ABSTRACT

NOROUZI, AMIRHOSSEIN. Investigation of Specimen Geometries for the VECD Model

and Calibration of the LVECD Program for Fatigue Performance of Asphalt Pavements.
(Under the direction of Dr. Youngsoo Kim.)

Fatigue cracking is one of the complex distresses that is dependent of pavement

structure, asphalt mixture properties and environmental conditions. During the last decades,
many asphalt agencies have conducted significant researches to investigate the fatigue
cracking characterization. However, fatigue performance is still difficult to predict not only
due to models and parameters but also because this phenomenon itself is not well understood.
The key point in fatigue performance prediction is which model to use and how to
find the correct parameters for the selected model by using the simplest but the most reliable
testing method. The modulus is one of the primary asphalt mixture properties used for the
mechanistic performance prediction of asphalt pavements. Dynamic modulus testing is a
common method of measuring mixture modulus as a function of loading frequencies and
temperatures. Despite the numerous researches that have been carried out to evaluate mixture
stiffness, it is still necessary to establish a practical dynamic modulus test method that is
compatible with the field cores which are mostly less than a few inches. One of the
objectives of this dissertation is to present the results of a ruggedness study of dynamic
modulus testing in indirect tension mode to evaluate the factors that are most likely to affect
the final results. Specimen thickness, air void content, gauge length, test temperature, and
horizontal strain level, that are the critical factors that affect the dynamic modulus of asphalt
concrete, were selected for the ruggedness analysis. According to the findings, air void
content was found to be a major factor that affects the dynamic modulus values.
To investigate the fatigue life of the pavement, valid cyclic fatigue testing data which
truly represents the mixture behavior seems necessary besides the mixture stiffness. With
regard to direct tension fatigue testing, one of the common problems that substantially
influence the mixture fatigue behavior is the failure at the ends of the asphalt specimens.
During testing, it was observed that as more and more material was cut from the top and
bottom of the gyratory-compacted specimens, the likelihood of failure in the middle of the
specimen greatly increased. Therefore, fabricating shorter test specimens that are cored and
cut from taller gyratory-compacted specimens can produce test specimens that have more
uniformly distributed air voids such that middle failure occurs within the gauge length of the
linear variable differential transformer (LVDT) in direct tension tests. As a part of study, the
optimum specimen geometry of 100 mm diameter and 130 mm height was introduced
through the experimental testing and numerical simulation.
The Simplified Viscoelastic Continuum Damage (S-VECD) model, a continuum
damage mechanics-based model that is known as one of the effective models, has been
applied to predict the performance of asphalt concrete mixtures under different loading
conditions. Besides, energy-based fatigue failure criterion (GR) has been proved to be able to
predict the fatigue life of asphalt concrete mixtures across different modes of loading,
temperatures, and strain amplitudes. This dissertation presents the application and
calibration of the Layered ViscoElastic pavement analysis for Critical Distresses (LVECD)
program which is based on both S-VECD and GR method to evaluate 33 pavement sections
from different locations inside the United States, Canada, South Korea, and China. The
capability of the LVECD program to capture crack initiation, crack propagation and, the
damage in the pavement sections is investigated by comparing the simulation results with the
field observations. In this regard, LVECD was found to effectively predict the fatigue
cracking propagation in the pavement sections since reasonable agreement was obtained
between the program simulations and field observations. Finally, predicted damage-to-field
cracking transfer function was developed to correlate the predictive damage to the measured
cracking.
© Copyright 2015 by Amirhossein Norouzi
All Rights Reserved
Investigation of Specimen Geometries for the VECD Model and Calibration of the LVECD
Program for Fatigue Cracking Performance of Asphalt Pavements

by
Amirhossein Norouzi

A dissertation submitted to the Graduate Faculty of

North Carolina State University
in partial fulfillment of the
requirements for the degree of
Doctor of Philosophy

Civil Engineering

Raleigh, North Carolina

2015

APPROVED BY:

_______________________________ ____________________________________
Dr. Youngsoo Kim Dr. Akhtarhusein Tayebali
Distinguished University Professor Associate Professor
Committee Chair

________________________________ ____________________________________
Dr. Murthy N. Guddati Dr. Cassandra Castorena
Professor Assistant Professor
 DEDICATION

To the magnificence of God, to my father in Heaven, to my lovely mother and my

two sisters. Without their support and love, none of this would be possible. I am truly blessed.

ii
 BIOGRAPHY

Amirhossein was born as a second child of three children to Aliasghar and Masoumeh
Norouzi in Tehran, Iran on September 14, 1986. He attended Kowsar High school for
talented students and graduated in 2004. He entered Amirkabir University of Technology
(AUT) in 2004 and earned a Bachelor’s degree in Civil Engineering in 2008. Mr. Norouzi
studied Geotechnical Engineering as a Masters student at Khaje Nasir Toosi University of
Technology (KNTU) in 2008 under supervision of Prof. S.N. Moghaddas Tafreshi. He
focused on the experimental and numerical investigation of the effects of shredded rubber on
the bearing capacity of foundations. In August 2011, he enrolled in North Carolina State
University to pursue his Doctorate of Philosophy in Transportation Materials under the
guidance of Prof. Youngsoo Richard Kim.

iii
 ACKNOWLEDGMENTS

First of all, I would like to express heartfelt gratitude to my advisor Prof. Youngsoo
Richard Kim for his guidance, patience, motivation, encouragement, and immense
knowledge not only on my research but also on my life and mental growth. I would also like
to thank my former advisor, Prof. S. N. Moghaddas Tafreshi, for his directions, support and
friendship. I know that any success that I achieve in future will be a direct result of my
advisors’ recommendations and advices.
Secondly, I would like to thank all of the committee members, Prof. Murthy Guddati,
Dr. A. A. Tayebali, Dr. Cassandra Castorena, and Dr. Min Liu for their invaluable
recommendations. Special thanks for Dr. Shane Underwood and Dr. Andrew Lacroix who
helped me to quickly get involved into the big research group when I first came in. I would
like to specially thank my friend and colleague, Dr. Mohammadreza Sabouri for providing
the foundation of my research and also for the continuous scientific and technical assistant he
provided.
Also, I really feel grateful to work with Performance-Related Specification (PRS)
project team including Dr. Yeong Tae Choi, Dr. Jongsub Lee, Dr. Luis Alberto do
Nascimento, Dr. Andrew Wargo, Dr. Dahae Kim and Behrooz Keshavarzi. I also extend my
thanks to the current and former colleagues: Dr. Hong Joon Park, Dr. Junghyuk Im, Dr. Wei
Cao, Dr. Afshin Karshenas, Dr. Mehran Eslaminia, Dr. Arash Dehghan Banadaki, Dr. Javon
Adams, Dr. Amirshayan Safavizadeh, Dr. Mohammad Ilias, Seong hwan Cho, Mehdi
Mashayekhi, Yizhuang Wang, Farhad Yousefi Rad, Michael Elwardany, Nasrin Sumee,
Sonja Pape, Kangjin Lee, Mary Rawls, and Allen Piper. My former colleague, Diane
Gilmore, also deserves much appreciation. I also appreciate the help from my friends in
Mann Hall, Dr. S. Alireza Abbasian, Ashtad Javanmardi, Dr. Milad Hallaji, Dr. Farnam
Ghasemzadeh, and Amirhossein Mazrooei.
Last but not least, my work could not have been done without the financial support
from the FHWA DTFH61-08-H-00005 project, Hot Mix Asphalt Performance-Related
Specification Based on Viscoelasticity Continuum Damage (VEPCD) Models.

iv
 TABLE TENTS
E OF CONT

LES ..................................................................................................................... ix 

T OF TABL
LIST
RES ..................................................................................................................... x
T OF FIGUR
LIST

APTER 1 IN
CHA NTRODUCT TION .................................................................................................1 
1.1 Introductiion ........................................................................................................................ 1 
1.2 Objectivees .......................................................................................................................... 3 
1.3 Dissertatiion Organizaations ................................................................................................. 3

CHA
APTER 2 TH HEORETIC CAL BACK KGROUND ....................................................................5 
2.1 Linear Viiscoelastic Theory
T ................................................................................................ 5 
2.2 Viscoelasstic Continuuum Damage (VECD) Thheory.......................................................... 9 
2.3 Simplificaation of the VECD
V Mod del Formulatiion for Cycllic Loading .......................... 10 
2.4 Fatigue Failure Criterrion (GR App proach) ....................................................................... 12 
2.5 Layered Viscoelastic
V Critical Disttresses (LVE ECD) Prograam ....................................... 17

CHA
APTER 3 EX XPERIMEN NTAL PRO OGRAM .......................................................................20 
3.1 Mixture Sampling
S ocedure and Equipment ................................................................. 20 
Pro
3.2 Specimen n Fabricationn....................................................................................................... 21 
mples ................................................................................................ 21 
3.2.1 Cyliindrical Sam
3.2.2 Indiirect Tension
n Testing (D Disk-Shaped)) Samples ................................................ 22 
3.3 Determining Air Void ds for Perforrmance Testss .............................................................. 23 
3.4 Materials and Test Seections ............................................................................................. 24

CHA
APTER 4 RU UGGEDNE ESS STUDY Y OF DYNA AMIC MOD DULUS TES STING OF
ASPH HALT CON NCRETE IN N INDIREC CT TENSIO ON MODE ..........................31 
4.1 Abstract ............................................................................................................................ 31 
4.2 Introductiion ...................................................................................................................... 31 
4.3 Theoretical Backgrou und ................................................................................................... 33 
4.3.1 Lineear Viscoelaastic Solution n for IDT Sppecimens ................................................. 33 
4.3.2 Rugggedness Expperiments ........................................................................................ 36 
4.4 Laboratorries, Materiaals, and Equiipment ......................................................................... 37 
4.5 Factors th
hat Affect ID DT Dynamic Modulus Teesting ...................................................... 39 
mperature ............................................................................................................ 39 
4.5.1 Tem
4.5.2 Air Void Conten nt..................................................................................................... 40 
4.5.3 Horrizontal Straiin ..................................................................................................... 40 
4.5.4 Speccimen Thick kness................................................................................................ 40 
4.5.5 Gauuge Length .......................................................................................................... 41

v
 4.6 Results annd Discussio on .................................................................................................... 41
4.6.1 Temmperature ............................................................................................................ 45 
4.6.2 Air Voids ................................................................................................................. 45 
4.6.3 Gauuge Length .......................................................................................................... 49 
4.6.4 Straain Levels ........................................................................................................... 50 
4.6.5 Thicckness ................................................................................................................ 51 
4.7 Summary y and Conclu usions .............................................................................................. 51

CHA
APTER 5 DE ETERMINIING SPECIIMEN GEO OMETRY O OF CYLIND DRICAL
SPEC CIMENS FOR DIREC CT TENSIO ON FATIGU UE TESTIN NG OF
ASPH HALT CON NCRETE ........................................................................................53 
5.1 Abstract ............................................................................................................................ 53 
5.2 Introductiion ...................................................................................................................... 54 
5.3 Research Objectives and a Methodss .................................................................................. 56 
5.4 Theoretical Backgrou und ................................................................................................... 57 
5.5 Linear Viiscoelastic Prroperties ......................................................................................... 57 
5.6 Viscoelasstic Continuu um Damage Properties .................................................................. 57 
5.7 Failure Crriteria for Faatigue Perforrmance Preddiction ..................................................... 59 
5.8 Experimeental Program m ..................................................................................................... 60 
5.8.1 Matterials ................................................................................................................. 60 
5.8.2 Speccimen Fabrication.............................................................................................. 61 
5.8.3 LVDDT Gauge Length L .............................................................................................. 61 
5.8.4 Dynnamic Modulus Tests ......................................................................................... 61 
5.8.5 Direect Tension Tests T ............................................................................................... 62 
5.9 Determination of Speecimen Geom metry and LV VDT Gauge Length for L Linear
Viscoelasstic Propertiees ..................................................................................................... 62 
5.10 Viscoelaastic Continu uum Damagee Finite Elem ment Program m in C++ ............................ 65 
5.11 Determin nation of Sp pecimen Geo ometry and L LVDT Gauge Length forr Damage
Characterrization ............................................................................................................... 68 
5.12 Specimeen Geometry y and Failuree Location ................................................................... 70 
5.13 Investigaation of Imp proved Midd dle Failure foor the New S Specimen Geeometry Usinng
Various Mixtures
M ............................................................................................................. 71 
5.14 Conclusions .................................................................................................................... 74

CHA
APTER 6 MECHANIST
M TIC EVALU UATION O OF FATIGU UE CRACKIING IN
ASPH HALT PAV VEMENTS ....................
. .................................................................75 
6.1 Abstract ............................................................................................................................ 75 
6.2 Introductiion ...................................................................................................................... 76 
6.3 Test Methhods and Mo odels ................................................................................................ 77 
6.4 Linear Viiscoelastic Model M ............................................................................................... 78 
6.5 Simplifiedd Viscoelasttic Continuum m Damage ((S-VECD) M Model .................................. 79 
6.6 Layered Viscoelastic
V Critical Disttresses (LVE ECD) Pavem ment Analysiis Program ...... 81 
6.7 Materials .......................................................................................................................... 82

vi
 6.8 Test Resuults ...................................................................................................................... 86
6.8.1 Dynnamic Modulus Tests ......................................................................................... 86 
6.8.2 S-V
VECD Modell Characterizzation Curvees ............................................................. 89 
6.8.3 Failure Criteria ........................................................................................................ 93 
6.9 LVECD Model
M Simulation Results ................................................................................ 96 
6.10 Summarry and Concllusions .......................................................................................... 104

CHA
APTER 7 NU UMERICAL L EVALUA ATION OF P PAVEMEN NT DESIGN N
PARRAMETERS S FOR THE E FATIGUE E CRACKIN NG AND R RUTTING
PERRFORMANC CE OF ASP PHALT PAV VEMENT .............................................105 
7.1 Abstract .......................................................................................................................... 105 
7.2 Introductiion .................................................................................................................... 106 
7.3 Test Protoocols................................................................................................................. 108 
7.3.1 Dynnamic Modulus Testing.................................................................................... 108 
7.3.2 Cycclic Testing Using U the S-VECD Moddel .......................................................... 109 
7.3.3 Fatigue Failure Criterion (G GR Method) ................................................................ 110 
7.3.4 Permmanent Defo ormation (TS SS testing) ................................................................. 111 
7.4 Materials and Test Methods ........................................................................................... 112 
7.4.1 Speccimen Fabrication............................................................................................ 113 
7.5 Test Resuults and Disccussion........................................................................................... 116 
7.5.1 Dynnamic modullus Values ..................................................................................... 116 
7.5.2 S-V
VECD Characcterization Curves C ....................................................................... 117 
7.5.3 Fatigue Failure Criterion Lines ............................................................................ 118 
7.5.4 Triaaxial Stress SweepS Tests ................................................................................. 119 
7.6 Pavementt Analysis ......................................................................................................... 121 
7.6.1 Fatigue Perform mance ............................................................................................. 121 
7.6.2 Ruttting Perform mance ............................................................................................. 125 
7.7 Conclusio ons .................................................................................................................... 128

CHA
APTER 8 AP PPLICATIO ON OF VIS SCOELAST TIC CONTIINUUM DA AMAGE
APPROACH TO O PREDICT T THE FAT TIGUE CRA RACKING
PERRFORMANC CE OF BIN NZHOU PER RPETUAL PAVEMEN NTS ................130 
8.1 Abstract .......................................................................................................................... 130 
8.2 Introductiion .................................................................................................................... 131 
8.3 Overvieww of Binzhou u Test Sectio ons ............................................................................. 133 
8.4 Material Modeling
M ......................................................................................................... 135 
8.4.1 Lineear Viscoelaastic Continu uum Damagee Theory ................................................ 135 
8.4.2 Failure Criterion n .................................................................................................... 137 
8.5 Experimeental Charactterization....................................................................................... 140 
8.5.1 Dynnamic Modulus Test ......................................................................................... 140 
8.5.2 Direect Tension Cyclic C Fatig gue Test ..................................................................... 142 
8.6 Fatigue Performance Predictions ....................
. .............................................................. 146 
8.6.1 The Finite Elem ment Packagee - LVECD ................................................................ 146

vii
 8.6.2 Sim
mulation of Faatigue Perfo ormance and Discussionss ........................................ 147
8.7 Conclusioons .................................................................................................................... 152

CHA
APTER 9 CA ALIBRATION OF FAT TIGUE CR RACKING M MODEL FO OR LVECD D
PRO OGRAM ............................................................................................................153 
9.1 Introductiion .................................................................................................................... 153 
9.2 Fatigue Cracking
C Pred diction Mod dels ............................................................................ 155 
9.2.1 Trad ditional Tenssile Strain-B Based Fatiguue Cracking M Model ............................... 155 
9.2.2 Energy-Based Fatigue F Crack king Model ............................................................... 156 
9.3 Calibratioon of Energy y-Based Fatig gue Criterion ........................................................... 156 
9.4 Field Craccking Data Collection
C ..................................................................................... 157 
9.4.1 Lonng Term Paveement Performance (LT TPP) ....................................................... 158 
9.4.2 Natiional Centerr for Asphaltt Technologyy (NCAT) ............................................. 158 
9.4.3 Man nitoba Infrasstructure and d Transportattion (MIT) ............................................ 159 
9.4.4 Fedeeral Highwaay Administrration Acceleerated Loadiing Facility (FHWA-AL LF)
.............................................................................................................................. 159 
9.4.5 Korrea Expressw way Corporattion (KEC) ............................................................... 160 
9.5 LVECD Program
P Simmulation Process........................................................................... 160 
9.5.1 Streess-Strain Caalculation Method in LV VECD Prograam ..................................... 160 
9.5.2 Dam mage Calculaation in the LVECD L Proogram ..................................................... 161 
9.6 Analysis ofo Measured d Fatigue Craacking in LT TPP Databasse ....................................... 164 
9.7 Fatigue Cracking
C Tran nsfer Function ............................................................................. 166

CHA
APTER 10 COMPARIS
C SON OF FA ATIGUE CR RACKING PERFORM MANCE
PRE
EDICTIONS S IN ASPHA ALT PAVEMENTS US SING PAVE EMENT ME E
ANDD LVECD PROGRAM P S ................................................................................183 
10.1 Abstractt ........................................................................................................................ 183 
10.2 Introducction .................................................................................................................. 184 
10.3 Backgroound .................................................................................................................. 186 
10.4 Data Collection and Field Sectio ons ............................................................................. 190 
10.5 Results and
a Discussiion ................................................................................................ 196 
10.6 Conclusions .................................................................................................................. 204

CHA
APTER 11 CONCLUSI
C IONS AND FUTURE W WORK ..................................................206 
11.1 Conclusions .................................................................................................................. 206 
11.2 Future Work
W ................................................................................................................. 209

ERENCES..........................................................................................................................210 
REFE

viii
 LIST OF TABLES

Table 3-1. Volumetric properties of mixtures used for this research proposal. ...................... 30 
Table 4-1 Factors and Levels of Ruggedness Analysis ......................................................... 37 
Table 4-2 Experimental Design for Ruggedness Testing ...................................................... 37 
Table 4-3 Mixture Properties ................................................................................................. 38 
Table 4-4 Coefficients for Calculating the Dynamic Modulus .............................................. 38 
Table 4-5. Summary of F-values for NE Mixture for Both Laboratories and All Factors. .... 43 
Table 4-6. Summary of F-values for VA Mixture for Both Laboratories and All Factors. .... 44 
Table 5-1 One-way ANOVA Results for Dynamic Modulus Tests at 5% Significance Level
......................................................................................................................................... 64 
Table 5-2 One-way ANOVA Results for Direct Tension Cyclic Tests at 5% Significance... 70 
Table 5-3 Experimental Verification of New Specimen Geometry for Improved Middle
Failure ............................................................................................................................. 73 
Table 6-1. Volumetric Properties of Different Mixtures Used in the Study. .......................... 85 
Table 7-1. Volumetric Properties of the KEC Mixtures ....................................................... 113 
Table 8-1. Asphalt concrete materials used in the Binzhou test sections. ............................ 134 
Table 10-1. Input Information for LVECD and Pavement ME Programs for Performance
Analysis of NCAT and ALF Test Sections................................................................... 195 

ix
 LIST OF FIGURES

Figure 2-1. |E*| values of different frequencies and temperatures............................................ 7 

Figure 2-2. Example of a temperature shift factor curve. ......................................................... 8 
Figure 2-3. Time-temperature shifted |E*| master curve. ......................................................... 8 
Figure 2-4. Schematic representation of total released pseudo strain energy WRC in the stress-
pseudo strain space. ........................................................................................................ 13 
Figure 2-5. History of WRC and its corresponding rate. ......................................................... 14 
Figure 2-6. Relationship between GR and Nf for mix S9.5C .................................................. 15 
Figure 2-7. WRC vs. N for a CX and a COS case with similar Nf. ......................................... 16 
Figure 2-8. GR vs. Nf for CS, COS and CX modes for VTe30LC mixtures. .......................... 16 
Figure 3-1. Loose mix sampling schematic. ........................................................................... 21 
Figure 3-2. Specimen Fabrication for uniform air voids: (a) coring; and (b) cutting. ............ 22 
Figure 3-3. Asphalt pavement structure for the FHWA-ALF project. ................................... 25 
Figure 3-4. Asphalt pavement structure for the NCAT project. ............................................. 26 
Figure 3-5. MIT pavement sections with thickness (mm) (a) WMA, (b) RAP. ..................... 27 
Figure 3-6. Asphalt pavement sections for the KEC project. ................................................. 28 
Figure 3-7. Binzhou pavement structure layout for each section. .......................................... 29 
Figure 4-1. Dynamic modulus mastercurves obtained from axial compression and IDT tests:
a) semi-log space and b) log-log space. .......................................................................... 42 
Figure 4-2. Comparison of dynamic modulus mastercurves with 6% and 8% air void
contents: (a) Case 1 and Case 2, (b) Case 5 and Case 6, (c) Case 3 and Case 4, and (d)
Case 7 and Case 8. .......................................................................................................... 47 
Figure 4-3. Variation of dynamic modulus values versus air void content. ........................... 48 
Figure 4-4. Comparison of dynamic modulus mastercurves with 50.8 mm and 101.6 mm
gauge lengths: (a) Case 1 and Case 7, (b) Case 3 and Case 5, (c) Case 2 and Case 6, and
(d) Case 4 and Case 8...................................................................................................... 49 
Figure 4-5. Comparison of dynamic modulus mastercurves with strain levels of 40  and 60
: (a) Case 1 and Case 7, (b) Case 3 and Case 5, (c) Case 4 and Case 6, and (d) Case 2
and Case 8. ...................................................................................................................... 50 
Figure 5-1. Experimentally measured dynamic modulus mastercurves: (a) in semi-log scale,
(b) in log-log scale, and (c) phase angle mastercurves. .................................................. 63 
Figure 5-2. (a) Vertical strain contour (Ezz), (b) vertical stress contour (Szz), and (c) stress
and strain measurements to simulate laboratory tests. .................................................... 66 

x
Figure 5-3. Simulation results of the dynamic modulus tests: (a) in semi-log scale, (b) in log-
log scale, (c) pseudo stiffness and damage, and (d) damage characteristic curves......... 68 
Figure 5-4. Damage characteristic curves for different specimen geometries from: (a) VECD-
FEP++ simulations and (b) experimental measurements................................................ 69 
Figure 5-5. Effect of end failure on two failure criteria: (a) Cf based on peak phase angle and
(b) Gᴿ vs. Nf. ................................................................................................................... 70 
Figure 5-6. Damage propagation contours of potential cracking (pseudo stiffness, C) for: (a)
75 mm x 150 mm, (b) 75 mm x 130 mm, (c) 100 mm x 150 mm, and (d) 100 mm x 130
mm specimen geometries. ............................................................................................... 72 
Figure 6-1. Asphalt concrete pavement structure for (a) FHWA-ALF, (b) NCAT, (c) MIT-
WMA, and (d) MIT-RAP mixtures. ............................................................................... 84 
Figure 6-2. Average dynamic modulus test results: (a) FHWA-ALF, (b) NCAT surface layer,
(c) NCAT intermediate layer, (d) NCAT base layer, (e) MIT-WMA surface layer, (f)
MIT-WMA bottom layer, and (g) MIT-RAP mixtures................................................... 88 
Figure 6-3. Damage characteristic curves for the FHWA mixtures. ...................................... 90 
Figure 6-4. Damage characteristic curves for the FHWA mixtures: (a) surface layer, (b)
intermediate layer, and (c) bottom layer. ........................................................................ 91 
Figure 6-5. Damage characteristic curves for the MIT mixtures: (a) WMA surface layer, (b)
WMA bottom layer, and (c) MIT-RAP. ......................................................................... 92 
Figure 6-6. Failure criterion envelopes (GR vs. Nf) for: (a) FHWA-ALF, (b) NCAT surface
layer, (c) NCAT intermediate layer, (d) NCAT base layer, (e) MIT-WMA surface layer,
(f) MIT-WMA bottom layer, and (g) MIT-RAP mixes. ................................................. 95 
Figure 6-7. Five-year simulation results obtained from the LVECD program for the FHWA
mixtures: (a) control, (b) terpolymer, (c) SBS, and (d) crumb rubber. ........................... 98 
Figure 6-8. Five-year simulation results obtained from the LVECD for the NCAT mixtures:
(a) NCAT-Control, b) NCAT-AW, (c) NCAT-FW, (d) NCAT-RW, (e) NCAT-R, and
(f) NCAT-O. ................................................................................................................... 99 
Figure 6-9. Five-year simulation results obtained from the LVECD program for the MIT-
WMA mixtures: (a) control, (b) Advera, (c) Evotherm, and (d) Sasobit. ..................... 100 
Figure 6-10. Five-year simulation results obtained from the LVECD program for the MIT-
RAP mixtures: (a) control, (b) 15R, (c) 50R, and (d) 50RSB....................................... 100 
Figure 6-11. (a) LVECD analysis results of ALF pavements, (b) field measurements from
ALF pavements, (c) LVECD analysis results of NCAT pavements, (d) field
measurements from NCAT pavements, (e) LVECD analysis results of MIT-WMA
pavements, (f) field measurements from MIT-WMA pavements, (g) LVECD analysis
results of MIT-RAP pavements, and (h) field measurements from MIT-RAP pavements.
....................................................................................................................................... 101

xi
Figure 7-1. Asphalt pavement sections at the KEC test road. .............................................. 115 
Figure 7-2. Dynamic modulus mastercurves for KEC test road mixtures: (a) in semi-log
space and (b) in log-log space. ...................................................................................... 116 
Figure 7-3. Averaged damage characteristic curves for KEC test road mixtures. ................ 117 
Figure 7-4. Fatigue failure criterion lines for the KEC mixture. .......................................... 118 
Figure 7-5. TSS test results of KEC mixtures: (a) ASTM, (b) PMA, (c) BB1, (d) BB3, and (e)
BB5. .............................................................................................................................. 120 
Figure 7-6. Effects of different parameters on fatigue cracking performance as obtained from
LVECD simulations: (a) base layer material, (b) base layer thickness, (c) subgrade
material, and (d) sub-base thickness. ............................................................................ 123 
Figure7-7. Effects of different parameters on pavement fatigue cracking in the field: (a) base
layer type, (b) base layer thickness, (c) subgrade material, and (d) sub-base thickness.
....................................................................................................................................... 125 
Figure 7-8. Effects of different parameters on rutting performance: (a) surface layer type, (b)
base layer type, (c) subgrade type, and (d) sub-base thickness..................................... 127 
Figure 8-1. Structure layout of test sections in Binzhou perpetual pavement project. ......... 134 
Figure 8-2. Schematic representation of the characteristic dissipated pseudo strain energy.138 
Figure 8-3. Dynamic modulus test results for the asphalt mixes in Binzhou sections. ........ 142 
Figure 8-4. Damage characteristic curves for the asphalt mixes in Binzhou sections. ......... 144 
Figure 8-5. Relations between ACDPSE and fatigue life for each mix in Binzhou sections.
....................................................................................................................................... 145 
Figure 8-6. C-contour of the Binzhou sections after 15 year simulations. ........................... 149 
Figure 8-7. The evolution of failure percentages for Binzhou sections after 15 year
simulations. ................................................................................................................... 151 
Figure 9-1. NCAT test track section area. ............................................................................ 159 
Figure 9-2. LTPP alligator cracking vs. asphalt layer thickness........................................... 164 
Figure 9-3. (a) Alligator cracking vs. MAAT ranges and (b) maximum alligator cracking. 165 
Figure 9-4. Schematic definition of damage area in LVECD. .............................................. 167 
Figure 9-5. Damage definition obtained from method 1 for FHWA-ALF: (a) top half of
asphalt layer, (b) bottom half of asphalt layer, and (c) field measurements. ................ 167 
Figure 9-6. Damage definition obtained from method 2 for FHWA-ALF: (a) top half of
asphalt layer, (b) bottom half of asphalt layer, and (c) field measurements. ................ 168 
Figure 9-7. Damage definition obtained from method 1 for NCAT: (a) top half of asphalt
layer, (b) bottom half of asphalt layer, and (c) field measurements. ............................ 169 

xii
Figure 9-8. Damage definition obtained from method 2 for NCAT: (a) top half of asphalt
layer, (b) bottom half of asphalt layer, and (c) field measurements. ............................ 170 
Figure 9-9. Damage definition obtained from method 1 for MIT-RAP: (a) top half of asphalt
layer, (b) bottom half of asphalt layer, and (c) field measurements. ............................ 171 
Figure 9-10. Damage definition obtained from method 2 for MIT-RAP: (a) top half of asphalt
layer, (b) bottom half of asphalt layer, and (c) field measurements. ............................ 172 
Figure 9-11. Damage definition obtained from method 1 for MIT-WMA: (a) top half of
asphalt layer, (b) bottom half of asphalt layer, and (c) field measurements. ................ 173 
Figure 9-12. Damage definition obtained from method 2 for MIT-WMA: (a) top half of
asphalt layer, (b) bottom half of asphalt layer, and (c) field measurements. ................ 174 
Figure 9-13. Predicted damage vs. measured fatigue cracking: (a) Pavement ME design
guide and (b) LVECD program. ................................................................................... 177 
Figure 9-14. Measured field cracking vs. LVECD program-predicted cracking at different
pavement thicknesses: (a) 10 cm, (b) 16 cm, (c) 20 cm, and (d) 30 cm. ...................... 178 
Figure 9-15. Plot of correlation between ((100/FC)-1) and (2.5D/100) for different pavement
sections. ......................................................................................................................... 179 
Figure 9-16. Variation of absolute values of  vs. pavement thickness............................... 180 
Figure 9-17. Comparison of observed versus predicted cracking with the LVECD transfer
function. ........................................................................................................................ 181 
Figure 9-18. (a) non-calibrated damage trends for MIT-WMA, (b) measured cracks for MIT-
WMA, (c) calibrated predicted damage for MIT-WMA, (d) non-calibrated damage
trends for NCAT, (e) measured cracks for NCAT, (f) calibrated predicted damage for
NCAT, (g) non-calibrated damage trends for MIT-RAP, (h) measured cracks for MIT-
RAP, and (i) calibrated predicted damage for MIT-RAP. ............................................ 182 
Figure 10-1. Schematics of asphalt pavement structures for the test sections: (a) NCAT
(asphalt layers), (b) FHWA-ALF (asphalt layers), (c) MIT-WMA (asphalt layers), (d)
MIT-RAP (asphalt layers), and (e) KEC (whole pavement structure). ........................ 194 
Figure 10-2. Measured and predicted fatigue cracking amount for NCAT and FHWA test
sections: (a) amount of cracking in the field (NCAT), (b) predicted damage area in
LVECD (NCAT), (c) predicted damage area in Pavement ME (NCAT), (d) maximum
amount of cracking at the end period of study (NCAT), (e) maximum predicted damage
area at the end of analysis with LVECD (NCAT),(f) maximum predicted damage area at
the end of analysis with Pavement ME (NCAT), (g) amount of cracking in the field
(FHWA), (h) predicted damage area in LVECD (FHWA), (i) predicted damage area in
Pavement ME (FHWA), (j) maximum amount of cracking at the end period of study
(FHWA), (k) maximum predicted damage area at the end of analysis with LVECD
(FHWA), and (l) maximum predicted damage area at the end of analysis with Pavement
ME (FHWA). ................................................................................................................ 198 

xiii
Figure 10-3. Damage factor (N/Nf) distribution of NCAT Test Track OGFC section at the
end of analysis period. .................................................................................................. 199 
Figure 10-4. Measured and predicted fatigue cracking amount for MIT-WMA and MIT-RAP
test sections: (a) amount of cracking in the field (MIT-W), (b) predicted damage area in
LVECD (MIT-W), (c) predicted damage area in Pavement ME (MIT-W), (d) maximum
amount of cracking at the end period of study (MIT-W), (e) maximum predicted damage
area at the end of analysis with LVECD (MIT-W),(f) maximum predicted damage area
at the end of analysis with Pavement ME (MIT-W), (g) amount of cracking in the field
(MIT-R), (h) predicted damage area in LVECD (MIT-R), (i) predicted damage area in
Pavement ME (MIT-R), (j) maximum amount of cracking at the end period of study
(MIT-R), (k) maximum predicted damage area at the end of analysis with LVECD
(MIT-R), and (l) maximum predicted damage area at the end of analysis with Pavement
ME (MIT-R).................................................................................................................. 202 
Figure 10-5. Effects of different parameters of fatigue cracking performance for KEC test
sections: (a)-(c) base layer material, (d)-(f) base layer thickness, (h)-(j) application of
anti-frost layer, and (k)-(m) sub-base thickness............................................................ 203 
Figure 10-6. Overall trends of predicted fatigue cracking vs. field measurements: (a) LVECD
program predictions and (b) Pavement ME program predictions. ................................ 204 

xiv
CHA
APTER 1 INTR
RODUCT
TION

1.1 Introducttion
Fatigue performance
p modeling iss one the inteeresting topiics in asphallt concrete
modeeling work. Fatigue
F crack
king due to repeated
r trafffic loading is one of thee major typess of
distreess in asphallt concrete paavements in North Amerrica. This foorm of distress results froom
the ap
pplication off repeated traaffic loading
g which causses failure inn flexible pavvements.
Depeending on thee failure mecchanisms and the locatioon of crack innitiation, paavement struccture
can be
b categorizeed into two major
m phasess: pre-localizzation and poost-localizattion. In general,
pre-lo
ocalization is manifested
d by the initiiation and prropagation oof micro-craccks, and postt-
failurre representss the formation and propagation of m
macro-crackss. Propagatioon of fatiguee
crack
ks in the asph
halt layer wiill eventually
y allow wateer to come innto the pavem
ment system
m, and
causee deterioratio
on of pavement structuree which reduuces its serviiceability. Despite signifficant
reseaarches on thee fatigue craccking topic, it has not beeen well undderstood becaause of the
comp
plexity of thee phenomeno
on. Thereforre, one of thee major conccerns amongg state highw
way
agenccies is which
h model to utilize
u to pred
dict the paveement crackiing.
The dynam
mic modulus is one of th
he most important asphaalt propertiess that has beeen
used to predict th
he amount off distresses, such
s as fatiggue crackingg and rutting,, in a pavem
ment
(Noro
ouzi and Kim
m 2015, Norrouzi and Sab
bouri 2014, Sabouri andd Kim 2014). During reccent
des, significaant research has been conducted to eevaluate mixxture stiffnesss (Majidzaddeh et
decad
al. 19
979, Witczak
k et al. 1996,, Shook et all. 1969); how
wever, it is sstill necessarry to establissh a
practical dynamicc modulus teest method th
hat is compaatible with thhe use of fielld cores (Kuutay
et al. 2009, Park et al. 2014, Kim
K and Kim
m 2015, Leee et al. 2015)). One of thee concerns
assocciated with th
he current dy
ynamic mod
dulus test prootocol, AASH
HTO TP 79,, is its
applicability to fo
orensic studiies and pavement rehabiilitation desiggn. This prootocol requirees
axial compression
n testing of asphalt
a conccrete specimeens that are 100 mm in ddiameter andd 150
mm in
i height. Su
uch specimen
ns are often impossible
i tto obtain from
m actual pavvements beccause
typical asphalt lay
yers are lesss than a few inches thickk. Therefore, indirect tennsion (IDT)
testin
ng of field co
ores seems to
o be approprriate for evalluating existting pavemennts, even thoough

1
controversial differences between axial compression test methods and IDT test methods
remain.
Besides material stiffness, a reliable test procedure must be adopted to determine the
damage properties of asphalt mixtures. Cyclic direct tension is one of the common test
procedures to identify the fatigue behavior of asphalt mixture; however a recent experimental
problem was encountered that involved failure at the ends of the asphalt specimens
(hereinafter called end failure). During testing, it was observed that as more and more
material was cut from the top and bottom of the gyratory-compacted specimens, the
likelihood of failure in the middle of the specimen (hereinafter called middle failure) greatly
increased. This outcome was due to the air void gradient along the diameter and height of the
gyratory-compacted asphalt mixture specimens. A study (Chehab et al. 2000) found that the
air void content is high near the mold wall and top and bottom of the specimen, and the
average difference in the air void content between the cored and cut specimens that are 75
mm x 150 mm (75 mm diameter and 150 mm height) and the gyratory-compacted specimens
that are 150 mm x 175 mm is about 2.5 percent. The uniformity of the air void distribution
throughout the test specimens is a much more critical factor in direct tension tests than in
axial compression tests. Higher air void contents at the ends of the specimen (due to the
removal of the ends of gyratory-compacted specimens of insufficient length) cause end
failure in direct tension tests, whereas in axial compression tests they may cause more
deformation, but not catastrophic failure. Therefore, fabricating shorter test specimens that
are cored and cut from taller gyratory-compacted specimens can produce test specimens that
have more uniformly distributed air voids such that middle failure occurs within the gauge
length of the linear variable differential transformer (LVDT) in direct tension tests.
To properly understand and model the fatigue cracking over the range of conditions
encountered in the field without performing a large number of experiments, it is essential to
employ mechanistic models for material characterization and pavement analysis. These
models should describe the material and pavement responses to repeated loading, particularly
the process of cracking. Viscoelastic continuum damage (VECD) theory has been used
successfully for asphalt mixture characterization in the United States under different modes

2
of loading and temperatures. Researchers at North Carolina State University applied the
VECD model into the Layered ViscoElastic pavement analysis for Critical Distresses
(LVECD) program to design the pavement structures. This tool has shown potential to be used
as a reliable performance prediction approach and serves as the basis for the Federal Highway
Administration’s (FHWA’s) Performance-Related Specifications for Hot-Mix Asphalt (HMA-
PRS). However, this tool is not verified for various conditions and the transfer function that
relates the predicted damage to the field cracking has not been developed yet.

1.2 Objectives
 Investigation of effective parameters on the dynamic modulus in indirect tension test
(IDT), and development of specification for characterizing the field cores using the
IDT testing.
 Geometry study for the S-VECD cyclic direct tension test to find the optimum
geometry to satisfy the specimen failure in the middle without the change in damage
parameters of the mixture.
 Implementation of the VECD model for the characterization of Performance Related
Specification (PRS) project and validate the modeling approach for mixtures
composed of a wide range of asphalt binders, gradation, and warm mix additives.
 Simulation of the fatigue performance of PRS asphalt pavement test sections under
real traffic loading using the S-VECD cyclic testing and the LVECD pavement
analysis tool.
 Validation of the LVECD program for pavement fatigue performance analysis, and
development of predicted damage to field cracking transfer function to calibrate the
LVECD design tool using the PRS field cracking data.

1.3 Dissertation Organizations

This dissertation consists of 10 chapters. Chapter 1 presents the introduction and
objectives of the study; this is while Chapter 2 mostly focuses on the theoretical background
information regarding asphalt material, pavement distresses, and pavement structure analysis.

3
 Chapter 3 is mainly related on the explanation of experimental design, material
selection, sample fabrication, and pavement sections selection. Chapter 4 provides the details
about ruggedness analysis of dynamic modulus testing in indirect tension mode (IDT).
Chapter 5 introduces the new specimen geometry for the cyclic fatigue testing through both
experimental investigation that was all done by the author and numerical analysis using
FEP++ software performed by Dr. Jong-Sub Lee. Mechanistic evaluation of real pavement
sections was performed through LVECD simulations; the results of various pavement
conditions are presented in Chapter 6. Chapter 7 investigates the effects of base layer
thickness, asphalt layer thickness, anti-frost layer, and base layer material on the amount of
fatigue cracking. Verification of the LVECD has been done in Chapter 8, and the LVECD
simulation-to-field cracking transfer function is developed in Chapter 9. Finally, Chapter 10
demonstrates the comparison between Pavement ME simulations done by Yizhuang Wang
versus the LVECD simulation that was performed by the author.

4
CHA
APTER 2 THEO
ORETIC
CAL BAC
CKGROU
UND

Asphalt concrete show

ws different behavior in comparisonn to the otherr materials, ssince
it is composed
c off aggregate which
w is treaated as an elaastic materiaal and asphallt binder thatt is
consiisted of long
g and flexiblee polymer ch
hains. In genneral, these m
molecular chhains are nott able
to disstort immediiately under external forcces. It is knoown that undder low or inntermediate
temperatures asph
halt binder is dominated
d by viscoelaasticity wherreas under hiigh temperattures
visco
oplasticity wiill become a considerablle componennt in responsse.
In this secction, linear viscoelastic theory, visccoelastic conntinuum dam
mage (VECD
D)
modeel which pred
dicts the pro
ogression of material dam
mage, fatiguee failure critterion based on
the reeleased pseu
udo-strain energy, and the theories annd assumptioons behind thhe LVECD
progrram will be discussed
d briefly.

2.1 Linear Viiscoelastic Theory

Hot mix asphalt
a (HMA) is a lineaar viscoelastiic material aat small strainn levels (i.e.., less
than approximate
a ely 75 microstrains). Forr linear elastiic materials,, the stress-sttrain relationnship
is desscribed using
g Hooke’s Law,
L i.e. the stress
s and sttrain are lineearly proporttional to eachh
otherr. For linear viscoelastic
v (LVE) mateerials, whichh is dependennt on time annd temperatuure,
their response is not
n only affeected by the current inpuut, but also thhe past inpuut history.
Thereefore, the strress-strain reelationship can be expresssed by the ffollowing tw
wo convolution
integrrals:
d
t
   E (t   ) d (2.1)
0
d

d
t
   D(t   ) d (2.2)
0
d

where
E (t ) = relaxation modulus,
m
D(t ) = creep complliance, and
 = integration variable.
v

5
 For LVE material and for the case of cyclic loading, the constitutive relationship can
∗
be written in a simple form using the complex modulus and phase angle ). If the
applied strain is
   0 sin( t ) , (2.3)
then the LVE stress response would be
   0 sin( t   ) (2.4)
where
0 = strain amplitude,

0 = stress amplitude,

 = loading frequency, and

 = the time shift between the stress and strain signals.
HMA is also thermorheologically simple (TRS), which means the effects of loading
frequency and temperature, seen in Figure 2-1, can be combined into a parameter called
reduced frequency to produce a single curve to describe the dynamic modulus. |E*| values
can be predicted for any reduced frequency within the measured parameters and extrapolated
for values outside the measured range. In the AASHTO T 342 standard, the single curve that
describes the |E*| as a function of temperature and frequency is called a master curve. The
master curve is represented by the sigmoidal function given in Equation (2.5).
b
log | E* | a  (2.5)
1
1
ec  d *log( fr )
where a, b, c, and d are optimized constants, and fr is the reduced frequency.
The time-temperature superposition (t-TS) principle states that unit response
functions (e.g., the |E*|) in TRS materials, such as HMA, can be shifted in the time or
frequency domain (i.e., along the horizontal axis) to produce a single, continuous master
curve. The shifted frequency, known as the reduced frequency, is calculated by multiplying

6
the measured frequency by a shift factor. Equation (2.6) presents the reduced frequency for
obtaining the master curves.
f r  aT f , (2.6)
where aT is the time-temperature shift factor, and f is the loading frequency in Hz.
The shift factor for this study is represented by a quadratic equation:
log(aT )  a1T 2  a2T  a3 , (2.7)
where a1, a2, and a3 are constants, and T is the temperature. The result is a curve that
provides the relationship between shift factors and temperature as seen in Figure 2-2 with a
reference temperature of 5C. As seen in Figure 2-3, the horizontal shifting of the |E*| values
produces a master curve.

100000

10000
|E*| (MPa)

4°C
1000 20°C

40°C

54°C
100
0.01 0.10 1.00 10.00 100.00
Frequency (Hz)

Figure 2-1. |E*| values of different frequencies and temperatures.

7
 1.0
y = 0.0004x2 - 0.1447x + 0.5728
0.0

-1.0

Log Shift Factor

-2.0

-3.0

-4.0

-5.0

-6.0

-7.0
0 20 40 60
Temperature (°C)

Figure 2-2. Example of a temperature shift factor curve.

100000
4°C

20°C

40°C
10000
|E*| (MPa)

54°C

1000

100
1.E-08 1.E-06 1.E-04 1.E-02 1.E+00 1.E+02
Frequency (Hz)

Figure 2-3. Time-temperature shifted |E*| master curve.

Since the dynamic modulus |E*| describes a constitutive relationship, it can be

mathematically transformed from a frequency-dependent property to time-dependent
properties, such as the relaxation modulus [E(t)] and creep compliance [D(t)]. The relaxation
modulus, with appropriate mathematical considerations, can be used to predict the stress
response to any applied strain. The creep compliance is used to predict the strain response to

8
any stress history. The predicted responses are accurate as long as the applied histories
remain in the linear viscoelastic range. The relaxation modulus can be expressed using the
Prony series representation given in Equation (2.8).
m
E (t )  E   Ei et / i (2.8)
i 1

where
E = the elastic modulus;
t = time; and
Ei and i = the modulus and relaxation time of the ith Maxwell element, respectively
(Prony fitting coefficients).

