac-
turer’s analysis software. The GPR outputs based on acquired images for the
pavement layer thickness and vertical and horizontal locations of embedded steel
object were confirmed by known design thickness or cylindrical cores after con-
struction. Since typical dielectric constants for asphalt and dry concrete are ranged
from 3.0 to 8.0, one arbitrary constant was initially picked in the range for each
material and gradation, adjusted, and calibrated in between 3.0 and 10.0 for the
materials. The constants 6.5 and 8.0 were set up for the corresponding design
thickness 13 cm asphalt and 31 cm concrete pavement layers after calibrations.
A dielectric constant 16.0 was used for the soil following GPR manufacturer’s
manual (Geophysical Survey System Inc.).

3 Test Conditions

A flexible pavement test section called high tire pressure (HTP) constructed at
NAPTF is on a CH clay subgrade known as DuPont clay. The pavement structure
consisted of 43 cm of econocrete as a stabilized base course (P-306) and a 13 cm
HMA surface layer (P-401) (Federal Aviation Administration 2011). Therefore, a
high EM velocity layer (P-401) was located above a low velocity layer (P-306) to
maximize reflections below the high velocity layer. The HMA layer was placed
with two different mix designs. The mixes used the same aggregate and aggregate
grading, but different asphalt binders. One mix had a straight PG 64-22 asphalt
binder and the other had a polymer modified PG 76-22 binder. A hydronic heating
system was embedded in the econocrete at its base to control the temperature of the
HMA at a temperature representative of high temperature airport operations. The
as-constructed dimensions of the HTP test pavement structure was 35 m long and
3.7 m wide. Figure 3 shows the dense gradations, Job Mix Formula (JMF) for
aggregates, used for both PG 64-22 and PG 76-22 asphalt mixtures with the FAA’s
P-401 specification requirements.
Dual tires with 137 cm spacing center to center mounted on one of the loading
modules on the NAPTF test vehicle were used for the test loading. The tires were
inflated to 1.45 and 1.69 MPa for the North and South wheels, respectively. Higher
tire pressures up to approximately 1.5 and 1.75 MPa were reached during testing.
Wheel loads were 23.8 and 27.8 MT with the load application pattern as indicated
in Fig. 4. The wander pattern of +18 cm (toward the South), 0 (equal distances
from the centerline to each tire), and −18 cm (toward the North) is shown in Fig. 4a
and the loading pattern is shown in Fig. 4b. The loads were applied at a trafficking
Use of Ground Penetrating Radar … 559

Fig. 3 Aggregate gradations used for HMA layer constructed at NAPTF

Fig. 4 Trafficking conditions for the high tire pressure test area a foot prints for wander pattern
with 18 cm spacing; b loading conditions (the dotted blue lines show the position of the transverse
profile measurements)
560 I. Song and A. Larkin

speed of 0.305 m/s from west to east as defined one pass number. The module was
unloaded at the east end and returned to the west end at 1.22 m/s, after which the
wander position was changed and the loading pattern repeated.

4 Data Collections

4.1 Ground Penetrating Radar System

The FAA owns GPR system, GSSI TerraSIRch SIR System-3000, was used for the
data collection presented in the paper. The system equipped with 900 MHz antenna
(Model 3101D) and RADAN 6.5 data analysis software. The penetration depth
setup was determined from surface to 91.4 cm to cover the layer(s) in bound
materials. Some of the setups for the antenna are as follows;
• Sample per scan: 512
• Resolution: 16 bits
• Vertical high pass filter: 225 MHz
• Vertical low pass filter: 2500 MHz
• Transmit rate: 100 kHz
The collected data was processed and interpreted in RADAN 6.5. Color was
customized in gray scale for the interpretations. Additional display options such as
Wiggle and O-Scope were selected to confirm the color-amplitude interpretations.

4.2 Transverse Pavement Surface Profiles for Monitoring

Permanent Deformation

It is well known that permanent deformation is one of the dominant deteriorations

leading to pavement failure in HMA pavements. The deterioration rates will be
controlled by, but not limited to, traffic conditions and material properties from each
pavement layer. The layers are composed of insulating materials in combinations of
aggregates, asphalt cement, and air, and generally experience compressive and
shear stresses during the trafficking.
Transverse profiles were measured during trafficking to monitor increasing
permanent deformation at the middle of each test item, 6.1 m apart, as shown in
Fig. 4b. GPR measurements were made at approximately every 21 passes or
whenever significant changes were observed. Since traffic testing at the NAPTF is
conducted on a pavement area 18.3 m wide, the transverse profiles were measured
with the FAA’s 20 m long truss profiler equipped with a non-contact vertical dis-
placement transducer as shown in Fig. 5.
Use of Ground Penetrating Radar … 561

Fig. 5 FAA truss type transverse profiler equipped with infrared laser and encoder

4.3 Cylindrical Cores

Cylindrical cores were taken from HMA surface layer before trafficking and after
the pavement failure. The 150 mm diameter cylindrical cores were taken at multiple
locations to measure bulk specific gravities. The results were used for HMA layer
density changes which directly related to variation of dielectric constants as well as
construction quality assurances.

5 Development of Dielectric Constant Sweep Test

5.1 Principle of Ground Penetrating Radar for Data

Interpretation

GPR uses EM waves to evaluate subsurface information by emitting a short pulse of

electromagnetic energy and examining the reflected energy from the pulse. The
success of GPR is based on, even though attenuation due to electrical conductivity
and dielectric losses may be significant, EM waves operating in the frequency range
where displacement currents dominate and losses associated with conduction cur-
rents are minimal (Reppert et al. 2000).
The GPR technology is based on Maxwell’s equations, which describe the
propagation of EM waves with a medium. Based on the equations, only one
562 I. Song and A. Larkin

parameter defines the interactions between EM waves and a nonmagnetic medium.

All pavement materials and subgrade soils and rocks are essentially nonmagnetic
medium. For virtually lossless materials, such as materials with very low electric
conductivity, which mostly applies to concrete and asphalt pavement materials in a
dry condition, the imaginary part ε″ can be neglected (Uddin 2006).

5.2 Dielectric Constant Changes with Airplane Traffic

Loading

The dielectric constant was assumed to be the only parameter related to monitoring
HMA asphalt layer density changes when the 900 MHz ground couple GPR
antenna was used. The assumption was based on the findings from the previous
research (Reppert et al. 2000).
The HMA layer in aggregates, asphalt cement, and air experience compressive
and shear stresses during the traffic loading as described earlier. Mechanism of
HMA pavement failure in conjunction with volumetric changes and material
shifting in the layer, which is directly related to changes of EM velocity, can be
explained by the stresses. However, current dielectric constant setup for data
acquisitions is determined by calibrations with extracted cores (material properties
and thickness) after construction or design thickness before opening traffic. The
determined constant before traffic loading is applied for the whole pavement life,
even though density changes are expected under traffic loading in the HMA layer. It
is difficult to monitor the subsurface conditions of airport pavement because of
safety and economic issues for airport operators, in closing an in-service runway or
taxiway pavement.
For GPR data collection during the full-scale airplane traffic loading, the con-
structed HTP pavement structure at NAPTF could isolate the HMA surface layer
placed above low strength concrete base layer causing relatively low EM velocity.
The HMA layer thickness is monitored using 900 MHz antenna by changing
dielectric constants from 1.0 to 10.0 going a 1.0 increment at each predetermined
pass number, 0, 273, 483, 672, 1155, and 1344. The pavement surface elevation
changes are also monitored by measuring transverse profiles to compute the HMA
layer thickness changes at the time of GPR measurements.
The measured transverse profiles were processed to compute HMA layer
thickness changes. Figure 6 illustrates the thickness changes with increasing pass
numbers. Each transverse profile data was collected at a traffic pass number and
plotted in a chart to show the pavement thickness changes. The changes can be
simply defined by the difference between two thicknesses, d1 and d2, at two dif-
ferent pass numbers because the econocrete base layer is not deformed by the
loadings in both 1.69 and 1.45 MPa tire pressure areas.
Assuming that a homogeneous medium is under stress without any material
property changes except thickness, GPR data should provide the same thickness
Use of Ground Penetrating Radar … 563

1.69 MPa Pavement Surface 1.45 MPa

Before Traffic
South Loading North
Depth (cm)

5.0

Width (m)
d1
d2
-13.0
Econocrete Base

Fig. 6 Transverse profile changes from the 1.69 and 1.45 MPa wheel paths at increasing pass
number
HMA Layer Thickness (cm)

d1

L2 L2' L1
d2

1 2

Dielectric Constant

Fig. 7 Dielectric constant sweep test using GPR in HMA layer

changes, as what was measured from transverse profile data with 0.25 cm accuracy
(Song et al. 2012). Since the HMA layer is inhomogeneous and unisotropic made of
mixtures in aggregates, asphalt cement, and air, the dielectric constant is expected
to change. Figure 7 shows the concept of dielectric constant sweep test including
564 I. Song and A. Larkin

the procedures to calculate the amount of dielectric constant changes. L1 is a

regression model at zero (0) traffic passes in exponential function developed with
dielectric constant setup values as independent variables and GPR measured HMA
thickness as a dependent variable. L2 can be developed after a certain number of
traffic passes in case of no dielectric constant changes like a homogeneous material.
Because of the constant changes, L2′ would be seen in HMA layer shifted from
regression line L2. The subtraction of d2 from d1 on the Y-axis showing HMA layer
thickness will represent the amount of HMA thickness reduction. The subtraction of
ε1 from ε2 on the X-axis showing dielectric constant will represent dielectric
constant increase relating to increase in HMA density during traffic loading.

6 Test Result and Analysis

6.1 Dielectric Constant Sweep Test

The GPR measurements were performed at the stations 215, 253, and 270 for both
higher and lower tire pressure loading areas following the transverse profile loca-
tions marked in dotted blue lines in Fig. 4b. Therefore, the GPR data were collected
at the different levels of tire pressures, loading amounts, and pavement temperatures
as well as multiple levels of traffic repetitions.
One of the dielectric constant sweep test (DCST) results based on the GPR
computed HMA thickness is presented in Fig. 8. Regression modeling was
developed for each traffic number in power low function as Saarenketo developed
for void content (Saarenketo 1997). The correlation coefficient (R2) for the all
regression models were above 0.9960.

Fig. 8 Dielectric constant sweep test result using GPR at station 270 (S)
Use of Ground Penetrating Radar … 565

6.2 Dielectric Constant Changes with Permanent

Deformation

The dielectric constant changes were compared at different levels of permanent

deformations. Figure 9 shows permanent deformations with increasing pass numbers
at different material properties and test conditions which marked the GPR measure-
ment points in circles. As expected, the solid dots labeled station 270 (S) representing
softer asphalt binder (PG64-22) at higher HMA temperature show rapid deformation
progress relative to the empty dots labeled station 215 representing stiffer asphalt
binder (PG76-22) at lower HMA temperature.
Relative dielectric constants, normalized constants based on the initial values,
compared to initial measurements were computed based on the developed DCST
and plotted in Fig. 10. Except the drastic constant drops after 273 passes, the
patterns generated by the relative dielectric constants are similar to the patterns
monitored for permanent deformation as shown in Fig. 9. Similar behaviors were
monitored with same material properties and loading levels. Station 270 (N) and
270 (S) were under identical test conditions except tire pressures, 1.69 and
1.45 MPa. No significant dielectric constant changes were identified from station
215.
The drastic constant drops after 273 passes occurred at every locations and they
are very close to the first inflection points in permanent deformation chart shown in
Fig. 9. In general, asphalt mixtures experience three deformation stages, primary,
secondary, and tertiary. Most plastic deformation occurs at primary and secondary
stages after the first inflection point. Especially, in the primary stage, most of the air
voids in the HMA layer will be removed due to compaction. The structural strength

Fig. 9 Permanent deformation and GPR measurements with increasing pass number
566 I. Song and A. Larkin

Fig. 10 Relative and GPR measurements with increasing pass numbers

would be a primary factor after the first inflection point, which might include
re-positioning aggregates and minor shear movement. The property changes at
around the first inflection point could generate the peak at 273 passes.

6.3 Comparison to ASTM Model and Bulk Specific Gravity

Changes

Equation (1) shows ASTM D4748-98 definition for the dielectric constant (εr) in
terms of two-way pulse travel time through layer (Δt) in ns, layer thickness in mm,
and speed of light in air in mm/ns (ASTM 1998).

er ¼ ½ðDt  cÞ=ð2  TÞ2 ð1Þ

Figure 11 shows the comparisons of εr collected from the DCST (εr—DCST)

with the computed ASTM model in Eq. (1) (εr—ASTM).
The extracted HMA cores before and after full-scale traffic testing were used for
bulk specific gravity to monitor the effect of traffic loading, which is directly related
to the dielectric constant changes monitored by GPR. The dotted lines with square
and circular data points represent bulk specific gravity changes with increasing pass
(traffic) numbers at PG 76-22 and PG 64-22 test items, respectively, in Fig. 12. The
solid lines showing dielectric constants monitored by GPR are compared with the
bulk specific gravity changes.
Use of Ground Penetrating Radar … 567

Fig. 11 Dielectric constant comparisons between dielectric constant sweep test (DCST) and
ASTM model

Fig. 12 Dielectric constant comparisons between dielectric constant sweep test (DCST) and
ASTM model
568 I. Song and A. Larkin

7 Conclusion

• GPR was successfully used to monitor typical permanent deformation patterns

based on the monitored dielectric constant changes under traffic loading.
• Significant dielectric constant changes are noticed after the primary stage in
HMA layer.
• The DCST methods are developed using GPR applications to full-scale asphalt
pavement test conditions.
• Dielectric constant changes at HMA layer during traffic loadings were moni-
tored and compared with computed dielectric constant values. The monitored
and computed dielectric constant values show strong correlations with R2 of
0.9542.
• The monitored dielectric constant changes at HMA layer during traffic loadings
were also compared with bulk specific gravity changes. The bulk specific
gravity was increased with increasing pass (traffic) numbers as the dielectric
constant values are increased.
• Further research analysis in the future using GPR data with increased dielectric
constant sampling rates could be adopted for HMA layer construction quality
control to replace current nuclear density and bulk specific gravity methods.
Using the developed methodology, only one or two cores could be needed for
the HMA pavement area with same mix design to get density and thickness
information from the placed HMA layer.

Acknowledgments The work described in this paper was supported by the FAA Airport
Technology Research and Development Branch. The contents of the paper reflect the views of the
authors who are responsible for the facts and accuracy of the data presented within. The contents
do not necessarily reflect the official views and policies of the FAA or SRA International, Inc. The
paper does not constitute a standard, specification, or regulation.

References

ASTM. (1998). D4748-98, Standard Test Method for Determining the Thickness of Bound
Pavement Layers Using Short-Pulse Radar, West Conshohocken, Pennsylvania.
Federal Aviation Administration. (2011). Advisory Circular 150/5370–10E, Standards for
Construction of Airports, Washington, DC (pp. 195–208, 230–232).
Geophysical Survey Systems, Inc. TerraSIRch SIR System-3000 User’s Manual, MN72–433
Rev D, North Salem, NH (p. 71).
Hironaka, M. C., Hitchcock, R. D., & Forrest, J. B. (1976). Detection of voids underground and
under pavements. Hueneme: Naval Civil Engineering Laboratory. (Report No. CEL-TN-1449)
Reppert, P. M., Morgan, F. D., & Tosoz, M. N. (2000). Dielectric constant determination using
ground-penetrating radar reflection coefficient. Journal of Applied Geophysics, 43, 189–197.
Saarenketo, T. (1997). Using ground-penetrating radar and dielectric probe measurements in
pavement density quality control. In Transportation Research Record, 1575, Journal of the
Transportation Research Board, Washington, DC (pp. 34–41).
Use of Ground Penetrating Radar … 569

Song, I., Hayhoe, G. F., & Aponte, R. (2012). Rut depth measurement method and analysis at the
FAA’s National Airport Pavement Test Facility (NAPTF). In 4th International conference on
accelerated pavement testing, Davis, California, September 2012.
Song, I., Larkin, A., & Gagnon, J. (2014). Monitoring Hot Mix Asphalt pavement density changes
using Ground Penetrating Radar at the FAA’s National Airport Pavement Test Facility. In 15th
international conference of ground penetrating radar (GPR), Brussels, Belgium.
Uddin, W. (2006). Ground penetrating radar study—Phase I technology review and evaluation.
Jackson: Mississippi Department of Transportation Research Division. (Report
No. FHWA/MS-DOT-RD-06-182)
 Part VII
Relating Laboratory Tests to Performance
Using Accelerated Pavement Testing
Development and Calibration
of Permanent Deformation Models

Fabricio Leiva-Villacorta, Adriana Vargas-Nordcbeck,

José Pablo Aguiar-Moya and Luis Loría-Salazar

Abstract As part of the first accelerated pavement tests completed in Costa Rica,
laboratory permanent deformation tests were performed on asphalt mixture, gran-
ular base and soil samples from the test sections. The objective of this study was to
calibrate prediction models using measured rutting data collected from instru-
mented flexible pavements. For the asphalt concrete the laboratory test was per-
formed using unconfined cyclic loading at three different temperatures. Permanent
deformation was found to be a function of the resilient strain, temperature and
number of loading cycles. For the granular base and soil samples, the test was
performed under repeated axial cyclic stress at different magnitudes and different
confining stresses. Permanent deformation was found to be a function of the con-
fining stress, deviator stress, moisture content, and number of loading cycles. Four
instrumented flexible pavements were subjected to heavy vehicle simulator testing.
Backcalculated moduli from multi depth deflectometers were introduced into an
elastic multilayer system to obtain pavement responses. The number of equivalent
standard axle load repetitions, the effective asphalt concrete mean temperature, and
the in-place moisture content along with computed responses were used to calculate
permanent deformation for the different layers. By comparing the measured rut
depths with the predicted values from laboratory based models, the optimum
combination of field calibration factors was determined so that the coefficient of
variation was minimal by means of ordinary least squares.

Keywords Permanent deformation
HVS
Instrumentation

F. Leiva-Villacorta (&)
A. Vargas-Nordcbeck
J.P. Aguiar-Moya
L. Loría-Salazar

LanammeUCR, CP-11501-2060 San José, Costa Rica
e-mail: [email protected]
A. Vargas-Nordcbeck
e-mail: [email protected]
J.P. Aguiar-Moya
e-mail: [email protected]
L. Loría-Salazar
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 573

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_37
574 F. Leiva-Villacorta et al.

1 Introduction

Calibration of pavement performance models is a systematic process to eliminate

the bias and minimize the difference between observed or measured performance
from the actual pavements and predicted results from an empirical or mechanistic
model (Von Quintus et al. 2007). The implementation of the Mechanistic Empirical
Pavement Design Guide (MEPDG) includes one of the largest efforts to calibrate
pavement performance models, first taking into consideration the various climatic
regions across the United States and more recently local or state calibrations
(NCHRP 2004). For its implementation the MEPDG not only focused on the
calibration of the different models but also the testing and development of those
models at the laboratory level. The idea behind the calibration process is to be able
to predict pavement performance from material properties and laboratory developed
models in a reliable form. In order to have a reliable model in a short period of time,
actual data from full scale accelerated testing facilities are currently being used
(Harvey and Kanekanti 2006; Ullidtz et al. 2006; Powell 2006).
Rutting is an undesired distress in a pavement for several reasons. For the road
users, it gives an increase of fuel consumption due to increased friction and
increased risk of hydroplaning in wet weather conditions. Rutting is caused by
several factors such as density of the layer (Lekarp 1997), stress conditions
(Werkmeister et al. 2003), and number of load applications, among others, and
occurs in different layers of the pavement. A rule of thumb is that the wider the ruts
are, the deeper in the pavement the permanent deformation has occurred.
There are numerous models for predicting the permanent deformation. The aim
of these models is to make a good forecast of the actual rut development. For
asphalt concrete, the NCHRP 1-37A mechanistic-empirical model was used in this
study. This model estimates the permanent strain in the asphalt layer as function of
the resilient strain, temperature and number of loading cycles. On the other hand,
the model developed by Guimaraes (2009) for unbound layers was used in this
study for a granular base and subgrade materials. This model estimates the per-
manent deformation in the unbound layer as function of the confining stress,
deviator stress and number of loading cycles.
In order to predict the actual behavior of the unbound material under loading
conditions, several testing methods are available. The idea behind these existing
testing devices is to simulate the stresses occurring in the field beneath the traffic
load. In this study only the repeated load triaxial test has been considered.
Finally, in order to perform the calibration from laboratory to field conditions,
four instrumented flexible pavements were subjected to Heavy Vehicle Simulator
(HVS) testing. Backcalculated moduli from multi depth deflectometers were
embedded into an elastic multilayer system to obtain pavement responses. The
number of equivalent standard axle load repetitions, the effective asphalt concrete
mean temperature, and the in-place moisture content along with computed
responses were used to calculate permanent deformation for the different layers. By
Development and Calibration of Permanent Deformation Models 575

comparing the measured rut depths with the predicted values from laboratory based
models, the optimum combination of field calibration factors was determined so that
the coefficient of variation was minimal by means of ordinary least squares (OLS).

1.1 Objective

The objective of this study was to calibrate permanent deformation prediction

models using measured rutting data collected from instrumented flexible pavements
under accelerated pavement testing conditions.

1.2 Test Sections

For the first stage of accelerated tests in Costa Rica the construction of 4 experi-
mental sections was performed in May 2013 (Fig. 1). The objective of this phase
was to perform a structural comparison in terms of thickness of the asphalt concrete
layer and base material type (granular vs. cement treated) (Aguiar-Moya et al.
2012). Table 1 shows the characteristics of the 4 sections with their respective layer
thicknesses obtained from Ground Penetrating Radar (GPR) measurements and
backcalculated layer moduli based on Falling Weight Deflectometer results.

Fig. 1 Test track distribution

576 F. Leiva-Villacorta et al.

Table 1 Test tracks in-place properties

Properties\section AC1 AC2 AC3 AC4
Asphalt concrete thickness (H1), cm 6.1 6.3 13.2 13.2
Granular base thickness (H2), cm 21.9 21.2 31.0 24.9
Subbase thickness (H3), cm 30.1 30.1 30.1 30.1
AC modulus (E1) @ 25 °C, MPa 3800 3800 3800 3800
Base modulus (E2), MPa 1200 170 170 1200
Subbase modulus (E3), MPa 140 140 140 140
Subgrade modulus (E4), MPa 70 70 70 70

The top layer consists of an asphalt concrete (AC) mixture with nominal max-
imum aggregate size of 19.0 mm with an optimum binder content of 4.9 % by total
weight of mixture. The cement treated base (CTB) was designed to withstand
35 kg/cm2 with an optimum cement content of 1.7 % by volume of aggregate and
with a maximum density of 2013 kg/m3. The base material and granular sub-base
were placed at a maximum density of 2217 kg/m3 with an optimum moisture
content of 8.6 %. The sub-base material had a CBR of 95 %. Finally, the subgrade
material was constructed for a maximum density of 1056 kg/m3 with an optimum
moisture content of 52 % and CBR of 6.6 %.
By September 2015, over a million load repetitions ranging from 40 to 80 kN
have been applied on each section. The total estimated equivalent 40 kN load
repetitions is over 50 million in a 2 years period of time.

1.3 Instrumentation

The experiment included not only the instrumentation integrated with the HVS
system but also embedded instrumentation in all four test sections. HVS onboard
sensors can record the applied load, tire pressure and temperature, position and
velocity of the load carriage. Embedded instrumentation include asphalt strain
gauges (PAST model sensors), pressure cells (SOPT model sensors), multi depth
deflectometers (MDDs), and moisture and temperature probes. These sensors were
chosen based on previous HVS owner’s experience (HVS 2015; Baker et al. 1994).
Additionally, the HVS was equipped with a laser profiler that can be used to create
a three-dimensional profile of the section and a Road Surface Deflectometer is
added to the testing equipment to obtain deflection basins at any location along the
test section (Leiva-Villacorta et al. 2013, 2015).
Figure 2 shows the instrumentation array used for the first series of experimental
sections. The PAST sensors were placed at the base/HMA layer interface in the
longitudinal or traffic loading direction and in the transverse direction. MDD sen-
sors were installed at four different depths to cover all four structural layers. The
thermocouples were placed at four depths: surface, middle depth of the AC layer, at
Development and Calibration of Permanent Deformation Models 577

30 cm 60 cm 90 cm

MDD MDD
Thermocouple

Section Length = 6.0 m

HMA
GB/CTB

Subbase
PAST
Subgrade Sensor

Fig. 2 Sensor array

the PAST sensors depth and 5 cm into the base layer. In the case of AC1 and AC4
sections the same gauge array was used while excluding PAST sensors.
Data collection of the 3D profile, strain, pressure, temperature and deflection is
performed based on load repetitions. At the beginning of the test, data is obtained at
short intervals. After 20,000 load repetitions, data is collected on daily basis.
Inspection of fatigue and reflective cracking, friction loss, loss of aggregate-asphalt
bond and any other surface damage is performed on daily basis during the HVS
daily maintenance work.

2 Permanent Deformation Models

For the asphalt concrete mixture the test was performed using unconfined cyclic
loading at three different temperatures: 46, 52 and 58 °C. Permanent deformation
was found to be a function of the resilient strain, temperature and number of loading
cycles. The model for permanent deformation of the asphalt concrete layer is given
in Eq. 1.
ep
¼ b1 T b2 N b3 ð1Þ
er

where,
εp: permanent deformation (mm/mm × 10−3)
εr: resilient deformation (mm/mm × 10−3)
N: number of load repetitions
T: Temperature (°C)
β1, β2, β3: regression coefficients
For the granular base and soil samples, the test was performed under repeated
axial cyclic stress at different axial loads and different confining stresses. Permanent
deformation was found to be a function of the confining stress, deviator stress,
578 F. Leiva-Villacorta et al.

moisture content, and number of loading cycles. The model for permanent defor-
mation of the unbound layers is given in Eq. 2. Each test was carried out for up to
5000 load cycles for 27 combinations of axial and confining stresses. The reason for
this choice mainly depended on having an analysis period compatible with labo-
ratory work constraints and a practical method to characterize the mechanical
behavior of these materials. Moreover, such an analysis period is in accordance
with the procedure based on the shakedown concept so the samples would not reach
the plastic creep limit (Werkmeister et al. 2003, 2005). The Plastic Creep stage is
achieved when the cumulative permanent strain rate is decreasing or constant.
Although the deformation is not totally resilient, the permanent deformation is
acceptable. In this case, the material could reach the failure after a very large
number of load cycles.
 b3  b4
b2 rd r3
pd ¼ b1  N    %w5 ð2Þ
q0 q0

where,
pd: permanent deformation, mm
N: number of load repetitions
σd: deviator stress (kg/cm2)
σ3: confining stress (kg/cm2)
ρ0: reference pressure (1 kg/cm2)
%w: moisture content (%)
β1, β2, β3, β4, β5: regression coefficients

3 Results

3.1 Development of Permanent Deformation Models

The laboratory-based model parameters obtained for the asphalt concrete, granular
base and subgrade materials are given in Table 2. It this case it was found that the
permanent deformation of the subgrade material was significantly affected by the
moisture content and this can be observed with the larger β5 coefficient compared to
the granular base. Regression analysis results are given in Table 3 for each model.

Table 2 Material parameters (laboratory)

Layer β1 β2 β3 β4 β5 R2
AC e−10.919 2.961 0.355 – – 0.835
Granular base 10−4.998 0.069 1.687 0.077 1.881 0.901
Subgrade 10−32.954 0.040 2.041 0.421 16.893 0.856
Development and Calibration of Permanent Deformation Models 579

Table 3 Models regression analyses

Layer Variable Coefficient Standard error T-stad. P value
AC Ln(β1) −10.919 0.098 −110.95 <0.05
Ln(β2) 2.961 0.024 124.27 <0.05
Ln(β3) 0.355 0.002 215.74 <0.05
Base Log(β1) −4.998 0.0160 −311.4 <0.05
Log(β2) 0.0685 0.000887 77.2 <0.05
Log(β3) 1.687 0.00440 383.4 <0.05
Log(β4) 0.0768 0.00352 21.8 <0.05
Log(β5) 1.881 0.0213 88.4 <0.05
Subgrade Log(β1) −33.0 0.0946 −348 <0.05
Log(β2) 0.0389 0.00101 38.6 <0.05
Log(β3) 2.04 0.00498 410 <0.05
Log(β4) 0.421 0.00741 56.8 <0.05
Log(β5) 17.0 0.0572 297 <0.05

30

25

20
p/ r

15 46 °C
52 °C
10 58 °C
5

0
0 200 400 600 800 1000 1200 1400 1600 1800
Load Repetitions

Fig. 3 Asphalt concrete permanent deformation test results

Figure 3 shows an example of the permanent deformation curves for the asphalt
concrete. Only the data at which tertiary flow begins were used to fit the model.
Figure 4 shows an example of the permanent deformation curves for the gran-
ular base at one stress combination and at different moisture contents. In particular,
for each test after the set confining pressure was reached, the axial stress with a
haversine wave frequency of 1 Hz was applied.
The test program, in terms of confining pressure σ3 and cyclic axial stress σd,
was selected according to European Standards (EN 13286-7 2004). In Table 4 a
detailed summary of the testing program is shown.
580 F. Leiva-Villacorta et al.

1.8
1.6
Permanent Deformation, mm

1.4
1.2
1
0.8
0.6
0.4
(Q1) %w=7,2 (Q2) %w=6,9 (Q6) %w=8,1
0.2
0
0 1000 2000 3000 4000 5000 6000
Load repetitions

Fig. 4 Granular base permanent deformation for σ3 = 40 kPa and σd = 60 kPa (Araya 2015)

Table 4 Unbound materials Granular material Subgrade material

test program
σd, kPa σ3, kPa σd, kPa σ3, kPa
60 40 40 30
80 60
120 80
80 69 60 40
138 80
205 100
155 103 75 50
206 100
309 150

3.2 Measured Permanent Deformation

Various types of data are collected during an HVS test. The most important data
from the viewpoint of developing permanent deformation models are the data
obtained from the Multi-Depth Deflectometer (MDD) system. MDD modules were
installed at predetermined depths in the pavement structures with a reference point
at the anchor.
Installation was done after pavement construction and the MDD modules were
placed at the layer interfaces, at mid depth of the base and subbase and near the road
surface. Two kinds of output were obtained from the MDD stack. Firstly, the
resilient deflection of each MDD module relative to the reference point at the
anchor is measured under the moving wheel load. This results were used to obtain
layer moduli by means of a backcalculation process. The second type of data
obtained from the MDD stack, is the permanent movement of each MDD module
relative to the reference point at the anchor for the duration of the HVS test. An
example of the permanent deformation data obtained in this study is illustrated in
Fig. 5. For this example, MDD modules were placed at the same depths for each
section and both sections are equal in thicknesses but section AC1 has a cement
Development and Calibration of Permanent Deformation Models 581

Permanent Deformation, mm

Permanent Deformation, mm
6 10
AC
AC 9 Base
CTB
5 Subbase 8 Subbase
Subgrade
Subgrade 7
4
6
3 5
4
2
3
1 2
1
0 0
0 200000 400000 600000 800000 1000000 0 200000 400000 600000 800000 1000000
Load Repetitions Load Repetitions
AC1 AC2

Fig. 5 Example of the measured permanent deformation

treated base (CTB) and AC2 has a granular base. The effect of the CTB layer with
respect to deformation can be observed. While section AC1 had lower deformations
overall, the layers underneath the CTB had almost the same level of deformation as
if the CTB layer was behaving as a concrete slab with equally stress distribution
along the test section. On the other hand, it can be observed that each layer of
section AC2 had different deformation levels with the highest at the top and the
lowest at the bottom.
The average measured deformation at eight million 40 kN equivalent axles for
section AC1 was 8.7 mm, 11.7 mm for section AC2, 5.1 mm for section AC3, and
2.1 mm for section AC4 (Fig. 6). These results followed the expected trend since
the section with the highest structural capacity was AC4 followed by AC3, then
AC1 and finally AC2. The permanent deformation curves exhibited in Fig. 6 were
generated from the HVS laser profiler. While this deformation considers the
maximum surface rut depth, the MDD rut depth is 25 mm into the asphalt concrete
and at the center of the wheel path. Therefore lower deformations were not only
expected but also obtained for the MDD stack.

14
Permanent deformation,

12
12.6

10 10.1 AC1
8
AC4
mm

AC2
6 6.125 AC3
4
2.57
2
0
0 2 4 6 8 10 12 14 16 18 20
MESALs

Fig. 6 Average measured permanent deformation

582 F. Leiva-Villacorta et al.

1
10000 HMA Base
Backcalculated Moduli, MPa
0.9
Subbase Subgrade

MDD Deflection, mm
0.8
0.7
1000
0.6
0.5
0.4
100 Measured MDD1-25
0.3 Predicted MDD1-25
0.2 Measured MDD1-250
Predicted MDD1-250
0.1 Measured MDD1-520
10 Predicted MDD1-520
0 1 2 3 4 0
0 1 2 3 4
Equivalent Load Repetitions (40 kN) Equivalent Load Repetirions (40 kN)

Fig. 7 MDD backcalculated layer moduli

3.3 Backcalculated Layer Moduli

MDD deflection data were used to determine the progression of the pavement layer
moduli. This was done by applying the method of equivalent thickness (Ullidtz
1987) whereby the thickness of the structure is transformed into a single layer. This
transformation is done using Odemark’s methodology and calculation of stresses,
strains and deflections were performed using Boussinesq theory.
Figure 7 shows an example of the backcalculated layer moduli for the different
layers for one of the test tracks as function of equivalent load repetitions in millions.
A good correlation (R2 > 0.8) obtained between measured and estimated deflec-
tions along with a small deviation from the equality were the criteria to perform
backcalculation and ensure that this methodology was successfully applied to each
particular data set.

3.4 Estimated Versus Measured Rutting

An example of the estimated and measured permanent deformation trend for each
layer is shown in Fig. 8. While measured rut depths show a typical variable trend
during the test, predicted rut depths follow a well defined trend. However, in all
cases, the model predictions did not fall within a established confidence interval of
the measured results; therefore a calibration process and adjustment of the model
coefficients was required. Only one MDD sensor was installed at the surface of the
asphalt layer providing four different data sets. For the granular base and subgrade
one or more MDD sensors were installed to capture the permanent deformation
providing at least six different data sets.
For asphalt concrete, the model underestimated the measured deformation at any
stage of the field test. For the granular base and subgrade materials the estimated
results were more scattered and the trend was to overestimate at the early stages
while throughout the end of the test the tendency was to do the opposite. This
Development and Calibration of Permanent Deformation Models 583

AC

Estimated Deformation, mm
Permanent Deformation, mm
6 14

5 Measured Estimated 12
AC
10
4 Equality
8
3
6
2
4
1 2
0 0
0 2 4 6 8 10 0 2 4 6 8 10 12 14
MESALs Measured Deformation, mm
Base

Estimated Deformation, mm
Permanent Deformation, mm

1.8 4
1.6
1.4 3 Base Equality
1.2
1
2
0.8
0.6
Measured Estimated 1
0.4
0.2
0 0
0 2 4 6 8 10 0 1 2 3 4
MESALs Measured Deformation, mm
Estimated Deformation, mm
Permanent Deformation, mm

SG 1.6 2
1.4 Subgrade
1.2 1.5 Equality
1
0.8 1
Measured Estimated
0.6
0.4 0.5
0.2
0 0
0 2 4 6 8 10 0 0.5 1 1.5 2
MESALs Measured Deformation, mm

(a) (b)

Fig. 8 Estimated versus measured permanent deformation, a section AC1, b all sections

behavior was an indication that the asphalt concrete model could be calibrated with
the adjustment of one coefficient while the remaining models could require a full
parameter calibration.
The cumulative distribution plot of the absolute relative errors for all
measured/estimated data points is given in Fig. 9. It can be observed that the
median relative errors for the AC, base and subgrade were 65 %, 60 % and 28 %
respectively. In addition, 90 % of the relative errors for the AC, base and subgrade
were below 90 %, 265 % and 180 % respectively. These large errors were expected
due to the high deviation of the estimated results with respect to the measured ones
(equality).
584 F. Leiva-Villacorta et al.

100
Cumulative Distribution

80

60
AC

40 Base
Subgrade

20

0
0% 50% 100% 150% 200% 250% 300% 350% 400% 450% 500%
Relative Error, %

Fig. 9 Cumulative distribution of the absolute relative errors before calibration

3.5 Calibration Process

By comparing the measured rut depths with the predicted values from laboratory
based models, the optimum combination of field calibration factors was determined
so that the coefficient of variation was minimal by means of OLS. The
laboratory-developed model coefficients were used to predict permanent deforma-
tion and the results either over-estimated or under-estimated this type of distress.
The resulting errors were either always positive or negative, meaning that they will
never cancel out, thus, contributing to the bias in the prediction model. To correct
these biases (systematic errors) in the model, the bias correction factors were
introduced (also known as the calibration coefficients).
The calibration process performed in this study started off by multiplying a
single correction factor (ki) to each individual regression coefficient (βi) and
completely adjust the initial model. As mentioned above, the process ended when
the coefficient of variation was minimal by means of OLS. Table 5 shows the
resulting calibration factors for each of the evaluated models. In the case of the
asphalt concrete model a single factor was applied. For the remaining models a full
calibration was required in order to correct the systematic errors in the model.
The cumulative distribution plot of the absolute relative errors for all
measured/estimated data points is given in Fig. 10. It can be observed that the
median relative errors for the AC, base and subgrade were after calibration 27 %,
9 % and 7 % respectively. In addition, 90 % of the relative errors for the AC, base
and subgrade were below 90 %, 100 % and 40 % respectively. These errors are
significantly lower than the results obtained before calibration. A further investi-
gation of this errors indicated that large errors were associated with low level rut
depths where a small change in the predicted value could result in errors up to
100 %. The final-calibrated models are shown in Eqs. 3–5.
Development and Calibration of Permanent Deformation Models 585

Table 5 Calibration factors

Layer k1 k2 k3 k4 k5
AC 16.256 – – – –
Granular base 1.048 3.841 0.608 1.911 0.873
Subgrade 0.346 5.728 −0.411 −0.631 0.393

100
Cumulative Distribution

80

60

AC
40
Base
Subgrade
20

0
0.0% 20.0% 40.0% 60.0% 80.0% 100.0% 120.0%
Relative Error, %

Fig. 10 Cumulative distribution of the absolute relative errors after calibration

ep ¼ 16:254  e10:919  T 2:961  N 0:355  er ð3Þ

pdGB ¼ 105:239  N 0:265  r1:025

d  r0:147
3  %w1:642 ð4Þ

pdSG ¼ 1011:424  N 0:229  r0:839

d  r0:266
3  %w6:677 ð5Þ

where,
εp: permanent deformation (mm/mm × 10−3)
εr: resilient deformation (mm/mm × 10−3)
N: number of load repetitions
T: Temperature (°C)
pd: permanent deformation, mm
σd: deviator stress (kg/cm2)
σ3: confining stress (kg/cm2)
ρ0: reference pressure (1 kg/cm2)
%w: moisture content (%)
586 F. Leiva-Villacorta et al.

4 Conclusions and Recommendations

Basic, conceptual models have been developed for the permanent deformation of
pavement layers from repeated triaxial load testing. The application of these models
to predict actual pavement deformation resulted in large errors which lead to a
calibration process. Calibration of the asphalt concrete permanent deformation
model was achieved with the use of a single factor at reasonable low errors. On the
other hand, calibration of the unbound materials required a full set of factors to
significantly decrease the error. However, this type of calibration provided minimal
coefficient of variation and lowest error by means of ordinary least squares (OLS).
The main source of data for developing these design models was the Heavy
Vehicle Simulator test data collected over 2 years period of time in Costa Rica for
one asphalt mixture type, one granular base and subbase type and one subgrade
type. Therefore, the application of these models is limited to the materials evaluated
until now.
This is the first attempt ever performed in Costa Rica to calibrate permanent
deformation models based on full scale accelerated test results. The process was
considered to be successfully applied to the collected data set. Therefore, this
process will be utilized and improved for future studies.
Because of the time involved in large scale accelerated pavement testing, it is
expected that a combination of laboratory testing and accelerated pavement testing
would yield the quickest results. Therefore, it should be possible to generate the
type of data needed to develop transfer functions for these materials from laboratory
test methods and then be verified by means of a limited number of full scale,
accelerated pavement tests.

References

Aguiar-Moya, J. P., Corrales, J. P., Elizondo, F., & Loría-Salazar, L. (2012). PaveLab and heavy
vehicle simulator implementation at the National Laboratory of Materials and Testing Models
of the University of Costa Rica. Advances in pavement design through full-scale accelerated
pavement testing, APT 2012.
Araya, Y. (2015). Desarrollo de modelos de deformación permanente para materiales granulares
y suelos. San José: Universidad de Costa Rica.
Baker, H. B., Buth, M. R., & Van Deusen, D. A. (1994). Minnesota road research project: Load
response instrumentation installation and testing procedures. Minnesota Department of
Transportation.
EN 13286-7. (2004). Unbound and hydraulically bound mixtures—Part 7: Cyclic load triaxial test
for unbound mixtures. European Committee for Standardization.
Guimaraes, A. C. (2009). Método mecanístico-empírico para la predicción de la deformación
permanente en suelos tropicales para pavimentos. Río de Janeiro: Universidad Federal de Río
de Janeiro.
Harvey, J., & Kanekanti, V. (2006). Calibration of mechanistic-empirical models using the
California heavy vehicle simulators. Transportation Association of Canada. Quebec:
International Society for Asphalt Pavements.
Development and Calibration of Permanent Deformation Models 587

Heavy Vehicle Simulator. (2015). Monitoring of test sections and instrumentation. Documento
consultado el 6 de abril del. http://www.gautrans-hvs.co.za/.
Leiva-Villacorta, F., Aguiar-Moya, J. P., & Loria-Salazar L. G. (2013). Ensayos acelerados de
pavimento en Costa Rica. Infraestructura Vial Vol. 15 Núm. 26.
Leiva-Villacorta, F., Aguiar-Moya, J. P., & Loria-Salazar, L. G. (2015). Accelerated pavement
testing first results at the Lanammeucr APT Facility. In Transportation Research Board 94th
annual meeting. Washington, DC.
Lekarp, F. (1997). Permanent deformation behaviour of unbound granular materials. Licentiate
Thesis. Kungliga Tekniska Högskolan.
NCHRP. (2004). Guide for mechanistic-empirical design of new and rehabilitated pavement
structures. National Cooperative Highway Research Program, Report 1-37A, March 2004.
http://www.trb.org/mepdg/guide.htm.
Powell, R. (2006). Predicting field performance on the NCAT pavement test track. Auburn:
Auburn University.
Ullidtz, P., et al. (2006). Calibration of incremental-recursive flexible damage models in CalME
using HVS experiments. Report prepared for the California Department of Transportation
(Caltrans) Division of Research and Innovation. University of California Pavement Research
Center, Davis and Berkeley.
Ullidtz, P. (1987). Pavement analysis. Development in civil engineering (Vol. 19). Amsterdam:
Elsevier.
Von Quintus, H. L., Darter, M. I., & Mallela, J. (2007). Recommended practice for local
calibration of the ME pavement design guide. Applied Research Associates, Inc.—
Transportation Sector.
Werkmeister, S., Dawson, A., & Wellner, F. (2003). Permanent deformation behavior of granular
materials and the shakedown concept. Journal of the Transportation Research Board, 1757,
75–81.
Werkmeister, S., Dawson, A., & Wellner, F. (2005). Permanent deformation behavior of granular
materials. Road Materials and Pavement, 6, 31–51.
Laboratory Tack Coat Investigation
of Factors Contributing to Debonding
Observed at the NCAT Test Track

Michael C. Vrtis and David H. Timm

Abstract Asphalt pavement structural integrity relies heavily on proper bonding at

lift interfaces. While it is intuitive that interface debonding leads to pavement
failure, the factors contributing to debonding in the field are not well understood.
The objective of this research was to further investigate layer interface debonding
that occurred in the 2012 research cycle at the NCAT Test Track, where debonding
was observed in several sections with relatively large amounts of recycled mate-
rials. Pavement construction and the tack coat application were closely monitored
and are not believed to have contributed to the debonding. Since debonding did not
occur immediately, it was suspected that environmental aging, moisture damage,
recycled binder content, and mixing temperatures may have been contributing
factors. This study investigated these factors by comparing the interface bond
strengths—measured with a shear collar in a Marshall stability tester, of
plant-mixed materials that were laboratory compacted—to field performance of the
full scale sections. The same non-tracking tack type and rate used in construction
was applied in the laboratory to fabricate specimens matching the field sections. To
investigate time and environmental effects, a set of specimens was subjected to
laboratory long term conditioning (AASHTO R30) and a set was subjected to
35 day aging. Specimens were then subjected to accelerated moisture damage
(ASTM D7870). Statistical analyses of the results from the sections allowed for
identification of debonding causal factors.

M.C. Vrtis (&)

Harbert Engineering Center, Auburn University, Auburn, AL 36849, USA
e-mail: [email protected]
D.H. Timm
Brasfield & Gorrie Professor of Civil Engineering, Harbert Engineering Center,
Auburn University, Auburn, AL 36849, USA
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 589

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_38
590 M.C. Vrtis and D.H. Timm

1 Background

Flexible pavements are constructed in multiple lifts and the bonds between lifts are
critical for the pavement to behave as a continuous system. Tack coats are used to
provide these bonds. Research has shown that loss of bonding between pavement
layers will reduce the life of the pavement system (Kruntcheva et al. 2005; Leng
et al. 2008) and therefore, sufficient interface bonding from the tack is imperative.
Debonding problems may result from construction errors, such as equipment
driving over a tacked surface that is waiting to be paved. However, debonding can
still occur even when construction practices and materials are closely monitored
(Willis and Timm 2007; Vrtis and Timm 2015).
Numerous laboratory studies have been undertaken to investigate tack coats. The
majority of these studies were aimed at finding the optimal tack application rate for
various surface types and tack materials (West et al. 2005; Mohammad et al. 2012).
Some previous studies have investigated the impact of moisture damage on bond
strength. Taylor used a vacuum saturation (similar to AASHTO T283 conditioning)
on laboratory-fabricated bond strength specimens to investigate bond resistance to
moisture damage (Taylor and Willis 2012). Zaniewski et al. also used accelerated
moisture damage on laboratory fabricated specimens to help identify factors
affecting bond strength in West Virginia (Zaniewski et al. 2015). Though earlier
studies have investigated bond strengths, replication of bonding conditions to
explain field performance was not evident in the literature and considered an area of
high need relative to understanding the importance of bonding.

2 Objective and Scope of Work

The objective of this research was to further investigate the observed field interface
bonding conditions from the 2012 National Center for Asphalt Technology
(NCAT) Test Track with laboratory testing to evaluate interface bond strength after
simulated aging and moisture damage. This was accomplished by using
plant-mixed samples collected during construction to create two-lift gyratory
specimens with a tack coat interface. The specimens were then split into four groups
for conditioning and testing. The first group was unconditioned and tested 3 days
after fabrication to establish baseline strength. The second group was also uncon-
ditioned but tested after 35 days to capture the strength change over time. The third
group was subjected to long-term oven aging, following AASHTO R30. Moisture
sensitivity was evaluated on the fourth group using the Moisture Induced
Susceptibility Test (MIST) following ASTM D7870.
Laboratory Tack Coat Investigation of Factors Contributing … 591

3 Field Sections and Performance

The NCAT Test Track is a 2.7 km (1.7 mile) closed-loop, flexible pavement testing
facility located in Opelika, Alabama. The Test Track is divided into 46 different
research sections. Of particular interest to this research was the 2012 Green Group
(GG) experiment. The objective of the GG was to evaluate performance of pavement
sections containing relatively large amounts of recycled materials (e.g., 30+%), such
as reclaimed asphalt pavement (RAP), recycled asphalt shingles (RAS), and ground
tire rubber (GTR) (not discussed in this paper). The GG sections were each inde-
pendently designed to optimize the amount of recycled material and performance. In
other words, more than one mix characteristic was changed from section to section.
During trafficking, layer interface debonding was observed in multiple sections.
Although the sections were designed to develop distresses within the 10 million
equivalent single axles loads (ESAL) applied, layer interface debonding was not an
intended failure mechanism and inhibited the evaluation of the entire pavement
system. Thus, as a result of the unanticipated debonding, further research was initiated
to compare mix properties of sections with debonding to those of sections without
debonding in an effort to identify factors that contributed to the observed debonding.
The four GG sections chosen for comparison were RAP, original high RAP,
RAP/RAS, and reconstructed high RAP. The as-built thicknesses and material
properties are displayed in Fig. 1. Each section was composed of 3 lifts with a total

Fig. 1 Average as-built thicknesses and materials

592 M.C. Vrtis and D.H. Timm

design thickness of 152 mm (6 in.) of asphalt concrete over pavement foundation

materials that have been used in previous Test Track research cycles. The aggregate
base was 152 mm (6 in.) of crushed granite from Columbus, Georgia often used by
the Alabama Department of Transportation. The subgrade has been classified as an
AASHTO A-4(0) metamorphic quartzite soil (Taylor and Timm 2009). Table 1
provides additional design and in-place properties. The High RAP sections and
RAP/RAS section had recycled binder ratios of approximately 0.44 in the inter-
mediate lift whereas the intermediate lift of the RAP section contained 0.35.
Warm-mix technology was used in all mixes; however, for the intermediate lift of
the Reconstructed High RAP section the plant temperature, 166 °C (330 °F), was
higher than warm mix temperatures despite the use of foamed asphalt.

Table 1 Design and in-place properties

Section Lifta NMAS Air voids, Asphalt Binder Plant
(mm)/ design/in-place content/effective ratio, total temperature
mix (%) asphalt content binder °C (°F)/
type (wt%) weight WMA type
RAP 1 9.5 4.0 5.2 0.17 132 (270)
SPb 8.4 4.7 Foam
2 19 4.0 4.6 0.35 141 (285)
SP 6.9 4.0 Foam
3 19 4.0 4.7 0.35 138 (280)
SP 6.5 4.1 Foam
Original high 1 9.5 4.0 5.8 0.16 135 (275)
RAP SMAc 6.9 5.3 Foam
2 19 4.0 5.0 0.44 138 (280)
SP 5.0 4.3 Foam
3 19 4.0 4.5 0.35 141 (285)
SP 6.3 3.9 Additive
RAP/RAS 1 9.5 4.0 5.5 0.2 135 (275)
SMA 7.9 5.1 Foam
2 19 4.0 4.9 0.43 138 (280)
SP 6.0 4.3 Additive
3 19 2.0 5.3 0.21 141 (285)
SP 3.5 4.8 Foam
Reconstructed 1 9.5 4.0 6.3 0.16 141 (285)
high RAP SMAd 6.5 5.8 Foam
2 19 4.0 5.0 0.44 165 (330)
SP 4.2 4.3 Foam
3 19 4.0 4.7 0.35 141 (285)
SP 7.2 4.1 Additive
a
Lift (1 surface, 2 intermediate, 3 base)
b
SP Superpave
c
SMA Stone Matrix Asphalt with 80 gyration Design
d
SMA Stone Matrix Asphalt with 50 blow Marshall Design
Laboratory Tack Coat Investigation of Factors Contributing … 593

All tack coat applications were calibrated and monitored during construction. No
tack coat issues were observed during construction and no discrepancies in the tack
application between sections were observed. A non-tracking tack, NTSS-1HM, was
applied at all interfaces at an undiluted rate of 0.227 L/m2 (0.050 gal/yd2). For the
Reconstructed High RAP section, the tack rate was increased to 0.364 L/m2
(0.08 gal/yd2) for the surface/intermediate lift interface and increased to 0.454 L/m2
(0.100 gal/yd2) for the intermediate/base interface.
These sections were selected for comparison because they exhibited an array of
field performance. A summary of the field performance is provided in Table 2.
The RAP, Original High RAP, and RAP/RAS sections were constructed during the
summer of 2012. Trafficking began on these sections in October 2012. In March
2013, cracking was observed in the Original High RAP section. The severity of the
cracking rapidly progressed and the predefined failure threshold of 20 % cracked
lane-area was exceeded within 2 weeks of the first crack being observed (after an
additional 200,000 ESALs). In Fig. 2a, photographed when the failure threshold
was reached, fine material that was pumped through the cracks can be seen in the
inside wheelpath. Forensic cores were then taken from the section and it was
apparent that debonding at the intermediate/base interface was leading to middle-up
cracking, as shown in Fig. 2b, c. In core 1 (furthest left), the interfaces between lifts
are distinguishable. In core 2, a crack can be seen developing at the interface of the
intermediate and base lift. In core 3, a crack begins to propagate towards the surface
from the lower interface and in core 4, the crack reaches the surface. In core 5 the
cracking has extended through the base lift resulting in a fully cracked structure.
At this point, a brief laboratory tack investigation was conducted and several
changes to the original design were implemented for reconstruction to ensure that
layer debonding did not reoccur. As mentioned previously, the tack rate and
plant-mixing temperatures were increased for the reconstructed high RAP section
(Vrtis and Timm 2015). The reconstructed high RAP section was opened to traffic
on May 23, 2013 and a crack was not observed in this section until after an
additional 4.2 million ESALs were applied.
The RAP/RAS section also showed signs of debonding at the intermediate/base
interface but the distress progression time from the first observed crack to the failure
threshold was roughly 10 times longer than the Original High RAP section.
Bottom-up fatigue cracking was identified as the failure mechanism in the RAP
section and no signs of debonding were observed.

Table 2 Green group field performance comparison

Section ESALs until 1st ESALs until failure Interface
crack threshold debondeda
RAP 2,461,914 4,070,349 NA
Original high RAP 2,222,506 2,439,373 2/3
RAP/RAS 3,412,759 5,721,386 2/3
Reconstructed high 4,233,868 NA NA
RAP
a
Interface Debonded (1 surface, 2 intermediate, 3 base)
594 M.C. Vrtis and D.H. Timm

(a) Original High RAP section at failure threshold

(b) Cores taken from Original High RAP section

(c) Cores taken from Original High RAP section with cracks highlighted

Fig. 2 Photographs of original high RAP section at time of failure

4 Sample Preparation and Conditioning

Plant-mixed samples collected during construction were used to create the bond
strength testing specimens. Specimen fabrication began by making 76.2 mm (3 in.)
tall Superpave Gyratory specimens from the base lift with air voids of 7.5 %
(±0.5 %). The tack material was then applied to the top surface at a rate of
0.227 L/m2 (0.050 gal/yd2). The tack was applied with a syringe and spread evenly
across the entire surface with a paintbrush. A balance was used to control the
amount of tack applied to each specimen. The same tack rate was applied to all
Laboratory Tack Coat Investigation of Factors Contributing … 595

specimens, even though the Reconstructed High RAP section utilized a tack rate of
0.454 L/m2 (0.100 gal/yd2) in the field. The tack rate was kept the same for all
sections to help determine if increasing the tack rate led to the improved perfor-
mance. The tack was allowed to break for at least 24 h before the specimen was
placed back into a gyratory mold and 76.2 mm (3 in.) of intermediate lift material
was compacted on top of the tack coat and base lift specimen. The intermediate lift
also had target air voids of 7.5 % (±0.5).
The specimens were then randomly split into four different treatment groups.
The “3 day” group was tested 3 days after specimen fabrication to establish the
unconditioned and unaged interface shear bond strength. The “35 day” group was
aged inside the laboratory for 35 days prior to bond strength testing to evaluate the
impact of time without any extreme temperature or moisture conditions. The “R30”
group specimens were subjected to long-term laboratory aging following AASHTO
R30-02, which conditions specimens in an oven at 85 °C (185 °F) for 120 h.
The “D7870” group was conditioned in the MIST device following ASTM
D7870. The MIST cycles hydrostatic pore pressure to force water in and out of the
specimen to evaluate moisture susceptibility. The MIST has been successfully used
to evaluate moisture susceptibility (Buchanan et al. 2004) and has been recom-
mended as a replacement for the highly variable moisture susceptibility evaluation
procedure in AASHTO T283 using the tensile strength ratio (Schram 2012).

5 Bond Strength Testing

The interface bond strength was evaluated using ALDOT-430 Standard Test
Method for Determining the Bond Strength Between Layers of an Asphalt
Pavement. The loading configuration is shown in Fig. 3a and a standard Marshall
Stability Tester was used to apply the displacement controlled load at
50.8 mm/min. (2 in./min). This loading scheme utilizes a guillotine-like collar
(Fig. 3b) to apply a shear load at the lift interface. All shear testing was conducted
at room temperature. The interface shear strength was determined by dividing the
peak load over the cross-sectional area.

6 Results and Statistical Analysis

The average bond shear strengths for each group are shown in Table 3 along with
the standard deviation, coefficient of variation and the percent change from the 3
day strength. All data were scrutinized for outliers following ASTM E 178
Standard Practice for Dealing with Outliers and no outliers were found. An
analysis of variance (ANOVA) is shown in Table 4. The two-way ANOVA shows
that both the sections and treatments uniquely affected the bond strength. In other
596 M.C. Vrtis and D.H. Timm

(a) Loading scheme for shear bond strength testing (West et al, 2005)

(b) Shear collar (c) Shear collar in testing press

Fig. 3 Shear bond strength testing apparatus

words, each treatment had a unique impact on the bond strength and each group
responded differently to the treatments.
The RAP section had the highest overall bond strength and smallest coefficients
of variation. The reconstructed high RAP section also had relatively low coeffi-
cients of variation. The D7870 group of the original high RAP section had the
lowest bond strength and highest coefficients of variation.
Long-term laboratory aging (R30) increased the bond strength from the 3 day
strength in all sections. MIST conditioning (D7870) decreased the bond strength in
all sections. The MIST device was successful in producing statistically significant
results to evaluate moisture damage, as observed by Zanieweski. The 35 day aging
treatment had the least impact of the treatments.
Laboratory Tack Coat Investigation of Factors Contributing … 597

Table 3 Shear testing results

Section Treatment Average Standard Coefficient Percent of 3
bond deviation, of variation day
strength, kPa kPa (psi) (%) strength
(psi) (%)
RAP 35 Day (361.1) (12.4) 3.4 90
D7870 (298.0) (8.2) 2.8 75
R30 (428.1) (12.6) 2.9 107
Original high 3 Day (341.4) (30.7) 9.0 100
RAP 35 Day (276.2) (20.3) 7.4 81
D7870 (168.8) (42.3) 25.0 49
R30 (401.7) (22.1) 5.5 118
RAP/RAS 3 Day (311.9) (56.5) 18.1 100
35 Day (321.3) (12.2) 3.8 103
D7870 (250.8) (15.5) 6.2 80
R30 (388.5) (13.3) 3.4 125
Reconstructed 3 Day (278.0) (20.9) 7.5 100
high RAP 35 Day (308.8) (9.5) 3.1 111
D7870 (253.3) (8.1) 3.2 91
R30 (363.8) (21.7) 6.0 131

Table 4 Shear testing results analysis of variance

Source of variation SS df MS F P value F crit
Section 42,463 3 14,154.5 22.820 4.34E−08 2.90
Treatment 138,343 3 46,114.4 74.348 1.64E−14 2.90
Interaction 22,226 9 2469.62 3.981 0.00173 2.188
Within 19,847 32 620.248
Total 222,881.5 47

The original high RAP with MIST conditioning had the greatest reduction from
the 3 day strength. This group retained only 49 % of its 3 day bond strength after
being subjected to the MIST. This indicates that the original high RAP section was
more susceptible to moisture damage than the other sections. Due to the high
moisture susceptibility it is likely that moisture damage contributed to the rapid
failure of the original high RAP section shortly after the first crack was observed.
With such a large deviation from the 3 day strength relative to the other specimens,
it is reasonable to assume that if this testing was done prior to construction this
result would stand out and would have been investigated prior to construction.
Further comparison of the shear strength results (Table 3) with the field per-
formance results (Table 2) does not yield any additional clarity. The RAP and
reconstructed high RAP sections, the two sections that did not debond in the field,
had the highest and lowest average bond strengths, respectively. Thus, a high or
low bond strength value is not directly indicative of debonding performance.
598 M.C. Vrtis and D.H. Timm

It must be pointed out that all specimens tested in this project exceeded the mini-
mum threshold used by ALDOT of 689 kPa (100 psi).

7 Conclusions and Recommendations

Plant-mixed samples collected during original construction at the NCAT Test Track
were used in the laboratory to create multi-lift gyratory compacted specimens that
were subjected to a variety of treatments. After treatment, the interface bond
strength was measured and strength results were compared to field performance.
Based on the work presented in this paper the following conclusions and recom-
mendations can be made:
• Careful application of tack to gyratory specimens in the laboratory produced
consistent specimens in each group.
• The MIST device (following ASTM D7870) was successful in creating statis-
tically significant moisture damage in the specimens.
• The original high RAP section was identified as being highly susceptible to
moisture damage which confirmed the observed poor field performance.
Moisture damage can drastically reduce the bond strength. Further research should
investigate whether the MIST device can be used to predict future debonding.

References

Buchanan, M. S., Moore, V., Mallick, R., O’Brien, S., & Regimand, A. (2004). Accelerated
moisture susceptibility testing of hot mix asphalt (HMA) mixes. In 83rd Transportation
research board annual meeting, Washington, DC.
Kruntcheva, M. R., Collop, A. C., & Thom, N. H. (2005). Effect of bond condition on flexible
pavement performance. Journal of Transportation Engineering, 131(11), 880–888.
Leng, Z., Ozar, H., Al-Qadi, I. L., & Carpenter, S. H. (2008). Interface bonding between hot-mix
asphalt and various Portland cement concrete surfaces: Laboratory asessment. Transportation
Research Record: Journal of Transportation Research Board, 2057, 46–53.
Mohammad, L., Elseifi, M. A., Bae, A., Patel, N., Button, J., & Scherocman, J. A. (2012). NCHRP
report 712: Optimization of tack coat for HMA placement. Washington, DC: The National
Academies Press.
Schram, S. (2012). Ranking of HMA moisture sensitivity tests in Iowa. Report RB00-012, Iowa
Department of Transportation.
Taylor, A. J., & Timm, D. H. (2009). Mechanistic characterization of resilient moduli for unbound
pavement layer materials. Report no. 09-06, National Center for Asphalt Technology, Auburn
University.
Taylor, A. J., & Willis, J. R. (2012). Effects of Nanotac additive on bond strength and moisture
resistance of tack coats. Report 12-02, National Center for Asphalt Technology, Auburn
University.
Vrtis, M. C., & Timm, D. H. (2015). Case study on premature pavement failure and successful
reconstruction of a high RAP section at the NCAT test track. In International airfield and
highway pavements conference 2015, Miami, FL.
Laboratory Tack Coat Investigation of Factors Contributing … 599

West, R. C., Zhang, J., & Moore, J. (2005). Evaluation of bond strength between pavement layers.
Report 05-08, National Center for Asphalt Technology, Auburn University.
Willis, J. R., & Timm, D. H. (2007). Forensic investigation of debonding in rich-bottom pavement.
Transportation Research Record: Journal of Transportation Research Board, 2040, 107–114.
Zaniewski, J. P., Knihtila, S. F., & Rashidi, H. N. (2015). Evaluation of the bond strength of asphalt
overlays. In International airfield and highway pavements conference 2015, Miami, FL.
Research Partnership Between
the National Center for Asphalt
Technology (NCAT) and the Minnesota
Department of Transportation MnROAD
Facility for a Nationwide Pavement
Performance Experiment

R. Buzz Powell and Ben Worel

Abstract Full-scale accelerated pavement testing (APT) has been conducted at the
National Center for Asphalt Technology (NCAT) Pavement Test Track since 2000.
Practical research in surface mix performance, structural pavement design, and
pavement preservation has been cooperatively funded in 3-year research cycles by
state departments of transportation (DOTs) located primarily in the southeastern
United States (US). The Minnesota DOT’s Road Research Facility (MnROAD) has
been conducting research in these same focus areas for northern states since 1994.
Although the two programs have collaborated informally for many years, no formal
experiments have been executed that simultaneously address the needs of both
northern and southern states. The 6th research cycle at the NCAT Pavement Test
Track and the 3rd phase of MnROAD research are for the first time officially
connected in a nationwide study through two group experiments that are cooper-
atively funded by numerous state DOTs from all over the country. The objective of
the preservation group (PG15) experiment is to quantify the benefit of pavement
preservation on both low volume and high volume roadways. The objective of the
cracking group (CG15) experiment is to identify laboratory tests that accurately
predict both top-down and thermal cracking in mixes produced with diverse
materials and a broad range of expected cracking susceptibilities. Construction of
experimental pavements and surface treatments for these two experiments are
planned for 2015 and 2016. An overview of the experiment design, constituent

R. Buzz Powell (&)

National Center for Asphalt Technology, 277 Technology Parkway,
Auburn, AL 36830, USA
e-mail: [email protected]
B. Worel
Minnesota Department of Transportation, Office of Materials and Road Research,
1400 Gervais Ave, Maplewood, MN 55109, USA
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 601

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_39
602 R. Buzz Powell and B. Worel

materials, mix designs, construction quality, baseline pavement condition mea-

surements, and analysis methodologies are summarized in this document.

Keywords Accelerated pavement testing
Pavement preservation
Top-down

cracking Thermal cracking

1 Background

The National Center for Asphalt Technology (NCAT) and the Minnesota
Department of Transportation (MnDOT) through its MnROAD facility have for-
mally partnered for the first time to execute pavement preservation and asphalt
mixture performance testing experiments with a nationwide implementation impact.
Positive experiences with implementable findings that reduce the life cycle costs of
pavements and facilitate rapid deployment of sustainable technologies have made
past research at both NCAT and MnROAD an outstanding investment for numerous
state DOTs. The partnership is expected to synergize the yield and broaden the scope
of implementable findings, while at the same time expand the capabilities of both
facilities on complementary research through close collaboration.
Recognizing that improved pavement performance is an immediate nationwide
goal, MnROAD and NCAT are focused on facilitating the validation of promising
technologies through accelerated pavement testing. NCAT and MnROAD are
full-scale test tracks that use real-world pavement construction, full-scale truck and
live interstate traffic (respectively), all under actual climate conditions that effect
pavement performance (MnROAD 2015; NCAT 2015). The combination of traffic
loading types and the range in climate conditions provide unique opportunities to
address pavement performance issues.
A closer relationship between NCAT and MnROAD is a logical progression in
developing and evaluating new sustainable technologies, pavement systems and
construction methods that lead to safer, quieter, lower cost and longer-lasting roads.
Both organizations have core staff members with extensive experience in the
operation and analysis of full-scale accelerated pavement testing. Furthermore, staff
is continually engaged in technical committees and conferences related to pavement
testing, analysis and performance, forming relationships with local, national and
international pavement engineering experts (AFD40 2015). This provides access to
breakthrough findings and technologies that can be utilized to conduct more
effective and efficient research.
Both test tracks monitor pavement performance routinely for rutting, cracking,
roughness, texture, friction, noise, and structural capacity. All test sections have
instrumentation to measure dynamic vehicle load and/or environmental response.
The data generated from the physical measurements and instrumentation is used for
the development of pavement response models and, ultimately, improved
mechanistic-empirical design procedures.
Research Partnership Between the National Center for Asphalt ... 603

1.1 NCAT Facility

The NCAT Pavement Test Track was originally constructed as a result of interest
and support from state Departments of Transportation (DOTs) who shared a con-
cern for building and maintaining safe and cost effective pavement infrastructure.
Forty-six 200-ft test sections are subjected to 10 million equivalent single axle
loadings (ESALs) of heavy truck traffic compressed into two years of fleet opera-
tions on the 1.7-mile closed oval. Test pavements are rebuilt every three years to
facilitate a new research cycle in which sponsors can focus on mill/inlay research on
perpetual pavements, structural research, or pavement preservation (TPF 2015). The
main objectives of research within these focus areas are to help state DOTs
implement positive change and to promote real innovation (Picture 1).

