Technical Report Documentation Page

Technical Report Documentation Page
1. Report No. 2. Government Accession No. 3. Recipient's Catalog No.

FHWA/TX-08/0-5798-1
4. Title and Subtitle 5. Report Date
A REVIEW OF PERFORMANCE MODELS AND TEST October 2007
PROCEDURES WITH RECOMMENDATIONS FOR USE IN THE Resubmitted: June 2008
TEXAS M-E DESIGN PROGRAM Published: September 2008
6. Performing Organization Code
7. Author(s) 8. Performing Organization Report No.

Fujie Zhou, Emmanuel Fernando, and Tom Scullion Report 0-5798-1
9. Performing Organization Name and Address 10. Work Unit No. (TRAIS)
Texas Transportation Institute
The Texas A&M University System 11. Contract or Grant No.
College Station, Texas 77843-3135 Project 0-5798
12. Sponsoring Agency Name and Address 13. Type of Report and Period Covered
Texas Department of Transportation Technical Report:
Research and Technology Implementation Office September 2006-August 2007
P. O. Box 5080 14. Sponsoring Agency Code
Austin, Texas 78763-5080
15. Supplementary Notes
Project performed in cooperation with the Texas Department of Transportation and the Federal Highway
Administration.
Project Title: Develop Test Procedures to Characterize Material Response Behavior and Transfer Functions
for TxDOT M-E Design
URL: http://tti.tamu.edu/documents/0-5798-1.pdf
16. Abstract
In the first year of this project, a comprehensive review was made of the available models for
predicting the major distresses in flexible pavements, including cracking of asphalt layers and chemically
bound layers, permanent deformation of asphalt layers, and permanent deformation of granular base and
subgrade layers. In conducting these reviews, the latest models under consideration in both national efforts
and various state development efforts were reviewed. The models identified for each of the major distresses
are described in this report. Additionally, the associated laboratory test procedures, which can be used to
provide TxDOT with the material properties needed as inputs to both the pavement response and
performance prediction models, were also identified and discussed. Finally, a detailed laboratory testing
plan was proposed for Year 2 study.
17. Key Words 18. Distribution Statement

Flexible Pavement Design, Overlay Test, Repeated No restrictions. This document is available to the
Load Test, Rutting, Fatigue Cracking, Tex-ME public through NTIS:
National Technical Information Service
Springfield, Virginia 22161
http://www.ntis.gov
19. Security Classif.(of this report) 20. Security Classif.(of this page) 21. No. of Pages 22. Price
Unclassified Unclassified 142
Form DOT F 1700.7 (8-72) Reproduction of completed page authorized
A REVIEW OF PERFORMANCE MODELS AND TEST PROCEDURES
WITH RECOMMENDATIONS FOR USE IN THE TEXAS M-E DESIGN
PROGRAM
by
Fujie Zhou
Assistant Research Engineer
Emmanuel Fernando
Research Engineer
and
Tom Scullion, P.E.

Senior Research Engineer
Report 0-5798-1
Project 0-5798
Project Title: Develop Test Procedures to Characterize Material Response Behavior
and Transfer Functions for TxDOT M-E Design
Performed in cooperation with the

Texas Department of Transportation
and the
Federal Highway Administration
October 2007
Resubmitted: June 2008
Published: September 2008
TEXAS TRANSPORTATION INSTITUTE

The Texas A&M University System
College Station, Texas 77843-3135
DISCLAIMER
The contents of this report reflect the views of the authors, who are responsible for the
facts and the accuracy of the data presented herein. The contents do not necessarily reflect the
official views or policies of the Texas Department of Transportation or the Federal Highway
Administration. This report does not constitute a standard, specification, or regulation. The
engineer in charge was Tom Scullion, P.E. (Texas, #62683).
v
ACKNOWLEDGMENTS
This project was made possible by the Texas Department of Transportation (TxDOT) in
cooperation with the Federal Highway Administration. In particular, the guidance and technical
assistance provided by the project director (PD) Joe Leidy, P.E., of TxDOT and the program
coordinator (PC) Darrin Grenfell, P.E., proved invaluable. The following project advisors also
provided valuable input throughout the course of the project, and their technical assistance is
acknowledged: Mark McDaniel, P.E., Construction Division, TxDOT; Billy Pigg, P.E., Waco
District; and Ricky Boles, P.E., Lufkin District.
vi
TABLE OF CONTENTS
Page
List of Figures .............................................................................................................................. viii
List of Tables ................................................................................................................................. ix
Chapter 1. Introduction ...................................................................................................................1
Chapter 2. Models for Predicting Fatigue Cracking of HMA Layers ............................................3
2.1 Background of Fatigue Cracking .........................................................................................3
2.2 Fatigue Cracking Modeling .................................................................................................4
2.3 Recommended Fatigue Model for Tex-ME.......................................................................27
Chapter 3. Models for Predicting Rutting in HMA Layers ..........................................................29
3.1 Introduction .......................................................................................................................29
3.2 Rutting Mechanisms ..........................................................................................................29
3.3 Rutting Prediction Models .................................................................................................37
Chapter 4. Permanent Deformation Models for Granular Base and Subgrades ...........................47
Chapter 5. Fatigue Cracking Models for Chemically Stabilized Materials...................................55
5.1 Definition of Chemically Stabilized Materials ..................................................................55
5.2 Fatigue Cracking Models for Chemically Stabilized Materials ........................................55
5.3 Model Input Requirements and Associated Laboratory Testing .......................................56
Chapter 6. Review of Laboratory Testing Procedures..................................................................59
6.1 Asphalt Rutting Testing .....................................................................................................59
6.2 Asphalt Cracking Testing ..................................................................................................64
6.3 Base and Subgrade Testing................................................................................................67
Chapter 7. Proposed Testing Program for Year 2 of Project 0-5798............................................73
7.1 Laboratory Test Program for HMA Mixes ........................................................................73
7.2 Laboratory Testing Program for Base/Subgrade Materials ...............................................77
References .....................................................................................................................................81
Appendix A. OT for Fracture Properties of HMA Mixes.............................................................91
Appendix B. VESYS Test Protocol for Asphalt Mixes................................................................97
Appendix C. Recommended Permanent Deformation and Resilient Modulus
Laboratory Test Protocols for Unbound Granular Base/Subbase Materials
and Subgrade Soils..................................................................................................................105
vii
LIST OF FIGURES
Figure Page
1. Three Modes of Crack Opening Displacement: (a) Model I – Opening Mode,
(b) Mode II – Shearing Mode, (c) Mode III – Tearing Mode ...........................................15
2. Non-dimensionalized Bending and Shearing SIF vs. Non-dimensionalized
Crack Length......................................................................................................................16
3. 8-node Quadrilateral and Quarter-point Triangular Elements...........................................17
4. Cohesive Cracking Model Analogy...................................................................................19
5. Case 9, Damage Field and Crack Pattern after 396,000 Load Applications .....................23
6. Predicted logk1 vs. Measured logk1....................................................................................25
7. Trench Profiles for 161 (Top) and 162 (Bottom) ..............................................................32
8. Pavement Cross Sections for (a) Test Tracks 1 and 2 (six test sections), (b) Test
Track 3 (three test sections), and (c) Test Tracks 4 and 5 (six test sections) ....................34
9. Typical Transverse Profiles for Section 4A.......................................................................35
10. Progression of Hump to Wheelpath Volume Ratio: Unmodified Mixes...........................36
11. Progression of Hump to Wheelpath Volume Ratio: Modified Mixes ..............................36
12. Log Kr1 Coefficient vs. Voids Filled with Asphalt ...........................................................40
13. Typical Permanent Deformation Behavior (After MEPDG Supplemental
Documentation Appendix GG-1).......................................................................................48
14. Thin Pavement Structures Considered in Test Plan...........................................................68
15. Permanent Strain Development under Repeated Loading .................................................68
16. Proposed Laboratory Test Program for Thin Pavements...................................................70
17. DM Test Setup and Test Specimens (Cylindrical and Prismatic) .....................................74
18. RLPD Loading Configuration............................................................................................75
19. The Hamburg Test Device and Test Specimen..................................................................75
20. The Overlay Tester and Specimen Setup...........................................................................76
21. Permanent Deformation Curves from Tests on Grade 1 Spicewood Base Specimens......79
22. Permanent Deformation Curves from Tests on Grade 2 Groesbeck Base Specimens ......79
viii
LIST OF TABLES
Table Page
1. Comparison of Fatigue Cracking Modeling Approaches ..................................................28
2. Summary of Ratios of Hump and Wheelpath Volumes.....................................................34
3. Fine Aggregate Angularity Index Used to Adjust Findex ....................................................41
4. Coarse Aggregate Angularity Index Used to Adjust Cindex................................................41
5. Summary of Permanent Deformation Prediction Models..................................................49
6. Minimum Values of 7 Days Unconfined Compressive Strength,
for Chemically Stabilized Materials in the MEPDG .........................................................55
7. Comparison of Various Test Methods for Permanent Deformation Evaluation................62
8. Performance Test Included in the Superpave Mixture Analysis System...........................62
9. Summary of Post SHRP Permanent Deformation Testing Research.................................63
10. Comparison of Test Methods for Cracking .......................................................................66
11. Proposed Laboratory Test Program for Thin Pavements...................................................71
12. Proposed Laboratory Testing Program for HMA Mixes ...................................................77
13. Test Matrix for Comparative Evaluation of Permanent Deformation Setups ...................78
ix
CHAPTER 1
INTRODUCTION
The objective of Texas Department of Transportation (TxDOT) Project 0-5798 is to

develop the framework for the development and implementation of the next level of
Mechanistic-Empirical Pavement Design Guide (MEPDG) for TxDOT (Tex-ME). As specified
in the project statement, this initial project, which is in the development process, will focus on
the following areas:
• Identify and evaluate test procedures that characterize material properties needed
to predict pavement response.
• Assemble existing performance prediction models (transfer functions), and
evaluate their feasibility of being implemented in Texas. Key considerations will
be the models’ performance in basic sensitivity analysis, the practicality of the
data input requirements, and their performance at simulating results from
accelerated pavements tests (APT).
• Calibrate the selected transfer functions with available performance data from the
LTPP databases, various test track studies, and whatever performance data is
available from the databases being assembled in Texas.
In the first year of this project, a comprehensive review was made of the available models
for predicting the major distresses in flexible pavements, including cracking of hot-mix asphalt
(HMA) layers, permanent deformation of HMA layers, and permanent deformation of granular
base and subgrade layers. In conducting these reviews the latest models under consideration in
both national efforts and various state development efforts were reviewed. The models
identified for each of the major distresses are described in next three chapters of this report.
Another very important aspect of this project is to identify laboratory testing procedures,
which can be used to provide TxDOT with the material properties needed as inputs to both the
pavement response and performance prediction models. As stated in the Texas Transportation
Institute (TTI) proposal, the eventual Tex-ME program (as being proposed in other ME
programs) will provide the user with various levels of flexibility when selecting material
properties. At the lowest level, default values will be available for all of the design items used
by TxDOT. However, Level 2 will be properties derived from the current specification and
acceptance/design tests that are run on a routine basis by TxDOT. Level 1 will be the highest
level where advanced materials characterization techniques will be used on all layers in the
pavement structure.
In Chapters 2, 3, 4, and 5, the models most appropriate for potential inclusion in the
future Tex-ME will be identified. Chapter 6 will provide a summary of laboratory test
procedures proposed to provide materials inputs for these models. The recommended material
characterization protocols for Level 1 inputs are provided in Appendix A, B, and C. Chapter 7
describes the detailed laboratory testing plan. Additionally, performance data is being assembled
1
from the National Center for Asphalt Technology (NCAT) test track and from the California
Heavy Vehicle Simulator (HVS) program, and response and performance data will be generated
on the instrumented sites currently being constructed in Texas. Materials from each of these
projects will be characterized later in the TTI laboratory.
2
CHAPTER 2
MODELS FOR PREDICTING FATIGUE CRACKING OF HMA LAYERS
This chapter is divided into the following major sections:
• Section 2.1 provides a description of the concept of fatigue cracking.

• Section 2.2 discusses the existing models available to predict fatigue cracking.
• Section 2.3 recommends models to be considered for inclusion in the Tex-ME.
2.1 BACKGROUND OF FATIGUE CRACKING
As noted by Suresh, the word fatigue originated from the Latin expression fatigare which
means “to tire” (1). Although commonly associated with physical and mental weariness in
people, the word fatigue has also become a widely used terminology in engineering vocabulary
for the damage and failure of materials under cyclic loads. Fatigue is defined as a term which
“applies to changes in properties which can occur in a metallic material due to the repeated
application of stresses or strains, although usually this term applies specially to those changes
which lead to cracking or failure.” This definition is also valid for fatigue of HMA concrete,
because asphalt pavements do not crack immediately after the traffic starts, and it usually takes
many years and millions of load applications.
For asphalt pavement fatigue cracking, two phases of the degradation process are
generally considered: crack initiation and crack propagation. During the process of the crack
initiation, microcracks grow from microscopic size until, as some research indicates, a critical
length of about 7.5 mm is reached (2). In the crack propagation process, a single crack or a few
cracks grow until the crack(s) reaches the pavement surface. Researchers noted that both
microcracks and macrocracks can be propagated by tensile or shear stresses or combinations of
both. Thus, in a pavement structure, microcracks can form and grow in any location where
tensile or shear stresses generated by traffic or environmental variations are sufficiently large.
Any tensile or shear stress applied to a field where microcracks exist may cause them to grow, to
reach critical size, and then to propagate as macrocracks.
The number of traffic load repetitions, Nf, to cause a crack to penetrate through the full
depth of the pavement surface layer is the sum of the number of load repetitions for crack
initiation, Ni, and the number of load repetitions required for macrocrack to propagate to the
pavement surface, Np.
Nf = Ni + Np (1)
It should be noted that all existing asphalt pavement thickness design programs do not
directly consider the Np, but indirectly consider it through the field calibration.
3
The physical evolution law governing each of the two phases (crack initiation and crack
propagation) may be quite different so that different approaches have been proposed to model
these two phases. More information is provided below.
2.2 FATIGUE CRACKING MODELING
Fatigue cracking is one of the major distress modes considered in asphalt pavement
designs and has been studied for several decades. In 1955, Hveem demonstrated the concept that
fatigue cracking has a higher propensity to occur on an asphalt pavement when the pavement
experiences a larger deflection and a higher loading frequency (3). Since then, different types of
fatigue cracking models have been proposed. Generally speaking, these models can be classified
into three categories: crack initiation models, crack propagation models, and crack initiation and
propagation models. The following subsections discuss these models.
2.2.1 Crack Initiation Models
Most of existing fatigue cracking models actually only describe the crack initiation phase
of asphalt pavement cracking, and the crack propagation is taken into account through field
calibration. These types of models can be further classified as strain-based fatigue model,
energy-based fatigue model, and damage-based fatigue model. Detailed discussion for each type
of models is provided below.
Strain-Based Fatigue Models
This type of model has been implemented in many of the existing asphalt pavement
design procedures. The most commonly used model form to predict the number of load
repetitions to fatigue cracking, as shown in Equation 2, is a function of the tensile strain and mix
stiffness (modulus).
k2 k3
⎛1 ⎞ ⎛ 1 ⎞
N f = k 1 ⎜⎜ ⎟⎟ ⎜⎜ ⎟⎟ (2)
⎝ εt ⎠ ⎝ S mix ⎠
where:
Nf = number of repetitions of load to cause fatigue cracking,
εt = tensile strain at the critical location,
Smix = stiffness of the material, and
k1, k2, k3 = regression coefficients obtained from laboratory testing.
Four examples of the strain-based fatigue models were developed by Shell Oil (4), the
Asphalt Institute (MS-1) (5), the SHRP A-003A-Berkeley model (6), and the MEPDG fatigue
model (7). These are shown below:
• Shell Oil fatigue cracking model
Because of the known impact between stress state and damage mechanism for different
thicknesses of asphalt layers, Shell Oil Company has developed fatigue damage prediction
equations for the two major forms of laboratory fatigue testing (4). In practice, the constant
stress equation would be recommended for thick asphalt layer design, whereas the constant strain
would be for thinner layers, although the transition from thick to thin is somewhat arbitrary. The
4
equations developed are presented as follows:
[ ]
Constant strain: N f = A f 0.17PI − 0.0085PI(Vb )+ 0.0454Vb − 0.112 5 ε −t 5 E − 1.8 (3)
Constant stress: N f = A f [0.0252PI − 0.00126PI(Vb )+ 0.00673Vb − 0.0167]5 ε −t 5 E − 1.4 (4)
where:
Nf = number of repetitions to fatigue cracking,

E = stiffness of the material,
Vb = effective asphalt content in volume (%),
Af = laboratory to field adjustment factor (default =1.0), and
PI = penetration index.
• Asphalt Institute (MS-1) model

⎛ Vb ⎞
4.84⎜⎜ − 0.69 ⎟⎟
Vb + V a ⎠ ε − 3.291 E − 0.854
Nf = 0.00432 * 10 ⎝
t (5)
where:

Vb = effective asphalt content in volume (%), and
Va = air voids (%).
Note that this MS-1 fatigue equation is based upon modifications to constant stress
laboratory fatigue criteria. The Asphalt Institute Ninth Edition of the MS-1 design manual uses a
field calibration factor of 18.4 so that predictions from the model can be matched to observed
field performance (5). This correction factor was developed for a 20 percent level of wheelpath
cracking; it was recommended by Finn in his classic NCHRP 1-10 study (8).
• SHRP A-003A-Berkeley fatigue model
−2.720
N f = 2.738 * 105 * e 0.077*VFBε t −3.624 * S o'' (6)
where:

VFB = percentage of voids filled with asphalt, and
So" = initial loss-stiffness of mix as measured in flexure (psi).
5
The SHRP A-003A research team established an integrated asphalt mix and asphalt
thickness design system based on the research results from the SHRP A-003A program. This
model has been recalibrated based on the HVS tests (9).
• MEPDG fatigue cracking model
The NCHRP 1-37A research team examined the Shell Oil and the MS-1 models for the
recently developed MEPDG. It was found that the Shell Oil models possessed more scatter and
did not possess any definite trends (10); also, the MS-1 model had much less scatter and resulted
in a definite trend. Thus, the MS-1 model was selected and implemented in the MEPDG. In
contrast to the models described above, the MEPDG fatigue cracking model actually includes the
following three models:
• The number of the load repetitions fatigue model
3.9492 1.281
⎛1 ⎞ ⎛1⎞
N f = 0.00432 * k1 ∗ C ⎜⎜ ⎟⎟ ⎜ ⎟ (7)
⎝ εt ⎠ ⎝E⎠
where:

hac = asphalt layer thickness (inches), and
k1, C = correction factors.
⎛ V ⎞
4.84⎜⎜ b −0.69 ⎟⎟
Vb +Va
C = 10 ⎝ ⎠ , (8)
1
k1 = (9)
0.003602
0.000398 +
1 + e11.02 − 3.49∗ hac
• Fatigue damage model
Fatigue damage caused by different traffic loads is calculated as the ratio of the applied
number of traffic repetitions to the allowable number of load repetitions (to some failure level)
as shown in Equation 10.
T
ni
D= ∑N
i =1 i
(10)
where:
D = damage,
T = total number of periods,
ni = actual traffic for period i, and
Ni = allowable failure repetitions under conditions prevailing in period i.
6
• Fatigue cracking amount model
Finally, another transfer function is used to calculate the fatigue cracking from the fatigue
damage, which was developed and calibrated using the LTPP data. The final fatigue damage
versus cracking amount model in the MEPDG is as follows:
⎛ 6000 ⎞ ⎛ 1 ⎞
FC = ⎜ ⎟∗⎜ ⎟ (11)
⎝ 1 + e C1 −C2 ∗log D ⎠ ⎝ 60 ⎠
where:
FC = percentage of fatigue cracking of total lane area,

D = damage (Equation 10),
C1 = -2*C2,
C2 = -2.40874-39.748*(1+hac)-2.85609, and
hac = asphalt layer thickness (inches).
Energy-Based Fatigue Models
Since the early work done by Van Dijk, the energy-based fatigue models have been
widely investigated (11). Various representations and applications of dissipated energy concepts
have been proposed and are presented below.
• Initial dissipated energy approach

Initial dissipated energy (IDE) is the area under the stress-strain curve between the
loading and unloading cycle measured during the initial loading cycles. Typically, in fatigue
testing, the first 50 cycles are regarded as the conditioning cycles, and the dissipated energy at
the 50th loading cycle is considered as the initial dissipated energy. Initial dissipated energy can
be a good indicator of fatigue performance for similar mix types (12). Baburamani and Porter
(13) also showed a good correlation between the initial dissipated energy and fatigue life.
Additionally, Ghuzlan in his thesis found the initial dissipated energy is one of the most
important factors that affect fatigue behavior of HMA mixes (14). Based on extensive bending
beam fatigue testing data, Tayebali et al. proposed the following as a surrogate model to relate
initial dissipated energy to fatigue life (6):
N f = 6.72e 0.049VFB (w0 )−2.047 (12)
where:
Nf = number of repetitions to failure,
VFB = percentage of voids filled with bitumen, and
w0 = initial dissipated energy.
One disadvantage of the initial dissipated energy approach is that it is not appropriate for
the whole loading range, especially when dealing with low strain fatigue tests. Shen and
Carpenter did not find any good correlation between the initial dissipated energy and fatigue life
(15).
7
• Cumulative dissipated energy approach
The cumulative dissipated energy is the summation of the dissipated energy experienced
by the material during the fatigue test, which relates the fatigue behavior to both initial and final
test cycles. A relationship between the cumulative dissipated energy and the number of loading
cycles to failure is characterized as:
( )z
WN = A N f (13)
where:
WN = cumulative dissipated energy to failure,