2.2 Viscoelastic Continuum Damage (VECD) Theory

The viscoelastic continuum damage (VECD) model is important for characterizing
the material behavior of HMA at strain levels beyond the range of LVE behavior. The VECD
model is based on three concepts: 1) Schapery’s work potential theory, 2) the elastic-
viscoelastic correspondence principle, and 3) the t-TS principle with growing damage.
According to Schapery’s work potential theory, damage is quantified by an internal
state variable that considers microstructural changes in the material. Once damage initiates
and grows the secant modulus or stiffness of the material decreases. For a viscoelastic
material, the correspondence principle is applied first to uncouple the time-dependency
associated with viscoelasticity. Then, the viscoelastic problem can be solved using the same
equations as used for the elastic materials after transforming the physical strain to the
pseudostrain. The viscoelastic continuum damage theory is composed of the following basic
equations:
the pseudo strain energy density function,
W R
 f ( R , S ) (2.9)
the stress-pseudo strain relationship,
W R
 (2.10)
 R

9
and the damage evolution law,
dS W R 
 ( ) (2.11)
dt S
where
WR = the pseudo strain energy density,
R = pseudo strain,
S = the damage parameter (internal state variable), and
 = the damage evolution rate.
For uniaxial mode of loading, Equation (2.9) can be written as:
1 R 2
WR  ( ) C (2.12)
2
where C (the effective stiffness) is the only variable that is a function of damage S. When it
is substituted into Equation (2.11), the damage evolution law becomes:
S 1 C 
 (  ( R ) 2 ) (2.13)
t 2 S
Lee and Kim (1998) used the chain rule (Equation (2.14)) to solve the above damage
evolution law by substituting it into Equation (2.13).
dC dC dt
 (2.14)
dS dt dS
After simplification, the damage calculation equation is given in Equation (2.15). In
this equation the time step term, Δt, is replaced by the reduced time interval, Δξ, due to the
verification of the time-temperature superposition principle with growing damage.
 1
1
dSi  (  ( R )i2 Ci )1 .(  )i 1 (2.15)
2
where
 = reduced time interval.

2.3 Simplification of the VECD Model Formulation for Cyclic Loading

The rigorous VECD model requires all the parameters such as the pseudo strain,
effective stiffness, and damage to be calculated for the entire loading history. For instance, an

10
average test with, 10,000 cycles to failure and 100 data points per cycle would then require
the analysis of 100,000 data points which is not impossible by using the advanced computers,
but it is cumbersome and time consuming. The calculations of C and S are slightly different
for the cyclic loading rather than the calculations of a monotonic, constant crosshead
displacement test performed for the research in the previous references. Hou et al. (2010)
present a simplified approach that assumes the rigorous Equation (2.15) can be simplified to
Equation (2.16) for the damage calculation.

1
 1 R 2 * 1 1
dSi    ( 0,ta )i Ci  .( p )i .  K1 1
1
(2.16)
 2 
Compared with Equation (2.15), Equation (2.16) replaces pseudo strain  R with
pseudo strain tension amplitude  R0,ta, pseudo stiffness C with cycle-based value C*, and
time step Δ with reduced pulse time interval Δp. This simplified calculation implicitly
assumes that pseudo strain is some constant value within a cycle. The errors of this
assumption are adjusted by multiplying by K1, an adjustment factor valid when damage
growth within an individual cycle is small.
f
1
  f   
2
K1  d , (2.17)
 f  i i

where  i and  f are the reduced starting time and ending time of the tensile loading
for a given cycle, respectively.
The simplified VECD model used in this study takes advantage of both the rigorous
approach and the simplified approach as suggested by Underwood et al. (2009a), where
pseudo strain is calculated piecewise. For the first loading path where damage growth may be
substantial, the rigorous calculation is used. For the remaining cycles, the simplified
calculation is used, i.e., Equation (2.18). As a result of the piecewise definition of pseudo
strain, the pseudo stiffness and damage are also calculated piecewise, as defined in Equations
(2.19) and (2.20). A detailed explanation of the simplified model formulation can be found in
the papers by Underwood et al. (2009a and 2010).

11
 
 1 d
   E (   ) d   p
R

 ER 0 d
 
R
(2.18)
1  1

 R
 0,ta  cycle i 
ER 2
. 
 0, pp  . | E* |LVE    p

 
C   R .DMR   p

C (2.19)

C*  R 0,ta   p
  0,ta .DMR

  1 R 2 C 
(dSTransient )timestep j       j  .(d ) j   p
  2 S 
dS   
(2.20)
  1 R 2 C * 
 Cyclic cyclic i  2 0,ta i S  .  d p  .  K1 
( dS )   ( ) .   p


2.4 Fatigue Failure Criterion (GR Approach)

Zhang et. al (2013) proposed the GR approach in which energy is evaluated in a
cumulative sense instead of looking at the hysteresis area inside the loop, which is depending
on the variation in phase angle. During the cyclic loading, the maximum amount of stored
pseudo strain energy within each cycle appears at the point of peak stress. This point also
corresponds to the point of maximum pseudo strain. According to Equation (2.21), the
maximum stored pseudo strain energy at cycle can be calculated as:
1 1
R
(Wmax )i  ( max )i ( max
R
)i  ( 0,ta )i ( R 0,ta )i (2.21)
2 2
Hence, the maximum stored pseudo strain energy at cycle can be re-written as:
1
R
(Wmax )i  (C ) i ( R 0,ta ) i 2
2 (2.22)
The maximum stored pseudo strain energy at each cycle reflects the material’s ability
to store energy at that particular time. As the damage accumulates, the material loses the
stored energy for the same magnitude of applied pseudo strain due to the reduction in pseudo
stiffness. The difference between the current stored maximum pseudo strain energy and the
corresponding undamaged state is referred to as the total released pseudo strain energy,

12
which represents the cumulative loss of pseudo strain energy due to the damage process, and
is denoted as WCR .

No damage, 1

With damage

Figure 2-4. Schematic representation of total released pseudo strain energy WRC in the stress-
pseudo strain space.

The released energy during the second step is the total released pseudo strain energy,
which is presented by the shaded triangular area in Figure 2-4. Its formulation is also given in
Equation (2.23).
1 R
  0,ta i (1  Fi )
2
(WCR )i  (2.23)
2
where
(WCR ) i = total released pseudo strain energy at cycle ,

( 0,Rta )i = pseudo strain amplitude at cycle , and

Fi = pseudo stiffness at cycle .

The total released pseudo strain energy is affected by two factors: one is the applied
pseudo strain amplitude,  0,Rta , and the other is the reduction in the material’s pseudo

stiffness. showed that for all the tests, the evolution of WCR is generally categorized into three

13
regions (Figure 2-5) and there is a significant stable region for the rate of WCR , during which
the rate of release of the pseudo strain energy, or released pseudo strain energy per cycle, is
almost constant. This constant slope is called GR.

3.0 0.012
Dissipated strain energy per cycle

2.5

2.0 0.008

RDEC
1.5

1.0 0.004
Stage I Stage II Stage III

0.5

G0R
0.0
0 200 400 600 800 1000 1200 1400
0 500 1000 1500
Number of Cycles Number of Cycles

Figure 2-5. History of WRC and its corresponding rate.

As Figure 2-6 shows, it was also shown that the GR and the fatigue life, Nf (number of
cycles to failure), are highly correlated and this correlation is not sensitive to temperature,
which seems to be a characteristic function of the given mixture. This failure criteria was
applicable for the cyclic controlled-crosshead (COX) fatigue testing, this is while it was not
compatible for the controlled on-Specimen (COS) strain.

14
 100,000
S9.5C-5C

10,000 S9.5C-19C
S9.5C-27C

1,000

R
G
100

10

0
1,000 10,000 100,000 1,000,000
Nf

Figure 2-6. Relationship between GR and Nf for mix S9.5C

A comparison of the WCR histories of the two CX and COS mode tests, which result
in about the same number of cycles to failure (Nf), as shown in Figure 2-7, indicates that in
the case of CX more energy is released at the end of the loading history when the material is
losing its structural integrity rapidly. However, in the COS case, where the specimen is
forced to experience constant on-specimen strain from the beginning of the test, more energy
is released at the outset. Thus, the new failure criterion is able to capture the effect of the
whole loading history. Therefore, the new term, GR, is defined as the rate of change of the
averaged released pseudo strain energy (per cycle) throughout the entire history of the test
and can be defined as Equation (2.24). In short, the GR still depends on the rate of change of
pseudo strain energy, but also captures the effect of the whole loading history.
Nf

W
R
R C
W
GR  C
 0 2 (2.24)
Nf Nf

Because the GR characterizes the overall rate of damage accumulation during fatigue
testing, it is reasonable to hypothesize that a correlation must exist between the GR and the

15
final fatigue life (Nf), because the faster the damage accumulates, i.e., reaches a higher WCR
value during fewer numbers of cycles, the quicker the material should fail.

4.E+06
CX
COS
WRC 3.E+06

2.E+06

1.E+06

0.E+00
0.E+00 1.E+04 2.E+04 3.E+04 4.E+04
N (Cycle)

Figure 2-7. WRC vs. N for a CX and a COS case with similar Nf.

1.E+04

1.E+03
y = 5.17E+07x-1.38E+00
R² = 9.97E-01
GR

1.E+02
VTe30LC-CX-7C
VTe30LC-CX-13C
1.E+01 VTe30LC-CX-20C
VTe30LC-CS-13C
VTe30LC-COS-13C
1.E+00
1.E+02 1.E+03 1.E+04 1.E+05 1.E+06
Nf (Cycle)

Figure 2-8. GR vs. Nf for CS, COS and CX modes for VTe30LC mixtures.

Because the characteristic relationship between GR and Nf relates to the fundamental

properties of the material in terms of damage tolerance and is independent of temperature and
mode of loading, the development of this relationship can be simplified by conducting tests

16
 in only one mode (i.e., CX) at only one appropriate temperature. Thus, this approach
significantly reduces the number of tests required for the calibration of the failure criterion.
The proper test temperature should be low enough to minimize the effect of viscoplastic
strain during the cyclic testing. Equation (2.25) presents a rule of thumb that can be used to
determine the maximum test temperature based on the PG of the base binder. Also, a test
temperature higher than 19C is not recommended.

High temperature binder PG grade  Low temperature binder PG grade

T ( C )  - 3  19 C (2.25)
2

2.5 Layered Viscoelastic Critical Distresses (LVECD) Program

The goal of the performance-related specification is to predict the material behavior
in the real structure in the field. Reasonable stress-strain analysis is a key component in
pavement design and predicting pavement life. Given the complexity of variables such as
pavement life, traffic loading, and temperature variations, various approximate methods are
used to predict pavement performance. Despite differences in assumptions, all of these
prediction methods aim to reduce analysis that takes millions of cycles over several years to a
few hundred analyses under a single cycle of loading.
The three-dimensional finite element method (3-D FEM) is one of the sophisticated
analysis tools for pavement performance analysis that can model the response of a 3-D
pavement under a moving load. Although the 3-D FEM is capable of including the
viscoelasticity and nonlinearity of pavement layers, fatigue cracking, and rutting effects, its
computational cost is prohibitively expensive. Therefore, more practical approaches are often
adopted to perform pavement performance analyses (Xu and Rahman 2008).
The basic approach for stress-strain analysis is the layered elastic analysis (LEA),
where the pavement layers is considered as the elastic material under the stationary
axisymmetric loading (Huang 2003). However, LEA is not the accurate tool for asphalt
pavements since asphalt concrete exhibits viscoelastic behavior, especially under traffic
loading. Layered viscoelastic analysis is an improvement over LEA in that the effect of
viscoelasticity is captured through Laplace transform, but the load is still assumed to be

17
stationary (Xu and Rahman 2008). Layered viscoelastic moving load analysis (LVEMA) is
considered as a more reliable procedure over the LVEA method in the viscoelastic domain
due to the efficient functions for the moving loads effects and viscoelasticity which makes
LVEA quite attractive (Eslaminia and Guddati 2010). Based on the S-VECD model, GR
method for failure criterion and with the consideration of LVEMA that considers the three-
dimensional stress evolution under the moving loads, Eslaminia et al. (2012) developed the
Layered ViscoElastic pavement analysis for Critical Distresses (LVECD) program which is
effectively able to capture the damage in the pavement structures.
As mentioned several times, both rutting and fatigue performance should not be
predicted solely from material level testing. A layered viscoelastic structural model is used to
calculate the stresses and strains necessary for predicting the fatigue and rutting behavior
using the VECD and VP models respectively. The LVECD program has both fatigue and
rutting models implemented. This program uses fast-Fourier transforms to perform the
viscoelastic calculations quickly without the need to account for the previous loading history
at a given point. The assumption behind the transforms is that the loading occurs at the same
repeated manner each time, so the behavior is cyclic and has a steady-state response. This is a
reasonable assumption necessary to characterize 20 years of pavement performance within a
few hours.

The LVECD framework is based on the following assumptions:

 The pavement structure is considered as an infinite layered system where the material
properties change along the pavement depth. This assumption is due the fact that the
pavement dimensions are very large in comparison with the tire size and pavement
thickness.
 Hourly temperature variation is usually slow, whereas the traffic loading changes
every time just within a few seconds. Therefore, pavement performance analysis is
performed at the fixed temperature profile.

18
  The temperature changes along the pavement profile as a function of depth. Its
variation is considered negligible at a specific level therefore temperature is constant
in all the points at a given depth.
 Yearly temperature variation can be obtained by the time-consuming computational
efforts; however the temperature profile is simulated as a cyclic function for a one-
year period. Therefore, the stress and strain calculations can be reduced to a
representative year.
 Since the traffic loading mechanism is complicated, the periodic loading with the
constant tire shape at the specific speed is assumed to be modeled for the traffic.
The proposed framework for the LVECD program can be summarized in the
following steps:
Step 1: The pavement performance analysis is conducted with the average Enhanced
Integrated Climate Model (EICM) temperature for a one-year period.
Step 2: Each representative year is divided into shorter periods of time called stages
which are 12 months for this study. However, more detailed and accurate analysis can be
performed by assigning shorter periods of time.
Step 3: Based on the hourly variation of traffic and temperature, each stage has been
split into three shorter periods called segments which are morning, afternoon, and evening.
The temperature profile is assumed to be constant during each segment.
Step 4: Pavement performance prediction is conducted under a single cycle of traffic
loading using layered viscoelastic analysis for each segment.
Step 5: Given the pavement response analysis results for each segment, the fatigue
cracking analysis can be carried out using the viscoelastic continuum damage (VECD) model
(Kim et al. 2009) combined with an efficient nonlinear extrapolation scheme.

19
CHA
APTER 3 EXPE
ERIMEN
NTAL PR
ROGRAM
M

3.1 Mixture Sampling

S Procedure
P e and Equ
uipment
Most mix
xes used in th
his research were obtaineed as loose m
mix. Therefoore it is
ose mix into sample sizedd quantities. Samples aree separated ffrom
necesssary to sepaarate this loo
5 galllon pails into
o cloth bags which each contain enoough mixturee to fabricatee a single tesst
speciimen. In the procedure, each
e 5 gallon
n pail producces approxim
mately 3 testt specimens.
When
n specimens are needed, the materials in the clotth bags are rre-heated andd compactedd at
the ap
ppropriate teemperature for
f that mateerial.
When sep
parating the loose
l mix in
nto cloth baggs, the oven ttemperature is set 10°C
below
w the compaaction temperrature. If thee compactionn temperaturre is less thann 100°C, thee
ovenss are set at th
he compactio
on temperatu
ure for separration. A schhematic diaggram of the
samp
pling procedu
ure is shown
n in Figure 3-1. The samppling proceddure begins w
with four 5-
gallon
n pails of maaterial and reeduces the quantity
q to 122 test samplees. To accom
mplish this
proceess, first, eacch of the four buckets is quartered byy pouring thee contents innto the centeer of
a box
x that holds four
f smaller sub-pans. These
T sub-panns are then rrandomly selected and
poureed into one of
o a second set
s of pans. Once
O the conntents of all four originaal buckets aree
quarttered, each of
o these secon
nd pans conttains approxximately onee-fourth of eaach original
buckeet. The conteents of each of these pan
ns are thorouughly mixed by hand usiing a scoop aand
smalll rake and sp yers of equaal thickness. Once the maaterial has beeen mixed aand
pread into lay
spreaad, the pan iss placed on top
t of a box with 12 sub--boxes insidde. The seconnd set of panns
has a sliding botttom that can be removed
d quickly to aallow the conntents to droop into thesee sub-
boxess. These sub
b-boxes contaain approxim
mately one-ffourth of the mass needed for a test
speciimen.

20
 Four 5-gallon Pails of Loose Mix

Bags for 12 Test Samples

Figure 3-1. Loose mix sampling schematic.

3.2 Specimen Fabrication

3.2.1 Cylindrical Samples

As mentioned in the section 3.1, cloth bags of loose mix was placed in the oven at the
compaction temperature for 10 minutes, then the bag is taken off and the asphalt mixtures is
poured into a pan . After leaving the pan in the oven for one hour, the loose mix is weighted
out to the required testing size sample weight by scooping from each pan. The individual
samples are then heated to appropriate mixing temperature and compacted. The compacted
specimens were wrapped and sealed in black plastic and stored in a stable, environmentally
controlled room with air temperatures less than 25ºC (77ºF).
All specimens were compacted at a height of 178 mm and a diameter of 150 mm
using the Superpave gyratory compactor. To obtain specimens of uniform air void

21
distribution, these samples were cored to a diameter of 100 mm and cut to height of 150 mm
for dynamic modulus, and 130 mm for cyclic tension testing, as shown in Figure 3-2.

(a) (b)
Figure 3-2. Specimen Fabrication for uniform air voids: (a) coring; and (b) cutting.

Before testing, for quality control, air void content for each specimen is measured
using the CoreLok method. All the testing specimens used in this study have met the target
air void content ±0.5%. To minimize the aging effect, specimens are sealed in plastic bag and
stored carefully in the cabinet if they are not tested immediately after fabrication. No
specimens are tested later than one week after they are cored and cut. All direct tension test
specimens were glued to metal plates at both ends using epoxy before they were placed in the
machine for testing.

3.2.2 Indirect Tension Testing (Disk-Shaped) Samples

All of the testing samples were compacted using the Superpave gyratory compactor
(SGC) to a diameter and height of 150 mm. Each sample was cut into two equal halves to
obtain two disk-shaped specimens with the thickness and diameter of 75 mm and 150 mm,
respectively. To obtain more consistent air void distributed samples, 25/37 mm of the top of
the upper half, and bottom of the lower half were cut to reach 50/38 mm thickness based on
the testing condition. The specimen air void measurement was performed using the CoreLok

22
method before testing and the specimens which met the target air void ± 0.5% were selected
for testing.

3.3 Determining Air Voids for Performance Tests

Since the goal of testing all of the mixtures is to compare the predicted fatigue
cracking obtained from the analysis to field results, the air voids of specimens should match
the field air voids. The air voids have a significant effect on the material behavior. The
challenge with using only the initial air voids comes from the fact that fatigue cracking does
not occur until after 5 years or more of the pavement life.
The National Center for Asphalt Technology, under the auspices of NCHRP 9-9,
conducted a comprehensive study to examine this issue by measuring the air void content of
asphalt pavements after 3, 6, 24 and 48 months of service (Prowell and Brown 2007). The
basic conclusion from this work was that asphalt pavements densify relatively quickly and
ultimately reach their final density after 2-4 years of service. Interestingly, most of this
density is gained in the first 3 to 6 months.
After extracting the data by the NCHRP 9-9, a correlation between the initial (AVini)
and final air void (AVf) contents has been found. Equation (3.1) presents the simple
relationship, however the efforts to cross-correlate this relationship with traffic level, climate,
or other factors did not improve the predictability.
AV f  0.7883  AVini  0.9687 (3.1)

The air void reduction is not as simple as using the relationship above for all layers
because most densification occurs in the top 100 mm. The NCHRP 9-9 researchers reference
a single study, Blankenship et al. (1993), that addresses this particular issue. In that single
study the researchers conclude that not much of a relationship at all exists between traffic and
densification below a depth of approximately 100 mm. Based on the observations made
above, the following steps are proposed to determine the air voids for different layers:
1. Use the NCHRP 9-9 derived relationship to predict the air void content at 22 mm
from the pavement surface. (The average lift thickness in the NCHRP 9-9 project cores is 44
mm.)

23
 2. Compute the change in air void content at this depth by subtracting the predicted
long-term air void content from the initial construction air void content.
3. Fix the air voids for depths greater than or equal to 100 mm in the pavement
structure at a value equal to the as-constructed values.
4. Assume a linear variation in air void between 22 mm and 100 mm and compute the
change in air void as a function of depth between 0 and 100 mm.
5. Compute the long-term air void by subtracting the reduction in air void from the
initial as-constructed air void.
6. Compute the average and representative air void content of a given layer by
numerically integrating and averaging the computed air void distribution across the section
depth.
The base layer does not have any air void changes. This suggests that the base
mixtures will perform worse than the intermediate mixture in both rutting and fatigue tests
due to the high air voids contents, if all other conditions are the same.

3.4 Materials and Test Sections

For this study as a part of the Performance-Related Specification (PRS) research
project for Federal Highway Administration, 32 field sections with 50 different mixtures
from the United States, Canada, South Korea and China were selected to investigate. Federal
Highway Administration Accelerated Load Facility (FHWA-ALF), NCAT Test Track,
Manitoba Infrastructure and Transportation (MIT), Korea Expressway Corporation (KEC),
Binzhou perpetual pavements, and NYDOT I85 section were the projects adopted for this
research proposal.
As presented in Figure 3-3, FHWA-ALF contains four different sets of mixtures with
three modified mixes: Crumb Rubber Terminal Blend (CR-TB), Styrene Butadiene Styrene
(SBS), and Ethylene Terpolymer (Terpolymer). In each of these mixtures, the aggregate
gradation, asphalt content, and the target air void level are identical.
The NCAT mixtures were taken from the surface, intermediate, and the base layer of
six sections as shown in Figure 3-4 from the test track in Alabama, United States. The

24
mixtures were labeled in such a way that indicates the type of modified binder and the
location within the pavement. The C mixtures have the virgin aggregate, binder and the
normal production temperature. The O1 mixture is an open graded friction coarse, this is
while O2 and O3 are consisted of the same materials as the control section. Two warm mix
asphalt (WMA) technologies, Double-Barrel Green foaming technology (FW) and Evotherm
additive technology (AW), have been investigated. The R mixes which contain 50% RAP,
has the same gradation as the control mixtures. Foaming technology has been applied to the
R mixtures to reduce the compaction temperature; these mixtures labeled as RW.

Control SBS-LG CR-TB Terpolymer

0

Terpolymer
5
Depth (cm)

CR-TB

SBS-LG

Control
10

15

Figure 3-3. Asphalt pavement structure for the FHWA-ALF project.

25
 C O FW AW R RW
0 NCAT-C3
NCAT-C2
NCAT-C1
NCAT-O3
5
NCAT-O2
NCAT-O1
NCAT-FW3
10 NCAT-FW2
Depth (cm)

NCAT-FW1
NCAT-AW3
NCAT-AW2
15 NCAT-AW1
NCAT-R3
NCAT-R2
NCAT-R1
20
NCAT-RW3
NCAT-RW2
NCAT-RW1
25

Figure 3-4. Asphalt pavement structure for the NCAT project.

MIT sections as demonstrated in Figure 3-5 are located on provincial highway 8 in

Manitoba, Canada. The total project length is about 17 miles that is divided to two different
parts. The first part includes eight mixtures from four sections which evaluates the effect of
three warm mix asphalt technologies; Advera (MIT-W-A), Sasobit (MIT-W-S) and
Evotherm (MIT-W-E). All the mixtures used for this part have the same gradation and binder
as the control mixture (MIT-W-C), but with the different warm mix additives (Figure 3-5
(a)). The second part of MIT as shown in Figure 3-5 (b) were constructed in September 2009
and consisted of two 2-inch layer with conventional HMA (MIT-R-C), 15% RAP (MIT-R-
15R), 50% RAP (MIT-R-50R) and 50% RAP with the softer binder (MIT-R-50RSB). The
pavement sections were laid on top of a 4-inch HMA with 50% RAP that was constructed
one year before (i.e., 2008), which is on top of a base and a subgrade.

26
 (a) WMA Section
0.5 km 3 km 1 km 3 km 1 km 3 km 0.5 km

50 mm 50 mm
HMA Advera HMA Sasobit HMA Evotherm HMA

+35% +35% +35% +35%

+35% RAP +35% RAP +35% RAP
RAP RAP RAP RAP

(b) RAP section

1-3 km 2.5 km 2.5 km 2.5 km

0% RAP 15% RAP 50% RAP 50% RAP

100 mm 100 mm
150/200 pen 150/200 pen 200/300 pen 150/200 pen
(PG 58-28) (PG 58-28) (PG 52-34) (PG 58-28)

70% RAP Base Layer

Figure 3-5. MIT pavement sections with thickness (mm) (a) WMA, (b) RAP.

The asphalt pavement structures of the KEC test road that presented schematically in
Figure 3-6 were designed to systematically examine the effects of base type, base thickness,
and sublayer type under actual road and environmental conditions. Five of the asphalt
mixtures from the KEC test road project were chosen for field investigation. Of these
mixtures, four are conventional mixes: a surface course (ASTM), an intermediate course
(BB5), and two base courses (BB1 and BB3). The fifth mixture (PMA) has the same gradation
as the ASTM mixture, but polymer-modified asphalt replaces the conventional asphalt.

27
 A1 A2 A2-2 A3 A4 A5 A5-2 A6 A7 A8 A8-2 A9 A10 A10-2 A11 A11-2 A12 A12-2 A13 A13-2 A14 A14-2 A15 A15-2

5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5

8 8 8
18 18 18
28 28 28

30 30 30
40
30 30 30
40
30 30 30
40

40 40 40 40
30 30 30 30
20 20 20 20
10

ASTM PMA BB5 BB3 BB1 AGG SUBBASE ANTIFROST SUBGRADE

Figure 3-6. Asphalt pavement sections for the KEC project.

The NYDOT pavement sections have three layers: surface, intermediate, and base.
The intermediate layer is composed of two lifts. Accordingly, four mixtures have been tested
for the NYDOT project to accommodate the surface layer, two lifts of the intermediate layer,
and the base layer.
The Binzhou perpetual pavement project consists of five testing sections of varying
thicknesses and material compositions. Six asphalt concrete mixtures are used in these
sections. For each section, the top three layers are the same: 12.5 mm nominal maximum
aggregate size (NMAS) stone matrix asphalt (SMA), 19.0 mm NMAS dense-graded mixture,
and 25.0 mm NMAS dense-graded mixture, respectively. The thickness of these layers
differs between sections as do the support layers for these mixes. Figure 3-7 presents the
different pavement structures for the Binzhou project. As seen, section 5 adopts the widely
used flexible pavement design with semi-rigid bases used in China. The major problem with
this type of pavement is reflective cracking that initiates in the semi-rigid base and
propagates through the asphalt layers above. Section 4 is considered to be an upgrade of
Section 5 in that an absorbing layer of large stone porous mixture (LSPM) is inserted

28
between the conventional asphalt mixtures and the semi-rigid base. LSPMs have been field
proven in China to have the potential to reduce reflective cracking as well as drain water
from pavement structures. Sections 1 through 3 are full-depth asphalt pavements that are
designed based on traditional perpetual pavement principles. As such, each section employs a
highly fatigue resistant mixture on top of the base/subgrade and a layer of LSPM just above
this bottom fatigue layer.
Table 3-1 presents the volumetric properties of all of the mixtures used in this study.

Figure 3-7. Binzhou pavement structure layout for each section.

29
 Table 3-1. Volumetric properties of mixtures used for this research proposal.
NMSA RAP Binder Air Compaction
Project Mix Additive Location
(mm) Content (%) Grade Void (%) Temp (oC)
Control 12.5 - 0 PG 70-22 Surface 4.0 150
CR-TB 12.5 Crumb Rubber 0 PG 70-28 Surface 4.0 150
FHWA-ALF
SBS-LG 12.5 SBS 0 PG 70-28 Surface 4.0 150
Terpolymer 12.5 Terpolymer 0 PG 70-28 Surface 4.0 150
NCAT-C1 9.5 - 0 PG 76-22 Surface 4.3 154
NCAT-C2 19 - 0 PG 76-22 Intermediate 6.1 154
NCAT-C3 19 - 0 PG 67-22 Base 7.4 143
NCAT-O1 9.5 - 0 PG 76-22 Surface 18.3 152
NCAT-O2 19 - 0 PG 76-22 Intermediate 6.1 154
NCAT-O3 19 - 0 PG 67-22 Base 7.4 143
NCAT-FW1 9.5 Foam 0 PG 76-22 Surface 4.9 118
NCAT-FW2 19 Foam 0 PG 76-22 Intermediate 6.0 118
NCAT-FW3 19 Foam 0 PG 67-22 Base 7.7 118
NCAT
NCAT-AW1 9.5 Evotherm 50 PG 76-22 Surface 3.9 104
NCAT-AW2 19 Evotherm 50 PG 76-22 Intermediate 6.2 104
NCAT-AW3 19 Evotherm 50 PG 67-22 Base 6.1 104
NCAT-R1 9.5 - 50 PG 67-22 Surface 4.7 143
NCAT-R2 19 - 50 PG 67-22 Intermediate 6.1 143
NCAT-R3 19 - 50 PG 67-22 Base 5.0 143
NCAT-RW1 9.5 Foam 50 PG 67-22 Surface 5.0 118
NCAT-RW2 19 Foam 50 PG 67-22 Intermediate 5.8 118
NCAT-RW3 19 Foam 50 PG 67-22 Base 5.8 118
MIT-W-C1 - - PG 58-28 Surface 3.9 129
MIT-W-C2 - 35 PG 58-28 Bottom 4.8 139
MIT-W-S1 Sasobit - PG 58-28 Surface 3.2 106
MIT-W-S2 Sasobit 35 PG 58-28 Bottom 4.9 118
MIT-WMA 16
MIT-W-E1 Evotherm - PG 58-28 Surface 3.8 106
MIT-W-E2 Evotherm 35 PG 58-28 Bottom 5.4 117
MIT-W-A1 Advera - PG 58-28 Surface 3.0 108
MIT-W-A2 Advera 35 PG 58-28 Bottom 5.4 106
MIT-R-C - - PG 58-28 Surface 5.4 134
MIT-R-15R - 15 PG 58-28 Surface 5.2 134
MIT-RAP 16
MIT-R-50R - 50 PG 58-28 Surface/Bottom 5.9 134
MIT-R-50RSB - 50 PG 52-34 Surface 5.7 129
ASTM 19 - ‐ PG 64-22 Surface 5.9 144
PMA 19 SBS ‐ PG 76-22 Surface 5.9 144
KEC BB1 40 - ‐ PG 64-22 Base 8.1 144
BB3 25 - ‐ PG 64-22 Base 5.9 144
BB5 25 - ‐ PG 64-22 Intermediate 7.9 144
NY9.5 9.5 - ‐ PG 64-22 Surface 3.2 143
NY19_L5 19 - ‐ PG 64-22 Intermediate 5.0 143
NYDOT
NY19_L3 19 - ‐ PG 64-22 Intermediate 6.1 143
NY25 25 - ‐ PG 64-22 Base 6.2 143
SMA 12.5 MAC Modified ‐ PG 76-22 Surface 3.7 173
S19 19 MAC Modified ‐ PG 76-22 Intermediate 6.0 173
S25 25 ‐ PG 64-22 Intermediate 7.5 143
Binzhou
LSPM 25 MAC Modified ‐ PG 70-22 Base 15.1 173
F-1 12.5 ‐ PG 64-22 Base 3.6 143
F-2 12.5 SBS ‐ PG 76-22 Base 3.6 163

30
CHA
APTER 4 RUGG
GEDNES
SS STUD
DY OF D
DYNAMIIC
MO
ODULUS TESTIN
NG OF ASPHALT
A T CONC
CRETE IIN
IND
DIRECT TENSIO DE 1
ON MOD

4.1 Abstract
The modu
ulus is one of
o the primary
y asphalt miixture properrties used foor the
mech
hanistic perfo
ormance prediction of assphalt pavem
ments. Dynam
mic modulus testing is a
comm
mon method of measurin
ng mixture modulus
m as a function off loading freqquencies andd
temperatures. This chapter prresents the reesults of a ruuggedness sttudy of dynaamic moduluus
testin
ng in indirectt tension mo
ode to evaluaate the factorrs that are m
most likely too affect the fiinal
resultts. Specimen
n thickness, air
a void conttent, gauge llength, test ttemperature, and horizonntal
strain
n level, whicch are the criitical factors that affect tthe dynamic modulus off asphalt conccrete,
were selected forr the ruggedn
ness analysiss. Two differrent asphalt mixtures wiith the
particcipation of tw
wo laboratorries were useed in the studdy. Based onn the selecteed values forr the
differrent variablees, air void content was found
f to be tthe significaant factor thaat affects
dynam
mic modulus testing and
d dynamic modulus
m valuues. The otheer factors didd not appear to
have a major imp
pact on the teest results; however,
h reassonable tolerances were obtained for the
otherr parameters investigated
d in this chap
pter.

4.2 Introducttion
The pavem
ment industrry in the Uniited States w
would like too improve thee quality of
pavem
ment constru
uction and ex
xtend the liffe of new andd rehabilitateed pavementts. One way to
impro
ove the quality of pavem
ment design and
a performance predicttions is to chharacterize thhe
asphaalt mixture properties
p as precisely ass possible. Thhe dynamic modulus is one of the m
most
important asphaltt properties that
t has been
n used to preedict the amoount of distrress, such as
ue cracking and
fatigu a rutting, in a pavemeent (Norouzi and Kim 20015, Norouzii and Sabourri

1
This chapter previiously was pu
ublished as: Norouzi,
N A., aand Y. R. Kim
m. Ruggednesss Study of
Dyynamic Modu ulus Testing of
o Asphalt Con ncrete in Indiirect Tension Mode. ASTM
M Journal of
Testing and Evaalaution.

31
2014, Sabouri and Kim 2014). During recent decades, significant research has been
conducted to evaluate mixture stiffness (Majidzadeh et al. 1979, Witczak et al. 1996, Shook
et al. 1969, Tafreshi and Norouzi 2015); however, it is still necessary to establish a practical
dynamic modulus test method that is compatible with the use of field cores (Kutay et al.
2009, Park et al. 2014, Kim and Kim 2015, Lee et al. 2015). One of the concerns associated
with the current dynamic modulus test protocol, AASHTO TP79, is its applicability to
forensic studies and pavement rehabilitation design. This protocol requires axial compression
testing of asphalt concrete specimens that are 100 mm in diameter and 150 mm in height.
Such specimens are often impossible to obtain from actual pavements because typical asphalt
layers are less than a few inches thick. Therefore, indirect tension (IDT) testing of field cores
seems to be appropriate for evaluating existing pavements, even though controversial
differences between axial compression test methods and IDT test methods remain.
Hondros (1959) derived elastic solutions for the IDT test using the plane stress
assumption. Pavement agencies have been following his solutions until Roque and Buttlar
(1992) proposed correction factors for the bulging effects found in IDT test specimens.
Because the stress and strain distributions in IDT specimens are biaxial, the state of the stress
and strain can cause errors when trying to determine material properties from IDT testing
unless the derivation of the properties is carefully handled. Kim et al. (2004) introduced
viscoelastic solutions for dynamic modulus IDT testing using the theory of linear
viscoelasticity. The modulus values obtained from Kim’s equations are in an agreement with
those determined from uniaxial tests of mixtures commonly used in North Carolina.
One of the reasons for poor precision in current test methods is the lack of adequate
control over the sources of variation in the test procedures. Generally, these sources of
variability are not controlled satisfactorily because they were not identified during the
development of the test procedure (Bonaquist 2008). According to Heyden et al. (2011), the
ruggedness of any procedure is a strong tool for identifying the capacity of the test procedure
to stay within a specific limit by testing small but intentional variations in the parameters.
This approach provides an indication of the reliability of the test method during normal
usage. Ruggedness analysis, which is always accompanied by an inter-laboratory study with

32
the purpose of increasing the precision of the test results, can be employed to determine the
controllable test conditions that most affect the results and to establish limits for their control.
In general, ruggedness testing is a critical part of test method development; therefore,
efficient statistical designs should be developed and standardized. ASTM C1067, Standard
Practice of Conducting the Ruggedness or Screening Program for Test Methods for
Construction Materials, describes a specific design for simultaneously evaluating the effects
of changes in the number of the operating conditions when there is no interaction among the
various conditions.
Considering the significance of IDT dynamic modulus testing for the pavement
industry, i.e., departments of transportation (DOTs) and agencies, the purpose of this chapter
is to identify the impact of important parameters that affect IDT test results with the goal of
developing limits in a specification for measuring dynamic modulus values using IDT tests
of disk-shaped specimens. The final draft has been sent to the Expert Task Group (ETG) for
further review and consideration for publication as a testing method standard.

4.3 Theoretical Background

4.3.1 Linear Viscoelastic Solution for IDT Specimens

Hondros (1959) presented the stress and strain relationship (Equation (4.1)) along the
diameter of an IDT specimen subjected to strip loading.
1
x  ( x   y ) (4.1)
E
where
E = modulus of elasticity and
 = Poisson’s ratio.
The vertical and horizontal stresses in Equation (1) can be obtained from Equations
(4.2) and (4.3).
2P  (1  x 2 / R 2 ) sin 2 1  x 2 / R 2  2 P
 x ( x)    tan 1  tan     f ( x)  g ( x) (4.2)
 ad  1  2( x 2
/ R 2
) cos 2   x 4
/ R 4
1  x 2
/ R 2
   ad

33
 2P  (1  x 2 / R 2 )sin 2 1 1  x / R
2 2
 2P
 y ( x)     tan  tan      f ( x)  g ( x) (4.3)
 ad  1  2( x 2
/ R 2
) cos 2  x 4
/ R 4
1  x 2
/ R 2
   ad

where
P = applied load,
a = loading strip width, m,
d = specimen thickness, m,
R = specimen diameter, m, and
 = radial angle.
Kim et al. (2004) derived the stress and strain equations for viscoelastic materials
subjected to sinusoidal loading, as follows:
1
x  ( x   y ) (4.4)
E*
E *  E *  e i (4.5)

P  P0 e iwt  P0 (cos wt  i sin wt ) (4.6)

where
E * = is the dynamic modulus,
P0 = the loading amplitude,
w = the angular frequency of sinusoidal loading, and
 = the phase angle between the load and displacement.
By substituting Equations (4.2), (4.3), (4.5), and (4.6) into Equation (4.4), and
integrating over the gauge length, the horizontal displacement, U(t), can be determined by
Equation (4.7).
l
2P0  l l

U (t )    x ( x, t )dx  e i ( wt  )
(1   )  f ( x)dx  (  1)  g ( x)dx (4.7)
l E * ad  l l 
Therefore, the dynamic modulus obtained from the horizontal displacement, U(t), and
vertical displacement, V(t), can be expressed as:

2 P0 sin( wt   )  
l l
E   (1   )  f ( x)dx  (  1)  g ( x) dx 
*
(4.8)
 ad  U (t )  l l 

34
 2 P0 sin( wt   )  
l l
E* 
 ad  V (t ) 
 (  1) 
l
n ( y ) dy  (1   ) 
l
m( y ) dy 

(4.9)

where

f x 
1  x 2
/ R 2  sin 2
(4.10)
1  2( x 2 / R 2 ) cos 2  x 4 / R 4

1  x 2 / R 2 
g  x   tan 1  tan   (4.11)
1  x / R
2 2


m y 
1  y 2
/ R 2  sin 2
(4.12)
1  2( y 2 / R 2 ) cos 2  y 4 / R 4

1  y 2 / R 2 
n  y   tan 1  tan   (4.13)
1  y / R
2 2


With the substitution of Equation (4.10) through Equation (4.13) into Equations (4.8)
and (4.9), and combining them, the following function can be expressed for the dynamic
modulus, as shown in Equation (4.14).
P0 sin(wt   ) 1 2   2 1
E*  2 (4.14)
ad  2V (t )   2U (t )
where
l l
1    n( y)dy   m( y)dy , (4.15)
l l

l l
 2   n( y)dy   m( y)dy , (4.16)
l l

l l
 1   f ( x)dx   g ( x)dx , (4.17)
l l

l l
 2   f ( x)dx   g ( x)dx . (4.18)
l l

The vertical and horizontal displacements can be introduced in sine functions as

Equations (4.19) and (4.20).
V (t )  V 0 sin( wt   ) (4.19)

35
 U (t )  U 0 sin( wt   ) (4.20)
V0 and U0 are the constant amplitudes of the vertical and horizontal displacements.
Finally, the ultimate form of the dynamic modulus is:
P0 1 2   2 1
E*  2 (4.21)
ad  2V0   2U 0

4.3.2 Ruggedness Experiments

As mentioned in the Introduction, the main goal of ruggedness testing is to improve
the test procedure by specifying the major test conditions and establishing limits for them.
Ruggedness analysis consists of testing multiple materials in different laboratories to
evaluate the significant factors that affect the test procedures; such tests require time and
resources.
Efficient statistical designs have been developed and standardized for ruggedness
analysis as a part of the test method development. ASTM C1067 requires seven factors to be
examined for each test; however, the design used in this study is based on five factors, and
one factor was repeated twice to adhere to the seven-factor Plackett-Burman design (Tanesi
et al. 2013). This design seems to be efficient for simultaneously evaluating the effect of
changes in a number of operating conditions that are being evaluated. According to the
specification, in order to investigate the effects of changes in the test parameters, each
condition normally is evaluated at two different levels.
Table 4-1 summarizes the five-factor design with each corresponding level. All of the
five parameters are presented such that the high-level factors are indicated with ‘1’, and the
low-level factors are shown with ‘-1’. In order to determine the effect of each factor on the
final results, the design requires combinations of changes to be applied to each of the variable
factors; therefore, sixteen determinations (i.e., two replicates for each of the specific
combinations, as presented in Table 4-2) were conducted.
Analysis of the data, as described in the standard, is straightforward. It involves
determining the effects for each factor included in a partial factorial design and determining
the variance of a single measurement. Statistical analysis (F-tests) can be used to assess the

36
sensitivity of each factor relative to the variance of a single measurement; therefore, F-tests
were conducted for this study.

Table 4-1 Factors and Levels of Ruggedness Analysis

Factor Variable Lower Level Value Upper Level Value
A Air Void (%) 6 8
B Gauge Length (mm) 50.8 101.6
C Thickness (mm) 38 50
D Temperature (C) -11, 9, 34 -9, 11, 36
E Horizontal Strain () 40 60
Note: * is microstrain.

Table 4-2 Experimental Design for Ruggedness Testing

Determination Number
Factor 1 2 3 4 5 6 7 8
Air Void (%) 1 1 1 1 1 1 1 1
Gauge Length (mm) 1 1 -1 -1 -1 -1 1 1
Thickness (mm) -1 -1 1 1 -1 -1 1 1
Horizontal Strain () 1 -1 -1 1 -1 1 1 -1
Temperature (C) 1 -1 -1 1 1 -1 -1 1
Temperature (C) 1 -1 -1 1 1 -1 -1 1
Temperature (C) 1 -1 -1 1 1 -1 -1 1

4.4 Laboratories, Materials, and Equipment

For this study, North Carolina State University (NCSU) and the National Center of
Asphalt Technology (NCAT) were the laboratories participating in the ruggedness testing.
Although ASTM C 1067 recommends two or more laboratories to participate in ruggedness
testing, using more than two laboratories does not necessarily increase the testing precision

37
significantly. Generally, because of high variability in the test data obtained from different
asphalt labs, an inter-laboratory study is not able to improve the testing procedure.
To evaluate the ruggedness of the tests, two types of mixtures, NE and VA, from New
England and Virginia, respectively, were selected for IDT dynamic modulus testing.
Table 4-3 presents the characteristics of the two types of mixtures used in this study. All of
the test samples were compacted using the Superpave gyratory compactor (SGC) to a
diameter and height of 150 mm. Each sample was cut into two equal halves to obtain two
disk-shaped specimens with the thickness and diameter of 75 mm and 150 mm, respectively.
To obtain samples with consistently distributed air void contents, 25 mm and 37 mm of the
top of the upper half and bottom of the lower half, were cut to 50 mm and 38 mm thickness,
respectively, based on the test condition. The specimen air void measurements were taken
using the CoreLok method prior to testing, and the specimens that met the target air void
content of ± 0.5% were selected for testing.

Table 4-3 Mixture Properties

Mixture NMAS (mm) Binder Grade RAP (%) Fabricating Laboratory

VA 12.5 PG 64-22 25% VCTIR*

NE 9.5 PG 52-34 0% NCSU

Note: *VCTIR stands for Virginia Center for Transportation Innovation and Research.

Table 4-4 presents the coefficients of 1, 2, 1, and 2 calculated for different
specimen diameters, gauge lengths, and loading widths based on the plane stress assumption.

Table 4-4 Coefficients for Calculating the Dynamic Modulus

Specimen Loading Gauge
Lab Diameter Width Length β1 β2 γ1 γ2
(mm) (mm) (mm)
50.8 -0.0134 -0.0042 0.0037 0.0116
NCSU 12.8
101.6 -0.0328 -0.0084 0.0053 0.0186
150
50.8 -0.0193 -0.0057 0.0051 0.0167
NCAT 18.6
101.6 -0.0462 -0.0106 0.0072 0.0262

38
 IDT dynamic modulus testing was carried out in compression under load-controlled
mode at -10°C, 10°C, and 35°C and at frequencies of 25, 10, 5, 1, 0.5, 0.1, 0.05, and 0.01 Hz.
Four loose-core type miniature linear variable differential transducers (LVDTs) were
attached on the front and back surfaces (vertical to each other on each side) using either 50.8
mm or 101.6 mm as a gauge length to capture both vertical and horizontal displacements. A
closed-loop servo-hydraulic material testing system (MTS) containing an environmental
chamber was used to apply vertical loading to the IDT samples. The load levels were set by
trial and error to obtain the horizontal strain amplitude of either 40  or 60  5  without
any damage to the samples. In order to check the temperature of the test specimens, a dummy
specimen of the same size as the test specimen with a thermocouple inside was used to
determine the temperature.

4.5 Factors that Affect IDT Dynamic Modulus Testing

The most important considerations for experimental designs are factors and levels. As
mentioned earlier, for this study, the specimen thickness, test temperature, LVDT gauge
length, sample air void content, and horizontal strain levels are the factors that were selected
based on the experience of DOTs and asphalt laboratories with dynamic modulus testing.
This section briefly discusses the rationale behind the selection of the two levels for each
experimental factor.