1.2 MnROAD Facility

MnROAD is owned and operated by the Minnesota Department of Transportation

in Albertville, Minnesota. Established in 1994, the test track consists of a 3.5 mile
Interstate (I-94) roadway and a 2.5-mile closed loop low volume road
(LVR) simulating rural roads. The Interstate diverts live traffic for its loadings and
the LVR uses a fully loaded 80 K 5-axle semi driven five days a week providing on
average 60 laps a day. MnROAD has 50 unique test sections and as test cells fail
others are built to replace them. Research is sponsored by state DOTs, Minnesota

Picture 1 NCAT pavement test track

604 R. Buzz Powell and B. Worel

Picture 2 MnROAD pavement test track

Local Road Research Board, FWHA and industry. MnROAD is in its 2nd phase of
research and developing its 3rd phase which includes this partnership but also will
also include a separate partnership (National Road Research Alliance/NRRA) that
will focus on other national research needs other than national pavement preser-
vation and development of a HMA performance test which is the basis for
MnROAD’s partnership with NCAT (NRRA 2015) (Picture 2).

2 Partnership Development

The general focus of research at both facilities has historically been rapid imple-
mentation of findings that reduce the cost of safe and sustainable transportation
infrastructure for state DOTs. Although materials are transported great distances to
optimize the relevance of research outcomes, ambient climate has been a limiting
factor in growing the cooperative partnerships that fund construction and operations
at both locations. Generally hot weather in Alabama has attracted participation at
the NCAT Pavement Test Track from primarily southern state DOTs, while gen-
erally cold weather in Minnesota has attracted participation at MnROAD from
generally northern state DOTs. Informal technical collaboration through the AFD40
Full-Scale Accelerated Pavement Testing (APT) Committee of the Transportation
Research Board (TRB) has been helpful for both research programs, who have
actively promoted each other to state DOTs with research needs that are more
suitable to a different climate and/or pavement type. This regional client model has
sustained important research at both facilities through multiple research cycles;
however, a multi-climate synergy possible through deliberately coordinated
research has not been achieved.
The 2014 calendar year presented a unique opportunity for a paradigm shift in
both research programs. An integral part of the planning process for future work at
Research Partnership Between the National Center for Asphalt ... 605

the NCAT Pavement Test Track is close interaction with existing DOT sponsors to
identify their pressing research needs. This effort was undertaken in 2014 in order to
plan for the 6th (2015) research cycle. Two research needs quickly rose to the top of
the list of priorities that were good candidates for cooperatively funded “group
experiments” in which state DOTs pool resources to achieve common goals.
The first group experiment was a continuation and expansion of the pavement
preservation study started in the 5th (2012) research cycle at NCAT. Extended data
collection was necessary in order to continue to quantify the life extending and
condition improving benefits of various standalone and combination pavement
preservation treatments placed on nearby Lee Road 159 in the summer of 2012.
Expansion was necessary in order to increase the scope of the study from a low
volume county road that provides one-way access to an aggregate quarry and hot-mix
asphalt plant with limited load spectra to a high volume roadway with diverse traffic.
The second group experiment was needed to identify a laboratory cracking test
(s) that could be used to screen asphalt mixes for near-surface cracking potential for
both mix design approval and production quality control purposes. Interest in uti-
lizing as much recycled and reclaimed material as possible has forced this issue to
the forefront for state DOTs who want to take advantage of potential savings and
sustainability advantages without compromising pavement durability. As the
experiment design process evolved, the national scope of these research needs for
both southern (hot) and northern (cold) climates became evident.
Concurrently, MnROAD was engaged in a similar process to prepare for its third
research cycle. Similar to NCAT, the first two research cycles at MnROAD had
addressed both pavement preservation and asphalt pavement cracking with a focus
on regional solutions. Both facilities needed to respond to the growing demand for
multi-climate research in these areas, but in a manner that did not duplicate each
other’s ongoing efforts. The obvious solution for the informal collaborators was to
engage in a formal partnership to execute research that addresses the pressing
national needs in both pavement preservation and asphalt mix cracking perfor-
mance. Initial discussions between the two facilities were encouraging, and the
process of developing a mutually beneficial formal partnership was initiated.
Within a series of meetings in 2014 in which researchers visited each facility to
brainstorm the possibilities, the nature and future of the partnership was developed.
It was decided that the partnership would be launched (i.e., planning, design,
construction, and performance in the near term) within the framework of the
Transportation Pooled Fund (TPF) that supports the 2015 NCAT Pavement Test
Track (http://www.pooledfund.org/Details/Study/496). This effort is funded
through the end of September in 2018, with the majority of research expenditures
paid out through the end of February of the same year. In March of 2018, funding
for long term monitoring of surviving pavement preservation test sections at both
(north and south) locations will be supported through a new TPF hosted by the
Minnesota DOT. Funding for non-pavement preservation research specific to each
facility will be supported by other means in 2018 (e.g., a new pooled to support
research exclusively on the NCAT Pavement Test Track will be hosted by the
Alabama DOT). With a framework for the formal agreement between the north and
606 R. Buzz Powell and B. Worel

south components of a nationwide experiment in place, the planning process for the
two studies continued.
The primary objectives of the formalized partnership are:
• Constructing CG15 test sections on the existing NCAT and MnROAD facilities
and add an additional off-Track PG15 high ADT roadway in Alabama to go
along with the existing Lee Road 159 low ADT roadway. Additionally, PG15
test sections will be constructed in Minnesota on off-track low and high ADT
roadways;
• Applying truck traffic to the NCAT test oval for approximately two years in a
highly controlled and precise manner and at the MnROAD facility where public
interstate I-94 traffic public will load the CG15 test sections. Additionally, traffic
will be carefully documented on off-Track sections both in Alabama and
Minnesota for the length of the PG15 study;
• Assessing/comparing the functional and structural field performance of traf-
ficked sections on a regular basis via surface and subsurface measures;
• Validating/calibrating new and existing M-E approaches to pavement analysis
and design using pavement surface condition, pavement load response, precise
traffic and environmental logging, and cumulative damage;
• Quantifying the life extending benefit of various pavement preservation alter-
natives as a function of pretreatment condition in a highly controlled experiment
that will provide state DOTs with the objective foundation to implement a
decision tree for their own maintenance program. These types of programs are
then refined over time using actual pavement management performance data;
• Correlating field performance with laboratory results and developing rational
test criteria for possible specification use; and
• Answering practical questions posed by research sponsors through formal (i.e.,
reports and technical papers) and informal (e.g., one-on-one responses to
sponsor inquiries) technology transfer. For example, can the same laboratory
tests be used to screen both virgin and high binder replacement (i.e., higher RAP
and RAS) mixes for cracking susceptibility for both mix design approval and
quality control test applications?
The scope of work required to meet these objectives includes:
• NCAT will haul materials to the project from offsite locations. Material dona-
tions are typically secured by state sponsors, while reasonable hauling expenses
are handled by the pooled fund. MnDOT is planning to use locally found
aggregates that represent the northern sponsors and will also partner with
industry for other maternal when it supports the study goals;
• Rebuilding test sections in accordance with sponsors’ directives via competi-
tively bid subcontracts administered by NCAT. It is anticipated that equipment
rental, aggregate hauling, liquid asphalt supply and delivery, plant production,
and mix placement may all be procured via competitively bid subcontracts.
Additional preservation treatment sections will be applied to nearby highways as
well as at the PG15 Minnesota test sections. MnROAD CG15 test sections are
Research Partnership Between the National Center for Asphalt ... 607

going to be rebuild using traditional design-bid-build with the state possibly

supplying the aggregate and binder to the contractor;
• Installing both environmental (i.e., multi-depth pavement temperature probes)
and response instrumentation (e.g., high speed stress and strain gages) in new
experimental sections;
• Operating a 5-truck heavy triple-trailer fleet on the NCAT test oval in order to
apply accelerated truck traffic following the completion of construction. Actual
human drivers operate the vehicles in order to provide realistic vehicle wander
as MnROAD will utilize public I-94 traffic;
• Measuring field performance to fully document the changes in surface condition
as a function of traffic and temperature. High-speed pavement response will also
be measured in the structural sections. Pavement performance will also be
documented though a number of tests for both the test tracks and the off-Track
pavement preservation sections;
• Conducting laboratory testing to quantify basic material and mix performance
properties, which will serve as the basis of performance model development;
and
• Comparing predicted and measured pavement response as well as predicted and
measured cumulative pavement damage in order to validate then calibrate
prevailing M-E methodologies and to quantify the life extending benefit of
various pavement preservation alternatives (Fig. 1).

Fig. 1 Current agency partners for the NCAT/MnROAD partnership

608 R. Buzz Powell and B. Worel

3 National Pavement Preservation Study

Pavement preservation techniques are very cost-effective when applied to the right
road at the right time, with benefit-cost ratios as high as 10:1 (APT 2013).
Accelerated testing at MnROAD and NCAT provides unique opportunities to
determine the field performance of breakthrough materials and pavement preser-
vation concepts without the risk of failure that local and state agencies are
unwilling/unable to accept. Each facility has a history of evaluating the perfor-
mance of pavement preservation treatments, including chip sealing,
micro-surfacing, crack sealing and thin overlays. To address needs in both northern
and southern climates, similar test sections are being developed at each facility to
address national issues. Recognizing the long-term nature of measuring the impact
of some pavement preservation treatments, both MnROAD and NCAT are pursuing
off-site test locations on existing roads and highways that can be easily monitored.
The Preservation Group (PG15) study is designed to encompass timely issues
that are important to the pavement community. Five sections from the 2009 Group
Experiment survived the previous two traffic cycles (NCAT 2012). Select preser-
vation treatments were installed on the NCAT Pavement Test Track in the spring of
2014 when predetermined distress levels were reached. Funding for the PG15 study
will provide continued traffic and monitoring of these sections in order to fully
quantify the life extending benefit of the pavement preservation treatments.
NCAT also placed 23 preservation treatments on nearby Lee Road 159 in the
summer of 2012 (PPJ, 2012). This county road provides dead end access to a quarry
and an asphalt plant. Unlike the NCAT Pavement Test Track, where significant
resources are invested in precisely controlled accelerated truck traffic, test sections
on Lee Road 159 are exposed to public traffic related to the quarry, the asphalt
plant, and a handful of local residents. A detailed record of the truck traffic is
provided continuously via data sharing agreements with the quarry and the asphalt
contractor. Regular testing by NCAT documents roughness, rutting, macrotexture,
cracking, and structural integrity of each pavement preservation section as a
function of traffic, age, and seasonal effects. As part of the PG15 study, the research
on Lee Road 159 will be extended into the new 2015 research cycle (Picture 3).
Additionally, new pavement preservation treatment sections were built in the
summer of 2015 on nearby US-280, which is a higher average daily traffic (ADT),
more diverse load spectra route. The purpose of the new sections on US-280 is to
quantify the life extending and condition improving benefits of pavement preser-
vation on a more typical highway in between the accelerated damage environment
of the NCAT Pavement Test Track and the low ADT environment of Lee Road
159. The treatment sections are 0.1 miles in length and include treatments and
treatment combinations such as:
• Chip seals (single, double, and triple layer);
• Scrub seal;
• Micro surface;
• Cape seals (conventional, HMA, FiberMat, and scrub);
Research Partnership Between the National Center for Asphalt ... 609

Picture 3 Lee road 159 test sections (PG15)

Picture 4 US-280 test sections (PG15)

• Thin asphalt overlays (virgin and mixes with RAP and RAS);
• CIR (foam and emulsion) and CCPR (foam and emulsion) under thin overlays;
and
• Fog seals (with and without rejuvenators) (Picture 4).
610 R. Buzz Powell and B. Worel

Fig. 2 Location of Minnesota and Alabama test sections

Preservation treatments on US-280 were selected by the PG15 sponsors.

Consensus treatments will also be selected by sponsors for placement in Minnesota
on off-track high and low volume roadways in Mille Lacs County. This is about
45 min north of MnROAD (Fig. 2).
Results from the expanded PG15 experiment will provide a rational starting
point for the implementation of a life cycle cost-based preservation treatment
selection process that can be refined over time with location-specific pavement
performance feedback. The other objective of the PG15 experiment is to develop
guide specifications and recommend guidelines for quality assurance
testing/inspection of pavement preservation treatments.

4 Development of a National HMA Performance Test

NCAT and MnROAD have long histories advancing asphalt technologies. Both
have been involved in comprehensive laboratory testing and analyses, the devel-
opment of new asphalt testing methods and field validation of laboratory findings
with full-scale accelerated pavement testing (MnROAD 2015; NCAT 2015).
Through this partnership, future asphalt technologies will be developed more effi-
ciently over a wide range of climate and traffic factors.
The Cracking Group (CG15) is the second focus area in the 2015 research cycle
for both NCAT and MnROAD. The objective of this experiment is to validate
laboratory cracking tests by establishing correlations between the test results and
measured cracking in real pavements (test sections). This experiment will require
Research Partnership Between the National Center for Asphalt ... 611

the construction and trafficking of a series of new test sections with asphalt mixtures
that have a range of expected cracking susceptibilities. All other factors that can
impact cracking, such as layer thicknesses and underlying layer stiffness will be
controlled to the highest degree possible. A battery of laboratory cracking tests will
be conducted on mixes to identify which test results best correlate with field
cracking. The following tests have been suggested for inclusion in the CG15 study,
with the actual methods determined by consensus among participating agencies:
• Indirect tensile work;
• Bending beam fatigue;
• Simplified viscoelastic continuum damage;
• Disc-shaped compact tension tests;
• Semicircular bend test; and
• Texas overlay tester.
The analysis of the laboratory cracking tests will also consider its variability,
utility and practicality of implementation for both mix design approval and quality
control testing. The following surface pavements were built on the 2015 NCAT
Pavement Test Track on a relatively thin (6 in. thick), highly flexible, and fatigue
resistant structure (built with base and binder mix produced with highly polymer
modified binder, or HiMA) with the expectation that top-down cracking results
would vary significantly under accelerated truck traffic at NCAT:
• 20 % RAP control in section N1 (i.e., the first section on the north tangent);
• High density control variant in section N2;
• Low binder content and low density control variant in section N5;
• Control +5 % RAS variant in section N8;
• Control +15 % RAP with PG58 variant in section S5;
• Control with HiMA variant in section S6; and
• 15 % RAP Arizona rubber in section S13.
MnROAD is planning to place eight 500’ test sections in 2016 to support the
CG15 study. These eight test sections will consist of 6 in. layer of HMA, 29 in. of
granular base, over a clay subgrade. Agency sponsors are currently focusing on
developing mixes that will help demonstrate a lab performance test that will predict
future cracking in the field. The northern sponsors are focusing on low temperature
cracking. Mixes are being developed that are expected to have either a low or high
cracking potential and also focus on current key mixture issues facing each agencies
which matches nicely with the southern sponsors. Some of these issues include use
of highly modified asphalt, percentage of recycled asphalt pavements with and
without shingles, crack prone gradations, low asphalt and density mixes. The
sponsors also want to have a control mix to compare the mixes that will be used.
Table 1 shows the current mixes that are being considered at MnROAD in 2016.
The focus on low temperature cracking has been a focus area for research for a
number of years for most of the northern sponsors of this study. Figure 3
demonstrates the effects of pre-PG graded binders have on cracking performance. It
612 R. Buzz Powell and B. Worel

Table 1 MnROAD CG15 possible mix designs

Mix description Binder Polymer Crack
potential
Typical surface mix (20 % max RAP) PG 58-34 Yes Low
Typical surface mix (20 % max RAP) PG 64-22 No Med/high
Mix w/20 % FRAP + 3 % RAS PG 64-28 Yes Med/high
Mix w/30 % FRAP + 5 % RAS PG 64-22 No High
Mix w/highly mod. asphalt PG 76-34 + perf reqs Yes Low
(<15 % RAP)
Typical surface mix (20 % max RAP, silo PG 64-22 No High
aging)
Mix w/cracking prone gradation PG 64-34 Yes Med/high
Mix w/9.5 mm aggregate PG 64-22 No Med/high
Mix w/ > 30 % RAP and softer AC PG 52-34 with high ? Med/high
RAP?
Typical surface mix w/high density TBD
Typical surface mix w/low AC, low
density
Sponsor recommendation

Fig. 3 MnROAD phase-I thermal cracking

shows test sections at MnROAD where two PG graded binders were placed (PG
58-28 and PG64-22) both on “thinner” HMA (5-year) and “thicker” HMA (10-year)
test sections. After a cold spell in February 1996 where the air temperature reached
−39 °C, all the test sections cracked and the linear feet of cracking remained
constant over the remaining years. Only the severity of the cracking changed.
MnROAD learned that thickness did not impact thermal cracking and the PG
Research Partnership Between the National Center for Asphalt ... 613

Fig. 4 Fracture energy versus measured cracking from MnROAD and agency test sections

grading system worked but there was more going on than just the asphalt binder to
control cracking. The rest of the mix components also played a role in cracking
performance.
MnDOT has since lead two pooled fund studies that continued to investigate the
effects the asphalt mix has on cracking potential. These two studies used MnROAD
test sections and a number of agency provided test sections and determined that
asphalt binder grade, modifiers, asphalt content, aggregate type and gradation,
in-place density all play an important role to reduce thermal cracking (Marasteanu
2007; Marasteanu 2012). The study also determined that a value of 400 J/m2
predicted if the asphalt mix would thermal crack or not and the disk-shaped
compact tension (DCT) test provided the most consistent results. Figure 4 shows
some of the results from measure thermal cracking and the fracture energy deter-
mined from the asphalt mix.
MnDOT and other states have begun work to implement these findings into their
specifications but some resent questions have been raised. Can fracture energy
predict other forms of cracking? Is the DCT test still the best test to predict cracking
after others have developed modifications to the test and is fracture energy still the
best predictor of cracking potential in a mix? Agencies are also confronted with
new mix designs, materials and products that are sold to add additional life to our
roadways but remain unproven unless test sections are built and after a number of
years the performance can be calculated. These questions are the basis for the need
for the northern states to join in this CG15 study and study the development of a
national HMA performance test.
614 R. Buzz Powell and B. Worel

5 Pavement Monitoring

Test pavements built for the CG experiment will be inspected for any sign of new or
impending cracking using a combination of automated and manual (both real time
and post processed) inspections. Crack maps will be developed as a function of
traffic, time, temperature, and seasonal effects in a format that facilitates analysis
using a number of methods common to state DOTs (AASHTO, LTPP, etc.).
Roughness, rutting, and macrotexture will also be measured. Other key parameters
(e.g., deflections, surface friction, etc.) will also be measured regularly in order to
establish relationships with traffic, time, temperature, seasons, and pavement con-
dition, albeit at some reduced frequency to optimize project resources. The same
measurements will be made at both the north and south locations, with data stored
on a common digital platform, in order to fully connect the research efforts at
NCAT and MnROAD.
Likewise, similar analysis methodologies will be used to scrutinize results. Data
collection will generally be less frequent on the pavement preservation sections
because they will be installed on open roadways. Additionally, data analysis will be
completely different for the PG15 experiment aimed at quantifying the life
extending and condition improving benefits of pavement preservation as a function
of varying pretreatment condition. Each 0.1-mile section will be subdivided into at
least 30 subsections, each with a unique level of pretreatment condition. The
objective of the CG15 study is to validate laboratory testing in specimens fabricated

50%
Total Lane Area Cracked (%)

40%

30%

20%

10%

0%
0 2,000,000 4,000,000 6,000,000 8,000,000 10,000,000
Equivalent Single Axle Loadings (ESALs)

N5 S5 S6 S13 Trigger

Fig. 5 Example of the 2012 NCAT pavement test track performance data collected
Research Partnership Between the National Center for Asphalt ... 615

with actual plant produced mix by comparing it to actual test section performance
and discrete subsection monitoring will not be necessary. Figure 5 shows an
example of the performance plot of cracking measured at the 2012 NCAT
Pavement Test Track.

6 Implementation Plans

“Phase I” laboratory testing to aid in the test section planning process began in the
fall of 2014. Preparations to rebuild the NCAT Pavement Test Track began in
March of 2015. Construction of new experimental pavements on the Track test oval
and US-280 was completed in the summer of 2015. Truck traffic on the NCAT test
oval began in October 2015 and will be complete by the fall of 2017. Sections for
both the PG15 and CG15 experiments will be constructed at MnROAD in the
summer of 2016. Documentation of traffic and performance on off-Track sections
and at the MnROAD facility will begin as soon as pavement preservation sections
have been placed and continue until the end of the 3-year research cycle.
Onsite sponsor meetings are hosted every 6 months in order to share preliminary
findings, seek feedback from sponsors’ technical representatives, and promote
implementation. As a function of the partnership between NCAT and MnROAD,
the location of 6-month meetings will rotate between the southern site (NCAT) and
the northern site (MnROAD). The first two meetings will be held at NCAT in the
fall of 2015 and the spring of 2016. The next meeting will be hosted at MnROAD
in the fall of 2016 (after the construction of test pavements and placement of
preservation treatments). Meeting location will rotate thereafter between MnROAD
and NCAT.
In past research cycles, it has been observed that implementation occurs in three
distinct phases of accelerated pavement testing projects executed for state DOTs. In
the first phase, short-term lessons learned during construction are sometimes
implemented immediately. This may be the case for a new material source (e.g., an
aggregate quarry with a higher LA abrasion loss), a change in practice with a
perceived level of risk that is too high to prove on the open road (e.g., higher
percentages of reclaimed and/or recycled materials), or a new technology (e.g.,
using HiMA to reduce pavement thickness). A second intermediate phase of
implementation occurs following proven good performance at the end of the first
full summer following construction/placement (e.g., providing an agency with an
assurance of good rutting performance). Finally, a third phase of implementation
occurs after some length of proven long-term performance (e.g., providing an
agency with an assurance of adequate protection from moisture damage).
A conference is hosted near the end of each 3-year research cycle in which all
aspects of the project are presented to hundreds of attendees from government,
industry, and academia with an emphasis on implementation of short-term, inter-
mediate, and long-term findings. In addition to a series of technical papers that are
written throughout the 3-year project cycle, a draft synthesis report that summarizes
616 R. Buzz Powell and B. Worel

findings in a single convenient reference will be prepared and distributed to research

sponsors for review by March 2018. The final report will be published by the summer
of 2018 after responding to sponsors’ feedback. It is expected that sections will
transition into long term monitoring via a separate, lower cost pooled fund in 2018.

7 Conclusions

The development of this partnership between the National Center for Asphalt
Technology (NCAT) and the Minnesota Department of Transportation MnROAD
Facility demonstrates the benefits that can be obtained when two test tracks work
together on national high priority research needs. To-date 14 states have joined in
this pooled fund to study both pavement preservation and the development of a
HMA performance test on a national level. Test sections are being built at both
facilities and Minnesota/Alabama with a common work plan and study goals. Each
facility will collect and store field performance data with similar methods and also
perform analysis on the data together. This groundbreaking partnership is expected
to provide insights on how research should be done in the future.
The PG15 experiment is designed to deliver two distinct and significant
implementable findings to state DOTs. One product will be the quantified rela-
tionship between pretreatment pavement condition and the time/traffic needed to
return to pretreatment condition for each preservation treatment. This approach
avoids any bias resulting from directly comparing the performance of treatments on
underlying pavement surfaces of varying condition. State DOTs can implement the
life extending benefit curves directly into decision trees to objectively select the
most cost effective treatments on their own networks solely as a function of mea-
sured pretreatment condition. Pavement management data from states can then be
used to calibrate these curves for local climate, materials, contractors, etc. New
preservation treatment sections on a high ADT state route will expand the research
scope to include different types of recycled pavements and to accurately quantify
the condition improving benefit curve of each treatment/treatment combination.
This second product will be a valuable tool for program managers who sometimes
struggle to prioritize investments in proactive preservation measures.
The CG15 group experiment is designed to validate laboratory cracking tests by
establishing correlations between the test results and measured cracking in real
pavements (test sections). This experiment will require the construction and traf-
ficking of a series of new test sections with asphalt mixtures that have a range of
expected cracking susceptibilities. All other factors that can impact cracking, such
as layer thicknesses and underlying layer stiffnesses, will be controlled to the
highest degree possible. A battery of laboratory cracking tests will be conducted on
mixes to identify which test results best correlate with field cracking.
Research Partnership Between the National Center for Asphalt ... 617

References

AFD40 Committee on Full Scale/Accelerated Pavement Testing. Transportation Research Board,

2015. https://sites.google.com/site/afd40web/home.
Applied Pavement Technology (APT). (2013). Cost effectiveness of the MDOT preventive
maintenance program final report, Michigan Department of Transportation.
Brown, R., Guo, X., Leiva Villacorta, F., Nelson, J., Powell, B., Robbins, M., et al. (2012).
Phase IV NCAT pavement test track findings. NCAT Report 12-10, National Center for
Asphalt Technology.
Kuennen, T. (2012). Lee road preservation study unique to NCAT. Pavement Preservation
Journal. 5(6), Winter.
Marasteanu, M. (2007). Investigation of low temperature cracking in asphalt pavements: National
Pooled Fund Study—Phase I. Minnesota Department of Transportation−Research Services
Section. http://www.lrrb.org/media/reports/200743.pdf.
Marasteanu, M. (2012). Investigation of low temperature cracking in asphalt pavements: National
Pooled Fund Study—Phase II. Minnesota Department of Transportation—Research Services
Section, http://www.lrrb.org/media/reports/201223.pdf.
MnROAD. (2015). Minnesota Department of Transportation. http://www.dot.state.mn.us/mnroad/
index.html.
National Road Research Alliance (NRRA). (2015). Minnesota Department of Transportation.
http://www.dot.state.mn.us/mnroad/nrra/index.html.
NCAT Pavement Test Track. (2015). National Center for Asphalt Technology. http://www.
pavetrack.com/.
Transportation Pooled Fund (TPF) Program. (2015). National Cooperative Highway Research
Program (NCHRP). http://www.pooledfund.org/Details/Study/496.
 Part VIII
Instrumentation for Accelerated Pavement
Testing
Development and Validation
of a Nondestructive Methodology
to Measure Subgrade Moisture Contents
at the NCAT Pavement Test Track

R. Buzz Powell

Abstract The Pavement Test Track is a full-scale, accelerated performance test

(APT) facility for flexible pavements managed by the National Center for Asphalt
Technology (NCAT) at Auburn University. Forty-six unique 200-foot test sections
are installed around a 1.7-mile oval and subjected to accelerated damage via a fleet
of tractors pulling heavy triple trailers in order to compress 10 million equivalent
single axle loadings (ESALs) of pavement damage into 2 years of fleet operations.
Methods and materials that produce better performance for research sponsors are
identified so that future pavements can be constructed based on objective life cycle
comparisons. Test sections are built on the NCAT Pavement Test Track with a
focus on surface mixes, structural pavement design, or pavement preservation.
Off-Track test sections with a focus on pavement preservation were built in the
summer of 2012 on a nearby low volume roadway where electronic instrumentation
was impractical; however, changing subgrade moisture content was needed in both
treated and untreated test sections in order to fully document the benefit of pave-
ment preservation. A nondestructive methodology was developed in order to
quantify changing subgrade moisture content without using imbedded sensors. An
overview of the measurement methodology and observed relationships between
cracking and subgrade moisture content measurements are summarized in this
document.

Keywords Accelerated pavement testing
Pavement preservation
Subgrade

moisture content Nondestructive testing

R. Buzz Powell (&)

National Center for Asphalt Technology, 277 Technology Parkway,
Auburn, AL 36830, USA
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 621

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_40
622 R. Buzz Powell

1 Background

The Pavement Test Track is a full-scale, accelerated performance test (APT) facility
for flexible pavements managed by the National Center for Asphalt Technology
(NCAT) at Auburn University. Forty-six unique 200-foot test sections are installed
around a 1.7-mile oval and subjected to accelerated damage via a fleet of tractors
pulling heavy triple trailers in order to compress 10 million equivalent single axle
loadings (ESALs) of pavement damage into 2 years of fleet operations. Methods
and materials that produce better performance for research sponsors are identified
so that future pavements can be constructed based on objective life cycle com-
parisons. An aerial photo of the NCAT Pavement Test Track is provided as Fig. 1.
Test sections are built on the NCAT Pavement Test Track with a focus on
surface mixes, structural pavement design, or pavement preservation. Off-Track test
sections with a focus on pavement preservation were built in the summer of 2012
on a nearby low volume roadway where electronic instrumentation was impractical;
however, changing subgrade moisture content was needed in both treated and
untreated test sections in order to fully capture the benefit of pavement preservation.
It is assumed that pavement preservation will protect pavements from moisture
intrusion, but to date there has not been data to document this benefit. A picture of
off-Track pavement preservation test sections on Lee Road 159 is provided as
Fig. 2.

Fig. 1 NCAT pavement test track

Development and Validation of a Nondestructive Methodology … 623

Fig. 2 Off-track pavement preservation test sections on nearby Lee Road 159

2 Pavement Preservation on Lee Road 159

Although pavement preservation treatments have been applied to test sections on

the NCAT Pavement Test Track, it is not known how the accelerated rate of traffic
damage will relate to off-Track roadways. A local county road was identified on
which treatments could be applied and studied under traffic conditions that would
directly relate to actual roadways. Lee Road 159 is a local 5½-in. thick dead end
county road that provides access to a quarry and an asphalt plant. The age of the
surface on this road was 14 years at the time of treatment. Like most rural roads in
the United States, Lee Road 159 was not designed at its existing pavement thick-
ness. It evolved from a single lane dirt road to a chip sealed two lane road to a road
paved with multiple layers of asphalt pavement.
Because trucks traverse the north/inbound lane unloaded and the south/outbound
lane loaded, and because the road was widened long ago from one lane to two,
surface condition varies significantly from shoulder to shoulder. This diversity of
pavement surface condition provided researchers with the opportunity to quantify
the life extending and condition improving benefits of pavement preservation when
used both proactively (on pavement surfaces in ideal condition) and reactively (on
pavement surfaces in less than ideal condition). Data sharing agreements with the
624 R. Buzz Powell

Fig. 3 Pretreatment crack map for Lee Road 159 with preservation treatment assignments

quarry and asphalt contractor provide truck tare and load weights that can be used
to relate changing post treatment pavement condition to both time and traffic.
After obtaining approval from the Lee County Commission to conduct the
research described herein, the existing condition of Lee Road 159 was carefully
documented. Researchers at NCAT worked closely with representatives from the
pavement preservation industry to select 23 study treatments or treatment combi-
nations (compared to two control sections on either end of the pretreatment con-
dition scale), and to assign them to ideal locations on Lee Road 159. Although all
sections contained a significant surface area in very good condition, more robust
treatments were applied to sections with at least some surface area in very bad
condition. One hundred foot long test sections (two lanes wide) were subdivided
into five foot by ten foot subsections that could be discretely monitored in order to
quantify the life extending and condition improving benefit of pavement preser-
vation as a function of varying levels of pretreatment condition. The pretreatment
crack map overlaid with final treatment assignments is shown in Fig. 3.

3 Moisture Content Measurement

Dry pavement structures are assumed to perform better than wet pavement struc-
tures. Pavement preservation treatments would be expected to prevent water
intrusion, which should result in dryer structures. Test sections on Lee Road 159
provide a unique opportunity to quantify the potential moisture reducing benefit of
pavement preservation. Electronic methods to measure subgrade moisture content
have not worked well at the NCAT Pavement Test Track. As seen in Fig. 4, time
domain reflectometry (TDR) measurements calibrated to site specific subgrade
materials and densities have been observed to drift impossibly high shortly after
installation (Brown et al. 2002). Subgrade moisture contents that appeared to have
increased to the mid-twenties were later found to still be near optimum (around
10 %) through destructive sampling. Additionally, electronics and data acquisition
Development and Validation of a Nondestructive Methodology … 625

Fig. 4 Poor performance of TDR moisture contents on the NCAT pavement test track

systems would be a liability on the shoulder of Lee Road 159 (which is an open
roadway).
A method is needed that does not require the installation of intrusively con-
founding sensors and roadside hardware. A review of methodologies inside and
outside the confines of conventional pavement research identified a system devel-
oped for use in agricultural applications that was available via donation by a vendor
interested in supporting research at the NCAT Pavement Test Track (http://cpn-intl.
com/503-elite-hydroprobe/). In this approach, a moisture sealed tube is permanently
installed in the roadway. The tube is positioned such that the tip is embedded two
inches into the subgrade. At regular intervals, a probe that contains both a nuclear
source and a detector are temporarily lowered into the tube. Post installation
measurements are directly proportional to moisture content via count ratio (test
count divided by the standard count) and factory calibration coefficients. Results are
made more accurate by correlating estimated moisture contents to measured
gravimetric measurements made in the laboratory using dry auger trimmings
recovered during sample tube installation.
Using this nuclear nondestructive method, it is not necessary to permanently
install intrusive sensors or roadside data acquisition hardware. The only research
asset on the roadway is an empty tube sealed with a recessed top that is temporarily
removed with a pull handle to facilitate periodic testing. The measurement device
must be operated by a technician certified in radiation safety, and a license is
required for ownership and storage.
626 R. Buzz Powell

4 Installation and Calibration

A small testing program was undertaken in the laboratory using bucket samples with
known moisture contents in order to become familiar with the technology and
experiment with different types of sample tubes. A picture of the laboratory setup in
which two different types of tubes were being evaluated is shown in Fig. 5. One and
a half inch diameter black iron tubes with brass tops were selected because the results
were reproducible, they were locally available, and they were thought to be robust
enough for long term use in the roadway. Tubes were installed using a dry auger in
the south/outbound lane of Lee Road 159 in the middle of each section in between
the wheelpaths. As each tube was installed, dry auger trimmings were placed in a
sealed container and taken back to the laboratory for moisture content determination.
Immediately after each tube was installed, a test count was run in order to establish a
baseline for periodic measurement that could be directly correlated to laboratory
moisture content measurements (thus, the calculation of a “spot bias” for each test
section). Results from this installation measurement correlation process are shown in
Table 1, where “A” and “B” are correlation coefficients that are used to compute
uncorrected gauge moisture contents using the industry standard
“moisture ¼ A  ðtest count=standard countÞ þ B” relationship. After test counts
were taken, brass caps were installed to seal out moisture as shown in Fig. 6.

Fig. 5 Laboratory measurements to verify tube and nuclear probe methodology

Development and Validation of a Nondestructive Methodology … 627

Table 1 Gravimetric calibration/correlation data for each test section on Lee Road 159
A: 25.95 B: 16.21 Std count: 11,335
Section number Oven moisture Gauge Count Uncor gauge Spot
(%) count ratio moisture bias
1 5.3 9117 0.8043 4.7 0.6
2 5.2 9525 0.8403 5.6 −0.4
3 4.2 9434 0.8323 5.4 −1.2
4 4.3 9888 0.8723 6.4 −2.1
5 5.4 9208 0.8124 4.9 0.5
6 6.5 8255 0.7283 2.7 3.8
7 9.8 8890 0.7843 4.1 5.6
8 5.6 8346 0.7363 2.9 2.7
9 5.9 9570 0.8443 5.7 0.2
10 6.5 9208 0.8124 4.9 1.6
11 6.2 9888 0.8723 6.4 −0.2
12 6.4 9570 0.8443 5.7 0.7
13 5.7 9979 0.8804 6.6 −0.9
14 5.1 10,115 0.8924 6.9 −1.9
15 5.5 9752 0.8603 6.1 −0.6
16 5.2 9706 0.8563 6.0 −0.8
17 6.2 10,976 0.9683 8.9 −2.7
18 4.1 9298 0.8203 5.1 −1.0
19 5.0 9162 0.8083 4.8 0.2
20 5.4 10,976 0.9683 8.9 −3.6
21 4.7 9208 0.8124 4.9 −0.1
22 4.5 9026 0.7963 4.5 0.0
23 4.7 9480 0.8363 5.5 −0.8
24 5.2 9208 0.8124 4.9 0.3
25 4.7 9797 0.8643 6.2 −1.6

5 Monthly Moisture Measurements

Each month, NCAT researchers remove the brass caps and record a test count for
all 25 test sections that is used to estimate subgrade moisture content. The differ-
ence between each laboratory measured moisture content and moisture content
estimated from the test count at the time of installation is used to fine tune the
general calibration in each test section. All subsequent periodic measurements are
made more meaningful using these site specific “spot bias” correlations. A picture
of the periodic data collection process showing the test probe temporarily inserted
as well as the pull handle used to remove the brass cap is provided as Fig. 7.
The difference between periodic and installation measurements is computed for
each set of monthly measurements and compared to the two control sections. A plot
628 R. Buzz Powell

Fig. 6 Installed tube with sealed and recessed top ready for traffic on Lee Road 159

Fig. 7 Monthly moisture measurements in uncapped tube on Lee Road 159

Development and Validation of a Nondestructive Methodology … 629

8
Change in Gravimetric Moisture Relative to L3 & L4 Controls (%)

6/11/2015
6

0
L1 L2 L5 L6 L7 L8 L9 L10 L11 L12 L13 L14 L15 L16 L17 L18 L19 L20 L21 L22 L23 L24 L25

-2

-4

-6

-8
Treatment Test Sections on Lee Road 159

Fig. 8 Monthly moisture measurements for Lee Road 159

for the 23 treatment sections is shown in Fig. 8. Test sections with a change in
moisture content that is greater than the control sections are plotted above zero on
the y-axis and test sections with a change in moisture content that is less than the
control sections are plotted below zero on the y-axis. The black dot compares each
section to the average of the two control sections, while the red line shows how
each section compares to the range of moisture content change for the two control
sections. Here it is seen that almost 3 years after the placement of pavement
preservation sections on Lee Road 159, moisture contents in most of the treatment
sections are still generally lower than the two control sections. It was not possible to
compare these results to TDR sensors due to the length and diameter of the sam-
pling tube.

6 Relating Moisture to Return of Pretreatment Distress

As previously mentioned, cracks are mapped discretely for all forty subsections in
each test section on Lee Road 159. Figure 9 is provided to illustrate how the square
footage of post treatment cracking (shown as solid blue lines) is compared to the
square footage of pretreatment cracking (shown as dashed red lines) as a means of
quantifying the life extending benefit of pavement preservation. Only the inbound
half of the example section is shown in Fig. 9. The general severity level of each
630 R. Buzz Powell

Fig. 9 Crack mapping within the inbound lane of a subsection on Lee Road 159

subsection is reflected in the shading, where green represents generally good post
treatment condition, yellow represents generally medium post treatment condition,
red represents generally poor post treatment condition, and black signifies the
subsection is not used for analysis because of some confounding factor (which in
this case is a water line under the pavement with a slow leak that had a negative
impact on cracking performance).
Figure 10 represents a high altitude view of these cracking data as a function of
time and truck traffic. In this graphic, the average of all the subsection cracked areas
for related treatments are all plotted as a function of time (and therefore, traffic).
Total cracking as well as the rate of cracking in the control sections is much greater
than the four treatments/combinations, with the more robust treatments providing
the most favorable performance relative to the control. Here, it is seen that cracking

Fig. 10 Example cracking as a function of time and truck traffic

Development and Validation of a Nondestructive Methodology … 631

Fig. 11 Example relationship between cracking and subgrade moisture

performance for the scrub seal section is approximately equal to cracking perfor-
mance in the chip seal section that was crack sealed beforehand. This may be
viewed as a confirmation of the common claim made by companies that promote
scrub seal materials and applications.
The assumption that pavement preservation treatments reduce subgrade moisture
content seems to be supported by these data from Lee Road 159; however, it should
be noted that the observed life extending benefit is diminished when cracking
ultimately returns. This diminishing effect is observed when change in moisture
content (relative to the control sections) is plotted against cracking (mapped at the
same time). In Fig. 11, it is seen that the moisture benefit drifts towards zero on the
y-axis as the average square footage of subsection cracking begins to increase
dramatically.

7 Conclusions and Recommendations

Some conclusions and recommendations that may be made based on the research
described herein include:
• Nuclear testing of subgrade moisture content in the manner described above is a
viable option for pavement researchers who share NCAT’s concern about the
inaccuracy of electronic measurements and/or the liability of roadside hardware;
632 R. Buzz Powell

• The application of some type of sealant to the sides of tubes prior to installation
may prevent leakage that negatively impact periodic testing. Several tubes had
to be removed and reinstalled in order to protect against water intrusion.
Generous use of sealant along the sides of the tubes prevented a recurrence of
leaking;
• Pavement preservation treatments in general were observed to reduce subgrade
moisture content, as assumed;
• More robust treatments were observed to produce greater reductions in subgrade
moisture content, as expected;
• Reductions in subgrade moisture content were observed to be diminished as
cracking increased; and
• Work to quantify the life extending and condition improving benefits of pave-
ment preservation, including reduction in subgrade moisture content as some
function of recurring cracking, will continue in the 2015 research cycle at the
NCAT Pavement Test Track, which includes a partnership with MnROAD to
execute a nationwide experiment in both pavement preservation and laboratory
testing for mix cracking performance.

Reference

Brown, E. R., Cooley, L. A., Hanson, D., Lynn, C., Powell, R. B., Prowell, B., et al. (2002). NCAT
test track design, construction, and performance. NCAT report 02-12, National Center for
Asphalt Technology.
Development of a New Pavement Strain
Coil Measuring System at CAPTIF

F.R. Greenslade

Abstract The Canterbury Accelerated Pavement Testing Indoor Facility (CAPTIF)

has been measuring plastic and elastic strain in unbound pavements for more than
40 years. The systems used for this type of measurement are based around
inductive coil sensors. CAPTIF’s current system relies on commercially produced
electronic circuit boards for LVDT instrumentation. These boards are being phased
out by the manufacturer. This has necessitated a need to develop a replacement
system. This paper documents the development of new instrumentation that will
replace CAPTIF’s current system. The criteria for the new system was to be
equivalent or better than CAPTIF’s current system providing stability for measuring
plastic strains over long periods and high resolution when measuring elastic strains
from dynamic loading events. The new instrumentation is based around a special
purpose integrated circuit for LVDT signal conditioning, the Analog Devices
AD698. By developing this instrumentation ‘in house’, CAPTIF has been able to
meet the criteria with a significantly lower cost system than in the past. In addition,
the new instrumentation is more adaptable to taking portable field measurements
because of its simpler setup and programming requirements.

Keywords CAPTIF
Unbound pavements
Pavement strain
Strain coils

AD698

1 Introduction

New Zealand’s roads are primarily built of thin surfaced unbound granular mate-
rials. One of the fundamental measurements that have been recorded at the
Canterbury Accelerated Pavement Testing Facility (CAPTIF) is the strain that
occurs in these pavement materials. Both elastic strains caused by transient moving

F.R. Greenslade (&)

New Zealand Transport Agency (NZTA), CAPTIF Road Research Centre,
2 McLeans Island Road, Christchurch, New Zealand
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 633

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_41
634 F.R. Greenslade

vehicles and permanent strains caused by repeated loads over time can be measured.
The sensors used for these measurements are inductive coil sensors. The coils are
free floating and cause much less disturbance to the materials than other types of
strain sensor. The coil pairs work in both the vertical and horizontal plane and can
be used to measure both compressive and tensile strains. The principle of operation
for these sensors involves the mutual inductance coupling of two co-axial sensors
embedded within the pavement layer. One of the coils is excited with an alternating
current (transmitter coil). The other coil (receiver coil) outputs an induced alter-
nating current signal due to the magnetic coupling between them. The coupling is
proportional to the axial spacing between the coils. Hence, the signal output from
the receiver can be used to quantify the strain occurring between the coils. CAPTIF
has modified and used a wide variety of strain coil measurement systems over the
years. The initial device used was the ‘Bison Gauge’ (Selig 1969). The most recent
system was based around off the shelf products from National Instruments
Corporation (Greenslade et al. 2012). Some of the National Instruments products
are being phased out so a new system has been developed based on an integrated
circuit (IC), the Analog Devices AD698.

2 Inductive Coils

The inductive coils consist of a circular plastic disk with a groove cut into the
perimeter in which copper wire is wound to form an inductive circuit (Fig. 1.) The
coils can be customised to any diameter with any number of turns of wire. CAPTIF
has settled on a configuration that uses an Acetal Copolymer plastic disk of 60 mm
diameter and 7 mm thick. A groove is cut 3.5 mm wide and 12.5 mm deep. Then
350 turns of 0.25 mm diameter enameled copper wire is wound into the groove.
A flying lead cable of a suitable length is attached and the completed coil is sealed
in marine grade polyurethane.
The inductive coils are used as a pair. One of the pair is used as a transmitter in
which an alternating current is driven from an electronic source. This has the effect
of generating an alternating magnetic flux field around the coil. The receiver coil is
placed within this magnetic flux field and this causes an alternating current to be
generated within the receiver coil. This phenomenon is known as mutual induction.
The magnitude of the generated signal is proportional to the distance between the
coils. The strength of the receiver coil output is measured by electronic detection
circuits. Thus, a pair of coils will provide a non-contact strain measuring device.
The magnetic flux occurring in the coil pair can be influenced by metallic objects
and external magnetic radiation. It is advisable to keep metallic objects at least
300 mm away.
Development of a New Pavement Strain Coil Measuring System … 635

Fig. 1 CAPTIF strain coil

3 LVDT Technology

The inductive coils function electrically in a very similar way to a displacement

sensor known as the Linear Variable Displacement Transformer (LVDT). LVDT’s
are commonly used in scientific and industrial applications. The electronic tech-
nology to measure LVDT’s has become very accurate and reliable. Because of this
similarity it is possible to use the electronic circuits that have been devised to work
with LVDT’s with the strain coils. Analog Devices Incorporated manufactures an
integrated circuit (IC), the AD698. The AD698 is described as a ‘Universal LVDT
Signal Conditioner’. This IC provides the essential functions to create an LVDT
measurement system and hence can be adapted to drive and receive the strain coil
technology. The AD698 is the heart of the new CAPTIF system. An electronic
circuit (Fig. 2) was designed with the AD698 IC being supported by some
peripheral components. External components are required to set the transmitter coil
636 F.R. Greenslade

Fig. 2 Strain coil excitation and signal conditioning circuit diagram

frequency and excitation. A nominal frequency of 10 kHz was chosen with an

excitation voltage of 9VAC (pk–pk). Driver amplifiers (LT1010) are used to buffer
the output from the AD698 to the low impedance of the transmitter coil. An
instrumentation amplifier (OPA2277) is used to amplify the very small signal
returned from the receiver coil.

4 Relay Multiplexor

At CAPTIF strain coils a placed in the pavement as an array (Fig. 3). Any of the
coils can be designated as a transmitter or a receiver. Each coil pair is measured
synchronously therefore a multiplexor must be used to route the correct transmitter
and receiver coils to the signal conditioning circuit. A ‘reed relay’ component is
used with a driver IC (Fig. 4). A digital control signal is sent to turn on each relay
as required.
Development of a New Pavement Strain Coil Measuring System … 637

Fig. 3 A typical strain coil array used at CAPTIF

638 F.R. Greenslade

Fig. 4 Coil selection relay circuit

5 Data Acquisition and Control

The data acquisition and control hardware used for this system are products from
National Instrument’s CompactDAQ range. The CompactDAQ system consists of
modules which plug into a chassis. Each module has a defined function.
CAPTIF has initially chosen the CDAQ9174 chassis which communicates to a
control PC via the USB bus. Three modules are used:
1. The NI9239 is a 24 bit analog to digital converter (ADC) module. This is used to
measure the voltage signal coming back from the receiver coil via the signal
conditioning circuit board.
2. The NI9403 is a digital output module. This is used to select the correct coils via
the relay circuit.
3. The NI9215 is an 8 channel 12 bit ADC. This is used for peripheral associated
devices such as a displacement sensor when calibrating, temperature sensors and
triggers.
CAPTIF uses National Instruments ‘Labview’ software to control the data
acquisition system.

6 Calibration

CAPTIF settled on a nominal spacing of 75 mm when installing the coils in the

pavement. Therefore, a range of 70–80 mm is used for calibration. CAPTIF
developed a specialised jig (Fig. 5) which makes calibrating the coils relatively
quick and easy. The jig has a hand operated cam. The cam is an eccentric cylinder
that can apply a 10 mm displacement on the receiver coil with a 180° rotation.
Both the voltage output from the receiver coil and the displacement transducer
are recorded at a rate that gives approximately 2000 samples over the cam range.
The data can then be processed to give an XY plot (Fig. 6) and then a unique
calibration equation is calculated.
Development of a New Pavement Strain Coil Measuring System … 639

Fig. 5 Coil calibration jig

Fig. 6 XY plot of calibration curve

640 F.R. Greenslade

The relationship between the coil voltage and displacement is not linear because
the magnetic flux density gradient is non-linear. At CAPTIF the data is processed
using a ‘2nd order Polynomial Fit’ function which returns an equation of the form:

y ¼ ax2 þ bx þ c

where y = mm, and x = volts.

7 Dynamic Strain Measurements

Transient loading events cause a dynamic resilient strain that can be captured with
the strain coils. The data acquisition hardware is set up to acquire data at the rate of
5000 samples per second for a period of 2 s. The data is post processed with a 16th
order low-pass Butterworth Filter at a cut-off frequency of 10 Hz. Typically, strains
of less than 20 uϵ are discernible (Fig. 7). This is quite an improvement on
CAPTIF’s previous (NI) system where 50 uϵ was the accepted working limit.

Fig. 7 Transient strain event

Development of a New Pavement Strain Coil Measuring System … 641

8 Static Strain Measurements and Drift

The strain coils can be used to measure plastic strains by recording data over a
period of time with the pavement undergoing repeated loading events.
The ability of the system to accurately measure these long term plastic strains is
compromised by ‘electronic drift’ due to ambient temperature change. Ambient
temperature can alter the operating characteristics of many of the electronic com-
ponents in the system. These changes can be cumulative and will appear as an
‘apparent’ strain in the pavement. An experiment was set up to quantify the elec-
tronic drift in the system.
A strain coil pair was mechanically fixed apart with a plastic spacer. This
fixed pair was then put inside a laboratory oven and held at a constant temperature
of 30 °C so that there could be no ‘real’ strain acting on the coils. The electronic
measurement system was set up beside the oven and left subject to ambient air
temperature changes. A temperature sensor was mounted on the AD698 signal
conditioning circuitry. Data was logged over period of 5 days recording the
ambient temperature of the electronics and the apparent change in strain (Fig. 8.)
A line of best fit was calculated from the data with an equation of 19.9 uϵ change
per degree Celsius. This is less than half the drift CAPTIF experienced with its NI
system. The apparent displacement change over this range was 73.8739–
73.8927 mm. The equation could be used to apply an uncertainty factor to mea-
sured static strains or it could be used to correct data if temperatures readings were
recorded simultaneously. (Note: This does not affect dynamic strains as temperature
change is relatively slow.)

Fig. 8 Strain ‘drift’ due to temperature change of the electronic circuitry

642 F.R. Greenslade

9 Flexibility and Improvements

By using the AD698 LVDT IC, the user now has greater control over the signal
output from the coils. CAPTIF uses a standard configuration of 60 mm diameter
coils at 75 mm spacing. The range of output from the AD698 is set to be around
0–10 V for a range of 70–80 mm coil movement. The 0–10 V output is matched to
the NI9239. It is possible to change the output to suit any preferred strain mea-
surement range. For example, a user may wish to use 100 mm diameter coils at
200 mm spacing with a 5 mm coil movement. The output of the AD698 can be set
up to do this or any other configuration. It can also output a different voltage range
to suit other types of data acquisition equipment.
CAPTIF will include a temperature sensor on board inside the system so that it
can monitor temperature change and compensate for temperature drift. This com-
pensation will be applied to our static measurements. In the future CAPTIF may
look to provide some kind of temperature stabilisation in the form of heating or
cooling elements that will keep the electronics at a fixed temperature to eliminate
this drift.
To date CAPTIF has produced a portable field unit powered by batteries. This
unit is capable of 16 transmitter and 16 receiver coils. The next stage is to
implement an Ethernet based system inside our APT facility. The system will
consist of 4 stations with each station having a capacity of 32 transmitter and 32
receiver coils. There will be one central software program that controls all four
stations.
The NI9239 24 bit analog input device has four channels. This new system
multiplexes the receiver coil output on to one of those channels. The NI9239’s four
channels operate simultaneously; therefore we could take advantage of this feature
and measure different coil pairs simultaneously. This would require a more complex
multiplexing circuitry. The coil pairs would have to be far enough apart that they
didn’t interfere with each other. This has benefits in the field where loading events
are not repeatable as they are in the APT facility.

10 Conclusions

The new CAPTIF strain coil measurement system based on the AD698 IC and the
NI CompactDaq hardware has proven successful on all counts. The AD698 circuit
is a working alternative at a fraction of the cost of the LVDT modules that are to be
phased out. The AD698 circuit has a very flexible output and when combined with
a 24 bit Analog to Digital converter gives a significantly better strain resolution than
the NI system. The cost of the CompactDaq hardware is also a lot less than the PXI
hardware described in that system. The ability in the future to develop a simulta-
neous channel field measurement device is an exciting possibility.
Development of a New Pavement Strain Coil Measuring System … 643

References

Greenslade, F. R., Alabaster, D. J., Steven, B. D., Pidwerbesky, B. D. (2012). The CAPTIF
unbound pavement strain measurement system. Advances in pavement design through
full-scale accelerated pavement testing (pp. 113–120).
Selig, E. T. (1969). A new technique for soil strain measurement. Buffalo, NY: State University of
New York.
Effects of Surface Condition
on Non-uniform Contact Stresses
of APT Tests

W. JvdM Steyn and J. Maina

Abstract Traditionally, uniform circular contact stresses were assumed between

tires and the pavement surfacing in pavement analysis. In recent years, the mea-
surement of static tire-road surfacing contact stresses on a rigid support with a
uniform texture led to an improved understanding of these interactions, as well as
improved analysis of the data. Furthermore, the measurement of tire surfacing
contact stresses on surfaces with a range of textures has led to an improved
understanding of the effects of surface textures on contact stresses and subsequently
pavement response. In this paper the effect of such a range of surface textures on
scaled APT tire-surfacing contact stresses is evaluated and the potential conse-
quences discussed.

1 Introduction

Vehicle loading is transferred to the pavement surfacing through tires. The tire
footprint or contact patch properties depend on the load conditions, tire properties,
and pavement surfacing properties. Improved understanding of this contact area and
its properties is required to enhance the understanding and modeling of pavements
by engineers and researchers. Various researchers have published on this topic,
evolving from uniform circular contact patches to non-uniform, non-circular contact
patches (Poulos and Davis 1974; Muki 1956; Tielking and Roberts 1987;
Thompson et al. 1987; Maina and Matsui 2004; Al Qadi and Wang 2009; Steyn
2009; Maina et al. 2013). Steyn (2012) also identified the increased incorporation of
vehicle–pavement interaction effects (including improved load and contact stress
models) as one of the future significant issues for APT research.

W. JvdM Steyn (&)
J. Maina

University of Pretoria, Pretoria, South Africa
e-mail: [email protected]
J. Maina
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 645

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_42
646 W. JvdM Steyn and J. Maina

Traditional loading models assumes circular uniform contact stresses between

the tire and the pavement surfacing. As indicated, the understanding that this is not
reality has long been developed, mainly through various measurements of the actual
tire-pavement surfacing patches and contact stress values. Up till recently the
majority of these measurements were conducted on pavement surfacings (or sim-
ulated surfacings) with a constant texture condition. The effect of pavement sur-
facing texture on the tire is well-known (Steyn and Haw 2005; Brach 2006;
Sandberg et al. 2011; Chatti and Zaabar 2012; Dunford 2013).
The tire-road surfacing contact stress is dependent on issues such as the tire load,
tire inflation pressure, tire type and material, temperature, and pavement surfacing
properties. Although some of these parameters have been evaluated before, and are
well described in literature (i.e. load and tire inflation pressure), the effect of the
pavement surfacing texture and profile on the contact stress distribution has
apparently not been studied to the same level of detail.
Recently, tire-road surfacing contact stress measurements were conducted by the
University of Pretoria using a pressure sensitive film sensor and the scaled Model
Mobile Load Simulator 3 (MMLS3) with a standard diamond tire and four sur-
facing types under static conditions. One load level and two tire inflation pressure
levels were used. As a scaled Accelerated Pavement Testing (APT) device, the
MMLS3 is used in numerous research studies (Hugo and Steyn 2015), and a good
understanding of the potential effects of the texture of the pavement surfacing on
the tire-pavement contact stresses is thus vital. Although data were also collected
under moving conditions [3600 revolutions per hour (r/h)], the focus of the data
analyses in this paper is on the static data. The four road surfacing types consisted
firstly of a smooth concrete, with the other three surfacings consisted of seal Road
Patches provided by AJ Broom Road Products. A 4.75, 9.5 and 19.0 mm seal Road
Patch was used in the tests. These provided for a wide range of surface textures.
A similar experiment was conducted in 2014 on an in-service road with two tex-
tures to evaluate the contact stresses between a car tire and a number of surfacing
types (Steyn and Ilse 2015).

2 Methodology

In order to address the stated objective, the following general methodology was
followed:
• Seal Road Patches were used to define three surface textures, with the fourth
being a smooth concrete surfacing;
• The Roadlab MMLS3 was used at static and 3600 r/h settings, with one load
(2.7 kN) and two tire inflation pressures (690 and 800 kPa) on the standard
diamond tire;
• The surface profile and texture for each of the four options were measured using
standard tests;
Effects of Surface Condition on Non-uniform Contact Stresses of APT Tests 647

• The Tekscan pressure sensitive sensor technology were used with two different
mat resolutions, and
• Data were collected and analyzed with the view of comparing the applied
contact stresses at the four different surfacing textures.
The MMLS3 used in the tests was provided and operated by Roadlab in
Germiston (Fig. 1). It is a standard MMLS used for APT. The MMLS was operated
at a load of 2.7 kN and both a static (0 r/h) and slow (3600 r/h) speed setting. The
standard diamond tires were used during the tests (Fig. 1). Two tire inflation
pressures were used, 690 and 800 kPa (two tires at each inflation pressure).
The sensors were placed under the MMLS, in line with the tire track (Fig. 2) and
protected (as is normal practice) using 1 mm thick rubber pads. Two types of
sensors were used. The BigMat (5350 N) sensor, which has a surface area of
439.9 mm × 480.1 mm and uses 2112 sensels (resolution of 1 sensel/cm2–
100 mm2 per sensel). The 5101 sensor, which has a surface area of 111.8 mm ×
111.8 mm and uses 1 936 sensels (resolution of 15.5 sensels/cm2–6.45 mm2 per
sensel). A sensitivity of 8 was used for the BigMat and 1 was used for the 5101
sensor. The sensors were used simultaneously to collect data from the same tire
consecutively. Data from the sensors are generated in raw format. Calibration of the
sensors was done using the actual applied load at a selected location. Data are
converted to pressure/contact stress values for each of the sensor locations. Data
were collected at a frequency of 100 Hz for each of the tests. A representative frame

Fig. 1 MMLS setup with standard diamond-pattern tire used for the tests
648 W. JvdM Steyn and J. Maina

Fig. 2 Tekscan pressure sensitive mats in place for the tests

of data was selected for each of the static and dynamic measurements for further
analyses.
In order to allow for a range of surfacing textures to be evaluated in a short time
and under the same controlled conditions, three surface seal Road Patches were
used. These were for a 4.75, 9.5, and 19.0 mm aggregate. A smooth concrete
surface was used as the control surface. All seal Road Patches were placed on top of
the concrete slab during testing, and thus no deflection of the pavement was allowed
for these tests. A rubber mat (1.0 mm thick) was placed on top of the surface seal
Road Patches to protect the sensors against the aggregate, especially for the 9.5 and
19.0 mm surfacing seal Road Patches. The textures of the three surfacing seal Road
Patches were measured using the standard MMLS profilometers, as well as sand
patch tests. The test program for the experiment focused on a combination of
surfacing textures, tire inflation pressures, and load speed (Table 1).

Table 1 Test matrix for Test number Surfacing Speed

experiment
1 Concrete Static
2 3600 r/h
3 4.75 mm Static
4 3600 r/h
5 9.5 mm Static
6 3600 r/h
7 19 mm Static
8 3600 r/h
Effects of Surface Condition on Non-uniform Contact Stresses of APT Tests 649

3 Data Collection and Analysis

3.1 Introduction

Data analysis consisted of a process of validating the collected data, followed by

analysis of the data with relation to the surface textures on which the tests were
conducted and the tire inflation pressure ranges. In this paper the 3D view of the
contact stresses are not analyzed in detail. A brief analysis of these 3D data did
indicate that, due to the one load level and two relatively similar tire inflation
pressures used in the tests, there were not major differences in the contact stress
patterns caused by the load and/or tire inflation pressure. However, the move from
the smooth concrete to the rougher textured seal surfacings did cause a much less
uniform 3D stress distribution.

3.2 Basic Data

The basic data originating from the pressure sensors consist of a collection of
stresses/pressures as measured on each of the sensel locations on the sensor mat. In
Fig. 3 a view is shown of the contact stress data for the four tires, both static and
dynamic for the 5101 sensor on the concrete surfacing. This serves as an example of
the typical data collected. The green graphs indicate the load cycles. Each of the
squares represents one sensel on the sensor mat. Although these graphs provide a
visually appealing format for the data, it is not necessarily easy to analyze the trends
in the data. The data indicate a range of colors indicating the level of contact stress
measured. It also shows the area of the large sensor that was loaded during the
measurement, as well as the total load on the sensor.

3.3 Contact Stress Distributions

A standard analysis procedure of the contact stress data is to evaluate the contact
stress distributions for the different conditions. If a uniform distribution is observed
(as was the assumption for most mechanistic analysis procedures) a uniform dis-
tribution of stresses, all equal to each other will be seen. As soon as the contact
stresses start to be affected by non-linearity in the tire-road surfacing contact area,
the distribution will change. In Fig. 4 the cumulative distribution of the data col-
lected using the 5350 N (BigMat) and the 5101 sensors are shown. The statistics of
the two sets of data are shown in Table 2 for all the measurements taken combined.
The data indicate that a similar trend is visible for the two sets of overall data,
although the BigMat sensor shows higher contact stress values in the middle of the
distribution than the 5101 sensor. This is mainly attributed to the larger sensels of
650 W. JvdM Steyn and J. Maina

Fig. 3 Example of full dataset for a set of measurements (5101 sensor, concrete, 4 individual
tires)

Fig. 4 Cumulative distribution of contact stress data for all tests, BigMat (BM) and 5101 sensors
Effects of Surface Condition on Non-uniform Contact Stresses of APT Tests 651

Table 2 Statistics for all BM data (all) 5101 data

tests for BigMat and 5101 (all)
sensors
Mean 245 356
Standard error 4 2
Standard deviation 226 307
Range 1279 1159
Minimum 24 14
Maximum 1303 1173
Count 2553 35,599
Coefficient of variance 92 86
(%)

the 5350 N (BigMat) sensor compared to the 5101 sensor (1 cm2 compared to
0.067 cm2). The minimum and maximum stress values corresponded well for both
690 and 800 kPa tire inflation pressures.

3.4 Texture Data

The texture depths of the surfacing seal Road Patches were measured using a
standard sand patch test. Transverse profiles of the surfaces were also measured
using the standard MMLS profilometer. The profilometer is measured using a solid
wheel with a 25 mm diameter measuring at intervals of 1.5 mm. This means that
the measurement will filter finer texture values due to the radius of the wheel, as is
evident from the data in Table 3 (Road Patch surface texture data). Analysis of the
data indicates that the MMLS profilometer texture depths for the coarse texture are
about 45 % that of the sand patch texture depths, and about 25 % for the finer
texture depths.

Table 3 Surface texture data: sand patch

Surface texture Sand patch based texture Average MMLS profile depth
description (mm) (mm)
Concrete 0 0
4.75 mm 1.3 0.3
9.5 mm 2.5 0.6
19.0 mm 5.0 2.3
652 W. JvdM Steyn and J. Maina

3.5 Contact Stresses for Different Textures

The focus of these tests was to determine the relationship between the surface
texture data and the contact stresses measured. In order to be consistent, majority of
the analysis was conducted on the 5101 sensor data (higher resolution). Similar
trends are observable for the BigMat sensor.
In Fig. 5 the cumulative contact stress distribution is shown for the high reso-
lution sensor (5101) and static loads. The distribution is shown for each of the four
textures. An additional line (Uniform contact stress) indicates the stresses calculated
from directly using the measured applied load and measured loaded area for each of
the tests. This distribution thus ignores the actual distribution due to unevenness
and texture of the seal. The trend that is observed is that the smoother sections
(concrete, 4.75 mm seal Road Patch) has a more consistent distribution of contact
stresses over the full range of observed contact stresses (50th percentile of 360–
385 kPa compared to a mean of 356 kPa (Table 2)). The rougher texture sec-
tions (9.5 and 19.0 mm) show a skewed distribution of contact stresses with the
50th percentile significantly lower than the mean (50th percentile between 176 and
222 kPa compared to a mean of 356 kPa (Table 2)).
Coarser textures thus give rise to more peak contact stresses between the tire and
the surfacing. These stress distributions also indicate that the coarser surfacing had
a higher percentage of the peak (>900 kPa) stresses, which are concentrated where
the coarse stones are situated. The contact stress shapes are thus not dependent on
the typically reported n- or m-shapes (for the realistic tire inflation pressures used)
(De Beer et al. 2006), but much more dependent on the location of peak stones.

Fig. 5 Cumulative contact stress distribution for different textures (5101 sensor, static)
Effects of Surface Condition on Non-uniform Contact Stresses of APT Tests 653

Fig. 6 Comparison between texture depth and average, 90th percentile and maximum vertical
contact stress data

In Fig. 6 the relationship between the average, 90th percentile and maximum
vertical contact stresses are shown. These statistics were calculated using the full
distribution of contact stresses measured at each of the loading/texture conditions. It
indicates that while the average contact stress decreases as the texture depth
increases (rougher seal surfacing), the 90th percentile and maximum values remains
relatively constant. If this figure is analyzed on its own, it almost indicates that the
texture depth does not play a role in the contact stress data. However, it is important
that the data in Fig. 5 are evaluated, as it is only with the full distribution of the
stresses that the differences due to the texture depths become visible. These dif-
ferences include a reduction in mean contact stress but an increase in the 90th
percentile of the contact stresses as these textures becomes coarser.

3.6 Tire Inflation Pressure Effects

Data were collected at two different tire inflation pressures. As indicated, the focus
of the analyses in this paper is on the cumulative distribution of the vertical contact
stresses. A comparison of the data collected at tire inflation pressures of 690 and
800 kPa indicates that there is a general increase in the average, 90th percentile and
maximum vertical contact stresses as the tire inflation pressure increases (Fig. 7).
However, it is not a major increase in these values. It should be appreciated that the
two inflation pressures used are relatively close to each other and thus a major
difference in the contact stress data is not expected.
654 W. JvdM Steyn and J. Maina

Fig. 7 Comparison between tire inflation pressure and vertical contact stress statistics

In Fig. 8 the vertical contact stress distributions obtained from the 690 and
800 kPa tire inflation pressure tests are compared. The tests on the four types of
surfaces (concrete, 4.75, 9.5 and 19 mm seal Road Patch) are shown in four dif-
ferent colors, with solid (690 kPa) and dashed (800 kPa) lines. The data indicate

Fig. 8 Comparison between 690 and 800 kPa tire inflation pressure contact stress distributions
Effects of Surface Condition on Non-uniform Contact Stresses of APT Tests 655

that the tire inflation pressure has a much smaller effect on the contact stresses
measured than the surfacing type. It is accepted that the distribution of the contact
stresses over the area of the tire patches will be affected, as was often shown in
literature, however, these data indicate that, if the loads are being applied in a
wandering pattern, the pavement surfacing will probably experience a very similar
input stress over time regardless of the tire inflation pressure (for the two inflation
pressures evaluated in this case).