A, z = experimentally derived mix coefficient, and
Nf = number of load cycles to failure.
Van Dijk was one of the earliest researchers who did an extensive study on fatigue of
HMA materials based on the dissipated energy concepts (11, 16). He found that there is a strong
relationship between the cumulative dissipated energy and the number of loading cycles to
failure. This relationship is not affected by the loading mode (controlled-stress or controlled-
strain), the effects of frequency (between 10Hz and 50Hz) and temperature (between 10ºC to
40ºC), and the occurrence of rest periods. However, it is highly material dependent and has to be
mix specific to be applied.
Pronk and Hopman suggested the dissipated energy per cycle/period is responsible for
the fatigue damage (17). The total dissipated energy combined with Wöhler’s curve was used to
develop the fatigue equation. Additionally, Tayebali et al. introduced two terms: the stiffness
ratio, which is the ratio of the stiffness at load cycle (i) to the initial stiffness; and the dissipated
energy ratio, which is defined as the ratio of cumulative dissipated energy up to load cycle (i) to
the cumulative dissipated energy up to fatigue life (18). Their work showed there is a unique
relationship between the stiffness ratio and the dissipated energy ratio, but not necessarily
between cumulative dissipated energy and fatigue life, which is also verified by SHRP A-404 (6)
and later by Fakhri (19). This relation was also found to be mix and temperature dependent.
• Work ratio approach
This approach was first introduced by Van Dijk and Visser (16) and further developed by
Rowe (12). The work ratio, ψN1, is defined as the ratio between the product of the initial
dissipated energy in cycle 1 and N1 divided by the cumulative dissipated energy, as shown in
Equation 14.
w0 N1
ψ N1 = (14)
WN 1
8
where:
w0 = initial dissipated energy,
N1 = number of load cycles to crack initiation, and
WN1 = cumulative dissipated energy at cycle N1.
Work ratio can be calculated in terms of the initial rheological property of the HMA mix
and the mode of loading factor, Γ, as follows:
⎧ ⎫
⎪⎪ 2Sinφ0 ⎪⎪
ψ N1 = ⎨ ⎬ (15)
⎪ Sinφ0 + ΓSin ⎡a E60
⎪⎩ ⎢⎣
*
( )
(0.224 − 0.222 log a ) ⎤⎪
⎥⎦ ⎪⎭
where:
⎡100 − A − B ⎤
Γ=⎢
100 ⎥ = mode of loading,
⎣ ⎦
⎡ ε 0 − ε 60 ⎤
A=⎢ ⎥ × 100 = percent change in strain,
⎣ ε 60 ⎦
⎡ σ − σ 60 ⎤
B=⎢ 0 ⎥ × 100 = percent change in stress,
⎣ σ 60 ⎦
⎡ log φ 0 − 0.244 log E 0* ⎤
⎢ ⎥
1− 0.222 log E 0* ⎥⎦
a = 10 ⎢⎣
ø0 = initial phase angle,
E0* = initial extensional complex modulus (equivalent to bending stiffness),
E60* = 60 percent reduction in initial extensional complex modulus,
ε60 = 60 percent reduction in the initial strain, and
σ60 = 60 percent reduction in the initial stress.
Rowe (12) found that the work ratio can be used effectively to predict the fatigue life to
crack initiation through Equation 13. The crack initiation (Equation 16) is assumed to occur at
60 percent reduction of original extensional complex modulus.
N1 = 205Vb 6.44 w0−2.01ψ 1N.64

1 (16)
where:
N1 = number of load cycles to crack initiation,
Vb = volume of binder (%),
w0 = initial dissipated energy, and
ψN1 = work ratio.
• Dissipated energy ratio approach
Carpenter and Jansen first initiated an improved implementation of the dissipated energy
concept for HMA fatigue analysis, in which a dissipated energy ratio was used as a parameter to
relate to fatigue life (20). This approach believes that not all the dissipated energy is responsible
for material damage. For each cycle, the loss of energy due to material mechanical work and
9
other environmental influence remains almost unchanged. Therefore, if the dissipated energy
starts to change dramatically, it could be explained as the development of damage. Later, this
approach was examined and refined by Ghuzlan and Carpenter (14, 21), and Carpenter et al.
(22). It is found that the relationship between dissipated energy ratio and fatigue life is
fundamental in that it is independent of loading level, loading mode, and mix type (22).
This dissipated energy ratio approach was further improved by Shen and was renamed as
the ratio of dissipated energy change (RDEC) approach considering the fact that it is using the
ratio of the amount of dissipated energy change between different loading cycles to represent the
damage propagation (23). The distinctiveness of the RDEC approach is the relationship between
the energy parameter, plateau value (PV), and the fatigue life (Nf). This relation, as presented in
Equation 17, is unique for all HMA mixes, all loading modes (controlled stress and controlled
strain), all loading levels (normal and low damage levels), and various testing conditions
(frequency, rest periods, etc.) (23).
N f = 0.4801(PV )−0.9007 (17)
where:
Nf = fatigue life, and
PV = plateau value.
Furthermore, Shen also developed the following equation to estimate the energy
parameter, PV.
PV = 2.612 × 10 −10 (IDE )2.758 S 2.493 (VP )3.055 (GP )−2.445 (18)
where:
PV = plateau value,
IDE = initial dissipated energy,
S = the flexural stiffness of HMA mix from the laboratory fatigue test, MPa,
AV
VP = volumetric parameter, VP = ,
AV + Vb
AV = air voids, %,
Gmb × Pac
Vb = the asphalt content by volume, Vb = 100 × ,
Gb
Gmb = bulk density, %,
Pac = percent of asphalt by total weight of mix,
Gb = bulk specific gravity of the asphalt binder, assuming Gb=1.03,
PNMS − PPCS
GP = aggregate gradation parameter, GP = ,
P200
PNMS = percent of aggregate passing the nominal maximum size sieve,
PPCS = percent of aggregate passing the primary control sieve, and
P200 = percent of aggregate passing #200 (0.075mm) sieve.
As an energy-based approach, the RDEC is fundamental and has been demonstrated valid
for different testing methods such as flexural bending beam fatigue testing (14, 15, 21, 22) and
10
uniaxial tension testing (24), and various materials including both HMA materials and Portland
cement concrete materials (25). While it is more fundamentally correct to use dissipated energy
rather than tensile strain, current design systems are based on multi-layer elastic system and the
viscoelastic dissipated energy cannot be easily estimated. Recognizing this limitation, Shen also
developed an alternative equation, which is a strain-based fatigue prediction equation, as
presented in Equation 19 (23).
PV = 61.336ε 5.052 S 2.749 (VP )1.643 (GP )−0.094 (19)

where:
ε = tensile strain, and

all other parameters are the same as those in Equation 18.
Damage-Based Models
Two types of damage-based fatigue models have been proposed: viscoelastic continuum
damage mechanics model and CalME damage-based fatigue model. The main difference
between these two approaches is how to interpret the load reduction and fatigue damage during
the fatigue test. The CalME damage-based fatigue model interprets the load reduction as fatigue
damage, but for the visco-elastic continuum damage mechanics approach, the load reduction is
caused by both viscoelastic property of the HMA mix and fatigue damage. Thus, the viscoelastic
continuum damage mechanics approach uses the concept of “pseudo-stiffness” to define fatigue
damage. Note that the pseudo-stiffness is defined as the ratio of a stress value to a pseudo-strain
value at the peak pseudo-strain of each cycle. More discussion is presented below for each
model.
• Viscoelastic continuum damage mechanics approach
Continuum damage theory was originally developed by R.A. Schapery for analyzing the
response of solid rocket fuels and similar viscoelastic materials (26, 27, 28). Lytton, Kim, and
Little later applied Schapery’s work to asphalt concrete (29). Their work was extended and
refined by Y. Richard Kim, Daniel, Lee, and Yong-Rak Kim (30, 31, 32, 33, 34, 35). Practical
application of this continuum damage theory has been made by Lee et al. (35), and Christensen
and Bonaquist (36). The brief discussion presented below largely follows the development of
Christensen and Bonaquist (36).
Schapery defined uniaxial pseudo-strain as follows:

t
∫ E (t − t ) ∂t
1 ∂ε
ε (t ) =
R '
'
dt ' (20)
ER
0
where:
ε = strain,
εR(t) = pseudo-strain at time t,
E = relaxation modulus,
t΄ = time at which loading begins, and
11
ER = an arbitrary reference modulus, often set at unity.
The above definition is very similar to that for linear viscoelastic (LVE) stress:
t
∫ ( ) ∂t dt
∂ε
σ (t ) = E t − t ' '
'
(21)
0
where σ(t) is stress at time t.
From Equations 20 and 21, under LVE conditions, we have:
σ (t )
ε R (t ) = (22)
ER
That means that the pseudo-strain is equal to the stress resulting from an applied strain
history. To quantify damage accumulation, Kim et al. used the concept of pseudo-stiffness
defined by Equation 23 (33):
σ
C = max
R
(23)
ε max
where C is the normalized pseudo-stiffness; normalization meaning that adjustments are made in
the calculation of C for individual specimens so that the initial value (undamaged) is always
unity.
For fatigue testing, a specimen is subjected to a given strain-controlled loading. With the
damage accumulating during the fatigue test, the resulting stress σmax for every cycle will
gradually decrease compared to the pseudo-strain. Thus, Equation 23 simply defines
pseudo-stiffness as the ratio of the non-linear modulus to the initial LVE modulus. The
constitutive equation for uniaxial loading of a viscoelastic material with damage is given below
(31):
R
σ max = Cε max (24)
The applicable stress-pseudo-strain relationship is as follows (36):
α
⎛ ∂W R ⎞
σ max = ⎜⎜ R ⎟
⎟ (25)
⎝ ∂ε max ⎠
where WR is the pseudo-strain energy density function. The time dependent growth of damage
can be given by the following equation (36):
α
dS ⎛⎜ ∂W R ⎞
⎟
= − (26)
dt ⎜⎝ ∂S ⎟
⎠
where S is a variable characteristic of the amount of internal damage in a material, and α is a

material constant, which usually has a value close to 2.0. Equations 24 and 25 can be combined
and integrated to yield the following relationship (36):
12
( )
W R = 0.5C ε max
R 2
(27)
Regarding the relationship between pseudo-stiffness C and the internal damage parameter S, Lee
and Kim proposed a form of generalized power law (32):
C = C10 − C11 (S )C12 (28)
where C10, C11, and C12 are constants describing the rate of damage accumulation of a specimen
under cyclic loading. It should be noted that this equation would become negative at some value
of S, which means that the damaged modulus would also be negative, and an applied tensile
strain would result in a compressive stress. Knowing the limitation of Equation 28, Christensen
and Bonaquist suggested a better function in a simple exponential form (36):
C = exp(C2 S ) (29)
where C2 is a constant indicative of the rate of damage accumulation in a specimen under cyclic
loading. Now, substitute Equation 29 into Equation 27, and differentiate with respect to S; the
following relationship results:
∂W R
∂S
= 0.5C2 exp(C2 S ) ε max
R
( )2
(30)
Then, substitute Equation 30 into Equation 26 and integrate to solve for t:

S =t
2α exp(− αC2 S )
t= (31)
α (− C2 )1+α ε max
R
S =0
Now, if the reference modulus ER = 1, then, ε R (t ) = σ (t ) . In addition, for sinusoidal

loading, the maximum tensile stress is equal to:
σ max − σ min
= σ 0 = ε 0 × E LVE (32)
2
where:
E LVE = LVE complex modulus,
σ0 = maximum tensile stress (or stress amplitude), and
ε0 = maximum tensile strain (or strain amplitude).
Note that the number of loading cycles N is loading time t times frequency f (Hz).
Equation 30 can then be solved over the given integration limits and given in the
following form:
Nf =
[ ( ) ]
2α f exp − αC 2 S f − 1 ⎛ 1 ⎞
⎜⎜ ⎟⎟
2α
(33)
α (− C2 )1+α E LVE ⎝ ε 0 ⎠
2α
where Sf is the value of internal damage variable S at failure. It is clear that fatigue life (Nf) is a
13
function of the damage evolution characteristics of the material (C2), viscoelastic material
properties (α, E ), fatigue test conditions (f, ε0), and a failure criterion (S1f).
The main advantage of using continuum damage mechanics to predict fatigue behavior of
HMA mixes is that the time-temperature superposition principle can be employed to shift the
characteristic curve determined at one temperature to different temperatures. In that way, it is
possible to save considerable testing time and materials. The disadvantage of this approach is
that it needs sophisticated laboratory tests and data analysis techniques. Generally, this approach
is still under development. Application of this approach to predict fatigue cracking of asphalt
pavement has not been seen yet in the literature.
• CalME damage-based fatigue cracking model

Another approach of considering fatigue damage caused by repeated loading is through
the stiffness ratio (SR) proposed by Tsai, Harvey, and Monismith (37). Different from
viscoelastic continuum damage approach, this approach assumes that stiffness (or load)
reduction is caused by fatigue damage. This approach has been used in the CalME design
program. In the present version of CalME, the SR is predicted in the following equation (38):
(
SR = exp − α × N β ) (34)
where:
SR = stiffness ratio, defined as the ratio of stiffness at repetition n over the initial
stiffness (taken at about 50 repetitions),
N = number of load applications, and α and β are assumed on the format:
α = exp(αA + αB × t + αC × ln (w) + αD × t × ln (w))

β = β A + βB × t + β C × ln (w)
where:
t = temperature in ºC,
w = internal energy density (½×ε2×E), and
αA, αB, αC, αD, βA, βB and βC = constants determined from 4-point bending beam
fatigue tests under controlled strain.
The use of SR damage-based approach has several advantages: 1) stiffness is easy to

measure both in the laboratory and in the field, and 2) stiffness is often utilized as an input for
linear layered-elastic programs for pavement analysis, thus making it useful for programming
fatigue performance prediction. However, it is worth noting that no asphalt thickness design
program but the CalME program used this approach. Actually, Monismith and his associates are
continuously developing this model. More research is still needed to refine this model (38).
2.2.2 Crack Propagation Models
Different types of models have been developed to characterize fatigue crack propagation.
Models reviewed here include the classical fracture mechanics model, the cohesive crack model,
14
and the non-local continuum damage model.
Classical Fracture Mechanics Model
Since Majidzadeh et al. introduced fracture mechanics concepts into the field of asphalt
pavements, the fracture mechanics approach has been widely used in predicting pavement
cracking (39). In contrast to continuum mechanics, the fracture mechanics approach focuses on
crack propagation. The crack propagation process can be caused by Modes I, II, III, or a
combination of the three modes of loading (Figure 1):
• Mode I loading (opening mode, KI) results from loads that are applied normally to the
crack plane (thermal and traffic loading).
• Mode II loading (sliding mode, KII) results from in-plane shear loading, which leads to
crack faces sliding against each other normal to the leading edge of the crack (traffic
loading).
• Mode III loading (tearing mode, KIII) results from out-of plane shear loading, which
causes sliding of the crack faces parallel to the crack leading edge. Compared to Modes I
and II, Mode III is rare and is often neglected for simplicity.
The fact that the mechanisms of fatigue cracking (bending and shearing) discussed
previously can be exactly modeled by fracture Modes I and II makes the fracture mechanics
approach very attractive for modeling fatigue crack propagation.
(a) (b) (c)

Figure 1. Three Modes of Crack Opening Displacement: (a) Mode I − Opening Mode, (b)
Mode II − Shearing Mode, (c) Mode III − Tearing Mode (40).
The generally accepted crack propagation law was proposed by Paris and Erdogan in the
form of Equation 35 (41). It has successfully been applied to asphalt concrete by many
researchers for the analysis of experimental tests and prediction of reflection cracking and low
temperature cracking.
dc
= A ∗ (ΔK )
n
(35)
dN
where:
15
c = crack length,
N = number of loading cycles,
A, n = fracture properties of asphalt mixture determined by the experimental
test, and
ΔK = stress intensity factor (SIF) amplitude, depending on the geometry of
the pavement structure, fracture mode, and crack length.
The number of load cycles Nf needed to propagate a crack through an asphalt layer of
thickness h can be estimated by numerical integration in the form of Equation 36.
h
dc
Nf = ∫ A(ΔK )
0
n
(36)
The use of Paris’ law (Equation 35) for the description of the crack growth process in
viscoelastic materials, such as HMA mixes, has been theoretically justified by Schapery (42, 43,
44). However, it is apparent that both SIF and HMA fracture properties (A and n) must be
known in order to predict fatigue crack propagation. In the following paragraphs, the focus will
be placed on the SIF calculation and HMA fracture properties determination.
• Calculation of SIF
Since there is a singularity at the crack tip in the stress field, a finite element (FE)
program is needed to compute the SIF. Two special SIF computation programs for pavements
have already been developed for crack propagation. The first one named CRACKTIP was
developed for thermal cracking by Lytton and his associates at TTI in 1976 (45). The
CRACKTIP is a two dimensional (2-D) FE program, and it models a single vertical crack in the
asphalt concrete layer via a crack tip element (45). This program has been successfully used to
develop the SIF model and predict the cracking propagation. Figure 2 shows the SIF of bending
(SIFb) and SIF of shearing (SIFs) versus crack length relationship. It is interesting to note that
there is a “neutral axis” where bending stresses no longer cause crack propagation. Its location
depends on the level of load transfer and the moduli of the pavement layers. This neutral axis
must be considered in order to accurately predict reflection cracking.
16
Figure 2. Non-dimensionalized Bending and Shearing SIF vs. Non-dimensionalized Crack
Length (46).
Although 2-D FE programs run much faster than the three dimensional (3-D) FE
programs, it is common knowledge that the SIFs computed from 2-D plane strain conditions are
overestimated because of the difference between plane strain conditions and the 3-D nature of a
cracked geometry and loading conditions. In order to balance the accuracy of 3-D FE (3-D
nature of the cracked pavement geometry and the loading condition) and fast running time of
2-D FE, a semi-analytical FE program named SA-CrackPro was recently developed at TTI by
Zhou et al. (47). This SA-CrackPro provides for adequate accuracy and efficient analysis of
crack propagation in an asphalt layer. The SA-CrackPro program uses a single quarter-point
triangular singular element to produce the stress singularity at the crack tip as shown in Figure 3
(48). The SIFs (KI and KII) can then be elegantly determined based on Equations 37 and 38
proposed by Ingraffea and Manu, if the displacements of the crack tip nodes computed by FE
analysis are correlated to those predicted by theory (49).
b d
Z(u)
c e
Y(v)
Figure 3. 8-node Quadrilateral and Quarter-point Triangular Elements.

2π G
KI = [4(vB − vD ) + vE − vC ] (37)
L κ +1
2π G
K II = [4(uB − uD ) + uE − uC ] (38)
L κ +1
where:
G = shear modulus,
κ = (3 − υ ) / (1 + υ ) for plane stress,
κ = 3 − 4υ for plane strain, and
L = element length.
The inputs to the SA-CrackPro program are the same as those used in the multilayer
elastic program for calculating tensile strain at the bottom of asphalt layer. Furthermore,
regression equations for bending and shearing SIFs are under development. Once SIF regression
17
equations are developed, it becomes possible to practically consider crack propagation in the
structural design of asphalt pavement.
• HMA fracture properties: A, n
Laboratory tests characterizing HMA fracture properties (A and n) have been conducted
for a long time (39, 50, 51, 52, 53, 54, 55, 56, 57, 58). Among them, the most systematic
laboratory studies on fracture properties A and n were conducted by Molenaar and his associates
(53, 54, 55, 56, 57, 58). The most often used test to quantify these parameters is the repeated
direct tension test. However, this test method is relatively complicated and has not been widely
accepted in the field of asphalt pavement. Recently, Zhou et al. developed a very simple, quick
test procedure to determine fracture properties of HMA mixes (A and n) using the TTI overlay
tester (OT) (47). This procedure can be routinely used to determine HMA fracture properties.
More information about using the OT to determine fracture properties of HMA mixes is
presented in Appendix A.
In summary, the classical fracture mechanics-based fatigue crack propagation model is

conceptually sound, and the mechanisms of fatigue cracking (bending and shearing) can be
easily described with a fracture mechanics-based model. Furthermore, this type of model, as
discussed previously, has been successfully employed to predict the reflection cracking in
asphalt overlays by different researchers. Moreover, the two difficult aspects of application of
fracture mechanics: SIF calculation and fracture properties (A and n), as noted above, have been
solved. Thus, a fracture mechanics-based crack propagation model is mature enough to be
implemented in any mechanistic-empirical structural pavement design system.
Cohesive Crack/Zone Model
It is a well known fact that asphalt concrete is a non-linear elastic material, and its
fracture behavior is very complicated. Uzan and Levenberg discussed the phenomenology of
asphalt concrete fracture and provided an overview of the cohesive crack model (CCM) (59).
There is a strongly non-linear fracture process zone (FPZ) around the crack tip in asphalt
concrete as shown in Figure 4. It is important to mention that in some situations, for HMA
mixes, the FPZ can extend to considerable lengths, up to a few centimeters (60). In order to
account for a relatively large plastic yield zone ahead of a crack tip, Dugdale (61) and Barenblatt
(62) proposed a “correction” for the classical linear elastic fracture mechanics. Their model
approximated an elastic-plastic material behavior by applying closure stresses at the crack’s tip.
Hillerborg et al. proposed a similar model to account for the relatively large FPZ that have been
encountered in concrete failure (63). The above models are generally considered cohesive
cracking models, because the models employ cohesive closure stresses near the crack tip region.
18
Figure 4. Cohesive Cracking Model Analogy (59).
The three fundamental hypotheses of the standard cohesive crack model are as follows:
• The properties of the materials outside the process zone are governed by the
undamaged state.
• A crack length can be divided into two separate regions (see Figure 4): a traction free
length and a cohesive part. In the cohesive part, crack opening resisting tractions
exist, and there is still stress transfer between its faces, which is done by introducing
closure stresses. The CCM postulates that the cohesive part of the crack begins to
form at a “point” when the maximum principal stress at that “point” reaches the
tensile strength of the material (and the crack propagation perpendicular to the
maximum stress direction) (64). Actually, this postulation is a crack initiation
criterion.
• Meanwhile, the stress transfer capability of the cohesive part follows a descending
path, from full transfer capability (when the cohesive crack faces just begin to depart
[say peak stress conditions]) down to zero transfer capability as the displacement
between the two cohesive crack faces reach a critical opening. This representation
constitutes the CCM’s crack propagation criterion. During the crack propagation
analysis, the traction free crack is incrementally advanced whenever the calculated
displacement reaches the critical opening in size. The stress transferred between the
faces of the crack is described by a post-peak function (softening function). In the
case of the opening mode, the function is:
σ = f (w) (39)
19
where σ is the tensile stress and w is the crack opening displacement. This softening
curve of the material is considered to be a main component of the cohesive crack
model. Although each material has its unique softening curve, determined only by
experiments, Petersson first found that the softening curve is similar in shape for
different mixtures of Portland cement concrete when the softening curves are plotted
in a non-dimensional form (65).
Jenq and his associates first applied the CCM to simulate crack initiation and propagation
in asphalt concrete mixtures (60, 66). However, their work got little attention until the
Superpave model team started to develop an advanced asphalt concrete mixture material
characterization model (67). Then, Uzan and Levenberg developed a laboratory experimental
test (direct tension test) to determine the CCM parameters (59). Similar work was later done by
Seo et al. (68). Soares et al. considered the heterogeneity in crack modeling of asphalt concrete
mixtures (69). The latest research in this field is being led by Buttlar and their associates (70,
71, 72, 73). Their research focus was on developing a laboratory test such as a disk-shaped
compact tension test to determine the CCM parameters and associated numerical simulation.
In general, the application of the CCM to HMA mixes is still in the preliminary stage. All
studies discussed above only applied the CCM to cracking under monotonic loading. To extend
the CCM to repeated loading (such as reflection cracking), additional material parameters
describing damage accumulation under unloading and reloading are needed. However, no work
on this has been done yet. Therefore, the CCM is very promising, but it is not mature yet. More
research is still needed.
Non-Local continuum Damage Mechanics Model for Crack Propagation (74)
Wu et al. proposed another approach for modeling crack propagation (74). Continuum
damage mechanics (CDM) allows one to describe the heterogeneous microprocesses involved
during the straining of materials and structures at the macroscale. The basic theory of CDM can
be found in papers by Chaboche (75, 76). However, the application of CDM to asphalt concrete
mixes was pioneered by Lee and Kim (77, 78), followed by many other researchers, and it is still
under development. The ultimate state of local CDM corresponds generally to macroscopic
crack initiation upon which it becomes a crack propagation problem and should be considered in
the framework of fracture mechanics. If local CDM is used to describe crack propagation, the
spurious mesh dependency then comes into play. Fortunately, this mesh-dependency can be
avoided by introducing non-local mechanics. A non-local continuum is a continuum in which
the stress at a point depends not only on the strain history of the same point, but also on the
strain history of the point’s neighbor. Bazant and Jirasek gave a comprehensive, state-of-the-
research review of non-local formulations and provided a series of causes as well as motivations
for introducing non-local continuum (79).
Non-local CDM is essentially an “enhancement” of local-CDM. Thus, the local-CDM is

the first to be introduced below.
• Local CDM
20
The stress-strain relationship for a linear elastic material with isotropic damage can be
written as:
σ = (1 − ω )C : ε (40)
Eν Eν
where Cijkl = λδ ijδ kl + μ (δ ik δ jl + δ il δ jl ) with λ = and μ = is the elasticity
(1 +ν )(1 − 2ν ) 2(1 +ν )
tensor and the scalar ω ∈ [0,1] represents the damage. Damage is defined such that ω= 0
represents the initial, undamaged material, and ω= 1 represents a state of complete loss of
integrity.
Equation 40 is complemented by the damage evolution law:
& = g (ω, ~ε ) ~ε&