4.5.1 Temperature
Asphalt mixtures are considered to be viscoelastic material; therefore, their
viscoelastic behavior includes both elastic and viscous properties that are strongly dependent
on loading time and temperature. The current tolerance for temperature for cylindrical
gyratory specimens in the AASHTO TP 79 test protocol is  0.5C from the target
temperature. Because the purpose of this ruggedness study is to establish a practical test
procedure for disk-shaped samples, the research team decided to use  1.0C as the tolerance
to investigate the effects of temperature on the dynamic modulus.

39
4.5.2 Air Void Content
Air void content is one of the main parameters that must be controlled precisely.
Although it is possible and practical to prepare samples with an air void tolerance of  0.5%
by following the current specimen fabrication procedure, it is desirable to reach a higher
tolerance to reduce the time and resources needed for sample preparation. Because the typical
air void content for the field sections is around 7%, in order to capture the effect of air void
content on the dynamic modulus values, the specimens prepared in both labs had air void
contents of 6% or 8% (7%  1%), depending on the test conditions.

4.5.3 Horizontal Strain

The dynamic modulus test results for numerous mixtures tested at NCSU suggest that
strain levels between 50  to 75  provide accurate measurements of linear viscoelastic
properties without being affected by any damage to the specimens. Given the average
Poisson’s ratio of 0.35, the horizontal strain level varies between 17  to 25 , which are
relatively low levels to obtain a reliable conclusion due to very low quality horizontal
displacement patterns. AASHTO TP 62 proposes a wide tolerance range of 50  to 150 
for vertical strain without any damage to the sample. Therefore, with the consideration of a
conventional Poisson’s ratio (e.g., 0.3) for the asphalt material, the horizontal strain
amplitudes of 40  and 60  were adopted for investigation in this study.

4.5.4 Specimen Thickness

Thickness is considered to be one of the major parameters that affects the
representative volume element (RVE) properties. Generally, thicker specimens are better
representatives of the asphalt mixture properties than thinner specimens, especially for
mixtures that contain larger aggregate particles. However, because a typical asphalt layer is
less than a few inches thick, obtaining undamaged thick field cores is always a concern.
Furthermore, the plane stress assumptions used for the proposed viscoelastic solutions are not
valid for testing very thick specimens. ASTM D 6931 recommends various thicknesses based
on the diameter of the specimen: 50 mm for 150-mm diameter specimens and 38 mm for
101.6-mm diameter specimens in order to assess the IDT strength of bituminous mixtures.

40
For this ruggedness study, 38-mm and 50-mm thick specimens were evaluated to find the
maximum tolerance of thickness for disk-shaped field cores.

4.5.5 Gauge Length

In addition to specimen thickness, gauge length is another parameter that affects the
RVE properties. If the gauge length adopted is not sufficiently long, this critical factor will
lead to errors in data acquisition (Buttlar and Roque 1994). Changing the gauge length
significantly impacts the dynamic modulus values, especially when the nominal maximum
aggregate size (NMAS) is relatively high (e.g., 25 mm). However, in this study, because of
lack of materials, it was impossible to assess the mixtures that contain large aggregate
particles. So, to perform dynamic modulus tests of cylindrical asphalt specimens (100 mm in
diameter and 150 mm in height), a 75-mm (3-inch) gauge length was adopted. Based on the
literature review, 2-inch and 4-inch gauge lengths (3 1 inches) were used for the analysis to
check the effect of gauge length on the dynamic modulus of the IDT samples (Witczak
2000).

4.6 Results and Discussion

Figure 4-1 presents the dynamic modulus mastercurves obtained from both the axial
compression tests and IDT tests for the NE samples with the same air void content. The key
observation is that the mastercurve developed from the IDT test using the biaxial linear
viscoelastic solutions nearly collapses with the mastercurve obtained from the axial
compression test, with consideration of sample-to-sample variation. These results confirm the
findings by Kim et al. (2004) regarding good agreement between mastercurves obtained from
IDT and axial compression tests for twelve mixtures commonly used in North Carolina.

41
 16000
Axial Compression (a)

IDT-38mm Thickness
12000
IDT-50mm Thickness

|E*| (MPa)
8000

4000

0
1E-08 1E-05 1E-02 1E+01 1E+04
Reduced Frequency (Hz)

100000
Axial Compression (b)

IDT-38mm Thickness

10000 IDT-50mm Thickness

|E*| (MPa)

1000

100
1E-08 1E-05 1E-02 1E+01 1E+04
Reduced Frequency (Hz)
Figure 4-1. Dynamic modulus mastercurves obtained from axial compression and IDT tests:
a) semi-log space and b) log-log space.

In order to determine the statistical values that correspond to each parameter,

ruggedness analysis was conducted according to ASTM C1067. Table 4-5 and

42
 Table 4-6 present the distribution of the F-values for the NE and VA mixtures tested
at the NCSU and NCAT laboratories. The F-value of 5.32 represents significance for a factor
at a probability of 5 percent. A parameter was defined as statistically significant when its F-
value was higher than 5.32. The impact of each factor is discussed briefly in the following
section.

Table 4-5. Summary of F-values for NE Mixture for Both Laboratories and All Factors.
10 Hz 1 Hz 0.05 Hz

Temperature
Factor NCSU NCAT NCSU NCAT NCSU NCAT
(°C)
Thickness 6.26 0.09 4.05 0.88 2.11 1.18
Air Void 66.17 104.62 39.87 34.87 29.61 38.97
-10 Gauge Length 0.7 5.35 0.18 3.62 0.66 11.22
Temperature 5.17 2.94 2.54 1.34 2.31 1.90
Strain Level 0.90 8.04 0.08 1.77 0.20 1.16
Thickness 3.87 0.02 0.05 0.01 0.42 0.04
Air Void 17.67 82.2 23.76 16.84 26.3 33.44
10 Gauge Length 0.21 1.22 0.53 0.02 0.25 0.76
Temperature 2.68 12.8 0.5 0.5 0.09 1.17
Strain Level 0.03 0.50 0.14 1.17 0.03 1.02
Thickness 0.63 2.17 4.41 1.33 4.34 0.33
Air Void 8.20 0.37 28.64 3.54 9.80 4.71
35 Gauge Length 0.11 1.42 3.96 0.22 3.40 0.03
Temperature 0.03 0.84 0.20 0.12 0.90 0.23
Strain Level 0.61 0.43 0.45 0.05 2.34 0.13

43
Table 4-6. Summary of F-values for VA Mixture for Both Laboratories and All Factors.
10 Hz 1 Hz 0.05 Hz
Temperature
Factor NCSU NCAT NCSU NCAT NCSU NCAT
(°C)
Thickness 2.99 1.93 2.75 1.67 0.42 0.89
Air Void 47.52 106.25 34.13 41.44 11.61 17.12
-10 Gauge Length 2.1 7.6 2.13 4.82 0.16 2.11
Temperature 1.68 6.22 1.65 2.69 2.46 7.25
Strain Level 0.1 1.38 0.04 0.63 0.13 1.42
Thickness 0.36 0.3 0.65 0.27 0.7 0.03
Air Void 10.01 9.71 0.71 7.2 0.03 3.14
10 Gauge Length 0.09 1.25 0.21 2.5 0.51 2.37
Temperature 0.17 2.98 0.51 4.25 2.85 9.77
Strain Level 0.22 1.09 0.3 0.95 0.04 1.81
Thickness 0.06 0.05 0.99 0.1 2.98 1.86
Air Void% 0.07 1.24 1.91 0.01 4.29 3.18
35 Gauge Length 0.03 0.01 4.17 0.45 8.16 0.84
Temperature 1.43 2.79 3.99 4.02 2.7 5.73
Strain Level 1.22 0.30 1.28 0.82 16.33 1.35

44
4.6.1 Temperature
Although the F-values for the test temperature were higher than 5.32 in one case for
NE mixture and four cases for VA mixture, generally temperature was found to be
insignificant, as shown in Table 4-5 and

Table 4-6. The acceptable tolerance for temperature is  0.5C for dynamic modulus
tests, especially in axial compression mode. Therefore, based on the results of the ruggedness
study, it can be concluded that this tolerance can be increased to  1C without any major
change in the ultimate results.

4.6.2 Air Voids

Figure 4-2 shows the dynamic modulus mastercurves for the NE and VA mixtures.
Each condition was specified according to the determination number shown in Figure 4-3.
For example, a condition labeled Case 1 refers to determination 1, which has 38 mm
thickness, 8% air void content, 101.6 mm gauge length, 60 , and test temperatures of -9C,
11C, and 36C.

45
 A comparison of the dynamic modulus mastercurves presented in Figure 4-2 clearly
shows that changing the air void content has a pronounced effect on the dynamic modulus
values; i.e., the material stiffness for the 6% air void specimens is relatively higher than that
of the 8% air void samples. The results of the ruggedness analysis confirm this outcome,
because the corresponding F-values for the low and intermediate temperatures are
significantly higher than the critical F-value of 5.32. However, the effect of air void variation
appears to be negligible at high temperatures in comparison to the low and intermediate
temperatures. In order to determine an acceptable limit for the air void content, 10% lower
and higher of the dynamic modulus value for each temperature was specified by developing
the linear relationship between the dynamic modulus measurements in the lab testing; then,
the corresponding air void content to each value was calculated via interpolation. According
to the results shown in Figure 4-3, which indicate the variation of dynamic modulus value
versus air void content for both the VT and VA mixtures at different temperatures, gauge
lengths, thicknesses, and frequencies,  0.5% air void can be considered as a reasonable
tolerance based on the current specimen fabrication procedure.

46
 30000 30000
VA-Case 1-Rep 1 VA-Case 5-Rep 1

25000 VA-Case 1-Rep 2 25000 VA-Case 5-Rep 2

VA-Case 2-Rep 1 VA-Case 6-Rep 1

20000 20000
|E*| (MPa)

|E*| (MPa)
VA-Case 2-Rep 2 VA-Case 6-Rep 2

15000 15000

10000 10000

5000 5000

(a) (b)
0 0
1E-06 1E-03 1E+00 1E+03 1E+06 1E-06 1E-03 1E+00 1E+03 1E+06
Reduced Frequency (Hz) Reduced Frequency (Hz)
20000 20000
NE-Case 3-Rep 1 NE-Case 7-Rep 1

NE-Case 3-Rep 2 NE-Case 7-Rep 2

15000 15000
NE-Case 4-Rep 1 NE-Case 8-Rep 1

|E*| (MPa)
|E*| (MPa)

NE-Case 4-Rep 2 NE-Case 8-Rep 2

10000 10000

5000 5000

(c) (d)
0 0
1E-06 1E-03 1E+00 1E+03 1E+06 1E-06 1E-03 1E+00 1E+03 1E+06
Reduced Frequency (Hz) Reduced Frequency (Hz)

Figure 4-2. Comparison of dynamic modulus mastercurves with 6% and 8% air void
contents: (a) Case 1 and Case 2, (b) Case 5 and Case 6, (c) Case 3 and Case 4, and (d) Case 7
and Case 8.

47
 600 32000
Mixture: VA Mixture: VA
Gauge Length: 50 mm Gauge Length: 100 mm
500 Thickness: 38 mm Thickness: 50 mm
Temperature: 35C 30000
Temperature: -10C
Frequency: 0.05 Hz Frequency: 10 Hz
400 28000

|E*| (MPa)

|E*| (MPa)
300 26000

200 24000

100 Target Target

22000

0 20000
5.50 6.00 6.50 7.00 7.50 8.00 8.50 5.50 6.00 6.50 7.00 7.50 8.00 8.50
Air Void (%) Air Void (%)

13000 1500
Mixture: VA Mixture: VA
Gauge Length: 50 mm Gauge Length: 100 mm
Thickness: 50 mm Thickness: 38 mm
12000 Temperature: 10C Temperature: 35C
Frequency: 1 Hz 1300 Frequency: 1 Hz
|E*| (MPa)

11000

|E*| (MPa)
1100
10000

900
9000 Target Target

8000 700
5.50 6.00 6.50 7.00 7.50 8.00 8.50 5.50 6.00 6.50 7.00 7.50 8.00 8.50
Air Void (%) Air Void (%)

9000 18000
Mixture: NE Mixture: NE
Gauge Length: 50 mm Gauge Length: 100 mm
17000 Thickness: 38 mm
Thickness: 38 mm
8000 Temperature: 10C Temperature: -10C
Frequency: 10 Hz 16000 Frequency: 1 Hz
|E*| (MPa)

15000
|E*| (MPa)

7000
14000
6000 13000

12000
5000 Target Target
11000

4000 10000
5.50 6.00 6.50 7.00 7.50 8.00 8.50 5.50 6.00 6.50 7.00 7.50 8.00 8.50
Air Void (%) Air Void (%)

500 900
Mixture: NE Mixture: NE
Gauge Length: 50 mm Gauge Length: 100
Thickness: 50 mm mm Thickness: 50 mm
400 Temperature: 35C 800 Temperature: 35C
Frequency: 0.05 Hz Frequency: 1 Hz
|E*| (MPa)
|E*| (MPa)

300 700

200 600

100 Target 500 Target

0 400
5.50 6.00 6.50 7.00 7.50 8.00 8.50 5.50 6.00 6.50 7.00 7.50 8.00 8.50

Air Void (%) Air Void (%)

Figure 4-3. Variation of dynamic modulus values versus air void content.

48
4.6.3 Gauge Length
According to the findings from the ruggedness tests and the trend observed in the
dynamic modulus mastercurves shown in Figure 4-4, gauge length variation from 2 inches
(50.8 mm) to 4 inches (101.6 mm) did not seem to change the dynamic modulus values of the
mixtures greatly. This point cannot be concluded for certain, because the mixtures used in
this study did not contain large aggregate particles (i.e., 9.5 mm and 12.5 mm); therefore,
mixtures with larger aggregate particles should lead to more confident results. However, the
pavement thickness for the mixtures containing large aggregate particles (e.g., 25 mm or
larger) should be more than 150 mm; consequently, obtaining 150-mm height field cores is
possible for performing dynamic modulus tests in the axial loading mode.

30000 30000
VA-Case 1-Rep 1 VA-Case 3-Rep 1

25000 VA-Case 1-Rep 2 25000 VA-Case 3-Rep 2

VA-Case 7-Rep 1 VA-Case 5-Rep 1

20000 20000
|E*| (MPa)

|E*| (MPa)

VA-Case 7-Rep 2 VA-Case 5-Rep 2

15000 15000

10000 10000

5000 5000

(a) (b)
0 0
1E-06 1E-03 1E+00 1E+03 1E+06 1E-06 1E-03 1E+00 1E+03 1E+06
Reduced Frequency (Hz) Reduced Frequency (Hz)

20000 20000
NE-Case 2-Rep 1 NE-Case 4-Rep 1

NE-Case 2-Rep 1 NE-Case 4-Rep 2

15000 15000
NE-Case 6-Rep 1 NE-Case 8-Rep 1
|E*| (MPa)
|E*| (MPa)

NE-Case 6-Rep 2 NE-Case 8-Rep 2

10000 10000

5000 5000

(c) (d)
0 0
1E-06 1E-03 1E+00 1E+03 1E+06 1E-06 1E-03 1E+00 1E+03 1E+06
Reduced Frequency (Hz) Reduced Frequency (Hz)

Figure 4-4. Comparison of dynamic modulus mastercurves with 50.8 mm and 101.6 mm
gauge lengths: (a) Case 1 and Case 7, (b) Case 3 and Case 5, (c) Case 2 and Case 6, and (d)
Case 4 and Case 8.

49
4.6.4 Strain Levels
When comparing the mastercurves shown in Figure 4-5, the horizontal strain level
was found not to be a significant factor. In general, the statistical analysis in ASTM C1067
suggests that, except for one case of the VA mixture, the F-values are considerably lower
than 5.32, which means that strain amplitude variation from 40  to 60  did not
significantly impact the material stiffness. However, considering the average Poisson’s ratio
of 0.35 for asphalt material, a horizontal strain level of 60  may damage the specimen,
because the corresponding vertical strain will be close to 180 , which is above the strain
limit specified in AASHTO TP 62.

30000 30000
VA-Case 1-Rep 1 (a) VA-Case 3-Rep 1 (b)

25000 VA-Case 1-Rep 2 25000 VA-Case 3-Rep 2

VA-Case 7-Rep 1 VA-Case 5-Rep 1

20000 20000
|E*| (MPa)
|E*| (MPa)

VA-Case 7-Rep 2 VA-Case 5-Rep 2

15000 15000

10000 10000

5000 5000

0 0
1E-06 1E-03 1E+00 1E+03 1E+06 1E-06 1E-03 1E+00 1E+03 1E+06
Reduced Frequency (Hz) Reduced Frequency (Hz)

20000 20000
NE-Case 4-Rep 1 (c) NE-Case 2-Rep 1 (d)

NE-Case 4-Rep 2 NE-Case 2-Rep 2

15000 15000
NE-Case 6-Rep 1 NE-Case 8-Rep 1
|E*| (MPa)
|E*| (MPa)

NE-Case 6-Rep 2 NE-Case 8-Rep 2

10000 10000

5000 5000

0 0
1E-06 1E-03 1E+00 1E+03 1E+06 1E-06 1E-03 1E+00 1E+03 1E+06
Reduced Frequency (Hz) Reduced Frequency (Hz)

Figure 4-5. Comparison of dynamic modulus mastercurves with strain levels of 40  and 60
: (a) Case 1 and Case 7, (b) Case 3 and Case 5, (c) Case 4 and Case 6, and (d) Case 2 and
Case 8.

50
4.6.5 Thickness
Modulus calculations for circular specimens are based on the diameter-to-depth ratio.
However, the ruggedness study shows that the dynamic modulus was not affected by
changing the specimen thickness from 38 mm to 50 mm.

4.7 Summary and Conclusions

An IDT dynamic modulus ruggedness analysis was carried out to identify the
significant factors that can affect the test results and their tolerance ranges. The test
temperature, air void content, specimen thickness, gauge length, and horizontal strain level
are the factors evaluated in this study, based on the information available in current
literatures. This inter-laboratory study was conducted at NCSU and the NCAT to establish a
precision statement. According to the analysis results, air void content was the factor that was
found to be statistically significant, especially at -10C and 10C. However, the impact of air
void content on material stiffness (i.e., dynamic modulus values) at 35°C was found to be
statistically insignificant. In general, the other factors, i.e., specimen thickness, gauge length,
strain level, and test temperature, did not have a considerable impact on the dynamic
modulus values; however, the following limits were specified for dynamic modulus testing
based on the observations made in this study:
The temperature tolerance range of  1.0C seems adequate because the change in
dynamic modulus value was within the acceptable range; however, to maintain consistency
with other specifications, the tolerance range of  0.5C ( 1F) is recommended for the IDT
specification.
The horizontal strain level should be limited to 50  5 . Although the ruggedness
analysis did not show major change in the dynamic modulus values when the strain level
changed from 40  to 60 , displacement data quality was affected by the noise for strain
levels below 40 . On the other hand, testing at high strain levels (60  5 ), especially at
high temperatures, could damage the specimen because of the high vertical strain level.
An air void content tolerance of  0.5% should be reasonable because the dynamic
modulus variation was less than 10% for almost all the cases investigated.

51
 Changing the thickness of the specimen did not appear to have an impact on dynamic
modulus as long as correct coefficients are used in the biaxial linear viscoelastic solutions.
Therefore, the thickness of 38  3 mm is recommended for IDT samples, which is a practical
thickness for obtaining field cores from pavement sections.
The gauge length limit must be an appropriate representative volume element (RVE),
especially for mixtures that contain large aggregate particles; so, the gauge length of 101.6 
3 mm is recommended for IDT testing.

52
CHA
APTER 5 DETE
ERMINIING SPE
ECIMEN GEOME
ETRY OF
CYL
LINDRIC
CAL SPE
ECIMEN
NS FOR DIRECT
T TENSIION
FAT
TIGUE TESTING
T G OF AS
SPHALT RETE 2
T CONCR

5.1 Abstract
This chap
pter presents a specimen geometry sttudy of cylinndrical specim
mens used inn the
axial compression
n dynamic modulus
m testiing and the ddirect tensioon cyclic fatigue testing oof
asphaalt concrete using
u an asp
phalt mixturee performancce tester. Thhe current speecimen geom
metry
for diirect tension
n cyclic testin
ng is 100 mm
m in diameteer and 150 m
mm in heightt with a lineaar
variab
ble differenttial transducer (LVDT) gauge
g lengthh of 100 mm
m in the midddle of the
speciimen. In ordeer to use the displacemen
nt data for m
mechanistic ffatigue perfoormance
modeeling, specim
men failure must
m occur within
w the lenngth of the L
LVDT gaugee. However,
recen
nt experimen
nts using stiff
ff mixtures have
h shown tthat failure ooften occurs outside the
LVDT gauge length. This speecimen geom
metry study w
was conductted to determ
mine the
speciimen geomettry that enhaances the pro
opensity of thhe failure innside the gauuge length
witho
out sacrificin
ng the advan
ntage of the direct
d tensionn testing thaat provides uuniform stressses
and strains
s in the middle of th
he specimen
n. Viscoelasti
tic continuum
m damage finnite elementt
progrram (VECD--FEP++) anaalysis and laaboratory expperiments w
were performeed on cylinddrical
speciimens of diffferent geomeetries (i.e., different
d diam
meters and leengths) and different LV
VDT
gaugee lengths. Teest specimen
n diameters of
o 75 mm annd 100 mm aand specimenn heights of 130
mm and
a 150 mm were used in
i this study with three ddifferent gauuge lengths oof 70 mm, 755
mm, and 100 mm
m. Compariso
on of measured dynamicc modulus daata from varrious specim
men
geom
metries resultts in the speccimen of 100
0 mm in diam
meter and 1550 mm in heeight with thee
LVDT gauge length of 70 mm
m as the reco
ommended ggeometry. Based on the VECD-FEP++
simullation resultss and experim
mental resullts, the speciimen geomettry recommeended for thee

2
This chapter previiously was pu
ublished as: Lee,
L J.S., A. N Norouzi, and YY. R. Kim (2015). Determ mining
Speecimen Geom metry of Cylinndrical Specim
mens for Direect Tension Faatigue Testingg of Asphalt
Conncrete. ASTMM Journal of Testing
T and Evvaluation, In press.

53
direct tension cyclic testing is 100 mm in diameter and 130 mm in height with a 70-mm
gauge length. The recommended specimen geometry is applicable when the gyratory-
compacted specimen geometry is 150 mm in diameter and more than 178 mm in height.

5.2 Introduction
In order to characterize the properties of composite materials through laboratory
testing, the representative volume element (RVE) concept must be applied for the smallest
volume of a material that can represent the global properties of the material. In this study,
viscoelastic continuum damage mechanics coupled with dynamic modulus and direct tension
fatigue tests were applied to attempt to characterize the fatigue cracking performance of the
asphalt mixture using macroscale observations of the RVE.
For asphalt mixtures, the minimum ratio of the nominal maximum aggregate size
(NMAS) to the specimen diameter must be 1 to 4, and the minimum ratio of the specimen
diameter to the specimen height must be 1 to 2, according to ASTM D 3497. Therefore,
ASTM D 3497 recommends the cylindrical specimen size of 101.6 mm (4 inches) in
diameter and 203.2 mm (8 inches) in height for the determination of the dynamic modulus of
asphalt concrete. The NCHRP 9-19 project, Superpave Support and Performance Model
Management, evaluated specimen geometry and concluded that the RVE requirement for
axial compression tests, such as dynamic modulus and flow number tests, can be satisfied by
adopting the specimen geometry of 100 mm in diameter and 150 mm in height and a
frictionless end condition. The use of 0.3-mm thick latex membranes separated by silicone
grease at the bottom and top of the specimens can create frictionless boundaries (ends) during
axial compression tests.
The NCHRP 9-19 study also showed that the RVE requirement for axial tension tests,
such as constant crosshead rate direct tension monotonic tests and direct tension cyclic tests,
can be satisfied using the specimen geometry of 75 mm in diameter and 150 mm in height.
Because gluing the specimens to steel end plates causes end confinement at the ends of the
specimens during axial tension tests, the conservative diameter-to-height ratio of 1 to 2 was
recommended instead of the diameter-to-height ratio of 1.5 (100-mm diameter and 150-mm

54
height) that is used in the axial compression tests (Witczak et al. 2002). In recent studies, the
specimen geometry of 102 mm in diameter and 102 mm in height (the diameter-to-height
ratio of 1) was used to characterize fatigue cracking of asphalt mixture using a repeated
direct tension test (Luo et al. 2013). The reason to choose the specimen geometry of 102 mm
x 102 mm is not addressed in the paper. Also, the specimen geometry of 38 mm in diameter
and 110 mm in height (the diameter-to-height ratio of 2.9) is successfully determined to use
field cores for the fatigue characterization (Li and Gibson 2013).
The NCHRP 9-29 report, Simple Performance Tester for Superpave Mix-Design,
indicates that the asphalt mixture performance tester (AMPT) can be used not only for the
dynamic modulus and flow number tests for the mechanistic-empirical pavement design
program (or MEPDG, a product of NCHRP 1-37A), but also for direct tension cyclic fatigue
tests, such as the one standardized in AASHTO TP 107 (Bonaquist and Christensen 2008).
For these NCHRP studies, all of the cylindrical specimens were cored and cut from
gyratory-compacted mixture specimens that were 150 mm in diameter and 160 mm to more
than 180 mm in height in order to remove high air void content regions along the
circumferences and ends of the gyratory-compacted specimens. Coarse-graded mixtures
sometimes require more height in the gyratory compacted specimens in order to produce
uniform air voids for the specimen height of 150 mm and to ensure smooth, uniform ends
with minimal or no surface irregularities, in accordance with AASHTO PP 60.
With regard to direct tension fatigue testing, a recent experimental problem was
encountered that involved failure at the ends of the asphalt specimens (hereinafter called end
failure). During testing, it was observed that as more and more material was cut from the top
and bottom of the gyratory-compacted specimens, the likelihood of failure in the middle of
the specimen (hereinafter called middle failure) greatly increased. This outcome was due to
the air void gradient along the diameter and height of the gyratory-compacted asphalt
mixture specimens. A study (Chehab et al. 2000) found that the air void content is high near
the mold wall and top and bottom of the specimen, and the average difference in the air void
content between the cored and cut specimens that are 75 mm x 150 mm (75 mm diameter and
150 mm height) and the gyratory-compacted specimens that are 150 mm x 175 mm is about

55
2.5 percent. The uniformity of the air void distribution throughout the test specimens is a
much more critical factor in direct tension tests than in axial compression tests. Higher air
void contents at the ends of the specimen (due to the removal of the ends of gyratory-
compacted specimens of insufficient length) cause end failure in direct tension tests, whereas
in axial compression tests they may cause more deformation, but not catastrophic failure.
Therefore, fabricating shorter test specimens that are cored and cut from taller gyratory-
compacted specimens can produce test specimens that have more uniformly distributed air
voids such that middle failure occurs within the gauge length of the linear variable
differential transformer (LVDT) in direct tension tests.
In this study, it is assumed that the end failure occurred due to two factors: 1) high air
void content and 2) stress concentration (end confinement) at the end of the specimens.
Regardless of the high air void content at the end of the specimen, the stress concentration at
the end of the specimen, which was due to being glued to the end plates as required for direct
tension tests, may have caused the end failure, depending on the specimen geometry.
Therefore, the VECD-FEP++ program was used to investigate the damage initiation and
propagation according to the stress concentration for each specimen geometry in terms of the
reduction in pseudo stiffness. The simulation results show how the end confinement affected
the middle or end failure, depending on the specimen geometry used during the direct tension
fatigue testing.
In order to propose a new specimen geometry that can meet the RVE requirement and
ensure middle failure, laboratory mechanical tests and finite element analysis were conducted
using different specimen geometries (i.e., different diameters and lengths) and different
LVDT gauge lengths. For the mechanical tests, dynamic modulus and direct tension tests
were conducted to characterize the linear viscoelastic (LVE) and viscoelastic continuum
damage (VECD) properties, respectively.

5.3 Research Objectives and Methods

The objectives of this chapter are to: (1) determine the appropriate specimen
geometry for dynamic modulus tests that are applicable for use with the AMPT using

56
laboratory tests and statistical analysis and (2) determine the specimen geometry that best fits
the RVE concept and causes middle failure within the LVDT gauge through laboratory
testing. To accomplish these objectives, the following methods are used in this study: (1)
investigate the confinement effect of viscoelastic materials during direct tension fatigue
testing using finite element analysis and (2) verify the recommended specimen geometry for
the improved middle failure using laboratory testing of various asphalt mixtures.

5.4 Theoretical Background

The LVE and VECD properties are used in constitutive equations that explain the
relationship between stress and strain with damage. These properties can be used in
mechanistic pavement analysis models, such as finite element analysis, along with a proper
fatigue failure criterion to predict the fatigue performance of asphalt pavements.

5.5 Linear Viscoelastic Properties

Asphalt mixtures are viscoelastic materials that are sensitive to temperature and
loading frequency. The temperature and loading frequency dependency is captured in the
dynamic modulus (|E*|) mastercurve using the time-temperature superposition (t-TS)
principle, as described in AASHTO TP 62, TP 79, and PP 61. The relaxation modulus and
creep compliance are then obtained by applying theoretical relationships between the
dynamic modulus and these properties.

5.6 Viscoelastic Continuum Damage Properties

In order to model the viscoelastic behavior of asphalt materials with growing damage,
the elastic-viscoelastic correspondence principle, t-TS principle, and Schapery’s work
potential theory are used in the VECD model. First, the elastic-viscoelastic correspondence
principle allows elastic solutions to be applied for viscoelastic materials by incorporating the
pseudo strain concept (Kim et al.1995). Second, the t-TS principle allows the derivation of
the constitutive relationship between stress and strain with growing damage using the t-T
shift factor derived from the construction of the LVE dynamic modulus mastercurve (Chehab
2002). Finally, Schapery’s work potential theory uses the concept of an internal state variable

57
that quantifies any microstructural changes based on the change in stiffness of the material
(Underwood et al. 2010). The main output of the VECD model is the so-called damage
characteristic curve that describes the deterioration of the material integrity that is
represented by pseudo stiffness, C, as the damage, S, grows. Here, the material integrity of
the pseudo stiffness parameter is derived based on continuum damage mechanics, which
regards a damaged body with some stiffness as an undamaged body with reduced stiffness.
The damage parameter (D) represents the damaged condition of a material as an effective
stress or modulus value. By using this damage parameter, a fundamental constitutive model
can be applied to the damaged material, as expressed by Equation (5.1). The viscoelastic
continuum damage theory adopts pseudo stiffness (C) as the damage parameter, which is a
function of an internal state variable (S), as shown in Equations (5.2) through (5.5).

 / (1  D )   
    (D = 0: no damage, D = 1: full damage) (5.1)
E E E (1  D ) E
  f (S ) E R R  C (S )  E R   R (5.2)

  C (S ) R , C (S )  (5.3)
R
1
W R  C (S )( R )2 (5.4)
2

 W R 
S     (5.5)
 S 
where
= strain,
D = damage parameter,
 = stress,
 ' = effective stress,
E ' = effective modulus,
E = modulus,
 = pseudo strain,
f(S) = effective modulus,

58
 ER = reference modulus,

W R = total dissipated pseudo strain energy,

S = internal state variable (damage), and
 = damage evolution rate.

The VECD model is used for direct tension monotonic testing, whereas the simplified
viscoelastic continuum damage (S-VECD) model is designed to be used with direct tension
cyclic tests in order to reduce computational time and to overcome AMPT limitations.
Details regarding the S-VECD model can be found in the literature (Daniel and Kim 2002,
Underwood et al. 2012).

5.7 Failure Criteria for Fatigue Performance Prediction

Traditionally, the fatigue failure criterion is defined as the point where the modulus
value reduces to 50% of the initial modulus value (Reese 1997). This criterion is known to
underestimate the fatigue life of polymer-modified asphalt mixtures. The change in slope of
the phase angle time history has been found to be a reliable failure criterion regardless of
mixture type. However, the phase angle reduction after the material’s failure is related to the
distortion of the stress and strain time histories due to macrocrack development and thus
cannot be predicted from constitutive relationships that are based on continuum assumptions,
such as those on which the VECD model is based.
Another failure criterion that was introduced by Hou et al. employs the pseudo
stiffness value at failure (Cf). Hou et al. introduced Cf-based failure envelopes that depend on
mixture type, temperature, and loading frequency. However, the scatter in the Cf values was
quite extensive, and multiple fatigue tests had to be performed to establish the failure
envelope.
Recently, a new failure criterion was developed to predict fatigue performance using
the S-VECD model. This failure criterion is the power-law relationship between the average
of dissipated pseudo strain energy rate (Gᴿ) and the number of cycles to failure (Nf) at the

59
peak phase angle (i.e., when the slope of the phase angle time history changes). The Gᴿ is
calculated as shown in Equation (5.6).
Nf

W
R
R
W
GR   0 2 (5.6)
Nf Nf

where
G R = dissipated pseudo strain energy rate, and
f = number of cycles to failure.

This new method is independent of mode of loading and temperature and, therefore,
can be determined from direct tension cyclic testing at a single temperature, at a single
loading frequency, and at three to four different strain amplitudes (Sabouri et al. 2014).

5.8 Experimental Program

Dynamic modulus and direct tension cyclic tests were carried out for experimental
verification using different specimen geometries and LVDT gauge lengths. Analysis of
variance (ANOVA) was conducted to determine whether the LVE and fatigue damage
properties measured using the different specimen geometries and LVDT gauge lengths were
statistically significant or not (Ott and Longnecker 2010). Finally, the new specimen
geometry that increases the propensity for middle failure was investigated by performing
direct tension cyclic tests using a broader range of asphalt mixtures that were associated with
end failure.

5.8.1 Materials
This study used a Superpave 9.5 mm mixture consisting of granite aggregate, 19%
reclaimed asphalt pavement (RAP), and 5.2% PG 70-22 virgin binder. The binder from the
RAP was found to be PG 64-22 and 1.2% by weight of the RAP mixture. This mixture was
selected for this study because direct tension cyclic tests of this mixture using the 75 mm x
150 mm specimens resulted in mostly end failure. The reason this mixture was prone to end
failure is not clear; however, it provides a good testing ground to identify specimen

60
geometries that yield middle failure. Finally, 28 different asphalt mixtures based on the
Superpave 9.5 mm mix design (four different specimen geometries: 75 mm x 150 mm, 75
mm x 130 mm, 100 mm x 150 mm, and 100 mm x 130 mm, times seven test specimens; and
three replicates for the dynamic modulus tests and four replicates for the direct tension
fatigue tests) that were prone to end failure were tested to confirm the improved propensity
for middle failure using the recommended specimen geometry. Those test results are
discussed in the section, Determination of Specimen Geometry for Linear Viscoelastic
Properties and Determination of Specimen Geometry for Damage Characterization. After
determining the proposed specimen geometry, additional direct tension fatigue tests were
conducted to check the improved middle failure for 31 different asphalt mixtures using the
different mix designs described in the section, Investigation of Improved Middle Failure for
the New Specimen Geometry Using Various Mixtures.

5.8.2 Specimen Fabrication

All test specimens were fabricated using a Servopac Superpave gyratory compactor
manufactured by IPC Global. Because a high air void content in the top and bottom of the
cylindrical specimens is critical for the failure modes of middle or end failure, the maximum
gyratory specimen height of 178 mm, which could be fabricated using the Servopac gyratory
compactor, was used. Therefore, all of the test specimen geometries (75 mm x 150 mm, 75
mm x 130 mm, 100 mm x 150 mm, and 100 mm x 130 mm) used in this study were cored
and cut from 150 mm x 178 mm gyratory specimens.

5.8.3 LVDT Gauge Length

Different LVDT gauge lengths of 70 mm, 75 mm, and 100 mm that correspond to the
specimen geometries specified in NCHRP 9-19, AASHTO TP 79, and AASHTO TP 107
were used to run the dynamic modulus and direct tension fatigue tests.

5.8.4 Dynamic Modulus Tests

(NCHRP 9-19, TP 62 and TP 79). Dynamic modulus tests were conducted using
cylindrical specimens with the following four geometries: (1) 75 mm x 150 mm and (2) 75
mm x130 mm with 75-mm LVDT gauge length tested in tension-compression cyclic stress-

61
controlled mode, and (3) 100 mm x 150 mm and (4) 100 mm x 130 mm with 70-mm LVDT
gauge length tested in compressive cyclic stress-controlled mode, according to AASHTO TP
79 and PP 61. In accordance with AASHTO TP 79, the tests were completed for all specimen
geometries at 5°C, 20°C, 40°C, and 54.4°C and at 25 Hz, 10 Hz, 5 Hz, 1 Hz, 0.5 Hz, and 0.1
Hz. The load levels were determined by a trial and error process so that the resulting strain
level was between 50 and 75 microstrains. It should be noted that the AASHTO TP 79
procedure does not include the test condition at -10°C due to limitations of the AMPT
temperature chamber.

5.8.5 Direct Tension Tests

(NCHRP 9-19 and AASHTO TP 107). Direct tension tests were conducted using
cylindrical specimens with the following four geometries: (1) 75 mm x 150 mm and (2) 75
mm x130 mm with 100-mm LVDT gauge length tested in monotonic tension with constant
crosshead rate, and (3) 100 mm x 150 mm and (4) 100 mm x 130 mm with 75-mm LVDT
gauge length in cyclic tension with constant crosshead displacement amplitude according to
AASHTO TP 107. In this study, the constant strain loading rate of 0.0004 strain per second
that was used for the direct tension monotonic tests and the cyclic loading frequency of 10
Hz with different amplitudes of actuator displacement caused fracture at about 1,000, 10,000,
and 100,000 cycles for the direct tension cyclic tests at an intermediate temperature of 19°C.

5.9 Determination of Specimen Geometry and LVDT Gauge Length for

Linear Viscoelastic Properties
Figure 5-1 (a) and (b) show the experimentally measured dynamic modulus
mastercurves as a function of reduced frequency in semi-log scale and log-log scale,
respectively. The semi-log scale can better demonstrate the modulus values at high frequency
loading and low temperature conditions, and the log-log scale can more clearly distinguish
the modulus values at low frequency loading and high temperature conditions. Specimens
with 75 mm diameter were tested in tension-compression stress-controlled mode with 75 mm
LVDT gauge length, whereas 100 mm diameter specimens were subjected to axial
compressive loading in stress-controlled mode with 70 mm gauge length as specified in

62
AASHTO TP 79 and PP 61. Figure 5-1 (c) presents the phase angle results as a function of
reduced frequency in semi-log scale.

30000 100000
75x130 75x150 75x130 75x150
25000
100x130 100x150 100x130 100x150
20000 10000
|E*| (MPa)

|E*| (MPa)
15000

10000 1000

5000
(a) (b)
0 100
1E-08 1E-06 1E-04 1E-02 1E+00 1E+02 1E+04 1E-08 1E-06 1E-04 1E-02 1E+00 1E+02 1E+04

Reduced Frequency (Hz) Reduced Frequency (Hz)

45

40 75x130 75x150
Phase Angle (deg)

35 100x130 100x150

30

25

20

15

10 (c)
5

0
1E-08 1E-05 1E-02 1E+01 1E+04

Reduced Frequency (Hz)

Figure 5-1. Experimentally measured dynamic modulus mastercurves: (a) in semi-log scale,
(b) in log-log scale, and (c) phase angle mastercurves.

Using the experimental data, one-way ANOVA was conducted to verify whether the
specimen geometry of 100 mm x 130 mm or 100 mm x 150 mm leads to different moduli
values than the other geometries of 75 mm x 130 mm and 75 mm x 150 mm at 5%
significance level. If the p-value was less than the significance level, the assumption that all
the specimen geometries have the same dynamic modulus values at different reduced loading
frequencies and pseudo stiffness values at different damage states was rejected.
The ANOVA results are presented in Table 5-1. The temperature and frequency
combinations that resulted in p-values less than 5% are shaded in grey. In general, the

63
specimen geometry effect becomes greater as the temperature and loading frequency become
lower. In addition to the specimen geometry effect, it should be noted that the p-values
presented in Table 1 are affected by other test conditions, such as slight changes in the test
temperature, specimen air void content, and strain level (nonlinear viscoelastic behavior).
Table 5-1 (a) shows that the specimen geometries of 75 mm x 130 mm, 75 mm x 150 mm,
and 100 mm x 150 mm have statistically the same LVE properties, because all the p-values
are greater than the 5% significance level. Table 5-1 (b) shows that the geometry of 100 mm
x 130 mm produces statistically different LVE properties than the other geometries at 5°C
and at 0.5 Hz and 0.1 Hz. These experimental observations suggest that the specimen height
of 130 mm is too short. Among the three geometries of 75 mm x 130 mm, 75 mm x 150 mm,
and 100 mm x 150 mm, 100 mm x 150 mm geometry is preferred because it has a greater
chance of satisfying the RVE requirements for large aggregate sizes. In conclusion, the
findings from this study support the use of the specimen geometry specified in AASHTO TP
79 and PP 61, i.e., 100 mm diameter, 150 mm height, and 70 mm gauge length, in measuring
the dynamic modulus of asphalt mixtures.

Table 5-1 One-way ANOVA Results for Dynamic Modulus Tests at 5% Significance Level
Dynamic Modulus Tests (p-value), α = 0.05
(a)
75x130, 75x150, 100x150 25 Hz 10 Hz 5 Hz 1 Hz 0.5 Hz 0.1 Hz
(mm)
5°C 0.1947 0.294 0.4283 0.8164 0.7036 0.3841
20°C 0.4908 0.3434 0.2772 0.2375 0.2538 0.3253
40°C 0.2681 0.2796 0.2826 0.2612 0.2428 0.1931
54°C 0.3417 0.2878 0.2489 0.1747 0.1497 0.1048
(b)
75x130, 75x150, 100x130 25 Hz 10 Hz 5 Hz 1 Hz 0.5 Hz 0.1 Hz
(mm)
5°C 0.2532 0.5468 0.7237 0.126 0.0499 0.0243
20°C 0.1239 0.0806 0.0736 0.1127 0.1554 0.3187
40°C 0.1405 0.2297 0.3083 0.4154 0.398 0.2688
54°C 0.2872 0.434 0.4652 0.2982 0.2149 0.0945

64
5.10 Viscoelastic Continuum Damage Finite Element Program in C++
Because the ends of the test specimens need to be glued to steel end plates for direct
tension tests, the effects of end confinement should be taken into account in order to
determine the proper specimen geometry and LVDT gauge length. In this study, the effects
of specimen geometry and gauge length on damage characteristic curves were investigated
using finite element analysis and the experimental measurements obtained from the direct
tension cyclic fatigue tests.
For the finite element analysis, the VECD finite element program in C++ language,
referred to as VECD-FEP++, developed at North Carolina State University, was used to
implement the VECD material model in the finite element structural model. VECD-FEP++ is
a two-dimensional axisymmetric structural analysis tool that can determine the stress and
strain values at any nodal point in the specimen structure with proper consideration of
accumulated damage in the specimen. As the damage accumulates, the pseudo stiffness (C)
also can be calculated to predict the cracking potential throughout the entire cylindrical
specimen (Kim et al. 2004).
The end confinement was simulated by fixing the boundary nodal points in the x-axis
and z-axis directions at the bottom of the specimens and the boundary nodal points in the x-
axis direction at the top of the specimens. Because the purpose of the VECD-FEP++ analysis
is to simulate the experimental setup in direct tension testing, axial strains are calculated by
dividing the deformation between two nodal points that represent the LVDT gauge points by
the LVDT gauge length, as is done in actual experiments. Also, the stress in the axial
direction is calculated by averaging the stresses at all the nodal points at the top of the
specimen. Figure 5-2 (a) and (b) show the symmetric vertical strain and stress distribution
about the center in the z-axis direction through the cylindrical specimen. Figure 5-2 (c)
describes the methods used for the calculation of the stresses and strains in the VECD-
FEP++ simulations.

65
 Figu
ure 5-2. (a) Vertical
V strain contour (E
Ezz), (b) verrtical stress ccontour (Szzz), and (c) strress
and strain measurem ments to simuulate laboratoory tests.

Before thee VECD-FE

EP++ analysiis was used ffor the speciimen geomettry study, the
VECD
D-FEP++ prrogram was tested for co
onvergence aand accuracyy. The mesh size and
analy
ysis time step
p needed forr VECD-FEP
P++ analysiss were determ
mined by obbserving the
conveergence of th
he stress and
d strain outpu
uts simulatedd from the ddynamic moddulus and
mono
otonic direct tension fatig
gue tests. Th
he regular quuadrilateral ccell size of 5 mm and tim
me
step of
o 0.01 secon
nd were seleected based on
o the conveergence studdy. The seconnd step in the
study
y was to verify the accuraacy of the VECD-FEP++
V + analysis. T
This task waas accomplished
by ussing the expeerimentally measured
m maaterial propeerties (i.e., |E
E*| and C vs. S relationshhip)
as inp
puts to the VECD-FEP+
V ++ and then checking
c if th
the resultant simulated bbehavior mattched
the ex
xperimentallly measured behavior.
One of the material prroperties neeeded for the VECD-FEP
P++ analysis, but not
measured in the TP
T 107 direcct tension tessting, is the P
Poisson’s rat
atio. The Poissson’s ratio of
asphaalt mixtures changes as a function off temperaturre and loadinng frequencyy. The FHWA
A-
NJ-20
008-004 proj
oject develop
ped an equatiion that expllains the relaationship bettween Poissoon’s
ratio and the dynaamic modulu
us (Maher an
nd Bennert 22008). This rrelationship is shown in

66
Equation (5.7). In this study, Equation (1) was used to calculate the Poisson’s ratios of the
asphalt mixtures from the dynamic modulus tests, which in turn were used as input values for
the VECD-FEP++.
ν = -0.0505 × LN(|E*|) + 0.864 (5.7)

where ν is Poisson’s ratio and |E*| is the dynamic modulus value in psi.
Dynamic modulus test was simulated using the displacement-controlled mode, which
is performed by controlling the displacement of the top end plate, with the zero-minimum
displacement time history. Figure 5-3 shows that the dynamic modulus and direct tension
monotonic test results that were simulated using VECD-FEP++ are in good agreement with
the results obtained from the laboratory tests. Figure 5-3 (a) and (b) present a comparison of
the dynamic modulus values in semi-log scale and log-log scale, respectively. Figure 5-3 (c)
and (d) present the simulation results for pseudo stiffness (C) and damage (S), and the
combined damage characterization curve, respectively. Some discrepancies can be observed
in the predictions of the C and S values individually in Figure 5-3 (c), but the damage
characteristic curve shown in Figure 5-3 (d) coincides well with the experimental results.
The VECD-FEP++ program does not consider the internal structural effects that may
be related to the heterogeneity of an asphalt mixture, but it is able to capture the end
confinement effect of the viscoelastic material properties using continuum damage
mechanics according to the different specimen geometries. In a future study, it may be
interesting to perform microstructural analysis to evaluate the combined effect of end
confinement and air void gradient on the damage propagation that is related to the failure
mode (middle failure or end failure).