3.7 Comparison to Stress-in-Motion Data

Previously a similar project was conducted where the Stress in Motion

(SIM) system was used to measure the MMLS contact stresses (De Beer et al.
2006). There are differences in the methodology and approach used for the SIM and
the Tekscan measurements. These included the need to roll the tire over the SIM
systems to enable measurements to be conducted (static on the Tekscan system) and
the rigid base of the SIM system compared to the flexible nature of the Tekscan
system. In these tests the Tekscan was used on top of a concrete surface, and thus
the rigidity of the system should not affect data. Further, the SIM measures contact
stress data orthogonally and the Tekscan only vertically.
Although SIM data were collected on a very large matrix of options, the com-
parison with the current Tekscan data is limited to a set of similar load and tire
conditions. Two such data sets were extracted from the SIM database (one for a
700 kPa tire inflation pressure and one for an 800 kPa tire inflation pressure, both at
around 1 km/h and both at 2.7 kN and using a Diamond tire). The cumulative
vertical contact stresses were calculated for the two SIM data sets (not the
Maximum Vertical Contact Stress (MVCS) that is typically reported in the SIM
literature), and in Fig. 9 these data are compared to similar data obtained from the
Tekscan measurements for the concrete as well as the three Road Patch surfacings.
Analyses of these data indicated that the two sets of SIM data are similar to the
9.5 mm Road Patch data measured using the Tekscan system. It should be noted
that SIM measuring pin heads have a diameter of 9.7 mm. Similar distributions of
data were observed for the range of vertical contact stresses by Sime and Ashmore
in their tests as reported by Epps Martin et al. (2002) using a Quartz sensor lon-
gitudinal bridge sensor array. It is however noteworthy that their measured peak
maximum stresses were slightly less than those measured by the Tekscan system.
This is ascribed to the difference between the two measurement systems, with the
latter monitoring system being finer. It is also noteworthy that the values measured
by Sime and Ashmore (1999) were made at speeds of 7200 r/h. This lends support
to the finding mentioned above that speed appeared not to influence the contact
stresses significantly. Further measurement and analysis of MMLS contacts stresses
at a range of speeds is proposed. When comparing these data to other datasets and
conclusions regarding contact stresses, it is important to ensure that related statistics
are compared. Some analysis procedures report actual stress data (such as this
656 W. JvdM Steyn and J. Maina

Fig. 9 Cumulative vertical contact stress distribution for MMLS Diamond tire—Tekscan and
SIM data

report) and actual maximum values (e.g. Sime and Ashmore 1999), while others
compare average stresses and others such as SIM reports, the MVCS values (e.g.
De Beer et al. 2006).

4 Conclusions and Recommendations

Based on the discussions and data presented in this paper, the following conclusions
are drawn:
• Coarser textures give rise to higher contact stresses between the tire and the
surfacing;
• The effect of tire inflation pressure on the measured contact stresses is, rela-
tively, much smaller than the effect of surfacing type, and
• SIM measuring pin heads, with the diameter of 9.7 mm, provides a surface
texture similar to the 9.5 mm Road Patch surfacing based on the distribution of
measured contact stresses.
On the basis of the discussions and data presented in this paper, it is recom-
mended that the potential effect of these different tire contact stress distributions on
the performance of the pavement under investigation be evaluated through
mechanistic-empirical analyses to assist in the quantification of these distributions
on pavement material response.
Effects of Surface Condition on Non-uniform Contact Stresses of APT Tests 657

Acknowledgments This work is partly based on the research supported in part by the National
Research Foundation of South Africa. The Grantholder acknowledges that opinions, findings and
conclusions or recommendations expressed in any publication generated by the NRF supported
research are that of the authors, and that the NRF accepts no liability whatsoever in this regard.
The support and assistance of the following parties are greatly appreciated and acknowledged: AJ
Broom Road Products for supplying the seal Road Patches for the project; SANRAL for use of the
Tekscan equipment; Roadlab for supplying and operating the MMLS as well as the sand patch
tests, and Prof Fred Hugo for discussion regarding analysis and interpretation of data.

References

Al Qadi, I., & Wang, H. (2009). Perpetual pavement responses under various tire configurations:
Accelerated pavement testing and finite element modeling. Journal of the Association of
Asphalt Pavement Technologists, 78, 645–680.
Brach, R. M. (2006). Adhesion, hysteresis and the peak longitudinal tire force. PhD thesis,
University of Norte Dame. Indiana
Chatti, K., & Zaabar, I. (2012). Estimating the effects of pavement condition on vehicle operating
cost. NCHRP Report 720, Washington, DC.
De Beer, M., Fischer, C., & Coetzee, C. H. (2006). Tyre-pavement contact stresses of the model
mobile load simulator Mk3 (MMLS3) compared with those of the Heavy Vehicle Simulator Mk
IV + (HVS IV+). Technical Memorandum CR-2005/30, CSIR Built Environment, Pretoria,
South Africa.
Dunford, A. (2013). Friction and the texture of aggregate particles used in the road surface
course. PhD thesis, University of Nottingham.
Epps Martin, A., Ahmed, T., Little, D. C., Hugo, F., Poolman, P., & Mikhail, M. (2002).
Performance prediction with the MMLS3 at WesTrack. 9th International conference on
asphalt pavements, Copenhagen.
Hugo, F., & Steyn, W. J. M. (2015). A synthesis of applications of the MLS as an innovative
system for evaluating performance of asphalt materials in pavement engineering. 11th
Conference on asphalt pavements for Southern Africa (CAPSA), 16–19 August 2015, Sun City,
South Africa.
Maina, J. W., De Beer, M., & Van Rensburg, Y. (2013). Modelling tyre-road contact stresses in
pavement design and analysis. 32nd Southern African transportation conference, Pretoria,
South Africa.
Maina, J. W., & Matsui, K. (2004). Developing software for elastic analysis of pavement structure
responses to vertical and horizontal surface loadings. Transportation Research Records, 1896,
107–118.
Muki, R. (1956). Three dimensional problem of elasticity for a semi-infinite solid with a tangential
load on its surface. Transactions of the Japan Society of Mechanical Engineers, 22(119),
468–474.
Poulos, H. G., & Davis, E. H. (1974). Elastic solutions for soil and rock mechanics. New York,
NY: Wiley.
Sandberg, U., Ejsmont, U. A., Bergiers, A., Goubert, L., Zöller, M., & Karlsson, R. (2011). Road
surface influence on tire/road rolling resistance. Report MIRIAM_SP1_04.
Sime, M., & Ashmore, S. C. (1999). Measurement of contact stresses under MMLS3 tires at
WesTrack under axle load of 2.1 kN by tire pavement interface pressure patterns. Federal
Highway Administration (FHWA) (Unpublished).
Steyn, W. J. M. (2009). Evaluation of the effect of tire loads with different contact stress patterns
on asphalt rutting. GeoHunan 2009 conference, Changsha, Hunan, China. ISBN:978-0-
7844-1042-4.
658 W. JvdM Steyn and J. Maina

Steyn, W. J. M. (2012). Full-scale accelerated pavement testing, 2000 to 2011. NCHRP Synthesis
433. National Cooperative Highway Research Program (NCHRP), Transportation Research
Board (TRB), National Research Council, Washington, DC.
Steyn, W. J. M. (2015). Measurement of MMLS3 contact stresses on various surfacings. Project
Report UP/WJVDMS/01/2015, University of Pretoria, South Africa.
Steyn, W. J. M., & Haw, M. (2005). Transport challenges for 2010: The effect of road surfacing
condition on tire life. 24th Annual Southern African transport conference. Pretoria. 11–13 July
2005.
Steyn, W. J. M., & Ilse, M. (2015). Evaluation of tire/surfacing/base contact stresses and texture
depth. International Journal of Transportation Science and Technology, 4(1), 107–118.
ISSN:2046-0430.
Thompson, J. C., Lelievre, B., Beckie, R. D., & Negus, K. J. (1987). A simple procedure for
computation of vertical soil stresses for surface region of arbitrary shape and loading.
Canadian Geotechnique Journal, 24, 143–145.
Tielking, J. T., & Roberts, F. L. (1987). Tire contact pressure and its effect on pavement strain.
ASCE Journal of Transportation Engineering, 113(1), 56–71.
Electronic Upgrade of a Standard
Benkelman Beam to Enable Capture
of Full Bowl Deflections

F.R. Greenslade

Abstract The New Zealand Transport Agency (NZTA) decided it should revisit
the design of electronic instrumentation that could be attached to standard
Benkelman Beams to measure surface deflection. This has been attempted several
times in the past but resulting devices were often complex and expensive. A low
cost solution would encourage uptake by pavement construction companies that
have easy access to the Benkelman Beam apparatus. Information gathered from the
Benkelman Beam has been restricted to a few data points that must be read by an
operator from a dial gauge and recorded manually. This requires some degree of
skill and is prone to subjectivity. By upgrading existing Benkelman Beams with
electronic measurement sensors and data recording equipment then full deflection
bowl data could be acquired with improved reliability when compared to manual
methods of recording. The Benkelman Beam test requires two measurements to be
taken, pavement deflection and distance from the loading vehicle. A suitable
deflection sensor was chosen. The criteria for choosing a distance sensor was that
there would be no mechanical connection to the loading vehicle. A laser sensor was
found with a suitable range and resolution. A low cost data acquisition module with
software was successfully developed. The completed kit can be built for under
$2000USD in parts. This is a fraction of the cost of purpose built electronic beams.
The whole test can be completed by one operator.

Keywords Benkelman beam
NZTA
Pavement deflection bowl
Electronic

sensors

F.R. Greenslade (&)

CAPTIF Road Research Centre, New Zealand Transport Agency (NZTA),
2 McLeans Island Road, Christchurch, New Zealand
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 659

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_43
660 F.R. Greenslade

1 Introduction

The Benkelman Beam was developed at the Western Association of State Highway
Organizations (WASHO) Road Test in 1952. It is a simple device that is used to
measure the rebound deflection of a flexible pavement. The Benkelman Beam is
used with a loaded truck—typically on a single axle (8.2 t) with dual tyres inflated
to give a surface contact area of 0.048 m2. Measurement is made by placing the tip
of the beam between the dual tyres and measuring the pavement surface rebound as
the truck is moved away.
The standard Benkelman Beam is supplied with a mechanical Dial Gauge which
is suitable for measuring peak deflections but not easily used for measuring
deflection bowls. The advantage of a full deflection bowl is that the shape of the
bowl can be analysed. Generally, broad bowls indicate that upper layers are stiff
with respect to deeper layers. Narrower bowls indicate weak upper layers relative to
the sub-grade (Tonkin and Taylor 1998).
NZTA’s CAPTIF Road Research Centre investigated various methods of
improving the instrumentation used with the Benkelman Beam in order to capture
full deflection bowls. The aim of the investigation was to find suitable components
that could be easily added to existing in-service Benkelman Beams. This paper
describes instrumentation developed at CAPTIF that provides a high resolution data
trace of the full deflection bowl.

2 Method

The method consists of replacing the mechanical Dial Gauge with an electronic
displacement sensor (LDS 05). This sensor provides a dc voltage output with
pavement deflection. This voltage is amplified with an electronic op-amp circuit.
The truck movement is measured with a laser range sensor (DT35). The laser sensor
outputs a voltage with distance measured to a reflector on the truck. The signal from
each of these sensors is fed into a Data Acquisition module (NI-USB6008). The
data acquisition module is read via a USB port on a Windows computer running a
NI-Labview software program.

3 Equipment

3.1 Displacement Transducer

The displacement transducer chosen was a model LDS-05 manufactured by L D

Sensors Ltd. (UK). It has a range of 5 mm. When attached to the beam (Fig. 1) it
Electronic Upgrade of a Standard Benkelman Beam … 661

Fig. 1 LDS-05 mounted on Benkelman Beam

will measure a pavement rebound range of 10 mm due to the 2:1 leverage afforded
by the beam setup. This sensor is a full bridge type sensor and was chosen for its
linearity, low noise and ease of integration to the NI-USB6008.

3.2 Distance Sensor

The distance sensor chosen was a SICK DT35-B15851 manufactured by Sick AG

(Germany). This is an infrared laser sensor. It has a range of 12 m and an analog
output voltage of 0–10 V dc that can be input directly into the NI-USB6008.
A suitable reflector is required to be mounted on the beam truck (Fig. 2).

3.3 Data Acquisition Module

The NI-USB6008 is a low cost data acquisition module manufactured by National

Instruments Corporation. It communicates via a USB interface. Three analog input
channels are used. One for each sensor and one to monitor the power supply to the
LDS sensor. Each channel is configured as a differential input channel with a range
of 0–10 V dc.
662 F.R. Greenslade

Fig. 2 SICK DT35 attached to tripod aimed at reflector board on back of beam truck

3.4 Signal Conditioning Electronics

The signal from the LDS sensor is amplified using an AD623 instrumentation
amplifier (ANALOG DEVICES USA). The gain of the amplifier is set to 67.6 by
means of a 1500 O resistor. The LDS sensor is powered from the NI-USB6008 5 V
supply. The Laser distance sensor is powered from an external 12VDC supply.
The signal conditioning circuit (Fig. 3) should be constructed by an experienced
person with an understanding of electronic circuits. At CAPTIF, a simple circuit
board was designed which was integrated into an enclosure with the NI-USB6008.
The enclosure included plugs and sockets to allow the sensors, modules and
computer to connect to each other. CAPTIF can supply the digital artwork for the
printed circuit board.

3.5 Software

The NI-USB6008 requires the National Instruments driver software NI-DAQ to be

installed and setup on the computer. NI-DAQ is available with the purchase of the
NI-USB6008. CAPTIF has written a Labview software program to capture and
process the data. This software has been compiled as a runtime version and can be
installed on any computer running Windows. The software is also suitable for running
on a touch screen system. The data is recorded as a tab delimited text file (Table 1)
which can then be loaded into Excel or similar program for further processing.
The runtime software is available for free distribution from NZTA’s CAPTIF
office. The software has four user interface pages (Fig. 4a–d).
Electronic Upgrade of a Standard Benkelman Beam … 663

Fig. 3 Signal conditioning electronics

3.5.1 Setup Sensors Page

This page is used to set up the sensors. The sensors are monitored live and adjusted
to suitable positions. Once the sensors are in position a zero reading is taken.

3.5.2 Run Test Page

A stop distance is set then the Run button is pressed. The program acquires data
from the deflection and distance sensors until the stop distance is reached.

3.5.3 Process Page

Data that has been recorded from the sensors is now processed. The data that has
been recorded at fixed time interval of 1000 Samples per second. The data set is the
interpolated so that a data point for a fixed distance interval (nom. 10 mm) is saved.
664 F.R. Greenslade

Table 1 Data file format Date 2 Apr 15

Operator Frank Greenslade
Location CAPTIF track station 30
Surface temperature (°C) 13
Max deflection 0.886
Bowl corrected False
Distance (m) Deflection (mm)
0.0013 −0.1916
0.0113 −0.1995
0.0213 −0.2066
0.0313 −0.2115
0.0413 −0.2136
0.0513 −0.2138
0.0613 −0.2132
0.0713 −0.2124
0.0813 −0.212
0.0913 −0.2118
0.1013 −0.2115
0.1113 −0.2113
0.1213 −0.2116
0.1313 −0.2133
0.1413 −0.2168
0.1513 −0.2217
0.1613 −0.2277
0.1713 −0.2343
0.1813 −0.2409
0.1913 −0.2479
0.2013 −0.256
0.2113 −0.2655
0.2213 −0.277
0.2313 −0.2907
0.2413 −0.306
0.2513 −0.3224
0.2613 −0.339

The interpolated data set can also be filtered using a moving average function. The
final metre of data capture is averaged and used as a zero baseline from which the
peak deflection is calculated. The ‘Bowl Correction’ procedure defined in NZTA
T/1 (NZTA 2013) is implemented.
Electronic Upgrade of a Standard Benkelman Beam … 665

Fig. 4 a Benkelman Bowl software ‘Setup Sensors’ page. b Benkelman Bowl software ‘Run
Test’ Page. c Benkelman Bowl software ‘Process’ Page. d Benkelman Bowl software ‘Save Page’

3.5.4 Save Page

The processed data is now saved. A suitable file name is chosen and any notes can
be added here to the file.
666 F.R. Greenslade

Fig. 4 (continued)

3.6 Equipment Parts List and Budget

The equipment, prices and source are tabled below (Table 2).

4 Equipment Set-Up

The equipment is assembled and connected to the beam and computer (Fig. 5).
Electronic Upgrade of a Standard Benkelman Beam … 667

Table 2 Components
Part NZD Source
(March 2015)
1. LDS-05 535 http://www.ldsensors.co.uk/
2. SICK DT35 720 http://www.sick.com/au
3. NI-USB6008 (OEM) 290 http://nz.ni.com/
4. Signal conditioning and hardware 300 http://nz.rs-online.com

Fig. 5 Sensors, electronics, beam and computer connected

The beam should be placed between the dual tyres of the truck as far forward of
the axles as is practically possible. Normally 300–400 mm is sufficient.
The displacement sensor is set somewhere about mid range. This can be adjusted
by looking at the live software graph.
The laser distance sensor is fixed with a tripod or some appropriate type of
mounting arrangement so that the sensor is reading a distance of less than 2 m from
the truck. This will allow approximately 10 m of range to be measured.
Once each of the sensors is stable a zero position is recorded with the software.
The truck is then driven away at a crawl speed. The data is captured until the
distance sensor is out of range. The data is then processed and saved.
668 F.R. Greenslade

Fig. 6 Overlay of 10 consecutive deflection bowls at same test point

5 Trials

Trials were carried out at CAPTIF with a standard Benkelman truck supplied by a
local contractor. Ten consecutive runs were conducted at the same test spot on a
section of asphalt driveway (Fig. 6).
All ten runs over the same spot were recorded to a data file. The peak deflections
(mm) for each run were processed and compare (Table 3).

6 Future Enhancements

In order to have a more seamless data set other sensors could be integrated into the
hardware and software. A GPS sensor could be attached to the laptop. A temperature
sensor could be connected to the Data acquisition module. The software could be
modified to read these sensors and the data recorded in the data file. It should be
possible to have a remote interface (wireless) that enables the test to be monitored by
the truck driver.
Electronic Upgrade of a Standard Benkelman Beam … 669

Table 3 Peak Deflections Samples 0.301

from 10 runs over same Test
0.314
Spot
0.341
0.312
0.297
0.335
0.318
0.326
0.322
0.334
Mean 0.320
Standard deviation 0.015

7 Conclusions

A method has been developed to measure the deflected bowl shape of pavement
surfaces using electronic instrumentation that can be added to existing Benkelman
Beams. The equipment is relatively in-expensive and will produce accurate reliable
data. While the equipment is arguably a bit more time consuming to set up than a
standard mechanical beam, this negative attribute is outweighed by ability of a
single operator to capture the detail of the full deflection bowl and the removal of
dial gauge reading errors.

References

NZTA. (2013). Standard test procedure for Benkelman Beam deflection measurements.
Wellington: New Zealand Transport Agency.
Tonkin and Taylor Ltd. (1998). Pavement deflection measurement and interpretation for the
design of rehabilitation treatments. Road Research Report 117, New Zealand Transport
Agency, Wellington.
Evaluation of Weight in Motion Sensors
on the IFSTTAR Accelerated Testing
Facility

Pierre Hornych, Jean-Michel Simonin, Jean-Michel Piau,

Louis-Marie Cottineau, Ivan Gueguen and B. Jacob

Abstract This paper presents an experiment, performed on the IFSTTAR circular

accelerated testing facility, to evaluate the performance of different weight in
motion (WIM) sensors. This project, carried out with the French ministry of
transport, aims at improving the performance of these WIM sensors, to allow
automated overload control. Four different types of sensors were evaluated in the
project: piezo-quartz sensors (2 types), piezo-ceramic sensors and piezo-polymer
sensors. The full scale experiment was performed on a thick bituminous pavement.
A total of 10 WIM sensors were installed, as well as other sensors (temperature
probes, vertical displacement transducers, geophones). Accelerometers were also
used to monitor dynamic load variations. The response of the sensors was evaluated
under a wide variety of loading conditions (different loading speeds, temperatures
and lateral positions of the wheels). In the first phase of the experiment, the 4 arms
of the carrousel were all equipped with single wheels (with wheel loads of 45 or 55
kN). In a second phase, different wheel configurations were tested, on each arm of
the testing facility (single wheel, dual wheels, tandem, tridem). A large database, of
about 30000 WIM signals, was collected. The full scale tests have allowed to
evaluate the repeatability of the different types of sensors, and the influence of
different loading conditions (temperature, load type, loading speed..). The results
indicate that the sensitivity of the different types of transducers to loading condi-
tions is very different. Recommendations for correcting and improving WIM sensor
response have been made.

Keywords Accelerated loading Weight in motion sensors Piezo-electric sensors

Metrological evaluation

P. Hornych (&)
J.-M. Simonin
J.-M. Piau
L.-M. Cottineau
I. Gueguen
B. Jacob
IFSTTAR, LUNAM Université, Route de BOUAYE CS4,
44341 Bouguenais Cedex, France
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 671

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_44
672 P. Hornych et al.

1 Introduction

Heavy vehicle overloads contribute to premature deterioration of infrastructure and

increase road unsafety and unfair competition between transport modes and oper-
ators. Public authorities and road operators must therefore implement an efficient
checking system to enforce weights and dimensions at an affordable cost.
Weight in motion (WIM) technologies are widely used for screening overloaded
vehicles, and sometime to issue warnings to the violators. However, these systems
are mostly not yet approved by the Legal Metrology (OIML R134 2006) for
automated overload control and fining. With the traffic volume increase, the safety
constraints and the downsizing of traffic officer staff, it becomes necessary to
introduce automated overload control by WIM.
This paper presents work carried out in the ongoing Automated Overloads
Control (AOC) project, supported by the French Ministry of Transports (DGITM)
and carried out by IFSTTAR and CEREMA (Cottineau et al. 2015). This project,
started in January 2014, aims to demonstrate the feasibility of automated overload
enforcement with adapted existing WIM technologies, to be certified. The project is
carried out in partnership with WIM system manufacturers. It includes a laboratory
evaluation of different WIM sensors, tests on the accelerated pavement testing
facility of IFSTTAR in Nantes, and development of improved methods of calibration
and correction of the sensor measurement. Finally, a road trial is planned at the end
of the project, to demonstrate the feasibility of the system, in real conditions.

2 Objectives of the Experiment, and Characteristics

of the WIM Sensors

This paper presents one of the major steps of the Automated Overloads Control
project, which is the evaluation of the response and performance of different WIM
sensors, on the full scale accelerated testing facility, under controlled test condi-
tions. The objectives of this experiment were to evaluate the performance of each
type of sensor, and then to propose different solutions (calibration, application of
different corrections) for improving measurement accuracy, in order to satisfy the
tolerances required for direct enforcement. In the French context, the target toler-
ance on gross vehicle weight (GVW) for automated control at traffic speed was
fixed at 5 % for an isolated load and 7 % for a multiple load (tandem or tridem).
Three main types of WIM sensors, corresponding to different sensor technolo-
gies, were selected for the project: Piezo-quartz sensors, piezo-ceramic sensors, and
piezo-polymer sensors. Figure 1 presents cross-sectional views of the piezo-quartz
sensors (Kistler) piezo-ceramic sensors (Thermocoax) and piezo-polymer sensor
(MSI) as installed in the pavement.
The piezo-quartz sensors, produced by KISTLER, consist of an aluminium alloy
profile, in the middle of which quartz disks are fitted under pre-load, spaced every
Evaluation of Weight in Motion Sensors on the IFSTTAR … 673

piezo-quartz sensor

Piezo-ceramic sensor Piezo-ceramic sensor

Fig. 1 Cross-sectional views of the different sensor types, after installation in a pavement

10 cm. The lateral sides of the metallic profile are protected with expanded poly-
styrene and soft synthetic foam, to avoid transmission of horizontal forces from the
pavement to the sensor. Piezo-quartz sensors are the most elaborate type of
piezo-electric sensors, and present theoretically the best accuracy. Two type of
piezo-quartz sensors have been tested (F and G). The last version (G) is very similar
to the previous one (F).
The piezo ceramic sensor, from Thermocoax, consists of piezo-electric cable,
encapsulated in a plastic beam. It is simpler, and less expensive than the
piezo-quartz sensor, but the plastic beam is more flexible, and thus more sensitive to
the evenness and deformations of the sensor support. It is considered as less
accurate than the piezo-quartz sensors. The piezo-polymer sensor also consists of a
piezo-electric cable, similar to the piezo-ceramic sensor. The difference, in our
experiment, is that it was tested bare, without encapsulation, as proposed by the
manufacturer. It is a rather flexible cable.
The different types of sensors are installed in a slot cut in the pavement, and
sealed with epoxy resin. The surface is then ground, so that the sensor is perfectly
flush with the pavement surface.
Before the full scale accelerated test, laboratory loading tests were performed on
the piezo-quartz and piezo-ceramic sensors. These sensors were subjected to ver-
tical cyclic loading tests, with the sensor resting on a rigid support, and to cyclic 3
point bending tests, using a hydraulic machine. Similar tests could not be performed
on the piezo-polymer sensor, which presents itself as a very flexible cable. These
tests were used to determine the sensitivity of the sensors to different parameters:
674 P. Hornych et al.

position of the vertical load, conditions of support of the transducer, loading fre-
quency, temperature. The results have shown that:
• Both sensors present a longitudinal variation of their sensitivity, depending on
the point of application of the vertical load. This variation represents about
1.7 % for the piezo-quartz sensor, and 5.4 % for the piezo-ceramic sensor.
• The response of the piezo-quartz sensor is almost insensitive to the position of
the support points, loading frequency and temperature. By its design, this sensor
is sensitive only to the vertical stress applied to its upper side.
• The response of the piezo-ceramic transducer depends on the curvature of the
transducer, on loading frequency and temperature. In a pavement, the response
of this sensor should therefore be more influenced by the speed of the vehicles
and by the level of deflection of the pavement.

3 Programme of the Full Scale Experiment

The pavement testing facility of IFSTTAR Nantes is a full scale facility, designed to
study damage of real pavements, under accelerated heavy traffic (Fig. 2). This
equipment presents the following features:
• a circular test track, 120 m long and 6 m wide;
• a maximum loading speed of 100 km/h;
• four loading arms, with a maximum load level of 130 kN on each arm (65 kN on
a single wheel);

Fig. 2 The full scale pavement testing facility of IFSTTAR Nantes

Evaluation of Weight in Motion Sensors on the IFSTTAR … 675

• the facility comprises three test tracks, and the machine can be transferred from
one track to another in less than a week.
The experiment was carried out on a thick bituminous pavement consisting of a
7 cm thick asphalt concrete wearing course, over 34 cm of asphalt road base,
resting on a granular subbase. The deflection level was of about 100 µm, meeting
the requirements of class 1 (Excellent site) of the European specifications for WIM
by COST 323 (Jacob et al. 2002).
Ten WIM sensors, including four piezo-ceramic Thermocoax sensors, four
piezo-quartz Kistler sensors (of two different types, F and G), and two
piezo-polymer Meas-Spec sensors have been evaluated on the fatigue test track.
Table 1 summarises the characteristics of these different sensors. Figures 3 and 4
show the sensor positions on the test track. Only half of the track was used for the
experiment, the other half being dedicated to another test.
In addition to the WIM sensors, the test track was also equipped with 2 vertical
displacement transducers, 4 temperature sensors placed in the pavement, and 8
accelerometers installed on the 4 arms of the traffic simulator, to measure the
dynamic load variations (Table 2).
Three data acquisition systems have been used in parallel to record the different
measurements: 2 ground based systems, and one on-board system installed on the
traffic simulator, for the accelerometer measurements. To synchronize these 3 sys-
tems, four detection cells have been placed on the four arms of the traffic simulator,
to measure precisely the time of passage of the loads on the WIM sensors.
The experimental programme included 3 phases:
1. a first phase where the 4 arms were equipped with single wheels, loaded at 45 kN.

Table 1 Characteristics of the evaluated WIM sensors

N° Technology Manufacturer Type Length Sensitivity
(m) (pC/N)
1 Quartz Kistler 9195F421 1.75 −1.752
2 Quartz Kistler 9195F311 1.50 −1.761
3 Ceramic Thermocoax VB FW CI1/3.6 m/RG 3.60 0.97*
58GT/45 m
4 Ceramic Thermocoax VB FW CI1/3.6 m/RG 3.60 0.97*
58GT/45 m
5 Quartz Kistler 9195GC41 1.75 1.569
6 Polymer Meas-Spec Roadtrax BL 2.00 53
7 Quartz Kistler 9195GC31 1.50 1.694
8 Polymer Meas-Spec Roadtrax BL 2.00 61
9 Ceramic Thermocoax VB FW CI1/3.2 m/RG 3.60 0.97*
58GT/50 m
10 Ceramic Thermocoax VB FW CI1/3.2 m/RG 3.60 0.97*
58GT/50 m
676 P. Hornych et al.

25

Piezo-polymere sensors
20

15
Radius (m)

Piezo-quartz sensors

10

Piezo-ceramique sensors

0
-25 -20 -15 -10 -5 0 5 10 15 20 25

Radius (m)

Fig. 3 Positions of the WIM sensors on the fatigue test track

Fig. 4 View of the WIM sensors installed on the test track

2. a second phase, where the 4 arms were equipped with single wheels, but where
different loads (45 and 55 kN) and different tire pressures (7, 8.5 and 9 bars)
were applied.
Evaluation of Weight in Motion Sensors on the IFSTTAR … 677

Table 2 Influence of the loading speed on WIM sensor response (single wheel, wheel position 6,
temperatures 24 and 41 °C)
Temperature Speed Quartz F Quartz G Ceramic Polymer
(m/s) (volt.m)a (volt.m)a (volt.m)a (volt.m)a
41 °C 9 0.42 0.34 0.38 0.54
14 0.41 0.34 0.38 0.49
20 0.42 0.35 0.37 0.46
25 °C 9 0.45 0.36 0.36 0.37
14 0.44 0.37 0.35 0.35
20 0.45 0.38 0.36 0.32
a
The sensor measurements are expressed in volt.m. This value correspond to the integration of the
signal in volts of the sensor over the length of the tire surface. It is not expressed in load value,
because the sensor could not be calibrated accurately in situ

3. a third phase, where each arm of the traffic simulator was equipped with a
different wheel configuration: single wheel loaded at 45 kN, dual wheels loaded
at 65 kN, tandem loaded at 90 kN, tridem loaded at 135 kN.
During each phase, various tests have been performed, to evaluate the influence
of several parameters:
• the loading speed (about 30, 50 and 70 km/h)
• the transversal position of the wheels (11 positions, with a spacing of 10.5 cm)
• the pavement temperature, which was modified by running the tests at different
hours of the day: early morning (23–27 °C) or in the afternoon (40–44 °C).
All these measurements have been stored in a data base of more than 30,000
WIM sensor signals, corresponding to about 300 different measurement conditions.

4 Response of the WIM Sensors

To illustrate the response of the different sensors, measurements performed in the

same conditions with different types of sensors are presented below. All these
measurements have been performed during the third loading phase, with the fol-
lowing conditions:
• each arm of the carrousel was equipped with a different load, in this order: arm 1
—single wheel loaded at 45 kN; arm2—tandem with single wheels loaded at 45
kN per wheel; arm 3—dual wheels loaded at 65 kN; arm 4—tridem with single
wheels loaded at 45 kN per wheel.
• Loading speed: 4.5 rounds per minute, corresponding to about 30 km/h.
• Lateral position n°1 of the wheels, corresponding to a radius of 18.47 m.
• Measurements have been recorded during 13 rounds of the carrousel, leading to
13 measurement repetitions for each arm (each load type).
678 P. Hornych et al.

Piezoquartz sensors type F Piezoquartz sensors type G

3 2.5

2
2
1.5
Amplitude (V)

Amplitude (V)
1 1

0.5
0
0

-1 -0.5
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Times (s) Times (s)

Piezopolymere sensors Piezoceramic sensors

0.4 2

0.2 1
Amplitude (V)

Amplitude (V)

0 0

-0.2 -1

-0.4 -2
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Times (s) Times (s)

Fig. 5 Examples of signals measured by the 4 types of transducers under tridem axles, (speed
30 km/h; load position 1)

Figure 5 presents raw electrical signals of the different types of sensors, in the
case of the tridem wheels. Piezo-quartz and piezo-ceramic sensor signals are pro-
cessed with a charge amplifier which integrates the signal. For this reason, the shape
of the signals is quite different from the one of the piezo-polymer sensors. In this
case, the signal includes positive and negative alternations, which have to be
integrated. For the different sensors, the following remarks can be made:
• The Kistler sensors present the most accurate response. The signals present a
low noise (somewhat higher for type G), and return very rapidly to zero after the
application of each load. This means that the sensor is sensitive only to the
vertical load.
• The piezo-polymer sensor presents a different (almost sinusoidal) signal shape,
and also more noise. Care should be taken to integrate this noisy signal.
• The response of the piezo-ceramic sensor is different. It starts to respond well
before the first wheel passes on the transducer, and returns slowly to zero, after
the passing of the last wheel. This shows that this transducer is affected by the
lateral load, or the deformation of the pavement. This can be explained by the
Evaluation of Weight in Motion Sensors on the IFSTTAR … 679

high flexibility of this transducer, and the absence of protection from the
influence of lateral loads.
The processing of the transducer signals includes:
• Noise filtering, with a low pass filter (50 Hz) including a zero phase digital
filtering.
• For the piezo-ceramic sensors, a high pass filtering (1 Hz) with a zero phase
digital filtering;
• Integration of the polymer signal including a procedure which removes the
linear trend of the signal to obtain a positive peak similar to that of the other
sensors;
• Integration of the signals, over a time interval corresponding to the time of
application of each wheel. The signal area thus obtained is proportional to the
applied load. It is finally converted into a load value, using the coefficient of
sensitivity of the transducer.

5 Influence of Measurement Conditions

5.1 Influence of Temperature

Figure 6 compares signals obtained with the piezo quartz type G sensor, and the
piezo-polymer sensor, for a single wheel, passing at a speed of 50 km/h, in the
central position, for two different temperatures, 24 and 41 °C.
For each transducer, Fig. 6 shows 10 signals measured at each temperature (over
a time interval of about 5 min). For each measurement condition, the 10 signals are
practically superimposed, indicating a good repeatability. However, the response of
the piezo-polymer sensor presents a strong variation with temperature. This is due

-3
10
3
2.5
Temperature 41°C
Temperature 41°C
Température 24°C
2.5 Température 24°C
2

1.5

1.5
Amplitude (V)

Amplitude (V)

0.5
0.5

0
0

-0.5 -0.5
0 0.05 0.1 0.15 0 0.05 0.1 0.15

Times (s) Times (s)

Fig. 6 Example of influence of temperature on the measurements of a piezo-polymer sensor (left)

and piezo-quartz type G sensor (right)
680 P. Hornych et al.

Table 3 Mean value and coefficient of variation of the measurements of each individual
transducer (single wheel, speed 70 km/h, wheel position 6, temperature 41.5 °C)
Sensor Quartz F Ceramic Quartz G Polymer
Mean Coeff. of Mean Coeff. of Mean Coeff. of Mean Coeff. of
(volt. variation (volt. variation (volt. variation (volt.m) variation
m) (%) m) (%) m) (%) (%)
Sensor 1 0.421 2.0 0.366 1.6 0.349 0.7 0.000461 2.2
Sensor 2 0.369 1.1 0.294 1.0 0.386 1.2 0.000411 3.0

to the thermal sensitivity of the polymer material composing the sensor. On the
contrary, for the piezo-quartz sensor, the variation with temperature is not
significant.

5.2 Influence of Loading Speed

Table 3 presents results of measurements made with the different WIM transducers
for different loading speeds: 30, 50 and 70 km/h. The measurements have been
made under a single wheel, in central position, and for two different pavement
surface temperatures: 25 and 41 °C. The results show that the piezo quartz and
piezo-ceramic transducers present a very low sensitivity to loading speed. The
measurements of the piezo polymer sensor, on the contrary, are sensitive to loading
speed, and decrease when the speed increases. This is probably linked to the
sensitivity of the polymer sensor to the deformations of the pavement, which
change with loading speed (and temperature), due to the visco-elastic behavior of
bituminous materials. Some influence of temperature, in particular on the piezo
polymer sensor, can also be seen.

5.3 Influence of Dynamic Load Variations

A vehicle, rolling on a road, presents significant dynamic load variations, due to the
unevenness of the road profile, combined with the response of the vehicle sus-
pensions. For a typical heavy vehicle, these dynamic variations can be of the order
of 10 % for a very smooth road profile, and up to 80 % for a very rough profile.
These dynamic variations are one of the main sources of error associated with WIM
measurements. For this reason, on a real road, WIM stations generally include
multiple sensors, to eliminate these dynamic effects. For the same reason, the COST
323 action has defined minimum specifications for road evenness, for installation of
WIM sensors.
Evaluation of Weight in Motion Sensors on the IFSTTAR … 681

Similarly, on the accelerated testing facility, dynamic load variations cannot be

avoided, although a low stiffness, pneumatic, suspension system is used. To eval-
uate these dynamic load variations, Table 2 compares measurements made with
each of the two transducers of the same type, installed on the test track. These
“identical” transducers are separated by a distance of 0.8 m (quartz F and Ceramic),
or 2.4 m (Quartz G and polymer). The table presents measurements made with a
single wheel, for the central wheel position, at a speed of 70 km/h, and at a
pavement surface temperature of 41.5 °C. For each transducer type, Table 2 gives
the mean value measured by each transducer (for 13 measurement repetitions) and
the corresponding coefficient of variation. The results indicate that:
• For a single transducer, the coefficient of variation of the measurements does not
exceed 2 % for the piezo-quartz and piezo-ceramic sensors, and 3 % for the
piezo-polymer sensors, indicating a good repeatability for each individual
transducer.
• For the two transducers of the same type, however, the difference between the
means is much more important, between 10 and 20 %. The loading conditions
(temperature, speed, type of load) being identical, this difference is due mainly
to the dynamic load variations.

5.4 Influence of Load Position and Total

Measurement Error

Figure 7 presents the variation of the measurements of one of the most accurate
sensors, the piezo-quartz type G sensor, with the lateral position of the wheels, for
measurements made at 70 km/h and 41 °C. The results show that the wheel posi-
tion has a significant influence on the measurements (the variation is about ±5 %).
This had also been observed previously in the laboratory tests, which indicated a
variation of the sensitivity of the transducers with the position of the load (Otto
2015).
Figure 8 shows, for the same type G sensor the variations of the experimental
measurements obtained for a single wheel, for several successive load sequences,
with 3 different speeds, 2 different temperatures; and for the 11 wheel positions. The
experimental scatter, due to the variations of the 4 parameters (speed, temperature,
dynamic effects and wheel position) is about 10 %, which is too much for auto-
mated overload control. However, several solutions can be suggested, to improve
measurement accuracy:
• The detection of the lateral position of the wheels, and the application of a
correction for wheel position;
• The use of a grid of several sensors, to reduce dynamic effects;
• The application of temperature corrections (mainly for the piezo-polymer
sensors).
682 P. Hornych et al.

1.2
1.1
1
Normalized surface

0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 2 4 6 8 10 12
Position

Fig. 7 Response of the piezo-quartz type G sensor, for different wheel positions (speed 70 km/h,
temperature 41 °C)

1.2
1.1
1
Normalized surface

0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 2 4 6 8 10 12
Position

Fig. 8 Variations of the response of the piezo-quartz type G transducer, for different wheel
positions, 3 different speeds (30, 50, 70 km/h) and two different temperatures (24 and 41 °C)

5.5 WIM Sensor Repeatability

Finally, Table 4 below compares the repeatability of the measurements of the

different tested WIM sensors. This table gives the mean coefficients of variation
obtained for each sensor, for each type of load (single wheel, dual wheels, tandem,
Evaluation of Weight in Motion Sensors on the IFSTTAR … 683

Table 4 Repeatability of the measurements of each type of WIM sensor, expressed in percentage,
for constant loading conditions
Load type Piezo-quartz F Piezo-quartz G Piezo-ceramic Piezo-polymer
Coeff. of Coeff. of Coeff. of Coeff. of
variation (%) variation (%) variation (%) variation (%)
Single wheel 1.13 1.02 1.01 1.61
Dual wheels 1.16 1.04 1.12 1.51
Tandem (each 1.93 1.61 1.58 2.23
wheel)
Tridem (each 1.68 1.45 1.48 2.49
wheel)

tridem). Each “mean” coefficient of variation has been calculated by taking the
average of coefficients of variation obtained for different measurement sequences,
each having constant loading conditions (same speed, same temperature, same
wheel position).
For a single wheel and for dual wheels, the repeatability of the piezo-quartz and
piezo-ceramic sensors is similar. It is somewhat worse for the Piezo-polymer
sensors. For the tandem and tridem axles, the repeatability (for each wheel of the
group) is worse than for the single wheel. Nevertheless, in all cases, the repeata-
bility of the measurements, for constant loading conditions, is satisfactory. For the
piezo-quartz and piezo-ceramic sensors, in particular, this repeatability never
exceeds 2 %.

6 Conclusion

This paper presents results of a full scale pavement experiment, conducted to

evaluate the performance of 4 different types of WIM sensors. Thanks to the
characteristic of the IFSTTAR accelerated pavement testing facility, the senors have
been tested under a wide range of loading conditions: different load types (single
wheel, dual wheels, tandem, tridem), different temperatures, different load posi-
tions, different speeds. A large database, of about 30,000 measurements has been
obtained.
The results indicate that for constant loading conditions (constant speed, posi-
tion, temperature), the repeatability of the sensors, after installation in a pavement,
is satisfactory, less than 2 % for the piezo-quartz and piezo ceramic sensors, and
less than 3 % for the piezo-polymer sensors. Under variable loading conditions, the
total measurement error can increase to up to about 10 %. The main sources of error
are:
• The dynamic load variations
• The lateral position of the load
684 P. Hornych et al.

• And, only for the piezo-polymer sensors, the temperature and loading speed.
In the second part of the Automated Overload Control Project, the objective will
be to propose technical solutions to improve the measurement accuracy. From the
results of the full scale experiment, several improvements can already be suggested:
• Compensation of the dynamic effects, by using a grid of several WIM sensors,
or other types of additional sensors.
• Detection of the lateral position of the wheels, by means of a WIM sensor placed
diagonally in the road lane, and correction of the effects of the lateral position.
• Accurate calibration of sensor response with temperature (if possible in situ),
and correction of temperature effects.
Another full scale accelerated test, and a test on a real road, (motorway A4, in
the East of France), are planned at the end of the project to test and validate these
solutions.

Acknowledgements The authors would like to express their thanks to: the French Transport
Ministry’s DGITM Directorate and the CEREMA for their support of successive R&D projects on
weight in motion technology. The WIM system manufacturers, STERELLA and FARECO, for
their collaboration to this project.

References

Cottineau, L. M., Hornych, P., Jacob, B., Schmidt, F., Dronneau, R., & Klein, E. (2015).
Automated overload enforcement by WIM. 25th World road congress, Seoul, November 2015.
Jacob, B., O’Brien, E. J., & Jehaes, S. (2002). Weigh-in-motion of road vehicles—final report of
the COST323 action. LCPC, Paris (French edition, 2004).
OIML R134-1. (2006). Automatic instruments for weighing road vehicles in motion and
measuring axle loads. Part 1: Metrological and technical requirements—Tests; Edition 2006.
Influence of Tire Footprint Area
and Pressure Distribution on Pavement
Responses

Fabricio Leiva-Villacorta, Adriana Vargas-Nordcbeck,

José P. Aguiar-Moya and Luis Loría-Salazar

Abstract For most pavement analyses, it is assumed that the tire load is uniformly
applied over a circular area. Also, it is generally assumed that tire inflation and
contact pressures are uniform throughout the contact area. Several studies on this
topic have shown different non-uniform pressure patterns. Therefore, a full
understanding of the interaction between tires and pavement is necessary to obtain
more accurate pavement responses. The objective of this study was to evaluate the
effects of truck tire contact pressure on pavement responses at different loading
conditions. A tire footprint system was used to capture contact pressure patterns
statically and dynamically (low speed) at three inflation pressures and three wheel
loads. All testing conditions were performed using a Heavy Vehicle Simulator HVS
Mark VI with a five-rib tire type 11R22-5. A flexible pavement section instru-
mented with asphalt strain gauges, pressure cells and multi depth deflectometers
was used to measure pavement responses. Measured tire-pavement contact stress
data were input into a finite element analysis program to compute pavement
responses and compare them to the measured responses. The contact pressure
patterns obtained for the five-rib tire indicated that higher pressures were obtained
for the inner ribs based on the controlled variables. In general, the results indicated
that the contact area decreased for a given load as the inflation pressure was
increased. Statistical analysis confirmed that pavement responses were significantly
related to tire pressure distribution.

F. Leiva-Villacorta (&)
A. Vargas-Nordcbeck
J.P. Aguiar-Moya
L. Loría-Salazar

LanammeUCR, San José 11501-2060, Costa Rica
e-mail: [email protected]
A. Vargas-Nordcbeck
e-mail: [email protected]
J.P. Aguiar-Moya
e-mail: [email protected]
L. Loría-Salazar
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 685

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_45
686 F. Leiva-Villacorta et al.

1 Introduction

The contact footprint patterns of a vehicle tire on a pavement structure and its
corresponding pressure distribution related to the vehicle load variation may indicate
the tire quality and the state of the tire wear and tear. Accordingly, measurement of a
tire’s contact footprint pattern and pressure distribution is useful in determining
stress concentrations and determining possible causes of typical pavement distresses.
Pavement responses are closely related to long term pavement performance and
distress. Fatigue cracking and rutting, two major flexible pavements distresses,
could be related to immediate pavement responses and could be explained in a
mechanistic way. Horizontal tensile strains at the bottom of the asphalt concrete
layer can explain fatigue cracks that initiate at the bottom and progress to the
pavement surface. In contrast, pavement cracks can also start at the pavement
surface due to excessive tensile strains and progress downward (NCHRP 2004a, b;
El-Basyouny 2004). Similarly, vertical compressive strains at the top of the sub-
grade are considered closely related to pavement rutting due to compaction and/or
consolidation of the soil (Huang 1993).
The tire-pavement contact pressure distribution is significantly affected by tire
inflation pressure, tire type, tire load and tire tread patterns. Many measuring
systems have been developed to measure the tire-pavement contact pressure in the
last decade. The measured data clearly reveal that the tire-pavement contact pres-
sure distribution is non-circular, non-uniform and discontinuous (Roque et al. 2000;
Beer et al. 1999).
Pavement responses can be measured directly using in situ instruments
embedded in pavement structures. Over the past two decades, computer controlled
instrumentation technology has been used to acquire real-time measurements of
pavement responses to dynamic traffic loading. Mateos and Snyder (2002) tested
four sections at the Minnesota Road Research facility (Mn/ROAD) with a moving
load configured at various axle loadings and tire pressures and found that changes
in tire pressure did not significantly affect pavement behavior. Al-Qadi et al. (2002)
found that wide-base “super singles” were not more damaging to the Virginia Smart
Road sections than dual tires and they also reported that the radial tires reduced
strain at the bottom of the asphalt layer. Sebaaly and Tabatabaee (1992) found
increased pavement stresses caused by high tire pressure tires and the wide-base
“super singles”, but they found the effect of high tire pressure was insignificant to
pavement performance. Akram et al. (1993) from the Texas Transportation Institute
(TTI) tested two thin and thick road sections with varying tire pressures and vehicle
speeds and found that high tire pressures caused higher tensile strain at the bottom
of the AC layer but had no significant effect on vertical strain at the top of subgrade.
Although measurement of pavement responses in a road test can provide the
most direct real-time data, road tests are always expensive. Fortunately, aside from
real pavement on-site measurements, a theoretical analysis method can also be
employed to simulate pavement responses due to traffic loading. Two categories of
analytical programs have been used in flexible pavement analysis, the elastic
Influence of Tire Footprint Area and Pressure Distribution … 687

multilayer model and the finite element model. Machemehl et al. (2005) and Prozzi
and Luo (2005) used a multilayer program to compute asphalt pavement responses
due to measured non-uniform tire-pavement contact stress and found tire pressure
has significant effects on tensile strains at the bottom of the asphalt concrete layer,
but the effect of tire pressure on vertical strains at the top of the subgrade is
insignificant. Weissman (1999) used an elastic multilayer based program and
non-uniform contact pressure distribution to predict the distress patterns of a
pavement undergoing overloads at an accelerated loading facility. Siddharthan et al.
(2002) computed pavement responses using a measured non-uniform contact stress
as well as a uniform contact stress distribution and found the uniform method
overestimated pavement responses. Wang and Machemehl (2006) used a finite
element program to compute asphalt pavement responses due to measured
non-uniform tire-pavement contact stress and found increased truck tire pressure
can cause increased pavement distress for both cracking and rutting.
For most pavement analyses, it is assumed that the tire load is uniformly applied
over a circular area. Also, it is generally assumed that tire inflation and contact
pressures are uniform throughout the contact area. However, several of the studies
mentioned above on this topic have shown different non-uniform pressure patterns.
Therefore, a full understanding of the interaction between tires and pavement is
necessary to obtain more accurate pavement responses.

1.1 Objective

The objective of this study was to evaluate the effects of truck tire contact pressure
on pavement responses at different loading conditions at the full scale accelerated
pavement facility of the University of Costa Rica—PaveLab.
In order to achieve this objective, a tire footprint system was used to capture
contact pressure patterns statically and dynamically (low speed) at three inflation
pressures and three wheel loads. All testing conditions were performed using a
Heavy Vehicle Simulator HVS Mark VI with a five-rib tire type 11R22-5.
A flexible pavement section instrumented with asphalt strain gauges, pressure cells
and multi depth deflectometers was used to measure pavement responses. Measured
tire-pavement contact stress data were input into a finite element analysis program
to compute pavement responses and compare them to the measured responses.

1.2 Test Section

Table 1 shows the characteristics of the evaluated section with its respective layer
thicknesses obtained from ground penetrating radar (GPR) measurements and
backcalculated layer moduli based on Falling Weight Deflectometer results (FWD).
These are the layer moduli computed when the pavement structure was intact.
688 F. Leiva-Villacorta et al.

Table 1 Test Track in-place Properties\section AC3

properties
Asphalt concrete thickness (H1), cm 13.2
Base thickness (H2), cm 31.0
Subbase thickness (H3), cm 30.1
AC modulus (E1) @ 25 °C, MPa 3800
Granular base modulus (E2), MPa 170
Subbase modulus (E3), MPa 140
Subgrade modulus (E4), MPa 70

The top layer consists of an asphalt concrete (AC) mixture with nominal max-
imum aggregate size of 19.0 mm with an optimum binder content of 4.9 % by total
weight of mixture. The cement treated base (CTB) was designed to withstand a
compressive stress of 35 kg/cm2 with an optimum cement content of 1.7 % by
volume of aggregate and with a maximum density of 2013 kg/m3. The base
material and granular sub-base were placed at a maximum density of 2217 kg/m3
with an optimum moisture content of 8.6 %. The sub-base material had a CBR of
95 %. Finally, the subgrade material was constructed for a maximum density of
1056 kg/m3 with an optimum moisture content of 52 % and CBR of 6.6 %.

1.3 Instrumentation

The experiment included not only the instrumentation integrated with the HVS
system but also embedded instrumentation. HVS onboard sensors can record the
applied load, tire pressure and temperature, position and velocity of the load car-
riage. Embedded instrumentation include asphalt strain gauges (PAST model
H-shaped sensors), pressure cells (SOPT model transducer for soils), multi depth
deflectometers (MDDs), moisture and temperature probes. These sensors were
chosen based on previous HVS owner’s experience (HVS 2015; Baker Harris et al.
1994). Additionally, the HVS was equipped with a laser profiler that can be used to
create a three-dimensional profile of the section and a Road Surface Deflectometer
(RSD) is added to the testing equipment to obtain deflection basins at any location
along the test section (Leiva-Villacorta et al. 2013, 2015).
Figure 1 shows the instrumentation array used for the experimental section.
The PAST sensors were placed at the base/HMA layer interface in the longitudinal
or traffic loading direction and in the transverse direction. MDD sensors were
installed at 4 different depths to cover all four structural layers. The thermocouples
were placed at four depths: surface, middle depth of the AC layer, at the PAST
sensors depth and 5 cm into the base layer.
Data collection of the 3D profile, strain, pressure, temperature and deflection is
performed based on load repetitions. At the beginning of the test, data is obtained at
short intervals. After 20,000 load repetitions, data is collected on a daily basis.
Inspection of fatigue and reflective cracking, friction loss, loss of aggregate-asphalt
Influence of Tire Footprint Area and Pressure Distribution … 689

30 cm 60 cm 90 cm

MDD MDD
Thermocouple

Section Length = 6.0 m

HMA
GB/CTB

Subbase
PAST
Subgrade

Fig. 1 Sensor array

bond and any other surface damage is performed on a daily basis during the HVS
daily maintenance work.

1.4 Tire Footprint Acquisition System

A large high-resolution sensor, compatible with a wide range of tire sizes was used
to capture tire footprint pressure patterns. Tire footprints can be captured statically
or dynamically and are displayed as high resolution, multi-colored images of the
tire contact pressure pattern in real time. The system’s application specific graphing
and image analysis software enables quantitative and qualitative analysis of the tire
footprint. Figure 2 exhibits the pressure sensor, the sensor handle which is used to
transmit the sensor data to the Wireless/Datalogger Unit and a picture of the sensor
being used with the tire load configuration of Heavy Vehicle Simulator
(HVS) Mark VI. The software displays the pressure distribution data in multiple
formats and the user has the option to create and customize graphs from the cor-
responding “movie” data or export as an ASCII file for use with other programs.

Fig. 2 Tire footprint acquisition system

690 F. Leiva-Villacorta et al.

The sensor consists of a square matrix of 425 mm with a thickness of 0.381 mm

and 1936 pressure sensing elements (sensels). The maximum recordable pressure is
3447 kPa (500 psi). An equilibration device is used for improving accuracy and
lifespan of the sensor. During equilibration, the sensor is inserted between a flat
backing plate and an air filled bladder, which is inflated in order to apply a uniform
pressure to the active area of the sensor. The equilibration process allows the
software to compensate for any variation or uneven output across individual sensing
elements caused by manufacturing or repeated used of the sensor. Equilibration
devices are useful to perform quality assurance checks on the sensor and confirm
uniform output by the sensor.
The tire footprint acquisition system is a resistive-based technology. The
application of a normal force to an active sensor results in changes in the resistance
of each sensing element (sensel) in inverse proportion to the force applied.
A multi-point calibration procedure was performed at a range closed to the expected
tire pressure from 344 to 1379 kPa (50–200 psi).

2 Pressure Measurement

The tire-pavement contact pressure data were measured on a 11R22-5 tire which is
part of the dual load configuration of the Heavy Vehicle Simulator Mark VI. This
tire type was tested for nine load-inflation combinations: three tire loads and three
inflation pressures. The 11R22-5 tire was chosen because its size is among the most
typical in Costa Rica (Sibaja 2014). Figure 3 shows two examples of the tire
footprint and vertical pressure distribution for different loads and pressures. The five
tire treads or ribs are clearly defined and for all the cases evaluated in this study the
center tread had higher pressures.
The pressure data can be exported to any type of spreadsheet such that the results
can be analyzed. Figure 4 exhibits an example of the analyzed data. It was deter-
mined that the pressure distribution along the majority of the tire treads can be

Fig. 3 Tire footprint and pressure distribution. a 40 kN and 586 kPa and b 50 kN and 689 kPa
Influence of Tire Footprint Area and Pressure Distribution … 691

800-1000 1200
1000 600-800
Pressure, kPa

400-600 1000 R² = 0.8105

800

Pressure, kPa
200-400
600 0-200 800

400 600
200 400 R² = 0.7243

01 Outter Tread
7 200 Middle Tread
Inner Tread
Long. 13 0 R² = 0.9216
Distance, 19 0 5 10 15 20 25
cm Longitudinal Distance, cm

Fig. 4 Example of exported and evaluated pressure data

approximated with a parabolic function. On the other hand, further analysis was
performed to quantify the change in the footprint. Table 2 shows the change in the
maximum length of the wheel path (footprint), the maximum width and contact area
for all 9 load-inflation combinations. As expected the maximum length was located
in the center tread and tended to decrease as the load or pressure increases. For all
cases the maximum width remained the same at 22.1 cm. On the other hand, the
contact area determined by the number of sensing elements (sensels) was deter-
mined for static and dynamic loading (speed of 2 km/h). For static loading the
contact area tended to increase as the load or pressure was increased while the
opposite behavior, but to a lesser degree, was observed for dynamic loading.
Finally, the peak contact pressure was significantly higher for static loading and
tended to increase as the load or pressure was increased. The same trend, but in a
less significant degree, was observed for peak pressures under dynamic loading.

Table 2 Measured wheel path properties

Load Pressure Wheel max. length Contact area Peak pressure
(kN) (kPa) (cm) (cm2) (kPa)
Static/dynamic Static/dynamic Static/dynamic
40 586 23.1/22.5 292.7/233.4 1043/966
40 689 22.4/21.1 293.1/230.1 1296/1013
40 793 22.1/20.3 292.5/227.3 1537/1041
50 586 22.8/22.3 293.8/290.1 1050/960
50 689 23.8/21.6 303.6/285.1 1220/972
50 793 23.1/22.1 312.1/264.6 1516/1165
60 586 22.2/21.9 324.7/322.2 1099/961
60 689 25.4/25.1 335.3/317.6 1296/976
60 793 24.9/23.4 356.8/311.7 1434/1054
692 F. Leiva-Villacorta et al.

3 Modeled Pavement Responses

This part of the study involved prediction of pavement responses based on the
actual footprint pressure distribution and a circular load with uniform distribution
and then compare those results against measured tensile strain responses. In order to
do this, it was necessary to determine the layer moduli at the moment when the
tensile strains were obtained. Later, a finite element analysis software was used to
set the tire treaded shape and apply the parabolic pressure distribution.

3.1 Backcalculated Layer Moduli

Multi-depth deflectometer (MDD) data were used to determine the progression of

the pavement layer moduli. This was done by applying the method of equivalent
thickness (Ullidtz 1987) whereby the thickness of the structure is transformed into a
single layer. This transformation is done using Odemark’s methodology and cal-
culation of stresses, strains and deflections were performed using Boussinesq’s
theory.
Figure 5 shows an example of the backcalculated layer moduli for the different
layers for one of the test tracks as function of equivalent load repetitions in millions.
A good correlation was obtained between measured and estimated deflections and a
small deviation from the line-of-equality indicated the criteria to perform back-
calculation was sound and ensured that the methodology was successfully applied
to each particular data set. When the tensile strains were obtained and the
load-pressure was applied, the modulus of the asphalt concrete layer was estimated
to be 1400 MPa, the modulus of the granular base and subbase were 125 MPa and
the modulus of the subgrade was 50 MPa.

1
Backcalculated Moduli, MPa

10000 HMA Base

0.9
Subbase Subgrade
MDD Deflection, mm

0.8
0.7
1000
0.6
0.5
0.4
100 Measured MDD1-25
0.3 Predicted MDD1-25
0.2 Measured MDD1-250
Predicted MDD1-250
0.1 Measured MDD1-520
10 Predicted MDD1-520
0 1 2 3 4 0
0 1 2 3 4
Equivalent Load Repetitions (40 kN) Equivalent Load Repetirions (40 kN)

Fig. 5 MDD backcalculated layer moduli

Influence of Tire Footprint Area and Pressure Distribution … 693

1400 MPa

125 MPa

50 MPa

Pavement structure Model meshing Footprint and pressure design

Fig. 6 Example of the evaluated structure

3.2 Load and Structure Modeling

A finite element analysis was used to model the tire treaded shape and to apply the
parabolic pressure distribution. The three-dimensional (3D) finite-element based
software package EverStressFE 1.0 was used for the analysis of the flexible
pavement section subjected to the various wheel/axle load combinations. This
software takes advantage of symmetry and utilizes a 1/4 pavement configuration. In
this case, a dual tire load was simulated either with a parabolic approximation of the
actual treaded load and a circular load to estimate pavement responses. An example
of the modeled pavement structure, with the model meshing and the designed
footprint and pressure distribution is given in Fig. 6. All the modeled load and
pressure distribution for the treaded tire load where done under dynamic loading
results because the actual responses were obtained under dynamic loading.

4 Measured Versus Predicted Strains

Figures 7 and 8 are examples of the measured and predicted longitudinal and
transverse tensile strains obtained at the asphalt concrete/base interface. Measured
strain signals exhibit a viscoelastic behavior with a significant difference in the
shape of the signal (asymmetrical) when the wheel load approaches the sensor
location and when it moves away from the sensor. In contrast, predicted strain
responses based on layered elastic properties have a symmetrical behavior.
694 F. Leiva-Villacorta et al.

500
Predicted Long. Circular
400 Predicted Trans. Circular
Predicted Long. Treaded
300 Predicted Trans. Treaded
Microstrain

Measured Transverse
200 Measured Longitudinal

100

-100
0 1 2 3 4 5 6
Distance, m

Fig. 7 Measured versus predicted tensile strains at 40 kN and 689 kPa

650
Predicted Long. Circular
550 Predicted Trans. Circular

450 Predicted Long. Treaded

Predicted Trans. Treaded
Microstrain

350
Measured Transverse
250
Measured Longitudinal
150

50

-50

-150
0 1 2 3 4 5 6
Distance, m

Fig. 8 Measured versus predicted tensile strains at 60 kN and 689 kPa

Nevertheless, predicted strain responses tended to reproduce fairly well the shape of
the measured strain signal. Moreover, the treaded tire model exhibited a better
match of the measured signal for both longitudinal and transverse strain. The cir-
cular load with uniform pressure distribution tended to overestimate the peak
Influence of Tire Footprint Area and Pressure Distribution … 695

longitudinal strain and tended to underestimate the minimum compressive strain of

the transverse signal as shown in Figs. 7 and 8.
Figure 9 shows a comparison between measured and predicted responses. As
mentioned above, the uniform circular load tended to overestimate the peak lon-
gitudinal tensile strain, specifically at high load-pressure combinations, while the
non-uniform load provided a better agreement with actual peak values (closer to
equality). On the other hand, both load configurations tended to underestimate the
transverse peak tensile strain at low load-pressure combinations.
Another type of response measured in the field were deflections from MDD
sensors. These deflections obtained at the surface of the pavement and 370 mm
(bottom of the granular base) deep in the structure were also compared against
predicted deflections. In this case both load configurations tended to match fairly
well the actual deflections. Table 3 lists the critical horizontal strains at the bottom
of the AC layer and deflections at two different depths of the pavement.

650 110
Predicted Transverse MicroStra
Predicted Longitudinal MicroStr

600 Circular
100 Treaded
550 Equality
90
500

450
80

400 Circular
Treaded 70
350 Equality

300 60
300 350 400 450 500 550 600 650 60 70 80 90 100 110
Measured Longitudinal MicroStrain Measured Transverse MicroStrain
Longitudinal Microstrain Transverse Microstrain
1
Predicted Deflections, mm

Circular
0.9 Treaded
Equality
0.8

0.7

0.6

0.5
0.5 0.6 0.7 0.8 0.9 1
Measured Deflections, mm
Deflections

Fig. 9 Measured versus predicted responses

 696

Table 3 Predicted and measured responses

Load Pressure Load Modeled response Measured response
(kN) (kPa) type εlong εtrans Surface Deflection @ εlong εtrans Surface Deflection @
(×10−6) (×10−6) deflection (mm) 370 mm (×10−6) (×10−6) deflection (mm) 370 mm
40 586 Circular 416.8 65.42 0.636 0.511
40 689 Circular 422.2 64.9 0.635 0.512
40 793 Circular 425.8 65.1 0.634 0.513
40 586 Treaded 392.4 66.1 0.633 0.507 401.3 80.2 0.644 0.512
40 689 Treaded 409.6 66.54 0.639 0.501 415.4 80.9 0.648 0.512
40 793 Treaded 438.6 71.66 0.644 0.511 427.6 81.5 0.648 0.513
50 586 Circular 507.9 82.28 0.796 0.637
50 689 Circular 518.2 81.9 0.796 0.638
50 793 Circular 524.5 81.63 0.794 0.639
50 586 Treaded 432.6 80.48 0.778 0.629 415.5 94.9 0.785 0.611
50 689 Treaded 462.0 81.69 0.784 0.632 437.9 95.5 0.781 0.615
50 793 Treaded 493.0 83.01 0.791 0.634 456.3 96.8 0.788 0.621
60 586 Circular 593.7 99.31 0.952 0.762
60 689 Circular 608.2 98.25 0.955 0.764
60 793 Circular 618.3 98.4 0.956 0.766
60 586 Treaded 461.4 96.65 0.924 0.752 490.6 99.8 0.921 0.733
60 689 Treaded 488.2 93.85 0.927 0.753 512.2 100.1 0.926 0.737
60 793 Treaded 544.4 100.6 0.942 0.759 560.1 100.9 0.933 0.742
F. Leiva-Villacorta et al.
Influence of Tire Footprint Area and Pressure Distribution … 697

5 Statistical Analysis

Tire inflation pressure (586, 689 and 793 kPa), tire load (40, 50 and 60 kN) and
load type (circular and treaded) are the three loading parameters on which these
tire-pavement contact stress predicted responses were based. All these parameters
were analyzed and compared to study the effects of tire pressure on pavement
responses. An analysis of variance (ANOVA) with a two way interaction model
was used to compare sample means for all the different treatments and to evaluate
their effects on pavement responses. Table 4 shows the test results for the effects of

Table 4 ANOVA of predicted responses

Response Source df SS MS F p
Long. strain Load 2 54,602.7 27,301.4 1161.16 <0.5
Pressure 2 4821.5 2410.8 102.53 <0.5
Type 1 14,643.3 14,643.3 622.8 <0.5
Load × type 2 7616.9 3808.4 161.98 <0.5
Pressure × type 2 1697.4 848.7 36.1 0.003
Error 4 94 23.5
Total 17
S 4.8489 R2 99.89 % R2 (adj) 99.52 %
Trans. strain Load 2 2925.33 1462.66 1527.59 <0.5
Pressure 2 16.06 8.03 8.38 0.037
Type 1 0.64 0.64 0.67 0.46
Load × type 2 16.51 8.25 8.62 0.035
Pressure × type 2 20.35 10.17 10.63 0.025
Error 4 3.83 0.96
Total 17
S 0.9785 R2 99.87 % R2 (adj) 99.46 %
Surface Load 2 0.280602 0.140301 53166.72 <0.5
deflection Pressure 2 0.000149 0.000074 28.19 0.004
Type 1 0.00047 0.00047 178.19 <0.5
Load × type 2 0.000548 0.000274 103.85 <0.5
Pressure × type 2 0.000154 0.000077 29.2 0.004
Error 4 0.000011 0.000003
Total 17
S 0.00177 R2 99.99 % R2 (adj) 99.97 %
Deflection @ Load 2 0.187751 0.093876 19997.2 <0.5
370 mm Pressure 2 0.000059 0.00003 6.3 0.014
Type 1 0.000228 0.000228 48.47 <0.5
Error 12 0.000056 0.000005
Total 17
S 0.002166 R2 99.97 % R2 (adj) 99.96 %
698 F. Leiva-Villacorta et al.

all treatments on the horizontal tensile strains at the bottom of the AC layer and the
two deflection responses.
The results indicated that the variability observed for the predicted longitudinal
tensile strain, the transverse tensile strain and the surface deflection can be
explained by the statistical difference of load level, pressure, the type of load and
the interaction between type of load and load and pressure at a confidence level of
95 %. On the other hand, the variability of the predicted deflection at 370 mm deep
into the structure can be explained by the statistical difference of the three main
treatments at a confidence level of 95 %.
As expected, the tire load level is the main variable that can be used to explain
the differences in all the analyzed pavement responses. In second order, the load
type which is basically a comparison between uniform and non-uniform pressure
distribution. This variable did not affect significantly the transverse peak tensile
strain.