ω (41)
+
where ω& is the time derivative of damage ω, g (ω , ε~ ) is a non-negative function to enforce the
irreversibility of damage evolution, ε~ = f (ε ) is a measure of the strain that reflects its damaging
effect due to cracking, and . + denotes the Macaulay bracket, which is an average over a
representative volume. A popular definition of ε~ is given by Mazars and Cabot (80):
ε~ = ∑
3 2
i =1
εi +
(42)
where εi is the ith principal strain. When dealing with loading histories composed of well-defined,
discrete cycles, an evolution law in terms of the number of cycles and the loading amplitudes is
often considered more practical (81). Such a cycle-based damage evolution law can be obtained
from Equation 42 by integrating over one loading cycle resulting in a relation of the form (82):
∂ω
= G (ω ,ε~ a ) (43)
∂N
where N is the number of load cycles, ε~a is the amplitude of ε~ for the current load cycle, and G
is a non-negative function representing the damage accumulation property of the material.
• Non-local CDM
Numerous ways have been proposed to incorporate non-locality into the constitutive
relations of materials. The most successful ones fall into two categories: integral formulation and
implicit gradient formulation. The implicit gradient formulation was recommended since it is
much easier to implement in the FE code, and it is a special case of the integral formulation.
Implicit gradient formulation is proposed by Peerlings et al., in which a non-local strain

ε is introduced to replace the local strain measure ε~ in damage evolution Equations 41 and 42
(83). And ε and ε~ are related through an additional differential equation:
21
ε − c∇ 2ε = ε~ (44)
where c has a dimension of length and is related to the internal length scale, which should be
approximately equal to the maximum grain size of the material, and ∇ 2 = ∑i ∂ 2 / ∂xi2
is the Laplacian operator. Physically, Equation 44 implies that ε is a spacial average of ε~ and
the radius of the averaging domain is in proportion to c.
The introduction of Equation 44 leads to a coupled problem between the displacement

field and the non-local strain field. The non-local strain becomes an additional degree of freedom
for each node. The evaluation of a consistent algorithmic tangent at any Gauss point requires
only the current strain ε, damage ω, and non-local strain ε for that same point. In this sense, the
implicit gradient formulation is mathematically local and is much easier to be incorporated into
existing FE codes.
After developing the non-local CDM-based crack propagation model, the SHRP beam
fatigue tests were conducted to calibrate the model’s parameters. Frequency sweep test was used
to determine the Young’s modulus master curves of two HMA mixes. Fatigue tests provided
stiffness reduction curves that captured the material degradation process of the two asphalt
concrete mixes under repetitive loading. FE models were established to simulate the beam
fatigue test. Damage evolution law parameters were calibrated by matching the calculated and
measured stiffness reduction curves. Finally, the laboratory calibrated crack propagation model
was verified by simulating reflection cracking in an HVS test conducted on an asphalt concrete
overlay placed on a cracked and jointed concrete pavement. The model not only recovered the
most dominant crack pattern observed in the field, but it also predicted the reflection cracking
life of the overlay with reasonable accuracy. Figure 5 shows damage field and crack pattern
after 396,000 load repetitions. In conclusion, the implicit gradient non-local CDM, implemented
in a FE program, provides a promising mechanistic model for simulating crack propagation in
asphalt concrete overlays.
22
Figure 5. Case 9, Damage Field and Crack Pattern after 396,000 Load Applications (74).
In general, the non-local CDM reflection cracking model, similar to the CCM discussed
previously, is very advanced. Wu’s research results demonstrated this promising model to
predict reflection cracking in asphalt overlays over existing pavements (74). However, this non-
local CDM model is still under development and not ready for routine use.
2.2.3 Integrated Crack Initiation and Crack Propagation Model
The first integrated crack initiation and crack propagation model was proposed by Lytton
et al. under SHRP A-005 study (2). Since then, significant research efforts led by Lytton have
been made at TTI to study fatigue behavior of HMA mixes. The most comprehensive study just
finished by Walubita et al. further expanded the SHRP A-005 approach (84). The new name for
the expanded approach is Calibrated Mechanistic approach with Surface Energy (CMSE).
Practically speaking, this CMSE approach is still under development, and significant work is still
needed to refine and expand it in order to practically apply this approach for pavement design
and analysis. Alternatively, Zhou et al. took the concept of crack initiation and propagation and
developed an OT-based fatigue cracking prediction approach (47). Detailed information about
the OT-based approach is discussed below.
As noted previously, fatigue cracking is the combination of crack initiation and crack
propagation process. The number of traffic load repetitions (Nf) to cause a crack to initiate and
propagate through the asphalt surface layer is the sum of the number of load repetitions needed
for micro-cracks to coalesce to initiate a macro-crack (crack initiation, Ni) and the number of
load repetitions required for the macro-crack to propagate to the surface (crack propagation, Np).
23
N f = Ni + N p (45)
In the OT based approach, both Ni and Np are estimated from the fracture properties (A
and n), which are determined from the OT.
Estimation of Ni
It is well known that the traditional fatigue models established based on bending beam
fatigue tests mainly address the crack initiation stage. Thus, the traditional fatigue model shown
in Equation 46 is proposed to estimate Ni.
k2
⎛1⎞
N i = k1 ⎜ ⎟ (46)
⎝ε ⎠
The key issue of estimating Ni is to establish a “bridge” between fracture properties (A

and n) and fatigue parameters k1 and k2. Based on fracture mechanics, Lytton et al. found the
following relationships between these parameters (2).
⎛ n⎞
⎜ 1− ⎟
d⎝ 2⎠ ⎡ ⎛ c ⎞ (1− nq ) ⎤
k1 = ⎢1 − ⎜ 0 ⎟ ⎥ (47)
Ar n (1 − nq )E n ⎢⎣ ⎝ d ⎠ ⎥⎦
k2 = n (48)
Equation 47 indicates that parameter k1 (or logk1) is a function of k2 (= n), A, and E:
log k 1 = f (k 2 , E, A ) (49)
As reported by Schapery (28), Molenaar (53), Jacobs (55), Lytton et al. (2), and Erkens et
al. (57), the fracture property A is highly related to parameters n (= k2) and log E. Thus, it is
reasonable to simplify Equation 47 as follows:
log k 1 = a 1 + a 2 k 2 + a 3 log E (50)
where a1, a2, and a3 are regression constants. It is worth noting that a very similar relationship
shown in Equation 50 can also be developed based on continuum damage mechanics (35).
Therefore, Equation 50 is theoretically sound. The key to estimating parameter k1 is to
determine regression constants a1, a2, and a3.
In order to do so, the results from historical fatigue test data were reviewed. It was found
that the bending beam fatigue test (BBFT) is the most often used method to characterize fatigue
behavior of HMA mixes. In this project, several sources of BBFT data were assembled and used
to develop the required regression parameters in Equation 50. After carefully reviewing the
available BBFT data, the following data sets were selected for modeling:
24
• SHRP A-003A fatigue data (6): 218 tests,
• Harvey et al.-1996 (9): 211 tests,
• Sousa et al.-1998 (85): 129 tests,
• Tsai-2002-WesTrack fatigue data (86): 150 tests,
• Ghuzlan and Carpenter-2003 (87): 478 tests, and
• Tsai and Monisimth-2005 (88): 162 tests.
The total number of available BBFT data sets was 1348. The test variables covered in
these 1348 sets of data include type of asphalt binder (conventional and modified), asphalt
contents, type of aggregates, type of HMA mixes (dense-graded, Superpave, and SMA), air void
contents, test temperatures, and aging conditions.
Using the “Solver” optimization technique in Microsoft Excel® by minimizing the sum
of squared errors between the measured and the predicted k1, the regression constants a1, a2, and
a3 were determined, and the final k1 equation is presented below. Figure 6 shows the predicted
and the measured logk1.
6.97001− 3.20145 k 2 − 0.83661 log E
k = 10 R2=0.99 (51)
1
With Equations 46, 48, and 51, Ni can be estimated provided that tensile strain at the
bottom of asphalt layer and modulus of asphalt layer are known.
-70
-60
-50
Predicted Log k1
y = 0.9906x - 0.1023
R2 = 0.9906
-40
-30
-20
-10
0 -10 -20 -30 -40 -50 -60 -70
0
Measured Log k1
Figure 6. Predicted logk1 vs. Measured logk1.
Estimation of Np
Theoretically, with known fracture properties A and n (from the OT) and SIF (from the
FE program or regression equations), Np can be estimated from Equation 52:
25
h
1
Np = ∫ A(ΔK )
c0
n
dc (52)
where c0 is the initial crack length and h is asphalt layer thickness. Based on micro-mechanics
theory and laboratory test results, Lytton et al. recommended an initial macro-crack length (c0) of
7.5 mm, which results from micro-cracks growth (2).
However, it is well known that one axle passing over a crack results in three loading
sequences: shearing (approaching to a crack), bending (loading on the top of the crack), and
shearing (leaving from the crack). These three loading sequences make it difficult to directly
estimate Np from Equation 52. In this project, an alternative approach was proposed.
Instead of estimating Np from Equation 52, the authors recommended calculating the
crack propagation length induced by one axle pass using the following form of Paris’ law.
Δc = A(ΔK )n × ΔN (53)
Note that for one axle pass, a crack should propagate three times: ∆cs, ∆cb, and ∆cs,
corresponding to the shearing, bending, and shearing loading sequence, respectively. Thus, the
crack propagation length (∆c) induced by one axle pass is the sum of ∆cs, ∆cb, and ∆cs.
[ n n
]
Δc = 2 × Δcs + Δcb = A × 2 × (ΔK Shearing ) + (ΔK Bending ) × ΔN (54)
Add more axle passes and repeat the above process until the accumulated crack length is
equal to asphalt layer thickness (h). Then, Np is the sum of all the number of passes.
OT-Based Fatigue Cracking Prediction Approach
Based on the information presented above, the OT-based fatigue cracking prediction
approach is proposed. The key steps of this approach are summarized below:
1. Run dynamic modulus test to develop dynamic modulus master curves of HMA
mixes.
2. Run the OT to determine HMA fracture properties: A and n.
3. Estimate traditional fatigue model parameters, k1, k2, and Ni from Equations 46, 48,
and 51.
4. Compute the SIF caused by traffic load from regression equations or FE programs.
5. Estimate Np with an initial macro-crack length (c0 =7.5 mm) using Equation 54.
6. Calculate Nf from Equation 45.
7. Calculate the damage caused by a specified number of load repetitions (n) using
Miner’s law (Equation 55).
n
Damage = ∑ ∗ 100% (55)
Nf
8. Predict fatigue crack amount using the model proposed in the MEPDG (10).
26
100
crack area(% ) = (56)
1 + exp(a1 + a2 ∗ log Damage )
where a1, a2 = calibration coefficients.
Note that Equation 56 is a sigmoidal function form, which is bounded with 0 percent
cracking as a minimum and 100 percent cracking as a maximum. Specifically, it was assumed
that a fatigue cracking value of 50 percent cracking of the total area of the lane theoretically
occurs at a damage percentage of 100 percent. This assumption clearly indicates the following
relationship:
a1 = -2×a2 (57)
In summary, based on theoretical review and 1348 sets of BBFT data, a “bridge”
(equations) between crack initiation model (traditional fatigue model) and crack propagation
model (Paris’ law) was developed in this section. An OT-based fatigue cracking prediction
approach including both crack initiation and crack propagation was then proposed.
2.3. RECOMMENDED FATIGUE MODEL FOR TEX-ME
Table 1 presents a comparison among different types of fatigue cracking models based on
several parameters, such as the capability of characterizing fatigue crack initiation and
propagation process and compatibility of the model to the existing TxDOT FPS framework. As
noted in Table 1, both energy and strain-based fatigue models consider only crack initiation of
fatigue cracking and ignore the crack propagation stage. The CalME considers the fatigue
damage, but this approach still focuses on the crack initiation stage. The authors believe that the
lack of focus on crack propagation is why the current “crack initiation” approaches require very
large field calibration factors in the order of 15 to 300.
The viscoelastic continuum damage mechanics model, cohesive crack/zone model, and
non-local continuum damage mechanics model are very advanced models, and the current status
of these advanced models is that they are still under development and are many years away from
implementation. Thus, the OT-based fatigue cracking model is currently thought to be the best
option for better modeling fatigue cracking, and it is recommended for inclusion in a future
Tex-ME program. Furthermore, this approach has proven to be a practical approach for
predicting fatigue cracking under TxDOT 9-1502 pooled-fund study project (89).
27
Table 1. Comparison of Fatigue Cracking Modeling Approaches.
Crack Propagation Compatible
Mechanisms with FPS
Development Combined
Fatigue Models Crack
Status Crack Mechanisms
Propagation Yes No
Initiation
Bending+Shear
Shell Oil Finished √ √
Strain- AI Finished √ √
based SHRP-A-
Model Finished √ √
003A
MEPDG Finished √ √ √
Energy-based Models Finished √ √ √
Viscoelastic
Continuum Under
Damage- √ √
Damage development
based
Model
Models
Under
CalME √ √
improvement
Cohesive Cracking Under
√ √
Model development
Non-local Continuum
Under
Damage Mechanics √ √ √
development
Model
TTI OT-based Fatigue
Finished √ √ √ √
Cracking Model
28
CHAPTER 3
MODELS FOR PREDICTING RUTTING IN HMA LAYERS
3. 1. INTRODUCTION
Rutting gradually develops with increasing number of load applications and appears as
longitudinal depressions in the wheelpaths accompanied by small upheavals to the sides. Rutting
is always a major concern for at least two reasons: 1) if the surface is impervious, the ruts trap
water, and at depths of about 0.2 in., hydroplaning (particularly for passenger cars) is a definite
threat; and 2) as the ruts progress in depth, steering becomes increasingly difficult, leading to
added safety concerns. Therefore, it is important to make efforts to minimize rutting. This
literature review focuses on the following aspects of asphalt rutting:
• rutting mechanisms,
• rutting prediction methodology, and
• laboratory testing to characterize rutting resistance of HMA concrete.
3.2. RUTTING MECHANISMS
Rutting occurs in flexible pavements because of the accumulation of small permanent

deformations in any of the pavement layers or the subgrade. Such deformations may be caused
by too much repeated stress applied to the pavement layers or by an HMA mix that is too low in
shear strength. In the first case, the rutting is considered more a structural or construction
problem. It is generally the result of an underdesigned or undercompacted pavement section or
of an unbound base or subgrade that have been weakened by the intrusion of moisture. In the
second case, the rutting is normally a mixture design or placement-related problem. When an
asphalt pavement layer has inadequate shear strength, a small but permanent shear deformation
occurs each time a heavy truck applies a load. A rut will then appear with enough load
applications. As noted below, most pavement surface rutting, at least for reasonably stiff
supporting materials, is confined to HMA layers.
Regarding HMA layer rutting, it is commonly accepted that rutting (permanent

deformation) is a manifestation of two different mechanisms and is a combination of
densification (volume change) and repetitive shear deformation (plastic flow with no volume
change). It is difficult to determine the relative amounts of rutting occurring in each HMA layer,
and the relative proportions of rut depth that can be attributed to densification and shear, because
many factors, such as binder type, binder content, mix type, load level, temperature, initial
compacted density, etc., have influence on rutting. The following paragraphs document field
trench studies on asphalt pavement rutting.
• AASHO Road Test-1962: Trenching studies performed at the AASHO Road Test (90)
and test-track studies reported by Hofstra and Klomp indicated that the shear deformation
rather than densification was the primary rutting mechanism (91). The importance of
29
placing materials at high densities in order to minimize the shear deformation was
emphasized.
Measurements at the AASHO Road Test indicated that the surface rut depth
reached a limiting value for asphalt concrete thickness of approximately 10 in. Thicker
layers did not exhibit additional rutting.
• Hofstra and Klomp-1972 (91): The deformation through the asphalt-concrete layer was
greatest near the loaded surface and gradually decreased at lower levels. Because rutting
is caused by plastic flow, such a distribution of rutting with depth is reasonable: more
resistance to plastic flow is encountered at greater depths and shear stresses are smaller
there as well.
• Uge and van de Loo-1974 (92): The deformation within an asphalt layer (thickness
reduction under the action of pneumatic tires) no longer increased with increasing layer
thickness beyond a certain threshold (130 mm in their case).
• Eisenmann and Hilmer-1987 (93): The rutting was mainly caused by deformation flow
without volume change, including two stages:
o In the initial stage of trafficking, the increase of irreversible deformation below

the tires is distinctly greater than the increase in the upheaval zones. In this initial
phase, therefore, traffic compaction has an important influence on rutting.
o After the initial stage, the volume decrement beneath the tires is approximately
equal to the volume increment in the adjacent upheaval zones. This is an
indication that compaction under traffic is completed for the most part and that
further rutting is considered to be representative of the deformation behavior for
the greater part of the lifetime of a pavement.
• Brown and Cross-1989 (94): Brown and Cross’s trench results showed that permanent
deformation is limited to the upper 100 mm (4 in.) of the mix. It also indicated that, at
least for reasonably stiff supporting materials, most pavement rutting is confined to the
asphalt pavement layer.
• UC-Berkeley HVS study-2000 (95): Air-void contents of cores taken in the wheelpath
after trafficking showed relatively little densification, except when the overlays were
poorly compacted, despite final rut depths of 15 to 24 mm. Note that the maximum rut
depth is defined for their study as the vertical distance between the bottom of the
wheelpath and the highest of the adjacent humps. The average proportion of rut depth
attributable to the shear flow as opposed to the densification varied between 19 to 100
percent, depending on the overlay type. The greatest shear flow occurred on the 38 mm
asphalt-rubber hot mix gap-graded (ARHM-GG) sections and the least on the dense-
graded asphalt concrete (DGAC) sections. These results indicated that rutting did not
consist of a process of densification to a very low air-void content followed by rapid
shear flow. Instead, it appears that rutting consists of simultaneous densification and
shear flow, with the rates of shearing and densification varying at different periods of rut
development. The performance of the poorly compacted ARHM-GG mixes indicated that
considerable shear flow occurred at high air-void contents.
30
On many of these test sections, HVS trafficking was continued well beyond the
failure rut depth of 12.5 mm (0.5 in.). Final rut depths ranged between 15 and 24 mm.
All of the sections subjected to trafficking by highway wheels/tires had an initial period
of rapid rut development, followed by a second period with a reduced rate of rutting that
continued until trafficking was stopped. The aircraft wheel test section (513RF) had a
slight reduction in rut rate as trafficking progressed, but much less than that of the other
test sections. None of the test sections showed any evidence of a “tertiary” period of rut
development in which the rate of rut development increases again after the second period
of reduced rutting rate. The lack of a tertiary rutting period, despite final ruts of 15 to
24 mm, suggested that the tertiary stage is either:
o a phenomenon that occurs only in the laboratory during triaxil repeated load
testing,
o a phenomenon that occurs in the field when temperatures (40-55ºC) or loads
exceed those previously experienced by the mix, or
o a phenomenon that only occurs when rut depths have already exceeded 24 mm.
Results of trenching and profilometer measurements at the top of the base