67
 30000 100000
Measurement of 75x150GL100
Simulation of 75x150GL100
10000
20000

|E*| (MPa)
|E*| (MPa)

1000

10000 Measurement of 75x150GL100

100 Simulation of 75x150GL100

(a) (b)
0 10
1.E-08 1.E-06 1.E-04 1.E-02 1.E+00 1.E+02 1.E+04 1.E-08 1.E-06 1.E-04 1.E-02 1.E+00 1.E+02 1.E+04

Reduced Frequency (Hz) Reduced Frequency (Hz)

1.0 700000 1
Measured Data (75x150GL100)

Pseudo Stiffness (C)

600000
Pseudo Stiffness (C)

0.8 0.8 Simulataion (75x150GL100)

500000

Damage (S)
GL: Gauge Length

0.6 S 0.6
C 400000

0.4 300000 0.4

Measuremnt of C 200000
Simulation of C 0.2
0.2
Measurement of S 100000
Simulation of S
(c) (d)
0.0 0 0
0 1 2 3 4 5 0 200000 400000 600000 800000 1000000
Loading Time (Sec.) Damage (S)

Figure 5-3. Simulation results of the dynamic modulus tests: (a) in semi-log scale, (b) in log-
log scale, (c) pseudo stiffness and damage, and (d) damage characteristic curves.

5.11 Determination of Specimen Geometry and LVDT Gauge Length for

Damage Characterization
Damage propagation is investigated to find the failure location for the different
specimen geometries. Direct tension monotonic tests were simulated using VECD-FEP++ to
evaluate the effects of end confinement on the constitutive behavior and damage evolution of
the specimens. Also, the direct tension cyclic fatigue tests were performed on specimens with
different geometries in accordance with AASHTO TP 107.
Figure 5-4 (a) and (b) present the simulation and the experimental results,
respectively. It is noted that the 70-mm gauge length was used in all the direct tension testing
shown in Figure 5-4 (b). Both the VECD-FEP++ simulation results and the experimental
measurements indicate good agreement among the damage characteristic curves from the

68
different specimen geometries. However, end failures were observed for all the 150-mm tall
specimens (regardless of diameter), whereas the 130-mm tall specimens exhibited middle
failures. The arrows in Figure 5-4 (b) indicate the failure points for the different specimen
geometries. It is noted that, although the 150-mm tall specimens experienced end failure,
their damage characteristic curves nonetheless collapse with those of the 130-mm tall
specimens with middle failure. ANOVA tests were performed for the experimentally
determined damage characteristic curves shown in Figure 5-4 (b) at the C values of 0.4, 0.5,
0.6, 0.7, and 0.8. According to the ANOVA results presented in Table 2, the p-values at the
pseudo stiffness levels of 0.8, 0.7, 0.6, 0.5, and 0.4 are greater than 5% significance level,
indicating that the damage characteristic curves for the studied specimen dimensions are
statistically identical. Also, the experimental results presented in Figure 5-4 (b) show that the
shapes of the damage characteristic curves are almost identical. The ANOVA results shown
in Table 2 support the good agreement visually shown in Figure 5-4 (b).

1 1
75x150GL100mm 75x150GL75mm
Pseudo Stiffness (C)

75x130 75x150
75x130GL100mm 75x130GL75mm
Pseudo Stiffness (C)

0.8 0.8
75x110GL100mm 75x110GL75mm 100x130 100x150

100x150GL70mm 100x130GL70mm
0.6 0.6
75x150
GL: Gauge Length
100x150

0.4 0.4
75x130

0.2 0.2

(a) (b) 100x130

0 0
0 200000 400000 600000 800000 1000000 0 200000 400000 600000 800000 1000000

Damage (S) Damage (S)

Figure 5-4. Damage characteristic curves for different specimen geometries from: (a) VECD-
FEP++ simulations and (b) experimental measurements.

It is noted that the fatigue performance predictions of asphalt mixtures and pavements
involve not only stiffness (how the material deforms) and the damage characteristic curve
(how the damage evolves), but also the failure criterion (the point at which the damage
reaches a critical value). The experimental data shown in Figure 5-4 (b) were used to

69
determine the two VECD-based failure criteria: the Cf-based failure criterion and the GR-
based failure criterion. The results are shown in Figure 5-5, which clearly shows that end
failures cause significant errors for both failure criteria. The finite element simulations and
experimental measurements shown in Figure 5-4 suggest that, although the end failures in the
direct tension cyclic fatigue tests do not affect the damage characteristic curves, the error in
the failure criteria shown in Figure 5-5 is too significant to accept, thereby indicating the
importance of determining a specimen geometry that increases the propensity for middle
failure without being affected by the end effects.

0.40 10000
End Failure
(a) End Failure (b) 100-130 (Middle Failure)
Pseudo Stiffness at Peak

0.35
100-150 (End Failure)
Phase Angle (Cf)

0.30 1000 75-130 (Middle Failure)

Middle Middle
Failure Failure 75-150 (End Failure)
0.25

GR
0.20 100

0.15

0.10 10

0.05

0.00 1
75x130 75x150 100x130 100x150 1.E+03 1.E+04 1.E+05 1.E+06
Specimen Geometry Nf (Cycle)

Figure 5-5. Effect of end failure on two failure criteria: (a) Cf based on peak phase angle and
(b) Gᴿ vs. Nf.

Table 5-2 One-way ANOVA Results for Direct Tension Cyclic Tests at 5% Significance
Cyclic Direct Tension Tests (p-value), α = 0.05
19°C/10 Hz C = 0.8 C = 0.7 C = 0.6 C = 0.5 C = 0.4
75x130, 75x150, 100x150 (mm) 0.6527 0.7147 0.7446 0.9099 0.9976
75x130, 75x150, 100x130 (mm) 0.5159 0.5664 0.3913 0.3091 0.0574

5.12 Specimen Geometry and Failure Location

Investigating damage propagation by observing where potential cracks initiate,
propagate, and cause failure allows the effects of the different specimen geometries to be
evaluated. Figure 5-6 (a) and (b) show that initial damage (indicated by lower C values)

70
occurs at the top and bottom-right edges of the specimen and then immediately starts again at
the center of the specimen. When tension cyclic loading is applied, the initial damage
propagates from the center of the specimen toward the middle of the specimen surface, and
then the specimen finally fractures in the middle of the specimen. This pattern of damage
propagation is almost identical for specimen geometries of both 75 mm x 150 mm and 75
mm x 130 mm.
Figure 5-6 (c) and (d) present the 100-mm diameter specimen geometries and indicate
similar damage propagation and patterns to those of Figure 5-6 (a) and (b) for the 75-mm
diameter specimens. The 100 mm x 130 mm specimen is least similar to the other specimens.
The shape of the initial damage in the center of the specimen and the failure areas at the right
edge are larger than for the other specimens. The simulation results indicate that the
specimen geometry of 100 mm x 130 mm is the most likely to cause middle failure, as
evidenced by the damage propagation from the center of the cylindrical specimen to its wider
range at the edge. Therefore, the simulation results used in this study indicate that all
specimens fracture in the middle as long as other mixture factors (such as air void
distribution) remain constant over the specimen diameter and height. The results indicate also
that the 100 mm x 130 mm geometry provides the greatest propensity for middle failure
among the geometries investigated in this study.
It is noted that the VECD-FEP++ simulations were not affected by the air void
gradients that were present in the experimental investigation because the same material
properties were used throughout the entire specimen. Therefore, the greater damage in the
middle of the specimen seen from the VECD-FEP++ simulations is due to the complex
constitutive behavior of the material with growing damage under the given boundary
conditions.

5.13 Investigation of Improved Middle Failure for the New Specimen

Geometry Using Various Mixtures
The finite element analysis and experimental test results indicate that the specimen
height of 130 mm increases the chance of middle failure. In order to verify the improved

71
perceentage of mid
ddle failure using
u specim
mens with a height of 1330 mm, four graduate
studeents tested 31
1 asphalt mix
xtures with different
d chaaracteristics,, such as diffferent NMAS,
asphaalt binder performance grades
g (PGs), percentagees of RAP ussed in three ddifferent reggions
(Man
nitoba, North
h Carolina, and
a Alabamaa), and warm
m mix asphallt (WMA) teechnologies.
The test results in Tab
ble 5-3 indiccate that the 100 mm x 150 mm geom
metry yields a
total average of middle
m failurre of 32 perccent. In particcular, the miixtures with RAP and sooft
bindeers had signiificant end faailure probleems. The new
w geometry of 100 mm x 130 mm
impro
oved the pro
opensity for middle
m failurre to a total average of 990%, clearly validating tthe
findin
ngs from thee finite elemeent analysis and experim
mental investtigation.

Figu
ure 5-6. Dam
mage propagaation contou
urs of potentiial cracking (pseudo stiff
ffness, C) forr: (a)
75 mm
m x 150 mm m, (b) 75 mm
m x 130 mm m, (c) 100 mmm x 150 mm m, and (d) 100 mm x 130 mm
speciimen geome tries.

72
 Table 5-3 Experimental Verification of New Specimen Geometry for Improved Middle Failure
Direct Tension Cyclic Fatigue Tests 100 mm x 150 mm Geometry 100 mm x 130 mm Geometry
Number of Number of Number of Number of
Binder % Middle % Middle
Mixture Information NMAS % RAP Middle Total Middle Total
PG Failure Failure
Failures Specimens Failures Specimens
Control-HMA 9,5 mm PG 76-22 0 3 5 60 3 3 100
Control-HMA 19 mm PG 76-22 0 1 4 25 1 2 50
Control-HMA 19 mm PG 67-22 0 1 3 33 3 3 100
Open-Graded
9.5 mm PG 76-22 0 - - - 3 3 100
Friction Course
WMA-Foam 9.5 mm PG 76-22 0 2 6 33 3 4 75
National WMA-Foam 19 mm PG 76-22 0 2 6 33 3 3 100
Center for WMA-Foam 19 mm PG 67-22 0 1 4 25 3 3 100
Asphalt
WMA-Evotherm 9.5 mm PG 76-22 0 0 3 0 4 4 100
Technology
WMA-Evotherm 19 mm PG 76-22 0 3 3 100 3 3 100
Test Track
in Auburn WMA-Evotherm 19 mm PG 67-22 0 1 3 33 3 3 100
RAP 9.5 mm PG 67-22 50 0 4 0 3 3 100
RAP 19 mm PG 67-22 50 1 4 25 3 3 100
RAP 19 mm PG 67-22 50 2 4 50 3 4 75
RAP+Foam 9.5 mm PG 67-22 50 0 4 0 2 3 67
RAP+Foam 19 mm PG 67-22 50 0 4 0 3 3 100
RAP+Foam 19 mm PG 67-22 50 1 3 33 3 3 100
Control-HMA 16 mm PG 58-28 0 0 6 0 5 5 100
Control-HMA 16 mm PG 52-34 35 - - - 4 5 80
WMA-Advera 16 mm PG 58-28 0 0 2 0 5 6 83
Manitoba WMA-Advera 16 mm PG 52-34 35 - - - 4 4 100
Institute of WMA-Evotherm 16 mm PG 58-28 0 0 2 0 4 4 100
Transportation WMA-Evotherm 16 mm PG 52-34 35 - - - 4 4 100
Field WMA-Sasobit 16 mm PG 58-28 0 0 2 0 5 5 100
Construction WMA-Sasobit 16 mm PG 52-34 35 -- - - 5 5 100
In Canada HMA-0% RAP 12.5 mm PG 58-28 0 0 2 0 4 4 100
HMA-15%RAP 12.5 mm PG 58-28 15 - - - 4 4 100
HMA-50% RAP 12.5 mm PG 58-28 50 - - - 4 5 80
HMA-50% RAP 12.5 mm PG 52-34 50 - - - 5 7 71
Control-HMA 9.5 mm PG 70-22 19 6 14 43 3 4 75
North Carolina WMA-Foam 9.5 mm PG 70-22 19 7 15 47 3 4 75
WMA-Evotherm 9.5 mm PG 70-22 19 6 13 46 3 4 75
Average 32 Average 90

73
5.14 Conclusions
The effects of different specimen geometries and LVDT gauge lengths on the
dynamic modulus, damage characteristic curve, and failure location were investigated using
VECD-FEP++ simulations and an experimental program. The results indicate that the
specimen geometry and gauge length (100-mm diameter, 150-mm height, and 70-mm gauge
length) specified in AASHTO TP 79 and PP 61 yield dynamic modulus values that are in
good agreement with the results from other specimen geometries and meet conventional RVE
requirements. Regarding direct tension cyclic fatigue testing, the specimen geometry of 100
mm in diameter and 130 mm in height leads to statistically the same damage characteristic
curves as the other geometries investigated and increases the propensity for middle failure.
This finding is supported by the VECD-FEP++ simulations and the experimental program
that involved a total of 31 different asphalt mixtures. Based on these observations, the
authors recommend using 100 mm x 130 mm specimens cored and cut from gyratory-
compacted specimens that are 150 mm in diameter and greater or equal to 178 mm in height
for direct tension fatigue testing.

74
CHA
APTER 6 MECHANIST
TIC EVA
ALUATIO
ON OF F
FATIGUE
E
CRA
ACKING
G IN ASP ENTS3
PHALT PAVEME
P

6.1 Abstract
Over the last
l several decades,
d sign
nificant reseearch has beeen conductedd to predict tthe
fatigu
ue cracking performance
p e of asphalt pavements.
p T
These researrch efforts haave resultedd in
various test meth
hods and mod
dels for fatig
gue performaance predictions. Recenttly, the
simpllified viscoeelastic contin
nuum damag
ge (S-VECD)) model wass developed aas an efficieent
method of characcterizing the fatigue perfformance of asphalt mixttures under a wide rangee of
loadin
ng condition
ns. Two important materrial propertiees that can bee determinedd from the S
S-
VECD
D model aree the damagee characteristic curve thaat defines hoow damage eevolves in a
speciimen and thee energy-based failure crriterion that ddefines whenn the specim
men fails. Thhese
two material
m funcctions are un
nique for a giiven mixturee regardless of temperatuure, mode off
loadin
ng, stress/strrain amplitud
de, and load
ding history. These S-VE
ECD model ffunctions cann
then be
b implemen
nted in the laayered visco
oelastic criticcal distressess (LVECD) pavement
analy
ysis program
m for the fatig
gue performaance predicttion of pavem
ments.
This sectiion presents the applicatiion of the LV
VECD progrram to the faatigue
perfo dictions of 18 pavement sections fro m different locations in the United
ormance pred
States and Canad
da. The capab
bility of the LVECD proogram to cappture crack innitiation, craack
propaagation, and damage in the
t pavemen
nt sections is investigatedd by comparring the
simullation resultss with field observations
o s. This studyy found reasoonable agreeement in trennds
betweeen the damage growth throughout
t the
t pavemennt cross-sectiions as prediicted by the
LVEC
CD program
m and the surrface crack growth
g as eviidenced by ffield observaations. Howeever,
transffer functionss would be necessary
n in order to mattch the magnnitudes of daamage obtainned
from the LVECD
D simulationss and the surrface crackinng found from the field oobservationss.

3
Thiss Chapter wass previously published
p as: Norouzi,
N A., and Y. R. Kiim (2015). Mechanistic
Evaaluation of Faatigue Crackin
ng in the Aspphalt Pavemennts. Internatioonal Journal of Pavement
Enggineering.

75
6.2 Introduction
The fatigue resistance of asphalt concrete refers to the mixture’s ability to withstand
repeated loading without fracture. In recent years, fatigue cracking has become the primary
distress in asphalt pavements in the United States. Therefore, the development of a reliable
and effective fatigue test method and performance model is needed in order to mitigate
fatigue cracking problems in the nation’s roadway infrastructure.
The simplified viscoelastic continuum damage (S-VECD) model is a viscoelastic
continuum damage mechanics-based model that has been applied effectively to predict the
fatigue performance of asphalt mixtures under different modes of loading (Daniel and Kim
2002, Kim et al. 2003, Underwood et al. 2006). This model is built on three principles: (1)
the elastic–viscoelastic correspondence principle, (2) the continuum damage mechanics-
based work potential theory, and (3) the time-temperature superposition (t-TS) principle with
growing damage to include the combined effects of time/rate and temperature. Underwood et
al. (2010, 2012) have demonstrated the ability of the S-VECD model to predict the stress-
strain behavior of asphalt concrete with growing damage and, therefore, its fatigue life.
The S-VECD model is composed of two primary material functions: the damage
characteristic curve and the energy-based failure criterion. The damage characteristic curve
defines how damage grows in a material as loading is applied to the material. The energy-
based failure criterion defines when failure occurs. The major strength of these two material
functions is that they are unique for a given mixture regardless of temperature, mode of
loading, stress/strain amplitude, and loading history (Norouzi et al. 2014, Zheng et al. 2014).
The uniqueness of these material functions allows the mixture’s behavior and fatigue life to
be determined under a wide range of testing conditions from a small set of experiments, thus
reducing the testing requirements significantly.
Reasonable and reliable stress-strain analysis is an essential element in pavement
design and performance prediction. Given the complexity of variables such as pavement
structure, traffic loading, and temperature variation, various approximate methods have been
utilized to predict pavement performance.

76
 The basic approach for stress-strain analysis is layered elastic analysis whereby the
pavement layers are considered as elastic material under stationary axisymmetric loading
(Huang 2003). Layered viscoelastic analysis is an improvement over layered elastic analysis
in that the effect of viscoelasticity is captured through Laplace transform, but the load is still
assumed to be stationary (Xu and Rahman 2008). Further, layered viscoelastic moving load
analysis is considered to be a more reliable procedure over the layered viscoelastic analysis
method due to the nature of traffic loading on a pavement system. Eslaminia et al. (2012)
developed the layered viscoelastic pavement analysis for critical distresses (LVECD)
program that can perform three-dimensional analysis under moving loads in an efficient
manner. The S-VECD model is implemented in the LVECD program, thus allowing the
fatigue performance prediction of asphalt pavements under realistic loading and
environmental conditions.
Therefore, the objective of this section is to simulate the fatigue performance of
actual pavements in service using the LVECD program and laboratory test results of the
mixtures used in the pavements and compare these simulation and laboratory test results
against the measured field fatigue cracking performance data.

6.3 Test Methods and Models

Two main tests are needed to characterize asphalt mixtures using the S-VECD model:
(1) the dynamic modulus test (AASHTO TP79 2011) to determine linear viscoelastic
characteristics using the Asphalt Mixture Performance Tester (AMPT) and (2) controlled
crosshead (CX) cyclic direct tension tests to determine the damage characteristic curve and
the energy-based failure criterion (AASHTO TP107 2014). The theories behind these
material functions are discussed in the following sections.

77
6.4 Linear Viscoelastic Model
Asphalt material is considered to be linear viscoelastic material at specific strain
levels. The dynamic modulus, |E*|, assumes a fundamental linear viscoelastic material
characteristic in the form of a mastercurve that exhibits frequency- and temperature-
dependent behavior. The dynamic modulus mastercurve can be fitted to obtain the
coefficients of a sigmoidal function, expressed in Equation (6.1), and the time-temperature
shift factor function, i.e., Equation (6.2), by employing an optimization process. To transform
a frequency-dependent property to a time-dependent property, the relaxation modulus, E(t),
of an analytical form can be used, as described by Equation (6.3). The relaxation modulus of
the Prony series form can be used as a unit response function in the convolution integral that
shows a viscoelastic material constitutive relationship between strain and stress, as expressed
in Equation (6.4).
b
log | E*| a  (6.1)
1
1
ed  g log( f R )
log(aT )  a1T 2  a2T  a3 (6.2)
m t

E (t )  E    Ei  e  i (6.3)
i 1


t
   E (t   ) d (6.4)
0


where
a, b, d, and g = optimizing constants,
fR = reduced frequency,
aT = time-temperature shift factor, and
a1, a2, and a3 = constants.

78
6.5 Simplified Viscoelastic Continuum Damage (S-VECD) Model
The S-VECD model is a rigorous mechanistic model that can describe the fatigue
behavior of asphalt concrete under a wide range of loading and environmental conditions.
This model is characterized using controlled crosshead (CX) cyclic direct tension tests that
are performed at 10 Hz at a single temperature based on the binder performance grade (PG).
All the tests are performed at three different strain amplitudes (high, medium, and low). The
strain amplitudes are selected in such a way to create a spread of numbers of cycle to failure
(Nf) in the range of 1,000 and 100,000 cycles. In addition to taking air void measurements,
fingerprint dynamic modulus tests are conducted at 10 Hz and 50 cycles before running the
CX cyclic direct tension tests to check the variability of the fatigue test specimens more
precisely. The dynamic modulus value measured from the fingerprint test is specified as
|E*|fingerprint and is used to calculate the dynamic modulus ratio (DMR) via Equation (6.5).
The designation |E*|LVE represents the dynamic modulus at the frequency and temperature
used in the fingerprint test and is measured for the same mixture from a different set of
specimens to develop a mastercurve. A DMR value in the range of 0.9 to 1.1 indicates that
the difference between the specimens used in the dynamic modulus mastercurve
development and the specimen used in the cyclic fatigue test is not significant enough to
affect the S-VECD analysis.
| E*| fingerpr int
DMR  (6.5)
| E*|LVE
For each sample tested in the CX cyclic tests in this study, failure was determined
using Reese’s approach (1997), which is based on the sudden drop in phase angle. This sharp
decrease typically occurs around the failure point, which makes the determination of the
number of cycles to failure accurate and consistent in laboratory testing. In order to minimize
the effect of viscoplasticity, Sabouri and Kim (2014) suggest using the PG of the base binder
to determine a proper testing temperature, as shown in Equation (6.6).

High temperature binder PG grade  Low temperature binder PG grade

T ( C)  -3  19 C (6.6)
2

79
 One of the two products of the S-VECD model is the damage characteristic curve,
which is the relationship between the pseudo stiffness (C) and the damage parameter (S).
This damage characteristic curve has been proven to be independent of temperature, mode-
of-loading, stress/strain amplitude, and loading history.
The other primary material function determined that can be from the S-VECD model
is the energy-based failure criterion. During cyclic fatigue testing, the maximum stored
pseudo strain energy can be determined at the stress peak points in each cycle. This amount
of energy at each cycle can be calculated using Equation (6.7).
1 1
W R
max i  ( max )i ( max
2
R
)i  ( 0,ta )i ( 0,Rta )i
2
(6.7)

where

W  = maximum stored pseudo strain energy at cycle i,

R
max i

( max )i = maximum stress at cycle i,

( max
R
)i = maximum pseudo strain at cycle i,

( 0,ta ) i = stress amplitude at cycle i, and

( 0,Rta )i = pseudo strain amplitude at cycle i.

The maximum stored pseudo strain energy at each cycle represents the material’s
ability to store energy at that particular time. The material loses its stored energy as the
damage grows for the same magnitude of applied pseudo strain due to the reduction in
pseudo stiffness. The difference between the maximum stored pseudo strain energy and the
corresponding undamaged state is referred to as the total released pseudo strain energy and
is denoted as WCR .
Therefore, a new term, GR, is defined as the rate of change of the averaged released
pseudo strain energy (per cycle) throughout the entire history of the test and can be defined
as Equation (6.8). Details regarding the GR method and its corresponding calculations can be
found in the paper by Sabouri and Kim (2014).

80
 Nf

W
R
R c
W
GR  c
 0 2 (6.8)
Nf Nf

The fact that the damage characteristic curve and the GR failure criterion are unique
for a given mixture regardless of temperature, mode of loading, stress/strain amplitudes, and
loading history makes the fatigue cracking characterization requirements much simpler than
for conventional fatigue tests. The S-VECD model protocol requires characterization tests at
only a single temperature and a single mode of loading, thereby significantly reducing the
time needed for testing. Using the proposed failure criterion and the S-VECD model
together, the fatigue life of asphalt concrete under different modes of loading and at different
temperatures and strain amplitudes can be predicted from dynamic modulus tests and CX
cyclic direct tension tests at three to four strain amplitudes. These testing requirements are
much simpler than for other fatigue tests. For example, to characterize a mixture as a
function of strain level and temperature using the bending beam fatigue test would require at
least nine test conditions (i.e., three temperatures and three strain levels). Considering that
three replicates would be needed for each condition, the total number of tests becomes 27
tests, which would take at least a month just for testing. The mechanistic nature of the S-
VECD model allows the determination of the same information within a mere two days of
testing.

6.6 Layered Viscoelastic Critical Distresses (LVECD) Pavement Analysis

Program
In this study, the LVECD program has been utilized to predict the long-term fatigue
performance of pavements under normal traffic loading. The LVECD program is able to
capture the effects of thermal stress and damage that are due to traffic loads and temperature
(Park et al. 2014). The LVECD program framework is based on the following assumptions:

 The pavement structure is an infinite layered system where the material properties
change throughout the pavement depth. This assumption is based on the fact that the

81
 pavement dimensions are very large in comparison with the tire size and pavement
thickness.
 The hourly temperature variation is usually slow, whereas traffic loading changes
within just a few seconds. Therefore, pavement performance analysis is performed at
the fixed temperature profile for each hour.
 The temperature changes throughout the pavement profile as a function of depth. This
variation is considered negligible at a specific level; therefore, temperature is
considered to be constant at all points at a given pavement depth.
 Yearly temperature variations historically have been obtained using time-consuming
calculations; however, the LVECD model allows the temperature profile to be
simulated as a cyclic function for a one-year period. Therefore, the stress and strain
calculations can be reduced to a single representative year.
 Because traffic loading mechanisms are inherently complicated, periodic loading with
a constant tire shape at a specific speed is used to model traffic in the LVECD
program.

6.7 Materials
For this study, as part of the performance-related specifications (PRS) research
project for the Federal Highway Administration (FHWA), 18 field sections with 34 different
plant-mixed loose mixtures from the United States and Canada were selected. These
pavement projects include the FHWA Accelerated Load Facility (FHWA-ALF), National
Center for Asphalt Technology (NCAT) Test Track, and Manitoba Infrastructure and
Transportation (MIT) warm mix asphalt (WMA) and reclaimed asphalt pavement (RAP)
projects.
As presented in Figure 6-1 (a), the FHWA-ALF sections, located in McLean, Virginia
in the United States, are composed of four different sets of mixtures that include three
modified mixes: crumb rubber terminal blend (CR-TB), styrene butadiene styrene (SBS), and
ethylene terpolymer (terpolymer). For each of these mixtures, the aggregate gradation,
asphalt content, and target air void level are identical.

82
 The NCAT mixtures were taken from the surface, intermediate, and base layers of six
sections, as shown in Figure 6-1 (b), at the NCAT test track in Auburn, Alabama in the
United States. The mixtures are labeled in such a way to indicate the type of modified binder
and the location within the pavement. The control (C) mixtures have virgin aggregate, binder,
and a normal production temperature. The O1 mixture is an open-graded friction course, and
the O2 and O3 mixtures consist of the same materials as the C2 and C3 mixtures from the
control section, respectively. Two WMA technologies, Double-Barrel Green foaming
technology (FW mix) and Evotherm additive technology (AW mix), were investigated. The
R mixes, which contain 50% RAP, have the same gradation as the control mixtures. The
foaming technology was applied to the R mixtures to reduce the compaction temperature;
these mixtures are labeled as RW.
The MIT sections shown in Figure 6-1 (c) are located on Provincial Highway 8 in
Manitoba, Canada. The total project length is about 17 miles, which is divided into two parts.
The first part includes eight mixtures from four sections, which were used to evaluate
the effects of three WMA technologies: Advera (MIT-W-A), Sasobit (MIT-W-S), and
Evotherm (MIT-W-E). All the mixtures used for this first part have the same gradation and
binder as the control mixture (MIT-W-C), but contain different warm mix additives. The
second part of the MIT sections, as shown in Figure 6-1 (d), were constructed in September
2009 and consist of two 2-inch layers with conventional hot mix asphalt (HMA) (MIT-R-C),
15% RAP (MIT-R-15R), 50% RAP (MIT-R-50R), and 50% RAP with a soft binder (MIT-R-
50RSB). The pavement sections were placed on top of a 4-inch HMA layer with 50% RAP
and were constructed one year earlier than the other MIT sections (i.e., in 2008) on top of a
base and subgrade.
Table 6-1 presents the mixtures’ properties, such as nominal maximum aggregate size
(NMAS), additives, binder performance grade (PG), compaction temperature, testing
temperature, etc.

83
 Control SBS-LG CR-TB Terpolymer C O FW AW R RW
0 0 NCAT-C3
NCAT-C2
NCAT-C1
NCAT-O3
5
NCAT-O2
Terpolymer NCAT-O1
5 NCAT-FW3

Depth (cm)
NCAT-FW2
Depth (cm)

CR-TB 10
NCAT-FW1
SBS-LG NCAT-AW3
NCAT-AW2
Control 15 NCAT-AW1
10 NCAT-R3
NCAT-R2
NCAT-R1
20 NCAT-RW3
NCAT-RW2
NCAT-RW1
(a) (b)
15 25
Control Sasobit Advera Evotherm Control 15% -R 50% -R 50% -RSB
0 0

MIT-W-A2
2.5 5
MIT-W-A1 MIT-R-50R

MIT-W-E2
MIT-R-50RSB

Depth (cm)
Depth (cm)

5 MIT-W-E1 10
MIT-R-15R
MIT-W-S2
MIT-R-C
MIT-W-S1
7.5 15
MIT-W-C2

MIT-W-C1
10 20

(c) (d)
12.5 25

Figure 6-1. Asphalt concrete pavement structure for (a) FHWA-ALF, (b) NCAT, (c) MIT-
WMA, and (d) MIT-RAP mixtures.

84
Table 6-1. Volumetric Properties of Different Mixtures Used in the Study.
NMAS RAP Binder Air Compaction Fatigue Testing
Project Mix Additive Location
(mm) Content (%) Grade Void (%) Temp (oC) Temp (C)

Control 12.5 - 0 PG 70-22 Surface 4.0 150

CR-TB 12.5 Crumb Rubber 0 PG 70-28 Surface 4.0 150

FHWA-ALF 19
SBS-LG 12.5 SBS 0 PG 70-28 Surface 4.0 150

Terpolymer 12.5 Terpolymer 0 PG 70-28 Surface 4.0 150

NCAT-C1 9.5 - 0 PG 76-22 Surface 4.3 154

NCAT-C2 19 - 0 PG 76-22 Intermediate 6.1 154

NCAT-C3 19 - 0 PG 67-22 Base 7.4 143

NCAT-O1 9.5 - 0 PG 76-22 Surface 18.3 152

NCAT-O2 19 - 0 PG 76-22 Intermediate 6.1 154

NCAT-O3 19 - 0 PG 67-22 Base 7.4 143

NCAT-FW1 9.5 Foam 0 PG 76-22 Surface 4.9 118

NCAT-FW2 19 Foam 0 PG 76-22 Intermediate 6.0 118

NCAT-FW3 19 Foam 0 PG 67-22 Base 7.7 118

NCAT 19
NCAT-AW1 9.5 Evotherm 50 PG 76-22 Surface 3.9 104

NCAT-AW2 19 Evotherm 50 PG 76-22 Intermediate 6.2 104

NCAT-AW3 19 Evotherm 50 PG 67-22 Base 6.1 104

NCAT-R1 9.5 - 50 PG 67-22 Surface 4.7 143

NCAT-R2 19 - 50 PG 67-22 Intermediate 6.1 143

NCAT-R3 19 - 50 PG 67-22 Base 5.0 143

NCAT-RW1 9.5 Foam 50 PG 67-22 Surface 5.0 118

NCAT-RW2 19 Foam 50 PG 67-22 Intermediate 5.8 118

NCAT-RW3 19 Foam 50 PG 67-22 Base 5.8 118

MIT-W-C1 - - PG 58-28 Surface 3.9 129

MIT-W-C2 - 35 PG 58-28 Bottom 4.8 139

MIT-W-S1 Sasobit - PG 58-28 Surface 3.2 106

MIT-W-S2 Sasobit 35 PG 58-28 Bottom 4.9 118

MIT-WMA 16 12
MIT-W-E1 Evotherm - PG 58-28 Surface 3.8 106

MIT-W-E2 Evotherm 35 PG 58-28 Bottom 5.4 117

MIT-W-A1 Advera - PG 58-28 Surface 3.0 108

MIT-W-A2 Advera 35 PG 58-28 Bottom 5.4 106

MIT-R-C - - PG 58-28 Surface 5.4 134

MIT-R-15R - 15 PG 58-28 Surface 5.2 134 12

MIT-RAP 16
MIT-R-50R - 50 PG 58-28 Surface/Bottom 5.9 134

MIT-R-50RSB - 50 PG 52-34 Surface 5.7 129 7

85
6.8 Test Results

6.8.1 Dynamic Modulus Tests

Figure 6-2 (a) presents the dynamic modulus (|E*|) mastercurves for the FHWA-ALF
mixtures. The key observation here is the significant difference between the control mixture
stiffness and that of the other modified mixtures at high reduced frequencies (cooler
temperatures and/or higher loading frequencies). However, the control mixture exhibits the
same stiffness values as the other mixtures at lower reduced frequencies (higher temperatures
and/or lower loading frequencies). All the modified mixtures show similar stiffness values
for the entire mastercurve.
The dynamic modulus test results for the NCAT test track mixtures are presented in
Figure 6-2 (b) through (d). One interesting point is the impact of the addition of 50% RAP on
the control mixture. The control mixture’s stiffness values are shown to increase significantly
at both high and low reduced frequencies, especially for the intermediate and base layers. It
is presumed that the utilization of the aged RAP increased the modulus values of the mixture.
In some instances, as shown in Figure 6-2 (c) and (d), the modulus values for the high RAP
mixtures are nearly double the values of a nearly equivalent non-RAP mixture.
The addition of the WMA additives does not seem to have changed the dynamic
modulus values of the mixtures (FW1, AW1, FW2, AW2, FW3, and AW3) greatly, whereas
the foaming technology used for the RAP mixtures (RW1, RW2, and RW3) significantly
reduced the stiffness values. In general, the intermediate and base mixtures have higher
dynamic modulus values than the surface layer mixtures. Because the surface and
intermediate layers have the same binder PG, the difference in modulus values is probably
due to their different asphalt contents (6.1% versus 4.4%) and the aggregate gradations used
in these two layers.

86
 Figure 6-2 (e) and (f) show the undamaged dynamic modulus mastercurves for all
eight mixtures at a reference temperature of 5°C for both the surface and the bottom layers.
The most noteworthy point is the increase in stiffness values for the MIT-W-S1 mix in
comparison with those of the other mixtures. In a publication by Wu and Zeng (2012), the
authors show that the addition of Sasobit to virgin binder increases the binder PG and makes
the binder stiffer. In the current study, because the aggregate gradation, binder content, and
virgin binder PG are the same for the control mixture and Sasobit WMA, it is presumed that
the addition of Sasobit is the main reason for this increase in mixture stiffness. As observed
in the NCAT mixture results, mixtures containing Advera and Evotherm exhibit very similar
stiffness values to those of the control mixture.
The comparison of the dynamic modulus mastercurves in Figure 6-2 (g) shows that
incorporating RAP into the MIT mixtures had a pronounced effect on the dynamic modulus
values. In this case, the dynamic modulus values increased significantly when 15% RAP was
added to the mixture. One possible reason for this increase could be the greater difference in
the PGs of the RAP virgin binder in the MIT mixtures. The MIT RAP binder is PG 76-10
and the virgin binder grade is PG 58-28.

87
 40000 30000
Control (a) NCAT-C1 (b)
35000
Terpolymer 25000 NCAT-O1
30000 SBS NCAT-R1
20000

|E*| (MPa)

|E*| (MPa)
25000 Crumb Rubber NCAT-RW1
20000 15000 NCAT-AW1
NCAT-FW1
15000
10000
10000
5000
5000
0 0
1E-08 1E-06 1E-04 1E-02 1E+00 1E+02 1E+04 1E-08 1E-06 1E-04 1E-02 1E+00 1E+02 1E+04

Reduced Frequency (Hz) Reduced Frequency (Hz)

30000 30000
NCAT-C2 (c) NCAT-C3 (d)
25000 NCAT-O2 25000 NCAT-O3
NCAT-R2 NCAT-R3
20000 20000

|E*| (MPa)
|E*| (MPa)

NCAT-RW2 NCAT-RW3

15000 NCAT-AW2 15000 NCAT-AW3

NCAT-FW2 NCAT-FW3
10000 10000

5000 5000

0 0
1E-08 1E-06 1E-04 1E-02 1E+00 1E+02 1E+04 1E-08 1E-06 1E-04 1E-02 1E+00 1E+02 1E+04

Reduced Frequency (Hz) Reduced Frequency (Hz)

20000 20000
MIT-W-C1 (e) MIT-W-C2 (f)
MIT-W-A1 MIT-W-A2
15000 15000
MIT-W-E1
|E*| (MPa)

|E*| (MPa)

MIT-W-E2
MIT-W-S1
10000 10000 MIT-W-S2

5000 5000

0 0
1E-08 1E-06 1E-04 1E-02 1E+00 1E+02 1E+04 1E-08 1E-06 1E-04 1E-02 1E+00 1E+02 1E+04

Reduced Frequency (Hz) Reduced Frequency (Hz)

20000
MIT-R-C (g)

16000 MIT-R-15R

MIT-R-50R
|E*| (MPa)

12000
MIT-R-50RSB

8000

4000

0
1E-08 1E-06 1E-04 1E-02 1E+00 1E+02 1E+04

Reduced Frequency (Hz)

Figure 6-2. Average dynamic modulus test results: (a) FHWA-ALF, (b) NCAT surface layer,
(c) NCAT intermediate layer, (d) NCAT base layer, (e) MIT-WMA surface layer, (f) MIT-
WMA bottom layer, and (g) MIT-RAP mixtures.

88
6.8.2 S-VECD Model Characterization Curves
Figure 6-3 shows the damage characteristic curves for the FHWA-ALF mixtures. As
shown, the characteristic curves for the SBS and terpolymer mixes collapse almost on the
same line, whereas the crumb rubber mix stays a little above them. The last point on each
curve indicates the pseudo stiffness value at failure (Cf). The results for the ALF mixtures
strongly confirm the observations by Hou et al. (2010) that suggest that the material integrity
at the failure point increases as the material becomes stiffer. The higher Cf values indicate
that stiffer mixtures can tolerate less damage before cracking. According to the dynamic
modulus test results, the control mixture has the highest stiffness values at both the low and
high reduced frequencies, so it would be expected that the control mixture curve should stay
above the curves of all the other mixtures.
The damage characteristics curves for the NCAT test track mixtures are summarized
in Figure 6-4 (a) through (c). As presented, the damage characteristic curves for the mixtures
containing a high percentage of RAP are above the other lines for the surface, intermediate,
and base layer mixtures due to higher dynamic modulus values. The difference between the
R mixtures and the control mixtures is more significant for the intermediate and base layers
than for the surface layer because these mixtures have a larger NMAS than the surface
mixture (9.5 mm vs. 19 mm) and lower asphalt contents. Therefore, it is concluded that these
mixtures are more susceptible to fatigue cracking. The lines for the AW, FW and control
mixtures collapse almost on the same curve, which confirms that WMA additives have no
major impact on fatigue behavior.
Figure 6-5 (a) and (b) demonstrate how fast the pseudo stiffness (C), which represents
the material integrity, decreases as the damage (S) increases for the MIT-WMA mixtures. All
of the characteristic lines follow a similar trend, except for that of the MIT-W-S1 mix, which
is located above the other lines. It seems that the warm mix additives have no major effect on
the fatigue resistance of the mixtures in the bottom layers.

89
 For the MIT-RAP mix curves shown in Figure 6-5 (c), the non-RAP mix and also the
mix with soft binder show a rapid decrease in material integrity with an increase in damage
compared to the mixtures with 15% and 50% RAP content. This outcome is due to the
mixture becoming more brittle once RAP is incorporated into the mix.
Generally, a comparison of only the damage characteristic curves cannot yield
reliable information about different mixtures’ fatigue resistance, because the energy that is
input by mechanical force is consumed not only in creating and propagating cracks in the
material, but also in deforming the material. Therefore, it is important to include both
stiffness and damage characteristics of a material when determining fatigue cracking
resistance.

1.0
Control
Terpolymer
0.8 SBS
Crumb Rubber
0.6
C

0.4

0.2

0.0
0.0E+00 2.5E+05 5.0E+05 7.5E+05
S

Figure 6-3. Damage characteristic curves for the FHWA mixtures.

90
 1.0
(a) NCAT-C1
NCAT-R1
0.8
NCAT-O1
NCAT-FW1
0.6 NCAT-AW1

C
NCAT-RW1
0.4

0.2

0.0
0.0E+00 2.5E+05 5.0E+05 7.5E+05 1.0E+06
S

1.0
(b) NCAT-C2
NCAT-R2
0.8
NCAT-O2
NCAT-FW2
0.6 NCAT-AW2
C

NCAT-RW2
0.4

0.2

0.0
0.0E+00 2.5E+05 5.0E+05 7.5E+05 1.0E+06
S

1.0
(c) NCAT-C3
NCAT-R3
0.8 NCAT-O3
NCAT-FW3
0.6 NCAT-AW3
C

NCAT-RW3
0.4

0.2

0.0
0.0E+00 2.5E+05 5.0E+05 7.5E+05 1.0E+06
S

Figure 6-4. Damage characteristic curves for the FHWA mixtures: (a) surface layer, (b)
intermediate layer, and (c) bottom layer.

91
 1.0
(a) MIT-W-C1
MIT-W-A1
0.8
MIT-W-E1
MIT-W-S1
0.6

C
0.4

0.2

0.0
0.0E+00 2.5E+05 5.0E+05 7.5E+05
S

1.0
(b) MIT-W-C2
MIT-W-A2
0.8
MIT-W-E2
MIT-W-S2
0.6
C

0.4

0.2

0.0
0.0E+00 2.5E+05 5.0E+05 7.5E+05
S

1.0
(c) MIT-R-C
MIT-R-15R
0.8
MIT-R-50R
MIT-R-50RSB
0.6
C

0.4

0.2

0.0
0.0E+00 2.5E+05 5.0E+05 7.5E+05
S

Figure 6-5. Damage characteristic curves for the MIT mixtures: (a) WMA surface layer, (b)
WMA bottom layer, and (c) MIT-RAP.

92
6.8.3 Failure Criteria
An important result that can be obtained from fatigue testing is the characterization of
the failure criterion for the different study mixtures. Each single data point in Figure 6-6
represents the result of characterizing GR versus Nf for each replicate for a specific mixture.
One of the benefits of this failure criterion is its ability to predict failure at different
temperatures and strain levels without the need to test at multiple temperatures. As presented
in Figure 6-6 (a) for the modified binders, the SBS mixture exhibits the longest fatigue life
for the same GR value in comparison with the others. The crumb rubber and terpolymer
mixtures show similar fatigue resistance, and the control mixture seems to be the least
fatigue-resistant mixture among the FHWA-ALF mixtures.
One interesting point that can be seen in Figure 6-6 (b) through (d) is the effect of
RAP on the fatigue resistance of the mixtures. In general, the inclusion of RAP leads to the
deterioration of fatigue life, but the trend observed from the NCAT mix failure criterion lines
indicate that, except for the surface layer, the addition of RAP improved the mixture’s fatigue
resistance. Although the R2 mix failure criterion line stays below that of the control mixture
in Figure 6-6 (c), the R2 mix slope is less steep than the line for the control mixture, which
means that for longer cycles the fatigue resistance of the R2 mix is better than that of the
control mixture. This improvement in the fatigue failure criterion line is probably due to the
use of the lower binder grade for all of the RAP mixtures.
However, Figure 6-6 (g) demonstrates that incorporating a high percentage of RAP
into the mix deteriorates the mixture’s fatigue resistance, indicated by the failure criterion
line for the 15% RAP mix being a little below the virgin mix line. However, the line for the
50% RAP mix deviates from those two considerably, suggesting that at the same GR value,
the mix with 50% RAP results in a smaller number of cycles to failure. This outcome is
probably due to the increase in the virgin binder stiffness as 50% aged binder is added to the
mix, which makes the pavement system experience lower strain levels and consequently
lower GR values under the same loading history.

93
 The effects of warm mix additives are presented in Figure 6-6 (b) through (f) for the
NCAT and MIT-WMA mixtures. The utilization of WMA deteriorated the pavement
performance, as indicated by the fact that all of the GR vs Nf lines are located below those of
the control and intermediate mixtures, except for the bottom layer NCAT mixtures.
Similar to the case for the C versus S curves, a comparison of only the failure
criterion lines is not altogether fair, because in some instances, such as the R section in the
NCAT sections, a stiffer mixture will have a shorter fatigue life at a given strain level but
will have smaller strain levels in the same structure due to the greater stiffness of the mixture.
So, in order to obtain accurate performance predictions, the whole pavement structure should
be taken into consideration.

94
 10000 10000
(a) (b)

1000 1000

GR
NCAT-C1

GR
100 100
NCAT-O1
Control NCAT-R1
10 Terpolymer
10 NCAT-RW1
SBS
NCAT-AW1
Crumb Rubber
1 NCAT-FW1
1
1.E+03 1.E+04 1.E+05 1.E+06
1.E+03 1.E+04 1.E+05 1.E+06
Nf (Cycle) Nf (Cycle)

10000 10000
(c) (d)

1000 1000

GR
GR

100 NCAT-C2
100 NCAT-C3

NCAT-R2 NCAT-R3

NCAT-RW2 10 NCAT-RW3
10
NCAT-AW2 NCAT-AW3
NCAT-FW2 NCAT-FW3
1 1
1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+03 1.E+04 1.E+05 1.E+06
Nf (Cycle) Nf (Cycle)

10000 10000
(e) (f)

1000 1000
GR

GR

100 100
MIT-W-C1 MIT-W-C2
MIT-W-A1 MIT-W-A2
10 10
MIT-W-E1 MIT-W-E2
MIT-W-S1 MIT-W-S2
1 1
1.E+03 1.E+04 1.E+05 1.E+06 1.E+03 1.E+04 1.E+05 1.E+06
Nf (Cycle) Nf (Cycle)

10000
(g)

1000
GR

100
MIT-R-C
MIT-R-15R
10
MIT-R-50R
MIT-R-50RSB
1
1.E+03 1.E+04 1.E+05 1.E+06
Nf (Cycle)

Figure 6-6. Failure criterion envelopes (GR vs. Nf) for: (a) FHWA-ALF, (b) NCAT surface
layer, (c) NCAT intermediate layer, (d) NCAT base layer, (e) MIT-WMA surface layer, (f)
MIT-WMA bottom layer, and (g) MIT-RAP mixes.