6 Conclusions

The following conclusions were based on the results of this study:

• In general, the results indicated that the contact area for dynamic loading
decreased for a given load as the inflation pressure was increased.
• The contact pressure patterns obtained for the five-rib tire indicated that higher
pressures were obtained for the inner ribs based on the controlled variables for
the type of tire evaluated in this study.
• Analytical pavement modeling can be significantly enhanced by using measured
tire-pavement contact pressure distribution. With measured tire-pavement con-
tact pressure data, current available pavement analysis programs can more
accurately predict actual pavement responses.
• The conventional method that assumes uniform contact stress over a circular
area tends to overestimate the pavement responses at high tire pressures.
• Statistical analysis confirmed that pavement responses were significantly related
to tire pressure distribution. The ANOVA tests showed that the horizontal strain
in the longitudinal direction at the bottom of the asphalt concrete layer and
pavement deflections obtained at two different depths are highly sensitive to the
pressure distribution.
• Overall, the correct tire contact pressure shape and distribution are needed in
order to accurately predict pavement responses. It is recommended to use
analytical tools based on more representative material properties (viscoelastic
and non-linear properties) to obtain an even better prediction.
Influence of Tire Footprint Area and Pressure Distribution … 699

References

Akram, T., Scullion, T., & Smith, R. E. (1993). Using the multidepth deflectometer to study tire
pressure, tire type, and load effects on pavements. Research report 1184-2 (Vol. 2). College
Station, TX: Texas Transportation Institute, Texas A&M University.
Al-Qadi, I. L., Loulizi, A., Janajreh, I., & Freeman, T. E. (2002). Pavement response to dual tires
and new wide-base tires at same tire pressure. Transportation Research Record, 1806, 38–47.
Baker Harris, B., Buth Michael, R., & Van Deusen David, A. (1994). Minnesota Road research
project: Load response instrumentation installation and testing procedures. Minnesota
Department of Transportation.
Beer, D. et al. (1999). Towards improved mechanistic design of thin asphalt layer surfacing base
on actual tire/pavement contact stress-in-motion (SIM) data in South Africa. In 7th conference
on asphalt pavements for Southern Africa, South Africa.
El-Basyouny, M. M. (2004). Calibration and validation of asphalt concrete pavements distress
models for 2002 design guide. Ph.D. dissertation, Arizona State University.
Heavy Vehicle Simulator. (2015). Monitoring of test sections and instrumentation. Retrieved April
2015. http://www.gautrans-hvs.co.za/
Huang, Y. H. (1993). Pavement analysis and design. Englewood Cliffs, NJ: Prentice-Hall.
Leiva-Villacorta, F., Aguiar-Moya, J. P., & Loria-Salazar, L. G. (2013). Ensayos acelerados de
pavimento en Costa Rica. Infraestructura Vial, 15, 15–19.
Leiva-Villacorta, F., Aguiar-Moya, J. P., & Loria-Salazar, L. G. (2015). Accelerated pavement
testing first results at the Lanammeucr APT facility. In 94th annual meeting. Washington, DC:
Transportation Research Board.
Machemehl, R. B., Wang, F., & Prozzi, J. A. (2005). Analytical study of effects of truck tire
pressure on pavements using measured tire-pavement contact stress data. In CD-ROM
proceedings for the 84th TRB annual meeting. Washington, DC: Transportation Research
Board, January 9–13, 2005.
Mateos, A., & Snyder, M. B. (2002). Validation of flexible pavement structural response models
with data from the Minnesota Road Research Project. Transportation Research Record, 1806,
19–29.
NCHRP. (2004a). Guide for mechanistic-empirical design of new and rehabilitated pavements
structures. Final document, Appendix RR finite element procedures for flexible pavement
analysis. Washington, DC: National Cooperative Highway Research Program.
NCHRP. (2004b). Guide for mechanistic-empirical design of new and rehabilitated pavements
structures. Final report, part 3 design analysis. Washington, DC: National Cooperative
Highway Research Program.
Prozzi, J. A., & Luo, R. (2005). Quantification of the joint effect of wheel load and tire inflation
pressure on pavement response. In CD-ROM for TRB 2005 annual meeting. Washington, DC:
Transportation Research Board.
Roque, R., et al. (2000). Evaluating measured of tire contact stresses predict pavement response
and performance. Transportation Research Record, 1716, 73–81.
Sebaaly, P. E., & Tabatabaee, N. (1992). Effect of tire parameters on pavement damage and
load-equivalency factors. ASCE Journal of Transportation Engineering, 118(6), 805–819.
Sibaja, L. M. (2014). Determinación de presión de neumáticos de vehículos de carga típicos de
Costa Rica. Costa Rica: Trabajo Final de Graduación, Universidad de Costa Rica.
Siddharthan, R. V., Krishnamenon, N., El-Mously, M., & Sebaaly, P. E. (2002). Investigation of
tire contact stress distributions on pavement response. ASCE Journal of Transportation
Engineering, 128(2), 136–144.
Ullidtz, P. (1987). Pavement analysis. In Development in civil engineering (Vol. 19). Amsterdam,
The Netherlands: Elsevier.
700 F. Leiva-Villacorta et al.

Wang, F., & Machemehl, R. B. (2006). Mechanistic-empirical study of effects of truck tire
pressure on pavement using measured tire-pavement contact stress data. In 85th TRB annual
meeting. Washington, DC: Transportation Research Board.
Weissman, S. L. (1999). Influence of tire-pavement contact stress distribution on development of
distress mechanisms in pavements. Transportation Research Record, 1655, 161–167.
Instrumentation of an Innovative
Pavement Section on Motorway A10

Ngoc Son Duong, Juliette Blanc and Pierre Hornych

Abstract Recently, the French motorway company Cofiroute has decided to

reconstruct the slow lane of a section of the motorway A10, located west of Paris.
This section supports a very heavy traffic, of about 4000 trucks/day (corresponding
to 4800 french standard equivalent 130 kN axle loads). The solution adopted for the
reconstruction consists in retreating in situ, with a hydraulic binder, the existing
bituminous subbase, and adding new base and surface layers, incorporating 30 % of
recycled materials from the old pavement. To validate this novel solution, and in
particular the in situ retreatment, Cofiroute asked IFSTTAR (the French Institute of
Science and Technology for Transport, Development and Networks) to instrument
this pavement structure. Four sections were instrumented with classical strain gage
sensors, placed at different levels in the pavement (subgrade, treated subbase,
bituminous base), and temperature sensors. On a fifth section, a more innovative
instrumentation, including geophones, crack opening sensors, and temperature
sensors was installed, associated with a remote data acquisition system. The
instrumentation was installed during construction, in November 2011, and most
transducers are still functioning. The paper presents the monitoring of these
experimental sections. The strain gage measurements were used to fit a mechanical
model of the pavement structure, and to evaluate the modulus of the different
pavement layers, and their evolution with time. The results confirmed the good
performance of the in situ retreated subbase, and validated the technical choice of
the road owner. The geophones are very sensitive sensors, which measure vertical
displacement velocity. Different methods of treatment of the geophone measure-
ments were proposed, to monitor pavement deflections, and also to identify and
classify truck silhouettes, and estimate truck loads.

Keywords Pavement monitoring
Strains
Deflections
Geophones
Data

acquisition

N.S. Duong
J. Blanc
P. Hornych (&)

IFSTTAR, LUNAM Université, Route de BOUAYE CS4,
44341 Bouguenais Cedex, France
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 701

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_46
702 N.S. Duong et al.

1 Introduction

The motorway A10 constructed in 1972, and located west of Paris, supports a very
heavy traffic of about 4000 trucks/day. This motorway was widened to 2 × 4 lanes
in 1994. After almost 20 years of service, the slow lane presented important
deteriorations. After considerable investigations and local repairs (Hug et al. 2015),
the motorway company, Cofiroute, decided to reconstruct the slow lane. For eco-
nomic and environmental reasons, the chosen solution consisted in retreating
in situ, with a hydraulic binder, the existing bituminous subbase. This subbase was
then covered with new base and surface layers, incorporating 30 % of recycled
materials from the old pavement. To validate this novel solution, Cofiroute asked
IFSTTAR to instrument the pavement section, to monitor its behaviour and thus
reassure the project owner about its technical solution.

2 Instrumentation of the A10 Motorway

2.1 Pavement Structure

The structure of motorway A10 is composed of four pavement layers:

• 2.5 cm of very thin asphalt concrete (VTAC) (elastic modulus E = 3000 MPa at
15 °C, 10 Hz)
• 11 cm of high modulus asphalt concrete (HMAC) (E = 14,000 MPa at 15 °C,
10 Hz)
• 11 cm of high modulus asphalt concrete (HMAC) (E = 14,000 MPa at 10 Hz,
15 °C)
• 30 cm of in situ retreated material containing 7 % of hydraulic binder, called
Recyvia
The subgrade consists of a layer of sand treated with cement, over a sandy clay.
The pavement is designed for 20 years.

2.2 Pavement Instrumentation

Two types of instrumentation were installed on this pavement section: (Fig. 1).
• A classical instrumentation, consisting of strain gages and temperature sensors,
with on-site data acquisition.
• A more innovative instrumentation, comprising geophones, crack opening and
temperature sensors, with a remote data acquisition system.
Instrumentation of an Innovative Pavement Section … 703

Fig. 1 Instrumentation plan on the motorway A10. a instrumentation with gauges on four
sections PR 15+700, PR 15+900, PR 16+850 and PR 17+100; b instrumentation with remote
acquisition on section PR 16+016

Four sections were instrumented with strain gages. Each section was equipped
with the following sensors (Fig. 1):
• Two vertical strain gages, placed at the top of the subgrade.
• Two longitudinal strain gages, placed at the bottom of the Recyvia material,
treated with cement;
• Two longitudinal strain gages placed at the top of the Recyvia material.
• One longitudinal strain gage, placed at the bottom of the high modulus asphalt
base.
On a fifth section, a different instrumentation was installed on the top of the
cement treated Recyvia subbase, close to a pre-crack sawn in the upper part of this
subbase (Fig. 1). This instrumentation included:
• A crack opening sensor, consisting of a non-contact inductive displacement
transducer.
• Two geophones, placed vertically on each side of the crack, under the wheel
path.
• Two Temperature probes (PT100) installed at the top of the cement-treated layer
and at the top of the high modulus asphalt mix base.
On this section, the instrumentation was connected to a wireless data acquisition
system, called PEGASE developed by IFSTTAR (Sohm et al. 2012a).
The PEGASE platform consists of a “mother board” which integrates a computa-
tion processor (Analog Device Blackfin BF537), data storage, a Wi-Fi communi-
cation module and a small and low-power GPS receiver to ensure localization and
time synchronization. It is associated with a set of pluggable “daughter boards”
which include conditioners and analog to digital converters for the different
704 N.S. Duong et al.

Fig. 2 The wireless data

acquisition system PEGASE
developed by IFSTTAR

transducers (Fig. 2). The PEGASE system is very compact, modular, and has low
power consumption. It has been used for various applications: bridge inspection,
cable monitoring, pavement monitoring (Sohm et al. 2012b).

2.3 Characteristics of the Sensors

Figures 3 and 4 present some of the sensors installed in the pavement. The strain
gages installed in the subgrade and in the Recyvia layer were Kyowa sensors
KM-120-120-H211W1M3. The ones installed vertically in the subgrade were
equipped with disks (diameter 8 cm) at their extremities. Different sensors, con-
sisting of a thin aluminum stripe, equipped with strain gages, were placed in the
high modulus asphalt layer (Fig. 3). These aluminum stripes were chosen, because
of their better resistance to high temperatures (up to 180 °C) than the Kyowa
sensors. However, some of these aluminum gages failed during the compaction of
the HMAC layer, due to the very high compaction energy.

Fig. 3 Horizontal strain gage installed in the Recyvia layer (a) and vertical strain gage equipped
with disks installed in the subgrade (b)
Instrumentation of an Innovative Pavement Section … 705

Fig. 4 Non-contact inductive displacement sensor (a) and geophone (b)

Figure 4 presents the non-contact inductive displacement sensor and the geo-
phones installed on the section with remote data acquisition. The displacement
sensor (MDS-45-K-SA from Micro-Epsilon) consists of a transducer body, and a
small metallic target, placed horizontally on each side of the crack. The geophones
(Geo space GS-11D, sensitivity 100 V/m/s) were placed vertically at the top of the
Recyvia layer and were used to measure the vertical displacement velocity of the
pavement.

2.4 Pavement Monitoring

On the sections equipped with strain gages, measurements were made on site,
during four measurement campaigns: in November 2011, April 2012, December
2012, February 2014. Each time, the lane was closed to traffic, and loads were
applied to the pavement with a reference truck, with a known axle load. For the
three first campaigns, the reference vehicle was a deflectograph. It is a two-axle
truck, with the back axle loaded at 130 kN. For the last campaign, a five axle truck,
with tridem rear axles was used. The load of the second axle was 137 kN and the
load of the tridem rear axles was 260 kN.
During each measurement campaign, the truck passed several times over the
instrumented zone, at a slow speed of 3 km/h, with the right wheels positionned at
the top of the strain gages. The pavement temperature was measured during each
measurement sequence. On the section with remote data acquisition, the response of
the various transducers (geophones, crack opening sensor, temperature sensors), is
recorded continuously, under normal traffic. The geophone and crack opening
sensor measurements are recorded at a high frequency of 1 kHz, whereas the
temperature measurements are recorded at a low frequency, of one measurement
every 5 min.
706 N.S. Duong et al.

3 Mecanical Modelling of the Motorway A10

Pavement design calculations have been made to evaluate the characteristics of the
pavement layers. These calculations have been made with the french pavement
design software ALIZE. This software is based on the multi-layer linear elastic
model of Burmister (1943), and conforms to the french pavement design guide
(LCPC - SETRA 1994).
Figure 5 presents an example of vertical subgrade strains measured on section
PR 15+700 under the passage of a deflectograph at 3 km/h and 12 °C, in November
2011. Each strain signal presents two peaks, corresponding to the two axles of the
deflectograph. The maximum vertical strain is obtained under the rear axle.
The vertical strains measured at the top of the subgrade, on the four instrumented
sections, during three successive measurement campaigns are presented in Fig. 6.
The results show that:
• Sections PR 16+850 and PR 17+100, where the subgrade consists of fine sand,
present the highest vertical strains (between about 120 and 160 µstrain);
• Sections PR 15+700 and PR 15+900, where the subgrade is treated with cement,
present about three times lower strains (between about 20 and 40 µstrains);
• On the four sections, the strains generally decrease with time, indicating
probably a consolidation of the subgrade over time.
The main objective of the ALIZE calculations was to define a mechanical model
of the pavement, and estimate the modulus of the Recyvia layer, treated with a
hydraulic binder, and its evolution. These calculations were made with the fol-
lowing assumptions:

Fig. 5 Vertical strains measured at the top of the subgrade of section PR 15+700 under the
passage of a deflectograph at 3 km/h and 12 °C
Instrumentation of an Innovative Pavement Section … 707

Fig. 6 Vertical strains at the top of the subgrade of the four instrumented sections, during three
measurement campaigns

• For the subgrade; according to the results of Fig. 6, it was decided to separate
the sections with a natural soil, and the sections with a cement-treated soil. For
each case, ALIZE was used to fit the modulus of the soil, and its thickness.
• For the asphalt layers, as only the bottom of the first HMAC layer was
instrumented, and as several strain gages failed, it was not possible to fit the
moduli of the asphalt layers using the strain gage measurements. Therefore,
standard values of asphalt moduli given by the French pavement design standard
(NF P98-086) have been used. These standard values are given at 15 °C and
10 Hz and have been corrected, to take into account the real temperature and
frequency of the measurements.
• For the Recyvia layer, the longitudinal strains measured at the bottom of the
layer were used to fit the modulus.
In a first stage, two series of measurements, made in November 2011, at two
different temperatures, 2 and 12 °C were fitted with ALIZE, to determine the layer
moduli. The fitted model parameters are given in Table 1. The modulus obtained
for the natural soil is 100, and 800 MPa for the cement treated soil. The values of
modulus of the Recyvia are 6500 and 6000 MPa respectively, for the two sections.
Figure 7 presents a comparison between the measured and calculated longitu-
dinal strains at the bottom of the Recyvia layer. The ALIZE calculations are in good
agreement with the experimental measurements.
For the three last measurement campaigns, calculations were also performed
with the ALIZE model. In these calculations, the characteristics of the subgrade and
of the asphalt layers were kept constant (except for the temperature correction
applied to asphalt layer moduli), and only the modulus of the Recyvia layer was
varied. The elastic moduli fitted for the 4 measurement campaigns are presented in
708 N.S. Duong et al.

Table 1 Pavement layer moduli of motorway A10, fitted using the sensor measurements made in
November 2011
Layer Structure with natural soil Structure with cement treated soil
(2 °C, 0.5 Hz) (12 °C, 0.5 Hz)
Layer thickness e E (MPa) ν Layer thickness e E (MPa) ν
(m) (m)
VTAC 0.025 3220 0.35 0.025 1886 0.35
HMAC 0.11 11,294 0.35 0.11 7882 0.35
HMAC 0.11 11,294 0.35 0.11 7882 0.35
Recyvia 0.3 6500 0.35 0.3 6000 0.35
Soil 0.5 100 0.35 1 800 0.35

Fig. 7 Measured and calculated longitudinal strain at the bottom of the Recyvia layer
(measurements at 12 °C, 0.5 Hz)

Fig. 8. The results indicate that the moduli of the Recyvia layer increase with time,
varying from about 6000 to 6500 MPa during the first measurement campaign
(November 2011) up to 8000 to 8500 MPa in February 2014. This confirms that the
Recyvia material, treated with a hydraulic binder, is a slow setting material, and that
its modulus continues to increase slowly, 2 years after construction.
In addition, the results obtained in November 2011 (2 months after construction)
are in good agreement with the modulus measured in the laboratory, at 90 days,
which was equal to 6000 MPa. This confirms that the in situ properties of the
material are in good agreement with the laboratory mix design.
Instrumentation of an Innovative Pavement Section … 709

Fig. 8 Evolution of the elastic modulus of the Recyvia layer

4 Geophone Measurements

Geophones are sensors designed initially to record seismic movements. Their

principle is simple: they consist of a spring-mounted magnetic mass moving within
a wire coil. When the mass moves, an electrical current is induced in the coil. The
electrical signal is proportional to the displacement velocity (Oome 2009).
Geophones are characterised by their natural frequency, damping coefficient and
sensitivity. They have a passing band typically ranging from about 4 to 400 Hz.
Geophones are very sensitive sensors which present a good robustness, and a
relatively low cost. They are used in different domains, including measurement of
ground movements or machine vibrations. In this project, the idea was to use
gephones to measure the vertical displacement velocity of the pavement structure
under heavy vehicle traffic. This vertical velocity can then be integrated, to obtain
the vertical displacement of the pavement under load (similar to a deflection).
On the A10 motorway, two geophones have been installed at the top of the
cement treated subbase, about 30 cm apart, on each side of a pre-crack. The signals
of the geophones were recorded only when their amplitude exceeded a triggering
treshold. This led to record about 100 heavy vehicle signals per day. These signals
were then integrated to obtain the vertical displacement.
Figure 9 presents an example of signals recorded by the two geophones. The
upper part of the figure presents the raw signals (vertical velocity measurements)
and the lower part of the figure the integrated signals (vertical displacement mea-
surements). The results show that:
• The signals of the two geophones are very similar, indicating a good repeata-
bility of the measurements;
710 N.S. Duong et al.

Fig. 9 Example of raw signals and integrated signals of the two geophone under the passage of a
truck

• The integration leads to a displacement amplitude of 0.09 mm, comparable to

the pavement deflection, measured with a curviameter in April 2013, at a
temperature of 15 °C, which was also about 0.09 mm.
The geophones detect with a good accuracy the small vertical displacements
(about 100 µm) produced by the traffic passing on this rigid motorway section.
A first possible application of these measurements could be the detection, and
counting, of different truck silhouettes. This is illustrated on Figs. 10 and 11, which
show signals obtained for two different vehicles: a three axle truck, and a five axle
truck, with tridem rear axles. The different axles of each truck can easily be
identified.
The second possible application of geophones is the evaluation of pavement
displacements or deflections, by integration of the geophone signals. As for the
strain gage measurements, the integrated geophone signals obtained on the A10
section have been compared with displacements calculated with the ALIZE soft-
ware. Figure 12 presents a comparison of vertical displacements measured under
the passing of a deflectograph, (two-axle truck, with the rear axle loaded at
130 kN), and calculated with ALIZE. These measurements have been made on
November 15th, 2011, at a temperature of 12 °C and a speed of 3 km/h.
The experimental, peak-to-peak amplitudes of displacement measured by the
two geophones G1 and G2 are 6.8 mm/100 and 6.5 mm/100 respectively, under the
rear axle. The displacement calculated with ALIZE is 5.8 mm/100. These values
are of the same order of magnitude. However, the shape of the geophone signals
differs significantly from the calculated displacements. These signals present a
positive part, which seems unrealistic, because it corresponds to an upward dis-
placement (uplift) of the pavement. On the contrary, the calculated displacement is
always negative (downward displacement). This alteration of the geophone signals
Instrumentation of an Innovative Pavement Section … 711

Fig. 10 Integrated signals of the geophones under a three-axle truck

Fig. 11 Integrated signals of the geophone under a five-axle truck

may be due the response of the geophones, which is attenuated at low frequencies,
and also to errors introduced by the integration of the signal. Work is under way to
improve the correction and integration of the geophone measurements, to obtain
more accurate vertical displacement values.
712 N.S. Duong et al.

Fig. 12 Comparison of measured and calculated vertical displacements at the top of Recyvia
layer. Measurements made on 11/15/2011 (12 °C and 10 Hz)

The measurements of the geophones under real traffic have been recorded with
the remote data acquisition system between November 2011 and February 2014.
Figure 13 presents the evolution of the monthly average amplitudes of the vertical
displacements measured by the two geophones, and the temperature measured at
the top of the Recyvia layer. The results show that:

Fig. 13 Monthly average values of displacements measured by the geophones and temperatures
at the bottom of the HMAC layer (from November 2011 to February 2014)
Instrumentation of an Innovative Pavement Section … 713

• The average monthly displacement amplitudes vary between 10 mm/100 and

7 mm/100. The overall trend shows a progressive decrease of deflections over
the 2 year period. This is coherent with the progressive increase of the modulus
of the cement-treated layer.
• The difference between two geophone signals is small (about 0.5 mm/100).
• An increase of deflections is observed in the summer, when the pavement
temperature increases, due to the decrease of the moduli of the bituminous
layers.
These results show that it could be possible to use the average vertical dis-
placement amplitude measured by the geophones as an indicator of the overall
performance of the pavement, and of its evolution with time, similarly to the
deflection measured by an FWD or a deflectograph. An advantage of the monitoring
using the geophones is that the measurements are continuous.

5 Conclusion

This paper presents some results of the instrumentation of an experimental pave-

ment section recently reconstructed on motorway A10, near Paris. Four sections
have been instrumented with strain gages, to monitor the vertical strains in the
subgrade and the horizontal strains at the bottom of the bound layers. A fifth section
was instrumented with geophones, and a crack opening sensor.
This section of motorway A10 is also regularly surveyed by classical distress
measurements, and deflection measurements, using a deflectograph. For this very
rigid pavement, the deflectograph measurements are not sufficiently accurate to
estimate correctly the elastic modulus of the Recyvia layer. In addition, the project
owner did not allow to take cores from the site, to test the cement-treated Recyvia
layer in the laboratory. In this context, the in situ instrumentation, associated with
the ALIZE calculations allowed to verify the performance of the Recyvia layer, and
its evolution with time. The results confirmed the initial modulus of the Recyvia
material, which was very close to the value measured in the laboratory, and a gain
of modulus of about 30 % after 2 years. These results validated the choice of this
innovative technical solution.
The tests made with geophones placed on the top of the Recyvia layer indicated
that these transducers, which measure the vertical displacement velocity, are very
sensitive to small levels of displacement. These geophone measurements can be
used to identify the different truck silhouettes, and estimate vertical displacements,
by integration of their signals. Additional work is still needed to correct and validate
the measured displacements, by taking into account the real frequency response
function of the geophones, but it seems feasible to use geophones to make con-
tinuous measurements of pavement deflections under passing heavy vehicle traffic.
714 N.S. Duong et al.

References

Burmister, D. (1943). The theory of stresses and displacements in layered systems and application
to the design of airport runways. In Proceedings of the Highway Research Board (Vol 23).
Hug, C., Blanchard, J. Y., Hornych, P., Blanc, J., & Kerzrého, J. P. (2015), La route de 5ème
génération: le tronc commun A10-A11. Revue Routes/Roads, no. 366, April 2015.
Oome, A. J. J. A. (2009). Modeling of an electromagnetic geophone with passive magnetic spring.
Sensors and Actuators, A: Physical, 2009, 142–154.
LCPC - SETRA. (1994). Design of pavement structures. Technical Guide, December 1994.
Sohm, J., Kerzrého, J. P., Hornych, P., Blanchard, J. Y., Nicollet, P., Cottineau, L. M. (2012a).
Acquisition de données à distance pour évaluer l’état d’une structure de chaussée: la route
intelligente. Revue Générale des Routes et Aérodromes, no. 901, mai 2012.
Sohm, J., Hornych, P., Kerzrého, J. P., Cottineau, L. M., Le Cam, V., Hautière, N., et al. (2012b).
Remote monitoring of an experimental motorway section—an enabling technology of the 5th
generation road. International Journal of Pavement Research and Technology, 5(5), 289–294.
 Part IX
Accelerated Pavement Testing on Portland
Cement Concrete Pavements
Design, Instrumentation and Construction
of Bonded Concrete Overlays
for Accelerated Pavement Testing

Julio Paniagua, Fabian Paniagua, Angel Mateos, John Harvey

and Rongzong Wu

Abstract Bonded Concrete Overlay of Asphalt (BCOA) are being investigated for
rehabilitation of asphalt pavements in California’s dry climate as part of a coop-
erative project led by the University of California Pavement Research Center for
California Department of Transportation. Heavy Vehicle Simulator (HVS) is going
to be used to evaluate future performance and develop a design and construction
guide for Caltrans. This paper consists in a design of an Accelerated Pavement
Testing (APT) test track for analyzing the performance of the BCOA structures
under the use of a HVS. Besides it summarizes the findings from previous BCOA
experiences, the results from the concrete mix designs and initial modeling. Second,
a detailed design, construction and instrumentation plan is presented for the 15
sections that have been built in the test track. Construction of the test track already
occurred, therefore a brief summary of important facts and initial responses of the
instrumented sections are shown at the end.

Keywords Bonded concrete overlay
HVS
Rehabilitation

1 Introduction

Bonded Concrete Overlay of Asphalt (BCOA) are being investigated for rehabili-
tation of asphalt pavements in California’s dry climate as part of a cooperative
project of the California Department of Transportation, the Southwest Concrete
Pavement Association, and concrete materials suppliers and contractors. This paper
presents the results of identification of the critical variables affecting this type of
pavement from the literature, previous APT testing in California and trial slab
results, and the design, instrumentation and construction of an accelerated pavement

J. Paniagua
F. Paniagua
J. Harvey
R. Wu

University of California Pavement Research Center, Davis, CA, USA
A. Mateos (&)
Univesity of California Pavement Research Center, Berkeley, CA, USA
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 717

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_47
718 J. Paniagua et al.

testing (APT) test track to measure the performance of the BCOA structures using a
Heavy Vehicle Simulator (HVS). Fifteen test sections were built, from which 11
sections will be tested with the HVS and the remaining four will be monitored to
study environmental effects. The geometry of the concrete slabs varies in the sec-
tions and was established after performing finite element modeling for slabs
between 100 and 140 mm thick and the slab sizes of 1.8 × 1.8 m and 3.6 × 3.6 m.
The instrumentation installed in the structures includes: dynamic strain gages
near the surface and bottom of the concrete to measure horizontal strains under
traffic loading, static strain gages at two depths to measure shrinkage and tem-
perature effects, thermocouples at five different depths, moisture sensors at different
depths, two types of relative humidity sensors and joint displacement measuring
devices to measure vertical and horizontal joint displacement. The recently con-
structed test track has similar instrumentation as the trial slabs, which were con-
structed to obtain experience for this APT team in the construction process and
instrumentation requirements, which was successfully applied to the construction of
the test track.
The thin overlay structures have been constructed over previous HMA test
sections built and trafficked with the HVSs at the UCPRC-Davis facility as part of
previous experiments. Four different asphalt surface conditions were included in the
experiment because bonding of the concrete overlay to the asphalt below it has been
identified as a key parameter affecting the performance of BCOA. The conditions
were: milled, micromilled, no treatment of aged HMA, and overlay with gap-graded
rubberized hot mix asphalt (RHMA-G).

2 Objectives and Tasks

The primary objective of this research project is to develop recommendations and

guidance on the use of BCOA as a rehabilitation alternative in California.
Objectives and tasks required to meet this primary objective are listed in Table 1.
This paper summarizes the results of Tasks 1, 2, 4, 5 and the instrumentation,
construction and initial results of environmental loading of Task 6.

3 Determination of Possible Structures and Establishment

of Critical Factorial

The final set of variables in the plan for the APT sections have been selected after
comprehensive review and discussion with Caltrans and industry, and incorporate
Caltrans goals regarding materials, mix designs, and rehabilitation construction
window constraints, likely most viable industry practices, and the knowledge that
other agencies and researchers have developed from previous studies. An extended
Design, Instrumentation and Construction … 719

Table 1 Summary of objectives and tasks

Objective Task
Determine current knowledge and 1: Literature review
state-of-the-practice
Optimize concrete mix designs for BCOA in 2: Development of concrete mix designs
California
Determine the optimized rubberized asphalt 3: Development of mix designs for
concrete mix designs for use as base for rubberized asphalt for use as BCOA base
BCOA in California
Improve understanding of mechanisms of 4: Improved modeling of BCOA for cracking
BCOA cracking and develop design and development of recommended designs
recommendations in California
Obtain understanding of mechanisms of 5: Modeling of thin whitetopping for faulting
BCOA faulting and develop design and development of recommended designs
recommendations
Evaluate improved designs and estimate 6: Evaluation and performance estimates of
performance using APT improved designs using the heavy vehicle
simulator and mechanistic-empirical
analysis
Develop initial maintenance and 7: Development preliminary maintenance and
rehabilitation strategies, cost estimates and rehabilitation strategies, cost estimates, life
LCA inventory framework cycle assessment inventory framework
Evaluate improved designs under high-speed 8: Evaluation of improved designs on
highway traffic mainline highway at MnROAD (if
possible)
Provide final recommendations, guidelines 9: Final recommendations, report and
and report guidelines

literature review and analysis of previous BCOA projects around the United States
was performed. Based on that research four main variables were as being of highest
priority: concrete mix design, slab size, and concrete/asphalt interface surface
condition.
Caltrans initially identified four possible construction windows for BCOA, each
one leading to concrete mix requirements for the time required to reach the opening
time strength of 2.8 MPa (400 psi) (Rasmussen and McCullough 1998): 4 h for
overnight construction, 12 h for 24-h closure construction and continuous fast-track
construction, 12–24 h for weekend construction and 10 days for long-term lane
replacement closures as per Caltrans specifications for ordinary portland cement
concrete (PCC). These were later reduced to two opening times for the APT
experiment: four and 10 h, for which industry proposed a mix using Type II/V
portland cement with admixtures for the 10 h opening time and several mixes that
included rapid set cements with Type III portland cement or calcium-
sulfate-aluminum (CSA) for the 4 h opening time. Also included in the APT sec-
tions are a Type II/V internal cured concrete (ICC) mix with light-weight aggregate
(LWA) for 10-h opening time and another section using surface applied shrinkage
reducing admixtures (SRA) to reduce the drying shrinkage gradient. Industry and
720 J. Paniagua et al.

Prof. Jason Weiss (Oregon State University) were consulted regarding the LWA
and SRA sections. Industry supplied the mix designs that fulfill the agency
requirements (Roesler et al. 2008), which were then verified in the UCPRC
laboratory.
The slab geometry criteria were based on a configuration that could handle
medium- to high-traffic volumes. According to the literature review (Rasmussen
and Rozycki 2004; Burnham 2006; Vandenbossche and Fagerness 2002) and the
recommendations of University of Pittsburgh team members the recommended slab
size for BCOA overlays is about 1.8 by 1.8 m for thicknesses of around 140 mm
mainly, intended to keep the longitudinal joint far from the wheel path
(Vandenbossche and Barman 2010). Finite element analysis has shown that 1.8 m
square slabs have helped to reduce curling, warping and load-related flexural
stresses (Roesler et al. 2008). Modeling analysis showed that the influence of the
composite concrete/asphalt action would have been increased if the thickness was
reduced to 114 mm. Previous knowledge explains that there are two main failure
modes that mainly vary according to the slab size (Minnesota 2012). Mid-panel
longitudinal crack typically occurs in 1.8 m by 1.8 m slabs. On the other hand, for
3.6 by 3.6 m slabs cracking typically occurs in the transversal direction at mid-slab.
Use of widened slabs that keep heavy trucks off the right side edge of the slab is of
interest to Caltrans, therefore 2.4 by 2.4 m slabs with the HVS wheel kept 0.6 m
from the edge were also included. Even though stresses caused by temperature and
moisture gradients might increase with these slabs, the reduction of stress due to
traffic load is considerable. In summary, three slab sizes and two thicknesses were
included in the APT test track design.
According to previous research and project discussions, most agencies that have
used BCOA agree on the need for milling of the existing asphalt (Mu and
Vandenbossche 2011). Nevertheless, several authors stated that sweeping and
cleaning the surface without milling is enough to allow good bonding (Gucunski
1998) to an unknown degree. Therefore, additional treatments of micromilling and
milling as well as just cleaning were included. Previous experience had indicated
bad experience when placing BCOA on new asphalt mixes, mainly due to the lack
of shearing capacity of the new mix. Nevertheless, it was Caltrans desire to study
the effect of new rubberized asphalt surfaces (Harvey et al. 2014) between the
existing asphalt and concrete overlay which was also included in the test track
sections.
As mentioned previously, the variables include four concrete mix designs, two
slab thickness, three slab sizes, and four asphalt base conditions (new RHMA,
swept old HMA milled, micromilled). Establishing a full factorial design was not
affordable due to time, space and cost constraints. Therefore, finite element mod-
eling, expert criteria and previous experience were used to determine the APT
experiment by considering at the same time Caltrans specifications, industry inputs
and in-place condition of the test track. Figure 1 shows the comparisons that were
included in the final design for the APT test sections. Dowels and fiber reinforced
concrete were considered for the experiment, but were not included due to the space
constraints.
Design, Instrumentation and Construction … 721

Fig. 1 APT test section partial factorial design (All slab thicknesses are 140 mm, HMA thickness
is 120 mm, base type is swept existing HMA unless other noted)

These test sections were constructed on top of two standard-width lanes each 73
meters long. The test sections were built on top of previous HVS test sections with a
generally homogeneous full-depth recycled section with no stabilization (FDR-NS)
layer 250 mm thick with a previously trafficked asphalt overlay on top. The
recycled subbase was created from a previous pavement with 120 mm of warm-mix
asphalt concrete and 130 mm of granular base (Jones et al. 2014). The FDR-NS is
on top a prepared subgrade with a stiffness modulus of 60 MPa. These two sections
act as an identical structural bed for 15 sections with three different asphalt layers
that serve as the platform for the BCOA. Eleven of the 15 sections will be trafficked
by the HVS and four will be used to monitor structural response to drying shrinkage
and the environment to calibrate the models for shrinkage, warping and curling
effect. Three of the four environmental monitoring sections have an FDR sub-base
modified with 5 % asphalt emulsion (FDR-AE), however, the base type under the
asphalt should not affect the environmental monitoring results.
The final test track experimental design to complete the comparisons shown in
Fig. 1 was defined after the modeling described below, and is shown in Table 2,
which summarizes the structural and geometric characteristics of each of the 15
sections. In section M it was placed SRA (shrinkage reducer admixtures) before
placing the curing compound.
In sections A through E and L through O the base used is an old Caltrans Type A
Hot-Mix Asphalt with 19 mm aggregate gradation with 15 % reclaimed asphalt
pavement (RAP) paved on November 14, 2012. The thickness of the asphalt on
sections A through C is of 120 mm while for all the other sections with old HMA
the thickness is 60 mm. Sections F through K have a newly placed layer of 30 of
12.5 mm Rubberized Gap Hot Mix Asphalt on top of conventional 19 mm HMA
722 J. Paniagua et al.

Table 2 Summary of test section characteristics

Sec. Sub-base Base Interface Concrete Slab size Response
treatment
A FDR-NS Old HMA Micromilling PCC Type II/V 1.8 m × 1.8 m Structural
120 mm (10-H) 140 mm
B FDR-NS Old HMA Micromilling PCC Type II/V 1.8 m × 1.8 m Structural
120 mm (10-H) 114 mm
C FDR-NS Old HMA Milling PCC Type II/V 1.8 m × 1.8 m Structural
120 mm (10-H) 114 mm
D FDR-NS Old HMA Micromilling PCC Type II/V 2.4 m × 2.4 m Structural
60 mm (10-H) 114 mm
E FDR-NS Old HMA Micromilling PCC Type II/V 1.8 m × 1.8 m Structural
60 mm (10-H) 114 mm
F FDR-NS New RHMA Micromilling PCC Type II/V 1.8 m × 1.8 m Structural
30 mm (10-H) 114 mm
New HMA
90 mm
G FDR-NS New RHMA Sweeping RSC Type III 1.8 m × 1.8 m Structural
30 mm (4-H) 114 mm
New HMA
90 mm
H FDR-NS New RHMA Sweeping RSC CSA (4-H) 1.8 m × 1.8 m Structural
30 mm 114 mm
New HMA
90 mm
I FDR-NS New RHMA Sweeping PCC Type II/V 1.8 m × 1.8 m Structural
30 mm (10-H) 114 mm
New HMA
90 mm
J FDR-NS New RHMA Sweeping PCC Type II/V 3.6 m × 3.6 m Struc. + Env.
30 mm (10-H) 114 mm
New HMA
90 mm
K FDR-NS New RHMA Sweeping PCC Type 3.6 m × 3.6 m Struc. + Env.
30 mm II/V-ICC (10-H)
New HMA 114 mm
90 mm
L FDR-NS Old HMA Micromilling PCC Type II/V 1.8 m × 1.8 m Environmental
60 mm (10-H) 114 mm
M FDR-AE Old HMA Micromilling PCC Type II/V 1.8 m × 1.8 m Environmental
60 mm (10-H) 114 mm
N FDR-AE Old HMA Micromilling RSC T.III (4-H) 1.8 m × 1.8 m Environmental
60 mm 114 mm
O FDR-AE Old HMA Micromilling RSC CSA (4-H) 1.8 m × 1.8 m Environmental
60 mm 114 mm
Design, Instrumentation and Construction … 723

with PG70-10 binder (Kocunik 2015). Caltrans believes that using rubber modified
mixes could improve BCOA performance under traffic and environmental loads;
however, bonding between RHMA-G and concrete has not been previously
investigated.
The RHMA-G was placed in the test track on October 13th, 2015. This impli-
cates that the surface layer was exposed to aging for approximately four months
prior to construction. It has been suggested that using new pavements as bases for
BCOA could lead to premature cracking in the concrete due to permanent defor-
mation of the asphalt layer under shear loading from traffic (Sheehan et al. 2004).
As mentioned previously, preparation of the existing pavement surface is gen-
erally considered necessary for improved performance (Winkelman 2005). The
reference procedure for the experiment is micromilling which creates grooves about
6 mm in depth. For the sections with new rubberized asphalt base only one has
been milled to test the hypothesis that placement of BCOA on newly milled asphalt
will cause a 50 % increase in critical load-induced stresses (Roesler et al. 2008;
Tarr et al. 1998).
Thickness is other geometric property that has been chosen based on reliable
information of other agencies. Usually 100 mm thickness has shown to decrease
reflection cracking and to reduce load-related stresses in the bottom of the structure
(Burnham 2006; Vandenbossche and Barman 2010). The selection of the thickness
must balance a number of factors, including the anticipated traffic loading, the
strength and stiffness of existing layer, the properties of the overlay concrete and the
degree of load transfer (Vandenbossche and Barman 2010). For this project,
140 mm has been chosen as the reference thickness; nevertheless, modeling has
shown that this thickness may be somewhat too thick when placed on the thin
asphalt concrete (60 mm) in the experiment.
The dry California environment increases drying shrinkage in the structure
compared to other parts of the country, and the depth that can dry out leading to
moisture shrinkage is around 50 mm (Burnham and Koubaa 2002) which could
mean up to 40 % of the total thickness of the BCOA structures. The ICC mix used
in section K uses pre-wetted lightweight aggregates consisting of expanded shales,
clays and slates. Even though there is an estimated 10 % increase in cost, the
implementation of this mix could be worthwhile if a significant reduction of drying
shrinkage and cracking is achieved.

4 Modelling of BCOA Structures

The purpose of this initial modeling effort was to estimate the testing time that will
be needed for each of the test sections. Each of the sections was modelled to
determine the stresses in the concrete under different loading conditions using
EverFE (Davids et al. 1998) which is a finite element software conceived for the
724 J. Paniagua et al.

analysis of rigid pavements. It has five main inputs which are geometry, materials,
loading, joints and meshing parameters. Loading of an HVS structure was modeled
first assuming a 40 kN dual wheel load with wander and field loading was modeled
second assuming a dual wheel-single axle.

5 Modelling Parameters

5.1 Geometric Parameters

Regarding the geometry parameters, a 3 × 3 slab scenario is the biggest that can be
built in EverFE. Three different slab sizes were used: 1.8 × 1.8 m, 3.6 × 3.6 m and
2.4 × 2.4 m. For the modelling process of EverFE, the first size is modeled with 6
slabs (2 × 3), the second size is modeled with 3 slabs (1 × 3) and the third and last
size is modeled with 6 slabs (2 × 3) as can be shown in Fig. 2.
Due to the software restriction, the only size that was able to be modeled as
constructed in the track in terms of numbers of in-lane slabs and adjacent slabs is
the structure with 3.6 × 3.6 m slab size. For the configuration of 1.8 × 1.8 m, a set
of 2 slabs wide by 5 in length was built in the track and for the 2.4 × 2.4 m slabs, a
set of 2 slabs wide by 4 slabs in length was constructed.
With respect to the layers, it was worked with PCC, AC, unbound base and
subgrade. The asphalt concrete layer was modeled with two different thicknesses
reduction, 10 and 25 mm in order to capture processes performed in the field track.

5.2 Material Parameters

The 90 days flexural strength from Caltrans specifications (Caltrans 2015) was
assumed to be 4.5 MPa (650 psi), the coefficient of thermal expansion used during
the modelling was assumed to be 0.9 με/°C (1.6 με/°F) and the specific gravity was
assumed to be 2.4 for all mixes. A sensitivity analysis was performed for the

Fig. 2 Sections geometries used in EverFE (Note 2.4 × 2.4 slab assumes loading kept 0.6 m from
edge)
Design, Instrumentation and Construction … 725

material properties of the different concrete mixes in order to determine sensitivity

of stresses to the following variables:
• The stiffness was assumed to be 27,000 MPa and variation of ±15 % was
considered.
• The drying shrinkage for the concrete is obtained using the Bazan equation
(Bazant 2000) which is also recommended in the 2004 MEPDG (National
Cooperative Highway Research Program 2004). For the 10-h mix, the ultimate
drying shrinkage assumed is 560 με, for the ICC mix a reduction of 50 % is
assumed, and for the 4-h mixes a value of 840 με is assumed.
The stiffness of the existing asphalt surface was obtained from 4-point bending
beam frequency sweep tests performed in November 2012 (Jones et al. 2014) the
data from which was fitted with a sigmoidal master curve. Because the asphalt mix
has been aged from the end of 2012 to the end of 2015, the CalME aging model
(2015) was used to estimate the later stiffness values. For the recently placed gap
graded rubberized asphalt mix, the estimated stiffness was obtained from the
UCPRC database for a gap graded mix with the same maximum aggregate size
(12.5 mm) and base binder grade (PG 64-16).
Environmental loads such as daily and seasonal temperature curling and seasonal
drying shrinkage play an important role in creating tensile stresses in the bonded
concrete overlay structure. The seasonal environmental loading times are much
longer than those of traffic loading, and these loading time effects can be difficult to
estimate for modeling. Stiffnesses of 100 MPa for daily changes was ssumed.
Lower stiffnesses for seasonal changes were obtained from extrapolation of the
master curve, (Fig. 3), and are lower than the stiffness of the FDR-NS base. The
150 MPa stiffness of the FDR was therefore assumed for the asphalt structure.

Fig. 3 Asphalt mixtures dynamic modulus master curves (20 °C Ref. temperature)
726 J. Paniagua et al.

The CalME temperature model (Lea and Harvey 2012), which is a one
dimensional finite element and finite differences procedure for determining tem-
peratures at any depth, was used for determining the asphalt temperature to estimate
the dynamic modulus. The final output of this process was a temperature histogram
with four categories that determine the temperature difference at 1/3 of the depth of
the asphalt base layer. The temperature histogram was accessed using estimated
pre-calculated temperatures based on a temperature history and thermal properties
using the Enhanced Integrated Climatic Model (Zapata and Houston 2008). The
environmental data used correspond to the first-order weather station at the
Sacramento International Airport, which is to closest to the test site at the UCPRC.
Two other parameters that were used for the modelling of the asphalt structure
are the Poisson ratio of 0.3 and specific gravity of 2.3, and both parameters were
used without a sensitivity analysis.
The Poisson ratio and specific gravity for the unbound layers were assumed to be
0.35 and 2.1 respectively, without sensitivity analysis. As mentioned above, the
FDR without stabilization had an approximate stiffness of 150 MPa.

5.3 Loading Parameters

The tires of the dual wheel were assumed to have a contact area of 190 × 200 mm,
40 kN load and 720 kPa inflation pressure. The assumed distance between the dual
tires is 170 mm. The wheel path mean value location was assumed to be 450 mm
from the outside edge of the outside tire to the edge of the slab and for the traffic
wandering, a normal distribution was assumed with a 250 mm standard deviation
(National Cooperative Highway Research Program 2004). Seven different trans-
verse positions were modeled and results for other wander positions were
interpolated.
From CalME Temperature Model (Zapata and Houston 2008) and the
Momentum Equivalence approach (Janssen and Snyder 2000), a temperature his-
togram was created for the different thicknesses of PCC. The resulting histogram
has six different bins for each of the overlays thicknesses.
The differential shrinkage occurring in the upper part of the concrete causes a
similar effect as temperature gradients when the concrete is colder at the top than
the bottom. From previous experience at the UCPRC, specifically APT sections in
the desert at Palmdale (Roesler et al. 1999), the differential shrinkage was assumed
to follow a bilinear model recommended by Rasmussen and McCullough
(Rasmussen and McCullough 1998; Du Plessis et al. 2002). A similar bilinear
model is also assumed by MEPDG (National Cooperative Highway Research
Program 2004) for the first 50 mm of the concrete. According to the MEPDG
shrinkage equation, only a 50 % of the ultimate shrinkage is developed after a
month, and 94 % occurs after 18 months. For the structural analysis presented in
this paper, it was considered a time of 10 months which will represent 90 % of the
ultimate shrinkage.
Design, Instrumentation and Construction … 727

5.4 Joint Parameters

Transverse joint aggregate interlock was modelled in EverFE by a joint stiffness

shear spring model which has units of stress divided by a displacement. A value of
10 MPa/mm that will result in full aggregate interlock was chosen for the structural
analysis due to the fact that the result will be the same as if the parameter was zero
shear stiffness because the bonding of the concrete to the underlying asphalt gives
to the structure full load transfer efficiency. Even though all the sections are
modeled as having full bonding, section B (which can be seen in Fig. 1; Table 2), is
modeled with both full bonding and no bonding of asphalt and concrete layers in
order to compare the results.
All of the longitudinal joints were also modeled assuming full load transfer
efficiency. Similar as in the transverse joints, this occurs when there is full boding
of the concrete and asphalt.
Fatigue life of the BCOA sections is estimated in this study by using the Riley
model:
 0:217
SR10:24 logðRÞ
logðNf Þ ¼
0:0112

where
SR is stress ratio (σ/MR)
R is reliability.

6 Modeling Scenarios

Each of the sections was modeled for two types of load condition, HVS and field
testing. In the HVS analysis, the effects of the testing machine shading the section
were also considered (in Fig. 4 the label HVS’ assumes thermal stresses without
shading while the label HVS is calculated with shading. Case “a” considers the
effect of drying shrinkage and case “b” was performed with an additional tem-
perature gradient to account for the effects of shrinkage.

6.1 HVS Modeling

Figure 4 shows the fatigue life of the pavement structure for each of the four dis-
tresses analyzed: bottom-up transverse cracking, top-down transverse cracking,
bottom-up longitudinal cracking and top down longitudinal cracking. It can be seen
that the critical distress is the top-down transverse cracking when drying shrinkage is
considered and bottom-up transverse cracking if drying shrinkage is not considered.
728 J. Paniagua et al.

Fig. 4 Field versus HVS performance of section B

It can also be seen that the case of HVS′ (not shaded) is more critical because will
undergo more thermal stress than the ones that the shaded structures obtain.

6.2 Field Modeling

Modeling of the expected field temperatures for Sacramento and a 80 kN dual

wheel axle type shows the same pattern as the results for the HVS, in which the top
down cracking is the critical distress when shrinkage is considered and bottom up
transverse crack for the case in which it is not.
Similar analysis was performed for each of the sections with their specific
geometric characteristics. A relationship between fatigue life and axle load is shown
in Fig. 5, it can be seen that for a maximum fatigue life of 300,000 repetitions, the
load needs to be between 80 and 130 kN. Based on this figure and from previous

Fig. 5 Expected fatigue life of section B

Design, Instrumentation and Construction … 729

experience, the dual wheel load of the HVS will start at 40 kN and will be increased
20 kN every 50,000 cycles and the fatigue life will be in between 150,000 and
350,000 uniaxial repetitions. Taking into consideration that HVS could perform
10,000 unidirectional cycles per day, the test duration for the section will be as
much as 1 month approximately and the total for the 11 sections will be roughly
1 year, which is the available time.

7 Instrumentation for HVS Sections for Critical

Performance Parameters

Figure 6 shows the layout for the different instrumentation systems. The layouts are
only shown using slabs of 1.8 m × 1.8 m., nevertheless, the configuration for the
slabs 2.4 m × 2.4 m and 3.6 m × 3.6 m is the same, what varies is the space
arrangement in each slab.
The total number of sensors used in the 15 sections is summarized in Table 3.
A total number of 673 instruments were installed, from which 458 will be

Fig. 6 Instrumentation systems layouts

Table 3 Total numbers of Instrument Structural Environmental Total

instruments placed
Thermocouples 165 90 255
Static strain gages 0 56 56
Dynamic strain 150 0 150
gages
LVDT 143 33 176
Relative humidity C 0 12 12
Relative humidity P 0 12 12
Moisture sensor 0 12 12
Total 458 215 673
730 J. Paniagua et al.

Fig. 7 Test sections

connected to the HVS during testing periods and the remaining 215 were
plugged-into data acquisition systems to record the environmental response from
sections J, K, L, M, N and O.

8 Trial Slabs Development

Once the structural and instrumentational design was performed, and before the
construction of the actual test track, four trial slabs were built to provide a first step
in the learning curve. Four different separated 1.8 × 1.8 m slabs were built with full
sets of instruments for measuring environmental and structural responses. Figure 7
shows a picture of the four trial slabs that were cast, and the different curing
procedures that were applied to the slabs.

9 Construction and Preliminary Data

Construction of test track took place on February 23rd and 25th. Each of the days,
one lane of the test track was built; Fig. 8 shows the resulting test sections. Out of
the 215 environmental instruments that were installed, a total number of one static
strain gage and five thermocouples sensors were broken or are not recording ade-
quate data. This really small quantity of sensors that are not working is due to a
specific protection plan that was followed during construction. In front of each
instrument, a shovel was placed, so it was not hit by the concrete when coming out.
Once the concrete started to reach the instruments, the concrete will tend to wrap
the device and after that the shovel could be taken out. Further care was just needed
to be taken with the concrete crew that was with the vibration task. A roller screed
Design, Instrumentation and Construction … 731

Fig. 8 Constructed test track

vibrator was used which was less harmful for the instruments, but when the con-
crete mix was drier, additional vibration was required from traditional vibrator. In
this cases the operator was instructed to insert it vertically and not horizontally
because the instruments could be damaged.
For the QC/QA testing conducted, there temperature was tracked in beams,
cylinders and curing boxes from initial time until demolding time at 24 h. Flexural
beam, strength and modulus of elasticity cylinders and coefficient of thermal
expansion were prepared for testing at twice the opening time and 10 days after
construction. Even though it was known that Type III and CSA mix will set very
fast, the QC/QA crew faced the challenge of finishing the specimens before the
concrete got very stiff. When trying to place the thermocouple in a Type III beam it
was not achieved because it has already hardened.
Figure 9 is included as an example of the type of data collected in the envi-
ronmental sections. This figure shows the strain at the top and at the bottom of the
concrete slabs for three different materials. Strain in the figure is the result of
moisture related shrinkage and also temperature change since the concrete sets. The
autogenous shrinkage (self-desiccation) is reflected as a vertical shift of both strain
values, at the top and at the bottom, while drying shrinkage reflects as a higher
strain (in absolute value) at the top compared to the bottom. As expected, type III
cement results in the highest autogenous shrinkage, while CSA cement resulted in
the lowest values. Drying shrinkage was quite low for all sections two weeks after
construction, due to the rainy weather. Still, recorded data seems to indicate that the
shrinkage reducing admixture application on the type II cement concrete reduces
drying shrinkage. This preliminary conclusion is supported by the relative humidity
measured at the top (20 mm below the surface) of the slabs, as reflected in Fig. 10.
732 J. Paniagua et al.

Fig. 9 Moisture related and thermal shrinkage

Fig. 10 Relative humidity in the concrete

Design, Instrumentation and Construction … 733

10 Summary and Conclusions

• The concrete mix designs chosen to be studied in this project represents the
common practices from industries in California and requirements and standards
from CALTRANS
• The APT structural variables of the test track built include: concrete layer
thickness (100 and 140 mm), type of cement (Type III, Type II/V, CSA, and
ICC), slab size, base surface preparation (milling and micromilling), asphalt
base type (RHMA and HMA) and environmental effects (shrinkage, thermal
gradients and seasonal changes).
• The modeling process provided a great starting point in order to predict which
will be the fatigue life of the 11 sections that are going to be tested with the
HVS, due to time constraints; the total time of testing will be 12 months.
Modeling also help to determine the overlay thicknesses that could be tested
fulfilling time constraints.
• The well-functioning and success of the BCOA structures relies on having a full
bonding between the PCC and the AC, otherwise the fatigue life will be reduced
considerably. This is why it is important to perform an appropriate treatment to
the asphalt layer before casting the concrete.
• Above 550 devices were installed in the test track. The intention is to gather
mechanical response data to understand and to fill the lack of knowledge of
bonding of asphalt and concrete layers, and BCOA performance.
• The fabrication of trial slabs resulted in important building aspects that were
implemented during the construction of the test track. Instrument protection
tasks learned during trial slabs turned into be very important during actual test
construction in order to keep the instruments safe.

References

Bazant, Z. P. (2000). Criteria for rational prediction of creep and shrinkage of concrete. ACI
Special Publications, 194, 237–260.
Burnham, T. R. (2006). Construction report for MnROAD thin whitetopping test cells 60–63.
No. MN/RC-2006-18.
Burnham, T., & Koubaa, A. (2002). Determining the coefficient of thermal expansion and
shrinkage of jointed concrete pavement. 2nd Annual Mn/ROAD Workshop, Minnesota.
CalME User Manual.
Caltrans (2015). Standard specifications section 40. California Department of Transportation,
State of California.
Davids, W., Turkiyyah, G., & Mahoney, J. (1998). EverFE: Rigid pavement three-dimensional
finite element analysis tool. Transportation Research Record: Journal of the Transportation
Research Board, 1629, 41–49.
Du Plessis, L et al. (2002). HVS test results on fast-setting hydraulic cement concrete, Palmdale,
California Test Sections, South Tangent. University of California Pavement Research Center.
Research Report: UCPRC-RR-2002-03.
734 J. Paniagua et al.

Gucunski, N. (1998). Development of a design guide for ultra thin whitetopping (UTW).
No. FHWA NJ 2001-018.
Harvey, J., Vandenbossche, J., Burnham, T. R., Khazanavich, L., Wu, R., Jones, D., & Li, H.
(2014). Pavement preservation: Workplan for the development of improved guidelines and
designs for thin whitetopping. University of California Pavement Research Center.
Janssen, D. J., & Snyder, M. B. (2000). Temperature-moment concept for evaluating pavement
temperature data. Journal of Infrastructure Systems, 6(2), 81–83.
Jones, D., Wu, R., & Louw, S. (2014). Full-depth recycling study: Test track construction and first
level analysis of phase 1 HVS and laboratory testing. University of California Pavement
Research Center Research Report: UCPRC-RR-2014-03.
Kocunik, M. (2015). UCPRC paving rubber mix. Teichert Materials and Quality Assurance,
Sacramento, California.
Lea, J. D. & Harvey, J. (2012). The simplified thermal modeling approach used in CalME.
Transportation Research Board 91st Annual Meeting.
Minnesota, D. O. T. (2012). Whitetopping design procedure. Minnesota Department of
Transportation, Office of Materials & Road Research, Pavement Engineering Section, March.
http://www.dot.state.mn.us/materials/pvmtdesign/docs/Whitetopping_Design_Procedures.pdf.
Mu, F., & Vandenbossche, J. M. (2011). Development of design guide for thin and ultra-thin
concrete overlays of existing asphalt pavements. Task 2: Review and selection of structural
response and performance models. No. MN/RC 2011-25.
National Cooperative Highway Research Program (2004). AASHTO mechanistic-empirical design
guide. NCHRP Project 1–37a, Transportation Research Board, Washington, DC.
Rasmussen, R. O., & McCullough, B. F. (1998). A foundation for high performance jointed
concrete pavement design and construction guidelines. Austin, TX: Transtec Consultants.
Rasmussen, R. O., & Rozycki, D. K. (2004). Thin and ultra-thin whitetopping. NCHRP Synthesis
of Highway Practice 338, National Cooperative Highway Research Program, National
Research Council, Washington, DC.
Roesler, J., Harvey, J., Hung, D., du Plessis, L., & Bush, D. (1999, October). Evaluation of
longer-life concrete pavements for california using both accelerated pavement and laboratory
testing. In Proceedings, international conference for accelerated pavement testing, Reno,
Nevada.
Roesler, J., Ioannides, A., Beyer, M., & Wang, D. (2008). Design and concrete material
requirements for ultrathin whitetopping. Publication FHWA-ICT-08-016, Illinois Center for
Transportation, University of Illinois at Urbana Champaign, Urbana.
Sheehan, M. J., Tarr, S. M., & Tayabji, S. D. (2004). Instrumentation and field testing of thin
whitetopping pavement in Colorado and revision of the existing colorado thin whitetopping
procedure. Report CDOT-DTD-R-2004-12, Colorado Department of Transportation, Denver.
Tarr, S. M., Sheehan, M. J., & Okamoto, P. A. (1998). Guidelines for the thickness design of
bonded whitetopping pavement in the state of Colorado. Report No. DTD-R-98-10, Colorado
Department of Transportation, Denver.
Vandenbossche, J. M., & Barman, M. (2010). Bonded whitetopping overlay design considerations
for prevention of reflection cracking, joint sealing, and the use of dowel bars. Transportation
Research Record: Journal of the Transportation Research Board, 2155(1).
Vandenbossche, J. M., & Fagerness, A. J. (2002). Performance, analysis, and repair of ultrathin
and thin whitetopping at Minnesota road research facility. Transportation Research Record:
Journal of the Transportation Research Board, 1809(1).
Winkelman, T. J. (2005). Whitetopping performance in Illinois. Illinois: Illinois Department of
Transportation, Bureau of Materials and Physical Research.
Zapata, C. E., & Houston, W. N. (2008). Calibration and validation of the enhanced integrated
climatic model for pavement design. NCHRP Report 602, Transportation Research Board.
Performance of Thin RCC Pavements
Under Accelerated Loading

Zhong Wu, Moinul Mahdi and Tyson Rupnow

Abstract Thin Roller Compacted Concrete (RCC) surfaced pavement has a

potential to be used as a cost-effective design alternative for a low-volume road where
heavy truck trafficking is often encountered, such as oil and gas exploration, logging
and agricultural activities. In this study three full-scale RCC pavement test sections,
each having a structure of 4-, 6-, or 8-in. RCC surface over an 8.5-in. soil cement base,
were tested at the Louisiana accelerated loading facility. The objective was to eval-
uate the structural performance and load carrying capacity of thin RCC pavements
under accelerated pavement testing. A sequence loading of using dual-tire loads of 9-,
16-, 20-, 22-, and 25-kip was applied on each RCC section till pavement failure of
fatigue cracking. The cracking performance of tested RCC sections was analyzed.
Based on the measured surface cracking and the comparison results between mea-
sured and predicted load-induced pavement responses, a RCC fatigue model for thin
RCC pavements was developed, which can be used to predict pavement fatigue life
under different heavy loads for the thickness design of thin RCC pavements.

Keywords Roller compacted concrete
APT
Pavement performance

Non-destructive testing Fatigue analysis

1 Introduction

Roller-Compacted Concrete (RCC) is broadly defined as a stiff, low-water mix of

concrete placed using an asphalt paving equipment and compacted with vibratory
rollers. RCC has similar strength properties as the conventional concrete consisting of
the same basic ingredients, i.e., well-graded aggregates, cementitious materials, and
water. The major difference lies in that RCC contains a higher percentage of fine
aggregates with a lower water cement ratio, which allows for tight packing and

Z. Wu (&)
M. Mahdi
T. Rupnow

Louisiana Transportation Research Center,
4101 Gourrier Avenue, Baton Rouge, LA 70808, USA
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 735

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_48
736 Z. Wu et al.

consolidation. Due to its relatively coarse surface, RCC has traditionally been used for
pavements carrying heavy loads in low-speed areas, such as parking, storage areas,
port, airport service areas, intermodal and military facilities (Harrington et al. 2010).
RCC applications to public roads began in the mid-1980s with a few, relatively
short experimental sections on local roads and residential streets (Piggot 1999).
Since then dozens of RCC pavements on urban streets and intersections have been
completed Rupnow et al. (2015). Most of these public roads were constructed with
thin asphalt overlays for better riding quality compared to the inferior surface
texture of RCC. With improved paving and compaction methods as well as surface
texturing techniques, the recent applications that using RCC as a riding surface for
low volume roads have been reported in several U.S. states including South
Carolina (Delatte 2004), Colorado (Damrong et al. 2012), and Arkansas (Williams
and McFarland 2013).
The Louisiana Department of Transportation and Development (LADOTD) has
seen a rapid decline of low volume roadway serviceability in recent years due to oil
and gas exploration within the Haynesville Shale play near Shreveport, LA and is
expected to see similar results on the low volume roadway network within the
Tuscaloosa Shale play north of Baton Rouge, LA as exploration continues to
expand in upcoming years. This research project has been started to investigate the
alternate use of thin RCC surfaced pavement structure for low volume roadways to
mitigate the effects of exploration efforts. This project consists of a small laboratory
study to determine acceptable mixtures and a larger full scale loading study of
several test sections to determine appropriate design thicknesses and criteria for
future construction efforts.

2 Objective and Scope

The main objective of this study was to evaluate the structural performance and
load carrying capacity of thin RCC-surfaced pavement structures through the
accelerated pavement testing (APT).
The scope of this paper focuses on the following: the APT experimental design,
construction of RCC test sections, in situ non-destructive testing, fatigue analysis
and APT results from three RCC-surfaced pavement test sections.

3 APT Experimental Design

3.1 RCC Pavement Test Sections

Six RCC pavement test sections were constructed at the Louisiana Pavement
Research Facility (PRF) site in Port Allen, Louisiana, using normal highway
construction equipment and procedures. Figure 1 presents the plan view and
Performance of Thin RCC Pavements … 737

Fig. 1 Pavement structures of PRF test lanes

pavement layer thickness configurations of the six test sections. Each section was
about 13 ft. (3.96 m) wide and 71.5 ft. (21.8 m) long. As shown in Fig. 1, the
pavement structures for sections 1–3 generally consist of a RCC surface layer (4–8
in. or 101.6–203.2 mm thick), a 12-in. (304.8-mm) cement treated base layer and a
natural subgrade, whereas, sections 4–6 include a RCC surface layer of 101.6–
203.2 mm thick, an 8.5-in. (215.9-mm) soil cement base layer, a 10-in. (254-mm)
cement treated subgrade layer and a natural subgrade.

3.2 Materials

The RCC materials used in this experiment include a type I Portland cement, a #67
crushed limestone, and a No. 89 crushed limestone manufactured sand. The designed
RCC mix contains a well-graded aggregate blend of 57 % coarse and 43 % fine
aggregate by weight and 11.4 % cement with an optimum moisture content of 6.5 %.
Silty-clay embankment soil (A-6) was used in both soil cement and cement treated
soil layers as well as the treated subgrade. To meet the LADOTD’s roadway design
specification, an 8 % cement by volume was applied for the 215.9 mm soil cement
base, whereas, a 6 % cement by volume applied in the 304.8-mm cement treated base
layer, and a 4 % cement by volume was used for the 254-mm treated subgrade layer.

3.3 Instrumentation

Figure 2 shows the instrumentation layout of this experiment. Each test section was
instrumented with three earth pressure cells (Geokon 3500), two H-type asphalt
738 Z. Wu et al.

Fig. 2 Instrumentation plan

strain gages (Tokyo Sokki KM-100HAS), and two concrete strain gages (Tokyo
Sokki PML-60), which were placed at various locations and layer interfaces, Fig. 2.
Several moisture sensors (TDR CS-616) and thermocouples (T 108-L) from
Campbell Sceintific were also installed.
National Instruments DAQ hardware was utilized to collect the dynamic
responses from pressure cells and strain gages, and a Campbell Scientific data
logger for collecting data from thermocouples and TDR. LabVIEW ver. 12 and
Campbell Scientific PC400 software were used to convert the electronic signal and
to store the data for this experiment.

3.4 Accelerated Pavement Loading

A heavy vehicle loading simulation device—ATLaS30 was used in this experiment.

ATLaS30 is 65 ft. (19.8 m) long, 7 ft. (2.1 m) high and 10 ft. (3.05 m) wide,
constructed around two parallel steel I-beams (Fig. 3). The ATLaS wheel assembly
models one half of a single axle and is designed to apply a dual-tire load up to
30,000-lbf (66-kN) by hydraulic cylinders. With a computer-controlled loading
system, the weight and movement of traffic is simulated repetitively over a 40-ft.
(12.2-m) long loading area in a bi-directional mode at a top speed of 6 mph
(9.66 km/h). Within 24 h approximately 3900 bi-directional loading cycles (each
cycle = two loading passes) can be applied over the 40-ft. long loading area. In this
experiment, each test section was loaded by an incremental loading sequence of 9
Performance of Thin RCC Pavements … 739

Fig. 3 ATLaS30 device

(19.8), 16 (35.2), 20 (44), 22 (48.4) and 25 (55) kips (kN), each of 78,000 loading
passes. If a section was not failed by the first round of loading, continuous loading
of various loads would be applied till pavement failure of fatigue cracking.