indicates that less than 5 mm of the final average maximum rut depth occurred at the
surface of the aggregate base on any of the test sections. Note that the asphalt layer
thicknesses of these test sections range from 7.5 to 9.0 in. The measurements are not
precise because of noise caused by individual particles at the surface of the base.
Disturbance at the surface of the base was minimized during sawing and slab removal,
although some disturbance was inevitable due to penetration of the prime coat into the
base and adhesion of particles of the base to the asphalt layers when the slabs were
removed.
• Federal Highway Administration’s ALF-1999 (96)
o The decreases in air voids due to trafficking indicated that when the rut depth in
the asphalt pavement layer was 20 mm, the range in percent densification was
approximately 20 to 55 percent, which is 4 to 11 mm.
o Based on the rutting data from all pavements, rutting occurred in all asphalt
pavement lifts. No particular lift or group of lifts consistently rutted the most.
The rut depths used in this analysis consisted of both the rut depth due to
densification and viscous flow.
o By splitting the total rut depth into the percent rut depth in the asphalt pavement
layer and the percent rut depth in the underlying layers, it was found that the
percentage of rutting in the asphalt pavement layer decreased with increasing
total rut depth. The percentage of rutting in the underlying layer increased as the
asphalt pavement layer became thinner due to lateral shearing and flow.
o The reductions in air voids due to trafficking (densification) in the top and bottom
halves of the 200 mm thick asphalt pavement layer were not significantly
different at a 95 percent confidence level for any pavement test. Based on the
average densification in the top and bottom halves, it was found that the average
31
densification in the bottom half could be greater than, equal to, or less than the
average densification in the top half.
• TxDOT’s trench on SPS1-US281 sections-2001 (97): The two trench profiles indicated that
the rutting was coming primarily from the top 50 mm (2 in.) HMA layer. As shown in
Figure 7, the deep rutting accompanied the considerable lateral shear flow.
Figure 7. Trench Profiles for 161 (Top) and 162 (Bottom).
• National Center for Asphalt Technology (NCAT) 2000 test track-2004 (98): Results
from NCAT 2000 test track are summarized below:
o The amount of permanent deformation in all of the test sections was very low.
Permanent deformation essentially stopped when the air temperature was less
than 28ºC. The accumulation of permanent deformation in the second summer
was significantly less than the first.
o Under traffic, mixes containing PG 64-22 densified more than mixes containing
PG 76-22 binder. As expected, the amount of permanent deformation was over
60 percent less in the sections that contained PG 76-22 as compared to the
sections containing PG 64-22.
o Adding an additional 0.5 percent binder above optimum to the mixes produced
with PG 64-22 increased permanent deformation by approximately 50 percent.
However, there was no increase when an extra 0.5 percent binder was added to
mixes produced with PG 76-22. This may indicate that slightly more binder can
32
be added to mixes with two high temperature binder bumps to improve durability
without sacrificing rut resistance.
o The amount of permanent deformation calculated based on the pavement
densification in most cases exceeded the actual permanent deformation. This
supports the fact that most of the test sections had very stable mixes and that the
small amount of permanent deformation observed was mainly related to
densification or consolidation.
o The performance of the coarse-graded and fine-graded mixes was about the same.
Hence, this study indicates that similar performance would be expected for
coarse-graded and fine-graded mixes with respect to permanent deformation.
• NCAT 2003 test track-2006 (99): Results from NCAT 2003 test track are listed as
follows:
o After Phase I (2000) testing, 23 sections were left in place for Phase II. The
maximum rutting in any of these sections that were left in place and subjected to
20 million total equivalent single axle loads (ESALs) was 7 mm, which means all
mixes are very rut resistant.
o The factor that most affected rutting of HMA pavements was the asphalt binder
PG grade. The modified asphalt reduced the rutting by over 50 percent when
compared to unmodified asphalt, which basically confirmed the observation in
Phase I (2000) test track.
o SMA sections had more rutting than the Superpave sections, but neither had
significant rutting. It appears that initial rutting in the SMA was due to
densification and/or aggregate reorientation. After this initial rutting, little
additional rutting occurred. No cracking had occurred in any of the SMA sections.
o SMA mixes placed in 2003 test track were designed with 50 and 75 gyrations
with the Superpave Gyratory Compactor. These mixes have performed well,
which indicates that this lower compactive effort can be used to increase the
optimum asphalt content and produce improved durability.
o Laboratory air voids had a significant effect on dense-graded mixes designed
using an unmodified asphalt binder. However, the air voids had little effect on
performance of those mixes using modified asphalts.
o The asphalt pavement analyzer (APA) showed a good trend with rutting
performance. Additional work is needed with the APA along with other
performance tests to clearly develop the best relationships.
o Coarse-graded and fine-graded mixes were compared. When fine-graded mixes
were compared to coarse-graded mixes, they were equally resistant to rutting, less
likely to be permeable, quieter, similar in friction, possibly easier to compact, and
higher in optimum asphalt content.
• Florida HVS study-2005 (100): Florida DOT studied the influence of modified asphalt
binder on rutting using HVS. Figure 8 shows the pavement structures tested under HVS.
As noted in Table 2, the HVS testing was conducted at two temperatures: 50ºC and 65ºC.
Figure 9 shows a typical rutting development under HVS loading. The observations from
Florida’s HVS testing are summarized below.
33
Figure 8. Pavement Cross Sections for (a) Test Tracks 1 and 2 (six test sections), (b) Test
Track 3 (three test sections), and (c) Test Tracks 4 and 5 (six test sections).
Table 2. Summary of Ratios of Hump and Wheelpath Volumes.
34
o As expected, at the beginning of the test, densification would be the major factor
in rut development, as the initial level of air voids would decrease in the
wheelpath because of repeated wheel loads. However, from Figures 10 and 11, it
can be seen that even after only 100 passes, approximately 40 to 50 percent of
rutting can be attributed to shear flow. That is, volume of the asphalt concrete in
the humps is approximately half the volume of the wheelpath. This indicates that
rutting may be caused simultaneously by densification and shear flow of the
asphalt concrete. Figure 10 also shows that for unmodified asphalt mixtures, the
ratio of hump to wheelpath volume increases with an increasing number of HVS
wheel passes. This indicates that after an initial number of wheel passes, most of
the rutting occurs only because of shear flow. In comparison, for the modified
asphalt mixtures, however, the ratio of hump to wheelpath volume remains
somewhat constant, with approximately 40 percent of rutting caused by shear
flow.
o An important factor to be noted is that during the course of this study, there was
no permanent deformation within the limerock base layer as all the rutting was
confined to the asphalt layers alone. This stiff base layer may have affected the
formation of humps at the edge of the wheelpath.
o In conclusion, since the only variable between the two Superpave mixes
considered in this study is the asphalt binder type, the rut initiation (or initiation
mechanism) is primarily controlled by the stiffness of the asphalt binder.
Figure 9. Typical Transverse Profiles for Section 4A.
35
Figure 10. Progression of Hump to Wheelpath Volume Ratio: Unmodified Mixes.
Figure 11. Progression of Hump to Wheelpath Volume Ratio: Modified Mixes.
36
• Summary comments on rutting mechanism:
o Available results reviewed clearly show that, at least for reasonably stiff
supporting materials, most pavement rutting is confined to the asphalt pavement
layer. Furthermore, rutting or permanent deformation, in most cases, is limited to
the upper 100-150 mm (4-6 in.) of an HMA layer.
o HVS studies at high temperature conducted by UC-Berkeley and Florida
Department of Transportation indicated that rutting did not consist of a process
of densification to a very low air-void content followed by rapid shear flow.
Instead, it appears that rutting consists of simultaneous densification and shear
flow, with the rates of shearing and densification varying at different periods of
rut development. Also, it seems that in-place density and asphalt binder type
(modified and unmodified) affect the relative contribution of densification and
shearing to the rutting.
o Florida HVS test results also clearly indicated the significant effect of test
temperature on rutting development, as shown in Table 2 (Sections 1A vs.1B and
2A vs. 2B).
o Results from NCAT Phases I and II clearly showed that rutting performance of
HMA mix with PG 76-22 binder is significantly different from those mixes with
PG 64-22 binder. The observed differences are listed as follows:
¾ Mixes designed by Superpave volumetric design method are rut resistant.

This observation is also consistent with two national surveys of Superpave
mix performance conducted by NCAT.
¾ The modified asphalt reduced the rutting by over 50 percent when compared
to unmodified asphalt, which basically confirmed the observation in the
Phase I 2000 test track.
¾ For rutting resistance, mixes with PG 76-22 binder are not so sensitive to
asphalt binder content as those with PG 64-22 binder. Adding an additional
0.5 percent binder above optimum to the mixes produced with PG 64-22
increased permanent deformation by approximately 50 percent. However,
there was no increase when an extra 0.5 percent binder was added to mixes
produced with PG 76-22.
¾ Similarly, laboratory air voids had a significant effect on dense-graded mixes
designed using an unmodified asphalt binder. However, the air voids had little
effect on performance of those mixes using modified asphalts.
o It should be noted that the above observations are based on thick asphalt
pavement and may not be applicable to thin-surfaced asphalt pavements typically
used on stabilized granular base or low-volume roads.
3.3 RUTTING PREDICTION MODELS
There currently exist two broad approaches to the problem of designing against rutting or
permanent deformation. One approach is design procedures based upon an empirical correlation
of excessive deformations related to some predefined “failure” condition of the pavement. This
group may be further subdivided into procedures based upon empirical tests used to categorize
37
the material strength and those designs based upon the use of a limiting subgrade strain (or
stress) criteria from elastic layered theory. The major advantage of these procedures is the fact
that they currently are workable design tools for pavement analysis. The major disadvantage of
such an approach is that they cannot be used to predict the amount of deformation anticipated
after a given number of load applications.
The second approach encompasses procedures based upon the prediction of accumulated
deformations in each component of the pavement system. Obviously, for a more advanced or
rational design method, this group is preferred due to the ability to compute cumulative
deformations of any pavement system. Substantial research and development efforts have been
conducted to make these approaches into rational design procedures. More information about
each category is presented below.
3.3.1 Limiting Subgrade Strain
In this approach, the pavement layers are made thick enough to limit the vertical
compressive strain at the top of the subgrade to a value associated with a specific number of load
repetitions, this strain being computed by means of a layered-elastic analysis. The logic of this
approach, first suggested by the Shell researcher (101), is based on the assumption that, for
materials used in the pavement structure, permanent strains are proportional to elastic strains.
Limiting the elastic strain to some prescribed value will also limit the plastic strain. Integration
of the permanent strains over the depth of the pavement section provides an indication of the rut
depth. By controlling the magnitude of the elastic strain at the subgrade surface, the magnitude
of the rut is controlled. An equation of the following form is used to relate the number of load
applications to vertical compressive strain at the subgrade surface:
b
⎛1 ⎞
N = A⎜⎜ ⎟⎟ (58)
⎝ εv ⎠
where:
N = number of load applications,

εv = elastic vertical strain at subgrade surface, and
A, b = empirically determined coefficients.
This approach has been quantified by the back-analysis of pavements with known
performance but is semi-empirical in nature since it applies to a particular range of structures
with particular materials under particular environmental conditions. Values of the coefficients
have been derived for different locations and circumstances. Brown and Brunton reviewed the
use of this semi-empirical criterion in 1984 (102). While they improved its application to allow
for varying rut resistance of different asphalt mixtures, they made it clear that the parameter is
only an indicator of the potential for critical rutting to develop as a consequence of permanent
deformations developing in all layers. A common misconception is that the subgrade strain
criterion only refers to permanent deformation in the subgrade. The relationships between
allowable strain and numbers of standard wheel loads were developed from linear elastic
back-analysis of structures with known performance in relation to rutting. The parameter is not,
therefore, fundamentally based and cannot be expected to provide reliable design guidance for
38
pavements with characteristics that differ significantly from those used in the back-analysis. In
1997 Brown critically reviewed this approach again, and concluded that it is out of date and
should be replaced (103).
3.3.2 Quasi-Elastic Approach (or layer-strain method)
A rather recent approach used for the prediction of permanent deformation is based upon
the use of elastic theory and the results of plastic strains determined by repeated load laboratory
tests on pavement materials. This approach is termed “quasi-elastic” due to the use of elastic
theory to predict a non-elastic response. The approach was initially introduced by Heukelom
and Klomp (104). Since then, research has been conducted by others such as Monismith (105),
McLean (106), Romain (107), Barksdale (108), and Morris and Hass (109) for soils, granular
materials, and asphalt concrete.
The fundamental concept of this approach is the assumption that the plastic strain εp is
functionally proportional to the elastic state of stress (or strain) and number of load repetitions.
This constitutive deformation law is considered applicable for any material type and at any point
within the pavement system. The response of any material must be experimentally determined
from laboratory tests for conditions (times, temperature, stress state, moisture, density, etc.)
expected to occur in situ.
Provided the plastic deformation response is known, elastic theory (either linear or
non-linear) is then used to determine the expected stress state within the pavement. By
subdividing each layer into convenient thickness (Δzj) and determining the average stress state at
each layer increment, the permanent deformation within the jth layer, δip may be found by
summing the (εip) * (Δzj) products. This process is done for each layer present in the pavement,
and the total permanent deformation is found from:
n
δ Tp = ∑ δ jp (59)
j =1
Obviously, such a summation process is done along a vertical axis (constant horizontal
plane coordinates). While different permanent deformation models have been proposed, only
four most widely promising models are presented: MEPDG rutting model (110), NCHRP 1-40B
rutting model (111), VESYS rutting model (112, 113), and WesTrack-shear-based rutting model
(114).
• MEPDG rutting model
The final MEPDG AC rutting model is presented below:
εp
= k1 × 10 − 3.4488 T 1.5606 N 0.479244 (60)
εr
where T is temperature (F), N is number of load repetitions, and k1 is depth adjustment
coefficient and defined as follows:
39
k1 = (C1 + C2 × depth ) × 0.328196depth (61)
2
C1 = −0.1039hac + 2.4868hac − 17.342 (62)
C 2 = 0.0172h ac 2 − 1.7331h ac + 27.428 (63)
• NCHRP 1-40B rutting model
NCHRP 1-40B rutting model has the same format as the MEPDG rutting model. The
enhancement is to adjust permanent deformation constants based on HMA volumetric properties.
εp
= k1 (10 k T k N k
r1 r2 r3
) (64)
εr
where k1 is depth adjustment function defined in the MEPDG rutting model. kr1, kr2, and kr3 are
material properties and defined below.
Constant kr1 is defined as follows:

[
k r1 = log 1.5093 ×10 −3 × K r1 × Va 0.5213 × Vbeff 1.0057 − 3.4488 ] (65)
where:
Vbeff = effective asphalt content in volume (%), and

Kr1= intercept coefficient (see Figure 12).
Figure 12. LogKr1 Coefficient vs. Voids Filled with Asphalt (%).
Constant kr2 is defined below:

0.25 1.25
⎛ Va ⎞ ⎛ Pb ⎞
k r 2 = 1.5606⎜ ⎟ ⎜ ⎟ Findex C index (66)
⎜ Va (design ) ⎟ ⎜ Pb (opt ) ⎟
⎝ ⎠ ⎝ ⎠
40
where:
Va(design) = design air voids,
Pb = asphalt content by weight,
Pb(opt) = design asphalt content by weight,
Findex = fine aggregate angularity index (Table 3), and
Cindex = coarse aggregate angularity index (Table 4).
Table 3. Fine Aggregate Angularity Index Used to Adjust Findex.

Fine Aggregate Angularity
Gradation – External to Restricted Zone
<45 >45
Dense Grading – External to Restricted Zone 1.00 0.90
Dense Grading – through Restricted Zone 1.05 1.00
Table 4. Coarse Aggregate Angularity Index Used to Adjust Cindex.

Type of Percent Crushed Material with Two Faces
Gradation 0 25 50 75 100
Well Graded 1.1 1.05 1.0 1.0 0.9
Gap Graded 1.2 1.1 1.05 1.0 0.9
41
Constant kr3 is presented below:
Pb
k r 3 = 0.4791× K r 3 × (67)
Pb (opt )
where:
Kr3 = slope coefficient:

Fine-graded mixes with GI<20 Kr3 = 0.40;
Coarse-graded mixes with 20<GI<40 Kr3 = 0.70;
with GI>40 Kr3 = 0.80.
# 50
GI = Gradation index = ∑ P − P(
i =3 / 8
i i 0.45 )
• VESYS rutting model
The VESYS rutting model is based on the assumption (or laboratory permanent
deformation law) that the permanent strain per loading pulse occurring in a material specimen
can be expressed by:
Δε p (N )
= μN −α (68)
ε
where:
∆εp(N) = vertical permanent strain at load repetition, N;

ε = peak haversine load strain for a load pulse of duration of 0.1 sec
measured on the 200th repetition; and
μ , α = material properties depending on stress state, temperature, etc.
The above equation assumes that ε remains relatively constant throughout the test, and
thus, the permanent strain increment, ∆εp(N), at any load cycle is:
Δε p (N ) = ε − ε r (N ) (69)
where εr(N) is the resilient or rebound strain taking place at cycle N. Then, the rut depth for any
single layer after N load cycles can be written as:
μ
RD = H × ε p = H × ε N 1−α (70)
1−α
where H is layer thickness.
The VESYS layer rutting model estimates the permanent deformation in each finite layer
as the product of the elastic compression in that layer and the layer material permanent
deformation law associated with that layer. The layer rutting model is expressed by:
42
N2 et n-1 N 2
RD= ∫ Us
+
μ sub N -α sub + ∑ ∫ (U i
+
- U i - ) μ i N -α i (71)
N1 es i=1 N 1
where:
Us + = deflection at top the subgrade due to single axle load,
Ui+, Ui- = deflection at top and bottom of finite layer i due to axle group,
et = strain at top of subgrade due to the axle group,
es = strain at top of subgrade due to a single axle,
μsub, αsub = permanent deformation parameters of the subgrade, and
μi, αi = permanent deformation parameters of layer i.
The major feature of the VESYS rutting model is to characterize layer properties rather
than global parameters used by MEPDG. For each layer, the VESYS rutting model requires
permanent deformation parameters: μ and αi.
• WesTrack shearing-strain rutting model
An alternative to the layer strain approach has been recently proposed to describe the
rutting behavior of the WesTrack test sections. In this approach, the pavement is modeled as a
multi-layered elastic system, with the asphalt concrete modulus determined from the repeated
simple shear test at constant height (RSST-CH) tests. Rutting in AC is assumed to be controlled
by shear deformations. Computed elastic shear stress and strain (τ, γe) at a depth of 50 mm
beneath the edge of the tire are used for rutting estimates. Densification of the asphalt concrete
is excluded in the rutting estimates since it has a comparatively small influence on surface
rutting.
In simple loading, permanent shear strain in the AC is assumed to accumulate according

to the following expression:
γ i = a × exp(bτ ) × γ e × n c (72)
where:
τ = shear stress determined at this depth using elastic analysis,

γe = corresponding elastic shear strain,
n = number of axle load repetitions, and
a, b, c = regression coefficients obtained from field data, RSST-CH laboratory
test data and the elastic simulations.
The time-hardening principle is used to estimate the accumulation of inelastic strains in

the asphalt concrete under in situ conditions. The resulting equations are as follows:
c
⎡ i ⎛⎜ 1 ⎞⎟ ⎤
⎢⎛⎜ γ j −1 ⎞⎟ ⎝ c ⎠ ⎥
γ j = aj ⎢
i
+ Δn j ⎥ (73)
⎜ aj ⎟
⎢⎝ ⎠ ⎥
⎣ ⎦
43
where:
j = the jth hour of trafficking,

γji = the elastic shear strain at the jth hour, and
Δn = the number of axle load repetitions applied during the j hour.
Rutting in AC layer due to the shear deformation is determined from the following:
RDAC = K ∗ γ ij (74)
For a 150 mm (6 in.) layer, the value of K is 5.5 where the rut depth (RD) is expressed in inches.
3.3.3 Rutting Accumulation Principle
To consider the effects of stresses of different magnitudes on the development of rutting,

which result from variations in traffic loads and environmental conditions, an accumulative
damage hypothesis is required, just as for fatigue. A “time-hardening” procedure appears to
provide a reasonable approach (114, 115).
For each season i, εip is computed from:
[
ε ip = ε ip (atN = 1) (N eqi + ni )S − N eqi
S
] (75)
where:
εip(at N=1) = permanent strain for i, first load repetition computed using the Vermeer
model,
ni = number of load repetitions during season i,
Neqi = equivalent total number of load repetitions at beginning of season i, and
S = slope of logεp –logN curve derived from test results.
The Neq is obtained for each element k with the time-hardening matching scheme as
follows:
N eq = 0
Season 1 (76)
ε 1p = ε 1p (N = 1)N 1S1
1
⎡ εp ⎤ S2
Season 2 N eq 2 = ⎢ p 1 ⎥ (77)
⎣ ε 2 ( N = 1) ⎦
[
ε 2p = ε 2p ( N = 1) (N eq 2 + n2 )S 2 − N eqS 22 ] (78)
1
⎡ εp ⎤ Sl
Season l N eql = ⎢ p l −1 ⎥ (79)
⎣ ε l ( N = 1) ⎦
44
[
ε lp = ε lp ( N = 1) (N eql + nl )S − N eql
l S l
] (80)
3.3.4 Comments on Rutting Models
• The layer-strain method is considered a reasonable approach for predicting rut depth,
at least for comparative purposes; it provides the added flexibility of allowing use of
either linear or nonlinear elastic theory.
• As noted by Brown et al., the development of permanent strain in pavement materials
is very difficult to predict from basic material characteristics because of the many
variables involved (116). This particular mechanical property, even more than
dynamic modulus or fatigue cracking resistance, requires a laboratory test to be
conducted on the material under consideration. This is because the aggregate
structure has a fundamental influence on the result. Consequently, details such as
particle surface characteristics, shape, and grading, together with packing and
orientation after compaction, are all influential and cannot reliably be predicted using
an empirical model.
For the present, it would be better to rely on good mixture design and testing to
limit rutting from the HMA layers and to use allowable stress criteria to deal with the
lower layers. Stress criteria could be based on accumulated research knowledge from
repeated load triaxial testing, which has identified “Threshold Stress” limits for many
materials (117).
• Regarding rutting prediction models, both MEPDG and NCHRP 1-40B rutting
models have specific parameters and do not need to run laboratory testing. While the
NCHRP 1-40B rutting model is an enhanced MEPDG model and considers many
more factors (e.g., asphalt binder content, angularity, gradation) influencing rutting,
asphalt binder PG grade (a parameter that most affects rutting of HMA pavement
based on accelerated load testing) is not directly considered in the NCHRP 1-40B
rutting model. It is worth noting that not requiring laboratory testing is both
advantageous and disadvantageous for these two models, because while it makes the
models simple to implement, not using laboratory characterization of HMA mixes
may lead to inaccurate rutting prediction. As noted above by Brown, HMA mixes are
very complex, and laboratory characterization of permanent deformation properties is
critical to adequately predict field rutting performance.
WesTrack shearing rutting model requires the RSST-CH to characterize

permanent deformation properties of HMA mixes and predict pavement rutting using
empirical shift factors. The feature of the WesTrack shearing rutting model is that
only the HMA layer located at 2 inches below the pavement surface, regardless of
how many HMA layers exist in the pavement structure, is required to be evaluated
under the RSST-CH. The disadvantages of the WesTrack shearing rutting model are:
• high variability of RSST-CH,

• very limited validation and calibration, and
• only UC-Berkeley uses this model.
45
The major feature of the VESYS rutting model is to characterize layer properties
rather than global parameters used by MEPDG. For each layer, the VESYS rutting
model requires permanent deformation parameters: αi and μi. Its disadvantage also is
acquiring these layer properties and running repeated load tests for each layer.
However, recognizing the complexity of HMA mixes, it is necessary to characterize
each HMA layer’s permanent deformation properties in order to make a more
accurate prediction.
46
CHAPTER 4
PERMANENT DEFORMATION MODELS FOR GRANULAR BASE AND