95
6.9 LVECD Model Simulation Results
Typical results of 20-year LVECD simulations for the 18 test sections are presented
in this section. The damage contours shown in Figure 6-7 through Figure 6-10 are the
contours for fatigue damage based upon Miner’s law, as expressed by Equation (6.9).
T
Ni
D
i 1
i 
N fi
(6.9)

where
D = damage,
T = total number of periods,
Ni = traffic for period i, and
Nfi = allowable failure repetitions under the conditions that prevail in period i.

The N/Nf values start from 0.0 in an intact condition and increase as the level of
damage increases. The value of N/Nf = 1 generally corresponds to the point at which the
strains localize and the material fails. The same color scale (between 0 (blue) and 1 (red)) is
used to present all the damage contours shown in Figure 6-7 through Figure 6-10. A high
damage ratio, which is represented by dark red in the figures, corresponds to the areas with
high levels of damage.
In order to quantify the LVECD simulation results, an index value, i.e., the damage
area (%), has been defined. This index is calculated by determining the percentage of the
damaged points across a pavement cross-section that have a damage ratio equal to 1 (N/Nf =
1) over the total nodes in the pavement structure.
It is important to note that real traffic data expressed in terms of equivalent single-
axle loads (ESALs) were used for all the pavement section simulations. In addition, the
environmental temperatures for the different points across the pavement section were
obtained from the Enhanced Integrated Climate Model (EICM) used in the Mechanistic
Empirical Pavement Design Guide (MEPDG).

96
 Field cracking measurements were taken based on the Long-Term Pavement
Performance (LTPP) criteria and specified into three different categories: low, medium, and
high intensities. To convert the field measurements into a single index, Equation (6.10) was
adopted from a FHWA report (FHWA 2003) and used to define a representative index for
fatigue performance. Because of the different amounts of cracking and multiple test sections
with different lengths, the parameters in Equation (6.10) can be estimated using Equation
(6.11). The results for the damage area (%) obtained from the LVECD simulations and the
crack area (%) obtained from the field measurements for the different projects are presented
in Figure 6-11.
 % Low   % Med   % High  
Crack area (%)  40      (6.10)
 350   200   75  
where
% Low = percentage of interval length, low severity
% Med = percentage of interval length, medium severity
% High = percentage of interval length, high severity
length of respective longitudinal cracking
percent int erval(%)  (6.11)
section length

97
 Damage Factor (N/Nf ) Distribution - @ January 1, 2034 Damage Factor (N/Nf ) Distribution - @ January 1, 2034
0 1 0 1
(a) (b)
1 0.9 1 0.9

2 0.8 2 0.8

3 0.7 3 0.7

4 0.6 4 0.6
Z (cm)

Z (cm)
5 0.5 5 0.5

6 0.4 6 0.4

7 0.3 7 0.3

8 0.2 8 0.2

9 0.1 9 0.1

10 0 10 0
-1.5 -1 -0.5 0 0.5 1 1.5 -1.5 -1 -0.5 0 0.5 1 1.5
X (m) X (m)

Damage Factor (N/Nf ) Distribution - @ September 1, 2033 Damage Factor (N/Nf ) Distribution - @ January 1, 2034
0 1 0 1
(c) (d)
1 0.9 1 0.9

2 0.8 2 0.8

3 0.7 3 0.7

4 0.6 4 0.6
Z (cm)

5 0.5 Z (cm) 5 0.5

6 0.4 6 0.4

7 0.3 7 0.3

8 0.2 8 0.2

9 0.1 9 0.1

10 0 10 0
-1.5 -1 -0.5 0 0.5 1 1.5 -1.5 -1 -0.5 0 0.5 1 1.5
X (m) X (m)

Figure 6-7. Five-year simulation results obtained from the LVECD program for the FHWA
mixtures: (a) control, (b) terpolymer, (c) SBS, and (d) crumb rubber.

98
 Damage Factor (N/Nf ) Distribution - @ July 1, 2029 Damage Factor (N/Nf ) Distribution - @ January 1, 2029
0 1 0 1

2 (a) 0.9 2
(b) 0.9

0.8 0.8
4 4
0.7 0.7
6 6
0.6 0.6
8 8

Z (cm)
Z (cm)

0.5 0.5
10 10
0.4 0.4
12 12
0.3 0.3

14 14
0.2 0.2

16 0.1 0.1
16

18 0 0
-1.5 -1 -0.5 0 0.5 1 1.5 -1.5 -1 -0.5 0 0.5 1 1.5
X (m) X (m)

Damage Factor (N/Nf ) Distribution - @ July 1, 2029 Damage Factor (N/Nf ) Distribution - @ July 1, 2029
0 1 0 1

2
(c) 0.9 2
(d) 0.9

0.8 0.8
4 4
0.7 0.7
6 6
0.6 0.6
8 8
Z (cm)

0.5 Z (cm) 0.5

10 10
0.4 0.4

12 0.3 12 0.3

14 0.2 14 0.2

16 0.1 16 0.1

0 0
-1.5 -1 -0.5 0 0.5 1 1.5 -1.5 -1 -0.5 0 0.5 1 1.5
X (m) X (m)

Damage Factor (N/Nf ) Distribution - @ July 1, 2029 Damage Factor (N/Nf ) Distribution - @ July 1, 2029
0 1 0 1

2
(e) 0.9 2
(f) 0.9

0.8 0.8
4 4
0.7 0.7
6 6
0.6 0.6
8 8
Z (cm)

Z (cm)

0.5 0.5
10 10
0.4 0.4

12 0.3 12 0.3

14 0.2 14 0.2

16 0.1 16 0.1

0 0
-1.5 -1 -0.5 0 0.5 1 1.5 -1.5 -1 -0.5 0 0.5 1 1.5
X (m) X (m)

Figure 6-8. Five-year simulation results obtained from the LVECD for the NCAT mixtures:
(a) NCAT-Control, b) NCAT-AW, (c) NCAT-FW, (d) NCAT-RW, (e) NCAT-R, and (f)
NCAT-O.

99
 Damage Factor (N/Nf ) Distribution - @ September 1, 2021 Damage Factor (N/Nf ) Distribution - @ September 1, 2021
0 1 0 1

1 (a) 0.9 1 (b) 0.9

2 0.8 2 0.8

3 0.7 3 0.7

4 0.6 4 0.6

Z (cm)

Z (cm)
5 0.5 5 0.5

6 0.4 6 0.4

7 0.3 7 0.3

8 0.2 8 0.2

9 0.1 9 0.1

10 0 10 0
-1.5 -1 -0.5 0 0.5 1 1.5 -1.5 -1 -0.5 0 0.5 1 1.5
X (m) X (m)

Damage Factor (N/Nf ) Distribution - @ September 1, 2021 Damage Factor (N/Nf ) Distribution - @ September 1, 2021
0 1 0 1

1 (c) 0.9 1
(d) 0.9

2 0.8 2 0.8

3 0.7 3 0.7

4 0.6 4 0.6
Z (cm)

Z (cm)
5 0.5 5 0.5

6 0.4 6 0.4

7 0.3 7 0.3

8 0.2 8 0.2

9 0.1 9 0.1

10 0 10 0
-1.5 -1 -0.5 0 0.5 1 1.5 -1.5 -1 -0.5 0 0.5 1 1.5
X (m) X (m)

Figure 6-9. Five-year simulation results obtained from the LVECD program for the MIT-
WMA mixtures: (a) control, (b) Advera, (c) Evotherm, and (d) Sasobit.
Damage Factor (N/N ) Distribution - @ August 1, 2029 Damage Factor (N/N ) Distribution - @ August 1, 2029
f f
0 1 0 1

2 (a) 0.9 2 (b) 0.9

4 0.8 4 0.8

6 0.7 6 0.7

8 0.6 8 0.6
Z (cm)
Z (cm)

10 0.5 10 0.5

12 0.4 12 0.4

14 0.3 14 0.3

16 0.2 16 0.2

18 0.1 18 0.1

20 0 20 0
-1.5 -1 -0.5 0 0.5 1 1.5 -1.5 -1 -0.5 0 0.5 1 1.5
X (m) X (m)

Damage Factor (N/N ) Distribution - @ August 1, 2029 Damage Factor (N/N ) Distribution - @ August 1, 2029
f f
0 1 0 1

2 (c) 0.9 2
(d) 0.9

4 0.8 4 0.8

6 0.7 6 0.7

8 0.6 8 0.6
Z (cm)

Z (cm)

10 0.5 10 0.5

12 0.4 12 0.4

14 0.3 14 0.3

16 0.2 16 0.2

18 0.1 18 0.1

20 0 20 0
-1.5 -1 -0.5 0 0.5 1 1.5 -1.5 -1 -0.5 0 0.5 1 1.5
X (m) X (m)

Figure 6-10. Five-year simulation results obtained from the LVECD program for the MIT-
RAP mixtures: (a) control, (b) 15R, (c) 50R, and (d) 50RSB.

100
 12 100
(a) (b)
10
80

Cracked Area (%)

Damaged Area (%)
8
60
6
40
Control
4 Control
CR-TB
CR-TB
20
2 SBS-LG
SBS-LG

Terpolymer Terpolymer
0 0
1.E+04 1.E+05 1.E+06 1.E+07 1.E+03 1.E+04 1.E+05 1.E+06
No. of Cycles No. of Cycles

10 60
(c) (d)
50
8
Damaged Area (%)

Cracked Area (%)

40
6
30
C C
4 O O
20
R R
RW RW
2
FW
10 FW
AW AW
0 0
1.0E+05 1.0E+06 1.0E+07 1.0E+08 1.E+07 1.E+08
No. fo Cycles No. of Cycles

10 12
(e) (f)
10
8
Damaged Area (%)

Cracked Area (%)

8
6
6
4
4
Control Control
2 Advera Advera
Evotherm 2
Evotherm
Sasobit Sasobit
0 0
1.0E+05 1.0E+06 1.0E+07 1.0E+05 1.0E+06
No. of Cycles No. of Cycles
5 14
(g) (h)
12
4
Damaged Area (%)

Cracked Area (%)

10
3
8

2 6
C
C 4
15R
1 15R
50R 2 50R
50RSB 50RSB
0 0
1.E+05 1.E+06 1.E+07 1.E+04 1.E+05 1.E+06
No. of Cycles No. of Cycles

Figure 6-11. (a) LVECD analysis results of ALF pavements, (b) field measurements from
ALF pavements, (c) LVECD analysis results of NCAT pavements, (d) field measurements
from NCAT pavements, (e) LVECD analysis results of MIT-WMA pavements, (f) field
measurements from MIT-WMA pavements, (g) LVECD analysis results of MIT-RAP
pavements, and (h) field measurements from MIT-RAP pavements.

101
 Several important observations can be made from the damage contours and field
measurements presented in Figure 6-7 through Figure 6-11.
 In general, most of the predicted simulation results indicate fairly good agreement
with the field performance observations.
 According to the results obtained from the dynamic modulus characteristic curves and
the failure criterion lines for the FHWA-ALF mixtures, it was expected that the SBS
mixture would perform the best and the control mixture would perform the worst out
of all the mixtures. The trends observed in the damage contours and field
measurements clearly show this prediction in Figure 6-11 (a) and (b). It also can be
observed for this pavement that the damage was concentrated in the bottom of the
asphalt concrete layer. This observation is in line with the generally accepted failure
mechanism for thin asphalt pavements, i.e., bottom-up fatigue cracking.
 Warm mix additives seem to deteriorate the mixtures’ fatigue resistance, although
they decrease the stiffness and aging during compaction. Figure 6-8 (b) and (c) and
Figure 6-9 summarize the damage contours for the Sasobit, Evotherm, Advera
mixtures and the foamed mixture. The field data support the simulation results
obtained from the LVECD program, as seen in Figure 6-11 (c) through (f).
 RAP is considered to be a material that decreases the fatigue resistance of a mixture
due to the increase in material stiffness. However, the simulation results suggest
otherwise and show that the inclusion of RAP improves the mixture’s resistance to
fatigue. Because the virgin binder grade used in this study is PG 76-22, which is
fairly high, it is possible that the utilization of RAP had no significant effect on
resistance to fatigue. Although the LVECD model rankings for the R and RW
sections do not match the field measurements, both the LVECD model results and the
field observations show a small amount of damage in Figure 6-11 (c) and (d).

102
 It is accepted that the addition of RAP up to 20% does not have any major effect on
virgin mixture performance, according to many asphalt agencies. The LVECD
simulation results for the MIT-RAP project presented Figure 6-10 (a) and (c) confirm
this finding, because no major differences in fatigue cracking can be observed for
both the LVECD model and field cracking results, whereas the utilization of RAP up
to 50% led to pavement performance deterioration, as seen in Figure 6-10 (c) and (d).
The use of soft binder led to some improvement in performance, as indicated in the
GR line for the MIT-R-50RSB mix compared with that of the MIT-R-50R mix;
however, no major change in performance was observed.
 The key point that emerges from the comparisons of the field measurements versus
the simulation predictions is that the magnitudes of the damage area (%) index are
quite different from the observed cracking areas in the field, although rankings among
the different sections match well. It is noted that the damage area index is calculated
from the amount of damage in the cross-section of a pavement, whereas the field
cracking areas are determined from the cracks on the pavement surface. This
difference emphasizes the significance of transfer functions. The need for such a
function is not unexpected for a simulation program such as the LVECD model. The
concept of transfer functions will be better defined as other complex phenomena,
such as asphalt material aging and healing, which are not included in the version of
the LVECD program used in this study, become incorporated for future releases of
the LVECD program.

103
6.10 Summary and Conclusions
This section presents the predictions of the fatigue cracking performance for 18
pavement sections composed of 34 different asphalt mixtures using experimental, analytical,
and computational methodologies. The 34 asphalt mixtures include WMA mixtures that
incorporate several different WMA technologies and RAP mixtures with different RAP
contents and virgin binders. The fatigue characterization at the material level was undertaken
using cyclic direct tension tests in accordance with the AASHTO TP 107 protocol and the S-
VECD model. The fatigue simulations at the pavement level were performed using the
LVECD program developed at NC State University. A comparison between the measured
and predicted performance results indicates that the growth trend of the ‘damage area’ in the
cross-section of a pavement agrees well with the cracking area measured from the pavement
surface. Although overall trends agree between the simulations and the field observations, it
is evident that transfer functions are necessary in order to match the magnitude of the damage
area and time scale in the LVECD simulations with the magnitude of the field surface
cracking area and time scale. The S-VECD characterization and the LVECD simulations of
the various modified mixtures, WMA mixtures, and RAP mixtures reveal the effects of
different material factors (e.g., different modifications, WMA technologies, RAP contents,
virgin binder grades) on the fatigue performance of asphalt mixtures as well as asphalt
pavements.

104
CHA
APTER 7 NUM
MERICAL
L EVALU
UATION
N OF PAV
VEMEN
NT
DES
SIGN PA
ARAMET
TERS FO
OR THE FATIGU
UE CRA
ACKING
AND
D RUTT
TING PER
RFORM
MANCE O
OF ASPH
HALT
PAV T4
VEMENT

7.1 Abstract
Over receent years, sig
gnificant reseearch has beeen conducteed to investiggate ways to
prediict fatigue crracking and permanent
p deformation
d ((rutting), whhich are two common
distreesses found in
i asphalt paavements. Th
hese distressses are affectted by materrial propertiees,
envirronmental co
onditions, an
nd the pavem
ment’s structuure. This secction investiggates commoon
pavem
ment design parameters,, including su
urface type, base layer thhickness, baase layer type,
sub-b
base layer thickness, and
d an anti-frosst layer, withh regard to thhe asphalt paavement
perfo
ormance of th
he Korea Ex
xpressway Co
orporation (K
KEC) test rooad. Test roaads are oftenn
regarrded as the most
m realisticc tools for ev
valuating thee effects of vvarious param
meters becauuse
they are
a subjected
d to real trafffic and environmental fa
factors. The K
KEC test roaad is 7.7 km long
and was
w construccted with thee aim of deveeloping a Koorean mechaanistic empirrical pavemeent
desig
gn guide usin
ng the newly
y developed Layered
L VisscoElastic paavement anaalysis for Criitical
Distresses (LVEC
CD) program
m.
According
g to the find
dings, the basse layer thickkness and baase layer maaterial were ffound
to afffect the fatig
gue cracking and rutting performancee, whereas thhe sub-base thickness annd
anti-ffrost layer were
w found no
ot to affect th
he amount oof distress siggnificantly. The LVECD
D
progrram was ablee to capture the effects of
o the changees in the aforrementionedd parameterss on
the am
mount of craacking and ru
ut depths. Reasonable aggreement waas found betw
ween the

4
This chapter
c was prreviously published as: Norou
uzi, A., D. Kim
m, and Y. R. K
Kim (2014). Nuumerical Evaluaation
ue Cracking annd Rutting Perfformance of Assphalt Pavements.
of Pavement Dessign Parameters for the Fatigu
Jou
urnal of Mateerials and Stru
uctures.

105
program simulations and the field data. However, it remains necessary to develop a
simulation of field transfer functions in order to obtain true field performance predictions.

7.2 Introduction
For decades, the pavement industry has been endeavoring to improve pavement
construction practices, extend the life of new pavements, and minimize the need for
pavement rehabilitation efforts. It is now well accepted that fatigue cracking and rutting are
two major distresses that occur in hot mix asphalt pavements. These phenomena have been
reported in many parts of the United States as well as Europe and other countries.
In pavement analysis, two strains are considered to be critical for design purposes: the
horizontal strain (t) at the bottom of the asphalt layer, which is instrumental in initiating
fatigue cracking, and the vertical compressive strain (v) at the top of the subgrade layer,
which causes permanent deformation in the asphalt pavement. A study by Behiry (2012)
showed that the base layer type, base layer thickness, and the subgrade resilient modulus are
the key elements that control the strain levels at the critical locations in the pavement
structure. Therefore, the key step for performance predictions and design purposes is to
utilize accurate methods that consider pavement life, traffic loading, and temperature
variations to obtain reasonable stress-strain analysis results for both the vertical and
horizontal directions throughout the depth of the pavement.
The basic approach for stress-strain analysis is layered elastic analysis, where the
pavement layers are considered to be the elastic material under stationary axisymmetric
loading (Huang 2003). However, layered elastic analysis is not an accurate tool for asphalt
pavement because asphalt concrete exhibits viscoelastic behavior, especially under traffic
loading. Layered viscoelastic moving load analysis, which is an improvement over layered
elastic analysis, is considered also to be more reliable than linear viscoelastic analysis in the
viscoelastic domain due to its efficient functions that account for the effects of moving loads
and viscoelasticity (Eslaminia and Guddati 2010).

106
 Generally, in order to predict asphalt performance, effective models that can reliably
represent fatigue damage growth and permanent deformation are still in demand in addition
to efficient tools for strain and stress calculations. The Simplified ViscoElastic Continuum
Damage (S-VECD) model is a mechanistic approach that has been applied effectively to
predict the performance of asphalt concrete mixtures during pre-localization stages under
different modes of loading (Hou et al. 2010, Chehab et al. 2003, Underwood et al. 2010,
Underwood et al. 2012). Zhang et al. (2013) developed an energy measure that represents the
rate of damage growth using the S-VECD model and can predict fatigue failure. However,
Zhang’s approach is based on the controlled crosshead (CX) mode of loading only. Sabouri
and Kim (2014) proposed a new energy-based failure criterion called the GR method, which
is independent of mode of loading, test temperature, and strain amplitude. This new criterion
has been verified successfully for multiple mixtures that contain reclaimed asphalt pavement,
warm mix additives, and modified binders (Norouzi et al. 2014, Sabouri et al. 2014, Tafreshi
and Norouzi 2012).
In this study, the rutting performance of the study mixtures was evaluated using a
permanent deformation model developed at North Carolina State University by Choi et al.
(2012). This so-called shift model is capable of expressing the permanent strain growth of
asphalt concrete in both the primary and secondary regions as a function of deviatoric stress,
load time, and temperature, based on the time-temperature and time-stress superposition
principles (Kim and Guddati 2011). These three factors (deviatoric stress, load time, and
temperature) are important in predicting rut depths in asphalt pavements because they vary
throughout the depth of the asphalt layer. The triaxial stress sweep (TSS) test used in this
study was developed to predict the rutting performance of the study mixtures by calibrating
the shift model.
However, performance predictions that are based solely on material testing cannot
provide a comprehensive picture of the pavement’s behavior. Because pavement is a layered
structure, the stress and strain distributions vary from point to point, which results in
complicated shear and flow zones (Gibson et al. 2009). The newly released LVECD program

107
is able to calculate linear viscoelastic pavement responses that can be used for both the shift
model and the S-VECD model to predict rutting and fatigue cracking, respectively. Norouzi
and Kim (2014) verified the LVECD program for multiple pavement sections in the United
States and found strong agreement between the simulation results and the field observations.
Given the above considerations, the purpose of this study is to gain a better
understanding of the effects of the design parameters, i.e., surface layer type, base layer
material properties, base layer thickness, sub-base thickness, and subgrade properties, on the
fatigue and rutting resistance of asphalt pavements using experimental material
characteristics and simulation software. This study used the following test protocols to find
the required parameters for the performance predictions: dynamic modulus tests (AASHTO
TP 79), uniaxial fatigue tests using the S-VECD model (AASHTO TP 107), and TSS tests
for permanent deformation. To verify the LVECD analysis results, field data were collected
from the tested sections and compared with the program simulations.

7.3 Test Protocols

7.3.1 Dynamic Modulus Testing

Asphalt material is considered to be linear viscoelastic material at specific strain
levels; therefore, viscoelastic materials have both viscous and elastic components. The
dynamic modulus, |E*|, can be expressed in the form of a mastercurve that exhibits
frequency- and temperature-dependent behavior. In this study, dynamic modulus testing was
performed in load-controlled mode in axial compression following AASHTO TP 79. The
tests were carried out for all the study mixtures at 4°C, 20°C, 40°C, and 54°C and at
frequencies of 25, 10, 5, 1, 0.5, and 0.1 Hz. The load levels were specified by trial and error
so that the strain amplitudes were between 50 and 75 microstrain to prevent damage to the
specimens. The dynamic modulus values were fitted for the coefficients of the sigmoidal
function and time-temperature shift factors by optimizing the dynamic modulus
mastercurves.

108
 After determining the shift factors, the dynamic modulus was converted to the
relaxation modulus. Finally, a power term, alpha (  ), used in viscoelastic continuum damage
(VECD) theory, was calculated from the maximum log-log slope, m, of the relaxation
1
modulus and time using the relationship   1  .
m

7.3.2 Cyclic Testing Using the S-VECD Model

The S-VECD fatigue performance model – which is the simplified form of the more
rigorous VECD model and can be used to characterize the fatigue behavior of asphalt
concrete using the elastic-viscoelastic correspondence principle, continuum damage
mechanics, and time-temperature superposition principle – has been proven to be
independent of loading. Controlled crosshead (CX) cyclic direct tension tests were performed
at 10 Hz at different temperatures based on the binder performance grade (PG) following
AASHTO TP 107. All the tests were performed at three different strain amplitudes (high,
medium, and low). The strain amplitudes were selected in such a way to create a spread of
numbers of cycles to failure (Nf) in the range of 1,000 and 100,000 cycles. In addition to
taking air void measurements, in order to check the variability of the fatigue test specimens
more precisely, fingerprint dynamic modulus tests were conducted at 10 Hz and 50 cycles
before running the CX cyclic direct tension tests. The dynamic modulus value measured from
this test is specified as |E*|fingerprint and is used to calculate the dynamic modulus ratio (DMR)
via Equation (7.1). A DMR value in the range of 0.9 to 1.1 guarantees that the linear
viscoelastic properties obtained from the dynamic modulus tests can be used effectively in S-
VECD analysis.

| E*| fingerpr int

DMR  (7.1)
| E*|LVE

109
 To determine failure for each sample, the corresponding cycle that is related to the
sudden drop in the phase angle, which typically happens around the failure point, has been
specified based on the Reese approach (1997). In order to minimize the effects of
viscoplasticity, Sabouri and Kim (2014) suggested using the PG of the base binder to
determine the proper testing temperature, as shown in Equation (7.2).

High temperature binder PG grade  Low temperature binder PG grade

T ( C)  -3  19 C (7.2)
2

7.3.3 Fatigue Failure Criterion (GR Method)

According to the S-VECD failure criterion (the GR method), the maximum stored
pseudo strain energy at each cycle represents the material’s ability to store energy at that
particular time. The material loses its stored energy as the damage grows for the same
magnitude of applied pseudo strain due to the reduction in pseudo stiffness. The difference
between the maximum stored pseudo strain energy and the corresponding undamaged state is
referred to as the total released pseudo strain energy and is denoted as WCR . According to
this criterion, a characteristic relationship exists between the rate of change of the averaged
released pseudo strain energy (GR) during fatigue testing and the final fatigue life (Nf). The
GR can be calculated using Equation (7.3). This failure criterion combines the advantages of
the S-VECD model and this characteristic relationship, which both originate from
fundamental mixture properties. Details regarding the GR method and its corresponding
calculations can be found in the paper by Sabouri and Kim (2014).

Nf
1
   1  F 
R 2
0,ta i
2
G  R 0
(7.3)
N 2f

110
 where
  R
0,ta i
= pseudo strain amplitude at cycle i, and
Fi
= pseudo stiffness at cycle i.

7.3.4 Permanent Deformation (TSS testing)

The TSS test is composed of two type of tests: a reference test at the high temperature
(TH) followed by three multiple stress sweep (MSS) tests at three different temperatures of
low, intermediate, and high (TL, TI, and TH), respectively. The reference test in this study
utilizes a 0.4-second pulse with a 10-second rest period. This reference test provides
permanent strain mastercurves by fitting the incremental model, which is expressed as
Equation (7.4).

 0  Nred
 vp  , (7.4)
 N I  Nred 


a p  p1 log( p )  p2 ,
(7.5)
a v  d1 ( v / Pa )d2  d3 .

where
Nred = reduced number of cycles at reference loading conditions,
p1, p2 = coefficients of reduced load time shift factor,
d1, d2, d3 = coefficients of deviatoric stress shift factor, and
Pa = atmospheric pressure to normalize stress.
The MSS tests consist of three loading blocks with increases in deviatoric stress level
(70, 100, and 130 psi) while the other loading conditions are kept constant. In this study, the
shift factors were obtained by shifting the permanent strain of an individual loading block
toward the permanent strain mastercurve, which was obtained from the reference test. The
reduced load time shift factor and deviatoric stress shift factor are shown in Equation (7.5).

111
 The physical number of cycles at a given condition was converted into a reduced
number of cycles using the total shift factor, which is the sum of the deviatoric stress shift
factor and the reduced load time shift factor. These two shift functions utilize temperature,
load time, and vertical stress to calculate the shift factors. Details regarding the TSS test
method and shift model can be found elsewhere (Choi et al. 2012).

7.4 Materials and Test Methods

For this study, experiments were performed using five laboratory-produced mixtures.
Of these mixtures, two types of asphalt mixtures were used at the surface to compare the
rutting and crack propagation; these mixtures were an ASTM mix and a 19-mm nominal
maximum aggregate size (NMAS) polymer-modified styrene butadiene styrene (SBS) mix,
which is designated as PMA throughout the study. The intermediate layer consisted of a 25-
mm NMAS BB5 mixture with 70 mm thickness. Mixes designated as BB1 (25 mm NMAS)
and BB3 (40 mm NMAS), which are frequently used in South Korea, were used for the base
layers.
Table 7-1 summarizes the volumetric properties of the mixtures used in the KEC test
road sections. The sublayers below the base layer are composed mostly of sub-base and anti-
frost layers that are placed on top of the subgrade. An anti-frost layer often is used to
compensate for the level difference due to the base; however, the anti-frost layer was omitted
from some sections for comparative purposes to evaluate the effectiveness of anti-frost layers
on pavement performance. Figure 7-1 schematically presents the KEC pavement structure.

112
 Table 7-1. Volumetric Properties of the KEC Mixtures
Type Surface Base Intermediate
Mixture ASTM PMA BB1 BB3 BB5
Styrene
Binder Type Unmodified Butadiene Unmodified
Styrene
Binder Grade PG 64-22 PG 76-22 PG 64-22
NMAS* (mm) 19 25 40 25
% Air Void
5.9 5.7 7.6 7.5
(S-VECD)
% Air Void
5.9 6.0 8.0 9.9
(Rutting)
Sieve Size Gradation, % Passing
37.5 mm 100.0 100.0 100.0 100.0
25.0 mm 100.0 100.0 88.6 100.0
19.0 mm 99.6 92.5 71.0 91.0
12.5 mm 84.9 72.9 51.1 67.5
9.5 mm 71.1 63.9 44.1 55.1
4.75 mm 49.3 48.5 38.1 31.2
2.36 mm 36.2 36.1 29.1 23.0
0.60 mm 18.1 18.0 15.1 12.8
0.30 mm 11.6 11.6 10.1 9.2
0.15 mm 7.4 7.3 6.8 6.7
0.075 mm 4.4 4.2 4.4 4.6
* Nominal maximum aggregate size.

7.4.1 Specimen Fabrication

To prepare the specimens, aggregate stockpiles were dried and sieved for batching.
The aggregate particles were then heated to the mixing temperature for more than six hours
before mixing. The asphalt binder and aggregates were mixed together using a bucket mixer.
Then, the mixtures were short-term oven-aged for four hours at the compaction temperature.

113
 All the test specimens were compacted to a height of 178 mm and a diameter of 150
mm using the Superpave Gyratory Compactor (SGC). To obtain specimens of uniform air
void distribution, these samples were cored to a diameter of 100 mm and cut to a height of
130 mm (Lee et al. 2014) for fatigue cyclic testing, and to a diameter of 100 mm with the
height of 150 mm for the dynamic modulus and the permanent deformation (TSS) testing.
The air void contents were measured using the CoreLok method for each specimen
prior to testing. All the test samples met the target air void content by  0.5 percent. The
direct tension test specimens were glued to metal plates using Devcon steel putty. Vertical
deformations were measured using four linear variable differential transducers (LVDTs) with
the gauge length of 70 mm at intervals of 90 degrees for both the dynamic modulus and CX
cyclic tension tests.

114
 A1 A2 A2-2 A3 A4 A5 A5-2 A6 A7 A8 A8-2 A9 A10 A10-2 A11 A11-2 A12 A12-2 A13 A13-2 A14 A14-2 A15 A15-2

5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5

7 PMA

8 8 8 ASTM

18 18 18

28 28 28 BB5

30 BB3
30 30

40
30 BB1
30 30

30 30 30 Agg

40

Sub-Base

40 40 40 40

30 30 30 Anti-Frost

20 20 20 20

10 Subgrade

* The numbers in the figure are all in centimeters (cm).

Figure 7-1. Asphalt pavement sections at the KEC test road.

115
7.5 Test Results and Discussion

7.5.1 Dynamic modulus Values

Figure 7-2 presents the averaged dynamic modulus values of the replicates for the
KEC mixtures, respectively, on the same graph. The comparison between the PMA and
ASTM mixes clearly shows that the PMA mix exhibits low stiffness values at high reduced
frequencies (low temperatures) and high stiffness values at low reduced frequencies (high
temperatures). In other words, it seems that the PMA mix presents favorable characteristics
for fatigue resistance at low temperatures and for rutting resistance at high temperatures. In
general, the other unmodified mixtures have similar stiffness values, given the sample-to-
sample variability.
25000
(a)

20000
|E*| (MPa)

15000

10000 ASTM
PMA
BB1
5000
BB3
BB5
0
1E-08 1E-06 1E-04 1E-02 1E+00 1E+02
Reduced Frequency (Hz)

100000
(b)

10000
|E*| (MPa)

1000

100
1E-08 1E-06 1E-04 1E-02 1E+00 1E+02
Reduced Frequency (Hz)

Figure 7-2. Dynamic modulus mastercurves for KEC test road mixtures: (a) in semi-log
space and (b) in log-log space.

116
7.5.2 S-VECD Characterization Curves
Figure 7-3 presents the averaged damage characteristic curves for the KEC test
mixtures. As shown, the corresponding curve for the ASTM mixture is slightly above that of
the PMA mixture. It can be concluded that both the PMA and ASTM mixes follow the same
behavior even though their dynamic modulus values are considerably different. This outcome
could be related to the higher asphalt content in the ASTM mix compared to the PMA mix.
Another reasonable explanation for this observation may be related to the physical nature of
the base binder and SBS modifier used in the PMA mix. The other important point is that the
curve for the BB1 mix is below the other curves, which could be due to its low asphalt
content and large aggregate size.
Generally, a comparison of the damage curves cannot yield reliable information about
different mixtures’ fatigue behavior, because the energy that is input by mechanical force is
consumed not only in creating and propagating the cracks, but also in deforming the material.
Therefore, it is important to include both stiffness and damage characteristics of the material
when determining a mixture’s fatigue cracking resistance.

1.0
ASTM
PMA
0.8 BB1
BB3
0.6 BB5
C

0.4

0.2

0.0
0.0E+00 2.5E+05 5.0E+05
S

Figure 7-3. Averaged damage characteristic curves for KEC test road mixtures.

117
7.5.3 Fatigue Failure Criterion Lines
The S-VECD failure criterion was applied to all of the mixtures in this study, and the
results are presented in Figure 7-4. The position of the failure criterion lines can be used to
make a relative comparison of a mixture’s expected fatigue resistance. Also, not only the
position but also the slope of the failure criterion line plays an important role in the pavement
system’s fatigue behavior. As observed in Figure 7-4, the line for the PMA mix is parallel to
that of the ASTM mix and is slightly above that of the control mixture. Because the two test
mixtures follow the same gradation, the difference in the failure criterion lines, which was
observed before in the characteristic curves, may be due to the modified binder that is used in
the PMA mix. The BB5 mix curve is located over the BB3 mix curve, which indicates that
the mixtures with smaller aggregate sizes perform better than the ones with larger aggregates.

10000
ASTM
PMA
1000 BB1
BB3
BB5
GR

100

10

1
1.E+02 1.E+03 1.E+04 1.E+05 1.E+06
Nf (Cycle)

Figure 7-4. Fatigue failure criterion lines for the KEC mixture.

118
7.5.4 Triaxial Stress Sweep Tests

The TSS test method requires four tests with two replicates for each test to calibrate
the shift model. The TSS test temperatures for the KEC mixtures are specified as follows:
22°C as the low temperature (TL), 36°C as the intermediate temperature (TI), and 46°C as the
high temperature (TH). Details regarding the temperature selection method can be found
elsewhere (Choi et al. 2012). Figure 7-5 presents the results of the TSS tests. The dotted lines
show the reference curves and the solid lines correspond to the averaged permanent strains of
the MSS tests at each temperature. The first important observation regarding the surface
mixtures is the difference in permanent strain levels between the ASTM and PMA mixtures.
The corresponding curves for the ASTM mix indicate higher permanent deformation levels
than the PMA mix because the PMA mix contains SBS-modified asphalt binder (PG 76-22),
which provides strong evidence for the benefits of polymer modification for rutting
resistance. For the base course (BB1 mix and BB3 mix) comparison, the BB1 mix exhibits
lower permanent deformation levels than the BB3 mix due to the smaller aggregate particles
(25 mm) and lower target air void content than the BB3 mix.
The averaged permanent strain levels presented in Figure 7-5 were used to
characterize the shift model. The model coefficients were then applied to the LVECD
program to evaluate rutting performance within a pavement structure.

119
 2% 2%
Permanent Strain (a) KEC-ASTM (b) KEC-PMA

Permanent Strain
Reference TH
Reference
1% 1% TH
TI
TI

TL TL

0% 0%
0 100 200 300 400 500 600 0 100 200 300 400 500 600
Number of Cycles Number of Cycles

2% 2%
(c) KEC-BB1 (d) KEC-BB3
TH TH
Permanent Strain

Permanent Strain
Reference

1% 1% TI
Reference TI

TL TL

0% 0%
0 100 200 300 400 500 600 0 100 200 300 400 500 600
Number of Cycles Number of Cycles

2%
(e) KEC-BB5
Permanent Strain

TH

1% Reference
TI

TL

0%
0 100 200 300 400 500 600
Number of Cycles

Figure 7-5. TSS test results of KEC mixtures: (a) ASTM, (b) PMA, (c) BB1, (d) BB3, and (e)
BB5.

120
7.6 Pavement Analysis
In this study, LVECD program simulations were used to evaluate the effects of the
pavement design parameters, i.e., the surface layer type, base layer thickness, base layer
material, sub-base thickness, and the impact of an anti-frost layer on pavement performance.
Eslaminia et al. (2012) developed the LVECD program to calculate the stresses and strains
throughout the pavement depth. The LVECD program, by combining time-scale separation
and layered viscoelastic analysis, uses fast-Fourier transforms to perform three-dimensional
viscoelastic calculations under moving loads in a rapid manner. The assumption behind using
the transforms is that loading occurs in the same repeated manner each time, so the behavior
is cyclic and has a steady-state response.
The asphalt layer is modeled as viscoelastic material with damage. Therefore, the
asphalt layer is represented by the Prony series of the dynamic modulus values, time-
temperature shift factors, S-VECD model coefficients, and TSS model parameters. The
aggregate base, anti-frost layer, and subgrade were modeled using linear elastic properties
with modulus values of 350 MPa, 88 MPa, and 75 MPa, respectively. The other inputs
required for the LVECD simulations are design time, pavement structure, traffic, and climate.
The design time for this study was assumed to be twenty years. A single tire with standard
loading of 80 kN at the center of the pavement was assumed. The average annual daily truck
traffic (AADTT) was 935 based on the traffic data obtained from the site. Pavement
temperatures were obtained from the Enhanced Integrated Climate Model (EICM) software.
The EICM program provides hourly temperatures of asphalt pavements in terms of pavement
depth.

7.6.1 Fatigue Performance

In order to evaluate fatigue resistance, the LVECD program calculates the damage
growth and the damage factor using Equation (7.6) based on Miner’s law. If the damage
factor is equal to zero, the element does not experience any damage. A damage factor of one
indicates failure of the element.

121
 T
Ni
Damage Factor  N
i 1 fi
(7.6)

where
T = total number of periods,
Ni = traffic for period i, and
Nfi = allowable failure repetitions under the conditions that prevailed in period i.

The fatigue cracking simulation results for 24 pavement sections are presented in
Figure 7-6. It should be noted that the fatigue performance predicted from the LVECD
program has not yet been fully calibrated with the field performance data, and the transfer
functions that convert the damage predicted from the LVECD software to the percentage of
cracking areas are still in process. To compare the fatigue life of the mixtures, the numbers of
failure points (elements with the damage factor of ‘1’) were counted and divided by the total
number of points throughout the section, and the resultant value was specified as an index for
the amount of damage (in percent, %) in the pavement section. To verify the cracking data
obtained from the LVECD program, field distress data from Seo (2010) were used.
Figure 7-6 (a) shows the amounts of damage for the different base layers (BB1, BB3,
and aggregate base mixtures). A comparison of the results clearly suggests better
performance for the pavements that have an asphalt material base layer than the ones that
have an aggregate base layer in terms of the different base layer thicknesses, which was
expected. Between the pavements that have an asphalt layer base, it seems that the BB1 mix,
which shows better fatigue resistance based on the failure criterion lines, exhibits less
damage than the BB3 mix. This outcome is in strong agreement with the field cracking data,
as presented in Figure 7-7 (a). Also, the difference in performance between the BB1 and BB3
mixes may be due to the lower air void content and smaller aggregate size of the BB1 mix,
which led to better aggregate interlocking in the mixture.

122
 50 50
(a) Base Layer Type (b) Base Layer Thickness
45 45
BB1 8 cm
40 BB3 40 18 cm
Damaged Area (%)

Damaged Area (%)

35 Agg Base 35 28 cm

30 30

25 25

20 20

15 15

10 10

5 5

0 0
8 cm 18 cm 28 cm BB1 BB3 Aggregate
Base Layer Thickness Base Layer Type
50 50
(c) Subgrade Type (d) Subbase Layer Thickness
45 45

40 Subgrade 40 30 cm
Damaged Area (%)

Damaged Area (%)

35 Anti-Frost 35 40 cm

30 30

25 25

20 20

15 15

10 10

5 5

0 0
8 cm 18 cm 28 cm 8 cm 18 cm 28 cm
Base Layer Thickness Base Layer Thickness

Figure 7-6. Effects of different parameters on fatigue cracking performance as obtained from
LVECD simulations: (a) base layer material, (b) base layer thickness, (c) subgrade material,
and (d) sub-base thickness.

Another interesting observation that can be made from Figure 7-6 (b) is the
interaction between two factors: the base layer material type and the base layer thickness. For
example, in the case of the aggregate base pavements, with an increase in the base layer
thickness from 8 cm to 28 cm, the damage decreased from 36% to 30%, whereas the same
increase in the BB1 mix layer led to a corresponding decrease in damage from 20 % down to
less than 5 percent. As expected, the BB3 mix was not as effective in resisting fatigue
cracking as the BB1 mix because the BB3 mix performed the worst among all of the
mixtures used in the KEC test road. However, a greater reduction in fatigue cracking was
observed with an increase in the base layer thickness, especially for the aggregate base

123
structure in the field, which was not captured effectively by the LVECD program, as shown
in Figure 7-7 (b).
The analysis results presented in Figure 7-6 (c) briefly indicate that the subgrade type
(anti-frost or subgrade) does not appear to play an important role in the fatigue behavior of
the entire pavement. In other words, similar pavement performance can be expected for
pavements that have either anti-frost or subgrade material. The field observations presented
in Figure 7-7 (c) also confirm this finding.
The effects of the sub-base layer thickness are demonstrated in Figure 7-6 (d). In
general, the pavements with a 30-cm sub-base layer behave similarly to the ones with a 40-
cm sub-base layer. It seems, unlike the base layer thickness that significantly alters fatigue
resistance, that the sub-base layer has no major impact on the amount of cracking. Although
this observed trend was not unexpected, the quantification of the changes in fatigue cracking
can be useful for an optimal pavement design.

124
 30 30
(a) Base Layer Type (b) Base Layer Thickness
25 BB1 25 8 cm
BB3 18 cm
Cracked Area (%)

Cracked Area (%)

Agg Base 28 cm
20 20

15 15

10 10

5 5

0 0
8 cm 18 cm 28 cm BB1 BB3 Aggregate
Base Layer Thickness Base Layer Type
30 30
(c) Subgrade Type (d) Subbase Layer Thickness
25 Subgrade 25 30 cm
Cracked Area (%)

40 cm

Cracked Area (%)

Anti-Frost
20 20

15 15

10 10

5 5

0 0
8 cm 18 cm 28 cm 8 cm 18 cm 28 cm
Base Layer Thickness Base Layer Thickness

Figure7-7. Effects of different parameters on pavement fatigue cracking in the field: (a) base
layer type, (b) base layer thickness, (c) subgrade material, and (d) sub-base thickness.

7.6.2 Rutting Performance

In order to help determine rutting resistance, the deviatoric stress and load time
underneath the pavement can be calculated using the LVECD program. The deviatoric stress
changes throughout the depth of the pavement depending on the load applied. The permanent
deformation is calculated for each layer based on the deviatoric stress.

125
 The permanent deformation value that is the sum of the permanent deformation of all
the layers is the total rut depth; therefore, the ultimate deformation in a given layer can be
obtained by multiplying the permanent strain by the total thickness.
As for the fatigue performance analyses, the LVECD rutting simulation results also
were compared to the measured rut depths in the field, as presented in Figure 7-8. Both the
measured and predicted rut depths obviously show better rutting resistance of the polymer-
modified mixture than the ASTM mixture, as shown in Figure 7-8 (a). Figure 7-8 (b) presents
the comparison of the rut depths of the different base layer types (BB1, BB3, and aggregate
base). As shown, the 8-cm aggregate base layer exhibits extreme rutting deformation,
whereas an increase in the thickness up to 18 cm corresponds to a decrease in the rut depth.
This outcome leads to the conclusion that a relatively thick aggregate base layer provides
reasonable rutting resistance.
However, increasing the thickness up to 28 cm did not decrease (lessen) the rut depth
compared to the 18-cm layer. The LVECD program was not able to simulate the poor
prediction of the thinner aggregate base layer because the rutting coefficient inputs in the
LVECD program were not accurate enough to capture the true performance of the aggregate
base. The utilization of the asphalt base mixtures (BB1 and BB3) provided better rutting
resistance than the aggregate base layers. Overall, the BB1 mix shows better rutting
resistance than the BB3 section, as demonstrated in Figure 7-8 (b).
The subgrade and anti-frost layers constituted the sublayers of the asphalt sections.
Three sections (1, 3, and 6) were selected among six sections that have no anti-frost layer (2,
4, and 8), and their rut depths were compared against those sections with an anti-frost layer.
Figure 7-8 (c) illustrates the effects (or lack thereof) of an anti-frost layer on rutting
performance. In general, no correlation between the anti-frost layers and rutting was found,
even though slightly greater rut depths were apparent at sections that had an anti-frost layer.
The LVECD prediction results show very close agreement with and without anti-frost layers.

126
 Figure 7-8 (d) presents the effect of the sub-base layer thickness on the rutting
performance of sections without an anti-frost layer. Although the rut depths in the 30-cm
base layer are slightly greater than those in the 40-cm base layer, there is no significant
relationship in terms of the sub-base layer thicknesses. The LVECD prediction results show
that the rut depths of the 30-cm and 40-cm sub-base layers are exactly the same. The rut
depths increase with an increase in base layer thickness because the deformation calculated
in a given layer is the permanent strain multiplied by the thickness.