4 Construction of RCC Test Sections

All RCC layers were constructed within a day using a set of specialized asphalt
paving and RCC mixing equipment. As shown in Fig. 4, the RCC placement
involves a number of steps: RCC production, transportation and placement, com-
paction, jointing, and curing. For the production of RCC, a Rapidmix 400C hori-
zontal twin shaft pugmill was used. It consists of one or two aggregate feeders, a
cement silo and feeder, a main feeder belt, a water supply system, a pugmill mixer,
a discharge belt, and a gob hopper at the end of the discharge belt. The pugmill
offers a number of advantages, such as rapid mobilization (takes 2–3 h from travel
mode to fully operational mode), self-contained with its own power source, reduced
transportation time for fresh concrete, high production rate (50–300 yd3/h or 38.3–
229.5 m3/h) and an efficient mixing system (Rapid 2014). During the startup
operation, the pugmill was calibrated and batching of RCC was monitored to fulfill
the specification in accordance with ASTM C94. The RCC mix was then trans-
ported to the jobsite in dump trucks. Precautions were taken to avoid excessive
moisture loss during transportation.
Before paving the RCC on the actual pavement sections, a 200-ft. (61-m) long
trial section was constructed on top of an existing roadway at the PRF site to
validate the design, rolling pattern and method of construction using the same
construction equipment. A Rollcon high density paver was used to place the RCC
over the prepared base layer to achieve high initial density and smoother surface.
A CAT CB-64 vibratory steel drum roller was then used to compact the RCC layer.
740 Z. Wu et al.

Fig. 4 Construction of RCC

Nuclear density gage was used to check the moisture and density right after the
paver and after the compaction using vibratory roller. The density ranges between
94 and 96 % of the target density right behind the paver and increased to 97–98 %
after compaction.
All RCC layers were placed in a single lift. Approximately two hours after
paving of RCC (when RCC was hard enough to withstand spalling damage during
sawing operations), transverse saw-cut joints were created on each RCC test sec-
tion. The saw-cut joints were typically spaced at 20 ft. (6.1 m), 15 ft. (4.6 m) and
10 ft. (3.1 m) interval for the 8-in. (203.2-mm), 6-in. (152.4-mm) and 4-in.
(101.6-mm) RCC layers, respectively, with the corresponding joint depths of 1.5 in.
(38.1 mm), 1.0 in. (925.4 mm), and 0.5 in. (12.7 mm). Finally, a white pigmented
Performance of Thin RCC Pavements … 741

water base concrete curing compound conforming to ASTM C309 was sprayed on
the finished RCC surfaces for curing.

5 In Situ Measurements

Cylindrical samples of RCC were prepared on site during the construction.

Cylinder RCC cores and saw-cutting beams were taken after the construction.
Laboratory test results indicated that the average compressive strength at 28 days
and average flexural strength of the constructed RCC layers were 5400 psi
(37.2 MPa) and 661 psi (4.6 MPa), respectively. More details on the laboratory test
results may be referred to elsewhere (10).
The sand patch test was used to determine the average macrotexture depth
(MTD) of the finished RCC surfaces, whereas, a Dynamic Friction Tester
(DFT) was used to measure the corresponding pavement surface friction. In addi-
tion, an ARRB Walking Profiler G2 was used to measure the centerline profilers of
the finished RCC surfaces in term of the International Roughness Index
(IRI) number for each RCC pavement section tested. Table 1 presents the surface
characteristic test results of RCC sections.
The DFT20 values shown in Table 1 represent the measured friction numbers at
20 km/h, which is often used as an indicator for surface micro-texture of a pave-
ment. In general, the measured DFT20 values of RCC test sections ranged from
0.22 to 0.42, which are in a similar range as those measured on the top of different
asphalt mixes (Wu et al. 2012). On the other hand, the measured MTD values
varied from 0.4 to 1.0 mm that are also similar to those of asphalt concrete surfaces.
Overall, it is believed that RCC pavement surfaces should have a similar surface
friction characteristic as asphalt concrete surfaces, in terms of both micro- and
macro-textures.
The IRI results for sections 1–3 in Lane 1 were significantly higher than the IRI
results for sections 4–6. This is most likely because of the different paver speeds
where Lane 1 was constructed in roughly 2/3 of the time as Lane 2 resulting in a

Table 1 Surface Section DFT20 MTD, in. IRI, in./mile

characteristics of RCC (mm) (m/km)
sections
Section 1 0.3 0.039 (1.0) 359 (5.6)
Section 2 0.4 0.028 (0.7) 469 (7.4)
Section 3 0.22 0.035 (0.9) 622 (9.8)
Section 4 0.3 0.015 (0.4) 190 (3.0)
Section 5 0.42 0.015 (0.4) 122 (1.9)
Section 6 0.28 0.017 (0.43) 168 (2.6)
742 Z. Wu et al.

rougher pavement surface. The IRI is also increased due to the frequent change in
pavement thickness as the lanes were paved. Each section is roughly 22 m in length
leading to frequent pavement thickness transitions. The authors believe that in an
actual construction project, IRI values would be equal to or lower than the results
form section 5. While diamond grinding or a thin asphalt overlay on top of the RCC
pavement would overcome the rough surface issue, on the other hand, it is sug-
gested that a finer aggregate gradation be used to achieve a smoother surface for
RCC-surfaced pavements (Rupnow et al. 2015).
In addition, falling weight deflectometer (FWD) tests were performed on the
finished surfaces of RCC test sections. Four FWD load levels (40-, 53.4-, 66.8-,
111.3-kN) were used. The moduli backcalculation results are presented in Fig. 5
below. As expected, the backcaluated moduli of RCC layers in sections 1–3 are
generally lower than those in sections 4–6, which is consistent with the compres-
sive strength test results obtained in the laboratory. The soil cement layers with an
8 % of cement content are stiffer in terms of the backcalucated moduli than the
cement treated soil layers of 6 % cement content. The backcalcuated subgrade
moduli are also found higher for those sections on lane 2 due to the 10-in.
cement-treated subgrade used.

Fig. 5 FWD backcalculated moduli of RCC test sections

Performance of Thin RCC Pavements … 743

6 Discussion of APT Results

6.1 Instrumentation Results

Figure 6 presents the comparison of typical stresses and strains measured at the
bottom of RCC slabs under different load magnitudes for the three RCC sections
considered. As expected, Fig. 6 shows that the measured stresses and strains are all
increased with the increasing of load magnitude and decreased with the increase of
RCC thickness.

6.2 Performance of RCC Sections

At the present, the APT tests were only completed on three RCC pavement sections
(i.e., sections 4, 5 and 6). The overall results generally confirmed that all tested
RCC pavements showed very high load carrying capacity. In the end, two test
sections (sections 5 and 6) were successfully failed by significant fatigue cracks on
pavement surfaces.

6.2.1 Section 6 (4-in. RCC+8.5-in. Soil Cement)

A total of 720,800 ATLaS load repetitions of various load magnitudes was applied
on this section and the estimated total number of ESALs to fatigue failure was 19.2
million. Figure 7 shows the cracking development under different load repetitions
observed on this section. FWD backcalculated subgrade moduli (Mr) at different
stations were also plotted on a vertical axis to the left side in Fig. 8. Neither visible
nor measurable distresses could be obtained on this section at the ends of 9-, 16-
and 20-kip of ATLaS dual-tire loading. At the beginning of the 22-kip loading, a
hairline longitudinal crack around Station+10 was noticed, which was in the middle
of one tire print, Fig. 7. With additional load repetitions, the longitudinal crack
propagated and expanded continuously, and resulted in some pumping fine

Fig. 6 Comparison of load-induced pavement responses

744 Z. Wu et al.

Fig. 7 Cracks versus load repetitions observed on section 6

Fig. 8 Cracks versus load repetitions observed on section 5

Performance of Thin RCC Pavements … 745

materials through the cracks and saw-cut joints on this section. After 480,000 load
repetitions, longitudinal cracks from outside the wheel path started to initiate.
Finally, the inside and outside longitudinal cracks connected to each other and a
punchout type failure occurred around Station+15 after a total of 706,500 passes of
ATLaS dual tire loading. Interestingly, the cracking failure was observed confined
only in the first half of the loading area on section 6, as shown in Fig. 7.
The following findings were be drawn from Fig. 7: (1) a longitudinal crack was
initialized in the middle of one tire print that seemed to be a bottom-up crack due to
high tensile stresses at the bottom of the 4-in. RCC slab; (2) The weaker subgrade
portion under the loading area caused a higher tensile stress under the slab than did
the stronger subgrade portion; (3) under the continuous load repetitions voids
seemed to be formed underneath the slab because of the pumping of fine materials,
which generated more deflections and cracks of the slab under loading; (4) due to
only 4 in. of slab thickness, the final cracking pattern was kept in a narrow area,
different from the cracking pattern observed on the 6 in. RCC section to be
described below.

6.2.2 Section 5 (6-in. RCC+8.5-in. Soil Cement)

It took much longer time and more loading repetitions to fail this section than
section 6. A total of 1,750,850 load repetitions of various loads was applied on this
section and the estimated ESALs to fatigue failure was 87.4 million. Figure 8
shows the development of cracking pattern under different load repetitions observed
on this section. Similar to that of section 6, the crack was also initiated in longi-
tudinal direction in this section. However, this time the longitudinal cracking was
initiated along the edge of a wheel tire, Fig. 8. Another significant difference
between the two failed sections is that the cracking pattern was much wider in
section 5 than section 6.
The following findings were observed from Fig. 8: (1) the initial longitudinal
cracking observed at the edge of a tire print seems to be a top-down fatigue cracking
due to the high shear stresses under the tire wall; (2) The more uniform subgrade
moduli resulted in a final cracking failure covering the entire loading area; (3) with
continuous load repetitions and pumping, voids could be formed underneath the
6-in. RCC slab, which generated more deflections and cracks of the slab under
loading; (4) due to the combination factors of a thicker slab thickness, more uniform
subgrade support and possibly high shear stresses under tire walls, the final
cracking pattern of section 5 was found much wider than that of section 6; (5) with
a 2-in. increase in RCC thickness, the load carrying capacity of a RCC pavement
could be significantly increased.
746 Z. Wu et al.

6.2.3 Section 4 (8-in. RCC+8.5-in. Soil Cement)

Only 392,500 load repetitions (approximately 11.3 million ESALs) were applied on
section 4. Overall no significant pavement damage could be observed after the
loading except two short, fine cracks appeared near one joint on this section. This
joint also showed moderate pumping action of fine materials after the test site
received more than 3 in. of heavy rains during the loading of 25-kip. After the rain,
the continuous loading could not generate any further damage on this section. Due
to a very high number of load repetitions required to fail section 5 (the 6-in. RCC
section), the research team then decided to discontinue testing on this section.

7 Fatigue Analysis of RCC Pavements

7.1 Pre-loading Fatigue Life Analysis

RCC pavements have engineering properties similar to those of conventional

Portland cement concrete (PCC) pavements. Thus, RCC pavement design methods
usually employ the same basic strategy as PCC pavements: keeping the pavements
flexural stress and fatigue damage caused by wheel loads within the allowable
limits (Harrington et al. 2010). A computer program called RCC-PAVE was
developed by the Portland Cement Association (PCA) for thickness design of RCC
industrial pavements, in which the fatigue equation shown in Eq. 1 was incorpo-
rated (Personal 2012). On the other hand, Eq. 2 below shows the fatigue model
used for PCC pavement thickness design developed by American Concrete Institute
(ACI 1995).

log Nf ¼ 10:2547611:1872ðSRÞ ð1Þ

Nf ¼ ð4:2577=ðSR0:4325ÞÞ3:268 for 0:45\SR\0:55

ð2Þ
log Nf ¼ 11:73712:077ðSRÞ for SR  0:55

where
Nf—the allowable number of load repetitions;
 
SR—Critical flexural stress ratio, SR ¼ CriticalFlexural
tensile stress at bottom of a slab
strength of RCC

In the pre-loading fatigue analysis, the critical tensile stresses at the bottom of
RCC slabs were predicted under different ATLaS dual tire loads using the finite
element program of KENPAVE (Huang 1993). Each RCC pavement section was
modelled as an RCC layer directly sitting over a solid foundation. First, based on
the FWD backcalculation results a modulus value of 4000 ksi was chosen for all
RCC layers to represent an overall average modulus of this material in a field
condition. Second, based on the results of matching predicted surface deflections to
Performance of Thin RCC Pavements … 747

Table 2 Predicted fatigue lives (number of load repetitions) of RCC test sections
Load Fatigue Section 1 Section 2 Section 3 Section 4 Section 5 Section 6
(kips) model (8” RCC) (6” RCC) (4” RCC) (8” RCC) (6” RCC) (4” RCC)
9 PCC ∞ ∞ 136,000 ∞ ∞ 420,000
RCC 95 million 640,000 13,000 115 8 million 27,000
million
16 PCC ∞ 12 202 ∞ ∞ 765
millions
RCC 6.5 124,000 33 9.3 220,000 113
million million
20 PCC ∞ 65,000 2 ∞ 145,000 6
RCC 960,000 7,000 0 1.5 12,250 1
million
22 PCC ∞ 57,500 1 ∞ 92,000 1
RCC 650,000 4500 0 1 million 9300 0
25 PCC ∞ 46,000 1 ∞ 12,250 1
RCC 168,000 600 0 284,000 1500 0

the FWD measured in an elastic analysis using KENPAVE a set of modulus values
of 30 ksi and 38 ksi were selected for the solid foundation layers on sections 1–3
and sections 4–6, respectively. The predicted critical tensile stresses at the bottom
of RCC slabs were then used in the pre-loading fatigue analysis. The predicted
fatigue lives of the six RCC test sections are presented in Table 2. In general, the
analysis results indicate that the pavement fatigue lives predicted by the RCC
fatigue model are significantly less than those predicted by the PCC model.
However, to fail an 8-in. RCC test section both models predicted large numbers of
allowable load repetitions, even under a 25-kip dual tire load, which simulates a
50-kip single axle load in reality. On the other hand, the predicted results also
showed that only a few number of 20-kip dual tire passes would fail a 4-in. RCC
pavement section.

7.2 APT Performance-Based Fatigue Models

It is obvious that the actual fatigue lives of those thin RCC pavement sections tested
in this experiment did not closely match to the predicted fatigue lives in the afore-
mentioned pre-loading fatigue analysis shown in Table 2. A further fatigue analysis
of RCC test sections was therefore conducted based on the actual performance of the
two thin RCC pavements tested (sections 5 and 6). The field critical tensile stresses
at the bottom of RCC slabs were at first estimated from the instrumentation
responses by assuming that the measured vertical stresses, longitudinal and trans-
verse strains at bottom of a RCC slab all representing the load-induced
responses for a single point at an interface of RCC-and-underneath-base-layer.
748 Z. Wu et al.

Fig. 9 Comparison between different fatigue models

By treating the RCC layer as a homogeneous elastic layer with a modulus value of
4000 ksi and a Poisons’ ratio of 0.15, the field tensile stresses under different ATLaS
dual tire loads may be estimated through solving three simultaneous Hook’s
stress-strain equations. With the estimated tensile stress under a specific ATLaS dual
tire load, the stress ratio, SR, can be obtained. Different SRs versus numbers of load
repetitions of various ATLaS dual tire loads would be accumulated to develop a
100 % fatigue damage for each section evaluated. Such information were grouped
together in an Excel spreadsheet and using the Solver function in Excel, the fol-
lowing two fatigue equations were developed in this study:

For section 6 : log Nf ¼ 9:071  12:729  SR ð3Þ

For section 5 : log Nf ¼ 9:507  12:597  SR ð4Þ

Figure 9 shows a comparison between the developed fatigue equations in this study
and those of RCC-PAVE and ACI fatigue models as listed in Eqs. 1 and 2. It can be
observed that the developed models of thin RCC pavements generally shift to the
left of the RCC-PAVE model developed for thick industrial pavements and the ACI
fatigue model for regular PCC pavements.

8 Summary and Conclusions

Six pavement sections of RCC over semi-rigid bases were constructed for an APT
experiment. The RCC slab thicknesses were relatively thin, varied from 4 in. to 8
in. A set of specialized paving and mixing equipment was selected in the RCC slab
Performance of Thin RCC Pavements … 749

construction. Each test section was instrumented with multiple earth pressure cells
and strain gages to measure the direct pavement responses under the wheel loading.
A heavy vehicle load simulator device with the bi-directional loading capability
was used in the experiment. The following observations and conclusions may be
drawn from the current APT results:
• With an improved RCC mix design and modem asphalt paving technology, it is
possible to build an RCC-surfaced pavement that can provide similar surface
micro- and macro-textures as asphalt concrete, and a riding surface as smooth as
section 5 in this study.
• Both sections 5 and 6 in this study showed an outstanding load carrying
capacity under various heavy truck loads indicating that a thin RCC pavement
(thicknesses of 4–6 in.), when properly constructed, would be suitable to be
used for low-volume roadways where often encountering significantly heavy
truck trafficking;
• Two RCC sections were successfully loaded to a fatigue cracking failure in this
experiment. The fatigue cracks were found to initiate longitudinally at either the
center of a tire print or a tire edge. Pumping of underneath fine materials at
saw-cut joints would produce voids below RCC slabs, which helped bend more
of RCC slabs then propagate cracks into a fatigue cracking pattern under the
continuous loading;
• The pre-loading fatigue analysis indicated that the pavement fatigue lives pre-
dicted by the RCC-Pave fatigue model were significantly less than those pre-
dicted by the PCC model, especially under a large load. Both models predicted
only a few number of 20-kip dual tire passes would fail a 4-in. RCC pavement
section, which were significantly different from the APT test results.
• Based on the APT performance of RCC sections in this study, a set of fatigue
prediction equations were developed. The developed models were found to
better suit in the predicting the fatigue damage of a thin RCC pavement.

Acknowledgments This study was supported by the Louisiana Transportation Research Center
and the Louisiana Department of Transportation and Development. The authors would like to
express thanks to all those who provided valuable help in this study.

References

ACI COMMITTEE 325.10R-95. (1995). State-of-the-art report on roller-compacted concrete

pavements, manual of concrete practice. American Concrete Institute.
Damrong, N., Liang, Y. C., & Yunping, X. (2012). Application of roller compacted concrete in
Colorado’s roadways. Report No. CDOT-2012-11, Colorado Department of
Transportation DTD Applied Research and Innovation Branch, October 2012.
Delatte, N. (2004). Simplified design of roller-compacted concrete composite pavement. In
Transportation Research Record: Journal of the Transportation Research Board, No. 1896,
Transportation Research Board of the National Research Council, Washington, DC.
750 Z. Wu et al.

Harrington, D., Abdo, F., Adaska, W., & Hazaree, C. (2010). Guide for roller-compacted concrete
pavements. National Concrete Pavement Technology Center, Institute for Transportation, Iowa
State University.
Palmer, W. D. (2005). Paving with roller compacted concrete (pp. 41–45). Public Works.
Personal email communication on 2012 with Mr. Wayne S. Adaska at Portland Cement
Association, Skokie, Illinois.
Piggot, R. W. (1999). Roller compacted concrete pavements—A study of long term performance.
RP366, Portland Cement Association.
Portland Cement Association. (1987). Structural design of roller-compacted concrete for
industrial pavements, IS233.01.
Rapid International Ltd. (2014). Rapidmix 400C, http://www.thomaco.com/pdf/rapidmix400c_en.
pdf. Last Accessed July 28, 2014.
Rupnow, T., Icenogle, P., & Wu, Z. (2015). Laboratory evaluation and field construction of roller
compacted concrete for testing under accelerated loading. Journal of Transportation Research
Record, No. 2504, Transportation Research Board of the National Research Council,
Washington, DC.
Williams, S. G., & McFarland, A. M. (2013). Roller compacted concrete for roadway paving.
Final Report TRC-1005, Arkansas Highway and Transportation Department, August 14, 2013.
Wu, Z., King, B., Abadie, A., & Zhang, Z. (2012). Development of a design procedure for
prediction of asphalt pavement skid resistance. Journal of the Transportation Research Board,
No. 2306, Transportation Research Board of the National Research Council, Washington, DC.
The Design, Construction and First-Phase
Heavy Vehicle Simulator Testing Results
on Full Scale Ultra-Thin Reinforced
Concrete Test Sections at Rayton,
South Africa

L. du Plessis, G.J. Jordaan, P.J. Strauss and A. Kilian

Abstract Ultra-Thin Reinforced Concrete Pavements (UTRCP) are successfully

being used in residential streets and low-volume road applications in South Africa.
Due to its popularity in this domain the Gauteng Provincial Department of Roads
and Transport (GPDRT), in conjunction with CSIR Built Environment, started a
research project to determine whether this type of technology can be used in
higher-order roads such as collector roads (bus routes) or secondary and tertiary
provincial roads where the traffic loading is significantly higher than what is found
on residential streets. The research project consists of the testing of full-scale test
sections with the Heavy Vehicle Simulator just outside Rayton in the Gauteng
Province. The bulk of this paper deals with the structural design and construction of
the various test sections. Four sections were constructed using different types of
substructure support and different configurations of steel reinforcement. The paper

L. du Plessis (&)
CSIR Built Environment, Meiring Naudé Street,
Brummeria, Pretoria, South Africa
e-mail: [email protected]
G.J. Jordaan
Tshepega Engineering (Pty) Ltd, PO Box 33783,
Glenstantia 0010, South Africa
e-mail: [email protected]
G.J. Jordaan
University of Pretoria, Pretoria 0001, South Africa
P.J. Strauss
Pieter Strauss CC, PO Box 588,
La Montagne 0184, South Africa
e-mail: [email protected]
A. Kilian
Gauteng Provincial Department of Roads and Transport,
Rayton, South Africa
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 751

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_49
752 L. du Plessis et al.

also includes a summary of the results of the first HVS test conducted at the Rayton
testing site.

Keywords Heavy vehicle simulator testing
Ultra-thin reinforced concrete

pavements (UTRCP) Modeling of UTRCP

1 Background

The upgrading of unsurfaced residential roads has become a priority for many
metropolitan areas in South Africa. This need is combined with a social expectation of
increased labor-intensive construction techniques. In 2007, the Gauteng Provincial
Department of Roads and Transport (GPDRT) initiated its Expanded Public Works
Program (EPWP). The goals and objectives of the EPWP are to alleviate unem-
ployment, targeting women, youth and the disabled specifically (Gautrans Provincial
Government, June 2007; Gauteng Province Department of Infrastructure
Development 2007). This goal will be achieved by creating work opportunities in the
following four constructive ways:
• Increasing the labor intensity of government-funded infrastructure projects;
• Creating work opportunities in public environmental programs;
• Creating work opportunities in public social programs, and
• Utilizing general government expenditure on goods and services to provide the
work experience component of small-enterprise learnership/incubation programs.
In line with these objectives, the GPDRT approached the CSIR to assess the
structural performance of Ultra-Thin Reinforced Concrete Pavement (UTRCP)
through accelerated pavement tests using the Heavy Vehicle Simulator (HVS); to
provide specialist advice on the application of this technology; and to facilitate the
transfer of knowledge and skills to other role-players in the industry. The research
project is a partnership between the CSIR, the Concrete Institute and the University
of Pretoria (UP).
The investigation is conducted in phases. During the first phase, short full-scale
test sections were constructed at a borrow pit (Roodekrans) to the west of Pretoria,
where loaded trucks travelled over the sections (Strauss et al. 2005). A weight
bridge at the site provided the truck volume and loading information. Following this
initial testing was Phase II, during which a number of UTRCP test sections were
constructed in the median of a highway (R80, north-west of Pretoria) which were
subjected to HVS loading (Du Plessis et al. 2010, 2011).
The Design, Construction and First-Phase Heavy Vehicle … 753

Eight HVS tests were completed during Phase II, representing various combi-
nations of support layers, with and without the application of water on the surface
of the UTRCP. Lessons learnt from these tests include:
• In the dry state, UTRCP offers exceptionally good performance. During HVS
testing, more than two million standard 80 kN axle load repetitions (E80s) were
applied, without any sign of noticeable damage.
• During wet/dry cycles, the performance of the pavement dropped. Through
limited testing, it was shown that under severe surface wetting conditions the
pavement sections failed after approximately 350,000 load applications. Given
the aggressive loading and watering regime, this is still considered adequate for
a low-volume, low-load residential street type application.
• UTRCP is prone to slab curling and warping along the longitudinal edge of the
slabs. Slab separation from the underlying base course was in the order of
3 mm.
From these test results it is concluded that even in the wet state, the pavement
sections should be able to carry the anticipated traffic (less than 1 million E80s) on a
residential township road over its design life (20+ years). This said, the pavement
system is sensitive to the ingress of water. Possible reasons for this include:
• the erodibility of the granular base;
• loss of slab stiffness due to the crack width and crack density of shrinkage cracks;
• the stiffness of the substructure (measured through surface deflections), and
• loss of slab stiffness due to construction-related problems such as poorly
compacted concrete, variable slab thickness and depth of the reinforcement. (It
is important to note that these initial tests were conducted using a concrete
thickness of 50 mm. This thickness has proven problematic to construct in
practice using labor-intensive methods. Consequently, the thickness of the test
sections constructed as discussed in this paper, was increased to 70–75 mm,
which also improved the correct placing of the reinforcement and assisted in the
performance of the UTRCP layer.)
The CSIR has produced a guideline document, which is available to road
authorities, pavement designers and consultants to assist them in all aspects of
constructing UTRCP-type pavements (CSIR 2011). The guideline is currently in
draft form and covers a wide range of topics, including mix design and construction
practices. The current limitation is that the structural design of UTRCP is not
adequately described in terms of the various traffic classes and the required sub-
structure support required for carrying the anticipated design traffic throughout the
design life of the pavement.
754 L. du Plessis et al.

2 Objectives of the Rayton HVS Testing

The HVS tests conducted on the UTRCP test sections located on the R80 proved
that the technology can be used successfully on low-volume residential streets
carrying a limited amount of heavy vehicles (design class ES0.3) or in dry regions
ES1 (maximum of 1 million E80’s) (Du Plessis et al. 2011).
The first objective of the testing at Rayton is to investigate whether the UTRCP
pavement system can also be improved for use on higher-order roads with higher
traffic volumes and higher percentages of heavy vehicles. In order to answer this
question, different UTRCP full-scale test sections were constructed on the
Rayton—Nooitgedacht rural road (north-east of Pretoria). These sections all have
different design strengths and the design structural capacity will be evaluated using
HVS testing in order to validate the structural design methods and to make rec-
ommendations with regard to the UTRCP pavement system and the applicability to
traffic and load conditions.
The second objective is to develop first-order performance curves for UTRCP
pavements. The aim is to develop transfer functions that will assist designers to
predict a realistic pavement life using inputs such as the anticipated traffic (number
of equivalent heavy vehicle loads) over the lifespan of the road and UTRCP
characteristics, for given support and climatic conditions. This objective will be
achieved by the analysis of all HVS data collected from the various testing sites as
well as all laboratory test data, which will feed into the development of appropriate
performance models. This objective is currently ongoing and not covered in this
paper.
The objective of this paper is limited to the structural design and construction of
the various test sections at the Rayton testing site. Four sections were constructed
using different types of substructure support and types of steel reinforcement. The
paper also includes a summary of the results of the first HVS test conducted at the
Rayton testing site.

3 Variables to Be Investigated

The variables to be evaluated under this series of HVS testing include the
following:

3.1 Different Steel Mesh Types for the UTRCP

Reinforcement

Currently, the UTRCP design (for a design traffic of less than 1 million E80s in
20 years) consists of a 50 mm thickness PCC with 200 mm × 200 mm 5.6 mm Ø
The Design, Construction and First-Phase Heavy Vehicle … 755

Fig. 1 The structural design of the different UTRCP HVS test sections at Rayton

steel mesh reinforcement. Firstly during the Rayton series of testing, the PCC
thickness was increased to 75 mm. This increased thickness adds structural strength
to the concrete and reduces deflection, which decreases the risk of pumping, and
also reduces construction problems as it proofed to be difficult to place the steel
mesh on the neutral axis if the PCC layer is only 50 mm thick. Secondly, different
steel mesh arrangements were tested. During one section, the longitudinal steel was
doubled (from 200 mm to 100 mm spacing). Figure 1 details the different steel
mesh configurations.

3.2 Different Support Conditions

In order to determine whether this technology can be used for higher-order roads
(traffic class ES 10, total design traffic 3–10 million E80s), two types of sub-
structure support were constructed. The first 50 m section consists of a cemented
subbase and cemented base (also called lower and upper subbase in the case of
concrete pavements) layers (South African material types C4 and C3) (Committee
of State Road Authorities 1985) to improve the structural support of the foundation
and to mitigate the negative effects of pumping during wet periods. The normal
UTRCP design for low-volume roads (only granular base and subbase layers
without cement stabilization) was used for the second 50 m section.
The UTRCP sections were constructed as shown in Fig. 1. The 100 m long
section was constructed with different design features as discussed above. Each
subsection is 25 m in length. The layer works for the various sections are also
shown in Fig. 1. The standard design for low-volume roads (design class ES1,
traffic less than 1 million E80s) is the last subsection shown on the right-hand side
756 L. du Plessis et al.

of Fig. 1 (with the exception of an increase in the thickness of the concrete layer as
discussed).

4 Structural Modeling of the Strong Section

As the standard design for low-volume roads has been verified through the previous
round of HVS testing (at the R80 highway) (Du Plessis et al. 2010, 2011), only the
strong section with a design life of 3–10 million E80s was modelled to determine if
the suggested structure will have the required capacity for the higher traffic loading
demand. Two empirical-mechanistic software packages were used to verify the
design. The first method used is the South African design software for concrete
pavements, cncPAVE (Strauss et al. 2007). In addition, the design was compared
with the results obtained using the South African (SA) mechanistic design for
flexible pavements through the use of mePADS, a software tool developed by the
CSIR (Theyse and Muthen 2000).

4.1 CncPAVE Modelling

cncPAVE is a computer simulation program which was developed to facilitate

competent decision-making when designing concrete pavements (Strauss et al.
2007). cncPAVE assumes that input values are uncertain. The approach is based on
the evaluation of consequences. The consequences of a certain pavement design are
expressed in terms of decision variables, viz.
• % shattered concrete surface (%SH);
• % pumping concrete surface (%PU);
• % faulting in the concrete pavement (%FA) (in the case of plain and dowel
concretes);
• crack spacing (in case of reinforced concretes);
• international roughness index (IRI); and
• life cost of the pavement, in SA rands per square meter.
cncPAVE treats input variables and all output items as random variables. The
program derives the probability distributions of the above decision variables by
means of Monte-Carlo simulation. The average values of the decision variables can
then be compared with certain decision criteria and the quality of design can be
judged accordingly. The formula, transfer functions and performance prediction
models used for the UTRCP modeling in cncPAVE is not part of paper but can be
found in various references (Du Plessis et al. 2007; Strauss et al. 2007).
The Design, Construction and First-Phase Heavy Vehicle … 757

Table 1 Failure criteria for UTRCP type pavements

Decision variable Good Acceptable Excessive
% shattered concrete below 0.2 % 0.2–0.5 % over 0.5 %
% pumping below 2 % 2–5 % over 5 %
Crack spacing 0.2–0.5 m 0.2–0.7 m below 0.2 m or over 0.7 m
IRI (m/km) below 1.5 1.5–3.0 over 3

In addition, probabilities of the decision variables exceeding certain limits can be

read off respective graphs. These probabilities can be used to establish the confidence
intervals of decision variables. For instance, the failure criteria for UTRCP are
summarized in Table 1.
The loading input is in the form of load spectra (as opposed to a standard 80 kN
axle design load). Several traffic distributions are available for the engineer
depending on the design life, traffic class and the degree of law enforcement to
control overloaded vehicles. An example of the load spectra input is shown in Fig. 2.
Output of cncPAVE showing the results of the analysis performed on the final
selected pavement design can be seen in Figs. 3a, b. The selected design is exactly
the same as shown in Fig. 1 with expected layer stiffness in agreement with the
various layer qualities and strengths as shown in Fig. 1.
A concrete thickness of 75 mm was selected (100 × 100 mm steel mesh) and a
T2 traffic distribution was chosen, which simulates total daily truck traffic (ADTT)

Fig. 2 Load spectra input used in cncPAVE

758 L. du Plessis et al.

Fig. 3 a Input variables and failure criteria results of the final accepted design. b Output from
cncPAVE
The Design, Construction and First-Phase Heavy Vehicle … 759

of 50 trucks per lane. T2 traffic loading is defined as a distribution of trucks, where

35–55 % of the trucks are longer than 18 m and consist of a three-chassis com-
bination (so-called interlink, consisting of a truck tractor, a semitrailer, and another
semitrailer at the end). In terms of E80s, this type of truck distribution equates to
2.82 E80s per heavy vehicle.
The results from cncPAVE analysis (Fig. 3a) indicates that after 10 years, a very
small percentage of the concrete area will be shattered (SH%) and the crack spacing
(Xm) will still be acceptable (both in the green bands). The total simulated truck
traffic converted to standard E80 axle loads is 4.17 million with a corresponding
surface deflection of 0.126 mm (Fig. 3b).
The results from cncPAVE indicated that the selected pavement structure should
be adequate to carry the design traffic (class ES 10: 3–10 million E80s) with
minimal damage after a 10-year evaluation period.

4.2 SA Mechanistic Design for Flexible Pavements

Due to the uncertainty on the construction quality of the PCC layer, the pavement
design was also modelled using the SA mechanistic design for flexible pavements
(Theyse and Muthen 2000). This was done to study the effect when the stiffness of
the PCC is reduced to stiffness values comparable to asphalt concrete and whether
the structural strength from the support (the base, subbase and subgrade) will render
sufficient support to still carry the design traffic. Using mePADS, the pavement
structure was modelled using the following input parameters:
• Asphalt Concrete thickness (75 mm)
• South African type C3 base (150 mm)
• South African type C4 subbase (150 mm)
• South African type G6 selected layer (150 mm)
• Prepared road bed G9 quality
Input stiffness values for the different material types are shown in Fig. 4.
The transfer functions and formulae used during analysis is not covered in this
paper but can be found in the reference on the SA mechanistic design for flexible
pavements (Theyse and Muthen 2000). The output of the analysis is shown in

Fig. 4 Input stiffness values for the various pavement layers

760 L. du Plessis et al.

(a)

(b)

Fig. 5 a Estimated layer-bearing capacity. b Output from mePADS

Figs. 5a, b. The critical layer (layer with the lowest calculated layer-bearing
capacity) is the base layer (C3). This is to be expected if the stiffness of the top layer
is reduced to an input stiffness of only 3500 MPa.
The estimated layer-bearing capacity of the base is 5.1 million standard 80 kN
axle loads during the crack initiation phase, which satisfies the design criteria of
3–10 million E80s. Note that the estimated deflection on top of the base is calcu-
lated as 0.249 mm (Fig. 5b).
Using the SA mechanistic design method for flexible pavements, it is concluded
that even if the quality of the UTRCP layer is significantly reduced, the suggested
pavement structure should be adequate to carry the design traffic (class ES 10:
3–10 million E80s).
The Design, Construction and First-Phase Heavy Vehicle … 761

5 Construction of the Test Pavement

The reporting on the physical construction of the UTRCP test sections falls outside the
scope of this paper and only limited in-field material properties are reported on here.
However, the construction was done implementing strict quality-control measures to
ensure that all design criteria were adhered to.
All the support layers have met or exceeded their respective specifications and
were accepted by the resident engineer. Layer thicknesses will be verified through test
pits and post mortem investigation after the completion of HVS testing. The average
concrete cube compressive strength was 25.3 MPa after 21 days and increased to
36.4 MPa after 28 days of curing. The average flexural beam strength was 6.25 MPa.
Figure 6 shows the partially constructed test section with the exposed steel mesh for
the last 25 m.

5.1 FWD Evaluation

FWD measurements were taken after the construction of the base material and
28 days after the casting of the concrete. The results are shown in Fig. 7. The red
blocks in Fig. 7 indicate where potential HVS testing areas were identified.

Fig. 6 UTRCP constructed sections at Rayton

762 L. du Plessis et al.

Fig. 7 FWD results on the base and concrete after curing

It is interesting to note that the average deflection on the strong base on top of
the concrete in the area of the HVS test section is in the order of 120 micron, which
is very similar to the calculated deflection from cncPAVE (0.126 mm). The
deflection on top of the C3 base material is also comparable with what was cal-
culated with the SA mechanistic design software (249 micron—see Fig. 5b).
Given the short test sections (25 m each), some variation in the quality and
strength of the layers can be expected, as shown by the variability in FWD data
(especially at chain age 0.02—0.025 and at 0.065). After the concrete was placed,
the deflections were more uniform. A second observation is that the effect of the
75 mm concrete with limited reinforcement on the comparative structural capacity
of the pavement structure is clearly evident from the measurements. As can be
expected, surface deflections on top of the concrete are significantly lower than the
deflections taken on top of the base layer (Fig. 7).

6 HVS Testing

Only limited HVS results are given as the analysis and testing have not been com-
pleted. The first test was done on the HVS testing area on the strong support on the
200 × 200 mm steel mesh (see 1st red block on the left side in Fig. 7). Two Multi
Depth Deflectometer (MDD) instruments were placed at 2 m from the HVS testing
starting point (MDD4) and at 4 m from the starting point (MDD8). MDDs were
The Design, Construction and First-Phase Heavy Vehicle … 763

Fig. 8 Cracked areas without pumping after 1.64 million repetitions

placed in the center of the wheelpath. Testing was started with a 40 kN dual wheel
load (simulating the damage caused by a 80 kN axle load). Due to the slow rate of
damage accumulation, the load was increased to 60 kN after 1 million repetitions
(simulating a 50 % overload). After some degree of cracking was observed, water was
added to the surface of the concrete. The water was added in a cycle: 2 days of adding
water, followed by 5 days of drying, and then a repetition of the cycle. The aim of this
cyclic application of water was to simulate the normal cyclic occurrences of rain on
the surface of the pavement.
Fine cracks developed across the whole surface of the tested area (to be expected),
but not to any degree that allowed water to penetrate (see Fig. 8). After the second

Fig. 9 Visible pumping only though the MDD hole after 1.64 million repetitions
764 L. du Plessis et al.

Fig. 10 Crack development at the end of the 80 kN phase

watering cycle (1.64 million repetitions), the only visible pumping was observed at the
MDD instrumentation hole (see Figs. 9, 10). This observation proved the success of
using a stabilized layer to mitigate the negative effects of water entering the pavement
and destroying the substructure. The loading was again increased to 80 kN (after 1.85
million repetitions), and finally to 100 kN (after 2.16 million repetitions).
At the end of the 80 kN phase, visible cracks were observed over the complete
length of the test section; however, apart from the pumping at the MDD holes, no
visible pumping was detected at any of the surface cracks (Fig. 9).

Fig. 11 Section failure after 2.59 million repetitions

The Design, Construction and First-Phase Heavy Vehicle … 765

The section finally failed during the 100 kN phase during a watering cycle
(Fig. 11) after 2.59 million repetitions. Using the accepted 4.5 exponential damage
law (which could be considered conservative (low) given previous HVS tests on
similar substructures), the section was subjected to approximately 40 million E80s.

6.1 Elastic Deflection Response

The small variations in the elastic deflections (Fig. 12) are mainly due to daily
temperature variations and the increase in loading as the test progressed. Figure 12
shows the deflections at the trafficking load captured by MDD4 at the surface and at
depths of 220 mm (bottom of the base), 380 mm (bottom of the subbase) and
540 mm (top of subgrade).
It is observed that, during the 40 and 60 kN phases, the deflections were very
low (<0.3 mm) and these only started increasing after the load was increased to
80 kN. During the 100 kN phase, the deflection increased to over 2 mm, which
lead to the cracking failures observed.
For comparative purposes a standard 40 kN load defection test is done at certain
repetition intervals even if the trafficking loading varied between 40, 60, 80 and
100 kN. Figure 13 shows the 40 kN defections as captured by MDD8. For clarity,
only the data from 1.5 million to the end of the test are shown as no increase or
deterioration in the pavement was observed before 1.5 million repetitions.

Fig. 12 MDD deflections at trafficking loads

766 L. du Plessis et al.

Fig. 13 MDD deflections at 40 kN test loads

Note that the real damage in terms of structural support only occurred after the
load was increased to an unrealistically high 100 kN (this represents an axle load of
200 kN!). At the end of the 80 kN phase, the surface deflections were still under
0.3 mm, and only after the load was increased to 100 kN, the deflections increased
to approximately 0.7 mm. The second observation is that the deflections recorded at
the bottom of the base (220 mm), subbase (380 mm) and subgrade (540 and
720 mm) did not increase significantly during the whole test which indicates that the
cemented base layer remained in its crack initiation phase (Fig. 4) and rendered good
resistance to water penetration and possible erodibility of the base material. This
observation proves that the design has been successful in protecting the subgrade
even in the event of the severe 100 kN loading during the last phase of the test.

7 Conclusions

This paper summarizes the structural analysis and initial testing of a strengthened
UTRCP pavement structure to be used on higher-order provincial roads in South
Africa. The initial structural analyses were done using two software packages,
cncPAVE and mePADS. Both predicted that the pavement would be able to
withstand the anticipated traffic loading during its design life. It is important to note
that during the mePADS evaluation, the bearing capacity of the pavement reported
in this paper is only for the crack initiation phase of the cemented layers. After this
The Design, Construction and First-Phase Heavy Vehicle … 767

phase, the cemented layers are considered cracked but still have additional bearing
capacity in its equivalent granular state. cncPAVE predicted pavement performance
as seen and measured under HVS testing. The observation that the field deflections
(recorded by FWD) were in agreement with the model prediction proves that the
calibration of the UTRCP model in cncPAVE has been a success.
Initial indications are that this pavement structure is not load sensitive (double
the load does not relate to double the damage or deflections) and showed excellent
resistance to water damage. The initial testing results show that this UTRCP
structure can be used for an ES10 traffic class (3–10 million E80s) with an
acceptable safety margin. The test was conducted simulating an aggressive amount
of rain and despite this, no accelerated damage was detected due to the ingress of
water and destruction of the supporting layers.
In conclusion, the importance of good quality control is highlighted in this study.
The UTRCP technology is aimed at using a high percentage of unskilled hand
labor. It is shown that with strict quality control, a quality pavement can be con-
structed that can have similar characteristics than a pavement constructed using
mechanical construction equipment.

Acknowledgments This HVS research project is supported by the Gauteng Provincial

Department of Roads and Transport. Their financial and technical contribution towards this study
is greatly acknowledged. The authors thank Mr Brian Perry, Managing Director of the Concrete
Institute, for his support and assistance.

References

Committee of State Road Authorities. (1985). Guidelines for road construction materials, TRH
14. Pretoria.
CSIR. (2011). Guidelines for the construction of a 50 mm thick ultra-thin reinforced concrete
pavement (50 mm UTRCP). Prepared for the Gauteng Department of Public Transport Roads
and Works, CSIR Pretoria, 2011.
Du Plessis, L., Strauss, P. J., Perrie, B. D., Rossmann, D. (2007). Accelerated pavement testing of
load transfer through aggregate interlock and the influence of crack width and aggregate type—
A case study. In International workshop on best practices for concrete pavements, Recife,
Brazil, October, 2007.
Du Plessis, L., Fisher, C., Louw, S. (2010). First-level analysis report: HVS testing of the
ultra-thin reinforced concrete test sections on Road R80 near Soshanguve Gauteng: Tests 457
to 464. Prepared for the Gauteng Department of Public Transport Roads and Work, CSIR,
Pretoria, 2010.
Du Plessis, L., Strauss, P. J., & Killian, A. (2011). Monitoring the behavior of thin reinforced
concrete pavements through accelerated pavement testing. In Emerging technologies for
material, design, rehabilitation, and inspection of roadway pavements, GeoHunan 2011
(pp. 43–50). ASCE Journal Geotechnical Special Publication (GSP). GSP 218, ISBN (print):
978-0-7844-7629-1,http://dx.doi.org/10.1061/47629(408)6.
Gautrans Provincial Government. (2007). Framework for monitoring and evaluation of the
expanded public works program.
Gauteng Province. Department of Infrastructure Development. http://www.did.gpg.gov.za/Pages/
GEPWPBackground.aspx. 2007.
768 L. du Plessis et al.

Steyn, W. J. vdM., Strauss, P. J., Perrie, B. D., & Du Plessis, L. (2005). The Roodekrans trial
sections: The role of structural support under very thin jointed and CRC pavements subjected
to heavy traffic. In The 8th international conference on concrete pavements. Colorado Springs.
Strauss, P. J., Slavik, M., Kannemeyer, L., & Perrie, B. D. (2007). Updating cncPAVE: Inclusion
of ultra-thin continuously reinforced concrete pavement (UTCRCP) in the mechanistic,
empirical and risk-based concrete pavement design method. In International conference on
concrete roads, Midrand, South Africa.
Theyse, H. L., Muthen, M. (2000). Pavement analysis and design software (Pads) based on the
South African Mechanistic-Empirical Design Method. In 19th South African transportation
conference (SATC 2000).
The Design, Construction and Heavy
Vehicle Simulator Testing Results
on Roller Compacted Concrete Test
Sections at the CSIR Innovation Site
and on a Full-Scale Test Road at Rayton

L. du Plessis, G. Rugodho, W. Govu, K. Mngaza and S. Musundi

Abstract Although the use of Roller Compacted Concrete (RCC) is not new in
South Africa, the use of it to construct roads is not that well known or studied. The
Gauteng Provincial Department of Roads and Transport (GPDRT) in conjunction
with CSIR Built Environment in South Africa and Cosal Consultants CC started a
research program on the use of RCC technology for roads. Whereas RCC is nor-
mally constructed with a relatively low labor component using heavy mechanical
equipment, one of the aims of this investigation was to evaluate the structural
performance of RCC constructed with a relatively high labor component using
hand-operated equipment. The evaluation was done using the Heavy Vehicle
Simulator (HVS) of the GPDRT. This paper briefly details two investigations.
One HVS RCC test was conducted at the CSIR innovation site and the other on a
full-scale test road at Rayton, Gauteng. Through HVS testing it has been shown that
this type of pavement performed well in the dry state, even when constructed on a
substandard support system. Test results indicate that this type of pavement
exceeded its predicted performance. The use of hand-labor for layer compaction is

W. Govu—Deceased.

L. du Plessis (&)
CSIR Built Environment, Meiring Naudé Street,
Brummeria, Pretoria, South Africa
e-mail: [email protected]
G. Rugodho
W. Govu
K. Mngaza
Gauteng Provincial Department of Roads and Transport,
Private Bag X3, Lynn East, Pretoria 0039, South Africa
e-mail: [email protected]
K. Mngaza
e-mail: [email protected]
S. Musundi
Cosal Consultants CC, PO Box 863, De Deur 1884, South Africa
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 769

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_50
770 L. du Plessis et al.

discouraged as this can lead to layer densities lower than acceptable standards, which
result in poor performance. The importance of proper RCC mix design to mitigate
the negative effects of shrinking and crack forming is highlighted in this study.

Keywords Heavy vehicle simulator testing Roller compacted concrete (RCC)



Labor-based construction Pavement performance evaluation

1 Background

The upgrading of unsurfaced residential roads has become a priority for many
metropolitan areas in South Africa. Coupled with this is the need to construct roads
using labor-intensive construction techniques. In 2007, the Gauteng Provincial
Department of Roads and Transport initiated its Expanded Public Works Program
(EPWP). The goals and objectives of the EPWP are to alleviate unemployment,
targeting especially women, the youth and the disabled (Gauteng Provincial
Government 2007). This goal will be achieved by creating work opportunities in the
following four constructive ways:
• Increasing the labor intensity of government-funded infrastructure projects;
• Creating work opportunities in public environmental programs;
• Creating work opportunities in public social programs, and
• Utilizing general government expenditure on goods and services to provide the
work experience component of small enterprise learnership/incubation programs.
In line with these objectives the EPWP embarked on a research program to
improve the living standards of many rural communities by upgrading their resi-
dential streets to surfaced standards and to provide meaningful sustainable job
opportunities by using labor-based pavement construction techniques. To address
these needs, two suitable technologies were evaluated by the CSIR: Ultra-thin layer
reinforced concrete pavement (UTRCP) (Du Plessis et al. 2011) and Roller
Compacted Concrete (RCC).
This investigation deals with the evaluation of RCC using the HVS in full-scale
field trails. Whereas RCC is normally constructed with a relatively low labor
component and with heavy mechanical equipment, the one aim of this project was
to evaluate the structural performance of RCC using a relatively high labor com-
ponent and hand-operated equipment.
The investigation was conducted in phases. During the first phase a short
full-scale test section was constructed at the Accelerated Pavement Testing
(APT) facility of the CSIR in Pretoria. The second phase involved the construction
and testing of a full-scale test pavement on an existing gravel road near Rayton,
Gauteng (Route D1814). This paper details some of the interesting results and
analyses of these tests.
The Design, Construction and Heavy Vehicle Simulator Testing … 771

2 Objectives

The primary objective of this project is to assess the performance of RCC under real
field conditions through APT coupled with a laboratory-testing program. The
outcomes of the project can be used to update the South African design method for
rigid pavements developed by the Cement and Concrete Institute (C&CI) (Strauss
et al. 2001) as also to develop a South African structural design guideline for the
use of RCC in pavement construction.
More specific on the testing done with the HVS as reported in this paper, the
objective focused on the success of using of high percentage of hand-labor during
the construction of RCC. This was done through comparative evaluation of sections
constructed by labor-based practices and full-scale mechanical equipment.
This study is aimed at building confidence in the use of RCC as a structural
layer, with due cognizance being paid to pavement structure, construction, climate
and traffic.

3 Description of Roller Compacted Concrete

RCC gets its name from the heavy vibratory steel drum and rubber-tired rollers used
in compacting the material to its final form. RCC has similar strength properties and
consists of the same basic ingredients as conventional concrete such as graded
aggregates, cement and water, but with different mixture proportions (Harrington
et al. 2010).
The biggest difference between RCC mixtures and conventional concrete mix-
tures is that RCC has a higher percentage of fine aggregates, which allows for tight
packing, low void content and consolidation. Fresh RCC is stiffer than typical
zero-slump conventional concrete. Its consistency is stiff enough to remain stable
under vibratory rollers, yet wet enough to permit adequate mixing and distribution
of paste without segregation. The use of RCC on roads can potentially offer mul-
tiple benefits compared with more conventional approaches.
RCC is typically placed with an asphalt-type paver equipped with a standard or
high-density screed, followed by a combination of passes with rollers for com-
paction. Final compaction is generally achieved within one hour of mixing. Unlike
conventional concrete pavements, RCC pavements are constructed without forms,
dowels or reinforcing steel. Joint sawing is not required, but when sawing is
specified, transverse joints are spaced further apart than with conventional concrete
pavements. The low water-cement ratio of RCC results in less shrinkage crack
development than ordinary PCC mixes.
RCC mixtures should be dry enough to support the weight of a vibratory roller
after placement, yet wet enough to ensure adequate hydration and even distribution
of the paste. Compaction is the process by which the aggregate particles in the RCC
mixture are forced closer together, reducing the amount of air voids in the mixture
and increasing the density of the layer. The increased density makes the pavement
772 L. du Plessis et al.

suitable in load-bearing applications. Rolling must occur before cement hydration

begins to harden the paste between the aggregate particles. Achieving proper
density during the rolling process helps to prevent non-uniform consolidation and
isolated weak areas. Depending on the specific mixture and construction equipment
used, external mechanical compaction by rollers may result in a 5–20 % reduction
in volume (Portland Cement Association 2004). Minimizing the air-void content in
the RCC mixture is crucial to the durability of RCC. Excess air voids allow the
penetration of air and water. Non-entrained air weakens the mixture, while
excessive water can cause materials-related distresses in the aggregates, low field
densities and insufficient early strength gain.

3.1 Possible Benefits

The use of RCC on roads can potentially offer multiple benefits by comparison with
more conventional approaches. The primary benefit of RCC is that it can be con-
structed faster and more cost-effectively than conventional concrete. Other beneficial
characteristics of RCC include the following (Portland Cement Association 2004):
• The lower paste content in RCC results in less concrete shrinkage and reduced
cracking from shrinkage-related stresses;
• RCC can be designed to have high flexural, compressive and shear strengths,
which allow it to support heavy, repetitive loads without failure such as in heavy
industrial, mining and military applications and to withstand highly concen-
trated loads and impacts;
• With its low permeability, RCC provides excellent durability and resistance to
chemical attack, even under freeze-thaw conditions;
• RCC provides chemical and rut resistance in industrial areas where point
loading from trailer dollies is a concern;
• Occasional light vehicles, such as cars and light trucks, can travel at low speeds
on RCC pavements soon after completion without causing damage;
• Construction of RCC at ambient temperatures is suitable for labor-based con-
struction and, hence, is ideally suited for Gauteng’s EPWP;
• Apart from the basic road-building equipment (grader, roller compactor, water
cart), only simple, inexpensive construction equipment is required;
• The existing subgrade and alignment can be used;
• RCC can be used as an overlay on existing roads where there is no limitation on
vertical alignment;
• A significant percentage of the cement is replaced in the mix by fly-ash, a waste
product from coal-fired power stations;
• RCC surfaces require less lighting energy at night than bituminous surfaces
because of the reflectivity of this type of surface;
• Repairs of potholes and utility cuts are simple using the same material, and
• Water demand is lower than that of conventional concrete, which is beneficial to
the environment.
The Design, Construction and Heavy Vehicle Simulator Testing … 773

4 Methodology

The general methodology that was followed for the laboratory and field evaluations
of the RCC mixes was as follows.
During the first phase of the project two full-scale trial sections with a total
length of approximately 65 m and a width of 3.6 m were constructed at the CSIR’s
innovation site. For the construction of the first section a normal 10-ton vibratory
roller was used for compaction. The second section was constructed using
hand-labor where compaction was done using hand-operated Bomag rollers, even
for the compaction of the RCC layer (Du Plessis et al. 2012). During the second
phase, RCC construction was done at Rayton (Route D1814) where normal 10-ton
rollers were used for the compaction of all the layers (Du Plessis et al. 2013).
The RCC was subjected to a range of loading conditions and moisture regimes
while being tested with the HVS in channelized, bi-direction trafficking mode in the
center of the slab (interior loading).

5 Pavement Construction

The RCC test section at the CSIR innovation site was designed for local and
provincial low-volume roads. The test pavements were designed for a South
African Category B pavement, ES3 traffic class, which is classified as inter-urban
collectors, and rural roads with a total cumulative traffic design of between one and
three million equivalent standard 80 kN axles (E80s) over a structural design period
of 20 years. The composition of an ES3 traffic class is: a maximum of 20 % heavy
vehicles per day over the 20-year design period. This relates to approximately 340
heavy vehicles per day and at an average of 1.2 E80s per heavy vehicle (50 % laden
and 50 % unladen) (Committee of State Road Authorities 1985). The designs for
both sites (at the CSIR and Rayton) were supplied by Cosal Consultants CC. The
South African pavement materials design manual classifies granular layers from the
highest (G1) to the lowest (G9) quality and cementitious materials from the
strongest (C1) to the weakest (C4) (Committee of State Road Authorities 1985).
The design specifications for the RCC test section at the CSIR innovation site
were:
• Subgrade: Min. CBR of 25 at 95 % Mod AASHTO, PI < 12, Max swell 1 %;
• Subbase: 150 mm thick in situ material compacted to 93 % Mod AASHTO;
• Base: 150 mm thick G5 quality imported material stabilized with 3 % cement
(of which 20 % was replaced with fly-ash), compacted to 95 % Mod AASHTO
(to conform with C4 specifications), and
• RCC: 150 mm thick layer mix design according to consultant’s specification.
The subgrade was ripped and re-compacted. After compaction, the in situ
material had an average California Bearing Ratio (CBR) of 58 and a field density of
774 L. du Plessis et al.

1 788 kg/m3 (94 % Mod. AASHTO), which was above the specified South African
limit of a minimum CBR of 25 at 95 % Mod AASHTO.
The density of the subbase was measured with a nuclear density gauge after
compaction. The subbase was only compacted to 90.7 % Mod AASHTO, 2.3 %
short of the target density of 93 % Mod AASHTO.
For the base material, the in situ material was used (instead of the specified G5
material) and stabilized with 2.4 % cement and 0.6 % fly-ash. Laboratory results
indicated that after stabilization, the base material had an unconfined compressive
strength (UCS) of 244 kPa. This is lower than the acceptable standard for a South
African C4 type stabilized base with a specified minimum UCS of 750 kPa. The
field compaction was 92.3 % Mod AASHTO, which was also lower than the
specified minimum limit of 95 % (Du Plessis et al. 2012).
The RCC layer was constructed with river sand and 9.5 mm crushed quartzite
stone. The mix design consisted of a cement:sand:stone ratio of 1:2:3. As in the
case of the stabilized base, 20 % of the cement content was replaced by fly-ash.
A CEM III A type (32.5 N) cement was specified. Apart from the fly-ash, no other
type of additive or air entrainment agent was used.
The design specifications for the RCC test section at the Rayton site were (Du
Plessis et al. 2013):
• 150 mm RCC;
• 150 mm C4 base material (G5 parent material) compacted to 97 % Mod
AASHTO;
• 150 mm G5 upper select subbase compacted to 95 % Mod AASHTO;
• 150 mm lower select subbase compacted to 93 % Mod AASHTO, and
• 125 mm G7 fill compacted to 90 % Mod AASHTO.
The subgrade investigation revealed that the top 150 mm should be reworked
and compacted to at least G9 material quality in order to satisfy the design
requirements for a ES1 to ES10 pavement design (between 300 000 and 10 million
equivalent 80 kN standard axles).
The RCC mix design as supplied by Cosal Consultants CC was:
• CEM II 32.5 N: 12.2 %
• Fly-ash: 5.2 %
• Sand: 45.4 %
• 13 mm stone: 37.2 %
• Water/Cement ratio: 0.53
The as-built pavement conformed to the design specifications with small vari-
ations. The concrete was supplied by a ready-mix supplier and no information on
the true field mix was made available for evaluation.
The Design, Construction and Heavy Vehicle Simulator Testing … 775

6 HVS Results

Due to space limitations only the most significant HVS results are reported. The
results of three HVS tests conducted on three different sections are summarized: the
section constructed with mechanical equipment, the section constructed using
hand-labor, and the full-scale section on Route D1814 at Rayton.

6.1 HVS Test 467A5 RCC Placed with Mechanical

Equipment

The section was tested in a dry state, ambient temperature and a standard half-axle
load of 40 kN for one million repetitions after which it was increased to 60 kN until
failure. The 8 m long HVS test pad on the 65 m section was selected so that the
HVS testing wheel ran over a shrinkage crack, as this was considered to be the
weakest area in that test section (Du Plessis et al. 2012). The pavement failed after
1.1 million repetitions, which equates to 7 million E80s using the accepted expo-
nential damage factor of 4.2. This surpasses the design life of the pavement which is
between 1 and 3 million E80s. Structural failure occurred in the form of mainly
transverse cracks and was concentrated around the shrinkage crack.

6.1.1 Elastic Deflection Response

Figure 1 shows both the surface deflections captured by the joint deflection mea-
suring devices (JDMDs) and multi-depth deflectometers (MDD). MDD 12 and
JDMD 1 and 2 were placed in close proximity of the shrinkage crack. Both
instruments (JDMD and MDD) recorded increasing deflections downwards
(although presented differently in the figure for the ease of distinguishing between
JDMD and MDD recorded results). MDDs were installed at the surface and in the
base (200 mm deep) and in the subbase (400 mm deep). The variations in the
elastic deflections (Fig. 1) are mainly due to daily temperature variations and the
increased load for 40–60 kN after a million repetitions. The sudden drop in
deflection at approximately 1.6 million repetitions is the result of aggressive
additional crack formation around the instrumented shrinkage crack towards the end
of the test. MDD data indicated that most of the deflections originated between the
surface and the base (200 mm from the surface).
It is obvious that a significant degree of separation between the bottom of the
concrete and the top of the base occurred during the 60 kN cycle. MDD deflections
776 L. du Plessis et al.

Fig. 1 Elastic deflections recorded with JDMDs and MDDs on the normally constructed section

on the surface were higher than 1.7 mm in comparison with base deflections where
the maximums were lower than 0.8 mm. Interesting to note that the permanent
deformation on the rest of the area away from the shrinkage crack was low (JDMD
3, 5 and 6).

6.1.2 Load-Transfer Efficiency

Load-transfer efficiency (LTE) across the shrinkage crack was recorded and are
shown in Fig. 2. LTE indicates the efficiency of how loading is transferred and
carried by the concrete on either side of the crack. It is defined as the relationship
between deflection of the unloaded slab (ΔU) and the deflection of the loaded slab
(ΔL), or: LTE (%) = ΔU/ΔL × 100. An LTE of 100 % is consistent with full-load
transfer, and an LTE of 0 % indicates zero load transfer (slabs fully separated by the
formation of a crack of significant crack width).
By considering the above, it is possible to quantify the effectiveness of the load
transferred through pavement deflection across a crack or a joint. The LTE was
relatively high (above 90 %) during the first 600,000 repetitions which dropped to
below 40 % after the load was increased to 60 kN. This drop in LTE (from over
90 % to below 40 %) signifies a significant loss in aggregate interlock across the
shrinkage crack after the application of approximately 1.2 million load applications.
The big variations in the data are due to day-night variations in LTE values due to
slab expansion (daytime) and contraction (nighttime).
The Design, Construction and Heavy Vehicle Simulator Testing … 777

Fig. 2 Load-transfer efficiency and relative movement results at test 467A5

6.1.3 Permanent Deformation

Concrete pavements do not traditionally fail in rutting, but due to the weak sub-
structure, the granular layers experienced a significant amount of permanent
deformation that was observed at the surface after significant crack formation

Fig. 3 Permanent deformation recorded at test 467A5

778 L. du Plessis et al.

around the original shrinkage crack. The permanent deformation recorded by both
JDMDs and MDDs is illustrated in Fig. 3. Similar to the deflection figure (Fig. 1),
both instruments recorded increasing deformation downwards (although presented
differently in the figure). Surface deformation of over 18 mm was recorded at the
end of the test in the vicinity of the shrinkage crack (JDMD 1 and 2). This caused
the slab to crack up to such an extent that no further testing was possible. Interesting
to note that the permanent deformation on the rest of the area away from the
shrinkage crack was low (JDMD 3, 5 and 6).

6.2 HVS Test 469A5 Hand-Constructed RCC Section

Test 469A5 was the first test conducted on the labor-intensive constructed section.
Testing was done in both the dry and wet states to simulate the effects of dry and
wet periods. The dry/wet cycles were based on five days of trafficking under dry
conditions, followed by two days of trafficking under wet conditions. This cycle
was then being repeated until failure (Du Plessis et al. 2012). Water was introduced
in this fashion to simulate the natural random occurrence of rain. Water was
introduced from 227,500 repetitions onwards. Testing was completed after 609,036
repetitions when the section was considered to have failed. The section lasted for
three wet cycles and failed within 37,000 dry repetitions after completion of the
third wet cycle. A significant degree of pumping was observed during the wet
cycles which caused cavities under the concrete that resulted in high surface
deflections, excessive block cracking and severe surface deformation. This caused
the early failure of this test section.
Reasons for the poor performance include:
• The subbase layer was not compacted properly as detailed in Sect. 5;
• The base layer (C4) was not constructed at optimum moisture content. Although
the laboratory mix design suggested that the optimum moisture content should
be 6.9 %, the field mix moisture content was on average between 3.08 % and
3.3 %. This low water content negatively influenced the ability of the cement in
the granular material to hydrate properly and lower material strengths were
achieved than theoretically possible. The cemented material had an average
UCS value of 245 kPa at field moisture contents. This value is significantly
lower than the minimum prescribed value of 750 kPa for a C4 type cemented
base material (Committee of State Road Authorities 1985), and
• RCC layer: The average field thickness of the concrete layer was 145 mm
instead of the prescribed 150 mm and the maximum 28-day strength of the field
mix was 15.63 MPa (field-cored samples). This is less than that expected from
CEM III 32.5 cement. A possible reason for the lower strength may be insuf-
ficient compaction of the RCC. Samples cored from the section showed big
voids in the RCC field mix. A picture showing the failure is presented in Fig. 4.
The Design, Construction and Heavy Vehicle Simulator Testing … 779

Fig. 4 Failure after 609,036 repetitions on the hand-constructed RCC layer

6.3 HVS Test 470A5 RCC Constructed at Rayton

The shrinkage crack pattern of the full-scale field test section constructed on Route
D1814 (at Rayton) is shown in Fig. 5 and includes the HVS test pad with respect to
the 100 m test section (Du Plessis et al. 2013).
The HVS was positioned over the first shrinkage crack which appeared within
10 days of construction approximately 19 m from the eastern end of the 100 m
RCC test section. It should be noted that the first 30 m (the grey area on the
right-hand side in the figure) was constructed with the aid of an asphalt paver, the
remaining 70 m being placed 5 days later using hand labor and a 10-ton vibrating
compaction roller. Ready-mix concrete was used for the whole 100 m section.

Fig. 5 HVS test pad placed across a shrinkage crack

780 L. du Plessis et al.

Although not accurately measured, the section constructed with the asphalt paver
had smaller crack widths (and more cracks per length of road) than those developed
during the hand-placed section (less cracks but with increased crack widths).
The first 1.207 million repetitions were done in the dry state using a 40 kN load
(simulating an 80 kN standard axle load). Due to budget and time constrains testing
was stopped after 2.5 million bi-directional repetitions. Because of the slow rate of
deterioration, the load was increased to 60 kN (simulating a 50 % overload) from
1.207 million repetitions onwards. In addition to this, water was introduced to the
surface from 1.76 million repetitions onwards. Water was introduced in the same
manner as the previous RCC test (five dry days followed by two wet days, then
repeat). Using the 4.2 exponential power damage law, a total of 8.3 million standard
80 kN axle loads had been applied without any signs of visual damage. This exceeds
the requirements of an ES3 design class (between 1 million and 3 million E80s).
The surface and in-depth deflections as measured by the MDDs are shown in
Fig. 6. The surface deflections are unrealistically high (over 1.7 mm) in the vicinity
of the shrinkage crack and the surface deflections away from the crack (not shown
in the figure) are low (less than 0.1 mm). The conclusion drawn from this is that a
significant amount of permanent transverse slab warping occurred at the shrinkage
crack, which caused a cavity and the concrete separated from the base. This con-
clusion is substantiated by the in-depth deflections as recorded by the MDD
modules placed in the base (170 mm) and subbase (330 mm): The deflection
module placed at the top of the base (at 170 mm in depth) recorded deflections less
than 0.1 mm in comparison with the surface deflections (over 1.6 mm). It is
obvious that the area surrounding the crack was structurally in a poor state in
comparison with the rest of the testing area.

Fig. 6 MDD surface and in-depth deflections at the facility of the shrinkage crack
The Design, Construction and Heavy Vehicle Simulator Testing … 781

Fig. 7 Deflection bowls showing the drop in load-transfer efficiency across a shrinkage crack

LTE was also investigated at the crack (Fig. 7). The figure shows the deflection
bowls captured by various JDMD instruments placed opposite each other, across
the shrinkage crack, during the hottest part of the day (pm) and during the coldest
part at night (am). After 580,000 repetitions a significant drop in the load transfer
can be seen, especially at night when a negative temperature differential (surface
cooler than the bottom of the concrete) caused a significant degree of slab curing.
The deflections were virtually zero during the day due to slab expansion and
downward curling effects.

7 Conclusions and Recommendations

This paper summarizes the most significant results and findings of three HVS tests
done on RCC pavement structures in South Africa. Although the sections have
lasted their design life, the differences in the behavior at shrinkage cracks in
comparison with uncracked areas are of concern. The importance of proper design
of all the pavement layers is highlighted in this study. Deviations from the design
specifications can lead to costly early failures. To meet the performance require-
ments, the RCC layer should be placed on top of a well-designed base course
material which should not be water-sensitive (should water enter the pavement
though surface cracks). Long-life pavement performance of concrete type structures
is governed, inter alia, by the following four important guiding principles:
782 L. du Plessis et al.