SUBGRADES
This chapter reviews permanent deformation models developed from previous research to
predict rutting in flexible pavement systems. Rutting generated from permanent deformation of
each layer is one of the major distresses associated with increasing roughness and reducing
overall serviceability. According to the Project 0-5798 proposal, there is a special concern in
dealing with thin flexible pavements. Permanent deformations of thin pavements are primarily
generated from the underlying materials. In addition, the proposal mentioned that while
calibrated permanent deformation models for the asphalt layer are relatively well-established,
models for underlying layers are not as developed. With this in mind, the researchers reviewed
existing permanent deformation models for underlying materials, with emphasis on identifying
models that are applicable for flexible pavements constructed under thin AC layers.
Based on a review of previous studies, Lekarp et al. identified the following factors that
need to be considered for predicting permanent deformation of unbound granular materials
(118):
• Effect of stress: Permanent deformation is directly related to deviator stress and

inversely related to confining pressure.
• Effect of principal stress reorientation: Permanent deformation increases with higher
principal stress reorientation.
• Effect of density, fines content, and aggregate type: Higher density generally gives less
permanent deformation development. Permanent deformation resistance is reduced as
the amount of fines increases. Angular aggregates provide better permanent deformation
resistance due to particle interlocking.
• Effect of number of load applications: Permanent deformation gradually accumulates
with the number of load applications. The rate of permanent deformation accumulation
varies with material properties and stress states as shown in Figure 13. Permanent
deformation can become unstable after the flow point.
• Effect of moisture content: Increasing moisture content leads to reduction of effective
stress because of excess pore water pressure resulting in loss of cohesion and bearing
capacity and generation of permanent deformations.
There are two representative approaches for computing rutting in flexible pavement
systems. The first approach is to calculate the permanent strain under the wheel load at different
depths along the same vertical line and to sum up the contributions of all layers. The other
approach is to compute the rate of rutting for each load application and integrate it over the
design life. To use these approaches, the appropriate permanent strains need to be estimated.
There are several models that have been proposed for predicting permanent strains. Table 5
summarizes models that can be used to predict permanent strains in materials underlying the
47
Figure 13. Typical Permanent Deformation Behavior (After MEPDG Supplemental
Documentation Appendix GG-1).
pavement surface. From Table 5, it is observed that permanent strains have been related to the
stress state, Mohr-Coulomb strength parameters, and the number of load applications,
respectively.
Recently, additional studies have been conducted to expand or improve model equations
through intensive laboratory and field test data. Ceratti et al. performed full-scale tests,
developed in a Brazilian pavement test facility, to investigate rutting of thin pavements (126).
They built five test sections with 1-inch seal coat over weathered basalt base materials, which
had three different thickness levels (6.3, 8.3, and 12.6 inches), two levels of strength index,
constructed on a 2-ft layer of clayey subgrade. Trafficking was simulated until 1-inch ruts
developed. Test results showed that the contribution of base rutting was most pronounced, and
the number of applications (N) for 1-inch rutting was related to axle load (L), pavement thickness
(T), and material strength index (IS). They obtained the following equation from the tests:
N = 10 5.25 L−3.97 T 4.12 IS 1.98 (81)
In addition, thicker pavements showed two distinct phases of rutting development: a rapid
growth at the beginning of the test, and after that a gradual rate of growth. In thinner pavements,
a third phase in which rutting grows exponentially was also identified. This finding indicates
that rutting of thin pavements accelerates after some number of applications due to weakened
structural capacity.
48
Table 5. Summary of Permanent Deformation Prediction Models.
Reference Model
σd ⎡⎛ σdRf ⎞ 1 ⎤
εa = / ⎢⎜⎜1 − ⎟ ⎥
⎣⎢⎝ 2(C cos φ + σ 3 sin φ ) ⎠ (1 − sin φ ) ⎦⎥
kσ 3n ⎟
εa = permanent axial strain

kσ 3 n = relationship defining the initial tangent
modulus as a function of confining pressure
Duncan and Chang (119)
(k and n are constants)
Rf = a constant relating compressive strength to an
asymptotic stress difference
C = cohesion
φ = angle of internal friction
σd = deviatoric stress
ε p = a + b log N
Barksdale (120)
εp = total permanent axial strain
N = number of load cycles
a and b = constants
ε p = IN S
εp = total permanent axial strain
Monismith et al. (121)
I and S = experimentally derived parameters
ε p (N )
= μN −α
εr
εp(N) = permanent strain due to single or Nth load
Kenis (122) application
εr = resilient strain at the 200th repetition
μ and α = VESYS parameters
⎛ σ d ⎞⎟
− 0 . 15
⎛σ ⎞ ⎛
⎜
⎛σ *⎞ ⎞
d ⎟⎟
⎜
ε p = ε 0 . 95 S • Ln ⎜ 1 − ⎜ d ⎟ + Ln (N )⎜
⎜ n / ⎜1 − m ⎜
d Sd ⎟ Sd ⎟ ⎜ S d ⎟ ⎟⎟
⎝ ⎠ ⎝ ⎠ ⎝ ⎝ ⎠⎠
Lentz and Baladi (123) Sd = static shear strength

ε0.95Sd = static strain at 95 percent of static strength
n = (0.809399+0.003769σ3)*10-4
m = 0.856355+0.049650Lnσ3
β
⎛ ρ⎞
ε a = ε 0 exp ⎜ − ⎟
Tseng and Lytton (124) ⎝ N⎠
ε0, β, and ρ = three parameter model constants
β
⎛σ ⎞ γ
ε p = α ⎜⎜ d ⎟
⎟ • N
⎝ Pa ⎠
Ullidtz (125) Pa = atmospheric pressure
α, β, and γ = constants
49
Park conducted permanent deformation testing with different base materials and clayey
subgrade in order to correlate material properties with the coefficients of the VESYS and three
parameter permanent deformation models (127). Consequently, he established equations to
estimate these parameters as functions of water content, dry density, confining and deviatoric
pressures, resilient modulus, first stress invariant, and total suction. He predicted rutting based
on an accelerated rutting model proposed by Lytton that considered moisture ingress into thin
low-volume pavements. Using liquefaction criteria and soil-water characteristic curves
expressed by Gardner’s equation, the VESYS α value was adjusted for saturated and unsaturated
conditions.
−k ( )t δ
(
θ ∂h ∂θ Ne ) ∂θ v
∂h
For saturated conditions, α = 1 − (82)

K (1 − n + θ )
(
fθ 2 ∂h )Ne
(
− k ∂h
∂θ
)t δ
v
For unsaturated conditions, α = 1 − ∂θ (83)

K (1 − n + θ )
∂h 1 Sγ w
= h (84)
∂θ 0.4343 γ d
where:
S = degree of saturation,
γw = unit weight of water,
γd = dry unit of weight of soils,
h = soil suction (cm),
θ = volumetric water content,
N = number of load repetitions,
k = Gardner’s unsaturated permeability,
tv = time between vehicles,
δ = a factor depending on the range of soil size (1/D102),
K = bulk stress of soils, and
f = unsaturated shear strength function (1/θ).
Generated rutting curves showed a rapid rut rate after around 100,000 load repetitions
due to the effect of moisture on the change of modulus.
The MEPDG uses a different set of equations to predict the parameters of the permanent
deformation model proposed by Tseng and Lytton (124). In their original work, the three
parameters were originally correlated with water content (Wc), deviatoric and bulk stresses, and
resilient modulus. Through calibrations of the model, NCHRP 1-37A (128) came up with a
revised set of equations, which eliminated the stress terms as given below. The guide
recommends that further calibrations be conducted for local practice.
50
β
⎛ρ⎞
εp ⎛ε ⎞ −⎜⎝ N ⎟⎠
= ⎜⎜ 0 ⎟⎟e
εr ⎝εr ⎠
⎛ε0
⎜⎜
(
⎞ 0.15e ρ
⎟⎟ =
β
)+ (20e ( ρ / 10 9 ) β
), (85)
⎝εr ⎠ 2
1
⎛ − 4.89285 ⎞ β
ρ = 10 9 ⎜⎜ ⎟ , and log β = −0.61119 − 0.017638W
9 β ⎟
( )
c
⎝ 1 − 10 ⎠
VESYS is a well-developed probabilistic and mechanistic flexible pavement analysis

computer program (http://www.volpe.dot.gov/sbir/sol03/docs/vesys-intro.doc). Two types of
rutting models are incorporated into the VESYS program, which are called the “layer rutting”
and “system rutting” models. The layer rutting model estimates the permanent deformation in
each finite layer as the product of the elastic compression in that layer and the layer material
permanent deformation law associated with the layer. The layer rutting model is expressed by:
N2 et n-1 N 2
RD= ∫U s
+
μ sub N -α sub + ∑ ∫ (U i
+
- U i - ) μ i N -α i (86)
N1 es i=1 N 1
where:
Us + = deflection at top of the subgrade due to a single axle load,

Ui+ and Ui- = deflection at top of and bottom of the finite layer i due to axle group,
et = strain at top of subgrade due to the axle group,
es = strain at top of subgrade due to a single axle load,
α sub, μsub = permanent deformation parameters of the subgrade, and
α i, μ i = permanent deformation parameters of layer i.
The system rutting formulation treats the pavement system as a whole and first calculates
an equivalent set of pavement system permanent deformation parameters, μsys and αsys, which
are determined as functions of load repetitions by least square regression analysis. Pavement
surface rut depth is estimated according to the equation:
N2
∫
R D = U μ sys N α sysdN
-
(87)
N1
where U is pavement surface deflection. Zhou and Scullion calibrated the VESYS rutting model
using the TxMLS data on US 281 (129). Rutting parameters (μ and α) were backcalculated and
used to calibrate the rutting parameters determined from laboratory tests conducted using the
protocol developed at TTI.
Uzan expanded the universal soil model to express the ratio of the accumulated
permanent strain to the resilient strain as a function of the bulk and octahedral shear stress for
granular materials as follows (130).
51
εp ⎡ ⎛ θ + k6 ⎞ ⎛τ ⎞⎤ ⎡ ⎛ θ + k6 ⎞ ⎛τ ⎞⎤
log = ⎢a 0 + a1 ⎜⎜ ⎟ + a 2 ⎜ oct ⎟⎥ + ⎢b0 + b1 ⎜ ⎟ + b2 ⎜ oct ⎟⎥ log N (88)
εr ⎟ ⎜ p ⎟ ⎜ p ⎟ ⎜ p ⎟
⎣⎢ ⎝ pa ⎠ ⎝ a ⎠⎦⎥ ⎣⎢ ⎝ a ⎠ ⎝ a ⎠⎦⎥
where a0 = −0.01891, a1 = −0.2514, a2 = 0.358983, b0 = 0.310932, b1 = −0.072703,

b2 = 0.098332, and k6 = 0.055 MPa.
He also proposed the equation below to predict the ratio of the accumulated permanent
strain to the resilient strain for clayey subgrades.
εp
log = −1.38164 + 0.324655 LogN (89)
εr
Zhou recently expanded Uzan’s model based on the HVS data from the cold regions
research lab including the water content term for A-2-4 and A-4 soils, respectively.
εp ⎛ θ ⎞ ⎛τ ⎞
log = 0.000851 − 0.246165⎜⎜ ⎟ + 3.418417⎜ oct + 1⎟ + 2.895124 log Wc + 0.16627 log N (90)
εr ⎟ ⎜ ⎟
⎝ pa ⎠ ⎝ pa ⎠
εp ⎛ θ ⎞ ⎛τ ⎞
log = 0.0112462 − 0.7195857⎜⎜ ⎟ + 3.8593287⎜ oct + 1⎟ + 0.6191976 log Wc + 0.2659706 log N (91)
εr ⎟ ⎜ ⎟
⎝ pa ⎠ ⎝ pa ⎠
Henning et al. presented LTPP studies in New Zealand (131). Since establishing 63
LTPP sites during 2000, they updated the database each year and currently manage 82 sites.
Using the LTPP and the Canterbury Accelerated Pavement Testing Indoor Facility (CAPTIF)
data, they tried to model pavement deterioration with respect to cracking and rutting. The
CAPTIF data was required since the LTPP data was limited in the number of sections that have
failed. From the developed rutting progression model data, the following findings were drawn:
• Three phases of rutting were identified: 1) initial densification; 2) rutting progression; 3)

initiation of accelerated rut progression. For strong pavements, only two phases, initial
densification and progression, were detected.
• There is no appropriate model to predict rutting progression at the current time.
• A linear logistic model was developed to predict the initiation of accelerated rut rate.
From analyses of the test data, researchers found that the initial densification (in mm) is
related to the structural number, SNP, according to the equation:
Rut initial = 3.5 + e ( 2.44 − 0.55SNP ) (92)
where SNP is the structural number derived from the falling weight deflectometer (FWD).
Researchers also found the rut progression in thin pavements (total layer thickness less than 6
inches) to be about 0.02 inches per year, and 0.01 inch per year for thick pavements. With
respect to modeling the initiation of accelerated rut progression, researchers came up with the
following equation:
52
1
Rut accel =
1+ e
( −7.568 *10
−6 [
* ESA + 2.434 * SNP − (4.426,0.4744 )for thickness = (0,1) ]
(93)
where ESA is equivalent standard axles. In the above, the thickness code is 0 for base layer
thickness less than 6 inches and 1 for thickness greater than 6 inches.
In thinner pavements, pavement maintenance in the form of surface overlays is

commonly performed to prevent further rutting from water intrusion. Mishalani and Kumar
studied the impact of overlays on pavement rutting with respect to design level and material
quality (132). They modified the rutting models proposed by Archilla and Madanat to take into
account the overlay effect on rutting development (133). The cumulative rut depth at a point in
time t is given by the summation of the incremental changes in each period s up to t as follows:
( )
t t
RDt = β 14 + ∑ a s exp(β 9 N s )ΔN s + ∑ m s exp β 13 N s' ΔN s' + μ + ε t (94)
s =1 s =1
a s = β 4 exp[− (β 1T1 + β 2T2 + β 3T3 )]exp[β 8 (TI s / 1000)] (95)

(γ 5 +γ 6 / GI )
⎛ VFA ⎞
m s = (γ 1 + γ 2 ⋅ GI ) + (γ 3 + γ 4 ⋅ GI )⎜ ⎟ (
⋅ TempDums + γ 8 AV exp γ 9 N s' ) (96)
⎝ 100 ⎠
where:
β14 = parameter that captures the initial rut depth immediately after
pavement construction,
β9, β13 = parameters capturing the hardening of the underlying and the AC layers,
respectively,
Ns = cumulative loading up to period s for rutting originating in the
underlying layers,
ΔNs = load applications during period s for rutting originating in the
underlying layers,
Ns’ = cumulative loading up to period s for rutting originating in the
AC layers,
ΔNs’ = load applications during period s for rutting originating in the
AC layers,
T1 = thickness of the AC layer,
T2 = thickness of the granular base layer,
T3 = thickness of the subbase layer,
TIs = thawing index during period s,
βi = model parameters (i = 1, 2, 3, 4, 8),
GI = gradation index of the AC mix,
VFA = voids filled with asphalt,
AV = in-place air voids,
TempDums = 1 if the mean maximum temperature during period s is greater than
28.6 °C, 0 otherwise,
γi = model parameters (i = 1, 2, 3, 4, 5, 6, 8, 9),
53
µ = time invariant random distance with zero mean and constant
variance reflecting unobserved heterogeneity across the pavement
sections, and
εt = error term with zero mean and constant variance reflecting
measurement errors and unobserved explanatory variables.
Overall rutting versus time was predicted using Equation 94 with three levels (low,
medium, and high) of design and material quality based on experimental design. Design levels
varied with thicknesses of each layer. The researchers selected three levels of GI, VFA, and AV
and noted the following findings:
• The rut predictions over the time varied with design level when overlays are applied
but eventually the rut depth converged to a constant value.
• The rut predictions over the time varied with material quality when overlays are
applied and the rut depth converged to different levels.
• The sensitivity in terms of absolute values of rutting to design levels decreased as
maintenance frequency increased.
• The sensitivity in terms of absolute values of rutting to material quality levels
increased as maintenance frequency increased, implying that for reaping the most
value from using a high material quality, a high maintenance frequency should be
adopted.
• The model needs to be further verified with field test data.
54
CHAPTER 5
FATIGUE CRACKING MODELS FOR CHEMICALLY STABILIZED
MATERIALS
This chapter reviews fatigue cracking models developed mainly from a PCA research
project (134) to predict fatigue cracking of chemically stabilized layers in flexible pavement
systems. In this chapter, the chemically stabilized materials are defined first before discussing
two fatigue models including the MEPDG and PCA-CTB (cement treated base) models. Finally,
the model input requirements and associated laboratory tests are also discussed.
5.1 DEFINITION OF CHEMICALLY STABILIZED MATERIALS

In both the MEPDG and the PCA-CTB programs, chemically stabilized layers are high
quality base materials that are treated with cement. These programs are intended for use with
“engineered” bases or sub-bases. An engineered base requires a formal laboratory design
procedure where both strength and durability criteria are achieved. Where a small amount of
chemical stabilizer is added to granular base materials to improve their strength, lower the
plasticity index, or decrease moisture susceptibility, this will not be considered an “engineered”
material unless a strength/durability test is performed. Without the use of strength and durability
criteria in the design process, the resulting bases should be considered unbound materials.
On the other hand, if these layers are engineered to provide structural support, then they
can be treated as chemically stabilized structural layers. To ensure durability and long-term
adequate performance of chemically stabilized materials, the MEPDG recommends the 7-day
UCS criteria shown in Table 6.
Table 6. Minimum Values of 7 Days Unconfined Compressive Strength, for Chemically

Stabilized Materials in the MEPDG.
Rigid pavements Flexible pavements
Base 500 psi 750 psi
Subbase, select material,
and subgrade 200 psi 250 psi
The numbers proposed in Table 6 are thought to be high, many DOT’s have recently
moved to designing stabilized bases with lower strength requirements. A common 7-day strength
requirement is 300 psi. In some DOT’s, 7-day strengths of between 200 and 250 psi have been
used with success. For the purpose of the PCA-CTB program, 7-day strength of lower than 250
psi may be used if the base also meets a moisture susceptibility requirement.
5.2 FATIGUE CRACKING MODELS FOR CHEMICALLY STABILIZED MATERIALS
Both the MEPDG and PCA-CTB models are presented as follows.

MEPDG Fatigue Cracking Models for Chemically Stabilized Materials
The fatigue relationship used in the MEPDG is a function of the stress ratio:
55
σt
(0.972βc1 − ( )
Mr
log N f = (97)
0.0825 * βc2
where
Nf = number of repetitions to fatigue cracking of the stabilized layer;
σt = maximum traffic induced tensile stress at the bottom of the stabilized layer
(psi);
Mr = 28-day modulus of rupture (Flexural Strength) (psi); and
βc1, βc2 = field calibration factors.
PCA-CTB Fatigue Cracking Models for Chemically Stabilized Materials
The PCA already have a fatigue relationship which they have used for many years to
design pavements containing cement treated bases. This relationship, which is also included in
the PCA-CTB program is also a function of the stress ratio but in an exponential form and is
shown below:
βc3 ⋅20
⎛ βc4 ⎞
Nf = ⎜ ⎟ (98)
⎝ σt Mr ⎠
where
βc3, βc4 = field calibration factor.
In a recently completed PCA study (134) Dr. Jacob Uzan calibrated these two models using the
accelerated pavement test data from PCA. He developed factors for two materials types: cement
treated base and fine graded soil cement. The fine graded soil cement would be equivalent to a
treated sub-base or subgrade layer. The final calibration factors for two types of cement treated
materials are presented below:
• For cement treated base:
βc1=1.0645, βc2=0.9003, βc3=1.0259, and βc4=1.1368
• For fine-grained soil cement:
βc1=1.8985, βc2=2.5580, βc3=0.6052, and βc4=2.1154
The PCA has evaluated these models and compared them with existing procedures. Both models
have been incorporated into simple design programs. The current plan is to incorporate both of
these design models into new PCA software to be released in mid 2008. The TTI research team
from study 5798 will review these models and other models developed for other stabilizers, and
consider them for incorporation into TexME.
5.3 MODEL INPUT REQUIREMENTS AND ASSOCIATED LABORATORY TESTING
The two major inputs required by the models are resilient modulus and modulus of
rupture of the chemically stabilized materials. Significant efforts have been made to evaluate
how to generate these two input parameters by Scullion under the PCA research project (134).
The final recommendation is presented as follows:
56
• Level 1 resilient modulus input: The standard resilient modulus test is not recommended
for routine use. The test is very difficult to run on cement treated materials. The strain
levels are very low, requiring accurate instrumentation. The biggest problem was that
even with careful sample preparation, problems are still encountered with the end
conditions - unlike other materials where a few seating loads will ensure good contact.
Seating loads do not ensure uniform contact with cement treated materials, where even
small unevenness of the surface causes major differences in strains measured on either
side of the test sample.
For the limited test program conducted by Scullion et al. (134), it appears that the seismic
modulus device is a better, more repeatable test for estimating the resilient modulus of
stabilized materials. However, samples for this test should have a minimum length to
diameter ratio of 1.5 to 1. The seismic modulus equipment used in the PCA study is
widely available within TxDOT; it was found in this study that the resilient modulus can
be estimated to be 75% of the measured seismic modulus.
Measuring Modulus of Rupture in the laboratory was performed in the PCA study
however this is also difficult particularly if the cement treated base has a relatively low
cement content. Level 1 testing was found to be very difficult with these materials.
• Level 2 resilient modulus input: The most attractive level for most users will be Level 2
where the design values are related to the standard 7-day UCS value. The recommended
relationships are given below:
For cement treated bases
28 day Modulus of Rupture (ksi) = 7.30* UCS (99)
Resilient Modulus (ksi) = 36.5* UCS (100)

For fine-grained soil cement
28 day Modulus of Rupture (ksi) = 6.32* UCS (101)
Resilient Modulus (ksi) = 31.6* UCS (102)

• Level 3 resilient modulus input: The default values are given below.
For cement treated bases
Resilient Modulus 1000 ksi, Modulus of Rupture 200 psi, Poisson’s Ratio 0.15
For fine-grained soil cement

Resilient Modulus 500 ksi, Modulus of Rupture 100 psi, Poisson’s Ratio 0.25
Note that the relationships and values proposed will be further reviewed in this study.
The level 2 approach seems reasonable. Necessary lab test will be conducted to evaluate these
UCS relationships for other chemical stabilizers. Additionally, the PCA software will also be
reviewed for consideration for use in Texas.
57
CHAPTER 6
REVIEW OF LABORATORY TESTING PROCEDURES
One critical part of any pavement design procedure is how to measure the material
properties that must be used as inputs to the design models. Many sophisticated models have
been developed, but few (or none) have been implemented for routine pavement design. One
difficulty is the high cost of obtaining the required rutting and cracking material properties for
each layer. DOTs run many tests to select the optimal asphalt content and gradation for HMA
layers, but these tests do not traditionally provide inputs to structural design models. In Texas,
the performance/acceptance test from HMA is the Hamburg wheel tracking test; more recently
the cracking potential of HMA layers is being assessed with an overlay tester. For granular base
materials, the different classes of base are routinely characterized by their triaxial strength, both
confined and unconfined after capillary soak. Up until now, the results of these tests have not
been used to provide inputs to thickness design programs.
In recent years, it has been proposed that multiple levels of material inputs may be
required for any new ME empirical design program. The research team believes that this
approach is reasonable for the future development of the Texas ME. Such an approach is
described below.
¾ Level 3 will be default values in look up tables for each specification item.
¾ Level 2 will be material properties inferred from TxDOT current design and acceptance
tests (Hamburg/Overlay tester), triaxial strength (Tex Method 117E or 143E), and
unconfined compressive strengths (Tex Method 121E). These will typically be
regression equations.
¾ Level 1 properties will be obtained from running advanced tests such as repeated load
triaxial, beam fatigue, etc.
In the remainder of this section, a description will be given of the different laboratory
tests available for characterizing the rutting and cracking potential of pavement layers. This will
be followed by recommendations on which tests to include in Year 2 of this project. These tests
will be run on materials obtained from test pavements being monitored to evaluate either
pavement response or pavement performance. The test pavements include the NCAT test track,
the recently completed HVS test track in California, the instrumented site being installed on SH
6 north of Calvert, and the thin instrumented site being constructed at Texas A&M’s Riverside
Campus.
6.1 ASPHALT RUTTING TESTING
HMA mixtures need to be designed to resist rutting (accumulation of permanent

deformation) under high tire contact pressures and from a large number of load repetitions.
Rutting is caused by a combination of densification (decrease in volume and AV) and shear
deformation (equal volume movement and increase in AV). For well-compacted HMA concrete
pavements, past research has indicated that shear deformation rather than densification was the
primary rutting mechanism. Resistance to permanent deformation or shearing stress has been
59
defined as a stability-related phenomenon. It is obvious that HMA mixtures must be designed
with adequate stability to ensure adequate performance. That is why stability is considered the
core aspect of HMA mixture design with respect to rutting.
Stability is affected by type/grade and amount of asphalt binder, aggregate properties