10 20
(a) Surface Layer Type (b) Base Layer Type
ASTM BB1

Total Rut Depths (cm)

Total Rut Depths (cm)

8
PMA 15 BB3
Aggregate
6
Measured
10
4 Predicted

5
2

0 0
8 cm 18 cm 28 cm 8 cm 18 cm 28 cm
Base Layer Thickness Base Layer Thickness
10 10
(c) Subgrade Layer Thickness (d) Base Layer Thickness
20 cm 8 cm
Total Rut Depths (cm)

Total Rut Depths (cm)

8 30 cm 8 18 cm
40 cm 28 cm

6 6

4 4

2 2

0 0
Subgrade Anti-Frost 30 cm 40 cm
Subgrade Type Subbase Layer Thickness

Figure 7-8. Effects of different parameters on rutting performance: (a) surface layer type, (b)
base layer type, (c) subgrade type, and (d) sub-base thickness.

127
7.7 Conclusions
In this study, the effects of common pavement design parameters on asphalt
pavement performance have been investigated through laboratory-produced asphalt mixture
testing, numerical simulations, and field verification. The fatigue cracking and rutting
performance of 24 asphalt pavement sections with different structures were evaluated using
validated models. The S-VECD model was used to evaluate the fatigue properties of the
mixtures, and TSS tests were performed to assess the rutting behavior of the mixtures. The
results were then input in the LVECD pavement analysis program to predict the long-term
performance of the test pavement sections. A summary of the findings and the conclusions
that can be drawn from this study are as follows:
 As expected, the utilization of the mixtures that contained modified binders improved
the pavement performance. As seen in Figure 7-8 (a), the sections with the PMA mix
exhibited less permanent deformation than the ASTM sections due to the higher
stiffness values of the PMA mix, especially at higher temperatures.
 One interesting observation is that the base layer material played the most important
role in affecting both rutting and fatigue cracking in the KEC test road sections. The
aggregate base layer exhibited greater rut depths and more cracking than the asphalt
base layers, whereas the sections with an asphalt base layer (either the BB1 mix or
BB3 mix) showed less distress than the aggregate base sections.
 The data obtained from the pavement analyses of the different base layer thicknesses
briefly indicate that an increase in the base layer thickness reduces the pavement
distresses, but the amount of improvement is dependent on the material type.
According to the observations made from Figure 7-7 (b) and Figure 7-8 (b),
increasing the aggregate base layer thickness from 8 cm to 28 cm can provide up to
70% reduction in pavement distress.

128
 The pavement simulations indicate no significant change in the fatigue resistance or
permanent deformation when using an anti-frost layer or increasing the sub-base layer
thickness. The field observations are in strong agreement with these findings, as
shown in Figures 7 (c) and (d).
 The key point that has emerged from the field measurements versus the simulation
predictions is that the magnitude of the damage area (%) index and the total rut
depths do not directly correspond to the observed field data. This finding suggests the
need for developing a simulation-to-field transfer function. The need for such a
function is not unexpected for a simulation program such as the LVECD software.

129
CHA
APTER 8 APPL
LICATIO
ON OF V
VISCOEL
LASTIC
CONTINUU
UM DAM
MAGE AP
PPROAC
CH TO P
PREDICT
T THE
TIGUE CRACKI
FAT C ING PER
RFORMA
ANCE O
OF BINZH
HOU
PER
RPETUA
AL PAVE S5
EMENTS

8.1 Abstract
Fatigue cracking
c in th
he asphalt paavements haas been know
wn as one off the complexx
distreesses that is dependent
d of asphalt mix
xture properrties, environnmental connditions, and
pavem
ment structu
ure. This pheenomenon is still difficullt to predict nnot only duee to models,
becau
use it has not been well recognized.
r One way to improve thee fatigue life is to increasse
the paavement thicckness to decrease the teensile strain at the bottom
m of the asphhalt layer. To
h this goal, perpetual pav
reach vements can be considereed as a reasoonable alternnative. In thhis
sectio
on, the Binzh
hou perpetuaal pavement test sectionss constructedd in Shandonng Province,
Chinaa, have been
n simulated for
f the long term
t fatigue performancce using the ffinite elemennt
packaage, LVECD
D (Layered Viscoelastic
V pavement annalysis for C
Critical Distrresses). In thhis
frameework, asphaalt concrete is modeled in
i the contexxt of linear vviscoelastic ccontinuum
damaage theory. A recently deeveloped uniified fatigue failure criteerion is also incorporatedd to
defin
ne the boundaaries of the applicable
a reegion of the theory. All m
material moddels are
charaacterized by dynamic mo
odulus tests and
a direct teension cyclicc fatigue testts with
cylindrical specim
mens in the laboratory.
l The
T 15 year fatigue perfoormance sim
mulations
recom
mmend the use
u of Sectio
on 1 and 3 ass the perpetuual pavementt design.

5
This chapter
c was prreviously published as: Cao, Wei.,
W A. Norou
uzi, and Y. R. Kim (2015). A
Application of
Viscoelastic Con
ntinuum Damag
ge Approach to
o Predict the Faatigue Crackinng Performancee of Binzhou
Perpetual Pavements. Journal of
o Traffic and Transportation
T Engineering, iin press.

130
8.2 Introduction
For the past decades, the most widely used pavement type in China has been the
semi-rigid asphalt pavement, which is characterized by semi-rigid base layers, such as
cement-stabilized aggregate and lime-fly ash treated soil, and overlays of asphalt concrete.
This type of pavement structure is cost effective and environmentally friendly as it is able to
consume a considerable amount of industry waste. On the other hand, the semi-rigid base
also provides a stable support necessary for the high traffic volumes (and even overloading)
typically seen in China. However, the semi-rigid materials are prone to fatigue cracking
under repeated traffic loading. Once crack initiates in the base layer, it propagates upward
into the flexible asphalt layers thereby leading to the reflective cracking and consequently the
structural failure of the pavement. Hence, the design life of such pavements is usually limited
to 10-15 years.
Inserting a stress-absorbing interlayer between the asphalt concrete and base materials
has been one of the major remedies for reflective cracking. Nevertheless, it can at most help
to delay the crack propagation when a high-quality material design and construction are
guaranteed for the interlayer; the base may crack as well. At the end, the rehabilitation or
even reconstruction is necessitated for the whole pavement structure involving all the layers.
Considering these unavoidable defects with semi-rigid materials, another alternative, albeit
expensive in the short term, would be the perpetual pavement. Perpetual pavement is
commonly designed to last for 40-50 years without structural failure; only the surface asphalt
layer requires rehabilitation or replacement when necessary. Therefore, in the long run
perpetual pavement could be a choice of high cost effectiveness and sustainability (Amini et
al 2011).
Traditionally, perpetual pavement is designed by invoking the concept of limiting
critical pavement responses. It is generally believed that if the imposed traffic loads produce
responses below certain threshold values then the structural damage does not accumulate.
The critical pavement responses of interest are the vertical compressive strain at the top of
subgrade and the horizontal tensile strain at the bottom of asphalt layers for structural rutting
and bottom-up fatigue cracking, respectively (Behiry 2012).

131
 In order to limit the rutting to upper few inches of the pavement, increase in the
structure thickness or stiffness of materials is required so that the vertical load can be
distributed to a greater extent before reaching the subgrade. Experimental evaluation tools for
the material rutting resistance include Asphalt Pavement Analyzer, Hamburg Wheel-
Tracking Device, and the Heavy Vehicle Simulator, to list a few. To simulate the material
macroscopic behavior and understand the underlying deformation mechanisms, readers are
encouraged to follow the work by Choi (2013), Cao (2015), and Darabi et al. (2012), for
example.
On the fatigue side, one way to decrease the probability of bottom-up cracking is to
increase the structure thickness as well, thereby reducing the maximum strain at the bottom
of asphalt pavement. The bending beam test has been conventionally used in the laboratory
to evaluate the fatigue resistance of asphalt mixtures and investigate the concept of fatigue
endurance limit which is the aforementioned threshold level of tensile strain. More advanced
explorations regarding the endurance limit are performed by Underwood and Kim (2009) and
Bhattacharjee et al. (2009) within the context of viscoelastic continuum damage theory.
Despite the wide acceptance, the above described design methods for perpetual
pavements are subject to questions. During the service life pavement experiences complex
traffic and environmental conditions, and thus the consideration of effects of temperature,
aging, and healing, for example, should be included in both material and structural design as
they all lead to variations in material properties.
Obviously, it is less persuasive and lacks rigor to approve or disapprove a mix design
by testing the material only at certain pre-specified conditions.
This chapter is focused on the material characterization and pavement performance
evaluation on the aspect of bottom-up fatigue cracking. A comprehensive analysis system is
presented and its applicability and versatility is demonstrated via simulations on the Binzhou
perpetual pavement test sections in Shandong Province, China. In this framework the effects
of temperature have been naturally incorporated through rigorous mechanistic modeling of
asphalt materials. As for the other factors such as aging and healing, continuous efforts are
being made in the on-going research.

132
8.3 Overview of Binzhou Test Sections
Six different asphalt mixes were designed for the Binzhou project. The material
designations and descriptions are listed in Table 8-1. Five different sections were
implemented, and the structure layout has been illustrated in Figure 8-1. As can be observed
from Figure 8-1, Section 1 through Section 3 are categorized as the full-depth asphalt
pavement which are the typical designs for perpetual pavement in Europe and North
America. In the project Section 1 was designed by using 70  as the critical value for the
tensile strain at the bottom of pavement, while Section 2 and 3 used 125 . Section 5 is the
semi-rigid asphalt pavement extensively adopted in China as mentioned previously. Section 4
used a similar design to Section 5 whereas the thickness of asphalt layers is considerably
increased by inserting a layer of large stone porous asphalt mixture (LSPM) right above the
semi-rigid base. Note that for all the sections the top 30 cm of subgrade was treated with lime
in field construction in order to increase the soil modulus.
As a relatively new type of asphalt mixture, LSPM is usually designed with a uniform
gradation and large particle sizes with the nominal maximum aggregate size (NMAS)
typically greater than 25 mm, which is intended for an interlocking aggregate structure. In the
meanwhile, a small amount of fine aggregate is used to fill the aggregate void to increase the
material’s stiffness and durability.
The air void content is usually targeted at 13-18% which gives the mixture excellent
draining capability (Zhao and Huang 2010). In addition, owing to the large particle size and
high air void content, LSPM is also able to function as a stress-absorbing interlayer and thus
possesses the potential to reduce the reflective cracking according to numerical simulations
and field observations in China (Luo and Xu 2007, Yufeng et al. 2008, Yin 2011).

133
 Table 8-1. Asphalt concrete materials used in the Binzhou test sections.
Designation Description
SMA Stone Matrix Asphalt (PG 76-22, MACa modified)
Superpave-19 19-mm NMAS Superpave (PG 76-22, MAC modified)
Superpave-25 25-mm NMAS Superpave (PG 64-22)
LSPM 25-mm Large Stone Porous Mixture (PG 70-22, MAC modified)
F-1 12.5-mm NMAS Superpave Fatigue Layer (PG 64-22)
F-2 12.5-mm NMAS Superpave Fatigue Layer (PG 76-22, SBSb modified)
Note: aMultigrade Asphalt Cement, bStyrene-Butadiene-Styrene.

Figure 8-1. Structure layout of test sections in Binzhou perpetual pavement project.

134
8.4 Material Modeling
Strictly speaking, asphalt concrete is a viscoelastic-viscoplastic material; which
mechanism dominates the material responses is dependent on the material intrinsic
properties, temperature, and loading rate. Since fatigue cracking usually occurs at lower
temperatures and the structure layer of interest is the bottom asphalt layer a few inches below
the surface, it is reasonable to confine the study within the context of viscoelasticity. In the
following the theory of linear viscoelasticity and continuum damage mechanics are briefly
reviewed. In the next section, the coupled viscoelastic continuum damage (VECD) theory is
then applied to laboratory testing to characterize the material stiffness and fatigue resistance.

8.4.1 Linear Viscoelastic Continuum Damage Theory

As just mentioned in the above, an intact linear viscoelastic material would
demonstrate a linear relationship between stress and pseudo strain in a monotonic loading.
When the microcrack damage initiates, this relation will evolve into nonlinear regions
resulting in a reduction in the instantaneous secant modulus, which is named as the secant
pseudo stiffness (C). As analogous to elastic cases, this pseudo stiffness is used to
characterize the material deterioration due to damage. Material is treated as a continuum with
microcracks being assumed as uniformly distributed within the body without coalescence. By
ignoring the effects of anisotropy, nonlinear viscoelasticity, and plasticity, if any, the VECD
framework hypothesizes that any reduction in the pseudo stiffness is caused by the material’s
internal damage (S), exclusively:

C (S )  (8.1)
R
where C is the secant pseudo stiffness, and S is the internal variable quantifying the
material damage status. It is assumed that between C and S there exists a single-valued
monotonically decreasing function, referred to as the damage characteristic relationship.
The pseudo stiffness can be directly calculated using Equation (8.1). Note that for the
intact state free of damage, the C-value is 1, equal to ER, as obtained by substituting Equation
(8.4) into Equation (8.1). Yet, the damage state variable can usually be approached only

135
through experimental investigations and theoretical hypotheses. In this case, Schapery’s work
potential theory, which is derived based upon thermodynamic principles, is employed to
characterize the damage evolution process:

dS  W R 
   (8.2)
dt  S 

where  is a material constant denoting the damage evolution rate, and WR is the
pseudo strain energy density function, which is given by analogy with linear elastic cases as
1
W R  C (S )  R 
2
(8.3)
2
Initially developed for material constitutive modeling under monotonic loading
conditions (Kim and Little 1990, Park et al. 1996), the VECD theory and its algorithms were
later on extended and improved to accommodate the cyclic loading scenarios (Daniel and
Kim 2002, Underwood et al. 2010, 2012), and it is then that the theory begins to see its
prosperity in modeling the fatigue characteristics in asphalt materials. In the laboratory, the
C-S curve is usually determined by running cyclic direct tension fatigue test with the
displacement controlled mode in a closed-loop servo-hydraulic test system. Afterwards, an
exponential function is usually used to fit the experimental curves in order for further
manipulations:
C  exp  aS b  (8.4)

where a and b are the fitting coefficients.

It is worth mentioning that the resulting damage characteristic relationship is a
material intrinsic property that is independent of temperature, loading type (monotonic or
cyclic), mode (stress, strain, or displacement controlled), and other conditions such as
loading amplitude and rate, as long as viscoelasticity is maintained as the dominating
mechanism in the material. Also note that the time-temperature superposition principle for
the intact state is still applicable to the material despite of a significant presence of damage
(Chehab et al. 2002, Underwood et al. 2006), and therefore the same time/frequency-
temperature equivalence relation can be used in the VECD framework.

136
8.4.2 Failure Criterion
By now, the damage characteristic relationship has been obtained. It prescribes the
path that the damage evolution of each material point in asphalt layers should follow.
However, the VECD theory does not provide a definition of the applicable region for the
damage curves as the fundamental assumption of the continuum would be violated as soon as
microcracks start to coalesce causing damage localization. Hence, a separate failure criterion
is called for to complete the VECD system.
Early explorations (Hou et al. 2010, Underwood et al. 2012) are focused on the
pseudo stiffness (C) by seeking a relationship to express the critical C-value at failure as a
function of mixture type (i.e., whether contains reclaimed asphalt pavement or not), NMAS,
and test conditions (e.g., temperature and load frequency). Such criteria are proved to be
unreliable because of high variability and dependence on the loading mode. The energy
based failure criterion developed by Sabouri and Kim (2014) presents a consistent and
unified approach to failure characterization that has been verified successfully for non-RAP
and RAP materials (Norouzi et al.,2014) . In the following the model is reviewed in a slightly
different way than the original presentation.
As shown in Figure 8-2, attention is placed on the space of stress and pseudo strain.
To construct the damage characteristic curves, the values of C and S are computed in a cyclic
manner in the VECD theory. Starting immediately after the stress peak in the first cycle, C is
redefined as the peak-to-peak pseudo stiffness, designated as F. Since the pseudo strain
energy dissipates slowly in the cyclic test, which is evidenced by the small area enclosed by
the hysteresis loops before failure, F is deemed as a good indicator of material damage status
or degree of integrity.
Material deterioration is exhibited via a progressive decline in the F-value, which is
accompanied with the loss in material’s ability to store the pseudo strain energy. For the
specific stress and (pseudo) strain history in a cycle i, the amount of the maximum pseudo
strain energy that could be stored if the material is intact is indicated by the triangular area
ADC, as shown in Figure 8-2, where the AC line has a slope of one, Rs denotes the
permanent pseudo strain, and R0,ta denotes the tensile amplitude of the pseudo strain in this

137
cyclee. Meanwhilee, the actual amount of th
he maximum
m pseudo strrain energy tthat can be sttored
in thee material fo
or this particu
ular cycle is indicated byy the area AD
DB.
C and denoteed as WCR , iss the
The differrence betweeen the two energy, indiccated by ABC
pseud
do strain eneergy that wou
uld have to dissipate
d forr the material to degrade from the inttact
state to the speciffic damage state
s in that cycle.
c There fore, for a giiven fatigue test, WCR caan be

consiidered as ano
other damage state variaable, and is nnamed as thee total releassed pseudo strain
energ
gy in the orig
ginal presenttation. Howeever, it shoulld be noted tthat this energy is not
physiically releaseed or dissipaated in the faatigue test, annd thereforee in this chappter it is referred
to as the characteeristic dissip
pated pseudo
o strain enerrgy (CDPSE)). Nevertheless, it providdes
an eff
ffective and convenient
c approach
a to tracking
t matterial status bby comparinng the damagged
to thee virgin statee in a cyclic manner. Lasstly, it is alsoo worth notinng that in thhe above
discu
ussion the rev
verse loading
g and the po
otential healiing effects arre neglectedd.

Fig
gure 8-2. Sch
hematic repreesentation of the charactteristic dissippated pseudoo strain enerrgy.

Acco
ording to Fig
gure 8-2, the CDPSE for cycle i is givven by
1 R 2
W  R
C i 
2
 0,ta i 1  Fi  (8.5)

138
 When plotted with cycle number until the cycle at failure Nf, the CDPSE exhibits
different profiles in terms of shape and magnitude depending on the test temperature and
loading mode. However, it is discovered that the following quantity GR, as defined in (8.6), is
consistently related to the fatigue life Nf :
Nf

W
R
R C ( N ) dN
W
G 
R
 C 0
(8.6)
Nf Nf2

where WCR is the averaged characteristic dissipated pseudo strain energy (ACDPSE)
in a fatigue test, and GR is named as the “rate of change of the average released pseudo strain
energy (per cycle)” throughout the entire history of the test (Sabouri and Kim, 2014). The
relationship found between GR and Nf from experimental data is well expressed via a power
function, or presented as a straight line in the logarithmic scales:
GR   N f (8.7)

where  and  are fitting parameters.

Some inappropriateness, however, is sensed regarding the physical significance
assigned to GR in the original model presentation as described above. Basically, according to
(8.6) GR is simply equal to the accumulated CDPSE that is averaged twice over the entire

fatigue life. It is difficult for the authors to relate the ratio of WCR to Nf with the rate of
change of ACDPSE. Instead, ACDPSE itself has a well-defined physical meaning and it may
be reasonably related to the total amount of pseudo strain energy that should be dissipated in
an intact material before reaching failure. Substituting (8.6) into (8.7) yields
WCR    N f  with     1 (8.8)

In this study, the unified failure criterion expressed in (8.8) is employed.

139
8.5 Experimental Characterization

8.5.1 Dynamic Modulus Test

To identify the relaxation modulus function, in the laboratory the dynamic modulus
test is often chosen preferably over the relaxation test given the concerns related to machine
compliance, capacity, and the accuracy in observing instantaneous responses. In this study,
for the dynamic modulus test the plant-mixed loose mixtures acquired from the project site
were compacted using a Servopac Superpave gyratory compactor and then cored and cut to
obtain cylindrical specimens for each mix with 100 mm in diameter and 150 mm in height.
All tests were conducted in accordance with AASHTO TP 62-10 testing protocol (AASHTO,
2010) using the Asphalt Mixture Performance Tester (AMPT). By virtue of the time-
temperature superposition principle, the resulting modulus data points at different
temperatures were then horizontally shifted in the space of modulus versus frequency to
construct a continuous and smooth master curve at a pre-specified reference temperature. The
amount of shift for each temperature is termed as the shift factor. The dynamic modulus
master curve is commonly expressed as a sigmoidal function:

log E *    (8.9)
1  exp     log  f r  

f r  f / aT (8.10)

log aT  1T 2   2T   3 (8.11)

where |E*| is the dynamic modulus, fr is the reduced frequency, f is the physical
loading frequency, aT is the shift factor at temperature T, and , , , , 1, 2, 3 are
parameters that can be identified in an optimization process in practice.

140
 In the laboratory test, a minimum of three replicates were employed and the averaged
dynamic modulus master curves are now presented in Figure 8-3, together with the phase
angle master curves and the shift factor functions. A few observations can be made as
follows. The LSPM mix shows the lowest stiffness at higher reduced frequencies (physically
representing lower temperatures and/or higher loading frequencies) and a relatively high
stiffness at lower reduced frequencies (physically representing higher temperatures and/or
lower loading frequencies), which is presumably due to the high air void content and the
interlocking aggregate structure formed by large particles, respectively. Conversely, the
Superpave-25 mix demonstrates the highest stiffness at higher reduced frequencies and a
relatively low stiffness at higher reduced frequencies. The F-1 and F-2 mixes are asphalt-rich
mixtures primarily designed to mitigate the fatigue and reflective cracking. They have the
same aggregate structure design but it can be seen that the master curve of F-2 consistently
lies above that of F-1 due to the use of SBS modified binder with a higher PG grade.
With the dynamic modulus test results at hand, the coefficients in the relaxation
modulus function, can be conveniently calculated by following the algorithms described in
Park and Schapery (1999). Basically, the dynamic modulus is converted into the relaxation
modulus by solving a system of linear algebraic equations for unknown Prony coefficients
(Ei), while the time constants (i) are specified a priori, usually a decade apart. The time-
temperature shift factor function, Equation (8.7), is applied to shift the obtained relaxation
modulus function from the reference temperature to the actual test temperature of each
fatigue test. Afterwards, the pseudo strain is calculated.

141
 30000 100000
SMA
25000 Superpave-19
Superpave-25
20000 10000
LSPM

|E*| (MPa)
|E*| (MPa)

F-1
15000
F-2
10000 1000

5000

0 100
1E-08 1E-06 1E-04 1E-02 1E+00 1E+02 1E+04 1E-08 1E-06 1E-04 1E-02 1E+00 1E+02 1E+04
Reduced Frequency (Hz) Reduced Frequency (Hz)

(a) Dynamic modulus master curves in arithmetic scales (b) Dynamic modulus master curves in logarithmic scales

40 2.0

35
0.0
30
Phase Angle (deg)

Log Shift Factor

25 -2.0
20

15 -4.0

10
-6.0
5

0 -8.0
1E-08 1E-06 1E-04 1E-02 1E+00 1E+02 1E+04 0 10 20 30 40 50 60

Reduced Frequency (Hz) Temperature (°C)

(c) Phase angle master curves (d) Shift factor functions

Figure 8-3. Dynamic modulus test results for the asphalt mixes in Binzhou sections.

8.5.2 Direct Tension Cyclic Fatigue Test

The fatigue resistance of asphalt mixtures is typically assessed by subjecting
cylindrical specimens in direct tension cyclic tests in accordance with AASHTO TP 107-14
testing protocol (AASHTO, 2014). The experiment is usually conducted in the displacement
controlled mode given the practical difficulties associated with other modes and the purpose
of minimizing the accumulation of plastic deformation. Each test is continued until the
specimen is pulled apart, and material failure and thus the fatigue life are determined as when
the calculated phase angle begins to drop (Reese, 1997). Obviously, extension of the VECD
theory to the material status with presence of macrocracks is invalidated.

142
 In the early practices the specimen geometry of 75 mm in diameter and 150 mm in
height was used with a strain-measuring gauge length of 100 mm in the middle along the axis
of the specimen. However, it has been observed that chances are good that specimens fail in
the proximity of ends beyond the gauge points. Such end failures may cause an inaccurate
measure of deformation, and more importantly the true fatigue life is not able to be
determined. Besides, the damage characteristic curves generated from tests with end failures
are usually terminated at lower S-values as higher levels of damage occur at locations beyond
the measurable range. Hence, for a complete material characterization the middle failure is
required in which case the material macroscopic status from intactness through failure is
fully captured. In order to raise the probability of middle failure, a comprehensive
exploration is accomplished on the study of various geometric dimensions and gauge lengths
by Lee et al. (2015) and it is recommended from finite element simulations and experimental
investigations that the geometry with 100 mm in diameter and 130 mm in height be used
along with a 70 mm gauge length. Nevertheless, it is worthwhile to point out that
experimental evidences also show an excellent overlap between damage characteristic curves
from both middle and end failures.
In this study, the above mentioned geometry is used for all fatigue tests. For each
mix, a minimum of two or three replicates were tested, depending on the material
availability. All tests ended with middle failures. In order to characterize the failure criterion,
Equation (8.8), three levels of fatigue life were targeted: Nf < 5,000, 5,000 < Nf < 25,000, and
Nf > 25,000, by properly choosing the displacement amplitude based on trial tests or
experience on similar mixes. All fatigue tests were performed at 19°C and 10 Hz. The
resulting damage characteristic curves were then fitted using Equation (8.4), and the relations
are presented in Figure 8-4.

143
 1.0
SMA
Superpave-19
0.8
Superpave-25
LSPM
C 0.6 F-1
F-2
0.4

0.2

0.0
0.E+00 2.E+05 4.E+05 6.E+05 8.E+05 1.E+06 1.E+06 1.E+06
S

Figure 8-4. Damage characteristic curves for the asphalt mixes in Binzhou sections.

As previously stated, C is an indicator of material integrity while S quantifies the

internal damage in the material. It is shown in Figure 8-4 that the material integrity of LSPM
deteriorates quickly with damage, which is attributable to the high air void content and the
open gradation. Conversely, for the F-2 mix, material maintains its integrity even when a
significant amount of damage is experienced, and this can be explained by the high asphalt
content and the use of SBS modified binder. In addition, a general observation can be made
by comparing all six mixes that fatigue resistance can usually be enhanced by the use of
smaller aggregate and modified asphalt binder. However, a better material property is not
always a guarantee of a better performance in the pavement since the pavement structure and
the material property in other layers should also be taken into consideration. Therefore, a fair
and rigorous comparison should be achieved by carrying the damage characteristic relations
obtained on the material level into a finite element structural analysis program, as will be
presented in the next section.

144
 For each fatigue test, the GR was calculated via Equations (8.5) and (8.6), and is
plotted with respect to its fatigue life as shown in Figure 8-5. Also illustrated in the figure are
the best fits for each material using Equation (8.8). As can be observed from Figure 8-5,
when a given amount of energy is to be consumed, LSPM would fail at early cycles, whereas
the F-2 mix is able to dissipate the energy in a slower pace so that the material can stand a lot
more cycles of loading before failure. The other four mixes demonstrate roughly the same
fatigue resistance with F-1 and SMA being slightly superior to Superpave-19 and 25. Note
that the line of Superpave-25 lies immediately above the line for Superpave-19 indicating a
marginal advantage in fatigue resistance, which is contrary to what is observed in Figure 8-4.
This inconsistency is presumably related to the variability in specimens, testing, and the
process of finding the turning point in phase angle to determine Nf. After all, the ranking of
fatigue resistance is expected to be the same using either the damage characteristic curves or
the failure criterion lines. Nevertheless, the material fatigue performance should be finally
assessed in the structural simulation.

1.E+07
SMA
Superpave-19
Superpave-25
1.E+06
LSPM
ACDPSE

F-1
F-2
1.E+05

1.E+04
1.E+03 1.E+04 1.E+05 1.E+06 1.E+07
Nf

Figure 8-5. Relations between ACDPSE and fatigue life for each mix in Binzhou sections.

145
8.6 Fatigue Performance Predictions

8.6.1 The Finite Element Package - LVECD

The finite element structural analysis package has been developed and improved for
pavement response and performance simulations along with the evolution of the VECD
theory at NC State University (Underwood et al. 2012, Cao et al. 2012, Norouzi and Kim
2015). The latest version is called the Layered ViscoElastic pavement analysis for Critical
Distresses (LVECD). In the following, the program input of material properties, traffic, and
climate data are explained, and some critical assumptions and simplifications are briefly
reviewed. For details in the algorithm, refer to Eslaminia et al. (2012).
All asphalt materials are treated in the framework of linear viscoelastic continuum
damage theory. The Prony coefficients are required for material stiffness, and the parameters
in Equation (8.4) and (8.8) are needed for predicting fatigue performance. The unbound
materials are considered in the linear elastic domain. The elastic modulus is assumed as
1,500 MPa for the lime fly-ash treated aggregate, 500 MPa for the lime fly-ash treated soil,
85 MPa for the 30 cm lime stabilized subgrade, and 35 MPa for the natural subgrade. For all
materials, Poisson’s ratio is deemed as a constant irrelevant to temperature and loading
conditions.
Traffic in LVECD is dealt with in a number of ways. Either a wheel (single/dual), an
axle (single/tandem/tridem), or a vehicle can be defined with each tire having its own load
and geometric characteristics. The traffic volume per day on the design lane is then given as
the total number of passes of the wheel, axle, or vehicle, whichever is defined in the previous
step. In the present simulation, a single axle with dual tires is employed as the unit for traffic.
To accommodate the potential overloading, the axle load is set as 100 kN with tire pressure
of 1.0 MPa. The number of passes of this traffic unit per day on the design lane is set at
10,000.

146
 Temperature profile for the pavement structure is input on the hourly basis, and can
be prepared using the climate information in the Enhanced Integrated Climate Model
(EICM). For simplicity, only one year temperature data are required in LVECD; this history
is then repeated for the remaining years of the whole design life. In this study, the climate
information for the pavement site in Shandong Province is approximated by the averaged
data over the state of Virginia in United States considering the geographic similarities.
In LVECD, adjacent layers are assumed as fully bound at the interface. The default
setting for the meshing gives 11 and 101 nodes for the vertical and transverse directions,
respectively, for each asphalt layer, which yields a total number of 1,111. The mesh size for
the unbound elastic materials is relatively coarse. The responses of stress, strain, and
displacement at each node under the moving loading of the traffic unit are conveniently
solved via the Fourier transform techniques in both time and space domains.

8.6.2 Simulation of Fatigue Performance and Discussions

The fatigue performance of each section is predicted with a 15 year analysis period.
As previously explained, the pseudo stiffness C is an indicator of material integrity, and thus
the C-contour can be utilized to examine and compare the performances of all sections, as
presented in Figure 8-6. Note that the Section Control shown in Figure 8-6 (e) is a virtual
section which is designed as the same with Section 4 except for the use of the 15 cm
Superpave-25 layer in place of the LSPM layer. The Section Control is added in the
simulation only to make a parallel and thus fair comparison with Section 4 to assess the
potential of LSPM in resisting fatigue cracking. Also note that in all contours, the maximum
C value is set at 0.8 instead of 1.0 considering inherent defects in virgin materials and certain
adverse effects due to compaction and aging during pavement construction.

147
 The fatigue performance can be compared by examining the lowest C-value shown
beside the color bar in each section. The ranking among all Binzhou sections from superior to
inferior is Section 1, 3, 2, Control, 4, and lastly Section 5. As expected, this sequence follows
the general ranking of the thickness of asphalt pavement. Section 1 has the largest thickness
and thus demonstrates the best fatigue performance at the end of the 15 year simulation. The
material damage is seen to concentrate in the layer of Superpave-19 as shown in Figure 8-6
(a). Owing to the use of SBS modified binder in the bottom asphalt layer, Section 3 exhibits a
slightly better performance than Section 2. Another noticeable difference between the two is
that the use of better materials also helps to shift the location of fatigue damage to upper
layers. In Section 2, fatigue damage is concentrated at the bottom of the fourth layer, whereas
in Section 3 the location of major damage has shifted to the second layer. This finding is
illuminating in the perpetual pavement design which is aimed to avoid structural failure at
lower pavement layers.

148
 0 0.8
5

12.5 0.795

Depth (cm)
23
0.79

33.5
0.785
42.5

50 0.78
-1.8 -1.5 -1 -0.5 0 0.5 1 1.5 1.8
X (m)
(a) C-contour for Section 1
0 0.8
5
0.79
Depth (cm)

12.5 0.78

0.77
21.5
0.76
30.5
0.75

38 0.74
-1.8 -1.5 -1 -0.5 0 0.5 1 1.5 1.8
X (m)
(b) C-contour for Section 2
0 0.8
5
0.79
Depth (cm)

12.5

0.78
21.5

0.77
30.5

38 0.76
-1.8 -1.5 -1 -0.5 0 0.5 1 1.5 1.8
X (m)
(c) C-contour for Section 3

0 0.8
4
0.75
Depth (cm)

10
0.7

18 0.65

0.6

0.55
33 0.52
-1.8 -1.5 -1 -0.5 0 0.5 1 1.5 1.8
X (m)
(d) C-contour for Section 4
0 0.8
4
0.75
Depth (cm)

10

0.7
18
0.65

0.6
33 0.58
-1.8 -1.5 -1 -0.5 0 0.5 1 1.5 1.8
X (m)
(e) C-contour for Section Control
0 0.8
Depth (cm)

4 0.6
9
0.4
15 0.2
-1.5 -1 -0.5 0 0.5 1 1.5
X (m)
(f) C-contour for Section 5

Figure 8-6. C-contour of the Binzhou sections after 15 year simulations.

149
 For Section 4 and Control, fatigue cracking would initiate at the bottom of asphalt
pavement according to Figure 8-6 (d) and (e). In comparing the lowest C-value in these two
sections, it appears that the utilization of LSPM at the bottom of asphalt layer would make a
poor design compared to Superpave-25 in the sense of fatigue cracking. Recall in Figure 8-5
and 6 that on the material level LSPM shows the least desired fatigue resistance due to its
special aggregate gradation. Therefore, the above observation in the parallel structural
comparison should be expected. However, Section 4 may actually perform better than
Section Control because the placement of LSPM immediately above the semi-rigid base may
potentially reduce the distresses caused by reflective cracking. Therefore, the previous
conclusion between the two sections is deemed valid only as far as bottom-up fatigue
cracking is concerned. In the current LVECD framework the mechanisms underlying
reflective cracking which initiates in unbound materials are not incorporated and thus a fair
comparison would require more advanced and comprehensive tools.
Section 5 is the conventional design of semi-rigid asphalt pavement that presents the
worst fatigue performance at the end of 15 year. As shown in Figure 8-6 (f), damage is
concentrated at the bottom of this thin asphalt pavement. The minimum C-value suggests that
Section 5 may not be able to maintain structural integrity up to 15 years due to fatigue
cracking, even if reflective cracking is not considered.
An alternative approach to fatigue performance inspection is to investigate the
percentage of material points with failure in the pavement cross section. The failure
percentage is simply calculated as the ratio of the number of material nodes with failure to
the total number of nodes in the cross section. The evolution of failure percentage with
respect to the number of axle passes is plotted in Figure 8-7 for each section.
The ranking of fatigue performance according to Figure 8-7 from superior to inferior
is Section 1, Control, 3, 2, 4 and Section 5, which is the same with the C-based ranking
except for the position of Section Control. This inconsistency is attributed to the fact that a
pavement having a lower minimum C-value may possibly show a smaller area of material
failure. In fact, a lower C-value may not necessarily represent a worse fatigue performance.
Recall that the process of material deterioration follows the damage characteristic curve, and

150
the C-contours shown in Figure 8-6 are generated based on this principle without considering
when failure would occur. Hence, the predicted C-values for the material nodes in the cross
section can be asymptotically decreasing to zero under repeated fatigue loading. Yet, beyond
failure the VECD theory is no longer applicable, which means that any further calculation of
the pseudo stiffness is actually invalid for any material point that has already failed. Given
these considerations, it is deemed that the comparison from the perspective of failure
percentage is more reasonable in ranking different structural designs.
Again, it should be emphasized that all the above simulations and conclusions are
conducted without involving the role of reflective cracking and other complexities such as
aging, healing and shearing. Judging purely from the fatigue performance, it is recommended
that Section 1 and 3 be the ideal candidates for perpetual pavement design. All other sections
are ruled out because of either severe damage or initiation of damage at lower structural
layers.

20
S1
18 S2
S3
Failure Percentage (%)

16
S4
Year = 10

14 Control
12 S5

10
8
6
4
2
0
0.E+00 2.E+07 4.E+07 6.E+07
Number of Axle Passes

Figure 8-7. The evolution of failure percentages for Binzhou sections after 15 year
simulations.

151
8.7 Conclusions
This section presents a systematic VECD framework of evaluating fatigue
performance for asphalt pavement. Asphalt mixture is treated as a linear viscoelastic
material, while the constitutive modeling of fatigue damage under repeated loading is
fulfilled by identifying the damage characteristic curve for each mix. The VECD framework
is then completed with the unified failure criterion which defines the applicable domain of
the theories. The model parameters are determined through dynamic modulus tests and direct
tension cyclic fatigue tests. All the material models are integrated in the finite element
package LVECD, which is used for fatigue performance predictions on Binzhou perpetual
pavement sections. The 15 year fatigue simulations suggest that Section 1 and 3 should be
used as the perpetual pavement design.

152
CHA
APTER 9 CALIIBRATIO
ON OF F
FATIGUE
E CRAC
CKING
MO
ODEL FO
OR LVEC
CD PROGRAM

9.1 Introducttion
Fatigue crracking that is due to rep
peated wheell loads is onne of the major distress tyypes
c occur in flexible pav
that can vement system
ms. The actiion of repeatted traffic looading inducces
both tensile and shear
s stressees in all the asphalt
a layerrs and leads tto a reductioon in the
structtural integritty of the pav
vement. Fatig
gue cracks innitiate at thee points wherre critical tennsile
strain
ns and stresses occur. Deepending on the dominannt mechanism
m, the devellopment andd
propaagation of faatigue crackss inside a pav
vement struccture can be categorizedd into two maajor
phasees: pre-localiization and post-localiza
p ation. In genneral, pre-loccalization is m
manifested bby
the in
nitiation and propagation
n of microcraacks, and poost-localizatiion is manifeested by the
formaation and pro
opagation off macrocrack
ks.
The locatiion of the crritical strain//stress is deppendent uponn several facctors. The moost
important one is the
t stiffness of the asphaalt layer. In aaddition, it sshould be reccognized thaat the
maxim
mum tensilee strain that develops
d witthin the paveement system
m might not be the mostt
critical or damaging factor, because
b the critical
c strainn is a functioon of the stiff
ffness of the mix
and th
he material’s integrity. Because
B the properties of asphalt mixxes in a layeered pavemeent
system
m change th
hroughout thee depth of th
he pavementt, these variaations will evventually afffect
the lo
ocation of the critical straain that causses fatigue ddamage.
It is comm
monly assum
med that fatig
gue crackingg typically innitiates at thee bottom of tthe
asphaalt layer and propagates to the surfacce. This so-ccalled ‘bottom
m-up’ crackking is due too the
bendiing action off the pavemeent layer thatt results in fl
flexural stresss developingg at the bottoom
of thee bound layeer. However,, various studies have alsso clearly deemonstrated that fatigue
crack
king may also initiate fro
om the top an
nd propagatee downwardd (top-down ccracking).

153
 The aforementioned common understanding of fatigue failure due to bottom-up
cracking has been challenged in recent years as more and more agencies have begun to report
another form of failure, top-down fatigue cracking (Myers et al. 1998, Pellinen 2002, Myers
and Roque 2001, Uhlmeyer 2000). At the time of this study, universal agreement has not yet
been reached as to the exact causes of this type of distress (Al-Qadi and Yoo 2007).
Nevertheless, it can still be stated with certainty that the overall driving conditions for top-
down fatigue appear to differ from those that cause the more common bottom-up fatigue
cracking. Thus, the material characteristics that are tied most closely to pavement
performance in terms of bottom-up and top-down fatigue cracking also may differ. For
example, pavements constructed with materials that exhibit a strong propensity for aging
may be significantly more sensitive to the top-down cracking phenomenon than pavements
constructed with materials that have a relatively weak propensity to age. However, because
aging is a top-down process, the bottom-up cracking phenomenon may be affected only
slightly.
Layered Viscoelastic pavement analysis for Critical Distresses (LVECD) computer
program that has been developed in North Carolina State University is capable of predicting
the fatigue cracking in both bottom-up and top-down modes using the Simplified
Viscoelastic Continuum Damage (S-VECD) model. However, this software has not been
fully calibrated with the field observations yet and the LVECD damage-to-field cracking
transfer function needs to be developed to truly predict the amount of cracking in the field.
Therefore, this chapter focuses on the calibration of the fatigue cracking model that is used in
the LVECD software. This section consists of an overview of traditional and energy-based
fatigue failure criteria, the field data collection procedure, the LVECD program simulation
process, and the transfer function development that is based on the field data and simulation
results.

154
9.2 Fatigue Cracking Prediction Models

9.2.1 Traditional Tensile Strain-Based Fatigue Cracking Model

The MEPDG utilizes the incremental damage concept in predicting alligator cracking.
The incremental damage concept simulates ways that pavement distresses in the field
develop incrementally over time. The Asphalt Institute’s (AI’s) fatigue cracking model is
employed in the MEPDG to predict alligator cracking, as shown in Equation (9.1). In
addition to the AI model, the MEPDG employs Miner’s law, as presented in Equation (9.2),
to convert the predicted number of cycles to failure (Nf) to the equivalent damage for
different combinations of traffic loading and environmental conditions.
 f 2k f 2  f 3k f 3
1  1 
N f  0.00432C  f 1k f 1     (9.1)
 t   | E* | 
where
C = 10M,
 Vb 
M = 4.84   0.69  ,
 Va  Vb 
f1, f2, f3 = local calibration factors, and
kf1, kf2, kf3 = model calibration factors.
T
ni
Damage   (9.2)
i 1 N fi

For a particular pavement structure with a HMA bottom layer, the pavement response
prediction models within the MEPDG employ linear elastic theory to predict the critical
tensile strain (εt) at the bottom of the asphalt layer for a given combination of traffic loading
and environmental conditions. Knowing the dynamic modulus (│E*│) value for the bottom
HMA layer, the critical tensile strain value, and some mixture volumetric information, the Nf
model shown in Equation (9.1) calculates the number of cycles that is required to fail the
pavement under specific combinations of traffic loading and environmental conditions.

155
 To calculate the amount of damage that the pavement has experienced, Miner’s law,
shown in Equation (9.2), is used to sum up all similar combinations of traffic loading and
environmental conditions and divide the summation by the Nf to calculate the percentage of
damage.

9.2.2 Energy-Based Fatigue Cracking Model

One of the primary material functions that can be determined from the S-VECD
model is the energy-based failure criterion. During cyclic fatigue testing, the maximum
stored pseudo strain energy can be determined at the stress peak points in each cycle. The
amount of energy at each cycle can be calculated using Equation (2.21).
The maximum stored pseudo strain energy at each cycle represents the material’s
ability to store energy at that particular time. The material loses its stored energy as the
damage grows for the same magnitude of applied pseudo strain due to the reduction in
pseudo stiffness. The difference between the maximum stored pseudo strain energy and the
corresponding undamaged state is referred to as the total released pseudo strain energy and
is denoted as WRC. Therefore, a term, GR, is defined as the rate of change of the averaged
released pseudo strain energy (per cycle) throughout the entire history of the test and can be
defined as Equation (2.24). Details regarding the GR method and its corresponding
calculations can be found in Section 2.4.

9.3 Calibration of Energy-Based Fatigue Criterion

In order to provide credibility to the new LVECD-based performance prediction
methodology, the theoretically predicted distress models must be calibrated to be applicable
to real asphalt pavements. The following steps were taken in this study to calibrate the
fatigue cracking model in the LVECD program.
 Field performance data, including the percentage of the fatigue cracking/length of
longitudinal cracks, were collected from the site projects and converted to the
percentage of cracking (% cracking) using the LTPP procedure.

156
  Materials were acquired based on the accuracy of the field data and binder/aggregate
availability. More than 33 pavement sections with different characteristics (e.g.,
modified binder, RAP content, structure, warm mix additive) from different locations
in the United States, Canada, South Korea, and China were selected for study.
 Samples were fabricated based on the field air void content of the asphalt layer.
Mechanical testing (dynamic modulus testing, S-VECD cyclic testing, and TSS
testing) characterization was performed to obtain the required parameters for LVECD
program analysis.
 Simulations were run using the experimental data, environmental conditions,
pavement structures, and traffic level. The climatic data were collected directly from
the EICM database.
 As a first step for verification, the predicted damage results obtained from each
simulation were compared to the measured cracking that was observed in the field.
 The predicted damage was correlated to the measured cracking in the field using the
sigmoidal function (as presented by many researchers as the best indicator of fatigue
cracking growth in real pavements).
The calibration data were collected at the same time for all types of fatigue cracking, as the
same sections were used for both bottom-up and top-down cracking calibration. In the
following sections, each step of the calibration process is discussed in detail.

9.4 Field Cracking Data Collection

The first step in the calibration process was to identify potential test sites. As part of
the FHWA’s HMA-PRS project, 33 pavement sections that are known to have experienced
sufficient cracking for the study were selected. These test sections cover a diverse range of
site features, including climate, traffic level, subgrade soil, and pavement structural cross-
section characteristics. All the data required for executing the models, including
measurements of fatigue distress, were extracted directly from the site, reviewed for accuracy
and completeness, and incorporated into the project database.

157
 These data elements include performance observations, material properties, traffic
and climatic characteristics, pavement cross-sections, and foundation information. The crack
measurement procedure is discussed in the following paragraphs.

9.4.1 Long Term Pavement Performance (LTPP)

For this study, fatigue cracking within a single site was divided into different
categories (i.e., low, medium, and high levels) in accordance with the procedure proposed by
the LTPP project. The recorded amounts of alligator and longitudinal cracking in the wheel
paths as obtained from the field surveys were converted to percentages based on the total
area of the site. In order to obtain the % cracking value, the crack length was converted to a
unit area by applying a standard width of wheel path (0.3 m or 1 ft) to the crack length, as
suggested by Jackson and Puccinelli (2006). The crack lengths were summed arithmetically,
regardless of the severity of the distresses, for the LVECD program fatigue cracking
calibration.