1. The concrete should be designed in such a way that any uncontrolled shrinkage
cracks are sufficiently small to ensure that good aggregate interlock is main-
tained and that the ingress of water into the lower layers is prevented;
2. The surface deflections should be as low as possible to prevent hydraulic action
which causes pumping of the support material under the influence of traffic and
water;
3. The loss of material as a result of pumping should be minimized through (1), as
well as through proper design of the base layer in order to ensure that the base
layer has sufficient resistance against water erosion, and
4. It should be realized that temperature plays an important role in the performance
of concrete pavements and that any concrete mix should be designed to with-
stand the build-up of stresses resulting from changes in temperature and tem-
perature gradients within the concrete.
The negative impact of using hand-labor on the structural life of RCC is illus-
trated in this study. Due to the low water-cement ratio and the granular structure of
RCC mixes, good compaction can only be achieved when heavy full-scale vibrating
rollers are used (as opposed to light-weight hand-operated compacters). A significant
drop in structural life was observed during the HVS test compacted by hand-
operated equipment. During a post-mortem investigation, cores taken from the
hand-compacted section revealed cavities and voids inside the concrete. This
obviously caused a significant reduction in the bearing strength of the concrete layer.
The hand-placed (but compacted by a 10-ton roller) section at Rayton proved to
be a greater success but care should be taken during the mix design and con-
struction to ensure that shrinkage cracking is well controlled, both in terms of crack
spacing and crack widths.

Acknowledgments This HVS research project is supported by the Gauteng Provincial

Department of Roads and Transport and their financial and technical contribution towards this
study is greatly acknowledged. The authors thank Mr. Steve Musundi, the managing director of
Cosal Consultants CC, who did the foundational work, was responsible for the design and con-
struction, and provided input during the execution of this research program.

References

Committee of State Road Authorities. (1985). TRH 14: Guidelines for road construction materials.
Pretoria.
Du Plessis, L., Fisher, C., & Leyland, R. (2013). HVS testing of roller compacted concrete on road
D1814, Cullinan: Draft preliminary report. CSIR contract report, December, 2013.
Du Plessis, L., Louw, S., Fisher, C., & Leyland, R. (2012). First level analysis report: HVS testing
of roller compacted concrete test 467A5 – 469A5. CSIR contract report, July, 2012.
Du Plessis, L., Strauss, P. J., & Kilian, A. (2011). Monitoring the behavior of thin reinforced
concrete pavements through accelerated pavement testing. ASCE Journal Geotechnical Special
Publication (GSP).
The Design, Construction and Heavy Vehicle Simulator Testing … 783

Gauteng Provincial Government. (2007). Framework for Monitoring and Evaluation of the
Expanded Public Works Programme, June, 2007. http://www.epwp.gov.za/documents.html.
Accessed 5 May 2016.
Harrington, D., Abdo, F., Adaska, W., & Hazaree, C. (2010). Guide for roller-compacted concrete
pavements. Iowa: Iowa State University, National Concrete Pavement Technology Center.
Portland Cement Association. (2004). Roller-compacted concrete density: Principles and
practices. Skokie, Illinois.
Strauss, P. J., Slavik, M., & Perrie, B. D. (2001). A mechanistically and risk based design method
for concrete pavements in South Africa. In Proceedings of the 7th International Conference on
Concrete Pavements, Orlando, Florida.
 Part X
Accelerated Pavement Testing to Evaluate
Functional Performance
Assessment of Flexible Pavement Response
During Freezing and Thawing
from Indoor Heavy Vehicle
Simulator Testing

Ahmed El-youssoufy, Guy Dore, Jean-Pascal Bilodeau

and Fritz Prophète

Abstract During the spring thaw period, pavements experience a significant

decrease of their bearing capacity, which leads to increased deterioration during that
season. The Ministry of Transportation of Quebec enforces load restrictions during
the thawing and the recovery of the pavement structures to ensure they are protected
from excessive thaw associated damages. The main objective of this paper is to
present the results of an investigation of the structural behaviour of pavement
structures during freezing and thawing under laboratory controlled conditions using
the Laval University Heavy Vehicle Simulator (HVS). The results of the study will
be used to develop criteria to assist the authority in the decision making process for
the implementation and the removal of load restrictions. A typical four-layer
flexible pavement structure was built inside an indoor concrete test pit at Laval
University. The pavement section was instrumented to monitor horizontal strains in
the asphalt concrete layer as well as vertical stress, vertical strain, and water content
in each unbound layer and in the subgrade soil. Temperature was also monitored in
all layers. The new HVS was used to impose 5000 kg (normal conditions), 5500 kg
(winter premiums) and 4000 kg (spring load restrictions) loads on a standard
dual-tire assembly. The load simulator was also used to induce freezing up to a
depth of 1.5 m in the pavement structure and to induce thawing. The results
obtained showed that during the thawing, the maximum strains occurred when the
thaw front reached a depth of 0.9 m. Finally, it was found that the spring load
reduction policies used in Quebec Province reduce the weighted annual damage by
approximately 13 % during thawing, the strains and stresses decreased rapidly
when the frost front reached subbase. Pavement damage caused by the standard
load in these conditions can be considered negligible.

Keywords Flexible pavements
Freezing and thawing
Heavy vehicle simula-

tor Structural response

A. El-youssoufy (&)
G. Dore
J.-P. Bilodeau

Laval University, Quebec City, Canada
e-mail: [email protected]
F. Prophète
Ministry of Transportation of Quebec, Quebec City, Canada

© Springer International Publishing Switzerland 2016 787

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_51
788 A. El-youssoufy et al.

1 Introduction

In cold regions, the pavement performance is influenced by environmental factors

and traffic loads (Doré and Zubeck 2009). During the freezing and the thawing
processes, the change of unfrozen moisture content and temperature plays a sig-
nificant role in the pavement response to loading. During winter, ice lens growth in
frost susceptible subgrades results in surface frost heave. In addition, the freezing of
the pavement structure is known to cause a significant increase of the pavement
bearing capacity. During the spring thaw period, the excess ice accumulated in soils
and in the pavement structure melts, causing an increases of the water content of
unbound layers and soils. As a result, pavements experience a drastic bearing
capacity reduction and a significant increase in permanent deformation and fatigue
damage under traffic loads (Cary and Zapata 2011; Erlingsson 2012; Salour 2015).
The objective of this project is to better understand the response of pavement
structures during freezing and thawing and to improve the prediction of the
gain/loss and recovery of the bearing capacity as a function of the evolution of the
freezing and the thawing processes as well as temperatures in the pavement. For
that purpose, an indoor flexible pavement section was built in an environmentally
controlled concrete pit at the geotechnical laboratory of Laval University.

2 Test Sites and Instrumentation

A typical four-layer flexible pavement structure was built in an indoor concrete test
pit at Laval University, the dimension of the test pit is 2 × 6 × 2 m3. The pavement
section was instrumented to monitor horizontal strains at the bottom of the asphalt

Fig. 1 Pavement structure and instrumentation layout

Assessment of Flexible Pavement Response During Freezing … 789

concrete layer as well as vertical stress, vertical strain, and water content in each
unbound layer and in subgrade soil. Temperature profile was also monitored in all
layers with a 2 m long thermistor string embedded in the pavement structure. The
pavement structure consists of a 100 mm asphalt concrete layer, a 200 mm granular
base, a 450 mm granular subbase resting on a silty sand subgrade (SM). The
detailed information on the pavement structure and on the instrumentation layout is
shown in Fig. 1.

3 Methodology

The Heavy Vehicle Simulator (HVS) was installed on top of the test pit. As shown
in Fig. 2, the simulator has a loading carriage equipped with a standard dual-tire
assembly to simulate real traffic loading conditions. The simulator is also equipped
with a heating and cooling system that allows inducing freezing and thawing
temperatures in the pavement structure by controlling the temperature at the

Fig. 2 a Laval University Heavy Vehicle Simulator (HVS); b test setup with insulation panels;
c loading carriage
790 A. El-youssoufy et al.

Table 1 Imposed conditions during the test process

Cycle Water ground Freeze or thaw Frost Loads frequency
level (m) temperature (°C) depth (m)
1st freeze 1.6 −10 1.5 5000 and 5500 kg
1st thaw 1.6 10 – 5000 and 4000 kg

pavement surface. The simulator has insulation side panels around the simulator to
help maintain the desired temperature. In order to freeze the pavement structure up
to a depth of 1.5 m, a temperature of −10 °C was imposed at the surface of the
pavement. Meanwhile, to represent the real in situ thermal regime, a temperature of
2 °C was maintained at the bottom of the tested pavement structure using a closed
circuit cooling glycol system embedded in the concrete slab at the bottom of the test
pit. In order to obtain moderate frost heaving, the groundwater level was adjusted to
1600 mm from the pavement surface.
During the testing process, the simulator was adjusted to apply 5000 kg on the
dual tire assembly, which corresponds to the legal load for half a single axle in
Quebec. The speed for the carriage through the loading procedure was set at 5 km/h
and the tire pressure was fixed at 700 kPa. The pavement response measurements
were collected periodically with the frost or thaw front progression. The frequency
of the measurement was initially five times a day, and was gradually reduced to two
or three times per week towards the end of the freezing period. When analyzing the
data, only peak response values were used to quantify the structural behaviour
variations of the experimental pavement. The temperature profile of the pavement
structure was collected at each response measurement to determine the frost line
position which was determined using linear interpolation between the position of
the two adjacent thermistors indicating temperatures above and below 0 °C. The
freezing process stopped when the frost line reached 1.5 m. In the thawing pro-
cedure, a constant temperature of 10 °C was applied at the surface until the whole
pavement structure was thawed and a stabilized mechanical response was obtained.
The conditions imposed on the pavement structure during this cycle are showed in
Table 1.

4 Results and Discussion

First, in order to better understand the mechanical behaviour variations of the

pavement structure, the strain and stress responses under loading at a given time t
during freezing or thawing were normalized by using the initial measurement before
freezing, which is considered to be optimal conditions. The normalized strain and
stress in each layer were calculated using the following equation:
Assessment of Flexible Pavement Response During Freezing … 791

ðe or rÞat time t during thawing

Normalized ðe or rÞ ¼ ð1Þ
ðe or rÞInitial ðbefore freezingÞ

where e ¼ microstrainðleÞ and r ¼ stressðkPaÞ.

4.1 Strain Variations

Figure 3 presents the evolution of strains for all layers in the pavement structure
during freezing with 5000 kg load as a function of the square root of the freezing
index. At the beginning, the frost depth increases gradually until all the asphalt
concrete becomes frozen. During that time, cooling of the asphalt concrete is
essentially responsible for the decrease in tensile strains and vertical strains at the
bottom of asphalt concrete and in the unbound layers, respectively. When the frost
front progresses through the base layer, the tensile strains in the asphalt concrete
and the vertical strains in the base layer are strongly reduced until they reach a low
and relatively stable level when the frost front reaches 0.45 m or when the freezing

Fig. 3 Strain variations in

the pavement structure during
freezing at 5000 kg
792 A. El-youssoufy et al.

index equals 25 °C days (5 (°C days)0.5). The vertical strains in the subbase and
in the subgrade decrease progressively and tend to stabilize when the frost
front reaches approximately 0.9 m or when the freezing index equals 144 °C days
(12 (°C days)0.5) which correspond to the depth at which all the instruments are in
the frozen zone.
The strains variations during the thawing process are presented in Fig. 4. It is
possible to notice on the figure that the strains of all the pavement layers increase
significantly as the thaw front enters the unbound layers. The strains in all layers
increased to their maximal value when the thaw front reached a depth of 0.9 m. The
increase in strain was caused by the increase of the temperature of asphalt concrete
and of the unfrozen moisture content in the unbound layers. After that, the strains
decreased until the complete recovery of the pavement, because of the decrease in
water content. At this time, water contents were 11, 7 and 6 % in the subgrade,
subbase and base respectively. The strains then tended toward a constant value
which represents optimal conditions during summer. During this thawing process,
the strain signal in the subgrade was lost.

Fig. 4 Strain variations in

the pavement structure during
thawing process at 5000 kg
Assessment of Flexible Pavement Response During Freezing … 793

Fig. 5 Stress variations in

the pavement structure during
freezing at 5000 kg

4.2 Stress Variations

Figure 5 shows the evolution of the stress in unbound layers and soil during
freezing as a function of the freezing index at 5000 kg. From this figure, it is
possible to observe that the stresses in the all layers decrease when the frost front
penetrates in the pavement structure. At first, the three measured stress values in the
granular and soil layers have a similar decrease rate when the frost front reached
0.30 cm. Then the stress in the base decreases rapidly when the frost front goes
through the base layer, but the stresses in the subbase and soil continue to decrease
with a similar rate. Finally, all the stresses approach constant values when the
pavement structure is completely frozen up to a depth of about 1.5 m.

4.3 Frost Heave

The frost heave measured during the freezing of the pavement structure is shown in
Fig. 6. It is possible to observe that there is a linear relationship between time
and frost heave. The average frost heave rate is approximately 0.02 mm/h.
Corresponding surface heaving was only observed when the frost front reached the
subgrade. Finally, the maximum frost heave was 13 mm, which was experienced
when the pavement structure was completely frozen up to a depth of 1.5 m.
The evolution of stresses during thawing is shown in Fig. 7. It is possible to
observe that the thawing of the asphalt concrete leads to the increase of base and
subbase stresses. The peak stress value was measured at a thawing index of
794 A. El-youssoufy et al.

Fig. 6 Frost heave

Fig. 7 Stress variations in

the pavement structure during
thawing at 5000 kg

90 °C days (9.49 (°C days)0.5) which corresponds to the time when the thawing
front reached the bottom of the instrumented zone. After the complete thawing of
the pavement structure, stress values gradually decrease until the complete recovery
of the bearing capacity. The stresses tend toward a constant value after the complete
recovery.
Assessment of Flexible Pavement Response During Freezing … 795

Fig. 8 Change in modulus during the thawing process

4.4 Modulus of Unbound Pavement Layers

The resilient modulus of the unbound layers was calculated by the following equation:
r
M¼ ð2Þ
e

where σ = stress (kPa) and e ¼ strainðleÞ. Also, the evolution or change of the
modulus was normalized by using the initial value before freezing. The evolution of
the unbound layers modulus is shown in Fig. 8. In general, the unbound layers
modulus decreases during the thawing process. When the pavement structure is
completely thawed, the unbound layers modulus value is only 50 to 80 % of the
initial value before freezing and thawing. That could be associated with some
damage that may have occurred in the granular layers during the thawing process.

5 Analysis

5.1 Fatigue Damage Model

The damage is calculated as the inverse of the allowable number of load repetitions
calculated from empirical transfer functions. During the pavement design process,
the annual damage calculated using pavements in optimal conditions (summer) is
796 A. El-youssoufy et al.

considered as reference damage. The allowable number of load repetitions was

calculated by using the Asphalt Institute model (Carpenter 2006). The allowable
number of loads was calculated by the following equation:
 kF2  kF3
1 1
NF ¼ C
KF1

ð3Þ
e E

where ε (strain (µε)) = horizontal strain at the bottom of the asphalt concrete, E
(MPa) = resilient modulus of the asphalt concrete and C, kf1, kf2 and kf3 are the
calibration coefficients of the model.
The results of the analysis are shown in Fig. 9, where it can be observed that the
damages caused by the 5000 kg load are more significant than the damages caused
by the 4000 kg load. Also, the damage increases with the thaw front penetration.
Finally, when the thawing of the pavement structure is completed, the pavement
structure gradually recovers its bearing capacity, and then the damage tends toward
a constant value that reflects the optimal conditions during summer.
The calculation of an annual total damage was made considering 6 months of
summer, 1.8 months of spring and 4.2 of winter, as well as with the area under
curve (AUC) for each season considered. Also the area under curve (AUC) during
winter is considered equal to zero because of the insignificant damage during this
season. The selected season lengths are representative of southern Quebec climatic
conditions. Equation (4) was used to calculate the total damage.

1:8  ðspring AUCÞ þ 4:2  ðwinter AUCÞ þ 6  ðsummer AUCÞ

Total damage ¼
12
ð4Þ

Fig. 9 Fatigue damage analysis during restrictions period

Assessment of Flexible Pavement Response During Freezing … 797

The damages caused by 5000 and 4000 kg loads are equal to 0.94 and 0.82
respectively. Therefore, for the case and conditions considered the spring load
restrictions of 20 % (5000–4000 kg) cause a decrease of total annual damage of
13 %.

6 Conclusion

The assessment of the flexible pavement response during freezing and thawing was
investigated using an indoor heavy vehicle simulator. During freezing, it was found
that the strain of asphalt concrete and base layer reached their maximal reduction
when the frost front reached 0.45 m (at the bottom of the granular base layer) at a
freezing index of 49 °C days. The strain for the subbase layer and subgrade reached
their maximal reduction when the frost front reached 0.98 m (underneath the
deepest strain sensor) at a freezing index of 193 °C days. During thawing, the
strains in all layers reached their maximum values when the thaw front reached
0.9 m (underneath the deepest strain sensor) at a thawing index of 90 °C days. In
addition, it was found that the maximum frost heave was 13 mm, which was
experimented when the pavement was completely frozen up to a depth of 1.5 m.
Finally, the calculation of the annual damage found that the damage caused by 5000
and a 4000 kg loads are equal to 0.94 and 0.82 respectively. Therefore, for the case
and conditions considered the spring load restrictions of 20 % (5000–4000 kg)
cause a decrease of total annual damage of 13 %.

Acknowledgments The research team of this project would like to thank the Quebec Ministry of
Transportation for their financial, technical and professional support for this project. The members
would like also to thank Sylvain Auger for his great help with the test preparation.

References

Carpenter, S. H. (2006). Fatigue performance of idot mixtures. Research Report

FHWA-ICT-07-007, p. 4.
Cary, C. E., & Zapata, C. E. (2011). Resilient modulus for unsaturated unbound materials. Road
Materials and Pavement Design, 12, 615–638.
Doré, G., & Zubeck, H. K. (2009). Cold regions pavement engineering. Reston, VA: American
Society of Civil Engineers Press.
Erlingsson, S. (2012). Rutting development in a flexible pavement structure. Road Materials and
Pavement Design, 13(2), 218–234.
Salour, F. (2015). Moisture influence on structural behaviour of pavements. Doctoral Thesis,
School of Architecture and the Built Environment, Royal Institute of Technology, Stockholm,
Sweden.
Development of IRI Models Based on APT
Data

José P. Aguiar-Moya, Pablo A. Torres-Linares,

Edgar Camacho-Garita, Fabricio Leiva-Villacorta
and Luis G. Loría-Salazar

Abstract As with most pavement deterioration models, pavement monitoring is

required during long time in order to properly capture deterioration trends. In this
sense, roughness modeling, whether based on subjective parameters such as ser-
viceability or indicators such as the International Roughness Index (IRI), requires
significant amounts of data on structural pavement distresses such as rutting,
patching, and cracking. Consequently, the initial roughness after construction, the
evolution of roughness and surface deterioration in time have to be properly doc-
umented. Accelerated pavement tests permit an enhanced roughness data generation
process. Furthermore, one of the most important factors in determining the future
roughness is roughness after construction that can be carefully monitored and
measured. Consequently, an IRI model has been developed based on results
obtained at the PaveLab facility containing an HVS equipped with dual profile
lasers for monitoring roughness along each test section. The model is based on four
experimental sections, each of which has been divided into several tracks to make
use of the collected data. The model was defined as a function of loss of support, as
measured by means of layer moduli, surface rutting, initial IRI and number of load
repetitions. Model estimation was based on a random effects model to account for
unobserved heterogeneity and using the aforementioned parameters to reduce
instrumental variable bias which can occur when using transfer function estimates

J.P. Aguiar-Moya (&)
P.A. Torres-Linares
E. Camacho-Garita
F. Leiva-Villacorta

L.G. Loría-Salazar
LanammeUCR, CP-11501-2060, San José, Costa Rica
e-mail: [email protected]
P.A. Torres-Linares
e-mail: [email protected]
E. Camacho-Garita
e-mail: [email protected]
F. Leiva-Villacorta
e-mail: [email protected]
L.G. Loría-Salazar
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 799

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_52
800 J.P. Aguiar-Moya et al.

as proxy for fundamental material properties in the model estimation process. The
estimated parameters showed an increase in efficiency, as compared to Ordinary
Least Squares estimated parameters.

1 Introduction

The functional performance of a pavement structure can be quantified in terms of

smoothness or roughness of the road surface, which in turn are related to ride
quality (Arhin et al. 2015). The concept of riding quality was first incorporated in
pavement design methodologies as a result of the American Association of State
Highway Officials (AASHO) Road Test (AASHO 1962) and was measured in terms
of serviceability, which is a subjective measure of how a panel of users perceive the
quality of the pavement (Present Serviceability Rating, PSR). Serviceability in turn
was related to several distress types found to play a major role in reducing func-
tional performance as follows:

PSI ¼ 5:03  1:91 logð1 þ SV Þ  0:01ðC þ PÞ0:5 1:38RD2 ð1Þ

where PSI = present serviceability index (an estimate of the mean panel service-
ability rating); SV = slope variance; C = major cracking in feet per 1000 ft2 area;
P represents patching in square feet per 1000 ft2 area; and RD = average rut depth
of both wheel paths in inches measured at the center of a 4 ft span in the most rutted
part of the wheel path.
The PSI prediction models were later modified to include variables such as rut
depth variance and depressions, in order to improve the predicting capabilities of
the model (Darter and Barenberg 1976; Al-Omari and Darter 1992).
Because of the subjectivity associated to PSR, an effort was performed to
develop other measures of pavement functional level. An important aspect
regarding PSR is that it was previously identified to be highly related to pavement
performance and roughness (Carey and Irick 1960). The relationship was later
demonstrated by several DOTs when relating their PSI measures or local PSR
ratings to roughness values measured by different means (Moore et al. 1987).
In the late 1970s, the NCHRP Report 228 related PSI to roughness by means of
the International Roughness Index (IRI). During the 1980s, IRI was adopted by the
World Bank as standard measure of roughness in units of in/mi, m/km, or an
equivalent unit (Sayers et al. 1986). IRI can be measured by several automated
equipment, but is typically measured by a laser profiler. Several relationships have
been developed by analyzing and comparing IRI and PSI data. One such example
was presented by Paterson (1986) and is structured as a direct nonlinear relationship
(IRI measured in inches per mile):
Development of IRI Models Based on APT Data 801

PSI ¼ 5 expð0:18
IRI Þ ð2Þ

A similar model was developed by analyzing data from Louisiana, Michigan,

New Jersey, New Mexico, Ohio, and Indiana (Al-Omari and Darter 1992):

PSI ¼ 5 expð0:26
IRI Þ ð3Þ

All of the previous models were developed by means of Ordinary Least Squares
(OLS), under assumptions of normality and no serial correlation. However, with the
current widespread use of profilers, which are capable of measuring at high speeds,
the need for developing IRI models as a function of pavement condition has
increased. This is the case of most current pavement design procedures that include
a functional failure criteria such as AASHTO’s Pavement Design ME, which was
originally proposed under NCHRP 1-37A (Von Quintus et al. 2001a, b; Von
Quintus and Yau 2001).

1.1 Current IRI Models

Because of the relatively constant rate of change of IRI with time, the current model
that was proposed under NCHRP 1-37A consists of a linear function of different
distress types and subgrade properties. The initial functional form proposed for IRI
was as follows,

IRI ¼ IRI0 þ DD þ DF þ DS ð4Þ

where IRI0 corresponds to the initial IRI after construction, DD is the effect on IRI
due to different distress types, DF is the effect due to frost-heave potential, and DS
is the effect associated to subgrade swelling potential. In order to identify the
parameters to be included in the model as instruments for DD; DF; and DS, stepwise
OLS regression was performed in order to derive an initial expression for IRI (Von
Quintus et al. 2001a; Mirza and Zapata 2003).
The current model included in Pavement Design ME was structurally simplified
to account for site specific properties. The model as presented in the
Mechanistic-Empirical Pavement Design Guide Manual of Practice (AASHTO
2008) is as follows:

IRI ¼ IRI0 þ 0:0150
SF þ 0:400
FCTotal þ 0:0080
TC þ 40:0
RD ð5Þ

where SF is a site factor that accounts for environmental, subgrade soil properties,
and the age of the pavement structure. FCTotal corresponds to the area of fatigue
cracking: combined alligator, longitudinal, and reflection cracking under the wheel
path in feet2. TC is the length of transverse cracking (feet-mile), and RD is the
802 J.P. Aguiar-Moya et al.

average rut depth measured in inches. The structural form associated to the site
factor is as follows:

SF ¼ Age
½0:02003
ðPI þ 1Þ þ 0:007947
ðPrecip þ 1Þ þ 0:000636
ðFI þ 1Þ

ð6Þ

where PI is the plasticity index of the subgrade soil, Precip is average annual
rainfall in mm, and FI is the average annual freezing index. The previous model
parameters were also calibrated based on Long Term Pavement Performance data
for the overall United States. The model itself does not directly include local
calibration parameters, but the previous can be accounted for in the cracking and
rutting transfer equations that are included in the design guide.
However, as per Aguiar-Moya et al. (2011) the model can be improved
accounting for several properties of the IRI data. Some of the observations on the
model are the following: (i) no constant term included in the model. Even though
the initial IRI is accounted for, a constant term can determine the presence of
omitted variable bias. If the constant parameter is significant, then the variables that
are being used to predict IRI might not be adequate. On the other hand, if the
parameter is not significant, it can be adequately removed from the model.
The previous is of importance when considering that the initial IRI used to
model (1) is not measured but extrapolated from the IRI field data observed up to
15 years after construction. Furthermore, it is noted by the authors that the distress
data used for calibrating the model does not necessarily match the time when IRI
was recorded. In order to account for the previous, Aguiar-Moya et al. (2011)
proposed the use of panel data estimation methods that can be used to determine
section specific parameters that can readily account for site-specific properties that
are not captured by the exogenous factors. One of the proposed model was the
following:

IRIit ¼ b0 þ b1
SFit þ b2
FCTotalit þ b3
TCit þ b4
RDit þ ui þ eit ð7Þ

which consists of a random-effects model where each factor is a function of site “i”
at time “t”, b0 is an average pavement initial condition (representing initial IRI), and
ui is a random component of the intercept intended to capture omitted-variable bias
that is constant for each pavement section in time. However, further advantage from
the data cross-sectional and time series data can be obtained when considering serial
correlation: the distress at time “t” is related to the distress at time “t − 1”, and
should therefore not be considered as independent.
An alternative approach to modeling IRI has been presented by Lin et al. (2003)
and George et al. (1998) where IRI is predicted as a function of different distress
types based on the use of Artificial Neural Networks.
Development of IRI Models Based on APT Data 803

1.2 Objective

The objective of this study was to calibrate an IRI prediction model using measured
profile, rutting, and stiffness data collected from instrumented flexible pavements
under accelerated pavement testing conditions.

1.3 Test Sections

The initial set of experiments performed at PaveLab in Costa Rica correspond to 4

structures that were constructed by March 2012 (Fig. 1) (Aguiar-Moya et al. 2012).
HVS trafficking on the sections began in July 2013 using a dual 11R22-5 tire, with
90 psi inflation pressure, applying a standardized 40 KN load. The objective of this
set of experiments consisted in the structural comparison of typical conditions in the
Country: use of granular versus cement-treated bases (CTB), and thin versus thick
HMA layers. Table 1 summarizes the thickness and material properties of the
analyzed sections. Layer thicknesses were verified by means of Ground Penetrating
Radar (GPR) and cores. Initial layer moduli was determined by means of back-
calculation based on Falling Weight Deflectometer (FWD) results (Leiva-Villacorta
et al. 2015).
The top layer consists of an asphalt concrete (AC) mixture with nominal max-
imum aggregate size of 19.0 mm with an optimum binder content of 4.9 % by total
weight of mixture. The CTB was designed to withstand 35 kg/cm2 with an

Fig. 1 Test track distribution

804 J.P. Aguiar-Moya et al.

Table 1 Test tracks in-place properties

Properties\section AC1 AC2 AC3 AC4
AC thickness (H1), cm 6.1 6.3 13.2 13.2
Base thickness (H2), cm 21.9 21.2 31.0 24.9
Subbase thickness (H3), cm 30.1 30.1 30.1 30.1
Initial AC modulus (E1) @ 25 °C, MPa 3800 3800 3800 3800
Initial base modulus (E2), MPa 1200 170 170 1200
Initial subbase modulus (E3), MPa 140 140 140 140
Initial subgrade modulus (E4), MPa 70 70 70 70

optimum cement content of 1.7 % by volume of aggregate and with a maximum

density of 2013 kg/m3. The base material and granular subbase were placed at a
maximum density of 2217 kg/m3 with an optimum moisture content of 8.6 %. The
subbase material had a CBR of 95 %. Finally, the subgrade material (MH, A-7-5)
was constructed for a maximum density of 1056 kg/m3 with an optimum moisture
content of 52 % (typical moisture content in Costa Rica) and CBR of 6.6 %. Both
the subgrade and subbase material properties are uniform for all test sections.

1.4 Instrumentation

The measurements were performed using the HVS integrated instrumentation and
embedded sensors in all four test sections. HVS onboard instrumentation record the
applied load, tire pressure and temperature, position and velocity of the load car-
riage. Embedded sensors include asphalt strain gauges, pressure cells, multi depth
deflectometers (MDDs), and moisture and temperature probes. The HVS was fur-
ther equipped with a laser profiler that can be used to create a three-dimensional
profile of the section and a road surface deflectometer to obtain deflection basins at
any location along the test section.
Figure 2 shows the instrumentation array used for the first set of experiments. The
asphalt strain gauges were placed at the base/HMA layer interface in the longitudinal
and transverse directions. Pressure cells were placed at the subbase/subgrade
interface. MDDs were installed at four different depths to cover all four structural
layers. The thermocouples were placed at four depths: surface, AC layer mid-depth,
at the asphalt strain gauges depth and 5 cm into the base layer. In the case of AC1
and AC4 sections the same gauge array was used while excluding PAST sensors.
Data collection of the 3D profile, strain, pressure, temperature and deflection is
performed based on load repetitions. At the beginning of each test, data is obtained
at short intervals. After 20,000 load repetitions, data is collected on daily basis.
Inspection of fatigue and reflective cracking, friction loss, loss of aggregate-asphalt
bond and any other surface damage is performed on daily basis during the HVS
daily maintenance work.
Development of IRI Models Based on APT Data 805

Fig. 2 Sensor array

Fig. 3 Test section example profile matrix

The data collected by the two profilers consist of a matrix of information that
stores the distance between the sensors and the pavement surface in the longitudinal
direction at 4 in intervals and in the transverse direction at every 1 inch. Therefore,
each test section is subdivided into 50 subsection in the longitudinal direction and
64 in the transverse direction for a total of 3200 subsections as per Fig. 3.
The IRI was estimated by means of the quarter-car vehicle math model for each
of the longitudinal data lines; the transverse measurements are independent.
A limitation associated to the IRI estimation is that the length of the section is 6 m,
and within this distance the effective measuring distance of the lasers is 5.1 m.
Therefore, the IRI corresponds to an average property within the aforementioned
distance. The previous differs to IRI measured by means of profilers in the road
network since the IRI is reported for a longer section length (50, 100, 200 m). The
difference is important since a longer averaging length reduces the impact of out-
liers or IRI values that deviate significantly from the mean. However, as opposed to
network level measurements, the quality of the data or possible outliers could be
verified using the 64 subsections for each experiment.
In order to correlate IRI to material deterioration, moduli backcalculation was
also performed using multi-depth deflectometers (MDD) data in order to determine
the reduction in stiffness that the different material layers are subjected to. Figure 4
shows the increase in deflection and reduction in backcalculated moduli associated
806 J.P. Aguiar-Moya et al.

Fig. 4 MDD backcalculated layer moduli for AC3

Fig. 5 Average measured permanent deformation

to trafficking for a particular test section. Furthermore, permanent deformation of

each test section was also recorded (Fig. 5). It can be noted that the section with
least permanent deformation is AC4 (thick HMA layer and CTB). It was also noted
that increased thickness in the HMA layer (and no CTB) outperformed the use of a
CTB with a thinner HMA layer.

2 APT IRI Data

The average IRI data for the evaluated test sections are shown in Fig. 6. The data
indicate that in general, the initial IRI after construction for most sections is in the
order of 1 mm/m. The exception corresponds to section AC1 which presented
significant construction variability in the CTB and HMA layers, and resulted in
higher initial IRI (over 1.7 m/km).
Development of IRI Models Based on APT Data 807

2.5
2.3
2.1
1.9
IRI (mm/m)

1.7
1.5
1.3
1.1
0.9
0.7
0.5
0 2,000,000 4,000,000 6,000,000 8,000,000 10,000,000 12,000,000 14,000,000
ESALs
AC1 AC2 AC3 AC4

Fig. 6 IRI in different test sections

In general, the IRI data exhibits a linear growth trend. The exception is the AC2
section which corresponds to the weaker structure: thin HMA layer and granular
base. The section showed a decreasing rate of IRI growth rate. For the initial 1.5
Million ESAL applications, the section exhibited a higher IRI increase; however,
after 1.5 Million ESALs, the IRI increase rate becomes linear.
An additional observation of interest is that the IRI growth rate for sections AC1
and AC4 is similar. This serves as a verification that IRI is highly dependent on
roughness of the underlying layers, which for the aforementioned sections is the
same (CTB layer underneath HMA). A similar case occurs for sections AC2 and
AC3, which correspond to the granular base sections.

0-1 1-2 2-3

3
IRI (mm/m)

7.7
MESALs

0
2.9
0.025
0.15
0.275

1.1
0.4
0.525
0.65

0.2
0.775
0.9
1.025
1.15

0
1.275
1.4
1.525

Transverse Section

Fig. 7 Variability in IRI for AC3

808 J.P. Aguiar-Moya et al.

However, it is important to highlight that the IRI displays considerable vari-

ability within each test section (both longitudinal and transverse directions). The
degree of variability can be exemplified in Fig. 7. As expected, the area which is
subjected to higher trafficking due to wheel wandering initially exhibits lower IRI
values but when damage occurs, the IRI rate of increase grows significantly. From
the modeling perspective, the previous is important since it suggests that the model
will need to account for heteroscedasticity in the data: variability in IRI as a
function of trafficking or service life. As such, traditional estimation by means of
OLS is inefficient.

3 IRI Model Estimation

An initial IRI model was calibrated using multiple linear regression by generalized
least squares (GLS), to account for heteroscedasticity. The structural form of the
model is the following:

DIRIit ¼ b0 þ b1 Ti þ b2 SHMA;it þ b3 SB;it þ b4 SSB;it þ b5 SSG;it þ b6 Rit þ b7 Nit þ eit

ð8Þ

where DIRIit corresponds to the change in IRI for section “i” at time “t” in m/Km in
reference to initial IRI after construction, T represents the initial thickness for the
overall pavement structure in mm, S is the stiffness of each pavement layer in MPa
at time “t”, R is the total rutting in mm at time “t”, N is number of load cycles in
ESALs at time “t”, b1 thru b7 are calibration parameters to be estimated, and eit are
unobserved factors that are not captured by the model. The estimated parameters are
shown in Table 2. Note that the model accounts for the incremental damage in the
different layers of the pavement structure.
The previous model attempts to correct for heteroscedasticity in the IRI model
because of correlation between the regressors (independent variables) and the error
term; however, better predictions of IRI can be obtained when accounting for
heterogeneity and the type of data being used in modeling of IRI. In estimating the
previous model, all the time series data (performance information for a given
pavement section through time) and the cross-sectional data (different pavement
sections at any given time) have been combined or pooled together. However, the
data corresponds to a panel data set: cross-sectional and time series data (Prozzi and
Madanat 2003).
A panel dataset refers to a data type that combines a temporary dimension with
another cross-section dimension for each unit of study: pavement sections are
monitored through time. Therefore, the bias in the model can be reduced by
accounting for unobserved variables that differ from one section to the next, such as
material differences, construction variability, or other site-specific variables that do
not change over time (heterogeneity).
Table 2 Estimated model parameters
Factor GLS RE FE
Coefficient SE t-Stat Coefficient SE t-Stat Coefficient SE t-Stat
T 0.0025381 0.0013665 1.86 0.0026736 0.0036325 0.74 – – –
SHMA −0.0001096 0.0000127 −8.64 −0.0002714 0.0000201 −13.5 −0.0003556 0.0000257 −13.86
SB 0.000037 0.0000379 0.98 −0.0014032 0.0001047 −13.4 −0.0018767 0.0001283 −14.63
Development of IRI Models Based on APT Data

SSB 0.0011749 0.0002783 4.22 −0.0004584 0.0002523 −1.82 0.0004147 0.0003104 1.34
SSG 0.001502 0.0003211 4.68 0.0160145 0.0005682 28.18 0.0182281 0.000635 28.71
R 0.0421306 0.004373 9.63 0.0838147 0.0019081 43.93 0.0868718 0.0019247 45.14
N 2.06 × 10−08 1.90 × 10−09 10.88 2.79 × 10−08 1.18 × 10−09 23.73 2.91 × 10−08 1.18 × 10−09 24.68
ΔIRIt−1 – – – 0.0026698 0.0004975 5.37 0.0025659 0.0004975 5.16
Intercept −0.4222819 0.1346159 −3.14 −0.7306446 0.2377246 −3.07 −0.5723572 0.0295232 −19.39
809
810 J.P. Aguiar-Moya et al.

The purpose of modeling based on panel data is to capture the unobservable

heterogeneity among the variables and time. The application of this methodology
allows the analysis of two important aspects that are part of unobservable hetero-
geneity: (i) the specific individual effects, which are those that unevenly affect each
variable contained in the sample, that directly affects its behavior, and (ii) tempo-
rary effects, which are those that equally affect all individual samples but vary over
time (Mayorga and Muñoz 2000).
In the panel data models, the time factor was captured by the number of loading
cycles (ESAL applications) and the dependent and independent variables are the
same that were used to generate the initial GLS model (8). The model includes an
auto-recursive [AR (1)] component (at time “t”, considers the IRI at time “t − 1”).
The general form of the proposed model is as follows:

DIRIit ¼ ai þ b1 Ti þ b2 SHMA;it þ b3 SB;it þ b4 SSB;it þ b5 SSG;it þ b6 Rit þ b7 Nit þ b8 DIRIit1 þ ui þ eit

ð9Þ

where ui represents the unobservable effects that differ between samples but not in
time, eit refers to the purely random error, and ai and b1 thru b8 are the calibration
parameters. DIRIit1 corresponds to the change in IRI associated to the previous
cycle [AR (1) component].
The model was estimated by means of fixed effects (FE) and random effects
(RE) techniques. The fixed effects analysis was performed in order to estimate
section specific intercepts (recognizing that omitted variables may lead to changes
in the intercepts either over time or among cross-sectional units) and for random
effects case, the model intercept is considered as a random variable accounting for
differences in test sections associated to heterogeneity. The results are shown in
Table 2. Figure 8 shows the predicted values versus observed values for the esti-
mated models.

6
Predicted IRI (mm/m)

0
0 1 2 3 4 5 6 7
Measured IRI (mm/m)
GLS RE FE

Fig. 8 Predicted versus observed IRI values

Development of IRI Models Based on APT Data 811

Based on the GLS model predictions, when considering heteroscedasticity, the

change in IRI can be mostly accounted for by changes in the stiffness of the HMA
layer, subbase layer, and subgrade layer; permanent deformation of the pavement
structure, and the traffic loading (most significant factor). Therefore, as expected,
the change in IRI can be related to the damage induced to the pavement structure as
a result of traffic loading: on average 1 Million ESAL applications can be associated
to an IRI increase of 0.02 mm/m.
The impact of permanent deformation of the pavement structure is considerable:
a 1 mm increase in rut depth can result in an IRI increase of 0.04 mm/m. Therefore,
for a permanent deformation failure criteria of 12.5 mm, as a minimum an increase
in IRI of 0.53 mm/m is expected, considering no other changes in material or
structural properties has occurred. Of slightly less statistical significance is overall
pavement thickness. However, based on the GLS model, on average thicker
pavement structures can be associated to lower IRI values.
As per Fig. 8, using the panel data models that account for sections specific
properties (heterogeneity) improves fit and reduces bias significantly. Because
pavement thickness is a particular property of each pavement section, its effect is
absorbed by the ai þ ui sections specific terms. As with the GLS model, changes in
material stiffness have an important effect in IRI, however, in contrast to the GLS
model, changes in base stiffness have an important effect in predicting IRI: a
1000 MPa change in stiffness can be related to an IRI increase of 0.1 mm/m.
Rut depth and traffic loading also continue to be highly significant factors in
explaining changes in IRI. Of special interest is the AR (1): DIRIit1 which cor-
responds to the predicted/measured IRI in the previous time period. Because the
factor is significant, it can be concluded that there is correlation between current IRI
and IRI deterioration history: changes in IRI in the past have an effect on current

2.5

2
IRI (mm/m)

1.5

0.5

0
0 2,000,000 4,000,000 6,000,000 8,000,000 10,000,000 12,000,000
ESALs

IRI GLS RE FE

Fig. 9 Predicted IRI values for AC3

812 J.P. Aguiar-Moya et al.

IRI. An example of the improvement in model fit using panel data models is shown
in Fig. 9 (estimation example for section AC3, similar results obtained for
remaining sections). Note that the models including the AR (1) term properly
predict the initial reduction in IRI that can be attributed to post compaction of the
pavement section.

4 Summary and Conclusions

Models for predicting IRI have been developed based on changes in roughness
resulting from APT results on several pavement test tracks. The models have been
defined as a function of traffic loading, but are also based on structural distress
captured by changes in layer stiffness and permanent deformations of the pavement
structure.
Considering current weaknesses in IRI models included in AASHTO’s
Pavement Design ME, the proposed models improves on efficiency accounting for
data heteroscedasticity and attempt to reduce omitted-variable bias because of
heterogeneity that is present in the data because of unobserved section-specific
variables. The latter can be achieved modeling IRI based on panel data techniques,
as opposed to standard OLS methods.
The panel data methods not only improve the models from a theoretical
standpoint but also result in changes in the estimated parameters: effect of variables
differs between OLS/GLS estimation and panel data estimates. The previous results
from the improvement in model consistency (reduction in bias) and efficiency. In
general the estimated suggest that changes in stiffness of the different material
layers, permanent deformation of the pavement structure, and traffic loading are
fundamental in predicting changes in IRI. The previous is expected since pavement
distress (rutting and cracking) have been historically identified as having an
important effect on IRI. The panel data models with an auto recursive term also
suggest that there is significant correlation between current IRI and previous IRI
measurements and therefore should be accounted for: IRI history is fundamental in
determining future IRI.
Finally, because stiffness data is not always readily available, the authors are
currently researching the use of surface deflections as instruments for material
stiffness. Deflection data can be easily obtained from PMIS data and can therefore
be included in future models.

References

AASHO. (1962). The AASHO road test. Special Report 61E. Washington, DC: American
Association of State Highway Officials, Highway Research Board.
AASHTO. (2008). Mechanistic-empirical pavement design guide. A manual of practice.
Washington, DC: American Association of State Highway Officials, Highway Research Board.
Development of IRI Models Based on APT Data 813

Aguiar-Moya, J. P., Corrales-Azofeifa, J. P., Loria-Salazar, L. G., & Elizondo-Arrieta, F. (2012).

PaveLab and heavy vehicle simulator implementation at the National Laboratory of Materials
and Testing Models of the University of Costa Rica. Advances in pavement design through
full-scale accelerated pavement testing.
Aguiar-Moya, J. P., Prozzi, J., & Smit, A. D. F. (2011). Mechanistic-empirical IRI model
accounting for potential bias. Journal of Transportation Engineering, 137(5), 297–304.
Al-Omari, B., & Darter, M. I. (1992). Relationships between IRI and PSR, interim report.
Publication No. UILU-ENG-92-2013. Springfield, IL: Illinois Department of Transportation.
Arhin, S. A., Williams, L. N., Ribbiso, A., & Anderson, M. (2015). Predicting pavement condition
index using international roughness index in a dense urban area. Journal of Civil Engineering
Research, 5(1), 10–17.
Carey, W. N. Jr., & Irick, P. E. (1960). The pavement serviceability performance concept.
Highway Research Board Bulletin 250, Washington, DC.
Darter, M. I., & Barenberg, E. J. (1976). Zero-maintenance pavements: Results of field studies on
the performance requirements and capabilities of conventional pavement. Report
No. FHWA-RD-76-105. Washington, DC: Federal Highway Administration.
George, K. P., El-Rahim, A. M. A., & Shekharan, A. R. (1998). Updates of pavement performance
modeling. In: Third International Conference on Road and Airfield Pavement Technology
Proceedings (pp. 402–4010), Beijing, China.
Leiva-Villacorta, F., Aguiar-Moya, J. P., & Loría-Salazar, L. G. (2015). Accelerated pavement
testing first results at the LanammeUCR APT facility. In Transportation Research Board 94th
Annual Meeting Proceedings, Washington, DC.
Lin, J. D., Yau, J. T., & Hsiao, L. H. (2003) Correlation analysis between international roughness
index (IRI) and pavement distress by neural network. In 82th Annual Meeting of the
Transportation Research Board Proceedings, Washington, DC.
Mayorga, M., & Muñoz, E. (2000). The technique of panel data: A guide for use and
interpretation. Central Bank of Costa Rica, Economic Division, DIE-NT-05-2000, San José,
Costa Rica.
Mirza, M. W., & Zapata, C. (2003). Appendix OO-3: Addendum—Estimation of distress quantities
for smoothness models for HMA surface pavements. Guide for mechanistic-empirical design of
new and rehabilitated pavement structures. Washington, DC: National Cooperative Highway
Research Program, Transportation Research Board.
Moore, R. K., Clark, G. N., & Plumb, G. N. (1987). Present serviceability-roughness correlations
using rating panel data. Transportation Research Record, 1117, 152–158.
Paterson, W. D. (1986). International roughness index: Relationship to other measures of
roughness and riding quality. Transportation Research Record, No. 1084.
Prozzi, J. A., & Madanat, S. M. (2003). Incremental nonlinear model for predicting pavement
serviciability. Journal of Transportation Engineering, 129(6), 635–641.
Sayers, M. W., Gillespie, T. D., & Paterson, W. D. O. (1986). Guidelines for conducting and
calibrating road roughness measurements. World Bank Technical Paper 46. Washington, DC.
Von Quintus, H. L., Eltahan, A., & Yau, A. (2001a). Smoothness models for hot-mix
asphalt-surfaced pavements developed from long-term pavement performance program data
(pp 139–156). Transportation Research Record No. 1764.
Von Quintus, H. L., & Yau, A. (2001). Appendix OO-2: Revised smoothness prediction models for
flexible pavement. Guide for mechanistic empirical design of new and rehabilitated pavement
structures. Washington, DC: National Cooperative Highway Research Program,
Transportation Research Board.
Von Quintus, H. L., Yau, A., Witczak, M. W., Andrei, D., & Houston, W. N. (2001b). Appendix
OO-1: Background and preliminary smoothness prediction models for flexible pavements.
Guide for mechanistic empirical design of new and rehabilitated pavement structures.
Washington, DC: National Cooperative Highway Research Program, Transportation Research
Board.
Permanent Deformation of an Unbound
Granular Pavement Subjected
to Accelerated Multiple-Axle Group Loads

Michael A. Moffatt

Abstract Road network owners are being faced with the need to make predictions
of the long-term effect of heavy vehicle loading changes on their networks.
Exploration of future mass-distance and incremental pricing for heavy vehicles
requires an understanding of the effects of different axle group loads and types on
pavement performance. The current Australian approach to assess the relative
damaging effects of different axle groups on road pavements is by comparison of
the peak static pavement deflection response under the axle groups. This ignores the
contribution to pavement damage made by the axles in the group which do not
correspond with the peak response. The assumption that the deflection response is
the most appropriate indicator of pavement damage is also open to question and is
not consistent with the current mechanistic design procedures, in which strains
rather than deflections are used to calculate the performance of pavement materials.
As part of a larger research exercise, the Accelerated Loading Facility was used to
apply single, tandem and triaxle (tridem) axle group loads to a typical Australian,
full scale, unbound granular pavement and subgrade. Analysis of permanent
deformation data was hampered by significant moisture change during the testing
period. However it was possible to demonstrate that the current standard load value
used for triaxle groups is appropriate. This standard load value assumes that the
interaction between axles of a multiple-axle group do not affect the relative per-
manent deformation damage caused by the same number of ungrouped axles.

Keywords Pavement design Accelerated loading facility Multiple-axle group




Granular materials Axles Design traffic

M.A. Moffatt (&)

ARRB Group, 500 Burwood Highway, Vermont South, VIC 3133, Australia
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 815

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_53
816 M.A. Moffatt

1 Introduction

Road travel is the major mode of Australian domestic transport and, significantly,
over 65 % of freight is carried by road (Austroads 2005), making Australia the
heaviest user of road freight on a tonne-kilometre per capita basis in the world
(Austroads 2005). It follows that the growth of Australia’s economy is heavily
linked to the movement of freight by road. Growth has been accommodated partly
by new infrastructure, but mostly by increases in the productivity of freight vehicles
including increased legal axle loads. This growth in freight demand is not expected
to diminish in the future, with the Bureau of Infrastructure, Transport & Regional
Economics (2011) predicting a near doubling of the freight task between 2010 and
2030.
A ‘user pays’ mass-distance-location charging process is being explored as a
potential means of linking increased maintenance expenditure by infrastructure
owners to increased productivity of freight movement. This prompts the obvious
question; how much should each user pay? A comprehensive understanding of the
incremental pavement damage resulting from different axle configurations and load
levels is required in order to be able to fairly attribute pavement deterioration to
individual heavy vehicles or to classes of vehicles. Two fundamental questions
would have to be answered, in the positive, before such a comprehensive under-
standing could be claimed:
• Is current understanding of the pavement damage resulting from different axle
configurations and load levels sufficient to reasonably attribute incremental
damage to a specific axle group type and load?
• Is the basis for this understanding sufficiently robust to allow the development
of a transparent, equitable mass-distance-location charging scheme, in which
each user pays their appropriate share?
In the Australian context at least, it is fair to say that these two questions have
not yet been satisfactorily answered. To date, Australian considerations of pave-
ment damage caused by different axle groups and loads within Australian policy,
planning and asset management studies and procedures have been based upon the
assumptions made within the pavement design procedures currently used
(Austroads 2012). As such, it is reasonable that the pavement design context is a
suitable place to consider the above questions in the first instance.
A recent research study investigated improved methods for assessing the
pavement damage caused by different multiple-axle group loads, and developed a
framework that can be used to quantify this pavement damage for use in Australia’s
flexible pavement design processes (Moffatt 2015a, b). This paper presents a
component of the study, focussing upon the common sealed flexible pavement
composition used in Australia; an unbound granular base with a sprayed seal
surfacing.
Permanent Deformation of an Unbound Granular Pavement … 817

2 Current Practice

In the current Australian design procedures (Austroads 2012) the allowable loading
for unbound granular pavements with thin bituminous surfacings is expressed as the
number of Equivalent Standard Axle (ESA) repetitions. The reference Standard
Axle is a single axle with dual tyres loaded with 80 kN. In order to express the
spectrum of different axle group load levels and types expected in the design traffic,
the design procedures use Eq. 1 to determine the number of ESA generated by a
given vehicle. Central to this process is the concept of equal damage loads
(EDL) for each axle group type. Each axle group type loaded to its EDL is con-
sidered to cause the same pavement damage as the Standard Axle. Table 1 lists the
EDLs used in the current design procedure.

Xn  
Li 4
ESA ¼ ð1Þ
i¼1
LEDL

where ESA = Number of ESA repetitions generated by the passage of the vehicle
Li = Load carried by axle group i
LEDL = Group equal damage load (EDL) for axle group i
n = Number of axle groups within vehicle
As documented by Jameson (2013), it is generally considered that the EDL
values shown in Table 1 were determined by Scala (1970), and were based on two
field studies. The work was based on the presumption, considered to be reasonably
supported by limited data from the AASHO Road Test, that axle groups (type and
load) that caused equal maximum deflection of the pavement surface caused equal
pavement damage. Scala measured surface deflection responses caused by different
axle group types and load levels on a number of sprayed seal and asphalt-surfaced
pavements.
However if two assumptions are made, the same EDL values as those shown in
Table 1 for tandem and triaxle groups can be calculated using Eq. 2. The two
assumptions are:
• with regards to pavement damage, there is no interaction between axles of a
multiple-axle group—i.e. the damage caused by each axle within the group is
completely independent of the damage caused by adjacent axles.
• for a single axle, damage increases with increasing load to the fourth power (in
the most recent, and also most definitive, study of the damage exponent for

Table 1 Equal damage loads Axle group type Load (kN)

for dual tyre axles
Single axle 80
Tandem axle group 135
Triaxle group 181
818 M.A. Moffatt

Australian unbound crushed rock pavements, Yeo et al. (2006) confirmed that a
value of four was appropriate).

   14
LEDL nn 4 1
n ¼ 1 or; LEDL ¼ 80  ð2Þ
80 n

where n = Number of axles in multiple axle group

LEDL = EDL for axle group (kN)
The present author is inclined to believe that Eq. 2 had a strong influence on
Scala’s determination of loads that cause equal damage given the highly scattered
experimental results obtained by Scala (1970). This implies that the loads used in
the Australian design procedure implicitly assume that interaction between the
loads on each axle within a multiple-axle group does not occur, and that each axle
of a multiple-axle group can be considered to be equivalent to a single axle.

3 Accelerated Pavement Testing

3.1 Test Equipment

As part of a larger study (Moffatt 2015a), an experiment was conducted with a view
to assessing the validity of the EDL values used in the current Australian method
for designing unbound granular pavements with thin bituminous surfacings
(Austroads 2012). The design method’s implicit assumption, that the surface
deformation damage generated by each axle within a multiple-axle group was not
influenced by the presence of other axles within the group, was also tested. The
Accelerated Loading Facility (ALF) was used to assess the deformation of a
sprayed seal-surfaced unbound granular pavement and subgrade under different
multiple-axle loads. The ALF machine is shown in Fig. 1.
The ALF is a full-scale pavement test system enabling the assessment of road
pavement performance within a short time scale. The half-axle load is applied to
12 m lengths of test pavement through a load assembly trolley. Prior to this study,
ALF could only simulate a (half) single axle. The machine was modified to allow
the application of tandem and triaxle (half) axle groups for this study.
A key aspect of the modified machine is that the load trolley assembly uses
identical and interchangeable axle modules (Fig. 2) which can be set up for single,
tandem or triaxle group trafficking. Within each module, the tyre rims are mounted
on a stub axle that is cantilevered from a swing arm. A standard shock absorber and
air spring is used to dampen the movement of the swing arm. Each axle module is
individually sprung and damped, and the modules are attached to a pivoting
attachment plate (Fig. 3a), the geometry of which ensures that the overall
self-weight of the assembly is shared evenly across each axle. When operating as a
Permanent Deformation of an Unbound Granular Pavement … 819

Fig. 1 Accelerated loading

facility

Fig. 2 Individual axle

module used in multiple-axle
assembly for ALF

single axle assembly, a lighter, non-pivoting, attachment plate is used (Fig. 3b).
When operated in tandem or triaxle configurations the drive motor is located on one
module and the remaining modules are freewheeling.
All ALF applied load levels described below refer to the load applied to the half
axles—these need to be doubled to reflect the equivalent load on a full axle or axle
group. The term ‘experiment’ is used within this paper to denote the ALF loading of

(a) Triaxle assembly (b) Single axle assembly with fixed mounting plate

Fig. 3 Triaxle and single assembly configurations showing individual models and attachment
plates
820 M.A. Moffatt

a specific 12 m length of pavement. Experiments are named with a numbering

scheme based upon location within test area.

3.2 Test Pavements

During testing the ALF machine was housed inside a building 54 m in length and
18 m wide. Available resources only allowed the test program to examine the effect
of multiple-axle group loads on a single pavement structure. As such, it was
important that the structure selected be representative of typical unbound granular
pavements. Additionally, as permanent deformation of the pavement was to be the
performance measure used during the experiments, it was important that at least
moderate levels of deformation occurred under the expected ALF loading in the
limited time available in which to conduct the experiments.
The pavement structure adopted was a 300 mm crushed rock base overlying
400 mm of imported clay subgrade material. Underlying this structure was 75 mm
crushed rock drainage layer overlying the natural clay subgrade—the top 300 mm
of which had been lime-stabilised as part of an earlier experimental program.
A 14/7 mm double-double latex modified sprayed seal was used to surface the test
pavements. Based on the Australian design procedure (Austroads 2012) the chosen
pavement structure would reach a terminal condition—being an average level of
deformation of 20 mm—after between 1.2 × 105 and 106 ESAs of loading,
depending upon the level of subgrade support provided by the imported subgrade.

4 Loading Applied

Each test location was only trafficked by a given axle configuration and load level.
The load on each axle configuration was adjusted so as to ensure that the same load
per axle was applied across the axle groups. Unfortunately, the self-weight of some
of the assembly components meant that the lightest load that the single axle
assembly could apply was 40 kN. Maintaining this axle load level throughout the
experimental program would have meant that the tandem axle configuration would
have required a load of 80 kN (which was achievable). However, the triaxle con-
figuration would have required an unachievable 120 kN. The loadings shown in
Table 2 were used to overcome this limitation. The unavoidable step change in
loading level was undesirable, but did offer, at least, the potential to link com-
parisons between group axles loaded with 30 kN, to those loaded with 40 kN.
The ALF machine was designed to enable trafficking at different transverse
locations within a wheelpath width of up to 1.2 m, to simulate traffic wander within
Permanent Deformation of an Unbound Granular Pavement … 821

Table 2 Axle group load Axle group Total group load (kN) Load per axle (kN)
levels for ALF experimental
program Single 40 40
Tandem 80 40
Tandem 60 30
Triaxle 90 30

a lane. For this study the ALF was programmed to conduct randomised passes of
the load wheels distributed according to a normal distribution within a 1.0 m width.
As the test pavement was constructed in indoor conditions, the sprayed seal was
not subjected to normal traffic or environmental hardening that would normally take
place following field placement. Additionally, the unbound granular base layer
could have been expected to have experienced a bedding-in or densification phase
under normal in-service loading conditions. To take account of this, a
bedding-phase process was adopted at the commencement of loading to simulate
the effects of in-service conditions. Each experiment utilised the 60 kN tandem axle
configuration for the 10,000 cycle duration of the bedding-in process.

5 Acquired Data

Before and during each ALF experiment, data was collected using non-destructive
methods in order to capture the change in deformation of the pavement resulting
from the application of cycles. At the end of each experiment more destructive
methods of pavement data collection were employed to explore the physical con-
dition of the base and subgrade.
To ensure that individual axles within groups were loading the pavement evenly
strain gauges were fitted to each of the assemblies’ stub axles. At the conclusion of
an experiment the dynamic load data collected at the start and end of trafficking
were analysed in order to make a decision about which pavement chainages were
not subject to full, even loading. Data at these locations was marked as invalid and
excluded from any subsequent analysis.
Surface deflection measurements were taken at the conclusion of the bedding-in
phase using a falling weight deflectometer (FWD). This data was used to
back-calculate estimates of pavement material moduli. In the back-calculation
analysis the crushed rock base was divided into three 100 mm thick sublayers, and
the imported clay subgrade into three 125 mm sublayers. In addition to the
back-calculated moduli, some additional parameters were defined to provide
aggregated stiffness parameters. These included simple averages, e.g. the average of
the three crushed rock sublayer moduli, and thickness weighted averages using
Odemark’s method of equivalent thickness approach (Ullidtz 1998).
Transverse profile data was collected progressively throughout each experiment
so as to track the gradual deformation of the pavement surface as a result of
trafficking. Transverse profiles were collected at spacings of 0.5 m along the length
822 M.A. Moffatt

of each experiment. The data was processed to calculate the deformation that
occurred due to trafficking at each 0.5 m measurement location. Deformation was
defined as the difference between the maximum profile depth at the time of mea-
surement and the maximum depth at the time that a datum set of profile readings
were taken. The surface profile recorded after the 10,000 cycles of bedding-in, as
described in Sect. 4, was used as the datum. All deformation data presented in the
paper represents the deformation that had accumulated since the end of this
bedding-in period. The total deformation of the test pavements that had occurred
from the start of the bedding-in period was obviously greater than the data pre-
sented in this paper.
After trafficking of each experiment was completed, samples of base material
extracted from both trafficked and untrafficked areas were sieved and particle size
distributions (PSD) were determined. The data did not indicate any significant
variation of PSD between the trafficked and untrafficked areas, or with depth within
the base material, indicating that aggregate particles did not significantly break
down under ALF trafficking.
Density testing was carried out during pavement construction as well as at the
conclusion of each experiment using a nuclear density meter in direct transmission
mode. Post-trafficking readings were collected in the ‘trafficked’ area as well as
along two outer ‘untrafficked’ offsets. Testing was conducted at 1 m intervals along
the length of the experiment. The data consistently demonstrated higher densities in
trafficked areas than the corresponding untrafficked areas. Dynamic cone pen-
etrometer (DCP) tests were conducted in the same locations.
After the conclusion of most of the experiments, transverse trenches were cut
across the trafficked areas. To observe pavement deformation changes with depth
within the pavement structure, the distances between pavement layer boundaries
(between sprayed seal, base and clay subgrade) were measured relative to an
overlying straight edge across the full width of the trench.
It proved difficult to clearly identify the interface between the granular material
and the fine-grained subgrade material, leading to an uncertainty in the accuracy of
the measurements of at least a few millimetres. Given the relatively low surface
deformations that occurred, any subgrade deformation would have been within this
measurement uncertainly. Additionally, even though the imported clay material was
finished with smooth drum rolling, the presence of indentations made by pad foot
rollers was clearly evident in the trench profiles, adding increased uncertainty of
measurement. It was concluded that there was no deformation in the top of the
imported clay material.

6 Variation in Deformation Results

For a given number of loading cycles the average of the 0.5 m deformation readings
recorded was called the ‘overall deformation’. Figure 4 plots the progression of this
overall deformation for each single axle and triaxle experiment. In some of the
Permanent Deformation of an Unbound Granular Pavement … 823

10 3500 10 3501
3507 3504
Overall deformation [mm]

Overall deformation [mm]

8 3511 8

6 6

4 4

2 2

0 0
0 100 200 300 400 0 100 200 300 400
Kilocycles of loading Kilocycles of loading

(a) Single axle – 40 kN (b) Triaxle group – 90 kN

Fig. 4 Progression of (post-bedding-in) overall deformation for single axle and triaxle
experiments

experiments it can be seen that the magnitude of overall deformation did not
smoothly change with increasing application of loading cycles: this is particularly
evident in Fig. 4b. This fluctuation in overall deformation is simply explained. The
raw deformation data collected at each chainage was examined to ensure that
inconsistent readings were identified and removed. This means that deformation
data for a given chainage might not be available for all cycle counts. As the overall
deformation parameter is simply an average of valid deformation data for all
chainages, the inclusion of data at a specific chainage in one cycle count, and its
exclusion in the subsequent cycle, can lead to the fluctuations evident in Fig. 4.
This indicates the need to examine deformation progression data at a
chainage-by-chainage level and not as a simple aggregate average such as the
overall deformation parameter.
A more significant issue that is highlighted by Fig. 4 is the widely different
performance exhibited by the different experiment locations subject to the same
loading. An analysis of the data sets described in Sect. 5 concluded that signifi-
cantly high rainfall during the conduct of the experiments had resulted in unex-
pected changes in groundwater conditions over the test pavement area and over the
3-year period over which the experiments were conducted. Estimates of the
California Bearing Ratio of the imported clay subgrade were made using DCP data,
and were found to vary significantly. Some variation of in situ dry density was also
found.
It was determined that none of the examined parameters could solely explain
variations in performance between experiment locations, and that there was con-
siderable variation in material properties within many of the experimental test
sections. It was considered that the best analysis approach to adopt would be to
824 M.A. Moffatt

develop a model that predicted the deformation at each individual test location
within an experiment based upon the material properties measured (e.g. moisture
content, density, deflections) or estimated (e.g. CBR, back-calculated moduli) at
that point. Such a model would allow the direct comparison of surface deformations
for a given set of material properties.
It became apparent that it was not possible to develop a generalised model able
to encapsulate the combined dataset of all experiments. Whilst there was variation
between many of the factors discussed above, the amount that each overlapped was
not consistent across the different axle group/loads applied. However it was found
(Moffatt 2015a, b) that test pavements trafficked by the 40 kN single axle and the
pavements trafficked by the 90 kN triaxle group spanned similar ranges of in situ
material properties. As discussed in Sect. 7, the same form of linear regression
model was found to be able to relate the deformations measured to the stiffnesses of
the crushed rock bases for both load types. Hence a direct comparison of defor-
mation developed under the two load types could be compared for a given level of
crushed rock base stiffness.

7 Comparison of 40 kN Single Axle and 90 kN Triaxle

Group

Regression models were developed to predict the number of ALF loading cycles of
an axle group that were required to reach a defined level of surface deformation.
A defined level of deformation of 4 mm (above the deformation achieved during
the bedding-in phase) was used. This level was selected as it was a level that was
achieved by the majority of the data point locations for both axle group types.
Stepwise regressions were undertaken to estimate the number of loading cycles
required to achieve this deformation level as a function of a combination of indi-
vidual deflection responses (e.g. d0, d200, d900 etc.) and derived aggregate stiff-
ness parameters. The same model form was found to best fit the single axle (Eq. 3)
and triaxle (Eq. 4) data. Trials testing whether incorporating additional parameters
would improve the fit of the model determined that any additional parameters were
statistically insignificant.

N40;4mm ¼ 149407 þ 1187  Erock ð3Þ

where N40;4mm = Cycles of the 40 kN single axle to reach 4 mm of deformation

Erock = Overall stiffness of the crushed rock base (kN)
 1 1 1 3
t E3 þ t E3 þ t E3
= 1 1t1 þ2t2 2þ t3 3 3
E1 ; E2 ; E3 = Back-calculated moduli of crushed rock sublayers (MPa)
t1 ; t2 ; t3 = Thickness of crushed rock base sublayers
Adjusted R2 = 0.88
Permanent Deformation of an Unbound Granular Pavement … 825

N90;4mm ¼ 217722 þ 1607  Erock ð4Þ

where N90;4mm = Cycles of the 90 kN triaxle group to reach 4 mm of deformation

Erock = Overall stiffness of the crushed rock base (kN)
Adjusted R2 = 0.87
In Fig. 5 the number of load cycles needed to reach the 4 mm deformation level
is plotted as a function of Erock , using Eqs. 3 and 4. It is evident, that for low values
of Erock , very few cycles of either load type were required to reach the 4 mm
deformation level and there was little difference between the cycles required by
either load group to cause the same damage. As the modulus of the base increases,
however, the regression models slowly separate, with more cycles of the 90 kN
triaxle group required to cause the same damage as the 40 kN single axle.
Expressing the difference between the numbers of cycles required to reach the
same deformation level as a ratio results in the load equivalency factor (LEF) in
Eq. 5.