(such as absorption, texture, shape of particle), gradation, compaction level, and temperature.
Higher stability is promoted by using hard aggregates with rough surface textures, dense
gradations, comparatively low asphalt binder contents, harder (stiffer) asphalts, and
well-compacted mixtures as long as the air voids do not fall below a certain level.
In the past, at least three laboratory tests – Hubbard-Field (135), Marshall (136), and
Hveem tests (136) – have been used to characterize the stability of HMA mixtures. Necessary
minimum values for the measured stability have been established in different HMA mixture
design methods (135, 136) to ensure adequate pavement stability. The minimum value
established will, of course, depend on the type of stability test, weight and volume of traffic, and
other factors such as climatic condition, type of underlying structure, and thickness of surfacing.
Because of the uncertainties of these factors and doubts about how to measure true pavement
stability, there is, quite often, a tendency to design for maximum stability. Sometimes, this is
done at the detriment of other very important design factors, such as cracking resistance and
durability.
With the renewed interest in HMA mixture design generated by the SHRP, several new
laboratory tests have recently been developed to characterize the permanent deformation
properties of HMA mixtures. Sousa et al. made an excellent review of available permanent
deformation tests for HMA mixture, which is shown in Table 7 (137). During the SHRP, the
series of performance-based tests listed in Table 8 were also developed (138). Christensen et al.
summarized the latest development after the SHRP which is described in Table 9 (139).
In summary, permanent deformation tests have evolved from purely empirical tests
(Hubbard-Field, Marshall, and Hveem tests) through simulation tests (such as Hamburg Wheel
Track Test [HWTT], Asphalt Pavement Analyzer, French wheel tracking test) to more
fundamental tests, such as the repeated load test (140). Furthermore, the HMA rutting model
recommended in Chapter 3 also requires the repeated load test to determine model parameters.
Therefore, it is highly recommended to use repeated load test to characterize permanent
deformation properties of HMA mixes and then make rutting prediction using VESYS rutting
model shown below:
Δε p (N )
= μN −α
ε (103)
where:
∆εp(N) = vertical permanent strain at load repetition, N;

ε = peak haversine load strain for a load pulse duration of 0.1 sec
measured on the 200th repetition; and
μ, α = material properties depending on stress state, temperature, etc.
60
However, it should be noted that the repeated load test is currently not a routine test like
the HWTT. Therefore, the researchers recommended a three-level HMA characterization of
permanent deformation:
• Level 1: the repeated load test,

• Level 2: the HWTT, and
• Level 3: default values recommended based on catalogued values from previous
lab testing.
The repeated load test procedure for HMA materials is provided in Appendix B of this
report, and the HWTT procedure is well documented in the Tex-242-F. The key issue here is to
establish the relationship between the repeated load test and the HWTT test, and to develop a
methodology for extracting the rutting parameter (μ, α) from the HWTT test. Chapter 7 presents
more discussion and a detailed lab test plan.
61
Table 7. Comparison of Various Test Methods for Permanent Deformation Evaluation (137).
Field Overall
Test Method Sample Shape Measured Characteristics Advantages and Limitations Simplicity
Simulation Ranking
Diametral Creep modulus vs. time Easy to implement
Static (creep) Permanent deformation vs. time Field cores can be easily obtained
Diametral Resilient modulus Shear stress field not uniform
Repeated 4 in diameter Permanent deformation vs. cycles State of stress is predominantly tension
3 1 2
2.5 in high Equipment is relatively simple in static test
Dynamic modulus
Diametral For repeated and dynamic tests, the complexity of
Damping ratio
Dynamic the equipment is similar to that of triaxial repeated
Permanent deformation vs. cycles
and dynamic equipment.
Almost all states of stress can be duplicated.
Dynamic axial modulus
1 in wall Capability of determining damping as a function of Not
Dynamic shear modulus
thickness frequency for different temperatures for shear as suitable
Hollow Axial damping ratio
18 in high well as axial 1 3 for
Cylindrical Shear damping ratio
9 in external Sample preparation is tedious. routine
Axial permanent deformation vs. cycles
diameter Expensive equipment use
Shear permanent deformation vs. cycles
Cores cannot be obtained from pavement.
Shear stress can be directly applied to the specimen.
Simple Shear Shear creep modulus vs. time
Cores can be easily obtained from existing
Static (Creep) Shear permanent deformation vs. time
pavement.
Simple Shear 4 in diameter Shear resilient modulus
Better expresses traffic conditions 2 2 1
62
Repeated 2.5 in high Shear permanent deformation vs. cycles
Shear dynamic modulus Capability of determining the damping as a function
Simple Shear
Damping ratio of frequency for different temperatures
Dynamic
Shear permanent deformation vs. cycles Equipment not generally available
Table 8. Performance Test Included in the Superpave Mixture Analysis System (138).
Distress Test Device
Frequency sweep at constant height
Repeated shear at constant stress
ratio
Permanent Deformation Superpave Shear Tester
Simpler shear
Uniaxial strain
Volumetric
Frequency sweep at constant height Superpave Shear Tester
Fatigue Cracking
Tensile strength Indirect Tensile Tester
Creep compliance
Low-temperature Cracking Indirect Tensile Tester
Tensile strength
Table 9. Summary of Post SHRP Permanent Deformation Testing Research (139).
Demonstrated Correlation
Test Device Criteria Advantages Disadvantages
with Measured Rutting
Dynamic Shear Superpave shear Preliminary based on Applicable to field cores Specimen preparation
High
Modulus tester sensitivity analysis and lab specimens Equipment cost and complexity
Applicable to field cores Specimen preparation
Repeated Shear Superpave shear Preliminary based on and lab specimens Equipment cost and complexity
High
Constant Height tester sensitivity analysis Stimulates traffic loading Available permanent deformation
Large deformation test model not widely accepted
Dynamic Modulus Compatible with the AASHTO
Simple performance Specimen preparation
High MEPDG MEPDG rutting model
test system Cannot test field cores
Active equipment development
Simulates traffic loading
Wide range of stress states
Flow Number High None possible
test system Cannot test field cores
Large deformation test
Very simple test
Wide range of stress states
High None possible
63
Flow Time test system Cannot test field cores
Large deformation test
High Temperature Superpave Gyratory
Potentially high based on Preliminary repeated Uses existing equipment
Indirect Tensile Compactor plus
correlation with repeated shear constant height Applicable to quality control Requires field verification
Strength AASHO T283 indirect
shear constant height test criteria testing
plus Compaction Slope tension test
Requires shear force measurement
Superpave Gyratory
Gyratory Shear To identify unstable Results available after capability
compactor with shear Fair
Resistance mixtures compaction Can only identify mixture
force capability
instability
Uses existing equipment
Superpave Gyratory Potentially high based on
Preliminary based on and low-cost indenter
Rapid Performance Compactor with correlation with Requires field verification
sensitivity analysis Applicable to quality control
indenter repeated load test
testing
Equipment cost
Asphalt pavement Location, facility,
High Intuitive test Extensive calibration to establish
analyzer and mix specific
local criteria
Wheel Tracking
Very simple test and applicable Equipment cost
Hamburg wheel
High Yes to Extensive calibration to establish
tracking test
field cores and lab specimens local criteria
6.2 ASPHALT CRACKING TESTING
Since the late 1950s, many efforts have been made to evaluate the fatigue cracking
resistance of HMA mixes. Table 10 presents the most often used laboratory tests to characterize
fatigue properties of HMA mixes (141). During the SHRP, UC-Berkeley chose the flexural
beam fatigue test for characterizing the fatigue properties of HMA mixes after a comprehensive
comparison among the flexural beam test, flexural cantilever test, and repeated diametral test.
Later, the flexural beam fatigue test was adopted by AASHTO (AASHTO T321) as a
performance test for evaluating fatigue cracking resistance of HMA mixes. Even today, the
flexural beam fatigue test still is the most widely used test for fatigue characterization of HMA
mixes.
In addition to the flexural beam fatigue test, alternative methods were developed for
evaluating fatigue cracking resistance. NCHRP 9-19 evaluated the dynamic modulus test and
IDT creep compliance test for characterizing fatigue resistance of HMA mixes (142). However,
there are not enough data published to support the conclusion that fatigue cracking is related to
dynamic modulus or creep compliance. Another significant effort was to use continuum damage
theory to analyze fatigue test results. With the leading research conducted at Texas A&M
University (143) and North Carolina State University (144), the continuum damage theory was
applied to the fatigue analysis of HMA mixes. Laboratory test protocols have been proposed.
However, sophisticated data analysis prohibits this approach from being used routinely for
characterizing fatigue resistance of HMA mixes. Further simplification is absolutely necessary
to this advanced approach of evaluating fatigue properties.
The latest development for fatigue cracking evaluation was the overlay tester-based
fatigue cracking characterization approach proposed by Zhou et al. (145). The overlay tester-
based fatigue cracking characterization and prediction approach was developed based on fracture
mechanics. The fundamental HMA fracture properties (A and n) determined from the overlay
tester were used to estimate fatigue life of asphalt pavements. However, not only is the fatigue
crack propagation considered, but the crack initiation (the number of load repetitions required to
form a macro-crack from micro-cracks) is also included in this approach. Therefore, the overlay
tester-based approach is different from traditional fatigue crack approaches (such as the flexural
beam fatigue-based models used in the MEPDG) in which the fatigue crack propagation stage is
not directly considered; it is also different from the traditional fracture mechanics approaches in
which the fatigue crack initiation stage is often ignored. This overlay tester-based approach is
not only a more fundamental test-based approach, but it can also be easily implemented in
DOTs. Furthermore, the approach has been validated through analyzing the FHWA-ALF fatigue
tests (145).
The overlay tester-based approach has the advantage over other crack tests-based
approaches such as the flexural beam fatigue test because it allows for evaluation of either
laboratory-molded gyratory samples or field cores. Furthermore, the overlay test takes less than
1 hour to run for traditional mixes. However, beam tests have been around for many years and
have been restricted to research institutions, and they have never been implemented as routine
design tests in DOTs. Additionally, beam tests use long samples that are difficult to fabricate or
extract from the field. The beam test itself can take many hours to run, and the test equipment is
64
very expensive. Therefore, the overlay tester is recommended for fatigue cracking
characterization.
The overlay tester procedure is described in Tex Method 248-F. A step-by-step

procedure for determining fracture properties (A and n) has been presented in Appendix A.
Additionally, Chapter 2 presents the overlay tester-based fatigue cracking prediction approach.
To provide input for the design equations, it will be necessary to run tests at three temperatures:
77°F, 59°F, and 41°F. HMA fracture properties at other temperatures can be interpolated based
on those at the three temperatures. A more detailed testing plan for the overlay tester is
presented in Chapter 7.
65
Table 10. Comparison of Test Methods for Cracking (141).
Simulation of
Application of Disadvantages and Overall
Method Advantages Field Simplicity
Test Results Limitations Ranking
Conditions
Well known, widespread

Basic technique can be used
Costly, time
for different concepts
Repeated Yes consuming
Results can be used directly 2 4 I
Flexure Test σb or εb ,Smix Specialized
in design
equipment needed
Options of controlled stress
or strain.
In the LCPC
methodology:
The correlations
based on one million
Need for conducting fatigue
Yes (through repetitions
Direct Tension tests is eliminated
correction) Temperature only at 7 1 I
Test Correlations exist with
σb or εb ,Smix 10ºC.
fatigue test results
Use of thickness of
bituminous layer for
1 million repetitions
only.
Simple in nature Biaxial stress state

Diametral
Yes Same equipment can be used Underestimates
Repeated Load 4 2 II
4σb and Smix for other tests fatigue life
Test
Tool to predict cracking
Accurate prediction
requires extensive
fatigue test data
Based on a physical
Dissipated Simplified
Φ, ψ, Smix and σb phenomenon
Energy procedures provide 3 5 III
or εb Unique relation between
Method only a general
dissipated energy and N
indication of the
magnitude of the
fatigue life.
At high temp., KI is
not a material
Strong theory for low constant.
Yes
Fracture temperature Large amount of
KI, Smix curve (a/h
Mechanics In principle the need for experimental data 5 7 IV
- N); calibration
Tests conducting fatigue tests needed
function (also KII)
eliminated. Only stable crack
propagation is
accounted for.
Compared to direct
Repeated
tension test, this is
Tension or
Yes Need for flexural fatigue tests time consuming,
Tension and 6 3 V
σb or εb ,Smix eliminated costly, and special
Compression
equipment is
Test
required.
Traxial
Costly and special
Repeated
Yes Relatively better simulation equipment is required
Tension and 1 6 VI
σd, σc, Smix of field conditions Imposition of shear
Compression
strains is required
Test
Note: the lower number, the better ranking.
66
6.3 BASE AND SUBGRADE TESTING
Based on the review of permanent deformation models, researchers propose to

investigate the application of the layer-strain approach to model the permanent deformation
behavior of thin pavements. This mechanistic-empirical approach provides a general method for
predicting rutting and will enable researchers to develop a unified framework for predicting
permanent deformation in flexible pavements (both hot-mix and thin-surface pavements). The
same approach is used in the MEPDG program developed from NCHRP Project 1-37A and in
the Texas VESYS program.
For the purpose of this investigation, thin pavements are considered to cover the
pavement structures shown in Figure 14, where the primary load bearing layer is the flexible
base. Using the layer-strain approach to predict the development of rutting in these pavements
will require characterizing the permanent deformation behavior of the flexible base and subgrade
materials comprising these pavements. This behavior is generally characterized using the
two-parameter VESYS model or the three-parameter model illustrated in Figure 15, where the
model parameters are determined using data from repeated load tests conducted on material
samples. The MEPDG program uses the three-parameter model for predicting permanent
deformation in flexible pavements. This same model was used in a recent project to verify the
load-thickness design curves in TxDOT Test Method Tex-117E. In that project, TTI researchers
used the three-parameter model to evaluate relationships between permanent deformation and
applied load based on plate bearing test data collected on field test sections. In these tests,
researchers monitored the displacements during the loading and unloading stages of the step
loads applied to the pavement sections. The three-parameter model was found to provide a good
fit to the unrecovered deformation data from plate bearing tests conducted under the different
step loads. Thus, researchers plan to evaluate this model, as well as the two-parameter model
used in the Texas VESYS program to investigate the permanent deformation behavior of thin
pavements. The results of this work will be useful in developing design guidelines for thin
pavements that supplement the existing triaxial design check with criteria based on the
development of permanent deformation under repeated loading. For this development,
researchers propose to conduct laboratory tests on selected base and subgrade materials
commonly found on thin pavements in Texas.
Figure 16 illustrates the laboratory test program for evaluating relationships and test
methods to predict rutting under repeated loading. This test program has the following major
elements:
• characterization of Mohr-Coulomb strength parameters based on the triaxial test
method proposed in the provisional TxDOT Test Method Tex-143E,
• characterization of permanent deformation behavior based on triaxial tests of
material samples under repeated loading, and
• evaluation of an alternate method of triaxial testing to investigate if a simpler test
procedure can be identified for characterizing strength properties and permanent
deformation behavior.
67
Figure 14. Thin Pavement Structures Considered in Test Plan.
Figure 15. Permanent Strain Development under Repeated Loading.
68
The test program recognizes that a simpler alternative to the repeated load permanent
deformation test would facilitate the implementation of design guidelines for thin pavements that
are based on the mechanism of rut development under repeated loading. It also recognizes that
the triaxial design check will likely continue to be used within TxDOT for flexible pavement
design. Thus, the proposed laboratory test program includes work to investigate an alternate
triaxial test, which is like the conventional geotechnical triaxial test except that the specimen is
subjected to a load and recovery cycle at the start of the test as illustrated in Figure 16.
Researchers plan to analyze the creep and recovery data from the load and recovery cycle to
investigate resilient and permanent deformation properties of the materials tested. These same
properties will then be compared against corresponding properties determined from repeated
load permanent deformation tests to assess the applicability of the alternate triaxial test for
characterizing properties needed to predict rutting on thin pavements. In particular, the load and
recovery cycle will provide data for evaluating the slope of the creep compliance curve, which
has been related (from theoretical considerations) to the parameter s of the VESYS model (see
Figure 15). In addition, the recovery data will provide an estimate of the amount of deformation
that remains after the specimen is unloaded, which may prove useful to evaluate relationships for
determining the intercept I of the VESYS model. Note that this parameter physically
corresponds to the permanent deformation at the first load cycle (N=1) and might prove useful in
a material specification for thin pavements to assess material durability on the basis of its
propensity to rut.
As Figure 16 indicates, researchers also plan on comparing the cohesion and friction
angles determined from alternate triaxial tests with corresponding parameters from tests based on
Tex-143E. By comparing test results with data from existing test procedures, an assessment will
be made of the applicability of using the alternate triaxial test to characterize material properties
for designing thin pavements on the basis of triaxial design and repeated load permanent
deformation criteria. Table 11 summarizes the proposed laboratory test program, which covers
tests on two flexible bases and two subgrade materials commonly found on thin pavements built
in Texas. These materials will be taken from the construction on SH 6 north of Calvert, which
has a Class 1 base on a sandy subgrade and from the experimental pavement constructed at the
TTI Riverside Campus, which has a clay subgrade and Grade 2 base material. The proposed
program includes tests to characterize resilient modulus. Researchers plan to run these tests on
the same specimens prepared for permanent deformation testing with the specimens first tested
for resilient modulus and then for permanent deformation behavior.
In terms of developing design guidelines for thin pavements, this project will also need
field test data to verify and calibrate rutting models, such as those used in the MEPDG and Texas
VESYS programs. For this purpose, researchers plan on reviewing and assembling relevant data
from accelerated pavement tests, from in-service surface-treated pavements, and from
instrumented field sections built and tested in this project. In this regard, researchers have built a
number of surface-treated and hot-mix asphalt pavement sections at the Riverside Campus, as
well as instrumented sections of ongoing construction projects to investigate the response of
flexible pavements under repeated loading. Researchers note that these field tests are not
intended to verify predictions of pavement service life as this verification is not realistic within
the time frame of this project. Rather, researchers plan to use the test data from instrumented
69
Figure 16. Proposed Laboratory Test Program for Thin Pavements.
70
field sections to verify predictions of pavement response, assess the relative contributions of the
different layers to the total pavement deformation, and rank the test sections with respect to
rutting potential. Thus, pavement response and rutting models for thin pavements will be
verified based on these factors.
With respect to verifying predictions of pavement life and calibrating the permanent
deformation model, researchers plan to use available data from accelerated pavement tests and
in-service pavements identified from ongoing TxDOT database development projects. The
extent to which this task can be accomplished will depend on the availability of good
performance data on thin pavements that have gone through or are close to the end of their life
cycles.
Table 11. Proposed Laboratory Test Program for Thin Pavements.
Triaxial Test Moisture-

Permanent Resilient Atterberg Soil Dielectric
Material density
Deformation Modulus Limits Suction Test
Standard Alternate Curve
Grade 1
Crushed Τ Τ Τ Τ Τ Τ Τ Τ
Limestone
Grade 2
Crushed Τ Τ Τ Τ Τ Τ Τ Τ
Limestone
Clay Τ Τ Τ Τ Τ Τ Τ Τ
Sand Τ Τ Τ Τ Τ Τ Τ Τ
71
CHAPTER 7
PROPOSED TESTING PROGRAM FOR YEAR 2 OF PROJECT 0-5798
Chapters 2, 3, 4, and 5 presented an overview of the performance models that are

available for consideration for inclusion in the Tex-ME. However, as described, many of the
models, particularly the advanced cracking models, are many years away from practical
implementation. Given the scope of this project, the research team recognized the importance of
obtaining material properties from existing TxDOT laboratory tests. After a realistic
appreciation of the complexity of merging test procedures used with pavement design as
compared with those required for thickness design, the research team strongly believes that the
most appropriate model for predicting the rutting in flexible pavement systems is a VESYS-type
model in which the repeated load test is required to determine HMA permanent deformation
properties. For HMA layers, the HWTT will be explored as an alternative test for evaluating
HMA permanent deformation properties. Meanwhile, the most practical fatigue cracking model
is the Overlay Tester-based fatigue model in which the Overlay Tester is used to determine
HMA fracture properties. This chapter provides the proposed laboratory testing program to
determine materials input parameters including modulus, permanent deformation, and fatigue (or
fracture) properties.
7.1 LABORATORY TEST PROGRAM FOR HMA MIXES

7.1.1 The Dynamic Modulus Test
Dynamic Modulus (DM) testing is an AASHTO standardized test method for
characterizing the viscoelastic properties of asphalt mixtures, measured in terms of the complex
modulus |E*|. A typical DM test is often performed over five different temperatures of 14, 40,
70, 100, and 130°F and six loading frequencies of 25, 10, 5, 1, 0.5, and 0.1 Hz for each test
temperature, respectively. DM is a stress-controlled test involving application of a repetitive
sinusoidal dynamic compressive-axial load (stress) to an unconfined specimen. Figure 17 shows
TTI’s Universal Testing Machine (UTM-25) setup that was used for conducting the DM test and
includes the loading configuration and test specimens.
The standard DM test specimen is cylindrically shaped with dimensions of 4 in. diameter
(φ) by 6 in. height (h). However, for most of the field cores with thinner layers (< 6 in.),
prismatic specimens as shown in Figure 17 were used. These prismatic specimens were cut
consistent with the procedure suggested by Dr. Jacob Uzan under the Report 0-4822-1 (146).
The minimum specimen dimensions were 2 in. breadth by 2 in. width by 5 in. in length to ensure
at least a minimum 1.5 aspect ratio and coverage of the nominal aggregate size. The stress level
for conducting the DM test was chosen to maintain the measured resilient strain (recoverable)
within 50 to 150 microstrain consistent with the AASHTO TP 62-03 test protocol (147). The
order for conducting each test sequence was from the lowest to the highest temperature and the
highest to the lowest loading frequency at each temperature to minimize specimen damage. For
each test sequence, the test terminates automatically when a preset number of load cycles have
been reached. During DM testing, the measurable parameters include the applied load (stress),
loading frequency, temperature, vertical axial deformations, phase angle, and the dynamic
modulus.
73
Figure 17. DM Test Setup and Test Specimens (Cylindrical and Prismatic).
7.1.2 The Repeated Load Permanent Deformation Test
The Repeated Load Permanent Deformation (RLPD) test is often used to characterize the
permanent deformation properties of HMA mixtures under repeated compressive haversine
loading. The detailed test procedure is presented in Appendix B.
RLPD is a stress-controlled test involving repetitive application of a haversine-shaped

compressive-axial load (stress) to an unconfined specimen, at a frequency of 1 Hz with 0.1 s
loading time and 0.9 s rest period, respectively, for up to 5000 load cycles. The RLPD tests are
often conducted at three test temperatures (77, 104, and 122°F). As with the DM test, TTI’s
UTM-25 setup was used for conducting the RLPD testing, and the loading configuration is
shown in Figure 18. During RLPD testing, the measurable parameters include the applied load
(stress), test temperature, time, number of load cycles, axial permanent deformation, and strains.
From a plot of the accumulative axial permanent microstrain versus load repetitions on a
log-log scale, permanent deformation parameters εr, a, b, alpha (α), and gnu (μ) were determined
consistent with the procedure described by Zhou and Scullion (129). These parameters constitute
the VESYS5 rutting input parameters (μ and α) for asphalt mixtures and are defined as follows:
• εr = axial resilient microstrain measured at the 100th load cycle,

• a and b = intercept and slope of the linear portion of the permanent microstrain
curve (log-log scale),
• alpha (α) = rutting parameter computed as α = 1 − b , and
• gnu (μ) = rutting parameter computed as μ = ab .
εr
74
Loading parameters:
Stress levels = 30 & 20 psi
Loading time = 0.1 s
Rest period = 0.9 s
Figure 18. RLPD Loading Configuration.
7.1.3 The Hamburg Wheel Tracking Test
The HWTT is a test device used for characterizing the rutting resistance of asphalt
mixtures in the laboratory including a stripping susceptibility assessment (moisture damage
potential). The loading configuration consists of a repetitive passing load of 158 lb-force
(705 N) at a wheel speed of 52 passes per minute and a test temperature of 122°F in a controlled
water bath. The HWTT test specimens are 2.5 in. thick by 6 in. diameter, with one trimmed edge.
Figure 19 shows the Hamburg test device with a specimen setup. During HWTT testing, the
measurable parameters include the applied load, temperature, number of load passes, and vertical
permanent deformation (rutting). The research team will evaluate the potential relationship
between the RLPD test and the Hamburg test through extensive laboratory testing.
Figure 19. The Hamburg Test Device and Test Specimen.
7.1.4 The Overlay Tester

75
The overlay tester is a simple performance test for characterizing HMA cracking
resistance. The test loading configuration consists of a cyclic triangular displacement-controlled
waveform at a maximum horizontal displacement of 0.025 in. and a loading rate of 10 s per cycle
(5 s loading and 5 s unloading). Typical overlay test specimens are 6 in. total length, 3 in. wide,
and 1.5 in. thick that can be conveniently cut by trimming a lab-molded Superpave gyratory
compactor (SGC) specimen or a 6 in. diameter highway core. Figure 20 shows the overlay test
setup and a test specimen. During the overlay testing, the measurable parameters include the
applied load, opening displacement, time, number of load cycles, and the test temperature.
Details of the overlay test for HMA fracture properties (A and n) are presented in Appendix A.
Generally, the overlay test is conducted at 77ºF. However, in this project researchers proposed
to run the overlay test at the following three temperatures: 77, 59, and 41ºF.
Figure 20. The Overlay Tester and Specimen Setup.
Proposed Laboratory Testing Program for HMA Mixes
The proposed laboratory testing program for HMA mixes is shown in Table 11. The
objectives of the laboratory testing are four-fold:
• Finalize the laboratory testing protocols proposed previously.