9.4.2 National Center for Asphalt Technology (NCAT)

The amount of cracking measured by the NCAT is reported in three separate
categories for each pavement section: right wheel path (RWP), left wheel path (LWP), and
percentage of the cracked area in a lane. According to the procedure adopted by the NCAT,
each crack exhibits influence for six inches. When multiple cracks are present, the connected
cracks have zones of influence that overlap, and simple cracking then becomes referred to as
cracking areas. For this study, simple and complex cracking patterns were converted to
cracking areas either by calculating the length of simple cracks and multiplying by 1 foot or
by directly calculating the area of influence of complex cracking, respectively.
The wheel paths at the NCAT test track are 3 feet wide and lie 1.5 feet from the
edge/centerline of the lane. Therefore, percent wheel path cracking is defined as the
distressed area over the total wheel path area. Percent lane also is specified as the total
cracked area by the total lane area in the section (150-ft length by 12-ft width) as presented in
Figure 9-1.

158
9.4.3 Manitoba Infrastructure and Transportation (MIT)
The crack length is reported in three different categories (low, medium, and high) for
the MIT pavement sections (500-ft length and 12-ft width). For this study, the LTPP program
procedure was employed to convert the crack length to the percent cracked area without
considering any weight for different crack types.

9.4.4 Federal Highway Administration Accelerated Loading Facility (FHWA-ALF)

All of the fatigue cracking distresses were measured and reported in terms of percent
cracked area for the FHWA-ALF pavement sections in accordance with the LTPP
specifications.

Right Wheel Path (RWP)

1.5 ft

1.5 ft
12 ft
1.5 ft

1.5 ft

Left Wheel Path (LWP)

Figure 9-1. NCAT test track section area.

159
9.4.5 Korea Expressway Corporation (KEC)
Similar to the FHWA-ALF project, the KEC defines fatigue cracking in terms of
percent cracked area according to the procedure proposed by the LTPP program.

9.5 LVECD Program Simulation Process

9.5.1 Stress-Strain Calculation Method in LVECD Program

Reasonable stress-strain analysis is a key component in pavement design and for
pavement life predictions. Given the complexity of variables such as pavement life, traffic
loading, and temperature variations, various approximate methods can be used to predict
pavement performance. Despite differences in their assumptions, all of these prediction
methods aim to reduce analysis that can take millions of cycles over several years to a few
hundred analyses for a single cycle of loading.
The most basic method that is used in the Pavement ME program is layered elastic
analysis (LEA), where the pavement is idealized as a layered elastic system under a
stationary axisymmetric load. In this method, the normal and radial stresses/strains often are
computed using a Fourier-Bessel transform (Huang 2003). However, LEA can lead to
inaccurate responses because (1) traffic loading (i.e., tire pressure) is neither stationary nor
circular in reality, and (2) asphalt concrete exhibits significant viscoelastic behavior,
especially under moving loads.
Layered viscoelastic moving-load analysis (LVEMA) is an improvement over LEA in
that the viscoelasticity and the moving load effects are handled efficiently with the help of
Fourier transforms. LVEMA is thus more appealing than LEA for pavement stress
analysis, although the stress redistribution effects due to damage still are not captured.
Taking LVEMA into consideration, the NCSU research team has developed a three-
dimensional layered viscoelastic FEM tool with moving loads to calculate the responses and
performance of asphalt pavements. The advantages of this analysis tool, i.e., the LVECD
program, can be summarized as follows:

160
  Asphalt concrete shows a significant amount of viscoelasticity, and thus, the behavior
of asphalt concrete depends on both loading frequency and temperature. Pavements
undergo wheel loading at a vast range of frequencies as well as different temperature
distributions throughout the depth of the pavement over the lifetime of the pavement,
such that the whole structural behavior of the pavement changes significantly over
time. The LVECD program is able to account for this loading frequency and
temperature dependency of asphalt concrete at different depths of the pavement.
 Each pavement layer in the analysis tool can be assigned its own material and
structural properties. For example, for granular material, the LVECD program can
assume anisotropic elasticity with the axis of symmetry in the depth direction.
 The LVECD program takes advantage of the assumptions of Fourier transform and
time-scale separation to reduce computational costs. This three-dimensional moving
load layer analysis considers damage (rutting and fatigue) and can still complete a
simulation within a relatively short time compared to other viscoelastic analysis
programs.

9.5.2 Damage Calculation in the LVECD Program

The LVECD program can evaluate fatigue cracking using the VECD model (Kim et
al. 2009) where asphalt concrete is modeled as a viscoelastic material with microcrack-
induced damage. The 3D-Move program also is capable of fatigue performance analysis, but
its framework is based on the conventional tensile strain-based model.
The VECD model (Kim et al. 2009) can be used to represent the effect of damage on
the material’s stiffness, where S is the damage parameter and  R is the pseudo strain. Then,
the damage characteristic curve can be obtained through laboratory cyclic testing. The
damage evolution law is given by
 
ds  w R   1 c 2
     ER   R   (9.3)
dt  s   2 s 

where α is a material model parameter.

161
 Equation (9.3) can be extended to the multiaxial case using the definition for a pseudo
strain energy density function.
1
WR  A11e2  A22ed2  2 A12ed e  A44   13R   23R   A44   13R  es   (9.4)
2

where
ev  11R   22R   33R ,

e
ed  33R  v , and
3
es   11R   22R .

For the multiaxial state of stress, the damage evolution law can be written as
 2
s 1 c  c   1 R R 
     ev  ed  (9.5)
t 2 s  s   3 
Using the chain rule and integrating with respect to time, the change in the
normalized stiffness value in a single cycle can be obtained easily, which becomes Equation
(9.6).
 2
1 c  c  Tcycle 1 R R 
c   
2 s  s  
0  ev  ed  dt
3 
(9.6)

where Tcycle is the time period of the cyclic moving load. The LVECD program calculates the
damage increment for each segment assuming constant pseudo stiffness at the beginning of
each life stage. Then, nonlinear extrapolation is conducted by solving the following ordinary
differential equation to calculate the damage during each stage.

N segments
 dc  ds 
c      ,
 ds  dt 
i 1 (9.7)
1  1 
1
1
c  c   log(c )   b
 c     ,
n  c0   log(c0 ) 

162
where C0 is the pseudo stiffness at the start of each life stage, n is a cycle in each analysis
segment, and N is the total number of cycles at the end of each analysis segment.
The LVECD program can calculate damage C for a large N and then calculates the

WRc .
1
WRc  1  c W R (9.8)
2
At the end of each stage, the total damage increment that is due to traffic and thermal
loading is calculated. The LVECD program uses weighted averages of the damage
increments of all the segments in each stage using Equation (9.9)
Segments

 N i  C i (C 0 )
(9.9)
Ccomb  i
Segments


i
Ni

Using the nonlinear extrapolation formula, the final damage value for each point can
be calculated using Equation (9.10). Note that there is no need to extrapolate C; the C value
needs to be found only at the end of the last cycle of loading. The LVECD program finds the
final C value for the mesh points. Because all the output data should be shown for the
evaluation points, the LVECD program uses interpolation to find the damage values.
1
1 (1 )(1 )
c  c   log  c   b
 Ccomb     (9.10)
n  c0   log  c0  
Damage increases as the number of cycles increases according to the damage
characteristic curve. As the damage increases, an asphalt element will fail when the element
reaches the ultimate state, which is defined by the GR versus Nf relationship. The LVECD
program finds the number of cycles to failure (Nf) in each element for a segment of applied
traffic and thermal loading condition by calculating the GR value for the given loading
conditions in the segment and by finding the corresponding Nf value to the GR value using the
GR versus Nf relationship. Then damage index value is calculated as the ratio between the
actual experienced numbers of cycles in that loading segment to the Nf value calculated for
that loading segment.

163
 The damaage index vaalues are cum
mulative and are added uup at the end of each life
stage according to Miner’s laaw. The dam
mage index iss calculated ffor each noddal point andd
varies between 0 (no damagee) and 1 (failure).

9.6 Analysis of
o Measurred Fatigue Crackin
ng in LTPP
P Databasse
To find th
he true param
meters that arre involved in the transffer function, the field datta
must be checked for general reasonablene
r ess, and anyy trends obseerved from thhese data shoould
be ex
xamined. Thee Pavement ME design guide
g providdes informatiion about the measured data
obtain
ned from thee LTPP dataabase.
As a first step, the thickness valuees of all the asphalt conccrete layers ffor each secttion
were extracted from the datab
base and theen added togeether to obtaain the total tthickness off the
asphaalt concrete layers
l for eaach section. Figure
F 9-2 shhows the ploots of the tottal fatigue
crack
king percentaages from th
he 640 LTPP
P sections thaat were evaluuated as a fuunction of assphalt
thickn
ness. This fiigure clearly
y indicates th
hat maximum
m cracking ggenerally occcurs at an aspphalt
layer thickness beetween four inches and six
s inches. Itt also showss that the am
mount of craccking
dram
matically decrreases when the asphalt layer
l thickneess is more tthan six inchhes.

Fiigure 9-2. LT
TPP alligato
or cracking vvs. asphalt laayer thicknesss.

164
 Another major
m factor that can affeect fatigue crracking is thhe mean annuual air
temperature (MA
AAT). As sho MAATs for thhe LTPP sections
own in Figurre 9-3 (a) annd (b), the M
uated ranged
evalu ximately 29F to approxximately 77F. An initiall expectationn of
d from approx
the allligator crack ore cracking would occurr in cold reggions and less
king results was that mo
crack
king would occur
o in hot regions.
r How
wever, the pllots indicatee that the occcurrence andd
amou
unt of alligattor cracking are very close for all reggions and inddependent of the MAAT
T
t). Thus, the MAAT app
(site environment
e pears to havee little to no significant eeffect on the
measured alligato
or cracking reported
r in th
he LTPP dattabase.

(a)

(b)
gure 9-3. (a) Alligator crracking vs. MAAT
Fig M rangees and (b) maximum alliigator cracking.

165
9.7 Fatigue Cracking Transfer Function
The next step of the calibration process is to derive an appropriate transfer function
that relates the predicted damage to the measured fatigue cracking. The performance index
used in the LVECD program is the so-called damage area (%), which is the number of nodes
where the asphalt mixture fails under the wheels (N/Nf = 1, as discussed in Section 10.5.2)
over the total number of nodes throughout all the asphalt layers expressed in percentage.
Each layer has 101 nodes in length and 11 nodes in depth, for a total of 1,111 nodes for each
layer.
To specify the appropriate procedure to define damage in LVECD program and to
introduce indices for the top-down and bottom-up cracking in the pavement sections, two
methods were adopted to check the fatigue cracking behavior in the LVECD simulations: (1)
the failure points in the area of 40 cm of right and left side of each wheel with the pavement
depth was counted and divided by the total points for the top and bottom halves of the layer
thickness as demonstrated in Figure 9-4, and (2) the averaged N/Nf values through the
asphalt layer was calculated and presented as percentage for both top and bottom halves of
the asphalt layer. Figure 9-5 to Figure 9-12 present the comparison of fatigue behavior
trends obtained from the aforementioned procedures for both halves of the pavement layer
with the field performance data.
According to the results obtained from the analysis, considering only top or bottom of
the asphalt layer cannot lead to the true performance prediction and it is necessary to
consider the whole asphalt layer to define the damage in LVECD. However, these two
methods can be identified as two different indices to predict the top-down or bottom-up
cracking within the asphalt layer.

166
 Asphalt Layer

80 cm
Base Layer

Subgrade

Figure 9-4. Schematic definition of damage area in LVECD.

12 30
N-Nf (a) N-Nf (b)
10 (Top) 25 (Bottom)
Damage Area (%)

% Damaged Area
8 20

6 15

4 Control
10 Control
CR-TB CR-TB
2 SBS-LG
5 SBS-LG
Terpolymer Terpolymer
0 0
1.E+04 1.E+05 1.E+06 1.E+07 1.E+04 1.E+05 1.E+06 1.E+07
No. of Cycles No. of Cycles

100
Field (c)
80
Cracked Area (%)

60

40
Control
CR-TB
20
SBS-LG
Terpolymer
0
1.E+04 1.E+05 1.E+06
No. of Cycles

Figure 9-5. Damage definition obtained from method 1 for FHWA-ALF: (a) top half of
asphalt layer, (b) bottom half of asphalt layer, and (c) field measurements.

167
 0.2 0.5
Average N/Nf (a) Average N/Nf (b)
(Top) (Bottom)
0.4
0.15
Average N/Nf

Average N/Nf
0.3
0.1
0.2
Control Control
0.05 CR-TB CR-TB
0.1
SBS-LG SBS-LG
Terpolymer Terpolymer
0 0
1.E+04 1.E+05 1.E+06 1.E+07 1.E+04 1.E+05 1.E+06 1.E+07
No. of Cycles No. of Cycles

100
Field (c)
80
Cracked Area (%)

60

40
Control
CR-TB
20
SBS-LG
Terpolymer
0
1.E+04 1.E+05 1.E+06
No. of Cycles

Figure 9-6. Damage definition obtained from method 2 for FHWA-ALF: (a) top half of
asphalt layer, (b) bottom half of asphalt layer, and (c) field measurements.

168
 35 60
N/Nf (a) N/Nf (b)
30 (Top) (Bottom)
50
Damaged Area (%)

Damaged Area (%)

25
40
20
30
C C
15
O O
20
10 R R
RW RW
5 FW
10 FW
AW AW
0 0
1.0E+05 1.0E+06 1.0E+07 1.0E+08 1.0E+05 1.0E+06 1.0E+07 1.0E+08
No. fo Cycles No. fo Cycles

60
Field (c)
50
Cracked Area (%)

40

30
C
O
20
R
RW
10 FW
AW
0
1.E+07 1.E+08
No. of Cycles

Figure 9-7. Damage definition obtained from method 1 for NCAT: (a) top half of asphalt
layer, (b) bottom half of asphalt layer, and (c) field measurements.

169
 0.6 1
Average N/Nf (a) Average N/Nf (b)
0.5 (Top) (Bottom)
0.8
Average N/Nf

0.4

Average N/Nf
0.6
0.3
C C
O 0.4 O
0.2
R R
RW 0.2 RW
0.1 FW FW
AW AW
0 0
1.0E+05 1.0E+06 1.0E+07 1.0E+08 1.0E+05 1.0E+06 1.0E+07 1.0E+08
No. fo Cycles No. fo Cycles

60
Field (c)
50
Cracked Area (%)

40

30
C
O
20
R
RW
10 FW
AW
0
1.E+07 1.E+08
No. of Cycles

Figure 9-8. Damage definition obtained from method 2 for NCAT: (a) top half of asphalt
layer, (b) bottom half of asphalt layer, and (c) field measurements.

170
 1 4
N/Nf (a) N/Nf (b)
(Top) (Bottom)
0.8

Damaged Area (%)

Damaged Area (%)

0.6
2
0.4
C C
15R 1 15R
0.2
50R 50R
50RSB 50RSB
0 0
1.E+05 1.E+06 1.E+07 1.E+05 1.E+06 1.E+07
No. of Cycles No. of Cycles

14
Field (c)
12
Cracked Area (%)

10

4 C
15R
2 50R
50RSB
0
1.E+04 1.E+05 1.E+06
No. of Cycles

Figure 9-9. Damage definition obtained from method 1 for MIT-RAP: (a) top half of asphalt
layer, (b) bottom half of asphalt layer, and (c) field measurements.

171
 0.06 0.25
Average N/Nf (a) Average N/Nf (b)
0.05 (Top) (Bottom)
0.2

Average N/Nf
Average N/Nf

0.04
0.15
0.03
0.1
0.02
C C
15R 0.05 15R
0.01 50R
50R
50RSB 50RSB
0 0
1.E+05 1.E+06 1.E+07 1.E+05 1.E+06 1.E+07
No. of Cycles No. of Cycles

14
Field (c)
12
Cracked Area (%)

10

4 C
15R
2 50R
50RSB
0
1.E+04 1.E+05 1.E+06
No. of Cycles

Figure 9-10. Damage definition obtained from method 2 for MIT-RAP: (a) top half of asphalt
layer, (b) bottom half of asphalt layer, and (c) field measurements.

172
 10 10
N/Nf (a) N/Nf (b)
(Top) (Bottom)
8 8

Damaged Area (%)

Damaged Area (%)

6 6

4 4
Control Control
2 Advera 2 Advera
Evotherm Evotherm
Sasobit Sasobit
0 0
1.0E+05 1.0E+06 1.0E+07 1.0E+05 1.0E+06 1.0E+07
No. of Cycles No. of Cycles
12
Field (c)
10
Cracked Area (%)

4
Control
Advera
2
Evotherm
Sasobit
0
1.0E+05 1.0E+06
No. of Cycles

Figure 9-11. Damage definition obtained from method 1 for MIT-WMA: (a) top half of
asphalt layer, (b) bottom half of asphalt layer, and (c) field measurements.

173
 0.5 0.4
Average N/Nf (a) Average N/Nf (b)
(Top) (Bottom)
0.4
0.3

Average N/Nf
Average N/Nf

0.3
0.2
0.2

Control 0.1 Control

0.1 Advera Advera
Evotherm Evotherm
Sasobit Sasobit
0 0
1.0E+05 1.0E+06 1.0E+07 1.0E+05 1.0E+06 1.0E+07
No. of Cycles No. of Cycles

12
Field (c)
10
Cracked Area (%)

4
Control
Advera
2
Evotherm
Sasobit
0
1.0E+05 1.0E+06
No. of Cycles

Figure 9-12. Damage definition obtained from method 2 for MIT-WMA: (a) top half of
asphalt layer, (b) bottom half of asphalt layer, and (c) field measurements.

Figure 9-13 (a) and (b) present the field measurements versus the predicted damage
obtained from the Pavement ME and LVECD programs, respectively. It seems that the same
trend is observed in both cases where there is a shift to the left with an increase in pavement
thickness. Therefore, it can be assumed that the same functional form that is in the Pavement
ME can be used to represent the field crack growth in the LVECD program. Figure 9-14
presents the fitted curves for predicting the field cracking using the sigmoidal function.
According to the observations made from Figure 9-13 (b) and Figure 9-14 (b), the following
assumptions can be made for the LVECD program transfer function:

174
  A sigmoidal function is the best representation of the relationship between cracking
and damage.
 The fatigue cracking value of 50 percent of the total area of the lane occurs at a
damage percentage of 40 percent in the LVECD program.
The fatigue cracking predicted damage transfer function is assumed to take the form
of a mathematical sigmoidal function that is similar to the Pavement ME transfer function.
The transfer function can be presented as Equation (9.11).
100
% FC  C1  C2 log 2.5 D
(9.11)
1 e
where
FC = fatigue cracking percentage in the field,
D = LVECD program-predicted fatigue damage in terms of percentage,
C1, C2 = regression coefficients.
To satisfy the second assumption, the C1 value must be equal to twice the negative
value of C2, as shown below:
C1  2C2 (9.12)
Following the aforementioned assumption, Equation (9.11) can be simplified as
Equation (9.16):
100
% FC  C2  2 log 2.5 D 
(9.13)
1 e
100
% FC  C2 (9.14)
 2.5 D 
log  
1 e  100 

100
% FC    2.5 D C2 
(9.15)
  100  e 
 loge .log10 
 
1 e  

100
% FC  0.43C2 (9.16)
 2.5D 
1  
 100 

175
 By considering 0.43C 2   , Equation (9.16) can be written as:

100
% FC   (9.17)
 2.5D 
1  
 100 
With more simplification of Equation (9.17), it will be found that there is a linear
correlation between the two logarithmic terms as presented in Equation (9.18). Figure 9-15
 100   2.5 D 
presents the plot of   1 versus   for different pavement sections.
 FC   100 

 100   2.5 D 
  1    (9.18)
 FC   100 
 100   2.5 D 
log   1    .log   (9.19)
 FC   100 

176
 (a)
60
10cm

16cm
50
20cm
Cracked Area (%)

40 30 cm

40 cm
30

20

10

0
0.1 1.0 10.0 100.0
Averaged Damage
e Area (%)

(b)
Fig
gure 9-13. Predicted
P dam
mage vs. meeasured fatiggue crackingg: (a) Pavemeent ME design
guide and (b) LVECD D program.

177
 25 80
10 cm Thickness (a) 16 cm Thickness (b)
70
20

Field Cracked Area (%)

Field Cracked Area (%) 60

50
15

40

10
30

20
5
10

0 0
5 10 15 20 25 30 5 15 25 35 45 55 65
LVECD Damaged Area (%) LVECD Damaged Area (%)
25 12
20 cm Thickness (c) 30 cm Thickness (d)

10
20
Field Cracked Area (%)

Field Cracked Area (%)

8
15

10
4

5
2

0 0
0 1 2 3 4 5 6 0 1 2 3 4
LVECD Damaged Area (%) LVECD Damaged Area (%)

Figure 9-14. Measured field cracking vs. LVECD program-predicted cracking at different
pavement thicknesses: (a) 10 cm, (b) 16 cm, (c) 20 cm, and (d) 30 cm.

178
 1000
10 cm
16 cm

y = 0.02x-2.47 20 cm
100 R² = 0.98 30 cm
y= 0.41x-5.64
(100/FC)-1 R² = 0.84
y = 0.09x-1.83
10
R² = 0.99
y = 1.45x-1.35
R² = 0.81

0
0.01 0.10 1.00 10.00
(2.5D/100)

Figure 9-15. Plot of correlation between ((100/FC)-1) and (2.5D/100) for different pavement
sections.

The rate at which cracks initiate and propagate in a pavement structure depends
mainly on the asphalt layer thickness. Therefore, the relationship that correlates fatigue
cracking to the damage in a pavement should include the effect of the rate of cracking that is
a function of the asphalt concrete layer thickness. To find the relationship between the C2 (or
the so-called  value) and the pavement thickness, the absolute values of  were plotted
versus the total section thickness, as shown in Figure 9-16. As shown, the exponential
function seems to be a suitable form to represent this relationship. However, it is still
necessary to modify the relationship between  and pavement thickness in a way that limits
the maximum and minimum values of  based on the pavement thickness.

179
 6

Absolute Value of 
y = 60.26x-1.74
4
R² = 0.91

0
0 2.5 5 7.5 10 12.5
Layer Thickness (in)

Figure 9-16. Variation of absolute values of  vs. pavement thickness.

Given all the aspects mentioned thus far, the most reasonable form that is able to
demonstrate the variations of  in terms of pavement thickness can be presented as Equation
(9.20).
     (1  hac ) (9.20)
After fitting the calculated data using Equation (9.21), the final equation is as follows:
  0.83  724(1  hac ) 3.103 (9.21)
where hac is the asphalt layer thickness.
Figure 9-17 demonstrates the plots of measured versus predicted cracking for
different pavement sections. The prediction capabilities presented in Figure 9-17 (R-square
of 0.81) are considered good for modeling the pavement fatigue cracking. The NCHRP
report 457 presents a synthesis of the AASHTO MEPDG implementation in the United
States, where R-square values above 0.65 are considered to have reasonable prediction
capabilities. The same report mentions that the fatigue cracking prediction capability of the
AASHTO MEPDG (United States national calibration) is poor, as it presents an R-square of 0.27.

180
 Figure 9-18 shows the comparison of the damage trends before and after fatigue cracking
calibration for the PRS pavement sections.
20 70
(a) (b)
4'' Pavement 6.5'' Pavement
60
Measured Cracking (%)

Measured Cracking (%)

15
50

40
10
30

20
5

10

0 0
0 5 10 15 20 0 10 20 30 40 50 60 70
LVECD Predicted Cracking (%) LVECD Predicted Cracking (%)
15 30
(c) 12'' Pavement (d)
8'' Pavement
25
12

Measured Cracking (%)

Measured Cracking (%)

20
9

15

6
10

3
5

0 0
0 3 6 9 12 15 0 5 10 15 20 25 30
LVECD Predicted Cracking (%) LVECD Predicted Cracking (%)
70
All Pavement Sections (e)
60
Measured Cracking (%)

50

40 y = 0.76x + 0.57
R² = 0.81
30

20

10

0
0 10 20 30 40 50 60 70
LVECD Predicted Cracking (%)

Figure 9-17. Comparison of observed versus predicted cracking with the LVECD transfer
function.

181
 30 20 20
(b) (c)
25
16 16

LVECD Crack Area (%)

Field Crack Area (%)

Damage Area (%)

20
12 12
15
8 8
10 Control
Control Control

Advera Advera Advera

4 4
5 Evotherm Evotherm Evotherm
Sasobit Sasobit Sasobit
0 0 0
1.0E+05 1.0E+06 1.0E+07 1.0E+05 1.0E+06 1.0E+05 1.0E+06
No. of Cycles No. of Cycles No. of Cycles

50 80 80
(d) (e) (f)
70 70
40

LVECD Crack Area (%)

Field Crack Area (%)

60 60
Damage Area (%)

30 50 50

40 40
C C C
20 30 30
R O O
O R R
20 20
10 FW RW RW
AW 10 FW 10 FW
RW AW AW
0 0 0
1.0E+06 1.0E+07 1.0E+08 1.E+07 1.E+08 1.E+07 1.E+08
No. fo Cycles No. of Cycles No. of Cycles

5 15 15
(g) (h) (i)

4 12 12
LVECD Crack Area (%)

Field Crack Area (%)

Damage Area (%)

3 9 9

2 6 6
C
C C
15R 15R
15R 3 3
1
50R 50R 50R
50RSB 50RSB 50RSB
0 0 0
1.E+05 1.E+06 1.E+07 1.E+05 1.E+06 1.E+05 1.E+06
No. of Cycles No. of Cycles No. of Cycles

Figure 9-18. (a) non-calibrated damage trends for MIT-WMA, (b) measured cracks for MIT-
WMA, (c) calibrated predicted damage for MIT-WMA, (d) non-calibrated damage trends for
NCAT, (e) measured cracks for NCAT, (f) calibrated predicted damage for NCAT, (g) non-
calibrated damage trends for MIT-RAP, (h) measured cracks for MIT-RAP, and (i) calibrated
predicted damage for MIT-RAP.

182
CHA
APTER 10 COM
MPARIS
SON OF F
FATIGU
UE CRAC
CKING
PER
RFORMA
ANCE PREDICT
P TIONS IN
N ASPHA
ALT PA
AVEMEN
NTS
USIING PAV
VEMENT
T ME AN
ND LVEC
CD PRO S6
OGRAMS

10.1 Abstractt
The perfo
ormance-baseed design off asphalt pavvements has been ongoinng since 2000.
Curreently availab
ble software for pavemen
nt performannce design inncludes the pprogram, thee
AASH
HTOware Pavement ME
E program th
hat was know
wn as Mechaanistic Empiirical Pavem
ment
Desig
gn Guide (M
MEPDG). The Pavement ME program
m allows useers to predictt pavement
distreesses by applying layered
d elastic theo
ory for the m
mechanical rresponses annd using
empirrical modelss for the distrress predictions. In addittion to the P
Pavement ME
E program, tthe
‘layerred viscoelastic pavemen
nt design forr critical disttresses’ (LV
VECD) progrram, which w
was
develloped recently at North Carolina
C Staate Universityy, can be useed to predictt the fatigue and
ruttin
ng performan
nce of pavem
ments using three-dimen
t sional viscoelastic finitee element
analy
ysis with mov
ving loads. The
T LVECD
D program em
mploys the w
well-known simplified
visco
oelastic contiinuum damaage (S-VECD
D) model as an effective tool for the fatigue
perfo
ormance pred
diction of asp
phalt mixturres under com
mplex loadinng and envirronmental
condiitions, in con
njunction wiith a newly developed
d ennergy-based fatigue failuure criterion.
This failure criterrion can pred
dict the fatig
gue life of assphalt mixturres independdent of loadiing
modee, frequency,, strain ampllitude, and teemperature.
y investigatees and compares the perfformance off 33 pavemennt sections frrom
This study
five different
d reseearch projects located in
n the United States, Canaada, and Souuth Korea using
both the Pavemen
nt ME and LVECD
L prog
grams. To veerify the resuults obtainedd from these two

6
This chapter was previously published as: Wang
g, Yizhuang., A
A. Norouzi, annd Y. R. Kim ((2015). Compaarison
of Fatigue Cracking Performance Predictions in Asphalt Pavvements Usingg Pavement ME
E and LVECD
Pro
ograms. Journaal of Transporttation Research
h Records, in ppress.

183
programs, the simulations were compared to the field performance data. In terms of ranking,
the LVECD simulations provided better agreement with the field performance data than the
Pavement ME simulations. For multilayer pavement sections, the LVECD program captured
the properties of all the layers for fatigue performance, whereas the Pavement ME program
considers only the bottom layer.

10.2 Introduction
It is now well accepted that fatigue cracking due to repeated traffic loading is one of
the most significant distresses found in the asphalt pavements. To determine the optimal
pavement design or management strategies, the origin of cracks that appear within the
pavement section needs to be identified. For achieving this goal, effective models and design
procedures that can reliably represent the fatigue damage growth are in demand to predict the
asphalt pavement performance (Nishizawa et al. 1997, Tayebali et al.1992, Norouzi et al.
2015).
Performance-based design is applied by most state highway agencies for pavement
construction, thereby replacing the empirical design procedures that were based on the
AASHTO test road data from the 1950s (Hveem 1955). Unlike empirical design,
performance-based design allows designers to see the predicted performance of a pavement
at the end of the design period and to select the most suitable pavement design accordingly
(NCHRP 2004). Designers can now predict the performance of a pavement using computer
programs such as AASHTOWare Pavement ME and the newer ‘layered viscoelastic
pavement design for critical distresses’ (LVECD) software.
The Pavement ME program was developed in 1996 based on the existing state-of-the
practice mechanistic-based models and became available in 2004 in the form of software
called the Mechanistic-Empirical Pavement Design Guide (MEPDG) (El-Basyouny and
Witczak 2004, Guo 2013).

184
 It applies layered elastic theory for the flexible pavements to calculate the required
stress and strain responses under traffic loading, and the responses are utilized in the
empirical models to link laboratory test results to field performance, including fatigue
damage and permanent deformation. However, it has been common to assume that fatigue
cracking normally initiates at the bottom of asphalt layer and propagates to the surface
(bottom-up cracking) due to the bending action of the pavement layer, which results in
flexural stresses to develop at the bottom of the bound layer (Myers and Rouqe 2001).
Consequently, top-down cracking cannot be explained by the traditional fatigue mechanisms
that are used to explain load-associated fatigue cracking that initiates at the bottom of the
pavement (Park et al. 2014). Furthermore, commonly used multi-layered elastic pavement
analysis is not capable of modeling the time dependent nature of asphalt concrete
appropriately (Park and Kim 2013).
The LVECD program was developed recently at North Carolina State University.
This program applies three-dimensional (3-D) finite element analysis with moving loads to
obtain the mechanical responses and uses mechanics-based models for the distress
predictions. For example, the simplified viscoelastic continuum damage (S-VECD) model is
employed for fatigue performance predictions (Guddati 2012, Guddati et al. 2013). The S-
VECD model is built on three concepts: (1) the elastic-viscoelastic correspondence principle
based on pseudo strain, (2) the continuum damage mechanics-based work potential theory,
and (3) the time-temperature superposition principle with growing damage to include the
combined effects of time/rate and temperature (Chehab et al. 2002, Park et al. 1996,
Underwood et al. 2012, Gibson et al. 2012). The model also includes a newly developed
energy-based failure criterion, referred to as GR, which is independent of temperature and
loading mode, to determine the failure of the mixture (Zhang et al. 2013, Sabouri and Kim
2014, Norouzi and Kim 2015). Therefore, LVECD does not require the assumption of crack
initiation a priori since the flexibility of S-VECD model allows cracks to initiate and
propagate wherever the fundamental law suggests, so that a realistic cracking condition can
be conducted through the simulation (Norouzi et al. 2014). As is the case with using the
Pavement ME program, designers who use the LVECD program are able to see the

185
performance of a pavement during the entire design period predicted from the mechanistic
models.
For this study, both the Pavement ME and LVECD programs were used to predict
fatigue cracking in 33 different pavement test sections from five different research projects.
The fatigue prediction results from the two programs are presented and compared in this
chapter. The chapter also includes descriptions of the tests conducted and the input
information needed for the two programs.

10.3 Background
Both the Pavement ME and the LVECD programs allow users to predict the long-
term performance of a pavement’s trial structure. Although similar input information is
required for both programs, the mechanisms of the two programs are very different. For
example, computational cost is an important factor for the long-term simulation of pavement
performance, and the two programs apply very different mechanisms to address the issue of
cost-effectiveness.
In the Pavement ME program, a multilayer elastic analysis program, JULEA, is used
to calculate the mechanical responses in order to reduce computational expenses. This elastic
analysis program is employed although the asphalt materials used in flexible pavements are
viscoelastic materials. To predict fatigue performance, the Pavement ME program applies the
Asphalt Institute Model, Equation (10.1), which is an empirical model that links the number
of cycles at failure and the applied strain. In this study, the fatigue properties of the different
mixtures used in the different pavement sections are taken into account through the material-
specific coefficients, kf1, kf2, and kf3. These material-specific coefficients were obtained from
cyclic direct tension fatigue tests performed on each mixture and the regression based on
further derivation (Equation (10.2)) from the S-VECD model.

186
 Using this equation, the number of cycles to failure under different loading strains
and temperatures can be calculated, and therefore, the exponential relationship between the
number of cycles to failure and the loading strain can be obtained through linear regression in
logarithmic scale. In this way, kf1, kf2, and kf3 for the different materials are computed.
Pavement ME program applies Miner’s law (Equation (10.3)) to accumulate the fatigue
damage. The calculated fatigue damage results are then converted to field performance data
by employing the transfer function presented in Equation (10.4). In this study, the local
calibration factors for North Carolina (Jadoun, 2011) is applied for all the pavement sections
regardless of the location of the projects.

 f 2k f 2  f 3k f 3
1 1
N f   f 1k f 1     (10.1)
 t  E
where
Nf = number of repetitions to fatigue cracking,
t = tensile strain at the critical location,
E = material stiffness,
k f 1, k f 2 , k f 3 = material-specific coefficients, and

 f 1 ,  f 2 ,  f 3 = local calibration factors.

  C12 1
 C12
  1 1 C12   1
2   C12  1 C11C12    0, ta 
2
   C  1 C   R
K1    C12 1

Nf   12 11
  0,Rta  



 (10.2)
   C12  1  C12 2 f R  2 
 
 

where
 R 0,ta = pseudo strain

fR = reduced frequency

K1 = loading shape function

187
 T
Ni
D
i 1
i 
N fi
(10.3)

where
D = damage,
T = total number of periods,
Ni = traffic for period i, and
Nfi = allowable failure repetitions under the conditions that prevail in period i.
 6000  1 
Alligator Cracking   ( C1 .C '1  C2 .C '2 .log(% Damage )  .  (10.4)
 1 e   60 
where
C1 '  2C2 '

C 2 '  2.40874  39.748 1  hac 

2.856

The LVECD program computes the amount of pavement distress from the beginning
to the end of the design period. In terms of the calculation of the mechanical responses, 3-D
finite-element viscoelastic analysis with moving loads is performed by the LVECD program.
In this analysis, fast Fourier transform is conducted twice to reduce the computational cost.
In terms of the computation of the fatigue damage, the S-VECD model (presented in
Equations (10.5) and (10.6)) is applied. Equations (10.5) and (10.6) characterize the damage
evolution in the asphalt mixtures where damage is a combination of microcrack initiation and
propagation. Equation (10.7) is the energy-based failure criterion (GR), which has been
proven to be able to determine the failure of mixtures; the GR versus Nf (number of cycles to
failure) curves among different types of mixtures (i.e., modified binder mixtures, high RAP
mixtures, and WMA mixtures) differ significantly. The S-VECD model is able to describe
the comprehensive constitutive fatigue behavior of an asphalt mixture. Each asphalt layer in
the cross-section of the pavement is divided into 1000 (100×10) elements with 11,111
(101×11) nodes, and the elements and nodes are used in the finite elements analysis for the
mechanical response and damage calculations. The whole design period also is divided into a
large number of life-stages (i.e., 240 life-stages for 20 years). At the end of each life-stage,

188
the damage state of the mixtures is updated. During the simulation, the LVECD program
applies Miner’s law, Equation (10.3), in order to accumulate the fatigue damage that is due to
different loading conditions. The performance index used in the LVECD program is the so-
called damage area (%), which is the number of nodes where the asphalt mixture fails
( 1) over the total number of nodes throughout all the asphalt layers.

Note that, even though the transfer function, which converts the ‘damage area’ into
the scale of the ‘cracking area’ in the field, has not been fully developed and thus has not yet
been applied in the LVECD program, the predicted results nonetheless match the field
measurements quite well, and the ranking of the fatigue performance of the pavement
sections is not affected by the inclusion of a transfer function. Therefore, with the lack of the
application of transfer function in LVECD and the local calibration factors for each section in
Pavement ME, this chapter focuses only on the ranking instead of the absolute values of the
prediction results. Furthermore, at the end of the computation process in the LVECD
program, the damage contours can be plotted to show the state of the fatigue damage of the
pavement and cracking location at any time in the design period; for example, see
Figure 10-3.

S  W R 
Damage Function:    (10.5)
t  S 

Damage Curve: 1  C11  S C 12

 (10.6)

Failure Criteria: G R   N f  (10.7)

where
S = damage parameter,
C = pseudo stiffness,
WR = work potential function,
C11, C12,, ,  = model parameters,
GR = energy release rate.

189
10.4 Data Collection and Field Sections
Experiments were performed and data were collected based on the input requirements
of the two programs. Several sections of input data are incorporated in both programs; these
data include general information (i.e., design period), traffic data, climate data, structural data
(i.e., layer thickness, flexible material properties, and unbound layer material properties), and
other information (i.e., local calibration factors). The traffic information in the Pavement ME
program is comprised of basic information (i.e., average annual daily truck traffic (AADTT),
vehicle speed), traffic volume adjustment (i.e., vehicle class distribution, hourly truck traffic,
and traffic growth factors), axle load distribution factors, and general traffic input (i.e., mean
wheel location, wander standard deviation). In the LVECD program, all the traffic data are
converted into the number of passes of a standard axle (two single tires with 80-kN axle load)
with constant speed (27 m/s) based on the design equivalent single-axle loads (ESALs). For
the climate data, both the Pavement ME and LVECD programs employ the Enhanced
Integrated Climate Model (EICM) database. Using this database, temperatures can be
extracted that are applicable throughout the pavement depth and design life. In addition, the
Pavement ME software extracts other information from the database, including moisture
profiles, frost depth, annual freezing index values, and mean annual number of wet days.
The structural data, including material properties, are the most important information
in the performance-based design process. For this study, the material properties of the asphalt
mixtures were obtained by performing dynamic modulus and fatigue tests, the results of
which are needed to run both Pavement ME Level I and the LVECD program. For both
programs, dynamic modulus tests were conducted, because material stiffness is important to
the simulations. In the Pavement ME program, the modulus values of the asphalt mixtures at
specific temperatures and frequencies are utilized directly by the program as the elastic
modulus of the mixtures under different conditions because no historical loading effects are
taken into account. However, in the LVECD program, the dynamic modulus test results were
analyzed based on the time-temperature superposition (Equation (10.8) and (10.9)).

190
 Coefficients of the Prony series (in Equation (10.10)) were calculated based on the
test results and input into the LVECD program in order to apply viscoelastic theory. In terms
of the inputs for the fatigue model, for the LVECD program, direct tension fatigue tests were
conducted based on AASHTO TP 107, and the required parameters of the S-VECD model
were computed and used in the program. For the Pavement ME program, the material-
specific fatigue coefficients kf1, kf2, and kf3 in Equation (10.1) were obtained from the same
tests. However, unlike the LVECD program, the Pavement ME program considers the fatigue
coefficients only for the bottom layer asphalt material. In addition to the mechanical
properties of the mixtures, the volumetric properties of the mixtures and shear modulus
values of the binders in the mixtures were input to the Pavement ME program. The material
properties of the unbound layers, i.e., the aggregate base and subgrade soil, also are needed.
Both programs take the mechanical properties of the unbound layer material, i.e., the
modulus values and Poisson’s ratio, into account. In addition to the mechanical properties,
other properties, such as soil type and gradation, also were input to the Pavement ME
software.

b
log | E* | a  (10.8)
1
1
ed  glog f R 
log  aT   1T 2   2T  3 (10.9)
m
E (t )  E   Ei .e  t / i (10.10)
i 1

where
fR = reduced frequency,

e = exponential function,
a , b, d , g = fitting coefficients of sigmoidal function,

aT = time-temperature shift factor,

1 ,  2 , 3 = fitting coefficients of time-temperature shift factor function,

191
 T = temperature,
E = long-term equilibrium modulus,
i = index of summation,
m = number of Prony terms,
Ei = modulus of Prony term number i,
t = time, and
i = relaxation time of Prony term i.

The performance prediction results from the two programs, Pavement ME and
LVECD, were compared with the field performance measurements of the different pavement
sections. Thirty-three pavement sections from five different research projects located in the
United States, Canada, and South Korea were used, and 38 different asphalt mixtures were
tested. The sections used in this study are from the National Center for Asphalt Technology
(NCAT) Test Track, the Federal Highway Administration Accelerated Loading Facility
(FHWA-ALF), the Manitoba Infrastructure and Transportation warm mix asphalt (MIT-
WMA) project, the MIT reclaimed asphalt pavement (MIT-RAP) project, and the Korean
Expressway Corporation (KEC) Test Road.
Six sections were selected from the NCAT Test Track: S9, S8, S10, S11, N10, and
N11. For this study, these section designations were abbreviated to C (control), O (open-
graded friction course surface with control intermediate/base), FW (control mixture with
foamed asphalt WMA), AW (control mixture with Advera additive WMA), R (50 percent
RAP mixture), and RW (50 percent RAP mixture with foamed asphalt WMA), respectively.
The structures of the NCAT pavement sections are all three-layer systems, and the
schematics are presented in Figure 10-1 (a).

192
 From the FHWA-ALF project, Lane 2, Lane 4, Lane 5, and Lane 6 were selected;
similarly to the NCAT sections, the ALF sections were renamed as control, SBS-LG (styrene
butadiene styrene with linear grafting), CR-TB (crumb rubber terminal blend), and
terpolymer (ethylene terpolymer), respectively, based on the materials used in the sections.
The structures of the ALF sections are presented in Figure 10-1 (b). For the MIT-WMA
project sections, three different types of WMA mixtures were compared to a hot mix asphalt
(HMA) control mixture. For the MIT-RAP project sections, four mixtures with different
contents of RAP were compared: control (zero percent RAP), 15%-R (fifteen percent RAP),
50%-R (fifty percent RAP), and 50%-RSB (fifty percent RAP with a soft binder). The
structural schematics for the MIT-WMA and MIT-RAP sections are shown in Figure 10-1 (c)
and (d), respectively. The KEC test road was built to evaluate the effects of the type of base
layer material, base layer thickness, an anti-frost layer, and the sub-base layer thickness. The
structural layout of the KEC project is presented in Figure 10-1 (e).
Among the selected projects, the NCAT and FHWA sections are full-scale
accelerated test sections, whereas the MIT and KEC projects are test roads. For the subgrade
inputs for the MIT and KEC sections, the soil properties were obtained from field core falling
weight deflectometer (FWD) backcalculations. Because the MIT and KEC projects are not in
the United States, their actual climate stations were not available in the EICM database.
Grand Forks, ND was selected to represent Manitoba, Canada, and Washington D.C. was
selected to represent South Korea, because the climate data are very similar to the locations
of the test roads in these two regions. The input traffic data are based on actual traffic
measurements. To determine the mixture properties, dynamic modulus tests and fatigue tests
were conducted for each mixture used in this study. For the Pavement ME program, only the
fatigue coefficients of the bottom layer material were input to predict the fatigue
performance. The loads applied in the full-scale test sections, i.e., the NCAT and ALF
sections, were converted into Pavement ME and LVECD program inputs. Details regarding
the traffic inputs, as well as other information, are presented in Table 10-1.

193
 C O FW AW R RW Terpolymer CR-TB SBS-LG Control
0 0
3.1 3.3 3.3 3.8 3.6 3
2.5
2.5
5
7.1 6.9 6.9 7.6
7.5 7.6 7.1
Depth (cm)

Depth (cm)
5 10 10 10 10

10

7.5
12.5
7.6 6.6 7.6 6.6 7.6 7.4
15
10
17.5
(a) (b)
20 12.5

Control Sasobit Advera Evotherm Control 15%-R 50%-R 50%-RSB

0 0

2.5 5 5 5 5 5 10 10 10 10

Depth (cm)
Depth (cm)

5 10

7.5 5 5 5 5 15 10 10 10 10

10 20

(c) (d)
12.5 25

A1 A2 A3 A4 A5 A6 A7 A8 A9 A10 A11 A12 A13 A14 A15

0
5
ASTM
10 7 Mixture
8 8 8 BB5
20 18 18 18 Mixture
28 28 28
30 BB3
Mixture
30 30 30 30
40 40 BB1
Depth (cm)

30 30 30 30 Mixture
50 40 Agg
30 30 30 30 Base
60 40
Sub
Base
70 40 40 40
30 30 30 Anti
80 20 20 20 Frost

Sub-
90 Grade
(e)
100

Figure 10-1. Schematics of asphalt pavement structures for the test sections: (a) NCAT
(asphalt layers), (b) FHWA-ALF (asphalt layers), (c) MIT-WMA (asphalt layers), (d) MIT-
RAP (asphalt layers), and (e) KEC (whole pavement structure).

194
Table 10-1. Input Information for LVECD and Pavement ME Programs for Performance Analysis of NCAT and ALF Test Sections
LVECD Pavement ME

NCAT ALF NCAT ALF

AADTT 3082 Tire Load 16000 lb.

Basic
Information
Vehicle Speed 45 mph Tire Pressure 120 psi
Standard Deviation of
Vehicle Class Distribution 50% Class 12+ 50% Class 13 5.25 in.
Wheel Wander
As distributed in the field based
AADT 14083 835
on shift changes, truck refueling,
Traffic Hourly Truck Traffic Monthly Repetitions 5208
driver breaks, and maintenance
Volume
stops
Adjustment

Traffic Monthly Adjustment Factors 1 for all months Annual Growth 0

Single
Traffic Growth Factors No growth Tire Location
Tire
Axle Load
single axle + tandem axle
Distribution
+ 5 single axles
Factors

No width: 8.5 ft.