N40 kNsingle ¼ LEF  N90 kNtriaxle ð5Þ

where N40 kNsingle = Estimated number of cycles of the 40 kN single axle needed to
reach 4 mm deformation level
N90 kNtriaxle = Estimated number of cycles of the 90 kN triaxle group needed to
reach 4 mm deformation level
LEF = Load equivalency factor, i.e. the factor which relates the equivalent
number of load applications to achieve the same level of damage
For locations that needed a minimum of 50,000 ALF loading cycles to reach the
4 mm deformation level, the LEF factor varied from 0.8 (at high effective stiff-
nesses) to 1.0 (at low effective stiffnesses) across the range of experimental data.
However, considering the spread of data points plotted in Fig. 5 it is difficult to
conclude that there is a substantial difference between the data sets. The confidence
intervals around the slope of the regression relationships shown in Fig. 5 were
computed—(959, 1415) for the 40 kN single axle, and (1138, 2076) for 90 kN
triaxle group—and were found to considerably overlap. It is concluded that both
load groups caused the same amount of deformation.
Section 2 demonstrated that, by assuming that axles act independently of each
other, and that damage increases with the fourth power of increase in load, that the
equal damage load for a full-width triaxle group should be 182 kN. Accepting that
the load-damage exponent for the pavements tested had a value of 4, and that the
results demonstrated that the equal damage load for a full-width triaxle group was
approximately 182 kN, it can be concluded that the interaction between axles does
not significantly affect deformation damage, and that the axles can be considered to
each contribute to the overall damage in isolation to each other.
Based on the lack of observed subgrade deformation it seems reasonable to
assume that the observed surface deformation due to ALF loading reflected
deformation limited to the upper component of the crushed rock base layer, at
826 M.A. Moffatt

Fig. 5 Number of 40 kN
300
single axle and 90 kN triaxle
group cycles required to reach

Cycles of ALF loading [kilocycles]

4 mm deformation as a 250
function of the effective
stiffness of the crushed rock 200
(Erock )
150

100

40 kN single
50
90 kN triaxle
40 kN single model
0 90 kN triaxle model

125 150 175 200 225 250 275 300 325

Effective stiffness of crushed rock base, Erock [MPa]

depths where the interaction between axles would be expected to be low. This
determination is supported by the similar conclusions reported by Gramsammer
et al. (1999) based upon the examination of the deformation performance of an
asphalt pavement under full-scale loading with a French accelerated loading device.

8 Conclusions

This paper has documented the measurement of the performance of a full-scale

unbound granular pavement structure subjected to full-scale multiple-axle group
loads. The experiments were conducted in a highly controlled manner; limiting the
degree of measurement error, minimising the variability of dynamic load levels, and
ensuring minimal variations in the condition of the constructed test pavement
sections. However, uncontrolled and unexpected variations in moisture condition of
the unbound granular base and subgrade materials during the conduct of the
experiments prevented the research from reaching as many definitive conclusions as
was initially hoped. Nevertheless, the experiments did allow the analysis to draw a
significant conclusion—the 90 kN loaded triaxle did not cause significantly dif-
ferent damage to the test pavements than the 40 kN single axle assembly.
The experimental results cannot be reasonably considered to provide enough
evidence that the currently used equal damage load for triaxles of 181 kN (full axle)
is inappropriate. The small difference in LEF to current practice (0.8 cf. 1.0) is
insignificant in comparison to the variation in experimental results and the variety
of assumptions made in the analysis.
Accepting that no significant interaction between axles occurs, the EDL for a
multiple-axle group, that is the load on the group that will result in the same
deformation damage as a single axle with dual tyres and a load of 80 kN, can be
calculated using Eq. 2.
Permanent Deformation of an Unbound Granular Pavement … 827

Whilst the use of Eq. 2 would result in only minor changes to current Australian
pavement design practice, the significance of the research described in this paper is
not trivial. The research represents the first known time that accumulated defor-
mation performance data resulting from isolated multiple-axle group loads has been
obtained for unbound granular pavements with thin bituminous surfacings. The
basis of current Australian design practice and special vehicle approvals has been
based upon limited response-to-load data and assumed behaviour rather than
accumulated damage observations. Important policy decisions need to be founded
upon an estimation of the damage of granular pavements resulting from different
multiple-axle group loads. The findings of this study will significantly increase the
confidence that can be held by pavement designers, asset managers and policy
makers that current Australian practice is a reasonable reflection of the true beha-
viour of unbound granular pavements.

References

Austroads. (2005). RoadFacts: An overview of the Australian and New Zealand road systems.
Austroads, Sydney, New South Wales, Australia.
Austroads. (2012). Guide to pavement technology: Part 2: Pavement structural design.
AGPT-02-12, Austroads, Sydney, New South Wales, Australia.
Bureau of Infrastructure, Transport & Regional Economics. (2011). Truck productivity: Sources,
trends and future prospects. Report 123, BITRE, Canberra, Australia.
Gramsammer, J. C., Kerzrého, J. P., & Odeon, H. (1999). The LCPC’s APT facility: Evaluation of
fifteen years of experimentations. In International conference on accelerated pavement testing,
1st 1999, Reno, Nevada, USA.
Jameson, G. (2013). Technical basis of Austroads guide to pavement technology: Part 2:
Pavement structural design. Research report ARR 384, ARRB Group, Vermont South,
Victoria, Australia.
Moffatt, M. (2015a). The influence of multiple-axle group loads on flexible pavement design.
Research report AP-R486-15, Austroads, Sydney, New South Wales, Australia.
Moffatt, M. A. (2015b). Improved consideration of multiple-axle loads in the structural design of
flexible pavements. Ph.D. thesis, Department of Civil Engineering, Monash University,
Clayton, Victoria, Australia.
Scala, A. J. (1970). Comparison of the response of pavements to single and tandem axle loads. In
ARRB Conference, 5th, 1970, Canberra, Australian Capital Territory, Australian Road
Research Board, Vermont South, Victoria, Australia, pp. 231–252.
Ullidtz, P. (1998). Modelling flexible pavement responses and performance. Lyngby: Polyteknisk
Forlag.
Yeo, R., Martin, T., & Koh, S. L. (2006). Investigation of the load damage exponent of unbound
granular materials under accelerated loading. Technical report AP-T73-06, Austroads,
Sydney, New South Wales, Australia.
Three Years’ Monitoring of IRRF
Instrumented Test Road

Leila Hashemian and Alireza Bayat

Abstract The IRRF’s test road, nearly 500 m in length, is unique to Western
Canada. Situated in Edmonton’s east end, the road consists of two test sections and
a control section designed to investigate the use of Tire Derived Aggregate as
embankment material and the performance of bottom ash and polystyrene board as
insulation layers in cold climate conditions. During construction, the road was
instrumented with more than 250 pavement and geotechnical sensors. The installed
geotechnical instrumentation monitors compression behaviour, internal tempera-
ture, and drainage characteristics of the test sections, while the pavement instru-
mentation monitors the mechanistic responses of the pavement to traffic and
environmental conditions. Data collected from these sensors is relayed in 15-s
intervals to the IRRF’s laboratory facilities, where it is used to study pavement
performance. Various activities are scheduled to study how seasonal changes affect
the pavement. Falling weight deflectometer tests are conducted monthly on the test
road to monitor seasonal changes in the pavement layers’ moduli, while the effects
of temperature, moisture variations, and traffic loads on the different pavement
layers are also evaluated. This paper summarizes the effect of seasonal variation on
pavement responses as a result of 3 years’ monitoring of the control section, tire
embankment and insulated sections of the IRRF instrumented test road. By ana-
lyzing several Falling weight deflectometer testing results and temperature data
collected from different instrumented sections during the time of monitoring, this
paper investigates the impact of freeze and thaw on subgrade soil resilient modulus
and the benefits of using insulation layers to protect the subgrade soil.

Keywords Test road
Instrumentation
Insulation layers
Seasonal variation

Waste material

L. Hashemian
A. Bayat (&)

University of Alberta, 7-243 Donadeo Innovation Centre for Engineering,
9211 116th Street, Edmonton, AB T6G 1H9, Canada
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 829

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_54
830 L. Hashemian and A. Bayat

1 Introduction

In cold regions such as Canada, pavement structures are subjected to extreme low air
temperatures and freeze-thaw cycles for long periods of time, resulting in pavement
distress, deterioration, and lower service life (Hass et al. 2004). Despite continual
efforts, roads in the country’s cold regions are deteriorating faster than maintenance
operations can prevent the damage. Consequently, road agencies face aging systems
with lower serviceability and higher maintenance costs. Several Canadian studies
have shown that environmental factors contribute significantly to roadway deterio-
ration, causing up to 70 % of road damage relative to traffic load (Dore et al. 2005;
Sottile 2015; El-Hakim and Tighe 2014). The same studies have identified increased
freeze-thaw cycles and changes in precipitation patterns resulting from ongoing
climate change as the most critical accelerants in pavement deterioration and decline
of long-term performance. Although there is a considerable amount of research
regarding different aspects of road and pavement engineering in relation to these
issues, most studies are based on either laboratory or field experiments designed to
address problems in warm to moderate climates. As the performance of pavement
varies greatly according to climate conditions, the results of these past studies cannot
be applied successfully to Canada’s cold regions.
Recently, in line with the move to other sectors of the economy, there is an
increased tendency in the pavement community for more sustainable and envi-
ronmentally friendly practices and materials. As a result, road highway agencies in
cold regions are urged to use waste and recycled materials as an economical
insulation layer in an ongoing trial-and-error process. Albertans generate more than
5 million waste tires and recycle more than 15 kg of tires per person per year (more
than any other provincial tire program; Alberta Recycling 2015). Scrap tires can be
shredded to produce Tire-Derived Aggregate (TDA), a lightweight and
free-draining fill material. TDA produced from Passenger and Light-Truck Tire
(PLTT) has been successfully used as fill embankment in Canada in a number of
small-scale pilot projects, but a rigorous evaluation of TDA properties in relation to
their potential use as embankment fill in cold climates has not been undertaken
(Humphrey et al. 2000). In addition to PLTT sources, heavy industry produces large
volumes of Off-The-Road (OTR) tires that have not been tested as fill materials.
In 2014, Alberta generated 68 % of its electricity from coal, which generates a
huge amount of ash (CAPA 2015). Several researchers have evaluated and com-
pared the performance of different economical and recycled thermal resistant
materials as insulation layers. Bottom ash, one material that can be used as an
insulating layer, is a byproduct of coal combustion when burnt in the boiler furnace
of an electric power plant. Bottom ash is mainly composed of silica, alumina, and
iron, and it is utilized in highway construction primarily in cold-mixed asphalt,
embankments, and base courses. Few studies have investigated the ability of bottom
ash mixed with the subgrade and base materials to limit frost depth. A study
conducted in Helsinki, Finland, over the course of three winters revealed that frost
depth in bottom ash sections was 40–60 % of that in gravel sections. This study
Three Years’ Monitoring of IRRF Instrumented Test Road 831

showed that carefully compacted ash is not susceptible to frost due to its hardening
and low permeability (Havukainen 2012). Another study showed that mechani-
cally- and chemically-stabilized bottom ash can be considered a high-quality base
material for highway applications. Based on experimental results, it was concluded
that most bottom ashes met several performance criteria, such as physical appear-
ance, gradation, and soundness, which makes them suitable for pavement con-
struction (Huang 1990). The application of bottom ash on top of saturated silt could
effectively mitigate frost heave and transverse cracking in Western Canada. In a
study conducted by Nixon and Lewycky (2011) in Edmonton, Alberta, pavement
sections comprising a 500 mm gravel base course and a 1100 mm bottom ash
insulation layer showed a maximum frost depth of 1.5 m, with no frost reaching to
the underlying silt layer. The thermal performance of bottom ash as an insulation
layer was investigated in another study and the results found that this layer
decreases the frost depth to 50 % (Tavafzadeh et al. 2014).
Plastic foam boards have been widely used as heat insulation layers in several
countries to extenuate frost penetration into the pavement foundation. Previous
studies have shown that using a 5 cm Styrofoam insulation layer can reduce frost
heave. Elevation measurements conducted at Wolf Creek Pass in Colorado indi-
cated that Styrofoam sections experienced 14 cm less frost heave compared to the
previous winter when no insulation layer was used (Hayden and Swanson 1972).
Insulating a test road section in Alaska showed that 5 and 10 cm thick polystyrene
insulation layers could noticeably reduce the thaw depth from 61 to 20 and 10 cm,
respectively. Additionally, settlement data collected during the Alaskan study from
July 1971 through September 1972 presented up to 9 cm less settlement in the
insulated sections when compared to the normal sections (Esch 1972). Research
investigating the frost-resistance capacity of the polystyrene insulated sections
showed that the water content of the underlying layers, as well as the thickness and
thermal conductivity of the polystyrene, play a significant role in the
frost-resistance capacity of the insulation layer (Gandahl 1982). Longitudinal frost
heave measurements collected from a test road in north Sweden indicated less
heave in the insulated sections compared to the non-insulated sections. Moreover,
results from a recent case study performed in Edmonton, Alberta, showed the
impact of a 5 cm thick Styrofoam Highload 40 extruded polystyrene insulation
board in reducing frost penetration into the subgrade by 40 % compared to the
non-insulated section. Field measurements of the frost line’s advancement over time
agreed with geothermal modeling predictions (Tatarniuk and Lewycky 2011).
Edmonton is located in hard freeze-dry climatic conditions, where the average
Freezing Index (FI) based on 30 years of historic weather data is 1365 °C days
(Environment Canada 2014). Frost penetration leading to frost heave and spring
weakening is a yearly challenge faced by the local highway agency. As reviewed
above, pavement insulation layers have been used to prevent deep frost penetration
into the subgrade of a few projects in Alberta.
As a result, the Integrated Road Research Facility (IRRF) was established in
2012 as a unique collaboration between the University of Alberta, Alberta
Transportation, Alberta Recycling, and the City of Edmonton. The IRRF was
832 L. Hashemian and A. Bayat

formed with the goal of answering important questions about the impact of cold
climates on pavement performance, as well as to test the feasibility of utilizing
waste and recycled materials in road construction. Through the application of
innovative methods and technologies, the IRRF advances engineering construction
and design practices with alternative and maintainable road solutions.

2 Overview of the IRRF’s Test Road Facility

The IRRF’s test road facility is the new access to the Edmonton Waste Management
Center (EWMC), located on the eastern edge of Edmonton, about 15 km from
downtown. Construction of the IRRF’s test road started in May 2012 and was
completed with the first stage of paving in August 2012 and the second stage in
October 2013. The test road has been opened since mid-October 2015. It is
two-lanes, approximately 500 m long, and is currently carrying 400–500 trucks per
lane each day.
As illustrated schematically in Fig. 1, the test road includes three main test
sections, including: (1) TDA fill embankment (three 20 m sections, totally 60 m),
(2) a pavement monitoring section (two separate 20 m sections), and (3) insulated
sections (three 20 m sections, totally 60 m). All three main test sections were
instrumented with a variety of environmental, geotechnical and pavement sensors.
The TDA fill embankment test sections were instrumented during construction with
environmental and geotechnical instrumentation. All the sensors within the TDA
embankment fill were wired to a low-speed data logger, CR1000 from Campbell
Scientific Corp. (CSC), programmed to collect the data from all the sensors at
15-min intervals.

Fig. 1 Schematic layout of IRRF test road

Three Years’ Monitoring of IRRF Instrumented Test Road 833

The northbound of the two pavement monitoring test sections was instrumented
with high-frequency instrumentation to capture the effect of traffic loading on the
pavement structure. A high-speed CR9000 data logger from CSC was programmed
to be triggered by the pass of a truck and continue to collect the data until the truck
passes the northern pavement test section. Environmental sensors were also
installed in the pavement monitoring sections, which were connected to the
low-speed data loggers. The environmental sensors are used to monitor moisture
flow and frost penetration across the pavement system. A Weigh-In-Motion
(WIM) system, as well as a traffic video camera, is installed at the southern section
of the road to continuously monitor and characterize the traffic. The EWMC’s
weather station, located approximately 1 km from the test road, is used to monitor
climatic indices. The insulation layer sections were instrumented with environ-
mental sensors across the pavement structure’s depth. Another CR1000 data logger
was used to continuously collect data for the insulation test section.
Each of the three data loggers is equipped with a spread spectrum Model RF401
radio to communicate with the antenna on the trailer. The antenna is connected to a
desktop computer in the trailer, which is programmed to automatically collect the
data from each data logger every day. The data from the onsite computer is then
retrieved at the University of Alberta through remote desktop access.

2.1 IRRF Pavement Materials Characteristics

The pavement structure for the test road, demonstrated in Fig. 2, comprises
250 mm of Hot Mix Asphalt (HMA), which in 2012 was placed in two stages of a
160 mm binder layer. In 2013, it was topped with another 90 mm wearing course.
The HMA mixtures’ properties are provided in Table 1. The HMA layers were
placed on 450 mm of granular base course (GBC) on 1 m of compacted
Clayey-Sand (SC) subgrade soil. The subgrade soil possessed a maximum particle
size of 0.5 mm, liquid limit of 25 and Plastic Index (PI) of 9 %, respectively. Based
on the Unified Soil Classification System (USCS), the subgrade soil was classified
as Clayey Sand (SC). Meanwhile, the GBC layer consisted of crushed aggregates
with a maximum size of 19 mm. Based on USCS, the GBC was classified as
Well-Graded Gravel (GW). The gradation of the GBC and subgrade material can be
found in Table 2.

Fig. 2 Pavement layers of

the test road
834 L. Hashemian and A. Bayat

Table 1 Physical properties of HMA mixes used in the test road

Property Asphalt physical properties
Binder course Wearing course
Max. aggregate size (mm) 25 12.5
Binder grade PG 58-28 PG 58-28
Reclaimed asphalt pavement (RAP, %) 20 10
Binder content by weight of mix (%) 4.58 5.3
Void in mineral aggregate (VMA, %) 13.1 14.3
Void filled with asphalt (VFA, %) 69.4 74.9
Air voids (%) 4.0 3.6
Density (kg/m3) 2355 2344
Marshal stability (kN) 17.7 16.9
Flow (mm) 2.25 2.5
Theoretical film thickness (µm) 6.7 7.1
Tensile strength ratio (TSR, %) 98 81.6

2.2 Tire Fill Embankment Section

The farthest east section of the test road is made of PLTT, which is the primary
source of TDA. The second section is constructed with OTR tires, and the third is
made of TDA-soil mixture (using native soil). The length of each section is 20 m.
Mixing TDA with soil can reduce the potential for internal heating and decrease the
compressibility of TDA (Meles et al. 2014).
The TDA test sections were designed based on recommendations and guidelines
available in Humphrey et al. (2000) and ASTM D6270 (2008). To minimize the
possibility of internal heating, the embankment was designed to contain two layers
of TDA (top and bottom), each with an uncompressed thickness of 3 m, which was
wrapped in a nonwoven geotextile (Geotex 351 by Propex) that separates the TDA
from the surrounding soil. A minimum of 0.5 m thick compacted mineral soil with
a minimum of 30 % fine was used as a separator between the two TDA layers, and
1 m thick, low-permeability soil cover (including a minimum of 30 % fines by
weight) was used on the top and at the sides of the embankment above the natural
ground. Both TDA layers were designed with 3 % longitudinal slope at the base to
allow water to drain out of the TDA and into the drain pipes installed at the bottom
corner of each layer.
Visual examination of samples taken during TDA production showed that TDA
made from PLTT-produced particles was mostly thin and plate-like in shape, and
TDA made from OTR tire-produced particles was thick and mostly irregular in
shape. Figure 3 presents pictures of the PLTT and OTR used in the project. The
samples were tested to confirm the gradation satisfied Type B TDA particle size
requirements and other criteria advised by the American Standard for Testing and
Materials (ASTM D6270 2008). The TDA (PLTT) and soil were mixed at 50:50
ratio by volume (22:78 ratio by weight measured by taking samples of the TDA-soil
Three Years’ Monitoring of IRRF Instrumented Test Road 835

Table 2 Grain size distribution of GBC, bottom ash and subgrade soil used in IRRF project
Subgrade soil GBC Bottom ash
Grain size, D Percent Grain size, D Percent Grain size, D Percent
(mm) finer (%) (mm) finer (%) (mm) finer (%)
0.250 98.0 19.000 99.15 4.75 93.89
0.150 83.9 12.500 75.17 2.38 84.24
0.075 46.9 9.500 61.23 1.19 71.40
0.044 37.1 4.750 42.73 0.425 52.50
0.032 32.5 2.380 32.08 0.297 45.64
0.023 27.3 1.190 25.61 0.149 27.57
0.012 24.1 0.595 18.93 0.075 11.23
0.009 22.4 0.297 7.09 – –
0.004 19.9 0.149 1.46 – –
0.001 15.0 0.075 0.51 – –

Fig. 3 Left Passenger and

Light-Truck Tires (PLTT);
right Off-The-Road (OTR)

mixture). The soil excavated from the site during the construction of the test
embankment was used. The volume ratio was selected based on past experience
with the TDA-soil mixture test for embankment by Yoon et al. (2006) and for ease
of mixing during construction. PLTT was selected over OTR for the TDA-soil
mixture because of its ample availability. Figure 4 presents typical particle size
distribution for both the PLTT and OTR, as well as soil.
The test embankment was instrumented with two types of geotechnical sensors
to monitor and evaluate the construction process, as well as the immediate and
long-term performance of the embankment. A total of 25 Vibrating Wire Liquid
Settlement Systems (model SSVW105), calibrated and supplied by RST
Instruments, were used to monitor the potential settlement of the embankment. Of
the 25 settlement plates, seven were installed in each of the PLTT and OTR
sections, six were placed in the TDA-soil mixture section, two were installed in the
control section, and the remaining three were installed on stable ground outside the
embankment section. These three settlement plates were used as references to
correct the measurements from the others for daily changes in the atmospheric
pressure. The settlement plates were installed at similar locations in the PLTT,
OTR, and mixture test sections as presented schematically in Fig. 5. As seen in this
figure, four settlement plates (SP 1–SP 4) were placed on the bottom TDA layer to
836 L. Hashemian and A. Bayat

Fig. 4 Cross-section of the

TDA fill embankment test
sections (units in m)

Fig. 5 Grain size distribution

of soil, PLTT and OTR

monitor the layer’s settlement, and another three (SP 5–SP 7) were installed on the
top TDA layer. A total of 18 thermistors, model 109AM-L supplied by Campbell
Scientific Corp. (Canada), were used to monitor the temperature changes at various
depths of the embankment.

2.3 Pavement Monitoring Test Section

This test section was instrumented with asphalt strain gage (ASG) (Model
CEA-06-125UT-350 from the CTL Group) and earth pressure cells (EPC) (Model
LPTPC12-S from rst instruments) at different depths in the unbound layers.
A high-speed CR9000X data logger from Campbell Scientific Corp. (Canada) was
used to collect the dynamic responses of both sections at 500 Hz under current
Falling weight deflectometer (FWD) tests, as well as future traffic. Both sections are
similarly instrumented at the bottom of the HMA layer with six ASGs laid in the
longitudinal direction (ASG-L), six ASGs laid in the transverse direction (ASG-T)
and six vertical ASGs (ASG-V). Figure 6a, b show the instrumentation layout,
which is replicated in the two sections at the IRRF’s test road. One array of ASG
Three Years’ Monitoring of IRRF Instrumented Test Road 837

Fig. 6 a Schematic of instrumentation. b Schematic cross-section and the EPC locations.

c Installation of sensors at the bottom of the HMA (all dimensions are in mm)

was laid along the Outer Wheel Path (OWP). To ensure repeatability of the mea-
surements and to provide redundancy, the arrangement of the ASGs along the OWP
was replicated in two additional lines, 600 mm to the right and 600 mm to the left
of the OWP. EPCs were installed at two locations: one on the OWP and one on the
inner wheelpath (IWP). As seen in Fig. 6c, each location includes three EPCs
installed at three different depths. EPC 1 and 2 were installed on the top of the
GBC, EPC 3 and 4 were installed at the top of the subgrade, and EPC 5 and 6 were
installed 1000 mm from the top of the subgrade layer to monitor the distribution of
load in the pavement under layers.

2.4 Insulated Test Section

The test road includes two sections consisting of different insulation layers placed
immediately underneath the GBC. Figure 7 shows the location of the bottom ash
layer and the polystyrene boards. The bottom ash layer was free of large lumps and
impurities. Table 1 shows the grain size distribution of the bottom ash, which had a
maximum particle size of about 5 mm. The amount of non-combusted coal particles
was less than 5 % of the material weight. To maintain the natural moisture content
of bottom ash as low as 35 %, the compaction of this layer and the placement of the
GBC were completed on the same day (July 31, 2012). The bottom ash layer was
wrapped in geotextile to avoid mixing with natural soil. Closed-cell Styrofoam
Highload 100 extruded polystyrene boards, provided by Dow Chemical Company,
838 L. Hashemian and A. Bayat

Fig. 7 Cross-sections and as-built depth of thermistors and moisture probes

were used in the next section. Based on the manufacturer’s data sheet, the poly-
styrene boards had a compressive strength of 690 kPa and a minimum flexural
strength of 585 kPa. The thickness of the boards was 5 cm; therefore, Poly-10
comprised two layers of polystyrene board while Poly-5 was constructed using a
single layer. Figure 7 shows the polystyrene boards’ location on the subgrade soil.
To monitor frost penetration and compare the effectiveness of bottom ash in
relation to the polystyrene and control sections, the sections were instrumented with
109AM-L thermistors and CS650 Time-domain reflectometers (TDR) from
Campbell Scientific Corp. (Canada), as shown in Fig. 7. All of the sensors were
wired to a CR1000 data logger from Campbell Scientific Corp. (Canada). The data
logger was programmed to collect the data from all of the sensors at 15-min
intervals and equipped with a spread spectrum Model RF401 radio used to com-
municate with an antenna Model L14221 installed on an onsite trailer. Through
remote desktop access, the data is retrieved at the University of Alberta from the
onsite computer.

3 Field Monitoring Results

3.1 Tire Sections

The result of three years’ monitoring of temperature variations inside the tire sec-
tions for top and bottom embankment layers are reflected in Figs. 8 and 9
Three Years’ Monitoring of IRRF Instrumented Test Road 839

Fig. 8 Temperature
variations on the top tire
embankment layers

Fig. 9 Temperature
variations on the bottom tire
embankment layers

respectively. As it can be seen in Fig. 4, selected thermistors are located in the

middle of layers. As Fig. 8 shows, the temperature changes in all top tire
embankment layers follow the air temperature variation trend, with a time lag. The
PLTT-soil mixed layer has the lower temperature compared to the tire layers, while
the difference becomes less pronounced after October 2013, when the second 9 cm
asphalt layers were paved. As expected, Fig. 9 shows that temperate variations in
bottom tire embankment layers are less. However, the temperature of PLTT-soil
mixture is still slightly lower than other tire filled embankments in this layer too.
Figure 10 shows the results of several Falling Weight Deflectometer (FWD) tests
on tire sections in different seasons. FWD is a non-destructive test that simulates the
dynamic load resulting from a moving vehicle on the pavement. It has become a
common method to evaluate the structural capacity of the pavement and
back-calculating the modulus of different layers of the pavement. During the mon-
itoring period, different FWD testing was performed on all sections of the IRRF’s
test road. Three stress levels of 26.7, 40.0, and 53.3 kN were applied at each station.
However, in this study, only the results of deflections under stress levels of 40 kN
are used for investigation purposes. Figure 10 shows the maximum deflection of tire
section under FWD testing at different times of the monitoring years. As evident in
this figure, all three tire sections have less or comparable deflections with the control
section. The deflection values in all sections are very dependent on temperature
changes, which could be related to HMA thick layer stiffness and tire embankment
temperature. It is clear that in all sections, maximum temperature occurs in July 2015
840 L. Hashemian and A. Bayat

Fig. 10 FWD maximum deflections in tire sections

when the asphalt temperature is around 48 °C, and minimum deflection occurs in
December 2015 when the asphalt temperature is around zero.

3.2 Insulated Sections

The effectiveness of the bottom ash and polystyrene layers in protecting the sub-
grade from freezing and thawing was compared to an adjacent, non-insulated
Control Section (CS), as shown in Fig. 11. Temperatures were recorded at the depth
of 1.7 m below the surface layer from October 2012 to June 2015. As can be

Fig. 11 Temperature variation of subgrade layer at the depth of 1.7 m below the surface
Three Years’ Monitoring of IRRF Instrumented Test Road 841

Fig. 12 FWD maximum deflections in insulated sections

concluded from this figure, the subgrade in the control section is frozen in the
winter time, while the subgrade under both insulation layers has been completely
protected from freezing in the monitoring years. It is also clear that temperature
variations under Poly-10 are less than those for the bottom ash layer.
To investigate the effectiveness of insulation layers on the resilient moduli of
different sections’ subgrade, FWD testing was conducted in thaw season (April and
May 2015) and again in September 2015, when the subgrade layer was expected to
be fully recovered from the extra moisture resulting from freeze-thaw. Figure 12
shows the maximum deflection under FWD loading. According to this figure, for all
sections, pavement temperature affects the maximum deflection. The maximum
deflection is consistently highest with FWD testing on May 25, when the asphalt
temperature (2 cm below the surface) is 31 °C. It is also concluded from Fig. 12
that the control section and bottom ash section always have lower deflections than
the polystyrene sections. The maximum deflection always appears in the poly-
styrene section with 10 cm board thickness, followed by the polystyrene section
with 5 cm board thickness. This difference is more pronounce in warmer months as
it can be seen in month of May with asphalt temperature of 31 °C, the deflection for
polystyrene sections are 37 and 34 % greater than control section for sections with
10 and 5 cm boards respectively. It can have a significant reduction in pavement
structural pavement capacity which should be considered in design procedure of
insulated pavements.
The back-calculation results of these FWD tests are presented and compared
with the month of September in Fig. 12. The results indicate that, during thaw
season, the subgrade resilient modulus of CS is considerably lower than September
(about 12 %), while there is no significant change in subgrade modulus in insulated
sections during the same time. Mr values in thaw season compared with September
show that the Mr variations during thaw season are negligible for polystyrene
sections, and there is a maximum 6 % decrease in the bottom ash section and 12 %
in the control section (Table 3).
842 L. Hashemian and A. Bayat

Table 3 FWD back-calculated subgrade modulus

5-Sep-14 21-Apr-15 5-May-15 25-May-15
CS 143.5 130.9 127.6 126.3
B.Ash 134.2 136.7 134.2 126.7
Poly-10 150.7 151.5 158.5 153.1
Poly-5 139.4 143.5 150.2 137.8
Comparison of subgrade modulus in thaw season with September
CS 1.00 0.91 0.89 0.88
B.Ash 1.00 1.02 1.00 0.94
Poly-10 1.00 1.01 1.05 1.02
Poly-5 1.00 1.03 1.08 0.99

4 Conclusions

Based on the investigation and analysis of the instrumentation data in the IRRF test
road, the following observations and conclusions were drawn:
• The construction of the test embankment was completed with conventional
construction equipment and without any major problems.
• No evidence of internal heating was detected by the thermistors during the
3 years’ monitoring time for all three TDA sections.
• The FWD test results showed that the tire embankment sections have less or
comparable deflection to the control section under the applied load.
• Three years’ temperature monitoring in the control section, as well as the
insulated sections, revealed that the subgrade layer under insulation layers is
fully protected from freeze/thaw effects.
• FWD testing results in insulated sections showed that the maximum deflection
in the control section and bottom ash section are comparable, while the poly-
styrene sections had greater deflections. The insulated section with 10 cm
polystyrene board had the greatest deflection in all seasons. However, the dif-
ference in warmer months is more pronounce and can increase to 37 % greater
than the control section that can negatively affect the bearing capacity of the
pavement.
• Subgrade FWD back-calculated modulus in the control section and insulated
sections showed that there was no significant decrease in the subgrade modulus
during thaw season in insulated sections, while there was a 12 % decrease in the
subgrade modulus in the control section.
• Monitored temperature data, as well as subgrade back-calculated modulus,
showed that Poly-10 had the best protection of subgrade soil between insulation
layers, followed by Poly-5 and bottom ash layers.
Three Years’ Monitoring of IRRF Instrumented Test Road 843

Acknowledgments The authors greatly appreciate Alberta Transportation, the City of Edmonton,
and Alberta Recycling for their financial and in-kind support of this project. The authors also
acknowledge Alberta Transportation for conducting the FWD tests for this study and ISL
Engineering, Land Services, and DeFord Contracting for coordinating the construction and
instrumentation activities of the IRRF’s test road. The authors would also like to thank Ms. Sheena
Moore for lending her editorial assistance in the review of this paper.

References

Alberta Recycling Management. (2015). Authority, http://www.albertarecycling.ca/tire-recycling-

program/eligible-tires.
ASTM D6270–08. (2008). Standard practice for use of scrap tires in civil engineering
applications. West Conshohocken, PA: ASTM International.
Canadian Association of Physicians for the Environments (CAPA). (2015). Backgrounder:
Phasing out Alberta’s coal plants. Toronto.
Dore, G., Drouin, P., Pierre, P., & Desrochers, P. (2005). Estimation of the relationship of road
deterioration to traffic and weather in Canada. Ottawa: Transport Canada.
El-Hakim, M., & Tighe, S. (2014). Impact of freeze-thaw cycles on mechanical properties of
asphalt mixes. Transportation Research Record: Journal of the Transportation Research
Board, 2444, 39–47.
Environment Canada. (2014). http://weather.gc.ca/canada_e.html. Accessed 1 Jan 2014.
Esch, D. C. (1972). Control of permafrost degradation beneath a roadway by subgrade insulation.
Juneau, Alaska: Alaska Department of Transportation and Public Facilities.
Hass, R., Falls, L., Macleod, D., & Tighe, S. (2004). Climate impacts and adaptations on roads in
Northern Canada. In Cold Climate Conference, Regina.
Havukainen, J. (2012). The utilization of coal ash in earth works. Advances in Mining Science and
Technology, 2(2012), 245–252.
Hayden, R. L., & Swanson, H. N. (1972). Styrofoam highway insulation on colorado mountain
passes. Colorado Division of Highways, Report Number CDOH-P&R-R&SS-72-6, Denver,
CO, USA.
Huang, W. (1990). The use of bottom ash in highway embankment and pavement construction. Ph.
D. Thesis, Purdue University.
Humphrey, D. N., Whetten, N., Weaver, J., & Recker, K. (2000). Tire shreds as lightweight fill for
construction on weak marine clay. In Proceedings of International Symposium on Coastal
Geotechnical Engineering in Practice (pp. 611–616), Balkema, Rotterdam, The Netherlands.
Meles, D., Bayat, A., Shafiee, M., Nassiri, S., & Gul, M. (2014). Investigation of tire derived
aggregate as a fill material for highway embankment. International Journal of Geotechnical
Engineering, 8(2), 182–190.
Nixon, D., & Lewycky, E. D. (2011). Edmonton experience with bottom ash and other insulating
materials for mitigation of frost heave induced damage. Presented at the Annual Conference of
the Transportation Association of Canada, Edmonton, Alberta, September 11–14.
Sottile, M. F. (2015). Effect of climate changes expected during winter on pavement performance.
Cold Regions Engineering, 617–628.
Tatarniuk, D., & Lewycky, C. (2011). A case study on the performance of high load polystyrene as
roadway insulation in Edmonton, Alberta. Presented at the Conference and Exhibition of the
Transportation Association of Canada, September 11, Edmonton, Alberta.
Tavafzadeh, N., Nassiri, S., Shafiee, M. H., & Bayat, A. (2014). Using field data to evaluate
bottom ash as pavement insulation layer. Transportation Research Record: Journal of the
Transportation Research Board of the National Academies, 2433, 39–47.
Yoon, S., Prezzi, M., Sidiki, N. Z., & Kim, B. (2006). Construction of test embankment using a
sand-tire shred mixtures as fill material. Journal of Waste Management, 26, 1033–1044.
 Part XI
Accelerated Pavement Testing to Evaluate
Sustainable Materials
Comparative Evaluation of Performance
of Warm Mix RAP Asphalt Under
Accelerated Unidirectional Wheelload
Trafficking

Yucheng Huang, Fred Hugo, W. JvdM Steyn, Haocheng Xiong

and Linbing Wang

Abstract The paper focuses on testing comparative evaluation of fatigue perfor-

mance of composite asphalt pavement layers constructed at the Turner-Fairbank
Highway Research Center (T-F) under full-scale trafficking with the ALF and the
one third scaled MMLS3 (also known as MLS 11, since January 2015) of Virginia
Tech in Blacksburg. The goal was to enhance economics of APT testing. Virgin
100 mm asphalt slabs of the composite 50 mm dual layers were extracted from the
in situ un-trafficked sections of the test lanes after unidirectional testing with the
ALF using a formalized process. The process was developed to overcome con-
straints due to traffic and temperature control as well as minimizing logistics during
field testing. After extraction, the slabs were transported by vehicle to Blacksburg,
VA where the MMLS3 testing procedure was formalized by means of a proof test
on the lower 50 mm of a slab from Lane 9. For this, a 25 mm neoprene slab was
used as interlayer between the asphalt and the underlying in situ concrete floor.
After successful completion of the proof test, four further tests were completed.
Two were done on the extractions from Lane 9B and 9A, and the other two were
done on the extraction from Lane 5. The latter three tests all had two 25 mm
neoprene interlayers. Aspects that were considered in the comparative evaluations
were rutting, strain response, seismic stiffness using the Portable Seismic Pavement
Analyzer (PSPA), cracking and contact stress under the MMLS tire. The effects of
the respective material characteristics were also discussed. From the findings it was
concluded that the monitored parameters found from the scaled tests were com-
parable to the related full-scale ALF test results in terms of intrinsic material
characteristics and pavement performance. This is similar to earlier reported com-

Y. Huang (&)
H. Xiong
L. Wang

Virginia Tech, Blacksburg, VA, USA
e-mail: [email protected]
F. Hugo
Stellenbosch University, Stellenbosch, South Africa
e-mail: [email protected]
W. JvdM Steyn
University of Pretoria, Pretoria, South Africa
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 847

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_55
848 Y. Huang et al.

parative studies. The authors consider this finding to be an economic benefit that
should be utilized in pavement engineering studies.

Keywords Collaborative research ALF MMLS3 RAP Warm mix Fatigue




performance APT economics

1 Introduction

Accelerated Pavement Testing (APT) provides an effective method to evaluate

various types of materials characteristics and pavement performance. APT helps
engineers and researchers to obtain the structural response and pavement perfor-
mance under a controlled and accelerated damage conditions in a compressed
limited time.
The paper focuses on comparative testing evaluation for fatigue performance of
composite asphalt pavement layers constructed at the Turner-Fairbank Highway
Research Center (T-F) under full-scale trafficking with the ALF and the one third
scaled MMLS3 (also known as MLS 11, since January 2015) of Virginia Tech in
Blacksburg. Eleven lanes of composite pavement comprising different content
percentage of Recycled Binder Ratio (RBR) Recycled Asphalt Pavement
(RAP) Warm Mix Asphalt (WMA), with two 50 mm asphalt layers on an unbound
base, were constructed as ALF test lanes. Each lane was constructed 11 m in length.
During construction a system was developed to enable slabs to be extracted from
the pavement layers for subsequent MMLS3 trafficking in a laboratory. The process
was presented at the ISAP international seminar in Sun City South Africa (Hugo
et al. 2015). It enables preselected sampling points to be established where slab
specimens can be extracted with or without a separation underlying platform to
enhance extraction. The former is placed in position prior to construction. The draft
extraction protocol was used and evaluated at T-F in conjunction with its extensive
ALF, APT test program. Three slabs were successfully extracted and transported to
Blacksburg for MMLS3 trafficking.
In order to achieve comparable stress conditions, the 100 mm slabs were sliced
into thicknesses of nominally 50 mm, as top lift and bottom lift for MMLS3 traf-
ficking. Each slab could then be placed on neoprene layers lying on a concrete slab
as test pavements. The neoprene is a substitute for the aggregate base with similar
modulus, which facilitates the pavement construction because it is easy to place or
replace the layer after the test. The smoothness of the slab was selected to be
±2 mm/2 m. A layer of sand was used to achieve this where necessary.
The MMLS3 testing procedure was formalized by means of a proof test on the
lower 50 mm of a slab from Lane 9. For this, a 25 mm neoprene slab was used as
interlayer between the asphalt and the underlying in situ concrete floor. Detailed
information regarding the collaborative research and the construction and imple-
mentation of field test lanes has been reported by Hugo et al. (2015) and Gibson
et al. (2016).
Comparative Evaluation of Performance of Warm Mix RAP Asphalt … 849

Table 1 Experimental design

ALF lane % RBR Virgin binder PG Production temperature (°C) WMA process
RAP
5 40 64-22 149–160 –
9 20 64-22 116–132 Water

After successful completion of the proof test, four further tests were completed.
Two were done on the extractions from Lane 9B and 9A, and the other two were
done on the extraction from Lane 5. The latter three tests all had two 25 mm
neoprene interlayers.
The four tested slabs include two slabs from Lane 5 and two slabs from Lane 9.
Lane 5 and Lane 9 respectively contained 40 and 20 % RBR RAP Warm Mix with
foamed PG64-22, respectively (see Table 1). 20 and 40 % RBR was achieved with
23 and 44 % RAP by weight. The four slabs were trafficked with MMLS3 under
comparable environmental conditions (at 20 °C) at Virginia Tech laboratory in
Blacksburg, Virginia.
The objective of this study was to compare the performance test results including
seismic modulus, strain under dynamic loads, rutting depth between full scaled
ALF and one-third scaled MMLS3 for the same HMA slabs. The object was to
validate the application of the reduced scale MMLS3 as effective experimental
equipment to test material under small scale prototype model with test conditions
comparable to the full scale field conditions. In comparison to ALF testing,
MMLS3 is more cost friendly and affordable. It has been shown to provide com-
parable test results for evaluation of pavement rutting and fatigue performance (Lee
and Kim 2004; Bhattacharjee et al. 2011; Hugo et al. 2015). It is therefore a useful
method for doing screen testing for establishing full-scale testing protocols or
scenarios as and when desired or required.

2 Test Protocol

This paper presents details of the test on Lane 9B bottom lift, Lane 9A top lift, Lane
5 top lift and Lane 5 bottom lift under repeated wheel loads using MMLS3, as a
continuation of the trafficking study of Lane 9B top lift that had been reported
earlier (Hugo et al. 2015). According to lessons from Lane 9B top lift test, the test
slabs were bonded on the neoprene top surface and the neoprene bottom side was
bonded on the concrete floor with emulsified asphalt to eliminate the small vertical
oscillations during MMLS3 trafficking. However, Lane 9B bottom lift was placed
on one layer of neoprene and subjected to 1.2 million load applications until the
seismic modulus dropped to half of the initial values, which indicated the fatigue
failure of the slab. In order to expedite the damage process, Lane 9A top lift, Lane 5
top lift and Lane 5 bottom lift were placed on top of two layers of neoprene under
MMLS3 trafficking. In this manner, the underside of the latter slabs would be
850 Y. Huang et al.

10 A. Thermocouple
E 11 B. Strain data logger
D 12 C. Profilometer beam
13 D. MMLS3 Tire
14 E. Opened hood
15
16
17
18
C 19
A 20
B 21
A

Fig. 1 Laboratory layout for MMLS3 test implementation

undergoing more tensile stress/strain than the condition with one underlying neo-
prene layer.
The test implementation and instrumentation are presented in Fig. 1. During the
trafficking of all four slabs, the load frequency was set at 7200 loads per hour from
the beginning to cumulatively up to 300,000 cycles. After that in order to increase
the damage and cracking manifestation on the slab, the traffic load frequency was
reduced to 3600 cycles per hour.
The slabs’ performances were compared in terms of rutting, strain response,
seismic stiffness using the Portable Seismic Pavement Analyzer (PSPA), and
cracking. Test parameters were presented as follows:
• Bogie load 2.7 kN.
• Tyre pressure 700 kPa.
• Profiling with profilometer at spacing of 2 mm.
• PSPA seismic stiffness at appropriate intervals to capture performance
characteristics.
• Temperature monitored continuously during the test (Temperature is controlled
at 20 ± 2 °C).
• Lateral wander setting was not used in the tests

3 Experimental Results and Analysis

3.1 Strain Measurement at the Underside of Slab

Four KYOWA-70-120-A1-11 L5M3R strain gauges were attached to the underside

of slab to measure and monitor strain status during trafficking. The Microstrain
G-link wireless node was used as data logger to measure and collect the strain
gauge data. The layout of the four strain gauges is shown in Fig. 2.
Comparative Evaluation of Performance of Warm Mix RAP Asphalt … 851

Fig. 2 Strain gauges instrumentation schematic at underside of slab

Due to the vulnerability of the strain gauges, in the measurement of Lane 9B

bottom lift, one transverse strain gauge (#4) failed to function after 50,000 loading
cycles, only one transverse and one longitudinal strain gauge (#1 and #2) were
available after 100,000 axle passes. After 200,000 loading applications, all four
strain gauges were damaged.
Apart from data reading of Lane 9B bottom lift at interval of 100,000 cycles, the
data logger was observing the strain history continuously from the beginning of
loading for Lane 9A top lift. However, all four strain gauges were damaged shortly
after 50 min of trafficking. As a result, this continuous monitoring produced
tremendous useless data. Due largely to these two crucial disadvantages, the
practice of continuous monitoring of strain was abandoned in the subsequent tests.
The strain history of the four test lifts are presented in Fig. 3.
In the test of Lane 5 top lift, the four strain gauges survived for a long life span.
Until the end of 500,000 axle loads, three out of the four strain gauges were still
available for measurement. In order to gather more detail of strain history, the data
acquisition was conducted at interval of 10,000 prior to 100,000 cycles, the sub-
sequent interval was increased to 25,000. The strain history of Lane 5 top lift along
with the entire trafficking process is shown in Fig. 4. It was noted that the strain
values decreased with the accumulation of number of load passes and the transverse
strain was much larger than the longitudinal strains.
During the trafficking of Lane 5 bottom lift, one transverse strain failed after
experiencing 20,000 cycles. One was damaged after 600,000 cycles and the other
two were working until subjected to 1 million loadings. The strain history output of
Lane 5 bottom lift is presented in Fig. 4 as well.
It is interesting to note that the magnitudes of two tensile strains located at 90°
i.e. in different directions were similar in the Lane 9B bottom lift test, however, in
the test of Lane 9A top lift and Lane 5 top lift, the transverse tensile strain was
much larger than longitudinal tensile strain which is in good agreement with the test
results of Bhattacharjee et al. (2011). However, the transverse strain gauges man-
ifested much higher resilient permanent viscoelastic strains compared to longitu-
dinal strain gauges. The tensile strains of both directions of Lane 9A top lift and
Lane 5 top lift were much higher than the strain of Lane 9B bottom lift, which is
reasonable since they were undergoing more strain on top of two layers of
neoprene.
852 Y. Huang et al.

(a) Longitudinal strain of Lane 9B bot lift (b) Transverse strain of Lane 9B bot lift

(c) Longitudinal strain of Lane 9A top lift (d) Transverse strain at of Lane 9A top lift

(e) Longitudinal strain of Lane 5 top lift (f) Transverse strain at Lane 5 top lift

(g) Longitudinal strain of Lane 5 bottom lift (h) Transverse strain at Lane 5 bottom lift

Fig. 3 Strain output

It is clear from the strain output history that the longitudinal strain consists of a
compressive part followed by a tensile part when the load is moving past the strain
gauge, while the transverse strain is entirely in tension. From the pattern shape of
Comparative Evaluation of Performance of Warm Mix RAP Asphalt … 853

Lane 5 top lift Lane 5 bottom lift

Fig. 4 Strain history of Lane 5 top lift and Lane 5 bottom lift

transverse strain output, the strain reduces at a decreasing rate during the unloading
phase. As expected the phase of relaxation after the strain reaching peak value. The
strain plots are comparable and similar to previous studies of strain history in both
full scale traffic loading and MMLS3 loading (Bhattacharjee et al. 2011).

3.2 Numerical Simulation

Both Finite Element Method (FEM) software ABAQUS and elastic multiplayer
analysis program BISAR were used to determine the tensile stress in the slab under
MMLS3 trafficking. The results were compared with test measured strain values for
verification. Asphalt concrete exhibits both linear elastic and viscoelastic properties,
and was considered as linear viscoelastic material in this study. Figure 5 shows the
master curve of the Lane 5 and Lane 9 material in terms of the dynamic modulus
test of the field cores (Gibson et al. 2016) while the Prony series were obtained via
the extra phase angle data. Based on the dynamic modulus test results, Lane 5 was
stiffer than Lane 9.
The sub-layers were treated as linear elastic materials. The neoprene was mod-
elled with a Young’s modulus of 350 MPa and a density of 1350 kg/m3 (Lee and
Kim 2004). The concrete floor was modeled with a thickness of 200 mm, a density

Lane 9 Lane 5

Fig. 5 Master curve of Lane 9 and Lane 5 asphalt materials

854 Y. Huang et al.

Table 2 Strain comparison at beginning phase of trafficking

Longitudinal strain at underside of slab (l strain) Test ABAQUS BISAR
Lane 9B bottom on one layer neoprene 72 136 123
Lane 9A top on two layer neoprene 152 145 159
Lane 5 top lift on two layers of neoprene 232 215 246
Lane 5 top bottom on two layers of neoprene 264 227 252

of 2400 kg/m3 and a Young’s modulus of 2700 MPa. The modeling simulated a
wheel passes the experimental section at the measured speed under the corre-
sponding temperature. In this paper, a UMAT subroutine program was written to
simulate moving load at speed of 2.8 m/s. For similarity, the tire load was assumed
to distribute uniformly over the contact area. Steyn and Maina (2016) used flexible
pressure sensors to measure the MMLS3 tire contact stress and contact area under a
2.7 kN load at static and 3600 cycles/h condition respectively. According to this
study, the average contact stress was around 500 kPa at static and 400 kPa at 3600/h
and the contact area was measured to be 6700 mm2. In the ABAQUS simulation
model, the contact stress was set to 330 kPa at a frequency of 7200 cycles/h and
400 kPa for a frequency of 3600/h. In the simulation analysis, the step time was set
to 0.1 s for the dynamic analysis of the slab response (Fig. 6). The BISAR solution
considered the asphalt slab as an elastic material and the load as a static load with a
circular contact area on the top of slab. Table 2 shows the comparative results
between ABAQUS and BISAR for the tensile strain at the underside of slab.

3.3 PSPA Seismic Modulus

The PSPA was used to measure seismic stiffness to indicate the integrity change of
the slab during the trafficking process. More information regarding PSPA could be

(a) Visualization of the model (b) Strain at underside of slab

Fig. 6 Numerical simulation in ABAQUS

Comparative Evaluation of Performance of Warm Mix RAP Asphalt … 855

Lane 9B bottom lift Lane 9A top Lift

Lane 5 top lift Lane 5 bottom lift

Fig. 7 Seismic modulus versus number of passes

found in Jurado et al. (2012). Based upon the test results of ALF field lanes (Gibson
et al. 2012) during the ALF trafficking, the FWD, LWD and PSPA were used to
measure the relevant moduli. Only the seismic stiffness of PSPA can effectively
manifest the degradation of asphalt integrity, while the measurements of FWD and
LWD manifested almost no change along with the cumulative axle loads.
Figure 7 shows the seismic modulus measurements plotted against number of
axle passes of Lane 9B bottom lift, Lane 9A top lift, Lane 5 top lift and Lane 5
bottom lift, respectively during the MMLS3 trafficking. In the diagrams, the black
vertical dash line indicated the fatigue failure. There are some abnormal data points
observed in the diagram, resulting from the early failure occurred during trafficking
or system error of the PSPA. The PSPA measure points were located at three
different positions at the centerline of wheel path in trafficking direction. The
measure point 200 mm off the slab end where the load begins were denoted as near
end, 700 mm off position as middle section, 1 m off position is denoted as far end.
In the middle section, an extra position 60 mm off the centerline of wheel path was
measured to serve as comparison base where the position was not undergoing any
trafficking.
It was observed that, for the Lane 9B bottom lift, the seismic modulus of three
positions dropped to half of initial values at 600,000 load cycles, however, after the
600,000 axle loads, the seismic modulus manifested some abnormal variations. It
was considered to be related to manifestation of large cracks and the damage at the
underside of the slab. In addition, according to the seismic waveform interpretation
856 Y. Huang et al.

after 700,000 axle loads, the slab shows multiple signs of shallow delamination or
cracking damage. Especially, the linear gradient waveform implies multiple
reflections/travel paths of a badly cracked dispersive medium.
In Lane 9A top lift test, the far end was the first to manifest damage after being
subjected to 300,000 load cycles resulting in the observed abnormal PSPA values.
The near end section PSPA value increased dramatically when approaching
400,000 load applications due to the fatigue cracking. The seismic modulus of the
three sections dropped to around 5 GPa after 500,000 cycles, which indicated the
failure of the whole slab.
The seismic stiffness of Lane 5 top lift had no substantial change during MMLS3
loading. It can be seen that, at the far end section, the abnormal seismic modulus
value emerged at 300,000 cycles, and reached to the 5.8 GPa at 400,000 cycles,
which is the measured lowest value of during the trafficking. There were no
extensive variations of the seismic modulus at near end and far end section. This
may have been as a result of the higher content of RAP at 40 %, compared to the
20 % RAP content of the other two slabs. As can be seen, Lane 5 bottom lift
sustained 2.25 million trafficking load cycles. It was aimed to observe more fatigue
cracks appeared on the top surface. After 900,000 loading applications, the seismic
modulus of all the three positions dropped to the half initial values, which indicated
that fatigue failure had occurred. Compared to the other three slabs, Lane 5 bottom
lift proved to have the most fatigue resistance.
Comparing the seismic modulus record under MMLS3 trafficking, Lane 9A top
lift and Lane 5 top lift reach failure and fatigue cracks appear much earlier than
Lane 9B bottom lift, which attributes to the two layers of neoprene underneath Lane
9A top lift. The two layers of neoprene have more deflection and the slab undergoes
more vertical displacement or vibration when the tire loads passes on the top side.
In general, the seismic stiffness was comparable to those measured under
full-scale trafficking of the ALF. Specific values are still subject to validated release
of the data by the Pavement Testing Facility (PTF) of the FHWA in McLean, VA,
USA.

4 Comparative Evaluations of the Tests Findings

4.1 Rutting Depth

The profilometer, with spacing of 2 mm, was used to measure the cross-section
profile of the slabs at different positions. Figure 8 illustrates the transverse profiles
and rutting depth of three test slabs along with the trafficking process. During the
test of Lane 9B bottom lift, after the measurement at zero axle loads, the pro-
filometer failed to work due to the damage of USB adapter. But, in order to save test
time, the trafficking continued until reaching 500,000, when the profilometer was
reinstated. The cross section profile without any trafficking served as baseline for
Comparative Evaluation of Performance of Warm Mix RAP Asphalt … 857

Lane 9B bottom lift Lane 9A top Lift

Lane 5 top lift Lane 5 bottom lift

Comparable Rutting depths

Fig. 8 Cross section profile

the subsequent rutting depth calculation. The curves above the baseline part means
the upheaval section of the rutting, denote as positive value in the diagram;
meanwhile the section lower than baseline means the depression part of the rutting,
which is denoted as negative value.
Table 3 shows the rutting depth and width of the three slabs after 500,000 load
cycles. It should be noted that the rutting width of Lane 9B bottom lift and Lane 9A
top lift are the same. In contrast, the rut depth of Lane 9B bottom lift was 3.1 mm at
500,000 loading cycles and the rut depth of Lane 9A top lift was 4 mm after
500,000 cycles. The difference is attributed to the placement of Lane 9A top lift on
two layers of neoprene. The rutting depth of Lane 5 top lift reached 2.5 mm after
500,000 load applications, while Lane 5 bottom lift was observed only 0.7 mm after
500,000 load cycles. Comparing the cross section profiles of the four test lifts along
with trafficking, both two Lane 5 lifts manifested lower rutting depth but wider
rutting width than other two slabs. This was taken to be as a result of the higher
858 Y. Huang et al.

Table 3 Rutting comparison between the three test lanes after 500,000 load applications
Lane 9B bottom Lane 9A top Lane 5 top Lane 5 bottom
lift lift lift lift
Rutting Depth 3.1 4 2.5 1.3
(mm)
Rutting Width 100 100 240 210
(mm)

stiffness of the asphalt of Lane 5. Furthermore, it correlated well with the initial
seismic modulus measurement results without any traffic load due to its higher
content of RAP. It could also have been due to an unplanned, unexpected drop in
the laboratory temperature to only 15 °C during the initial 200,000 load applica-
tions of Lane 5.
Another aspect that was considered, is the effect of the composition of the two
pavements that were tested. The pavement structure of ALF test lanes is composed
of 100 mm asphalt layers, 560 mm base layer and deep underlying subgrade. The
granular aggregate base that had been constructed in a previous project, was
reconditioned to be compacted using a smooth drum vibratory roller, and the
subgrade was not disturbed due to the good protection of the upper thick layers. In
contrast, the down-scaled pavement structure of the MMLS3 consisted of 50 mm
asphalt layer, 50 mm neoprene slab and underlying concrete floor. That is, the
pavement thickness of ALF test lanes is effectively at least 6.6 times that of the
MMLS3 pavement thickness. According to a 3D dimensional analysis, this yields a
vertical stress ratio of 2.8:1. Earlier studies have supported the hypothesis that the
resulting ruts would develop in the same ratio. This was compatible to the actual
performance observed. Although the stress is the primary factor affecting the rut
performance, other factor need to be taken into account. For instance, the stiffness
modulus of the base layer, GAB is substantially lower than the asphalt layer of the
ALF.
On the basis of the above analytical deduction, it was therefore concluded that
the rutting performance under MMLS3 trafficking and the f-s ALF, is compatible
under the prevailing conditions.

4.2 Crack Length and Relevant Fatigue Performance

Both the full scale ALF and MMLS3 tests were conducted at similar ambient
temperature of 20 °C. Therefore, in both test cases, the test conditions in terms of
traffic and temperature are similar. However, in terms of material characteristics and
pavement structure, they were different. In the case of Lane 9 of full scale testing,
the total linear crack length reached 2771 mm per meter length after 400,000 ALF
trafficking cycles. The total crack length of Lane 9B top lift under MMLS3 was
1566 mm after 300,000 trafficking axle loads, while 2352 mm was observed on
Comparative Evaluation of Performance of Warm Mix RAP Asphalt … 859

Fig. 9 Measured MMLS3

fatigue cracking of Lane 5
bottom lift

Lane 9B bottom lift after 1.2 million axle loads. The crack information of Lane 5
bottom lift is shown in Fig. 9. It can be seen, the cumulative linear length crack at
top surface approached 2961 mm after 2,250,000 load cycles. It should be noted
that these two slabs were placed on one neoprene layer of 25 mm for MMLS3
trafficking. It is concluded that the fatigue cracks of the MMLS3 were slightly less
than that under the ALF trafficking.
In comparison with the fatigue performance of Lane 9, the two lane 5 MMLS3
test lifts exhibited better fatigue resistance. This is based on the finding that the
crack length was less and the later cracks occurred after much more trafficking. This
is in contrast to findings from ALF. The hypothesis of the authors is that the reason
for this is related to the supporting material property under asphalt layer of Lane 5
being poorer than that under Lane 9. This would not manifest in the case of
performance under MML3 trafficking, is due to the shallow nature of the influence
test depth. In contrast, the relative performance of the two asphalt layers is in actual
fact more representative of the material properties. Accordingly, the authors con-
clude that the performance of the respective asphalt materials should be considered
to be similar or in fact, the performance of Lane 5 is better than that of Lane 9. On
the inspection on the slabs after extractions, it was also found the underside of slab
5 was smoother and denser than Lane 9, probably as a result of the supporting
platform that has been placed prior to paving. This could also have influenced the
differential performance. However this cannot be quantified specifically.

5 Conclusions

Based on the discussions and data in this paper, the following conclusions are
drawn:
1. The findings from the three MMLS3 comparative tests indicated that the
material integrity of the asphalt slabs had not been jeopardized after the
extraction from the ALF test lane and transportation to Blacksburg. Based on the
correspondence with FHWA, the initial PSPA value of Lane 9 is around
860 Y. Huang et al.

16 GPa, which is a slightly larger than the measurements of the un-trafficked

slabs in the laboratory. This is ascribed to differences in thickness and sup-
porting structure of the two pavements.
2. After 300,000 trafficked axles, the seismic modulus of Lane 9 dropped to 5 GPa
which indicated the onset of fatigue failure. This is comparable to the reported
in situ PSPA stiffness parameters of Lane 9 at T-F.
3. The initial indicators of performance provided useful comparative indicators
related to the slabs tested from Lane 5 and Lane 9. It was concluded that the
parameters that had been monitored served to validate the comparative perfor-
mance characteristics of the two structural systems and intrinsic material charac-
teristics. This finding is also comparable to earlier studies. Testing of extracted slabs
is continuing to complete the test matrix with further refinement of the test setup.
4. The strain value at the underside of the slab decreases with accumulation
number of passes and the fatigue damage.
5. The findings from one third scale MMLS3 test and full scale ALF test on
pavement fatigue and rutting performance are comparable in terms of rutting
depth, strain response, seismic stiffness and cracking. This validated the MMLS
to be an effective, economic and reliable trafficking tool to characterize rutting
and fatigue performance of pavement materials with due regard to the relative
structures. Accordingly, it can be used to enhance economic pavement designs
with unproven composite materials.
6. The Lane 5 HMA material with 40 % RAP has higher stiffness than the Lane 9
WMA material with 20 % RAP. It was noteworthy the former material had
better resistance to permanent deformation and fatigue cracking than the latter.
The is contrary to the ALF findings and was ascribed to factors related to the
structural composition of the tested structure.
7. The latter finding is considered to be important for the planning and execution of
tests that could be affected in a similar manner.

Acknowledgments Sincere appreciation is expressed to all the researchers that participated in the
respective collaborative projects efforts. The authors also thank all of the sponsors that supported
the project. Without this, it would not have been possible to achieve successful outcome of the
project. A particular word of thanks to Nelson H. Gibson, Ph.D., PE. Research Civil Engineer
(Highway) of the FHWA Office of Infrastructure R&D, who gave the permission to get the slab
extracted from FHWA and provided the fatigue performance test results after ALF trafficking.

References

Bhattacharjee, S., Gould, S. G., Mallick, R. B., & Hugo, F. (2011). An evaluation of use of
accelerated loading equipment for determination of fatigue performance of asphalt pavement in
laboratory. International Journal of Pavement Engineering, 5(2), 61–79.
Gibson, N., Carvalho, R., Li, X., & Andriescu, A. (2016). Fatigue cracking of recycled and warm
mix asphalts: Predicted performance considering structural variability influences. Journal of
the Association of Asphalt Paving Technologists.
Comparative Evaluation of Performance of Warm Mix RAP Asphalt … 861

Gibson, N., Qi, X., Shenoy, A., AI-Khateeb, G., Kutay, M., Andrisecu, A., et al. (2012).
Performance testing for superwave and structural validation. Technical Report. Report
No. FHWA-HRT-11-045.
Hugo, F., Huang, Y., Xiong, H., Wang, L., & Steyn, W. (2015). Lessoned learned during the first
application of MSP for extracting asphalt slabs in comparative testing of fatigue performance of
warm mix RAP asphalt MMLS3 trafficking. In ISAP 2015 seminar, Suncity, South Africa.
Jurado, M., Gibson, N., Celaya, M., & Nazarian, S. (2012). Evaluation of asphalt damage and
cracking development with seismic pavement analyzer. Journal of the Transportation
Research Board, 2304, 47–54.
Lee, S., & Kim, Y. R. (2004). Development of fatigue cracking test protocol and life prediction
methodology using the third scale model mobile load simulator. In Fifth international RILEM
conference on cracking in pavements mitigation, risk assessment and prevention (pp. 29–36).
Steyn, W. J. vdM. & Maina, J. W. (2016). Effects of surface condition on non-uniform contact
stresses of APT tests. In Paper submitted to the 5th international conference on accelerated
pavement testing, Costa Rica.
Evaluation of the Optimum Percentage
of RAP and RAS in Asphalt Mixes
in Texas Using Accelerate Pavement
Testing

Stefan A. Romanoschi, Tom Scullion, Fujie Zhou

and Reza Saeedzadeh

Abstract The first APT project at the University of Texas at Arlington took place
between May 2012 and August 2015. The project, funded by the Texas Department
of Transportation (TxDOT), aimed to validate the maximum allowable percentage
of Recycled Asphalt Pavement (RAP) and Recycled Asphalt Shingles
(RAS) allowed by TXDOT Specifications. To accomplish the validation, twelve
pavement structures were built and subjected to APT loading. The experiment was
a factorial combination of four asphalt surface mixes with different percentages of
RAP and RAS tested for their resistance to rutting, fatigue cracking and reflection
cracking. The project also aimed to validate that poor results on lab prepared asphalt
mixes containing RAP and RAS translates to poor field performance under
full-scale axle loads and to verify overlay design tools recently developed by Texas
A&M Transportation Institute (TTI) to predict mix cracking and rutting of new
asphalt overlays. The paper presents the development of the APT experiment,
briefly discusses the recorded performance of the studied mixes and provides a
summary of the results obtained in this experiment.

Keywords Recycled asphalt pavement Recycled asphalt shingles Balanced mix

design

Recycled Asphalt Pavement (RAP) has successfully been used in the United States
for more than 30 years; if designed under established mixture design procedures
and produced under appropriate quality control/quality assurance (QC/QA)

S.A. Romanoschi (&)
R. Saeedzadeh

Department of Civil Engineering, University of Texas at Arlington,
416 Yates St., Suite 425, Arlington, TX 76019, USA
e-mail: [email protected]
T. Scullion
F. Zhou
Texas A&M Transportation Institute, Texas A&M University System,
3135 TAMU, College Station, TX 77843-3135, USA

© Springer International Publishing Switzerland 2016 863

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_56
864 S.A. Romanoschi et al.

measures, asphalt mixtures incorporating RAP perform comparably to conventional

mixtures. The current national guideline for determining the binder grade adjust-
ment in HMA mixes incorporating RAP require:
• No change in binder grade for RAP content less than 15 %.
• The use of a virgin binder one grade softer than normal for RAP content
between 15 and 25 %.
• The determination of the virgin binder grade using blending charts for RAP
content more than 25 %.
However, many states have decided to change these limits to adapt to local
conditions. Several studies compared the performances of virgin and recycled
asphalt pavements and found them to be similar.
The use of Recycled Asphalt Shingles (RAS) in asphalt mixes has started rel-
atively recently. Therefore, there is no published information on in-service per-
formance of mixes incorporating RAS. Also, no mix design method specifically
addresses the incorporation of RAS alone, or with RAP, in recycled asphalt mixes.
RAP/RAS mixes are generally stiffer than virgin mixes. So it is anticipated that
RAP/RAS mixes are more rutting resistant but they may be prone to cracking.
Therefore, it is very challenging to design mixes with such stiff RAP/RAS mate-
rials. Either softening agents or rejuvenators should be considered to lower the
Performance Grade (PG) of the RAP/RAS binder. Another option to improve the
durability of these mixes is to use the Balance Mix Design (BDM) approach.
The BDM mix design approach was developed by the Texas A&M Transportation
Institute and was proposed to the Texas Department of Transportation (TxDOT) for
virgin mixes.
The current national guideline for determining the binder grade adjustment in
HMA mixes incorporating RAP is given in Table 1. (1) Some State transportation
departments have modified the range of percentages (e.g., increased the RAP
percentage that can be used before a softer binder grade must be chosen) based on
conditions in that area and/or additional testing.
For percentages of RAP greater than 25 %, procedures for developing a
blending chart are provided in the appendix of AASHTO M 323 (AASHTO 2010).
Two options are addressed:
I. When the desired final blended binder grade, the desired percentage of RAP,
and the recovered RAP binder properties are known, the required properties of
the appropriate virgin binder grade can be determined according to blending
chart procedures.