• Develop the default values for HMA engineering properties including modulus,
permanent deformation, and fracture properties.
• Establish the relationships between Level 1 and Level 2 inputs.
• Provide input parameters for calibrating models developed in this project.
In addition, the research team is working with NCAT and University of California at Davis
to test cores from their APT sites for performance model calibration. It is envisioned that the
same tests as listed in Table 12 will be conducted on those cores from both APT sites.
76
Table 12. Proposed Laboratory Testing Program for HMA Mixes.
E* Cracking Rutting
Binder
Mix Type Locations Aggregate Type 5 temp. OT RLPD Hamburg
Type
6 freq 3 temp. 3 temp. 1 temp.
SMA-C Limestone
SH 114 PG 70-28 Τ Τ Τ Τ
SMA-C IH 35 Traprock + gravel PG76-22 Τ Τ Τ Τ
SH 114 Limestone PG 70-22 Τ Τ Τ Τ
Superpave A PG 76-22 Τ Τ Τ Τ
IH 35-Laredo Traprock/gravel
PG 64-22 Τ Τ Τ Τ
Type B
IH 35-Waco Limestone PG 64-22 Τ Τ Τ Τ
Superpave B PG 76-22 Τ Τ Τ Τ
IH 35-Laredo Traprock/gravel
PG 70-22 Τ Τ Τ Τ
PG 70-22 Τ Τ Τ Τ
SH 114 Limestone
Type C PG 64-22 Τ Τ Τ Τ
SH 6 Gravel PG 64-22 Τ Τ Τ Τ
IH 35-Laredo Gravel PG 70-22 Τ Τ Τ Τ
Superpave C IH 35-San
Limestone PG 64-22 Τ Τ Τ Τ
Antonio
Riverside Limestone PG 76-22 Τ Τ Τ Τ
Type D
SH 12 Granite PG 76-22 Τ Τ Τ Τ
7.2 LABORATORY TESTING PROGRAM FOR BASE/SUBGRADE MATERIALS
The proposed method for characterizing the permanent deformation behavior of granular
materials and subgrades is provided in Appendix C. However, the research team feels that it is
critical to evaluate this procedure to ensure that representative samples are being tested and that
realistic permanent deformation properties are being obtained. There has been much discussion
on several issues, namely:
• the size of the sample–6 by 8 in. as opposed to 6 by 12 in.,