Growth No Growth Axle Configuration dual tire space: 13.5 in.
Growth
General tire pressure: 100 psi
Traffic Input
Mean Wheel Location Default value
Wander Standard Deviation Default value
Longitude -86.394
EICM Station:
Isotherma
Climate Montgomery Latitude 32.301 19 °C
l 19 °C
AL
Station Montgomery, Alabama
Dynamic Prony series coefficients for Dynamic
Tested results for all the mixtures, 4 temperatures and 6 frequencies
Modulus all asphalt mixtures Modulus
Asphalt
S-VECD Model
Material Fatigue Fatigue
Coefficients for all asphalt Regressed material-specific coefficients k1, k2, and k3 for only bottom layer mixture
Properties Properties
mixtures

Mechanical
Resilient Modulus or Young's Modulus; Poisson's Ratio
Unbound Properties
Layers Other
Soil Type, Gradation
Information

195
10.5 Results and Discussion
After the data were collected, the Pavement ME and LVECD programs were used to
calculate the performance prediction results for each pavement section. The prediction results
were compared between the two programs, and the predicted results were compared with the
measured ‘cracking areas’ in the field.
Comparisons between the prediction results and the field measurements of the NCAT
test track sections are presented in Figure 10-2 (a) – (f). According to the field
measurements, the open-graded friction course (OGFC) section and the foamed WMA
section exhibited the worst fatigue cracking performance (i.e., the least resistance to
cracking), and the ‘cracking areas’ of the control section and the two sections with RAP
material are seen to be relatively small. The LVECD program was able to capture the ranking
of the ‘cracking areas’, according to Figure 10-2 (d), and (e); however, the Pavement ME
program could not, according to Figure 10-2 (f) and (c). In the Pavement ME predictions, the
OGFC section has the same damage level as the control section, which is not the case in the
field. This outcome is due to the fact that the Pavement ME program considers only the
bottom asphalt layer to be critical for fatigue cracking, so only the fatigue cracking
coefficients of the bottom layer mixture were required by the program. However, the
intermediate and bottom layer mixtures in the OGFC section are the same materials as those
used in the control section. The only difference between the two sections is the application of
the OGFC mixture in the surface layer; however, because the OGFC is open-graded and has
a high air void content, cracking is more likely to occur using this mixture. By contrast, the
LVECD program accounts for the whole cross-section of the pavement, and so, damage is
computed throughout the pavement. In addition, the damage contours (presented in
Figure 10-3) also are provided by the LVECD program. From the damage contours of the
cross-section, top-down cracking is clearly critical in the OGFC section – not just bottom-up
cracking.

196
 In short, all the mixtures in a pavement must be taken into account, not just the
bottom layer material. Also, in this case, the damage contours can help the designer to
consider explicitly the ‘cracking area’ and severity of damage.
Figure 10-2 (g) – (l) presents the comparison between the predicted cracking damage
and the measured ‘cracking area’ in the field for the FHWA ALF sections. According to
Figure 10-2 (g) and (j), which present the cracks measured in the field, the control section
shows the worst fatigue performance (resistance to fatigue cracking), and the SBS-modified
section shows the best performance. The Pavement ME program generally was able to
predict the ranking of the ‘% crack area’ in the field, except that the rankings of the
terpolymer section and the SBS section are reversed. This ranking was better captured by the
LVECD program, as the shape of the damage area versus ESAL curve is very similar to the
curve in the plots for the field measurements when comparing Figure 10-2 (g) and (h) on a
semi-log scale.

197
 NCAT Performance Predictions and Field Measurements
60 50 8
(a) Field Measurement (b) LVECD Prediction (c) Pavement ME Prediction
50

LVECD Damage Area (%)

Alligator Cracking (%)

Field Cracked Area (%)
40
6
40
30
30 4
O
O C
20 FW
20 FW R
AW
AW O 2
C 10 FW C
10
RW AW RW
R RW R
0 0 0
1.E+07 1.E+08 1.0E+06 1.0E+07 1.0E+08 0 10 20 30
No. of Cycles No. fo Cycles Months

80 80 8
(d) Field Measurement (e) LVECD Prediction (f) Pavement ME Prediction
70 70

LVECD Damage Area (%)

Field Cracked Area (%)

Alligator Cracking (%)

60 60 6

50 50

40 40 4

30 30

20 20 2

10 10
0 0 0
O FW AW C RW R O FW AW C RW R O FW AW C RW R
Sections Sections Sections

FHWA Performance Predictions and Field Measurements

100 15 5
(g) Field Measurement (h) LVECD Prediction (i) Pavement ME Prediction
LVECD Damage Area (%)
Field Cracked Area (%)

80 4

Alligator Cracking (%)

10
60 3

Control
40 2
5 CR-TB
Control
Control Terpolymer
CR-TB CR-TB SBS-LG
20 1
Terpolymer SBS-LG
SBS-LG Terpolymer
0 0 0
1.E+04 1.E+05 1.E+04 1.E+05 1.E+06 1.E+07 0 20 40 60
No. of Cycles No. of Cycles Months

100 70 5
(j) Field Measurement (k) LVECD Prediction (l) Pavement ME Prediction
60
LVECD Damage Area (%)
Field Cracked Area (%)

80 4
Alligator Cracking (%)

50
60 40 3

30 2
40
20
20 1
10

0 0 0
Control CR-TB Terpolymer SBS-LG Control CR-TB Terpolymer SBS-LG Control CR-TB Terpolymer SBS-LG
Sections Sections Sections

Figure 10-2. Measured and predicted fatigue cracking amount for NCAT and FHWA test
sections: (a) amount of cracking in the field (NCAT), (b) predicted damage area in LVECD
(NCAT), (c) predicted damage area in Pavement ME (NCAT), (d) maximum amount of
cracking at the end period of study (NCAT), (e) maximum predicted damage area at the end
of analysis with LVECD (NCAT),(f) maximum predicted damage area at the end of analysis
with Pavement ME (NCAT), (g) amount of cracking in the field (FHWA), (h) predicted
damage area in LVECD (FHWA), (i) predicted damage area in Pavement ME (FHWA), (j)
maximum amount of cracking at the end period of study (FHWA), (k) maximum predicted
damage area at the end of analysis with LVECD (FHWA), and (l) maximum predicted
damage area at the end of analysis with Pavement ME (FHWA).

198
 0 1

2 0.9

0.8
4

0.7
6
0.6
8
Z (cm)

0.5

10
0.4

12 0.3

14 0.2

16 0.1

0
-1.5 -1 -0.5 0 0.5 1 1.5
X (m)

Figure 10-3. Damage factor (N/Nf) distribution of NCAT Test Track OGFC section at the
end of analysis period.

The predicted and field-measured fatigue damage results for the MIT-WMA project
are presented in Figure 10-4. In this case, while the rankings of the prediction results from
the Pavement ME program matched the measured results in the field, the rankings of the
Control and Advera sections were flipped in the LVECD program results. This outcome
could be due to the properties of the WMA mixtures. Previous studies have shown that,
under the same aging conditions, the stiffness of WMA mixtures increases more than that of
HMA mixtures, and in the LVECD program, the aging model has not been developed yet.
However, the aging effect of the WMA mixtures can be taken into account in the Pavement
ME program simulations using the Global Aging System (GAS) model.

199
 The predicted and field-measured performance results for the MIT RAP project
sections are presented in Figure 10-4 (a) – (f). According to Figure 10-4 (g), the sections with
materials with high RAP contents have significantly higher cracking damage values than
those with low RAP contents; however, the section paved with 15% RAP mixture does not
differ significantly from the control section. This finding confirms that it is reasonable to
allow about 20 percent RAP in asphalt mixtures, which most state highway agencies suggest.
The difference in performance between the high RAP mixtures and the low RAP mixtures is
captured in the LVECD program predictions. However, the Pavement ME program
prediction results indicate no significant difference among all the MIT RAP sections. Again,
this outcome is due to the fact that the Pavement ME program considers the fatigue
properties of only the bottom layer material, and all four sections have the same bottom layer
material. It is worth noting that though not much difference can be observed in the MIT RAP
sections based on the alligator crack curves predicted by the Pavement ME program, the four
sections have slightly different absolute crack percentage values at the end of the design
period. It can be observed that the ranking of the performance prediction results of the four
sections is the reverse ranking of the modulus values of the surface mixtures (which are 8694
MPa, 8680 MPa, 6801 MPa, and 6283 MPa for 50RSB, 50R, 15R, and Control mixtures
respectively at 20 and 10 Hz). Therefore, with identical base mixtures, the Pavement ME
program’s fatigue cracking predictions were affected strongly by the modulus values of the
mixtures rather than by their fatigue properties.
In brief, based on the prediction results and field measurements of the MIT RAP
section, the effects of the different RAP contents in this project can be characterized by the
LVECD program; in contrast, the Pavement ME program did not produce good results due to
the limited input information and mechanism of fatigue damage calculation.

200
 The KEC test road was constructed to evaluate ways that different factors affect
pavement. The factors used in this study include the base layer material, the thickness of the
base layer, the application of an anti-frost layer, and the thickness of the sub-base layer. The
evaluation results are presented in Figure 10-4. The effects of the base layer material are
presented in Figure 10-5 (a), (b), and (c). According to Figure 10-5 (a), aggregate base
pavements have a significantly higher cracking damage percentage than full-depth pavements
when the pavements have the same thickness of base layer. This effect was captured by both
the Pavement ME and LVECD programs. The effects of the thickness of the base layer are
shown in Figure 10-5 (d) to (f) where, for a given base layer material, a thicker base layer
results in less fatigue damage. This trend is matched in both programs, and the prediction
results of the two programs are also very similar to each other in this case. In terms of the
effect of an anti-frost layer and the effect of the thickness of the sub-base layer, according to
Figure 10-5 (g) to (i) and (j) to (l), respectively, the effects are not significant based on the
measurements in the field and the prediction results from the Pavement ME and LVECD
programs. In addition, according to Figure 10-5 (a) and (d), when the aggregate base is very
thin (8 cm), the crack damage percentage of the pavement section is very high; however, this
trend was not predicted by the Pavement ME and LVECD programs due to the application of
linear theory to the aggregate base. When the base layer is very thin, the strain and
deformation in the base layer are relatively high and exceed the linear elastic domain. When
the base layer response is in the domain of nonlinear elasticity, the crack damage propagation
is accelerated throughout the whole pavement. In short, based on the predictions and field
results, these changes in the base and sub-base layers can be predicted by both the LVECD
and Pavement ME programs. To improve the prediction programs, nonlinear elastic theory
should be applied for the aggregate base layers when the bases are thin or the strain level in
the base is relatively high.

201
 MIT-WMA Performance Predictions and Field Measurements
15 30 3
(a) Field Measurement (b) LVECD Prediction (c) Pavement ME Prediction
Sasobit
25

LVECD Damage Area (%)

Alligator Cracking (%)

Field Cracked Area (%)
Control

10 20 2 Advera
Evotherm
15

5 Sasobit
10 Control
1
Control
Advera
Advera 5 Evotherm
Evotherm Sasobit
0 0
0
1.E+05 1.E+06
1.0E+05 1.0E+06 1.0E+07 0 50 100 150 200
No. of Cycles
No. of Cycles Months
20 3
20 (e) LVECD Prediction (f) Pavement ME Prediction
(d) Field Measurement

LVECD Damage Area (%)

Alligator Cracking (%)

Field Cracked Area (%)

15 15
2

10 10

1
5 5

0 0 0
Sasobit Control Adverva Evotherm Sasobit Control Adverva Evotherm Sasobit Control Adverva Evotherm
Sections Sections Sections

MIT-RAP Performance Predictions and Field Measurements

20 5 4
(g) Field Measurement (h) (i) Pavement ME Prediction
Field Cracked Area (%)

Alligator Cracking (%)

15 3
Damage Area (%)

3
10 2

50RSB 2
50RSB
50R C
5 1 50R
15R
15R 1 15R
50R
Control Control
50RSB
0 0 0
1.E+05 1.E+06 1.E+05 1.E+06 1.E+07 0 50 100 150 200
No. of Cycles No. of Cycles Months

20 20 4
(j) Field Measurement (k) LVECD Prediction (l) Pavement ME Prediction
LVECD Damage Area (%)
Field Cracked Area (%)

Alligator Cracking (%)

15 15 3

10 10 2

5 5 1

0 0 0
50RSB 50R 15R Control 50RSB 50R 15R Control 50RSB 50R 15R Control
Sections Sections Sections

Figure 10-4. Measured and predicted fatigue cracking amount for MIT-WMA and MIT-RAP
test sections: (a) amount of cracking in the field (MIT-W), (b) predicted damage area in
LVECD (MIT-W), (c) predicted damage area in Pavement ME (MIT-W), (d) maximum
amount of cracking at the end period of study (MIT-W), (e) maximum predicted damage area
at the end of analysis with LVECD (MIT-W),(f) maximum predicted damage area at the end
of analysis with Pavement ME (MIT-W), (g) amount of cracking in the field (MIT-R), (h)
predicted damage area in LVECD (MIT-R), (i) predicted damage area in Pavement ME
(MIT-R), (j) maximum amount of cracking at the end period of study (MIT-R), (k) maximum
predicted damage area at the end of analysis with LVECD (MIT-R), and (l) maximum
predicted damage area at the end of analysis with Pavement ME (MIT-R).

202
 40 40 100
Field Measurement (a) LVECD Prediction (b) Pavement ME Prediction (c)

Alligator Cracking (%)

BB1 BB1
80

Damaged Area (%)

BB1

Cracked Area (%)

30 BB3 30 BB3
A13 Agg
BB3
Agg
60 Agg
20 20
40
10 A10 A1 A14 10
A11 A4 A15 20
A12 A7

0 0 0
8 cm 18 cm 28 cm 8 cm 18 cm 28 cm 8 cm 18 cm 28 cm
Base Layer Thickness Base Layer Thickness Base Layer Thickness
40 40 100
Field Measurement (d) LVECD Prediction (e) Pavement ME Prediction (f)
Base Thickness

Alligator cracking (%)

8 cm Base 8 cm Base 80

Damaged Area(%)
30 8 cm Base
Cracked Area (%)

30 18 cm Base 18 cm Base 18 cm Base

A13 28 cm Base
28 cm Base
60 28 cm Base
20 20
40

10 10
A1 A14 20
A10 A11 A4 A15
A12 A7
0 0
0
BB1 BB3 Aggregate BB1 BB3 Aggregate
BB1 BB3 Agg
Base Layer Material Base Layer Material Base Layer Material
10 100 100
Field Measurement (g) LVECD Prediction (h) Pavement ME Prediction (i)

Alligator Cracking (%)

8 w/o Anti-frost 80 80 w/o Anti-frost
Damaged Area (%)

w/o Anti-frost
Cracked Area (%)

w/ Anti-frost w/ Anti-frost w/ Anti-frost

6 A2 60 60

4 40 40
A5
A4 A8
2 20 20
A1
A7
No Data
0 0 0
8 cm 18 cm 28 cm 8 cm 18 cm 28 cm 8 cm 18 cm 28 cm
Base Layer Thickness Base Layer Thickness Base Layer Thickness
10 100 100
Field Measurement (j) LVECD Prediction (k) Pavement ME Prediction (l)

Alligator Cracking (%)

8 30 cm Subbase 80 30 cm Subbase 80 30 cm Subbase
Cracked Area (%)

Damage Area (%)

40 cm Subbase 40 cm Subbase 40 cm Subbase

6 60 60

4 40 40

2 20 20

0 0 0
8 cm 18 cm 28 cm 8 cm 18 cm 28 cm 8 cm 18 cm 28 cm
Base Layer Thickness Base Layer Thickness Base Layer Thickness

Figure 10-5. Effects of different parameters of fatigue cracking performance for KEC test
sections: (a)-(c) base layer material, (d)-(f) base layer thickness, (h)-(j) application of anti-
frost layer, and (k)-(m) sub-base thickness.

The overall trends of the prediction results from the LVECD and Pavement ME
programs are plotted against the field measurements in Figure 10-6; these results include
those of the NCAT, FHWA ALF, and MIT WMA and RAP projects. In terms of absolute
values of the prediction results and the field measurements, due to the lack of the application
of the transfer function and local calibration factors, the predicted damage values are not on
the same scale as the field measurements.

203
 However, the LVECD program results agree well with the results measured in the
field. In contrast, when the Pavement ME program prediction results are plotted against the
field measurements, the data are scattered and do not show good agreement, as seen in
Figure 10-6 (b).

70 8
(a) MIT-RAP (b) MIT-RAP

% Crack (Pavement ME)

% Damage (LVECD)

60 MIT-WMA 7
MIT-WMA
50 NCAT 6 NCAT
FHWA-ALF FHWA-ALF
5
40
4
30
3
20
2
10
1
0 0
0 20 40 60 80 100 0 20 40 60 80 100
% Crack (Field Measurement) % Crack (Field Measurement)

Figure 10-6. Overall trends of predicted fatigue cracking vs. field measurements: (a) LVECD
program predictions and (b) Pavement ME program predictions.

10.6 Conclusions
This study presents the fatigue performance prediction results for selected field test
sections using two programs, the LVECD and Pavement ME, and compares the prediction
results with the field measurements. Thirty-three pavement test sections from five research
projects were used in this study. The two programs apply different mechanisms to compute
the fatigue damage in the pavement sections. The following conclusions can be drawn based
on the comparison of the prediction results to the ‘cracking areas’ measured in the field:
 To obtain the mechanical responses of the pavements, the LVECD program employed
3-D finite element analysis with moving loads, and the Pavement ME program
employed multilayer elastic analysis. To predict fatigue cracking, the S-VECD model,
a constitutive damage model, was adopted in the LVECD program. The prediction
results were affected more by the stiffness of the asphalt mixtures in the Pavement
ME.

204
 Regarding
g the predicttion results, in
i terms of rranking, the LVECD proogram providded
better agrreement with
h the field meeasurementss than the Paavement ME
E program.
 For the multilayer
m pav
vement sectiions, the prooperties of thhe surface annd intermediaate
layer mateerials, not ju
ust the bottom
m layer mateerials, are takken into acccount in the
LVECD program.
p
 The damaage contourss of the crosss-sections prrovided by thhe LVECD pprogram cann be
useful in the
t design prrocess.
 The analy
yses perform
med in both programs
p didd not capturee the effects of the propeerties
of the agg
gregate base well. Nonlin
near elastic aanalysis willl be considerred for incluusion
in the LV
VECD prograam in the futu
ure.
 Although the LVECD
D program fo
ocuses on thee analysis off asphalt layers, the
Pavementt ME program
m considers more inform
mation in terrms of trafficc and climatte.
Similar feeatures will be
b added to the
t LVECD program in the future.

205
CHA
APTER 11 CON
NCLUSIO
ONS AN
ND FUTU
URE WO
ORK

11.1 Conclusiions
This reseaarch is divid
ded into five main steps: (i) statisticaal analysis off effective
param
meters in dyn
namic modu
ulus testing in
n indirect tennsion mode and developpment of testting
speciification for the
t field corres, (ii) geom
metry and gauuge length sstudy for the cyclic direcct
tensio
on testing to obtain the specimen
s faiilure in the m
middle, (iii) iimplementattion of the
simpllified viscoeelastic contin ge (S-VECD)) model couppled with thhe GR failure
nuum damag
criterrion for asph
halt mixture characterizat
c tion, (iv) vallidation of thhe LVECD ddesign tool ffor
pavem
ment analysiis based on field
f perform
mance observvations, and (v) definingg simulated
damaaged area-to--field crackeed area transffer function for the PRS pavement teest sections.
A summaary of the fin
ndings and th
he conclusionns drawn aree as follows::
 According
g to the rugg
gedness anallysis of dynaamic moduluus in indirectt tension moode,
air void content was found
f to be statistically
s significant, especially att -10C and 10C.
However,, the impact of air void content
c on m
material stiffnness (i.e., dyynamic moduulus
values) att 35°C was statistically
s in
nsignificant .
 In generall, specimen thickness,
t gaauge length,, strain level, and test tem
mperature, ddid
not have a considerab n the dynamiic modulus vvalues.
ble impact on
 o  1.0C seeems adequaate because tthe change iin
The temperature tolerrance range of
dynamic modulus
m vallue was with
hin the accepptable range; however, too maintain
consistenccy with otheer specifications, the toleerance rangee of  0.5C ( 1F) is
nded for the IDT specifiication.
recommen
 No signifi ved between the stiffnesss values at 40  and 60 .
ficant changee was observ
With the consideratio
c on of signal quality
q and ddamage to thhe sample, thhe horizontall
strain leveel should be limited to 50  5 .
 olerance of  0.5% shoulld be reasonnable becausee the dynam
An air void content to mic
modulus variation
v waas less than 10%
1 for almoost all the caases investiggated.

206
 Ruggedness analysis showed no major difference between the stiffness values for 38
mm and 50 mm thickness; therefore, the thickness of 38  3 mm is recommended for
IDT samples, which is practical for obtaining field cores from pavement sections.
 The gauge length limit must be an appropriate representative volume element (RVE),
especially for mixtures that contain large aggregate particles; so, the gauge length of
101.6  3 mm is recommended for IDT testing.
 Specimen geometry and gauge length (100-mm diameter, 150-mm height, and 70-mm
gauge length) specified in AASHTO TP 79 and PP 61 yield dynamic modulus values
that are in reasonable agreement with the results from other specimen geometries and
meet conventional RVE requirements.
 Regarding direct tension cyclic fatigue testing, the specimen geometry of 100 mm in
diameter and 130 mm in height leads to statistically the same damage characteristic
curves as the other geometries investigated and increases the propensity for middle
failure.
 SBS, as presented in the FHWA-ALF project, demonstrates the best performance for
both fatigue cracking and rutting since it decreases the mixture modulus at low and
intermediate temperatures, and increases the stiffness at high temperatures. Other
modified binders, crumb rubber, and terpolymer, also improve the pavement life in
terms of fatigue cracking.
 Warm mix additives seem to deteriorate the mixtures’ fatigue resistance, although
they decrease the stiffness and aging during compaction. Evotherm improves the
fatigue behavior in MIT; this is while it increases the amount of cracking in NCAT
test track project.

207
 It is accepted that the addition of RAP up to 20% does not have any major effect on
virgin mixture performance, according to many asphalt agencies. The results of MIT-
RAP project briefly admit this conclusion where utilization of RAP up to 15% did not
change the mixture performance. The mixtures containing 50% RAP show lower
fatigue life compared with the others.
 The pavement simulations indicate no significant change in the fatigue resistance or
permanent deformation when using an anti-frost layer or increasing the sub-base layer
thickness.
 One interesting observation is that the base layer material played the most important
role in affecting both rutting and fatigue cracking in the KEC test road sections. The
aggregate base layer exhibited greater rut depths and more cracking than the asphalt
base layers, whereas the sections with an asphalt base layer (either the BB1 mix or
BB3 mix) showed less distress than the aggregate base sections.
 The data obtained from the pavement analyses of the different base layer thicknesses
briefly indicate that an increase in the base layer thickness reduces the pavement
distresses, but the amount of improvement is dependent on the material type.
 For the multilayer pavement sections, the properties of the surface and intermediate
layer materials, not just the bottom layer materials, are taken into account in the
LVECD program.
 Regarding the prediction results, in terms of ranking, the LVECD program provided
better agreement with the field measurements than the Pavement ME program.
 The S-VECD model coupled with the GR failure criterion can be used for the accurate
fatigue life predictions of asphalt mixtures, even when very different asphalt binders
(unmodified, polymer-modified, crumb rubber-modified, WMA, RAP etc.) are
analyzed since there is a strong agreement between the LVECD simulations and field
observations.

208
  From both LVECD and Pavement ME observations, a sigmoidal function was chosen
for representing the relation between field cracked area and simulated damaged area.
 The proposed damage-to-field cracked area transfer functions were validated for both
the asphalt pavement new asphalt pavement projects. However, it should be pointed
out that the proposed framework has not been validated for the overlays and the
pavements with reflective cracking.

11.2 Future Work

The following topics are suggested for future work:

 Verification of the dynamic modulus and damage properties of small specimens since
these samples are more compatible with the field cores and they can be easily
fabricated in the asphalt labs.
 Development of fatigue failure criterion for the binder and finding the relationship
between binder and mixture fatigue life to reduce the amount of testing time in the
lab.
 Modification of VECD cyclic tension fatigue testing to monotonic tension mode for
more feasibility in mixture testing. A bridge which relates the failure criterion in
cyclic and monotonic mode is necessary to predict the fatigue life of the mixture.
 Verification and re-calibration of the LVECD program once mixture aging and
thermal cracking models are implemented in the new LVECD package.
 Calibration of LVECD software for the thin pavement (less than 4 in) and for the
perpetual pavements.

209
 REFERENCES

Abbasian-Hosseini, S. A., Leming, M. L., & Liu, M. (2015). Effects of Idle Time
Restrictions on Excess Pollution from Construction Equipment. Journal of Management in
Engineering, 04015046.

Abbasian-Hosseini, S. A., Liu, M., & Leming, M. (2015). Comparison of Least-Cost and
Least-Pollution Equipment Fleet Configurations Using Computer Simulation. Journal of
Management in Engineering, 04015003.

Abbasian-Hosseini, S. A., Nikakhtar, A., & Ghoddousi, P. (2014). Verification of lean

construction benefits through simulation modeling: A case study of bricklaying process.
KSCE Journal of Civil Engineering, 18(5), 1248-1260.

Al-Qadi, I. and P. J. Yoo. “Surface Tangential Contact Stresses Effect on Flexible Pavement
Response,” Journal of Association of Asphalt Paving Technologists, Vol. 76, pp. 663-692,
2007.

American Association of State Highway Transportation Officials (AASHTO) (2013).

AASHTO PP60-13, Preparation of Cylindrical Performance Test Specimens Using the
Superpave Gyratory Compactor (SGC).

American Association of State Highway Transportation Officials (AASHTO) (2013).

AASHTO PP61-10, 2013, Developing Dynamic Modulus Master Curves for Hot Mix
Asphalt (HMA) Using the Asphalt Mixture Performance Tester (AMPT).

American Association of State Highway Transportation Officials (AASHTO) (2014).

AASHTO TP 107, Determining the Damage Characteristic Curve of Asphalt Concrete from
Direct Tension Cyclic Fatigue Tests, AASHTO. Washington, D.C.

American Association of State Highway Transportation Officials (AASHTO) (2007).

AASHTO TP62-07, Determining Dynamic Modulus of Hot Mix Asphalt (HMA).

American Society for Testing and Materials (ASTM) (1985) ASTM D 3497-79, Standard
Test Method for Dynamic Modulus of Asphalt Mixtures. Annual Book of ASTM Standards,
Vol. 04.03, ASTM International, West Conshohocken, PA.

American Association of State Highway Transportation Officials (AASHTO) (2011).

AASHTO TP 79, Standard Method of Test for Determining the Dynamic Modulus and Flow
Number for Hot Mix Asphalt (HMA) Using the Asphalt Mixture Performance Tester
(AMPT). AASHTO. Washington, D.C.

210
American Society for Testing and Materials (ASTM) (1985). ASTM C1067-12, 2012.
Standard Practice for Conducting a Ruggedness or Screening Program for Test Methods for
Construction Materials. West Conshohocken, PA.

American Society for Testing and Materials (ASTM) (2012). ASTM D6931-12, 2012,
Standard Test Method for Indirect Tensile (IDT) Strength of Bituminous Mixtures.

Amini, A., Mashayekhi, M., Ziari, H, Nobakht, S., 2011. Life cycle cost comparison of
highways with perpetual and conventional pavements. International Journal of Pavement
Engineering, Vol 13, Issue 6.

Behiry, Ahmed Ebrahim. Fatigue and Rutting Lives in Flexible Pavements. Ain Shams
Engineering Journal, 2012, 3, pp. 367-374.

Bhattacharjee, S., Swamy, A.K., Daniel, J.S., 2009. Application of the elastic-viscoelastic
correspondence principle to determine the fatigue endurance limit of hot mix asphalt. Paper
submitted for 88th Annual Meeting. Transportation Research Board, Washington, D.C.

Bonaquist, R. National Cooperative Highway Research Program Report 629: Ruggedness

Testing of the Dynamic Modulus and Flow Number Tests with the Simple Performance
Tester. Transportation Research Board, Washington, DC, 2008.

Bonaquist, R and D. Christensen. Refining the Simple Performance Tester for Use in Routine
Practice. NCHRP Report 614, Transportation Research Board, National Research Council,
Washington, D.C., 2008.

Buttlar, W. G. and R. Roque. Development and Evaluation of the Strategic Highway

Research Program Measurement and Analysis System for Indirect Tensile Testing at Low
Temperature. Transportation Research Record 1454, TRB, National Research Council,
Washington, DC, pp. 163–171, 1994.

Cao, W., 2015. Experimental and analytical investigations of permanent deformation

behavior of asphalt mixtures under confining pressure. Ph.D. Dissertation, North Carolina
State University, Raleigh, NC.

Cao, W., Eslaminia, M., and Kim, Y.R., 2012. Fatigue performance modeling of Binzhou
perpetual pavements using viscoelastic continuum damage finite element program. In:
Proceedings of the International Society for Asphalt Pavements 2012, Nanjing, China.

Cao, W., A. H. Norouzi, and Y. R. Kim., 2015. Application of viscoelastic continuum

damage approach to predict the fatigue cracking performance of Binzhou perpetual
pavements. Journal of Traffic and Transportation Engineering, in press.

211
Chehab, G., E. O’Quinn, and Y. R. Kim, Specimen Geometry Study for Direct Tension Test
Based on Mechanical Tests and Air-Void Variation in Asphalt Concrete Specimens
Compacted by Superpave Gyratory Compactor. Transportation Research Record, No. 1723,
Transportation Research Board, National Research Council, Washington, D.C., 2000, pp
125-132.

Chehab, G., Y. R. Kim, R. A. Schapery, M. Witczack, and R. Bonaquist, Time-Temperature

Superposition Principle for Asphalt Concrete Mixtures with Growing Damage in Tension
State. Journal of Association of Asphalt Paving Technologists, Vol. 71, 2002, pp. 559-593.

Choi, Y. and Y. R. Kim. Development of Calibration Testing Protocol for Permanent

Deformation Model of Asphalt Concrete. Transportation Research Record: Journal of the
Transportation Research Board, No. 13-2555, 2012, pp. 34-43.

Daniel J. S. and Y. R. Kim, Development of a Simplified Fatigue Test and Analysis

Procedure Using a Viscoelastic Continuum Damage Model, Journal of Association of
Asphalt Paving Technologists, Vol. 71, 2002, pp. 619-650.

Darabi, M.K., Abu Al-Rub, R.K., Masad, E.A., Huang, C., Little, D.N., 2012. A modified
viscoplastic model to predict the permanent deformation of asphaltic materials under cyclic-
compression loading at high temperatures. International Journal of Plasticity, 35: 100-134.

Elwardany, Michael D. Performance of plant produced HMA mixtures with high RAP
content in terms of low temperature cracking, fatigue cracking, and moisture induced
damage. UNIVERSITY OF NEW HAMPSHIRE, 2012.

Eslaminia, M. and M. N. Guddati. Fourier-Finite Element Analysis of Pavements under

Moving Vehicular Loading. International Journal of Pavement Engineering, 2012.

Eslaminia, M., S. Thirunavukkarasu, M. N. Guddati, and Y. R. Kim. Accelerated Pavement

Performance Modeling Using Layered Viscoelastic Analysis. Proceedings of the 7th
International RILEM Conference on Cracking in Pavements, Delft, The Netherlands, pp. 20-
22, 2012.

Federal Highway Administration. Report FHWA-RD-03-031. Pavement Distress

Identification Manual for the NPS Road Inventory Program, June 2003.

Heyden, T., A. Nijhuis, J. Smeyers-Verbeke, B. Vandeginste, and D. Massart. Guidance for

Robustness/Ruggedness Tests in Method Validation, Journal of Pharmaceutical and
Biomedical Analysis, Vol. 24, No. 5–6, pp. 723–753, 2011.

212
Hondros, G. Evaluation of Poisson’s Ratio and the Modulus of Materials of a Low Tensile
Resistance by the Brazilian (Indirect Tensile) Test with Particular Reference to Concrete.
Australian Journal of Applied Science, Vol. 10, No. 3, pp. 243–268, 1959.

Hosseini, A., Nikakhtar, A., & Ghoddousi, P. (2012). Flow production of construction
processes through implementing lean construction principles and simulation. IACSIT
International Journal of Engineering and Technology, 4(4), 475-479.

Hosseini, S. A., Nikakhtar, A., Wong, K. Y., & Zavichi, A. (2012). Implementing Lean
Construction Theory into Construction Processes' Waste Management. In Reston, VA:
ASCEProceedings of the 2011 International Conference on Sustainable Design and
Construction| d 20120000. American Society of Civil Engineers.

Hou, T., B. S. Underwood, and Y. R. Kim. Fatigue Performance Prediction of North Carolina
Mixtures Using Simplified Viscoelastic Continuum Damage Model. Journal of the
Association of Asphalt Paving Technologists, Vol. 79, pp. 35-80, 2010.

Huang, Y. H. Pavement Analysis and Design. 2nd edition, Englewood Cliffs: Prentice-Hall,
2003.

Jackson, N. and J. Puccinelli. Long-Term Pavement Performance (LTPP) Data Analysis

Support: National Pooled Fund Study TPF-5(013)-Effect of Multiple Freeze Cycles and
Deep Frost Penetration on Pavement performance and Cost, Publication FHWA-HRT-06-
121. FHWA, U.S. Department of Transportation, 2006.

Kim, D. Modulus and Permanent Deformation Characterization of Asphalt Mixtures and

Pavements. Ph.D. dissertation, North Carolina State University, Raleigh, NC, 2015.

Kim, D. and Y.R. Kim. Determination of Dynamic Modulus of Asphalt Mixtures Using
Impact Resonance Testing of Thin Disk Specimens. Submitted to the ASTM Journal of
Testing and Evaluation, 2015.

Kim, Y.R., Little, D.N., 1990. One-dimensional constitutive modeling of asphalt concrete.
Journal of Engineering Mechanics, 116(4): 751-772.

Kim, Y. R., D. N. Little, and R. L. Lytton. Fatigue and Healing Characterization of Asphalt
Mixtures. Journal of Materials in Civil Engineering, Vol. 15, No. 1, pp. 75-83, 2003.

Kim, Y. R., Y. Seo, M. King, and M. Momen. Dynamic Modulus Testing of Asphalt
Concrete in Indirect Tension Mode. Transportation Research Record: Journal of the
Transportation Research Board, No. 1891, TRB, National Research Council, Washington,
DC, pp. 163–173, 2004.

213
Kim, Y. R., Y. Lee, and H. J. Lee, Correspondence Principle for Characterization of Asphalt
Concrete, ASCE Journal of Material in Civil Engineering, Vol. 7, No.1, 1995, pp. 59-68.

Underwood, B. S., Y. R. Kim, and M. N. Guddati, Improved Calculation Method of Damage

Parameter in Viscoelastic Continuum Damage Model, International Journal of Pavement
Engineering, Vol. 11, No. 6, 2010, pp. 459-476.

Kutay, M. E., N. H. Gibson, J. Youtcheff, and R. Dongre. Use of Small Samples to Predict
Fatigue Lives of Field Cores: Newly Developed Formulation Based on Viscoelastic
Continuum Damage Theory. Transportation Research Record: Journal of the Transportation
Research Board, No. 2127, National Research Council, Washington, DC, pp. 90-97, 2009.

Lacroix, A. and Y. R. Kim. Performance Predictions of Rutting for NCAT Test Track.
Transportation Research Record: Journal of Transportation Research Board, in press, 2014.

Lee, J. S., A. Norouzi, and Y. R. Kim. Determining Specimen Geometry of Cylindrical

Specimens for Direct Tension Fatigue Testing of Asphalt Concrete. Journal of Testing and
Evaluation, 2015, in press.

Lee, J.S., J.J. Lee, S.A. Kwon, and Y.R. Kim, Use of Cyclic Direct Tension Tests and Digital
Imaging Analysis to Evaluate Moisture Susceptibility of Warm-Mix Asphalt Concrete,
Transportation Research Record No. 2372, Transportation Research Board, National
Research Council, Washington, D.C., 2013, pp. 61-71.

Maher, A., and T. Bennert, Evaluation of Possion’s ratio for Use in the Mechanistic
Empirical Pavement Design Guide, Final Report of FHWA-NJ-2008-004, New Jersey,
http://www.state.nj.us/transportation/refdata/research/reports/FHWA-NJ-2008-004.pdf.

Majidzadeh, K., S. Khedr, and M. El-Mojarrush. Evaluation of Permanent Deformation in

Asphalt Concrete Pavement. Transportation Research Record: Journal of the Transportation
Research Board, No. 715, TRB, National Research Council, Washington, DC, 1979.

Mensching, David J., Jo Sias Daniel, Thomas Bennert, Marcelo S. Medeiros Jr, Michael D.
Elwardany, Walaa Mogawer, Elie Y. Hajj, and Mohammad Zia Alavi. Low-temperature
properties of plant-produced RAP mixtures in the Northeast. Road Materials and Pavement
Design 15, no. sup1 (2014): 1-27.

Moghaddas Tafreshi, S. N., and A. H. Norouzi. Application of Waste Rubber to Reduce the
Settlement of Pavement Systems. Journal of Geomechanics and Engineering, Vol. 9, No. 1,
2015.

214
Moghaddas Tafreshi, S. N., and A. H. Norouzi. Bearing Capacity of a Square Model Footing
on Sand Reinforced with Shredded Tire–An Experimental Investigation, Journal of
Construction and Building Materials, Vol. 35, Pages 547-556, 2012.

Myers, L. A., R. Roque and B. E. Ruth. Mechanisms of Surface-Initiated Longitudinal Wheel

Path Cracks in High Type Bituminous Pavements, Journal of Association of Asphalt Paving
Technologists, Vol. 65, pp. 401-432, 1998.

Myers, L. A. and R. Roque. Evaluation of Top-Down Cracking in Thick Asphalt Pavements

and the Implications for Pavement Design, Transportation Research Circular: Perpetual
Bituminous Materials. Washington, D.C, 2001.

NCHRP Report 602, Project 9-23. Calibration and Validation of the Enhanced Integrated
Climatic Model for Pavement Design. Transportation Research Board, 2008. ISBN 978-0-
309-09929-5.

Nikakhtar, A., Hosseini, A. A., & Wong, K. Y. (2012). Sensitivity analysis of construction
processes using computer simulation: a case study. Advanced Science Letters, 13(1), 680-684

Nikakhtar, A., Hosseini, A. A., Wong, K. Y., & Zavichi, A. (2015). Application of lean
construction principles to reduce construction process waste using computer simulation: a
case study. International Journal of Services and Operations Management, 20(4), 461-480.

Norouzi A., M. Sabouri, and Y. R. Kim. Evaluation of the Fatigue Performance of High RAP
Asphalt Mixtures. Proceedings of 12th International Society for Asphalt Pavements, Raleigh,
NC, 2014.

Norouzi, A., and Y. R. Richard Kim. Mechanistic evaluation of fatigue cracking in asphalt
pavements. International Journal of Pavement Engineering, 1-17, 2015.

Norouzi, A. H., and Y. R. Kim. Ruggedness Study on the Dynamic Modulus Testing of
Asphalt Concrete in Indirect Tension Mode. Journal of Testing and Evaluation, 2015, in
press.

Norouzi, A., D. Kim, and Y. R. Kim. Numerical Evaluation of Pavement Design Parameters
on the Fatigue Cracking and Rutting Performance of the Asphalt Pavements, Journal of
Materials and Structures, 2015, in press.

Park, H. J., M. Eslaminia, and Y. R. Kim. Mechanistic Evaluation of Cracking in in-Service

Asphalt Pavements. Journal of Materials and Structures, pp. 1–20, 2014.

215
Park, S.W., Kim, Y.R., Schapery, R.A., 1996. A viscoelastic continuum damage model and
its application to uniaxial behavior of asphalt concrete. Mechanics of Materials, 24: 241-255.

Park, S.W., Schapery, R.A., 1999. Methods of interconversion between linear viscoelastic
material functions. Part I – a numerical method based on Prony series. International Journal
of Solids and Structures, 36(11): 1653-1675.

Pellinen, T. Evaluation of Surface (top-down) Longitudinal Wheel Path Cracking in Indiana,

Joint Transportation Research Program. Purdue University, West Lafayette, IN, 2002.

Pierce, L. M., McGovern, G. (2014). Implementation of the AASHTO Mechanistic-

Empirical Pavement Design Guide and Software. Report NCHRP Synthesis 457.

Rad, Farhad Yousefi. Estimating blending level of fresh and RAP binders in recycled hot mix
asphalt. Diss. University of Wisconsin Madison, 2013.

Rad, Farhad, Nima Sefidmazgi, and Hussain Bahia. Application of Diffusion Mechanism:
Degree of Blending Between Fresh and Recycled Asphalt Pavement Binder in Dynamic
Shear Rheometer. Transportation Research Record: Journal of the Transportation Research
Board 2444 (2014): 71-77.

R. Reese, Properties of Aged Asphalt Binder Related to Asphalt Concrete Fatigue Life,
Journal of the Association of Asphalt Paving Technologists, Vol. 66, pp. 604-632, 1997.

Roque, R. and W.G. Buttlar, The Development of a Measurement and Analysis System to
Accurately Determine Asphalt Concrete Properties Using the Indirect Tensile Mode,
Proceedings, The Association of Asphalt Paving Technologist, pp. 304-333, 1992.

Sabouri, M. and Y.R. Kim, Development of a Failure Criterion for Asphalt Mixtures under
Different Modes of Fatigue Loading, Transportation Research Record, Transportation
Research Board, National Research Council, Washington, D.C., In press, 2014.

Sabouri, M., T. Bennert, J. S. Daniel, and Y. R. Kim. A Comprehensive Evaluation of the

Fatigue Behavior of Plant Produced RAP Mixtures Laboratory-Produced RAP Mixtures.
Journal of Road Materials and Pavement Design, 2014.

Sabouri M, T. Bennert, J. Daniel, and Y. R. Kim (2014). Evaluating Laboratory-Produced

RAP Mixtures in Terms of Rutting, Fatigue, Predictive Capabilities, and High RAP Content
Potential. Transportation Research Record: Journal of the Transportation Research Board.

216
Schapery, R.A. Correspondence principles and a generalized J integral for large deformation
and fracture analysis of viscoelastic media. International Journal of Fracture, 25: 195-223,
1984.

Seo, Youngguk. Distress Evolution in Highway Flexible Pavements: A 5-Year Study at the
Korea Highway Corporation Test Road. ID JTE 102107. Journal of Testing and Evaluation,
Vol. 38, No. 1, 2010.

Shook, J. F., B. F. Kallas, and B. F. McLeod. Factors Influencing Dynamic Modulus of

Asphalt Concrete. Journal of Association of Asphalt Paving Technologists, Vol. 38, 1969.

Tanesi, J., J. M. Gudimettla, G. L. Crawford, and A. A. Ardani. Ruggedness Study on The

Coefficients of Thermal Expansion of Concrete Test Method (AASHTO T336),
Transportation Research Record: Journal of the Transportation Research Board, No. 2342,
Transportation Research Board of National Academies, Washington, DC, pp. 54-60, 2013.

Underwood, B. S., C. M. Baek, and Y. R. Kim. Use of Simplified Viscoelastic Continuum

Damage Model as an Asphalt Concrete Fatigue Analysis Platform, Transportation Research
Record Issue 2296, Transportation Research Board, National Research Council, Washington,
D.C., 2012, pp. 36-45.

Underwood, B. S., Y. R. Kim, and M. N. Guddati. Characterization and Performance

Prediction of ALF Mixtures Using a Viscoelastoplastic Continuum Damage Model. Journal
of the Association of Asphalt Paving Technologists, Vol. 75, pp. 577-636, 2006.

Underwood, B. S., Y. R. Kim, and M. N. Guddati. Improved Calculation Method of Damage

Parameter in Viscoelastic Continuum Damage Model. International Journal of Pavement
Engineering, Vol. 11, 2010, pp. 459-476.

Wang, Yizhuang. Performance Evaluation of Warm Mix Asphalt Mixtures and Their
Incorporation into the AASHTOWare Pavement ME Program. Master thesis, North Carolina
State University, Raleigh, NC, 2014.

Wang, Y. D., A. Norouzi, and Y. R. Kim. Comparison of Fatigue Cracking Performance

Predictions in Asphalt Pavements Using Pavement ME and LVECD Programs", Journal of
Transportation Research Records, in press, 2015.

Witzak, M. W., K. Kaloush, T. Pellinen, M. El-Basyouny, and H. Von Quintus., Simple

Performance Test for Superpave Mix Design, NCHRP Report 465, Transportation Research
Board, National Research Council, Washington, D.C., 2002.

217
Witczak, M. W. and O. A. Fonseca. Revised Predictive Model for Dynamic Modulus of
Asphalt Mixtures. Transportation Research Record: Journal of the Transportation Research
Board, No. 1540, TRB, National Research Council, Washington, DC, 1996.

Witczak, M. W., R. Bonaquist, H. Von Quintus, and K. E. Kaloush. Specimen Geometry and
Aggregate Size Effects in Uniaxial Compression and Constant Height Shear Tests. Journal of
the Association of Asphalt Paving Technologists, Vol. 69, pp. 733–793, 2000.

Wu, C. and M. Zeng. Effects of Additives for Warm Mix Asphalt on Performance Grades of
Asphalt Binders. Journal of Testing and Evaluation, Vol. 40, No. 2, 2012.

Xu, Q. and M. S. Rahman. Finite Element Analyses of Layered Viscoelastic System Under
Vertical Circular Loading. International Journal for Numerical and Analytical Methods in
Geomechanics, Vol. 32, pp. 897-913, 2008.

Yin, Q., 2011. Analysis on primary road performance of LSPM, Value Engineering, 30(11):
62-63.

Zhang J., M. Sabouri, Y. R. Kim, and M. N. Guddati. Development of a Failure Criterion for
Asphalt Mixtures under Fatigue Loading. Journal of the Association of Asphalt Paving
Technologists, Vol. 82, 2013. In press.

Zhao, Y., Huang, X. Design method and performance for large stone porous asphalt
mixtures. Journal of Wuhan University of Technology – Mater. Sco. Ed., 25(5): 871-876,
2010.

218
219