Table 1 Binder selection guidelines for RAP mixtures according to AASHTO M 323 (1)
Recommended virgin asphalt binder grade RAP percent in the mix,
by weight
No change in binder selection <15
Select virgin binder one grade softer than normal (e.g., select PG 15–25
58-28 if PG 64-22 would normally be used)
Follow recommendations from blending charts >25
Evaluation of the Optimum Percentage of RAP and RAS … 865

II. When a specific virgin asphalt binder grade must be used and the desired
blended binder grade and recovered RAP properties are known, the allowable
percentage of RAP is determined according to blending chart procedures.
Currently, most states allow the use of RAP in HMA but not all states allow the
use of RAS. Texas is one of the states allowing both RAP and RAS, including
tear-off RAS. Table 2 gives the maximum allowable percentages of RAP and RAS
and the maximum allowed ratio of recycled binder to total binder in the recently
adopted Texas specification (TxDOT 2014). The Texas specification falls within the
range of the maximum allowable RAP/RAS usage but close to the maximum limits
(Romanoschi and Scullion 2014).
TxDOT’s current specifications have permitted the use of both RAP and RAS,
but there is little evidence that these mixes give comparable performance to mixes
that do not incorporate any recycled materials. Concerns have been raised about
mixes which are now “drier”, more difficult to compact and more susceptible to
both reflection and fatigue cracking. The concern is primary for the mixes used in
the asphalt surface layer or in asphalt concrete overlays.
In response to this need, the Texas Department of Transportation (TxDOT)
sponsored a research study that required the construction and testing of twelve
flexible pavement test sections using Texas mixes. The twelve typical TxDOT
pavement structures were subjected to accelerated pavement testing (APT) at the
recently built Accelerated Pavement Testing Facility (APTF) in Fort Worth, Texas.

Table 2 Texas specification

Mixture Maximum ratio of recycled Maximum allowable % (percentage by
description and bindera to total binder (%) weight of total mixture)
location Unfractionated Fractionated RASd
b c
RAP RAP
Surface mixese 35 10 20 5
Non-surface 40 15 30 5
mixesf < 8 in.
From final
riding surface
Non-surface 45 20 40 5
Mixesf > 8 in.
From final
riding surface
a
Combined recycled binder from RAP and RAS
b
Do not use in combination with RAS or fractionated RAP
c
May not be used in addition to unfractionated RAP; however, up to 5 % of fractionated RAP may
be replaced with RAS
d
May be used separately or as a replacement for no more than 5 % of the allowable fractionated
RAP
e
“Surface” mixes are defined as mixtures that will be the final lift or riding surface of the pavement
structure
f
“Non-Surface” mixes are defined as mixtures that will be an intermediate or base layer in the
pavement structure
866 S.A. Romanoschi et al.

1 Experimental Pavement Structures

The study focused only on surface mixes with different levels of RAP and RAS; the
goal was to validate the cracking and rutting potential of both control and
RAP/RAS modified mixes within a controlled APT program.
The configuration and location of the experimental pavement structures as well
as the type of mix used for each layer; they are given in Table 3 and Fig. 1.
Construction of the experimental pavements at the APRF facility was finalized in

Table 3 Configuration of experimental pavement structures at the APRF facility

Reflection Cracking Experiment Fatigue Cracking Experiment
Surface Intermediate Base Test Surface Base Subbase
Test 50 mm 50 mm 200 mm Section 75 mm 200 mm 200 mm
Section (2.0 in.) (2.0 in.) (8.0 in.) (3.0 in.) (8.0 in.) (8.0 in.)
A Type D Cement L Type D Cement
B High RAP (3.5%) E High RAP Bridgeport (2.0%)
C RAP&RAS Treated F RAP&RAS Rock Treated
Type C Base Subbase
D BMD G BMD
Type D – contains no RAP
Rutting Experiment High RAP = (19% RAP)
Surface Intermediate Base RAP&RAS = (15%RAP + 3%RAS)
Test 50 mm 150 mm 175mm BMD = Balanced Mix Design
Section (2.0 in.) (6.0 in.) (7.0 in.)
H Type D Cement
I High RAP (3.5%)
J RAP&RAS Treated
Type B Base
K BMD

Fig. 1 Test sections built at the APRF facility

Evaluation of the Optimum Percentage of RAP and RAS … 867

Table 4 Laboratory design information for the mixes used

Mix % Rock RAP % RAS TD AC PG LAS
BPD BPMS MCMS FS % % % %
Type D Control
D134 - TTI 61 - 30 9 - - 96.5 4.8 64-22 1
High RAP
D119 - TTI 46 29 - 6 19 - 96.5 4.8 64-22 1
RAP & RAS
D196 -TTI 48 29 - 5.6 15 3 96.5 5.0 64-22 1
+WMA
BMD
D196 - TTI 48 29 - 5.6 15 3 97.5 5.5 64-28 1

Type C
FT1C195 50 27.6 5 15 2.4 97.0 4.7 64-22 1
Type B
FT1B115 38 32 30 97.0 4.1 64-22 1

BPD – Bridgeport D rock BPMS – Bridgeport Manufactured Sand

MCMS – Mill Creek Manufactured Sand FS – Field Sand
OAC – Optimum Asphalt Content LAS – Liquid Anti-strip
TD – Target Density as percentage of Maximum Theoretical Density

mid-February 2013. Details of the asphalt mixes used are given in Table 4. Four
asphalt surfacing mixes were used and their design is given in Table 4. The Control
Mix was placed as the final surface on Sections A, L, and H.

2 APT Testing Conditions: Loading and Temperature

Control

The Accelerated Pavement Research Facility (APRF) is a research facility built on

the Fort Worth campus of the University of Texas at Arlington. The main testing
equipment is the Pavement Testing Machine (PTM). The PTM is a linear APT
testing device that can be transported easily to remote locations, if needed. It
consists of a steel frame that acts as a reaction frame and a moving bogie that travels
at speed of up to 8 km/h (5 mph). An axle attached to the bogie can apply a single
axle load of up to 162 kN (36,000 lbf). The central portion of the PTM is covered
by a temperature control chamber equipped with heating and cooling units
(Romanoschi and Scullion 2014). Figure 2 shows the exterior of the PTM.
Bi-directional trafficking was applied for the “fatigue cracking” and “rutting”
pavement sections while uni-directional trafficking was applied for the “reflection
cracking” pavement sections. In the uni-directional traffic, the axle is loading the
pavement sections only when it travels from south toward north. The experimental
pavement sections are loaded in pairs, with one wheel of the axle loading each
pavement section.
868 S.A. Romanoschi et al.

Fig. 2 Photos of the pavement testing machine (PTM)

An 81.6 kN (18,000 lb) single axle load and a tire inflation pressure of 724 kPa
(105 psi) was used throughout this experiment. The axle load and tire inflation
pressure was kept the same during the entire duration of the experiment. The tire
inflation pressure was checked every 4 weeks and it was found to be very stable.
The axle load was measured several times throughout the project using wheel load
scales. It was maintained constant by always selecting the same settings for the
hydraulic valve that controls the oil pressure in the hydraulic system that pushes the
loading axle down on the pavement.
The lateral position of the PTM machine was changed during testing such that it
follows a normal distribution with a standard deviation of 200 mm (8.0 in.) and a
maximum lateral position of 305 mm (12.0 in.). This way, the lateral wonder
follows a normal distribution truncated at the central 87th percentile (6.6 % on each
tail) (Romanoschi et al. 2015).
The “fatigue cracking” and “reflection cracking” pavement sections were tested
at the target temperature of ±20 °C (±68 °F) while the testing of the “rutting”
pavement sections was conducted at the target temperature of ±40 °C (±104 °F).
The heating and cooling units mounted on the PTM machine helped with tem-
perature control.
Evaluation of the Optimum Percentage of RAP and RAS … 869

3 The Performance of the Experimental Pavement

Sections

3.1 Permanent Deformation

The permanent deformation at the pavement surface was measured using an in-house
built profiler. Five transverse profiles were recorded for each sections during one
measurement session. Figure 3 shows the evolution of the average permanent
deformation (computed from the 5 transverse profiles measured in each section) for
the fatigue cracking and rutting sections. The reflection cracking sections did not
exhibit significant permanent deformations since they have a strong base and an
intermediate asphalt concrete layer. At the end of loading, the permanent defor-
mation was less than 0.3 in. Their failure was due to the extent of reflection cracks.
This failure mode was desired when the experimental test sections were designed.
The following was observed from the evolution of permanent deformation:
• After 700,000 passes of the PTM machine, the permanent deformation in the
rutting sections (H, I, J and K) was less than 5 mm (0.2 in.). This confirmed that
all asphalt mixes paved in this project have a very good rutting resistance.
• The largest permanent deformations were recorded for the fatigue cracking
sections G (BMD) and L (Type D), due to the soft foundation underneath the
asphalt concrete layer. As found during the post-mortem investigation, these
sections also had thinner asphalt concrete surface layers than the fatigue
cracking sections E (High RAP) and F (RAP&RAS). The foundation was
constructed soft with the intent of allowing the flexure of the asphalt concrete
surface layer to induce fatigue cracks. This failure mode was desired when the
experimental test sections were designed.

Fig. 3 Surface rutting recorded on the fatigue cracking and rutting sections
870 S.A. Romanoschi et al.

3.2 Reflection Cracking

During the construction of the reflection cracking sections A, B, C and D, three

full-width transverse saw cuts were made in the base and intermediate layers (Fig. 1).
During the accelerated loading, reflection cracks developed only above these saw cuts.
The position and the length of the reflected cracks was recorded periodically during the
accelerated loading. The length was divided by the full-width of the lanes to determine
the percent lane cracked shown in Fig. 4. The figure indicates that the Balanced Mixed
Design (BMD) approach leads to a mix with better resistance to reflection cracking;
section D performed the best. To determine the design binder content, the BMD
approach uses the results of the Hamburg Wheel Rut Tester (HWRT) and the Overlay
Tester (OT), a test developed at the Texas A&M Transportation Institute to determine
the fracture resistance of asphalt mixes (Zhou and Scullion 2005; Zhou et al. 2007). The
mix containing High RAP (Section B) and the mix containing RAP and RAS
(Section C) exhibited more reflection cracking than the mix with no recycled material
(Type D—Section A) and the BMD mix (Section D).

3.3 Fatigue Cracking

The surface of the APT loaded pavements was periodically inspected for surface
cracks. After the first crack was observed, cracking measurements were performed
periodically in order to record the location and extent of surface cracks. For this
purpose a rectangular constructed wood frame has been built. The frame has lon-
gitudinal and transverse strings at 150 mm (6 in.) spacing to form a grid with 6 in.
square openings. While the wood frame is positioned on the pavement, each
observed crack is recorded manually on a paper template using the string grid as the

Fig. 4 Percent length of reflected cracks

Evaluation of the Optimum Percentage of RAP and RAS … 871

Fig. 5 Measuring, recording, and calculating of fatigue cracking sections

Fig. 6 Percent of cracked area in the fatigue cracking sections

reference. The extent of cracking is determined by counting the number of 6 in.

squares in which the cracks are present (Romanoschi and Scullion 2014). Figure 5
shows how field fatigue cracking was recorded and measured.
The fatigue cracking percentage is calculated based on the wheel path width
which is about 1.2 m (4 ft). Figure 6 shows the summarized fatigue cracking area
curves for all fatigue cracking sections (G, L, E, and F).
872 S.A. Romanoschi et al.

4 Comparisons Between the Predicted and Measured

Performance Data

One of the objective of this research project was to validate the pavement performance
prediction models in the two software programs used by TxDOT for the design and
analysis of new flexible pavements and asphalt overlays. The two mechanistic-
empirical based programs, TxME for new pavement design and TxACOL for asphalt
overlay design were developed at the Texas A&M Transportation Institute (Hu et al.
2011, 2014; Zhou et al. 2009, 2010). The inputs used for the two programs included
the measured layer thickness, the backcalculated subgrade soil modulus, the tem-
perature at the mid-depth of the asphalt layers and the wheel load. The results of the
Overlay Tester performed in the laboratory on samples extracted from the built
sections were used to determine the fracture properties of the studied mixes. More
details on the laboratory testing and the two programs are given elsewhere (Hu et al.
2014; Romanoschi et al. 2015).
Since the experimental sections were loaded in pairs, the performance difference
for two sections of the same pair is mainly caused by the surface mixture because
the load/temperature conditions are very close to each other for the paired test
sections. Thus comparing the paired test sections field performance with the

Fig. 7 Comparison of fatigue cracking test sections E, F, G and L

Evaluation of the Optimum Percentage of RAP and RAS … 873

Fig. 8 Comparison of reflective cracking test sections A, B, C and D

predicted performance provides valuable information to judge the rationality of

model predictions. Figures 7 and 8 show the comparison results only for fatigue
cracking and reflection cracking; the permanent deformation recorded for the rut-
ting sections was minimal.
Figure 7 shows that TxME gave a very good prediction of the fatigue cracking
evolution of sections E, F, G and L. It predicted a slower fatigue cracking pro-
gression for section L (Type D) than for section G (BMD), and a better cracking
performance for section E (High RAP) than for section F (RAP&RAS). Therefore,
the prediction model for fatigue cracking was successfully validated.
Figure 8 shows that TxACOL predicted well the reflection cracking evolution of
sections B (High RAP) and C (RAP&RAS). However, it predicted a better
reflection cracking performance for section D (BMD) than for section A (Type D)
while the opposite was recorded in the full-scale testing. The reflection cracking
prediction uses the fracture properties determined with the Overlay Tester (OT).
The virgin mix used in Section A (Type D) showed a better performance in the
laboratory test than the mix used in Section D (BMD), which was designed fol-
lowing the Balanced Mix Design (BMD) concept. However, the full-scale perfor-
mance indicated that the BMD mix performs better than the virgin mix. This
874 S.A. Romanoschi et al.

suggests that the Balanced Mix Design concept can be effectively used to improve
the performance of mixes containing recycled materials.

5 Conclusions and Recommendations

The first full-scale accelerated pavement test conducted at the APT Facility of the
University of Texas at Arlington investigated the rutting, fatigue and reflection
cracking performance of asphalt surface mixes containing the maximum percentage
of recycled material allowed by the current specification of the Texas Department of
Transportation. Approximately 3 million passes of single axle load were applied to
twelve pavement test sections. The measured performance data included surface
permanent deformation and crack mapping. An extensive laboratory investigation
determined the as-constructed properties of the materials used in the test sections.
The measured data were then used to verify two pavement design and analysis
programs: TxME for new pavement design and TxACOL for asphalt overlay
design. The following major findings resulted from this research project:
• Rutting is not a problem for Texas mixes, regardless of containing RAP/RAS;
• With the same testing and pavement structural conditions, the virgin mix has
better cracking resistance. The reflective cracking performance of test sections A
(Type D) and D (BMD) is somewhat confusing since the expecting result is that
Section performs better than Section D, because the surface mixture of
Section A is virgin Type D (OT cycles 67) and Section D is BMD (OT cycles
12). However, the FWD deflection shows that test section D is much softer than
the other test sections, which resulted in higher damage for Section D.
• Generally, both TxME and TxACOL had reasonable prediction for performance
of these APT test sections. Except the reflective cracking test sections A (Type
D) and D (BMD), the predicted results matches the field measured results very
well when comparing paired test sections, which clearly shows that models are
rational and sensitive to material properties.
The success of the first series of APT testing clearly showed the value of APT
testing for validating performance models and evaluating new pavement materials.

References

American Association of State Highway and Transportation Officials. (2010). AASHTO M 323:
Standard specification for superpave volumetric mix design. In Standard specifications for
transportation materials and methods of sampling and testing (30th ed.). Washington, DC:
AASHTO.
Hu, S., Hu, X., Zhou, F., & Walubita, L. (2011). Development, calibration, and validation of a new
M-E rutting model for HMA overlay design and analysis. Journal of Materials in Civil
Engineering, 23, 89–99.
Evaluation of the Optimum Percentage of RAP and RAS … 875

Hu S., Zhou, F., & Scullion, T. (2014). Development of Texas mechanistic-empirical flexible
pavement design system (TxME). Report No. FHWA/TX-12/0-6622-2. Texas A&M
Transportation Institute, College Station, TX.
Romanoschi, S. A., & Scullion, T. (2014). Validation of the maximum allowable amounts of
recycled binder, RAP, & RAS using accelerated pavement testing. Interim report. Report
No. FHWA/TX-14/0-6682-1, The University of Texas at Arlington, May 2014.
Romanoschi, S. A., Zhou, F., Scullion, T., & Saeedzadeh, R. (2015). Testing and evaluation of
RAP and RAS experimental pavement sections. Final report. Report No. FHWA/
TX-14/0-6823-1, The University of Texas at Arlington, October 2015 (in press).
TxDOT. (2014). Standard specifications for construction and maintenance of highways, streets and
bridges. 2014 Edition, ftp://ftp.dot.state.tx.us/pub/txdot-info/des/spec-book-1114.pdf.
Zhou, F., & Scullion, T. (2005). Overlay tester: A simple performance test for thermal reflective
cracking. Journal of the Association of Asphalt Paving Technologists, 74, 443–484.
Zhou, F., Hu, S., Scullion, T., & Chen, D. (2007). The overlay tester: A simple performance test
for fatigue cracking. Transportation Research Record, No. 2001 (pp. 1–8).
Zhou F., Hu, S., Hu, X., & Scullion, T. (2009). Mechanistic-empirical asphalt overlay thickness
design and analysis system. Research Report FHWA/TX-09/0-5123-3, Texas A&M
Transportation Institute, College Station, TX, October 2009.
Zhou F., Fernando, E. G., & Scullion, T. (2010). Development, calibration, and validation of
performance prediction models for the Texas M-E flexible pavement design system. Research
Report FHWA/TX-10/0-5798-2, Texas A&M Transportation Institute, College Station, TX,
August 2010.
Field Performance Evaluations of Large
Sized Unconventional and Recycled
Aggregates for Subgrade Improvement

Hasan Kazmee, Erol Tutumluer and Debakanta Mishra

Abstract Illinois Department of Transportation has recently introduced new gra-

dation bands to accommodate large sized unconventional and recycled materials in
subgrade remedial applications. Existing standardized test protocols cannot char-
acterize these materials properly. To this end, an accelerated pavement testing study
was undertaken to evaluate rutting performance trends of pavements constructed
with such aggregate materials. An advanced field image segmentation technique
was implemented for the in situ characterization of aggregate size and morpho-
logical properties, i.e. shape, texture and angularity, texture. Four pavement
working platform test sections were constructed with railway ballast or riprap sized
virgin aggregates as well as large-sized concrete demolition waste capped with
densely graded dolomite and reclaimed asphalt pavement (RAP) materials.
Construction quality control was achieved through nuclear gauge density checks
and modulus measurements, the latter utilizing lightweight deflectometer and soil
stiffness gauge. Next, the full scale test sections constructed over weak engineered
subgrades were tested with unidirectional wheel loading in accelerated pavement
testing. Periodic rut measurements were taken to quantify the rutting progression.
Contributions of the subsurface layers to surface rut accumulation were assessed by
the use of the following field equipment: variable energy lightweight penetrometer,
ground penetrating radar (GPR), and geo-endoscopic probe for visualization of
layer compositions and interfaces. Despite higher modulus properties obtained,
RAP capped sections typically accumulated higher permanent deformation.
Geo-endoscopic imaging revealed that presence of shallow water table led to early

H. Kazmee
E. Tutumluer (&)

Department of Civil and Environmental Engineering, University of Illinois at
Urbana-Champaign, 205 North Mathews Ave., Urbana, IL 61801, USA
e-mail: [email protected]
H. Kazmee
e-mail: [email protected]
D. Mishra
Department of Civil Engineering, Boise State University, 1910 University Drive,
ERB 3137, Boise, ID 83725, USA
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 877

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_57
878 H. Kazmee et al.

failure in one of the test sections. According to the study results, current Illinois
subgrade stability design framework was found to be adequate for utilizing large
sized unconventional and recycled materials in subgrade remedial applications.

Keywords Large-sized aggregate Recycled Pavement working platform Rock





cap Accelerated pavement testing Lightweight deflectometer Variable energy


penetrometer Geo-endoscopic imaging

1 Introduction

Stable subgrade is a requirement both during pavement construction and throughout

intended service life. Despite the advent of new stabilization techniques, maintaining
required subgrade stability during spring thaw season has always been a major
hurdle for transportation agencies in Midwest and northern USA. In total, 19 US
states have spring load restrictions to minimize the damaging effects of freeze-thaw
on pavement structures (Waalkes 2003). Similarly, thirty percent of total surficial
soils in the state of Illinois is hydric in nature indicating permanent or seasonal
saturation with high organic content as encountered in wetlands (USDA 2009).
Typical Illinois soils are also rich in silt content which contributes to moderate
permeability and problematic matric suction for frost heave considerations. To
construct working platforms on top of such wet and very weak subgrade soils,
Illinois Department of Transportation (IDOT) commonly uses thick layers of
unbound aggregate materials based on minimizing equipment sinkage and providing
adequate compaction as outlined in the IDOT Subgrade Stability Manual. The
thickness design mainly considers in situ subgrade strength and ignores the effects of
aggregate physical characteristics (Bureau of Bridges and Structures 2005).
Tutumluer et al. (2009) conducted an experimental study on strength, modulus
and permanent deformation characteristics of four crushed and uncrushed aggregate
materials engineered in the laboratory to varying percentages of fines content,
moisture level and plasticity index. They concluded that aggregate morphology, i.e.
imaging based shape, texture and angularity properties, and plasticity of aggregate
fines had significant influences on the strength and deformation behavior of
unbound aggregates (Tutumluer et al. 2009). Later on, Mishra and Tutumluer
(2013) conducted a field study on the adequacy of different qualities of aggregates,
including three large-sized aggregate materials for weak subgrade stabilization,
which involved accelerated pavement testing of full scale pavement construction
platform test sections (Mishra and Tutumluer 2013). Accordingly, six full scale test
sections were constructed over a subgrade with engineered strength of CBR = 1 %
and tested for rutting performance. In comparison to the test sections with regular
size base course aggregates, the large rock sections survived higher number of load
repetitions. Inspired from these findings, IDOT recently introduced three different
large size crushed stone gradation specifications, namely CS01, CS02 and RR01,
respectively (Kazmee and Tutumluer 2015). These gradation bands have maximum
Field Performance Evaluations of Large Sized Unconventional … 879

aggregate sizes ranging from 76 mm (3 in.) to 203 mm (8 in.). Such an improved

subgrade constructed with large size unconventional virgin or recycled aggregates,
e.g. recycled concrete aggregate (RCA) and reclaimed asphalt pavement (RAP),
IDOT refers to as the “aggregate subgrade” layer.
Popularly known as rock cap, similar large size aggregate materials (76 mm or 3
in. top size) are used by most transportation agencies to provide a free draining frost
blanket (Mathis 1991; Uhlmeyer et al. 2003). Designing with rock cap materials
often revolves around the use of premium quality crushed aggregates. Therefore,
adequacy of large size recycled materials for improved subgrade stabilization is yet
to be investigated. Moreover, due to the sheer size of these crushed rocks, stan-
dardized laboratory test protocols are often insufficient. To this end, a field study
was recently undertaken at University of Illinois to investigate the rutting perfor-
mance trends of such large size unconventional materials. Six different aggregate
subgrade materials with varying source, size and composition were used to con-
struct twelve different full scale construction platform type test sections over weak
subgrade with controlled strength. Quality control tests such as nuclear density and
in place modulus measurement were conducted alongside the placement and
compaction of each granular layer. The test sections were then tested for rutting
potential using the Advanced Transportation Loading Assembly (ATLAS) at the
Illinois Center for Transportation (ICT). Transverse scanning with ground pene-
trating radar over the rut measurement locations helped to evaluate the subsurface
rutting patterns. In conjunction with periodic rut measurements, fixed and variable
energy dynamic cone penetration tests were carried out in the constructed test
sections to investigate linkages between strength properties and rutting performance
trends of construction platform test sections.

2 Objective and Scope

The objective of this study was to evaluate the comparative performances of selected
large size aggregate subgrade materials fitting certain IDOT gradation bands for
improved subgrade applications. To this end, six different aggregate subgrade and
two different capping materials with varying source, composition and gradation were
used to construct twelve full scale construction platform sections. The test sections
were then subjected to accelerated pavement testing and periodic rut measurements
were carried out at different number of wheel load applications. Considering the
limited scope of this paper, rutting performances of only four working platform
sections constructed with railroad ballast size riprap aggregates and large sized RCA
materials are presented herein. The study findings are complemented with the results
from a customized ground penetrating radar (GPR) used to assess the subsurface
rutting patterns in underlying layers, a variable energy PANDA penetration test
device to investigate linkages between layer strength and rutting characteristics, and
a state-of-the-art geo-endoscopic imaging technique used to visualize layer
compositions and interfaces as well as the depths of water table.
880 H. Kazmee et al.

3 Experimental Design of Working

Platform Test Sections

Figure 1 shows the three dimensional layout of proposed test sections. Two con-
secutive sections had similar aggregate subgrade (Type M ballast size riprap and
Type N rubblized concrete or RCA) with alternating capping layers (Type P
dolomite and Type O reclaimed asphalt pavement [RAP] aggregates). The sections
were designated with the type of capping stone followed by corresponding
aggregate subgrade material type. For example, the first section from the left hand
side in Fig. 1 had Type M riprap aggregates capped with Type P dolomite. Each of
the test sections was 4.6 m long and 2.7 m wide. The circular spot on the surface of
Section O-M indicates the tentative spot for quality control tests. Thicknesses of the
constructed layers were designed in accordance with IDOT Subgrade Stability
Manual so that the adequacy of the newly introduced gradation bands with the
unconventional large rocks could be evaluated. Henceforth, for an engineered
subgrade strength of CBR = 1 %, thicknesses of aggregate subgrade and capping
layer were selected to be 53.3 and 7.6 cm (21 and 3 in.), respectively.
The accelerated pavement testing machine at ICT comprised of a 25.9 m long
heavy vehicle simulator known as Advanced Transportation Loading ASsembly
(ATLAS) which was equipped with a super-single tire applying 44.5 kN (10 kip)
load uni-directionally at a constant speed of 8 km/h (5 mph) and exerting an
assumed uniform contact pressure of 758 kPa (110 psi). The four test sections were
enclosed with two 9.1 m long buffer zones for the ATLAS crawler placement and
speed stabilization. In addition, a 3.1-m long transition zone was provided between
the varying aggregate subgrade sections so that there was no interface effect during
the accelerated pavement testing. The transverse dash lines on the construction
platform surface denote the rut measurement locations; meanwhile the dotted lines
along the longitudinal direction illustrate the span of rutting measurement.

Fig. 1 Three dimensional view of the pavement construction platform test sections (scaling is
different in the vertical and horizontal direction)
Field Performance Evaluations of Large Sized Unconventional … 881

3.1 Aggregate Materials Used

The top left corner in Fig. 2 shows the particle size distributions of Type M through
Type P aggregates along with their target IDOT gradations. Capping materials,
namely Type O and P, were considerably finer than the aggregate subgrade (Type
M and N) materials. Type P crushed dolomite had the highest percentage of fines
(materials passing the No. 200 sieve or finer than 0.075 mm). The principal
motivation for regular base course type capping layer was to ensure proper
confinement/compaction effort for the large rocks which had large voids in the
granular matrix. Despite being well graded, Type O RAP material did not fit in the
conventional base course type CA06 gradation band. Owing to the clumping ten-
dency stemming from the presence of inherent binder and moisture, the RAP
aggregates had less than 1 % of fines content. Riprap Type M aggregates fitting the
RR01 gradation band was the coarsest among all the aggregate types. Even though
Type N rubblized concrete was selected to replicate representative CS01 gradation
aggregates, the coarser portion of that aggregate type did not meet the required
particle size distribution.
Besides the particle size distributions, aggregate shape properties, such as
imaging based quantifiable indices of angularity, surface texture and flat-elongated
ratio, were evaluated using the Enhanced University of Illinois Aggregate Image
Analyzer (E-UIAIA). Details of the aggregate imaging technique can be found
elsewhere (Maziar et al. 2013). The aggregate morphological indices are presented
in Fig. 2 in box whisker plots where the top and bottom part of the box stand for
3rd (75th percentile) and 1st (25th percentile) quartile of all the measured indices;
whereas, the diamond and the horizontal line inside the box represent the mean and

Fig. 2 Particle size distributions and morphological indices (angularity, surface texture and
flat-elongated ratio) of selected aggregates
882 H. Kazmee et al.

median index values, respectively. The recycled aggregates (Type N recycled

concrete and Type O reclaimed asphalt) exhibited lower surface texture and
angularity indices compared to the virgin crushed aggregate types (Type M riprap
and Type P dolomite). Based on the established ranges, Type P dolomite can be
categorized as angular and very rough with the highest shape indices. Type O RAP
was found to have the lowest shape indices among the four aggregate types
resulting in reduced aggregate interlock when subjected to repeated loading.

3.2 Modification of Subgrade Strength

The subgrade was engineered to a controlled California Bearing Ratio (CBR) through
a field preparation process. The existing soil was low plasticity clayey silt categorized
as CL-ML type according to the Unified Soil Classification System. According to the
compaction characteristics studied in the laboratory, to reach a corresponding CBR
value of 1 %, the field moisture content had to be maintained at 15 %. Accordingly,
the subgrade was graded on site and tilled to a depth of 30.5 cm (12 in.). Loose soil
samples were collected in grid like pattern and tested for moisture content in com-
pliance with ASTM D 4643. To achieve a uniform spatial distribution of moisture
content for the entire test road, any moisture deficient zones were sprayed with water
based on the volume of tilled earthwork and the required moisture percentage to reach
CBR = 1 % strength. After that, the tilled soil was compacted with a sheepsfoot
vibratory roller followed by dynamic cone penetration tests on the corresponding grid
spots. In place CBR values were determined from the empirical relationship estab-
lished by Kleyn et al. (1982). The entire procedure was repeated on a trial and error
basis until a uniform strength profile of CBR = 1 % was achieved all over the
designed test sections.

3.3 In-Place Density Assessment

The laboratory and field compaction characteristics of capping materials are shown in
Fig. 3. According to the laboratory test results, the maximum dry density of Type P
dolomite was approximately 17 % higher than that of Type O RAP. Similarly,
the optimum moisture content of dolomite was about 24 % higher than that of
Type O RAP. Moreover, Type O RAP had a minimal peak compared to Type P
dolomite with the change in moisture level. This was also reflected in achieved field
densities. For example, both Section O-M and O-N with RAP capping exhibited very
similar degrees of compaction; whereas, the remaining dolomite capped sections
showed very large discrepancy in achieved densities. Both the lowest and highest
degrees of compaction were recorded in the dolomite (Type P) capped sections.
Field Performance Evaluations of Large Sized Unconventional … 883

Fig. 3 Laboratory standard compaction characteristics and in place densities of capping materials
measured with nuclear gauge

3.4 In Situ Modulus Evaluation

In place stiffness of the compacted layers was measured with the use of lightweight
deflectometer (LWD) and soil stiffness gauge, aka Geogauge. Due to the difficulties
with rough surface on aggregate subgrade, only LWD reported modulus values of
aggregate subgrade are presented in Fig. 4. Type M riprap in Section O-M
exhibited the highest modulus response among all the sections. However, the same
section did not show any improvement upon the addition of dense graded capping
layer. Conversely, Section P-M, P-N and O-N exhibited 74, 46 and 30 % increase
in LWD modulus values due to the addition of thin capping layer, respectively. In a
recent study, Apeagyei and Hossain (2010) reported that excessive pore water

Fig. 4 Measured composite

surface moduli on aggregate
subgrade and capping layers
884 H. Kazmee et al.

pressure in soil could be generated under transient LWD loading and as a result,
misleadingly high modulus value would be observed if the pressure was locked
down in the soil (Apeagyei and Hossain 2010). In consideration to this, the high
modulus value in Section O-M can also be an indication of shallow water table.
Both LWD and Geogauge exhibited higher modulus values in the RAP capped
sections compared to their dolomite counterparts. Several other laboratory studies
investigated the resilient response of unbound granular materials with varying
percentage of RAP. Majority of those studies reported an increase in resilient
modulus with the increase in percentage of RAP. Henceforth, stiffness trends of
capping materials from the current study also conform to the finding of previous
literature. The lowest modulus was recorded in Type N (rubblized concrete)
aggregate subgrade. On the other hand, Section O-N had the highest modulus for
capping layers.

4 Rutting Performance Trends

Rutting is the governing mode of failure mechanism in unbound aggregate layers.

Henceforth, the magnitude of depression along the wheel path was periodically
monitored to investigate the performances of the large rock aggregate subgrade
materials. The stroke limit of ATLAS (76 mm or 3 in.) was deemed to be the
appropriate failure criterion in the construction platform sections. An average rut-
ting progression in the corresponding test sections is presented in Fig. 5. Despite
achieving higher relative density and in place modulus values, RAP capped sections
exhibited higher rutting potentials than those of the dolomite capped sections.
Several researchers have documented similar findings for RAP with respect to the
laboratory resilient modulus and permanent deformation trends (Garg and
Thompson 1996; Dong and Huang 2014; He 2006). Notably Edil et al. (2012)
investigated resilient modulus and permanent deformation trends of recycled

Fig. 5 Rutting
accumulations with number of
load applications for Sections
P-M through O-N
Field Performance Evaluations of Large Sized Unconventional … 885

concrete aggregates (RCA) and reclaimed asphalt pavement (RAP) materials from
eight different states. Consistent to the findings from this study, the researchers also
reported the highest magnitude of laboratory recorded plastic strains for the RAP
materials despite the exhibited high resilient modulus values when compared to
RCA and other virgin aggregate counterparts. Significantly low aggregate angu-
larity and surface texture might have been responsible for the reduced aggregate
interlock. Also, the viscous properties and creep susceptible nature of binder on
RAP particles might have contributed to higher accumulation of permanent
deformation under sustained repeated loading. As a result, the RAP capped sections
exhibited higher rutting amounts when subjected to accelerated pavement testing.
Section O-M constructed with riprap size aggregate subgrade and RAP capping
failed prematurely at only 80 passes. Similarly, the error bars on the average rutting
magnitudes of Section O-N also indicate failure at 1500 passes. Even though
Section P-M exhibited very similar slope of rutting progression for up to 150
passes, the accumulation stabilized after that when compared the rutting observed in
Section O-M.

5 Forensic Evaluations of Construction

Platforms

Rutting in unbound aggregate layers is greatly influenced by applied stress states

resulting from repeated wheel loads in comparison to material’s strength. Similarly,
depth of water table, thickness of granular layer, and aggregate shape properties can
also affect the rutting progression in unsurfaced pavements. In accordance, a
detailed forensic investigation was conducted in each of the test sections.
Transverse GPR scans were carried out before and after the accelerated pavement
testing to evaluate the relative contributions and modes of rutting in the underlying
layers. Strength profiles from penetration tests were also developed for the entire
depths of construction platforms using both a variable energy PANDA penetration
device and the dynamic cone penetrometer (DCP). Further, a geo-endoscope having
a lighted camera was lowered into the hole opened up by the PANDA penetration
test to visualize layer compositions and interfaces as well as the depths of water
table. Finally, a 1.5-m wide trench was excavated in each section to substantiate the
findings from transverse GPR scans and PANDA penetration testing. The ensuing
sections discuss the penetration testing based strength indices determined in the test
sections along with representative images of trenching in Section O-M.
886 H. Kazmee et al.

5.1 Variable Energy Penetration Testing

and Geo-Endoscopy

Figure 6 shows the strength indices with depth obtained from the penetration tests
in the constructed test sections. The gray colored solid horizontal lines designate the
average depths of layer interfaces determined from trenching of the test sections.
The depths of water table as identified from geo-endoscopic imaging are pinpointed
by the blue lines. The same figure also shows several of the geo-endoscopic images
recorded at different depths of the test sections. Cone resistances from variable
energy penetration testing are shown with the solid black lines; whereas, CBR
values from the DCP testing are presented with the dark grey dashed lines.
According to the strength depth profiles, significant improvement in subgrade
strength was observed with the utilization of aggregate subgrade materials.
Although the working platform sections were designed to have a subgrade strength
of CBR = 1 %, three out of the four sections exhibited 8–10 times higher than this
original very low strength. Note that this was the case even for fully saturated and
submerged conditions. Henceforth, it can be argued that some of the large rocks
penetrated deep into the weak subgrade and thus contributed to strength
improvement through individual particle contacts and the related load transfer
mechanism.

Fig. 6 Penetration test based strength indices in constructed test sections with geo-endoscopic
images
Field Performance Evaluations of Large Sized Unconventional … 887

As an important study finding, such penetration of the aggregate subgrade

material into weak soil proved to be quite beneficial. Accordingly, a quantifiable
extent of penetration and intermixing to be realized may be critical to successful
subgrade improvement remedial application. The variable energy penetration
PANDA testing would be more preferable to DCP for identifying such a zone of
intermixing since the preceding one exhibited gradual change in strength indices at
the layer interface. The geo-endoscopic imaging also indicated that a water pocket
might have been existed in Section O-M and section P-N. Despite having similar
depth of water table, Section P-N outperformed Section O-M because of the dis-
crepancy of strength in both capping and aggregate subgrade layers. As the Type N
crushed concrete had some intermediate and small size particles to fill up the voids
in aggregate skeleton, Section P-N exhibited higher strength because of that grain to
grain contact. The effect of voids in granular matrix was also evident from the
comparative assessment of strength indices in the capping layer and the aggregate
subgrade layer. For instance, the cone resistance in capping layer of Section P-N
was found to be in the range of 27–72 MPa. On the contrary, cone resistance in the
aggregate subgrade layer varied in between 3 and 109 MPa.

5.2 Subsurface Rutting Assessment with Ground

Penetrating Radar and Trenching

In addition to penetration testing, a proper diagnosis of the primary rutting mode

can provide more useful insight to establish the actual failure mechanism. Dawson
and Kolisoja (2006) documented three distinct rutting modes for unsurfaced
pavements or working platforms (Dawson and Kolisoja 2006). According to that
classification, mode ‘0’ refers to primary rutting in unsaturated granular materials
because of improper compaction. Failure mode ‘1’ is identified through the lateral
flow of unbound materials to the edge of wheel path alongside substantial amount
of heaving. Lastly, mode ‘2’ failure can be related to primary rutting in the subgrade
under repeated loads. This type of subgrade rutting can be associated with either
low shear strength or under-designed granular layers. Figure 7 shows the subsur-
face rutting trends revealed from trenching and transverse GPR scans in
Section O-M. The measured depths of granular layers over the western edge of the
trench is also presented in the same figure. The red, green and black lines denote the
surface, aggregate subgrade and subgrade interfaces respectively; meanwhile, the
dotted black lines stand for the tentative failure planes in the corresponding test
section.
Figure 7 indicates that the aggregate subgrade layer underwent bearing capacity
type failure with slippage planes intersecting at a certain angle in the subgrade.
Such a failure pattern can be construed as mode ‘2’ type rutting failure. Trench
image as illustrated in Fig. 7 clearly shows that lack of adequate aggregate interlock
in reclaimed asphalt pavement materials induced significant shearing in the capping
888 H. Kazmee et al.

Fig. 7 a Subsurface rutting observation from trenched section; b transverse GPR scan of
Section O-M

layer followed by notable densification of uniformly graded riprap. The resulting

failure further exacerbated due to the presence of shallow water table in the
subgrade.

6 Conclusions

This paper summarized accelerated pavement testing (APT) results obtained from
pavement construction platforms which utilized large sized unconventional virgin
and recycled aggregate materials. The full scale APT study consisted of twenty four
different test sections constructed at University of Illinois although the results from
only four test sections were presented in this paper. All of the test sections were
constructed over a weak subgrade with a controlled strength of CBR = 1 %.
Associated quality control tests were discussed in detail. Most notable contributions
to this APT study findings were obtained through the use of a very energy PANDA
penetration test device and its accompanying state of the art geo-endoscopic
imaging companion which was used to visualize large size aggregate penetrations
into weak subgrade and depth of water table during APT. Performances of the large
ballast or riprap size virgin crushed stone and recycled concrete aggregate materials
were assessed through the rutting accumulations in the constructed test sections
under unidirectional wheel loading. Rutting progression in the individual test sec-
tions was evaluated in the light of identified subsurface rutting modes from GPR
scans, penetration based strength indices, and trenching activities at the end of APT.
Based on the study findings, the following conclusions can be made:
Field Performance Evaluations of Large Sized Unconventional … 889

(a) Large size riprap and recycled concrete aggregate (RCA) materials were found
to be adequate for subgrade improvement applications provided that these
large rocks were capped with virgin crushed stones. Notably the in situ sub-
grade strength of approximately CBR = 10 % was significantly higher than
the design strength of CBR = 1 % indicating effectiveness of large rock
penetration into subgrade;
(b) Use of reclaimed asphalt pavement (RAP) as capping material should be
considered with caution since the RAP capped sections either accumulated
higher magnitude of rutting or failed at an earlier number of passes;
(c) Comparatively denser packing ensures better interlocking mechanism and
hence the somewhat well graded RCA outperformed uniformly graded riprap
virgin aggregate despite having lower morphological indices, i.e., aggregate
imaging based angularity index and surface texture index, and low magnitudes
of achieved density in the dolomite capped section.

References

Apeagyei, A. K., & Hossain, M. (2010). Stiffness-based evaluation of base and subgrade quality
using three portable devices. In Proceedings of the 89th transportation research board annual
meeting, Washington, DC.
Bureau of Bridges and Structures. (2005). Subgrade stability manual. Springfield, IL: Illinois
Department of Transportation.
Dawson, A., & Kolisoja, P. (2006). Managing rutting in low volume roads. In Rep. No. ROADEX III,
Swedish Road Administration, Northern Region, Luleå, Sweden.
Dong, Q., & Huang, B. (2014). Laboratory evaluation on resilient modulus and rate dependencies
of RAP used as unbound base material. Journal of Materials in Civil Engineering, 26(2),
379–383.
Edil, T., B, Tinjum, J., M, & Benson, C. H. (2012). Recycled unbound materials. In Rep. No. MN/
RC 2012-35, Minnesota Department of Transportation, St. Paul, MN.
Garg, N., & Thompson, M. R. (1996). Lincoln Avenue reclaimed asphalt pavement base project.
Transportation Research Record: Journal of the Transportation Research Board, 1547(1),
89–95.
He, W. (2006). Laboratory evaluation of unbound RAP as a pavement base material. Knoxville,
TN: Master of Science in Civil Engineering, University of Tennessee.
Kazmee, H., & Tutumluer, E. (2015). Evaluation of aggregate subgrade materials used as
pavement subgrade/granular subbase. In Rep. No. FHWA-ICT-15-013, Illinois Department of
Transportation, Springfield, IL.
Kleyn, E., Maree, J., & Savage, P. F. (1982). The application of portable pavement dynamic cone
penetrometer to determine in situ bearing properties of road pavement layers and subgrades in
South Africa. Transportation Research Record: Journal of the Transportation Research Board
(00390298).
Mathis, D. (1991). Rock cap: A true free draining base. In Proceedings of 42nd Annual Road
Builders’ Clinic (pp. 153–156). Coeur D’Alene, ID.
Mishra, D., & Tutumluer, E. (2013). Field performance evaluations of Illinois aggregates for
subgrade replacement and subbase—Phase II. In Rep. No. FHWA-ICT-12-021, Illinois
Department of Transportation, Springfield, IL.
890 H. Kazmee et al.

Tutumluer, E., Mishra, D., & Butt, A. (2009). Characterization of Illinois aggregates for subgrade
replacement and subbase. In Rep. No. FHWA-ICT-09-060, Illinois Department of
Transportation, Springfield, IL.
Uhlmeyer, J., Pierce, L., Lovejoy, J., Gribner, M., Mahoney, J., & Olson, G. (2003). Design and
construction of rock cap roadways: Case study in Northeast Washington State. Transportation
Research Record: Journal of the Transportation Research Board, 1821, 39–46.
USDA. (2009). Hydric soil distribution in Illinois. http://www.nrcs.usda.gov/Internet/FSE_
MEDIA/stelprdb1167160.jpg.
Waalkes, S. (2003). Cold weather and concrete pavements: Troubleshooting and tips to assure a
long-life pavement. In Proceedings of the transportation factor 2003: Annual conference and
exhibition of the transportation association of Canada (pp. 1–15), Newfoundland, Canada.
Structural Assessment of the Effect
of a Cement-Stabilized Base Combined
with a Cold Central-Plant Recycled Layer
at the NCAT Test Track

M.A. Díaz-Sánchez, D.H. Timm and B.K. Diefenderfer

Abstract Two full-scale pavement sections built in 2012 at the National Center for
Asphalt Technology (NCAT) Test Track were used to assess the effectiveness of
combining a cement-stabilized base and a cold central plant recycled (CCPR) layer,
from the perspective of the structural performance under accelerated traffic loading.
The two sections had similar asphalt concrete (AC) layers over CCPR while one
was built over an aggregate base and the other was built over a cement-stabilized
base resulting from a full-depth reclamation (FDR) process. The two test sections
performed adequately, presenting no evidence of pavement damage during a
two-year period (10 million ESALs) of accelerated traffic loading. Furthermore, ride
quality was not significantly affected with the application of traffic. Frequent
deflection testing over the duration of the study revealed that the section over the
cement-stabilized base exhibited much less temperature sensitivity and higher
moduli when compared to the section over the aggregate base. Similarly, based on
stress and strain measurements performed over the duration of the experiment, it
was inferred that the cement-stabilized base layer appears to cure over time, limiting
the strain levels and improving the structure over time.

M.A. Díaz-Sánchez (&)

Auburn University, Harbert Engineering Center, Auburn, AL 36849, USA
e-mail: [email protected]
D.H. Timm
Brasfield & Gorrie Professor of Civil Engineering, Auburn University, Harbert Engineering
Center, Auburn, AL 36849, USA
e-mail: [email protected]
B.K. Diefenderfer
Virginia Transportation Research Council, Charlottesville, VA 2290, USA
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 891

J.P. Aguiar-Moya et al. (eds.), The Roles of Accelerated Pavement Testing
in Pavement Sustainability, DOI 10.1007/978-3-319-42797-3_58
892 M.A. Díaz-Sánchez et al.

1 Introduction

In order to meet the needs of present road users without compromising future
generations, new and rehabilitated pavements must integrate innovative technolo-
gies and construction methods that take into account environmental, economic and
social indicators, targeting sustainable development. These many aspects lead to the
ultimate objectives of minimizing the consumption of non-renewable resources,
maximizing the reuse of existing materials, while generating a minimum of pol-
lutants and directly benefiting society, in the most cost-effective manner. Multiple
innovative pavement recycling and reclamation techniques, using common recy-
cling agents such as hydraulic cement and asphalt binder, meet these requirements.
The environmental, performance, and economic benefits of using reclaimed
materials have turned reclamation procedures into popular choices when building new
roads or rehabilitating distressed pavement structures. Cold central-plant recycling
(CCPR) and full-depth reclamation (FDR) have become popular solutions that allow
for reusing stockpiles of reclaimed asphalt pavement (RAP) and existing roadway
materials as enhanced base and subbase layers for a new pavement structure (Wirtgen
Group 2012). The most common recycling agents and additives include foamed
asphalt, emulsified asphalt, hydraulic cement, fly ash, and lime (ARRA 2001).
Cold central-plant recycling (CCPR) is a process whereby the existing asphalt
mixture is milled from the roadway and brought to a centrally located recycling plant
that incorporates recycling agents and additives into the material (Diefenderfer and
Apeagyei 2014). This approach allows for material removal from the roadway so that
the underlying foundation may be stabilized by a FDR process, or replaced as needed.
The milled RAP is subsequently reused as a recycled layer with very little added
virgin material. Furthermore, this process also allows for existing RAP stockpiles to
be used in new construction or rehabilitation projects (Diefenderfer and Apeagyei
2014). While there have been both laboratory and field investigations of CCPR
(Halles et al. 2013; Romanoschi et al. 2004; Loizos 2007; Jones et al. 2008; Fu et al.
2010), longer-term experiments under higher traffic volumes have been lacking for
CCPR specifically, and other recycled materials in general, and have limited its use.
To further investigate CCPR under higher traffic conditions, the Virginia Department
of Transportation (VDOT) began a field project in 2011 on Interstate-81 (I-81) to
evaluate CCPR, along with several other in-place recycling techniques (Diefenderfer
et al. 2012; Diefenderfer and Apeagyei 2014). Construction using CCPR proved
viable and the preliminary performance data through 34 months of service indicate
excellent performance (Diefenderfer and Apeagyei 2014). In conjunction with the
promising results obtained at the I-81 field project, VDOT developed a companion
study to evaluate the inclusion of a cement-stabilized base layer beneath the CCPR
layer. This companion study was meant to provide information regarding both the
CCPR and CCPR plus cement-stabilized materials under accelerated loading con-
ditions. To that end, VDOT sponsored the construction of two test sections at the
National Center for Asphalt Technology (NCAT) Pavement Test Track as part of the
2012 research cycle.
Structural Assessment of the Effect of a Cement-Stabilized Base … 893

2 Objectives and Scope of Work

The primary objectives of this investigation were to characterize the effect of a

cement-stabilized base layer on the field performance and structural characteristics
of a CCPR layer under accelerated traffic conditions at the NCAT Test Track.
Traffic was applied with triple-trailer vehicles and weekly performance measure-
ments and frequent falling weight deflectometer (FWD) testing were conducted.
The sections were also instrumented with strain gauges, pressure plates and tem-
perature probes to enable structural performance characterization.

3 Facility and Test Sections

The NCAT Test Track is a 1.7-mile closed loop full-scale flexible pavement test
facility located in Opelika, AL. Multiple two-hundred foot test sections are loaded
with approximately ten million equivalent single axle loads (ESALs) over a
two-year period. Frequent measurements of pavement response, deflection and
performance are completed during this period. Additionally, precise climate records
and traffic data, applied by a fleet of five tractor triple-trailers, are also kept during
the test cycle. Traffic operations for the fifth research cycle began in October 2012.
Two full-scale pavement sections containing a CCPR layer on top of a granular
base or a cement-stabilized base, respectively, were built as part of the fifth
reconstruction cycle at the Test Track during the summer of 2012. The two sections
included in this study were part of a larger study sponsored by VDOT in an effort to
evaluate the use of CCPR under accelerated traffic conditions. As shown in Fig. 1,
both sections consisted of approximately 4 in. of asphalt concrete (AC) over a 5 in.
thick CCPR layer. Each section featured a stone-matrix asphalt (SMA) surface layer
over one layer of a Superpave dense-graded mixture (SDG). The binder (Pb) and air
void (Va) contents of each AC layer are listed in Table 1. In the case of section N4,
the CCPR layer was constructed on top of a crushed granite base layer, while in
section S12 the layer beneath the CCPR consisted of a cement-stabilized base layer.
Both sections were built over the local subgrade found at the Test Track, previously
classified as an A-4 soil (Taylor and Timm 2009). According to the AASHTO
Classification system, an A-4 soil consists of a soil with over 35 % passing the
No. 200 sieve (0.075 mm), a liquid limit (LL) below 40 and a plasticity index
(PI) less than 10. The thicknesses for each layer were determined from survey data
at multiple random locations within each section. Therefore, the average as-built
thicknesses presented in Fig. 1 reflect the average thickness of each layer which
account for the natural inherent variation attributed to standard construction prac-
tices. Temperature probes were installed at an intermediate depth within the com-
bined AC and CCPR layers, as described by Timm (2009). Diefenderfer and Link
(2014) documented the laboratory characterization of the same CCPR material.
Extensive laboratory testing of the AC materials were also conducted but the results
894 M.A. Díaz-Sánchez et al.

Fig. 1 Pavement cross-sections

Table 1 As-built layer properties

Section N4 S12
Layer 1 description SMA: 19 mm NMAS with 12.5 % RAP and PG 76-22 binder
Binder content, % 6.0 6.1
Air voids, % 4.7 4.2
Layer 2 description Superpave: 19 mm NMAS with 30 % RAP and PG 67-22 binder
Binder content, % 4.6 4.7
Air voids, % 7.4 6.7
Layer 3 description CCPR: 100 % RAP with 2 % foamed PG 67-22 binder and 1 %
Type II hydraulic cement
Layer 4 description Granular base: Cement-stabilized base:
Crushed granite 6 in. CGAB + 2 in. subgrade stabilized
aggregate base in-place with 4 % Type II hydraulic cement
(CGAB)
Layer 5 description Track subgrade—AASHTO A-4 Soil

are beyond the scope of this investigation. The sections were opened to traffic on
October 23, 2012, and a program of regular FWD testing and performance moni-
toring began.

4 Performance Characterization

Rut depth and ride quality measurements were made with a high speed Automated
Road Analyzer (ARAN) vehicle, selected in an attempt to simulate the conditions
for true measurements on interstate pavements. Through the end of trafficking
(October 2014), no cracking was observed in either of the sections. Figure 2 shows
the evolution of rut depth over time. Both sections exhibited excellent performance
Structural Assessment of the Effect of a Cement-Stabilized Base … 895

Fig. 2 Rutting performance

over the research period, with maximum rut depths of 0.29 in. for section N4 and
0.23 for section S12, after approximately 10.05 million ESALs. Furthermore, very
little practical differences in performance were observed between the two sections.
The increasing rut depths noted between 3 and 4.5 million ESALs and between 8
and 9 million ESALs correspond to the increasing temperatures experienced during
the summer months of trafficking.
Pavement smoothness, expressed as the evolution of the International Roughness
Index (IRI) over the duration of the research period, is shown in Fig. 3. The data
indicates relatively little change in smoothness over time through October 2014.
Section S12, which included the cement-stabilized base layer, had an initial roughness
of nearly double the other section. This was caused by a localized low spot approx-
imately 40 to 60 feet into the section that was noted immediately after construction.
However, the IRI did not change appreciably over time, and S12 performed well in
terms of rutting and cracking. Overall, both sections maintained IRI values below the
maximum failure threshold of 170 in/mile recommended by the Federal Highway
Administration (FHWA 2015). Furthermore, these sections would be categorized as
“good” in terms of ride quality for an interstate pavement by VDOT (2014).

5 Structural Characterization

The structural characterization consisted of frequent FWD testing and weekly

monitoring of the strain gauges and pressure plates. The following subsections
describe each data collection activity and present the collected data.
896 M.A. Díaz-Sánchez et al.

Fig. 3 Ride quality

5.1 FWD Testing and Backcalculation

Each section was subjected to FWD testing, several times per month, in the inside
and outside wheelpaths and between the wheelpaths at four previously-selected
longitudinal stations. Testing was conducted using a Dynatest 8000 FWD with a
5.91-in radius split plate and nine deflection sensors, spaced at 0, 8, 12, 18, 24, 36,
48 and 72 in. from the load center. Although three replicates at four load levels
ranging from 6000 to 16,000 lbf were obtained through the research cycle, only the
data obtained for the second load level (approximately 9000 lb) were used for this
study. A corresponding mid-depth temperature was recorded during each FWD test.
Backcalculation was conducted using Evercalc® 5.0 and considered each sec-
tion as a three-layer structure, where the top layer was defined as the combination of
the AC and the CCPR, the aggregate base (N4) or cement-stabilized base (S12) was
considered as the second layer, and the subgrade represented the third layer.
Previous research had demonstrated that the CCPR would exhibit a behavior similar
to that of AC (Kim et al. 2009), so it seemed rational to combine the AC and CCPR
layers for the backcalculation process. Subsequent laboratory testing performed by
VDOT confirmed that the CCPR exhibited behavior consistent with AC materials
and supported the use of the combined AC/CCPR layer for backcalculation pur-
poses (Diefenderfer and Link 2014). Only backcalculated moduli corresponding to
a root mean square error (RMSE) of less than 3 % were considered for the analysis.
Figure 4 shows the influence of mid-depth temperature on backcalculated
AC/CCPR moduli. In both cases, the modulus clearly decreases with increasing
Structural Assessment of the Effect of a Cement-Stabilized Base … 897

temperature, indicating a direct influence of mid-depth temperature on the modulus.

The section on the unbound granular base layer (N4) indicates the strong influence
of mid-depth temperature on the modulus. The steep slope, exponential regression
equation and corresponding coefficients of determination (R2) observed for section
N4 are very similar to those reported previously for AC materials at the Test Track
(West et al. 2012; Timm et al. 2013; Vargas and Timm 2013). This behavior further
justifies considering the CCPR and AC materials as one combined layer for
backcalculation purposes. The section on the cement-stabilized base layer (S12)
shows a higher modulus and much less temperature sensitivity. The exponential
regression coefficient is less than one half the regression coefficient obtained for
section N4. It is believed that the increased modulus and lower temperature sen-
sitivity is an artifact of the backcalculation process whereby the AC/CCPR was
given a higher apparent modulus to adjust for smaller measured deflections on the
stabilized base section. Similarly, the wider range of backcalculated moduli values
observed for S12 may be attributed to backcalculation process and the relatively
small deflections obtained.
The backcalculated modulus values shown in Fig. 4 represent the FWD testing
results for the 9000 lb load level. However, to account for the variation of temperature
through the two-year testing period, the raw deflection data had to be temperature
corrected. For this, the modulus values were normalized to a reference temperature of
68 °F following previously-established procedures used in prior research at the Test
Track (Vargas and Timm 2013). Figure 5 shows the variation of normalized moduli
values over time. This may be considered as an indicator of possible pavement
damage over the duration of the study. At the same temperature, pavement damage
translates into a reduction of the backcalculated modulus over time. For the duration

Fig. 4 Backcalculated AC/CCPR modulus versus temperature

898 M.A. Díaz-Sánchez et al.

of this study, there was no evidence of pavement damage in the backcalculated data,
nor was any damage observed through routine pavement distress surveys.
A linear trendline was fit to the N4 data, with very little variation of the
AC/CCPR modulus and no pavement damage over time. The AC/CCPR modulus
for the stabilized base section (S12) clearly increased over time. This behavior is
believed to be the result of two different effects. On one hand, the backcalculation
process may be attributing certain cement-stabilized base properties to the
AC/CCPR layer. On the other hand, the cement stabilized layer may be curing over
time, as described by Diefenderfer et al. (2012), which in turn would result in
reduced pavement response measurements that will be explored in the next sub-
section. In any case, both sections appear to be healthy from a structural standpoint.

5.2 Pavement Response Measurements

The two sections were monitored once per week during the two year experiment,
measuring the stresses and strains induced by a minimum of three passes from each
of the five tractor triple-trailers on each section to account for natural wheel wander
and capture a direct hit over the gauge array. The data presented below represent the
“best-hit” on each section under the single axles on each date following
previously-established Test Track procedures (Timm et al. 2013). The mid-depth
pavement temperature at the time of data collection was also recorded.

Fig. 5 Backcalculated AC/CCPR modulus at 68 °F versus date

Structural Assessment of the Effect of a Cement-Stabilized Base … 899

Figure 6 shows the tensile strain response at the bottom of the CCPR layer
versus temperature. As expected, given the negative exponential relationship
between backcalculated modulus and temperature presented above, the strain
response was strongly correlated to temperature through exponential regression
equations. However, section S12 experienced relatively lower strains than section
N4 and demonstrated much less temperature sensitivity. The exponential regression
coefficient obtained for section S12 is approximately one half of that obtained for
section N4. Similarly, the strain magnitude is also significantly lower in S12 than
the other section. The differences are less pronounced at colder temperatures but
increase as temperatures increase. This is the combined effect of low temperature
sensitivity and the stiff base layer producing lower strain levels. The tensile strain is
also a function of the underlying supporting layer. In S12, this material is a
cement-stabilized base layer, while it is an unbound granular material in N4.
Therefore, it is reasonable to expect less strain in the section that has the stiffer
underlying material, which limits, to an extent, the tensile strain in the CCPR layer.
The measured tensile strain for section S12, with the cement-stabilized base
layer, is comparable to the strain obtained for perpetual pavements at the test track
(Willis and Timm 2009). Nonetheless, the continuation of traffic is needed before
making any final conclusion on the behavior of the pavement. Following the
temperature normalization procedure described previously for the backcalculated
moduli, the strain measurements were also corrected to a reference temperature of
68 °F. The normalized strains at the bottom of the CCPR layer over the duration of
the study are presented in Fig. 7. Linear trendlines have been assigned to each data
set to describe an overall trend. Section N4 shows an increase in strain over time.
The relatively higher slope observed for section N4 may indicate some damage may

Fig. 6 Tensile strain at the bottom of the CCPR layer versus temperature
900 M.A. Díaz-Sánchez et al.

Fig. 7 Tensile strain at 68 °F versus date/traffic

be occurring but has not yet been detected by FWD testing. On the contrary, the
smaller slope and lower R2 corresponding to section S12 indicate no appreciable
change over time and a healthy pavement structure.
The pressure at the top of the base and on top of the subgrade were also
measured as part of the study. Figure 8 shows the pressure at the top of the base
layer versus temperature, along with exponential regression equations for each
section. The measured base pressure for section S12, on the stabilized base, is
significantly lower than that obtained for section N4. Considering the temperature
normalization procedure described previously, Fig. 9 shows the base pressure
normalized to 68 °F versus time. Once again, the base pressure in section N4 seems
to be increasing over time, indicating some distress development that has not been
identified at the pavement surface. Conversely, as section S12 stiffens over time and
a corresponding net reduction in base pressure is evident. The subgrade pressure
measurements more clearly show the differences between the sections. Figure 10
illustrates subgrade pressure versus temperature. The subgrade pressure in section
S12 is significantly lower than that obtained for section N4. Similarly, section S12
seems to have a less significant temperature-dependency. The same
temperature-normalization procedure, described previously, was followed and the
results for the subgrade pressure are presented in Fig. 11. Section N4 shows very
little change in pressure over time while section S12, once again, trends toward
decreasing pressure as the section continues to stiffen over time as part of a curing
process.
Structural Assessment of the Effect of a Cement-Stabilized Base … 901

Fig. 8 Base pressure versus temperature

Fig. 9 Base pressure at 68 °F versus date/traffic

902 M.A. Díaz-Sánchez et al.

Fig. 10 Subgrade pressure versus temperature

Fig. 11 Subgrade pressure at 68 °F versus date/traffic

Structural Assessment of the Effect of a Cement-Stabilized Base … 903

6 Conclusions and Recommendations

This study was meant to investigate the effect of a cement-stabilized base layer on
the field performance and structural characteristics of a CCPR layer under accel-
erated traffic conditions at the NCAT Test Track. Based upon the data presented
above, the following conclusions and recommendations are made:
1. The two sections exhibited excellent performance over the duration of the study.
Although section S12 was originally built rougher than the other sections, the
change in smoothness over time in both sections was almost insignificant.
Cracking has not yet been observed in any of the sections. There were no
distinguishable surface-observable performance differences between the two
pavement sections.
2. The backcalculated AC/CCPR modulus for section S12, over the
cement-stabilized base, demonstrated much less temperature sensitivity and
higher values than that observed for section N4, over the granular base. It was
postulated that this behavior may be the result of two different effects; the
backcalculation process may be attributing certain cement-stabilized base
properties to the AC/CCPR layer, and the curing process of the
cement-stabilized layer. In any case, both sections appear to be healthy from a
structural standpoint.
3. Very little change in temperature-normalized modulus over time was observed
for section N4. Conversely, section S12 showed a relative increase in
temperature-normalized modulus over time, which was attributed to the curing
process of the cement-stabilized base layer. Nonetheless, the two sections did
not show signs of structural damage over time. Future investigations should
focus on laboratory evaluation of the cement stabilized material to determine its
curing characteristics.
4. The base pressure in section S12 was significantly lower than that observed for
section N4. In the case of section N4, normalized base pressure at 68 °F
increased over time. Conversely, the normalized base pressure at 68 °F
decreased over time for section S12, consistent with the stiffening of the section
due to the curing process of the cement-stabilized base layer. The
cement-stabilized base appears to improve the structure over time.
5. Subgrade pressure measurements captured differences between the two sections,
with the highest pressure reported in section N4 and the lowest in section S12.
Very little change in pressure at 68 °F over time was noted in section N4, while
section S12 again tended toward lower pressure over time.
6. Including the RAP contents in the asphalt mixtures, the entire pavement
structure of section S12 contained approximately 80 % recycled materials. The
testing conducted within this study demonstrates that new pavement structures
can be constructed using high recycled contents.
904 M.A. Díaz-Sánchez et al.

Acknowledgments This study was sponsored by the Virginia Department of Transportation.

Special recognition goes to Richard Schreck and Trenton Clark of the Virginia Asphalt
Association, David Stowell of Slurry Pavers, Marlin Hewitt from B&S Contractors and Richard
Ferron. Dr. Buzz Powell and Jason Nelson of NCAT deserve special recognition for their con-
tributions to this investigation.

References

ARRA. (2001). Basic asphalt recycling manual. Annapolis, MD: Asphalt Recycling and
Reclaiming Association.
Diefenderfer, B. K., & Apeagyei, A. K. (2014). I-81 in-place pavement recycling project. Report
no. 15-R1, Virginia Center for Transportation Innovation and Research, Virginia Department
of Transportation, Charlottesville.
Diefenderfer, B. K., Apeagyei, A. K., Gallo, A. A., Dougald, L. E., & Weaver, C. B. (2012).
In-place pavement recycling of I-81 in Virginia. In Transportation Research Record: Journal
of the Transportation Research Board, No. 2306 (pp. 21–27). Washington, DC: Transportation
Research Board of the National Academies.
Diefenderfer, B. K., & Link, S. D. (2014). Temperature and confinement effects on the stiffness of
a cold central-plant recycled mixture. In Proceedings of the 12th international society for
asphalt pavements conference on Asphalt pavements, Raleigh, NC.
Federal Highway Administration FHWA. (2015). Pavement smoothness. http://www.fhwa.dot.
gov/Pavement/smoothness/index.cfm. Accessed May, 2015.
Fu, P., Jones, D., Harvey, J. T., & Halles, F. (2010). Investigation of the curing mechanism of
foamed asphalt mixes based on micromechanics principles. Journal of Materials in Civil
Engineering, 22, 29–38.
Halles, F., Thenoux, G., & González, A. (2013). Stiffness evolution of granular materials stabilized
with foamed bitumen and cement. In Transportation Research Record: Journal of the
Transportation Research Board, No. 2363 (pp. 105–112). Transportation Research Board of
the National Academies, Washington, DC.
Jones, D., Fu, P., Harvey, J. T., & Halles, F. (2008). Full-depth reclamation with foamed asphalt:
Final report. UCPRC-RR-2008-07, University of California Pavement Research Center, Davis
and Berkeley.
Kim, Y., Lee, H. D., & Heitzman, M. (2009). Dynamic modulus and repeated load tests of
cold-in-place recycling mixtures using foamed asphalt. Journal of Materials in Civil
Engineering, 21(6), 279–285.
Loizos, A. (2007). In situ characterization of foamed bitumen treated mixtures for heavy-duty
pavements. International Journal of Pavement Engineering, 8(2), 123–135.
Romanoschi, S. A., Hussain, M., Gisi, A., & Heitzman, M. (2004). Accelerated pavement testing
evaluation of the structural contribution of full-depth reclamation material stabilized with foam
asphalt. In Transportation Research Record: Journal of the Transportation Research Board,
No. 1896 (pp. 199–207). Transportation Research Board of the National Academies,
Washington, DC.
Timm, D. H., Robbins, M. M., Willis, J. R., & Taylor, A. J. (2013). Field and laboratory study of
high-polymer mixtures at the NCAT test track: Final report. Report no. 13-03, National Center
for Asphalt Technology, Auburn University.
Vargas, A., & Timm, D. H. (2013). Physical and structural characterization of sustainable
asphalt pavement sections at the NCAT test track. Report no. 13-02, National Center for
Asphalt Technology, Auburn University.
Vargas-Nordcbeck, A., & Timm, D. H. (2013). Physical and structural characterization of
sustainable asphalt pavement sections at the NCAT test track. Report no. 13-02, National
Center for Asphalt Technology, Auburn University.
Structural Assessment of the Effect of a Cement-Stabilized Base … 905

Virginia Department of Transportation, Maintenance Division. (2014). State of the Pavement

Report 2014. Virginia: Richmond.
West, R., Timm, D., Willis, R., Powell, B., Tran, N., et al. (2012). Phase IV NCAT test track
findings. Report no. 12-10, National Center for Asphalt Technology, Auburn University.
Willis, J. R., & Timm, D. H. (2009). Field-based strain thresholds for flexible perpetual pavement
design. Report no. 09-09, National Center for Asphalt Technology, Auburn University.
Wirtgen Group. (2012). Wirtgen cold recycling manual (1st ed.). Windhagen: Wirtgen GmbH.