• the method of compacting samples–drop hammer or vibratory compactor,
• method of mounting sensors on sample to avoid slippage,
• the range of stress combinations needed for resilient modulus testing,
• integration of resilient modulus and permanent deformation testing,
• sample conditioning prior to testing (moisture and load conditioning), and
• problems associated with laboratory testing of low fines bases.
Substantial work has already been performed in each of these areas. For example,
Project 0-4358 measured resilient modulus at different sample sizes, and Project 0-3512
examined the differences between drop hammer versus vibratory compaction for base and
subgrade materials. The research team is well aware that TxDOT procedures for acceptance
have been in place for many years, and they will not be changed in the foreseeable future. The
research team is also aware that this level of testing on granular materials requires complex
equipment so that it is not anticipated that non-TxDOT standard test methods will be run for
77
routine pavement design purposes. For these materials, a correlation will be needed between the
routine test results and the required model inputs.
Prior to executing the test plan for thin pavements presented in Section 6.3, researchers
propose to conduct a comparative evaluation of permanent deformation test procedures to
establish the most appropriate methods for compacting the specimens and mounting the LVDTs
for the proposed permanent deformation tests in Section 6.3. Table 13 shows the test matrix for
this comparative evaluation. Sebesta, Harris, and Liu conducted permanent deformation tests on
granular base specimens prepared by impact hammer and vibratory compaction methods (148).
They found that permanent deformation curves of base materials varied significantly, as
illustrated in Figures 21 and 22, with vibratory-compacted specimens developing less permanent
deformation under repeated loading as compared to specimens of the same materials compacted
using the conventional impact hammer method. Other researchers have also reported problems
with slippage and rotation of LVDT holders pressed onto the specimen using clamps (149). In
view of these findings from reported research studies, the first phase of permanent deformation
tests in Year 2 of this project will cover the comparative tests presented in Table 13 to identify
the best approach for compacting and instrumenting the specimens. As proposed, two methods
of compaction, vibratory and impact hammer, will be compared, as well as two methods of
mounting the LVDTs—studs versus the traditional method of using rubber bands to press the
LVDT holders onto the specimen for testing. For these tests, researchers will prepare specimens
using samples of the base materials placed at the Annex test site and on the SH 6 project north of
Calvert in the Bryan District. Consistent with current TxDOT practice, researchers propose to
mold 6-inch diameter by 8-inch high specimens for testing, with specimens compacted at
optimum moisture contents. For these tests, researchers plan on using the permanent deformation
test protocols reported by Zhou and Scullion as a guide for running the comparative laboratory
tests proposed herein (129). Appendix C of this report presents these test protocols.
Table 13. Test Matrix for Comparative Evaluation of Permanent Deformation Setups.
Method of Mounting LVDTs
Compaction LVDT Holders Pressed LVDT Holders
Base Material
Method onto Specimen with Anchored to Specimen
Rubber Bands with Studs
Impact hammer Τ Τ
Annex
Vibratory Τ Τ
Impact hammer Τ Τ
SH 6
Vibratory Τ Τ
78
Figure 21. Permanent Deformation Curves from Tests on Grade 1 Spicewood Base
Specimens (148).
Figure 22. Permanent Deformation Curves from Tests on Grade 2 Groesbeck Base
Specimens (148).
Permanent deformation properties will be determined from the test data, and differences
in permanent deformation behavior will be examined to establish the test setup that provides the
79
most reasonable results. Researchers plan on examining the permanent deformation traces to
check for evidence of slippage. If slippage occurs, the specimen deformation will not be fully
transmitted to the LVDT resulting in lower resilient strains measured from the LVDT and higher
values of the permanent deformation parameter μ of the VESYS model. Changes in the
orientation of the LVDT holders will also be monitored to establish the effectiveness of studs at
minimizing the rotation of the LVDTs during testing. In addition, researchers propose to
perform Computed Tomography (CT) scans on the specimens to check for uniformity of
compaction, particularly at the vicinities of studs on specimens where these anchors for the
LVDT holders are used, and to compare these with corresponding scans obtained from
specimens tested where the LVDTs are mounted in the conventional manner. By examining and
comparing the test results from the different setups considered in the test matrix (Table 13), a
determination will be made as to the most appropriate test setup for running the permanent
deformation tests proposed in Section 6.3.
80
REFERENCES
1. Suresh, S., (1991). Fatigue of Materials. Cambridge University Press, Cambridge, UK, 1991.
2. Lytton, R.L., et al., (1993). “Development and Validation of Performance Prediction Model
and Specifications for Asphalt Binders and Paving Mixes.” The Strategic Highway Research
Program Project Report SHRP-A-357, Washington D.C.
3. Hveem, F.N., “Pavement Deflections and Fatigue Failures.” Bulletin 114 HRB, National
Research Council, Washington, D.C, pp. 43-87, 1955.
4. Bonnaure, F., Gravois, A., and Udron, J., A New Method of Predicting the Fatigue Life of
Bituminous Mixes. Journal of the Association of Asphalt Paving Technologists, Vol. 49, pp.
499-529.
5. Asphalt Institute. Thickness Design Manual (MS-1), 9th ed. The Asphalt Institute, College
Park, MD, 1981.
6. Tayebali, A.A., Deacon, J.A., Coplantz, J.S., Harvey, J.T., and Monismith, C.L., Fatigue
Response of Asphalt Aggregate Mixes. SHRP-A-404, Strategic Highway Research Program,
National Research Council, Washington, D.C., 1994.
7. NCHRP, Guide for Mechanistic-Empirical Design of New and Rehabilitation Pavement
Structures, Transportation Research Board of the National Academics, Washington, D.C.,
2004.
8. Finn, F.N., Sraf, C., Kulkarni, R., Nair, K., Smith, W., and Abdullad, A., (1977). “The Use of
Prediction Subsystems for the Design of Pavement Structures.” Proceedings of Fourth
International Conference on Structural Design of Asphalt Pavements, University of
Michigan, Ann Arbor, MI, pp. 3-38.
9. Harvey, J.T., Deacon, J., Tsai, B., and Monismith, C.L., Fatigue Performance of Asphalt
Concrete Mixes and Its Relationship to Asphalt Concrete Pavement Performance in
California. RTA-65W485-2, Asphalt Research Program, CAL/APT Program, Institute of
Transportation Studies, University of California at Berkeley, Berkeley, CA, 1996.
10. M., El-Basyouny, and Witczak, M., Calibration of Alligator Fatigue Cracking Model for
2002 Design Guide. Journal of Transportation Research Board 1919, Washington D.C.,
2005, pp. 77-86.
11. Van Dijk, W. (1975). “Practical Fatigue Characterization of Bituminous Mixes.”
Proceedings of the Association of Asphalt Paving Technologists (AAPT), Vol. 44, p. 38.
12. Rowe, G.M., (1993). “Performance of Asphalt Mixtures in the Trapezoidal Fatigue Test.”
Proceedings of Associations of Asphalt Paving Technologists, Vol. 62, pp. 344-384.
13. Baburamani, P.S., and Porter, D.W., (1996). “Dissipated Energy Approach to Fatigue
Characterisation of Asphalt Mixes.” Proceedings of Combined 18th ARRB TR Conference
Transit New Zealand Symposium, Part 2, pp. 327-347.
14. Ghuzlan, K., (2001). “Fatigue Damage Analysis in Asphalt Concrete Mixtures Based Upon
Dissipated Energy Concepts.” Ph.D. Thesis, University of Illinois at Urbana-Champaign,
Urbana, IL.
15. Shen, S., and Carpenter, S.H., (2005). “Application of Dissipated Energy Concept in Fatigue
Endurance Limit Testing.” Journal of Transportation Research Record: Transportation
Research Board, No. 1929, pp. 165-173.
16. Van Dijk, W., and Visser, W., (1977). “The Energy Approach to Fatigue for Pavement
Design.” Proceedings of Annual Meeting of the Association of Asphalt Paving Technologists
(AAPT), Vol. 46, pp.1-40.
81
17. Pronk, A.C., and Hopman, P.C., (1991). “Energy Dissipation: The Leading Factor of
Fatigue.” Proceedings of the Conference of the United States Strategic Highway Research
Program, pp. 255-267.
18. Tayebali, A.A., Rowe, G.M., and Sousa, J.B., (1992). “Fatigue Response of Asphalt
Aggregate Mixtures.” Proceedings of Asphalt Paving Technologists, Vol. 62, pp. 385-421.
19. Fakhri, M., (1997). “Characterisation of Asphalt Pavement Materials,” Ph.D. Thesis, The
University of New South Wales, Sydney, Australia.
20. Carpenter, S.H., and Jansen, M., (1997). “Fatigue Behavior Under New Aircraft Loading
Conditions.” Proceedings of Aircraft Pavement Technology in the Midst of Change, pp. 259-
271.
21. Ghuzlan, K., and Carpenter, S.H., (2000). “An Energy-Derived/Damage-Based Failure
Criteria for Fatigue Testing.” Transportation Research Record (TRR), No. 1723, pp. 131-
141.
22. Carpenter, S.H., Ghuzlan., K.A., and Shen, S., (2003). “A Fatigue Endurance Limit for
Highway and Airport Pavement.” Journal of Transportation Research Record (TRR), No.
1832, pp. 131-138.
23. Shen, S., Ph.D. Thesis, University of Illinois at Urbana-Champaign, Urbana, IL, 2007.
24. Daniel, J. S., Bisirri, W. M., et al., (2004). “Fatigue Evaluation of Asphalt Mixtures Using
Dissipated Energy and Viscoelastic Continuum Damage Approach.” Journal of the
Association of Asphalt Paving Technologists (AAPT), Baton Rouge, LA, Vol. 73, pp. 557-
583.
25. Daniel, J.S., and Bisirri, W.M., (2005). Characterizing Fatigue in Pavement Materials Using
a Dissipated Energy Parameter. Proceedings of the Geo-Frontiers 2005 Congress, Austin,
TX.
26. Schapery, R.A., A Theory of Mechanical Behavior of Elastic Media with Growing Damage
and Other Changes in Structure. Journal of Mechanics and Physics of Solids, Vol. 38, 1990,
pp. 215-253.
27. Schapery, R.A., On Viscoelastic Deformation and Failure Behavior of Composite Materials
with Distributed Flaws. Advances in Aerospace Structures and Materials, AD-01, ASME,
New York, NY, 1981, pp. 5-20.
28. Schapery, R.A., “Correspondence Principles and a Generalized J-Integral for Large
Deformation and Fracture Analysis of Viscoelastic Media.” International Journal of
Fracture, Vol. 25, 1984, pp. 195-223.
29. Little, D.N., Lytton, R.L., Williams, D., Chen, C.W., and Kim, Y.R., “Fundamental
Properties of Asphalts and Modified Asphalts—Task K: Microdamage Healing in Asphalt
and Asphalt Concrete.” FHWA Final Report, Vol. 1, Report No. DTFH61-92-C-00170,
Springfield, VA: National Technical Information Service, 1997.
30. Lee, H.J., Uniaxial Constitutive Modeling of Asphalt Concrete Using Viscoelasticity and
Continuum Damage Theory, Ph.D. Dissertation, North Carolina State University, Raleigh,
NC, 1996.
31. Kim, Y.R., Lee, H.J., and Little, D.N., “Fatigue Characterization of Asphalt Concrete Using
Viscoelasticity and Continuum Damage Theory,” Journal of the Association of Asphalt
Paving Technologists, Vol. 66, 1997, pp. 520-569.
32. Lee, H.J., and Kim, Y.R., “A Uniaxial Viscoelastic Constitutive Model for Asphalt Concrete
Under Cyclic Loading,” ASCE Journal of Engineering Mechanics, Vol. 124, 1998, No. 11,
pp. 1224-1232.
82
33. Kim, Y.R., Little, D.N., and Lytton, R.R., “Use of Dynamic Mechanical Analysis (DMA) to
Evaluate the Fatigue and Healing Potential of Asphalt Binders in Sand Asphalt Mixtures,”
Journal of the Association of Asphalt Paving Technologists, Vol. 71, 2002, pp. 176-199.
34. Daniel, J.S., and Kim, Y.R., “Development of a Simplified Fatigue Test and Analysis
Procedure Using a Viscoelastic, Continuum Damage Model,” Journal of the Association of
Asphalt Paving Technologists, Vol. 71, 2002, pp. 619-645.
35. Lee, H.J., Kim, Y.R., and Lee, S.W., Prediction of Asphalt Mix Fatigue Life with
Viscoelastic Material Properties, Journal of the Transportation Research Board, No. 1832,
2003, pp.139-147.
36. Christensen, Jr., D., and Bonaquist, R., Practical Application of Continuum Damage Theory
to Fatigue/Phenomena in Asphalt Concrete Mixtures, Journal of the Association of Asphalt
37. Tsai, B.W., Harvey, J.T., and Monismith, C.L. “Application of Weibull Theory in Prediction
of Asphalt Concrete Fatigue Performance.” Transportation Research Record No. 1832.
Transportation Research Board, National Research Council, Washington, D.C., 2003,
pp. 121-130.
38. Ullidtz, P., Harvey, J.T., Tsai, B.-W., and Monismith, C.L., April 2006. Calibration of
Incremental-Recursive Flexible Damage Models in CalME Using HVS Experiments. Report
prepared for the California Department of Transportation (Caltrans) Division of Research
and Innovation by the University of California Pavement Research Center, Davis and
Berkeley. UCPRC-RR-2005-06.
39. Majidzadeh, K., Kaufmann, E.M., and Ramsamooj, D.V., Application of Fracture Mechanics
in the Analysis of Pavement Fatigue, Proceedings of Association of Asphalt Pavement
Technologists, Vol. 40, 1970, pp. 227-246.
40. Monismith, C.L., “Reflection Cracking, Analyses, Laboratory Studies, and Design
Considerations,” Proceedings AAPT, Volume 49, pp. 268-313, 1980.
41. Paris, P.C., and Erdogan E., A Critical Analysis of Crack Propagation Laws, Journal of Basic
Engineering, Transaction of the American Society of Mechanical Engieering, Series D., Vol.
85, pp. 528-883, 1963.
42. Schapery, R.A., A Theory of Crack Growth in Visco-Elastic Media, Report MM 2764-73-1,
Mechanics and Materials Research Center, Texas A&M University, College Station, TX,
1973.
43. Schapery, R.A., A Theory of Crack Initiation and Growth in Visco-Elastic Media; I:
Theoretical Development; II: Approximate Methods of Analysis; III: Analysis of Continuous
Growth; International Journal of Fracture, Sijthoff and Noordhoff International Publisher;
Vol. 11, No. 1, pp. 141-159; Vol. 11, No. 3, pp. 369-388; and Vol. 11, No. 4, pp. 549-562,
1975.
44. Schapery, R.A., A Method for Predicting Crack Growth in Nonhomogeneous Visco-Elastic
Media, International Journal of Fracture, Sijthoff and Noordhoff International Publishers,
Vol.14, No. 3, pp. 293-309, 1978.
45. Chang, H.S., Lytton, R.L., and Carpenter, S.H., Prediction of Thermal Reflection Cracking in
West Texas, Texas Transportation Institute. Research Report 18-3, Study 2-8-73-18, College
Station, TX, March 1976.
46. Jayawickrama, P.W., and Lytton, R.L., Methodology for Predicting Asphalt Concrete
Overlay Life Against Reflection Cracking, Proceedings 6th International Conference on
Structural Design of Asphalt Pavements, Volume I, pp. 912-924, 1987.
83
47. Zhou, F., Hu, S., Scullion, T., Chen, D., Qi, X., and Claros, G., Development and
Verification of an Overlay Tester Based Fatigue Crack Prediction Approach, Journal of
Association of Asphalt Paving Technologist, March 2007, San Antonio, TX.
48. Barsoum, R.S., On the Use of Isoparametric Elements in Linear Fracture Mechanics,
International Journal Numerical Methods in Engineering, Vol. 10, p. 25, 1976.
49. Ingraffea, A.R., and Manu, C., Stress-Intensity Factor Computation in Three Dimensions
with Quarter-Point Crack Tip Elements, International Journal Numerical Methods in
Engineering, Vol. 12, p. 235, 1978.
50. Salam, Y.M., and Monismith, C.L., Fracture Characteristics of Asphalt Concrete,
Proceedings of Association of Asphalt Paving Technologists, Vol. 41, pp. 215-256, 1972.
51. Majidzadeh, K., Dat, M., and Makdisi-Ilyas, F., Application of Fracture Mechanics for
Improved Design of Bituminous Concrete, Vol. 2 Evaluation of Improved Mixture
Formulations, and the Effect of Temperature Conditions on Fatigue Models, Report No.
FHWA-RD-76-92, Final Report, June 1976.
52. Elmitiny, M.R.N., Material Characterization for Studying Flexible Pavement Behavior in
Fatigue and Permanent Deformation, Ph.D. Dissertation, The Ohio State University, 1980.
53. Molenaar, A.A.A., Structural Performance and Design of Flexible Road Constructions and
Asphalt Concrete Overlays, Ph.D. Dissertation, Delft University of Technology, The
Netherlands, 1983.
54. Molenaar, A.A.A., Fatigue and Reflection Cracking due to Traffic Loads, Proceedings of the
Association of Asphalt Paving Technologists, Vol. 53, pp. 440-474, 1984.
55. Jacobs, M.M.J., Crack Growth in Asphaltic Mixes, Ph.D. Dissertation, Delft University of
Technology, the Netherlands, 1995.
56. Jacobs, M.M.J., Hopman, P.C., and Molenaar, A.A.A., Application of Fracture Mechanics in
Principles to Analyze Cracking in Asphalt Concrete, Journal of the Association of Asphalt
Paving Technologists, Vol. 65, pp 1-39, 1996.
57. Erkens, S., Moraal, J., Molenaar, A.A.A., Groenendijk, J., and Jacobs, M., Using Paris’ Law
to Determine Fatigue Characteristics – A Discussion, Proceedings of the 8th International
Conference on Asphalt Pavements, Seattle, WA, pp. 1123-1142, August 10-14, 1997.
58. Medani, T.O. and Molenaar, A.A.A., Estimation of Fatigue Characteristics of Asphaltic
Mixes Using Simple Tests, Herron, TNO Building and Construction Research and The
Netherlands School for Advanced Studies in Construction, Vol. 45, No. 3, pp. 155-165,
2000.
59. Uzan, J., and Levenberg, E., Strain Measurements in Asphalt Concrete Specimens towards
the Development of a Fracture Model, International Journal of Pavement Engineering, Vol.
2, No. 4, pp. 243-258, 2001.
60. Jenq, Y.S., Liaw, C.J., and Liu, P., Analysis of Crack Resistance of Asphalt Concrete
Overlays − A Fracture Mechanics Approach, Transportation Research Record, No. 1388,
pp. 160-166, 1993.
61. Dugdale, D. S., Yielding in Steel Sheets Containing Slits, Journal of the Mechanics and
Physics of Solids, Vol. 8, pp. 100-108, 1960.
62. Barenblatt, G.I., The Mathematical Theory of Equilibrium Cracks in Brittle Fracture,
Advances in Applied Mechanics, Academic Press, Vol. VII, pp. 55-129, 1962.
63. Hillerborg, A., Modeer, M., and Peterson, P.E., Analysis of Crack Formation and Crack
Growth in Concrete by Means of Fracture Mechanics and Finite Elements, Cement and
Concrete Research, Vol. 6, pp. 773-782, 1985.
64. Bazant, Z.P. and Planas, J., Fracture and Size Effect in Concrete and other Quasibrittle
84
Materials, CRC Press, Boca Raton, FL, p. 616, 1998.
65. Petersson, P.E., (1981). “Crack Growth and Development of Fracture Process Zone in Plain
Concrete and Similar Materials,” Report No. TVBM-1006, Lund Institute of Technology,
Lund, Sweden.
66. Jenq, Y.S. and Perng, J.D., (1991). “Analysis of Crack Propagation in Asphalt Concrete
Using a Cohesive Crack Model,” Transportation Research Record, TRB, 1317, pp. 90-99.
67. Superpave model team (1999). “Advanced AC Mixture Material Characterization Models
Framework and Laboratory Test Plan Final Report,” NCHRP 9-19, Superpave Support and
Performance Models Management, Tempe, AZ.
68. Seo, Y., Kim, Y.R., Schapery, R.A., Witczak, M.W., and Bonaquist, R., A Study of Crack-
Tip Deformation and Crack Growth in Asphalt Concrete Using Fracture Mechanics, Journal
of Association of Asphalt Paving Technologists, Vol. 74, 2004, pp. 790-795.
69. Soares, J.B., de Freitas, F.A.C., and Allen, D.H., Considering Material Heterogeneity in
Crack Modeling of Asphaltic Mixtures, Transportation Research Record, No. 1832, pp. 113-
120, 2003.
70. Paulino, G.H., Song, S.H., and Buttlar, W.G., Cohesive Zone Modeling of Fracture in
Asphalt Concrete, Proceedings of the 5th International Conference on Cracking in
Pavements, May 5-7, Lemoges, France, pp. 63-70, 2004.
71. Kim, H., and Buttlar, W.G., Micromechanical Fracture Modeling of Hot-Mix Asphalt
Concrete Based on a Disk-shaped Compact Tension Test, Journal of Association of Asphalt
Paving Technologists, Vol. 75E, 2005.
72. Wagoner, M.P., Buttlar, W.G., and Paulino, G.H., Disk-shaped Compact Tension Test for
Asphalt Concrete Fracture, Journal of Society for Experimental Mechanics, Vol. 45, No. 3,
pp. 270-277, 2005.
73. Wagoner, M.P., Buttlar, W.G., and Paulino, G.H., Development of a Single-Edge Notched
Beam Test for Asphalt Concrete Mixtures, ASTM Journal of Testing and Evaluation,
Vol. 33, No. 6, pp. 1-9, 2005.
74. Wu, R., Harvey, J.T.,and Monismith, C.L., Towards a Mechanistic Model for Reflective
Cracking in Asphalt Concrete Overlays, Journal of the Association of Asphalt Paving
Technologists, Vol. 75, pp. 491-534, March 2006.
75. Chaboche, J.L., Continuum Damage Mechanics: Part I - General Concepts. ASME Journal of
Applied Mechanics, Vol. 55, No. 1, pp. 59–64, March 1988.
76. Chaboche, J.L., Continuum Damage Mechanics: Part II - Damage Growth, Crack Initiation,
and Crack Growth. ASME Journal of Applied Mechanics, 55(1):65–72, March 1988.
77. Lee, H.J., and Kim, Y.R., Viscoelastic Constitutive Model for Asphalt Concrete under Cyclic
Loading. Journal of Engineering Mechanics-ASCE, 124(1):32–40, 1998.
78. Lee, H.J., and Kim, Y.R., Viscoelastic Continuum Damage Model of Asphalt Concrete with
Healing. Journal of Engineering Mechanics-ASCE, 124(11):1224–1232, 1998.
79. Bazant, Z.P., and Jirasek, M., Nonlocal Integral Formulations of Plasticity and Damage:
Survey of Progress. Journal of Engineering Mechanics, Vol. 128, No. 11, pp. 1119–1149,
2002.
80. Mazars, J., and Pijaudier-Cabot, G., Continuum Damage Theory–Application to Concrete,
Journal of Engineering Mechanics-ASCE, Vol. 115, No. 2, pp. 345–365, 1989.
85
81. Peerlings, R.H.J., Brekelmans, W.A.M., de Borst, R., and Geers, M.G.D., Gradient-Enhanced
Damage Modelling of High-Cycle Fatigue, International Journal for Numerical Methods in
Engineering, Vol. 49, No. 12, pp. 1547–1569, 2000.
82. Lemaitre, J., A Course on Damage Mechanics. Springer, Berlin, 2nd edition, 1996.
83. Peerlings, R.H.J., de Borst, R., Brekelmans, W.A.M, and deVree, J.H.P., Gradient Enhanced
Damage for Quasi-Brittle Materials, International Journal for Numerical Methods in
Engineering, Vol. 39, No. 19, pp. 3391–3403, 1996.
84. Walubita, L.F., Epps-Martin, A., Glover, C.J., Jung, S.H., Cleveland, G.S., Lytton, R.L., and
Park, E.S., Application of the Calibrated Mechanistic Approach with Surface Energy
(CMSE) Measurements for Fatigue Characterization of Asphalt Mixtures, Journal of the
Association of Asphalt Paving Technologists, Vol. 75, 2006.
85. Sousa, J.B., Pais, J.C., Prates, M., Barros, R., Langlois, P., and Leclerc, A-M, Effect of
Aggregate Gradation on Fatigue Life of Asphalt Concrete Mixes, Transportation Research
Record, No. 1630, 1998, pp. 62-68.
86. Tsai, B.W., High Temperature Fatigue and Fatigue Damage Process of Aggregate-Asphalt
Mixes, Ph.D. Dissertation, University of California at Berkeley, June 2001.
87. Ghuzlan, K.A., and Carpenter, S. H., Traditional Fatigue Analysis of Asphalt Concrete
Mixtures, Transportation Research Board, CD-ROM, Washington, D.C., 2003.
88. Tsai, B.W., and Monismith, C.L., Influence of Asphalt Binder Properties on the Fatigue
Performance of Asphalt Concrete Pavements, Journal of the Association of Asphalt Paving
Technologists, Vol. 74, 2005.
89. Zhou, F., Hu, S., and Scullion, T., Development and Verification of the Overlay Tester Based
Fatigue Cracking Prediction Approach, FHWA/TX-07/9-1502-01-8, Texas Transportation
Institute, College Station, TX, January 2007.
90. Highway Research Board, The AASHO Road Test, Special Report 73, Publication No. 1012,
Washington, D.C., p. 117, 1962.
91. Hofstra, A., and Klomp, A.J.G., Permanent Deformation of Flexible Pavements Under
Simulated Road Traffic Conditions, Proceedings of the 3rd International Conference on the
Structural Design of Asphalt Pavements, Vol. I, London, England, 1972, pp. 613-621.
92. Uge, P., and van de Loo, P.J., Permanent Deformation of Asphalt Mixes, Koninklijke/Shell
Laboratorium, Amsterdam, The Netherlands, November 1974.
93. Eisenmann, J., and Hilmer, A., Influence of Wheel Load and Inflation Pressure on the
Rutting Effect at Asphalt-Pavements-Experiments and Theoretical Investigations,
Proceedings of the 6th International Conference on the Structural Design of Asphalt
Pavements, Vol. I, Ann Arbor, MI, 1987, pp. 392-403.
94. Brown, E.R., and Cross, S., A Study of In-Place Rutting of Asphalt Pavements. Proceedings,
Association of Asphalt Paving Technologists, Volume 58, 1989.
95. Harvey, J.T., and Popescu, L., Rutting of Caltrans Asphalt Concrete and Asphalt-Rubber Hot
Mix Under Different Wheels, Tires, and Temperatures–Accelerated Pavement Testing
Evaluation, Draft report prepared for California Department of Transportation Pavement
Research Center, CAL/APT Program, Institute of Transportation Studies, University of
California, Berkeley, CA. UCPRC-RR-2000-0, January 2000.
96. Stuart, K.D., Mogawer, W.S., and Romero, P., (1999). “Validation of the Superpave Asphalt
Binder and Mixture Tests that Measure Rutting Susceptibility Using an Accelerated Loading
Facility.” Rep. No. FHWA-RD-99-204, Federal Highway Administration, McLean, VA.
86
97. Chen, D., Bilyeu, J., Scullion, T., Lin, D., and Zhou, F., “Forensic Evaluation of Premature
Failures of Texas Specific Pavement Study-1 Sections.” Journal of Performance and
Construction Facility, Vol. 17, pp. 67-74, Nov. 2 2003.
98. Brown, E.R., Prowell, B., Cooley, A., Zhang, J., and Powell, R.B., “Evaluation of Rutting
Performance on the 2000 NCAT Test Track,” Preprints, Journal of the Association of
Asphalt Paving Technologists, March 2004.
99. David, T., West, R., Priest, A., Powell, B., Selvaraj, I., Zhang, J., and Brown, R., Phase II
NCAT Test Track Results, NCAT Report 06-05, National Center for Asphalt Technology,
December 2006.
100. Gokhale, S., Choubane, B., Byron, T., and Tia, M., Rut Initiation Mechanisms in Asphalt
Mixtures as Generated Under Accelerated Pavement Testing, Transportation Research
Record 1940, 2005, pp. 136-145.
101. Claessen, A.I.M., Edwards, J.M., Sommer, P., and Uge, P., Asphalt Pavement Design, The
Shell Method, Proceedings of the 4rd International Conference on the Structural Design of
Asphalt Pavements, Vol. I, Ann Arbor, MI, 1977, pp. 39-74.
102. Brown, S.F., and Brunton, J.M., Improvements to Pavement Subgrade Strain Criterion,
ASCE. Journal of Transportation Engineering, 110(6), 1984.
103. Brown, S.F., Achievements and Challenges in Asphalt Pavement Engineering, the ISAP
Distinguished Lecture series, delivered in Seattle, WA, August 1997.
104. Huekelom, W., and Klomp, A.J.G., Consideration of Calculated Strains at Various Depths
in Connection with the Stability of Asphalt Pavements, Proceedings of the 2nd International
Conference on the Structural Design of Asphalt Pavements, Vol. I, Ann Arbor, MI, 1967.
105. Monismith, C.L., Permanent Deformation Studies of Pavements, FCP Research Progress
Review Report, San Francisco, CA, 1973.
106. McLean, D.B., Permanent Deformation Characteristics of Asphalt Concrete, Ph.D.
Dissertation, University of California, Berkeley, CA, 1973.
107. Romain, J.E., Rut Depth Prediction in Asphalt Pavements, Research Report No. 150, Centre
de Recherches Routieres, Brasseks, Belgium, 1969.
108. Barksdale, R.D., Laboratory Evaluation of Rutting in Base Course Materials, Proceedings
of the 3rd International Conference on the Structural Design of Asphalt Pavements, Vol. I,
London, England, 1972.
109. Morris, J., and Haas, R.C.G., Designing for Rutting in Asphalt Pavements, Canadian
Technical Asphalt Association Annual Meeting, 1972.
110. Guide for Mechanistic–Empirical Design of New and Rehabilitated Pavement Structures.
Final Report, NCHRP Project 1-37A. Transportation Research Board, National Research
Council, Washington, D.C., March 2004. http://trb.org/mepdg/guide.htm.
111. Charles Schwartz, Improving Mechanistic-Empirical Models for Predicting HMA Rutting,
Presented at 2006 Symposium on Models Used to Predict Pavement Performance, Laramie,
WY.
http://www.petersenasphaltconference.org/download/Schwartz%20Improving%20Mechanis
tic-Empirical%20Models%20for%20Predicting%20HMA%20Rutting.pdf
112. Kenis, W.J., Predictive Design Procedure, VESYS User’s Manual: An Interim Design
Method for Flexible Pavement Using the VESYS Structural Subsystem. Final Report No.
FHWA-RD-77-154, FHWA, Washington D.C., 1978.
87
113. Kenis, W.J., and Wang, W., (1997) “Calibrating Mechanistic Flexible Pavement Rutting
Models From Full Scale Accelerated Tests, Proceedings of the Eighth International
Conference on Asphalt Pavements, Vol. I, pp. 663-672, Seattle, WA.
114. Epps, J.A., Hand, A., Seeds, S., Scholz, T., Alavi, S., Ashmore, C., Monismith, C.L.,
Deacon, J.A., Harvey, J.T., and Leahy, R.B., “Recommended Performance-Related
Specifications for Hot-Mix Asphalt Construction.” NCHRP Report 455, National
Cooperative Highway Research Program, Transportation Research Board, National
Research Council, Washington, D.C., 2002, 496 pp.
115. Lytton, R. L., et al. (1993). “Development and Validation of Performance Prediction Model
and Specifications for Asphalt Binders and Paving Mixes.” The Strategic Highway
Research Program Project Report SHRP-A-357.
116. Brown, S.F., et al., Independent Review of the Mechanistic-Empirical Pavement Design
Guide and Software, Research Results Digest 307, Transportation Research Board,
September 2006.
117. Brown, S.F., Design Considerations for Pavement and Rail Track Foundations, Ed. Correia
and Loizos, Geotechnics in Pavement and Railway Design and Construction, Athens, GA,
pp. 61-72, 2003.
118. Lekarp, F., Isacsson, U., and Dawson, A., State of the Art. II: Permanent Strain Response
of Unbound Aggregates. Journal of Transportation Engineering, ASCE, Vol. 126, No. 1,
pp. 76-83, 2000.
119. Duncan, J.M., and Chang, C.Y., Nonlinear Analysis of Stress and Strain in Soils. Journal
of the Soil Mechanics and Foundations Division, ASCE, 96(SM5), pp. 1629-1653, 1970.
120. Barksdale, R.D., Laboratory Evaluation of Rutting in Base Course Materials. Proc. 3rd
International Conference on the Structural Design of Asphalt Pavements, Vol. 1, Ann
Arbor, MI, pp. 161-174. 1972.
121. Monismith, C.L., Ogawa, L., and Freeme, C.R., Permanent Deformation Characteristics of
Subgrade Soils due to Repeated Loading. Transportation Research Record, No. 537,
Transportation Research Board, National Research Council, Washington D.C., pp. 1-17,
1975.
122. Kenis, W.J., Predictive Design Procedure, VESYS User’s Manual: An Interim Design
Method for Flexible Pavement Using the VESYS Structural Subsystem. Final Report No.
FHWA-RD-77-154, FHWA, Washington D.C., 1978.
123. Lentz, R.W., and Baladi, G.Y., Constitutive Equation for Permanent Strain of Sand
Subjected to Cyclic Loading. Transportation Research Record, No. 810, Transportation
Research Board, National Research Council, Washington D.C., pp. 50-53, 1981.
124. Tseng, K-H., and Lytton, R.L., Prediction of Permanent Deformation in Flexible Pavement
Materials. Implication of Aggregates in the Design, Construction, and Performance of
Flexible Pavements, ASTM STP 1016, ASTM, Philadelphia, PA, pp. 154-172, 1989.
125. Ullidtz, P., Modeling of Granular Materials Using the Discrete Element Method.
Proceedings of the Eight International Conference on Asphalt Pavements, Seattle, WA,
pp. 757-769, 1997.
126. Ceratti, J.A., Washington, P.N., Wai, Y.Y.G., and Jose, A., Rutting of Thin Pavements:
Full-Scale Study. Transportation Research Record, No. 1716, Transportation Research
Board, National Research Council, Washington D.C., pp. 82-88, 2000.
88
127. Park, S.W., Evaluation of Accelerated Rut Development in Unbound Pavement
Foundations and Load Limits on Load-Zoned Pavements. Ph.D. Dissertation, Texas A&M
University, College Station, TX, 2000.
128. Applied Research Associates. Guide for Mechanistic-Empirical Design of New and
Rehabilitated Pavement Structures. Final Report, NCHRP 1-37A, Champaign, IL, 2004.
129. Zhou, F., and Scullion, T., VESYS5 Rutting Model Calibrations with Local Accelerated
Pavement Test Data and Associated Implementation. Report No. FHWA/TX-03/9-1502-
01-2, Texas Transportation Institute, College Station, TX, 2002.
130. Uzan, J., Permanent Deformation in Flexible Pavements. Journal of Transportation
Engineering, ASCE, Vol. 130, No. 1, pp. 6-13, 2004.
131. Henning, T.F.P., Dunn, R.C. M., Parkman, C.C., and Brass, J., Long-Term Pavement
Performance (LTPP) Studies in New Zealand – A Progress Update. NZIHT & Transit NZ
8th Annual Conference, pp. 1-19, 2006.
132. Mishalani, R.G., and Kumar, A., Impact of Overlays on Pavement Rutting and Their
Interactions with Design and Material Quality. Transportation Research Record, No.
1869, Transportation Research Board, National Research Council, Washington D.C.,
pp. 97-105, 2004.
133. Archilla, A.R., and Madanat. S., Estimation of Rutting Models by Combining Data from
Different Sources. Journal of Transportation Engineering, ASCE, Vol. 127, No. 5,
pp. 379-389, 2001.
134. Scullion, T., Uzan, J., Hilbrich, S., and Chen, P., Thickness Design Systems for Pavements
Containing Soil Cement Bases, PCA R&D Serial No. 2863, Portland Cement Association,
Skokie, Ill, 2007.
135. Goetz, W.H., and Wood, L.E., Highway Engineering Handbook, 1st Ed., K.B. Woods, Ed.,
McGraw-Hill, New York, NY, 1960, pp. 18-52 to 18-92.
136. The Asphalt Institute, Mix Design Methods for Asphalt Concrete (MS-2), 6th Ed.,
Lexington, KY, 1997.
137. Sousa, J., Craus, J., and Monismith, C.L., Summary Report on Permanent Deformation in
Asphalt Concrete, SHRP A/IR/91-104, Strategic Highway Research Program, National
Research Council, Washington, D.C., 1991.
138. Cominsky, R.J., Huber, G.A., Kennedy, T.W., and Anderson, M., The Superpave Mix
Design Manual for New Construction and Overlays, SHRP-A-407, Strategic Highway
Research Program, National Research Council, Washington, D.C., 1994.
139. Christensen, D.W., Bonaquist, R., and Cooley, A., A Mix Design Manual for Hot-Mix
Asphalt, NCHRP 9-33, Draft Interim Report, May 2005.
140. Witczak, M.W., Kaloush, K., Pellinen, T., El-Basyouny, M., and Von Quintus, H., Simple
Performance Test for Superpave Mix Design, NCHRP Report 465, National Cooperative
Highway Research Program, Washington, D.C., 2002.
141. Tangella, S., Rao, C.S., Craus, J., Deacon, J.A., and Monismith, C.L., Summary Report on
Fatigue Response of Asphalt Mixtures, TM-UCB-A-003A-89-3, Prepared for Strategic
Highway Research Program, Project A-003-A, University of California, Berkeley, CA,
February 1990.
142. Witczak, M.W., Kaloush, K., Pellinen, T., El-Basyouny, M., and Von Quintus, H., Simple
Performance Test for Superpave Mix Design, NCHRP Report 465, National Cooperation
Highway Research Program, Washington, D.C., 2002.
89
143. Walubita, L.F., Epps-Martin, A., Jung, S.H., Glover, C.J., Park, E.S., Chowdhury, A., and
Lytton, R. L., Comparison of Fatigue Analysis Approaches for Two Hot Mix Asphalt
Concrete (HMAC) Mixtures, Draft TxDOT Technical Research Report FHWA/TX-05/0-
4468-2, October 2005.
144. Kim, Y.R., Lee, H.J., and Little, D.N., Fatigue Characterization of Asphalt Concrete Using
Viscoelasticity and Continuum Damage Theory, Journal of the Association of Asphalt
145. Zhou, F., Hu, S., Scullion, T., Qi, X., Chen, D., and Claros, G., Development and
Verification of Overlay Tester Based Fatigue Cracking Prediction Approach, Journal of
Association of Asphalt Paving Technologists, San Antonio, TX, March 12-14, 2007.
146. Scullion, T., “Perpetual Pavements in Texas: The State of the Practice,” Report
FHWA/TX-05/0-4822-1, Texas Transportation Institute, College Station, TX, 2006.
147. AASHTO. AASHTO Designation: TP 62-03, Standard Method of Test for Determining
Dynamic Modulus of Hot Mix Asphalt Concrete Mixtures, AASHTO Standards,
Washington D.C., 2001.
148. Sebesta, S., Harris, P., and Liu, W., “Improving Lab Compaction Methods for Roadway
Base Materials,” Research Report 0-5135-2, Texas Transportation Institute, Texas A&M
University, College Station, TX, 2006.
149. Andrei, D., “Development of a Predictive Model for the Resilient Modulus of Unbound
Materials,” Ph.D. Dissertation, Department of Civil and Environmental Engineering,
Arizona State University, Tempe, AZ, May 2003.
90
APPENDIX A
OT FOR FRACTURE PROPERTIES OF HMA MIXES
It is well known that HMA mixtures are complex materials. However, for simplicity and
practical applications, HMA mixtures are often assumed to be quasi-elastic materials represented
by dynamic modulus and Poisson’s ratio. With this assumption, the well-known Paris’ law
shown in Equation 1 can be used to describe crack propagation of HMA mixtures (1).
dc
= A(ΔK )n (1)
dN
where:
c = crack length,
N = number of load repetitions,
dc/dN = crack speed or rate of crack growth,
∆K = change of stress intensity factor (SIF), and
A, n = fracture properties of material.
In view of Equation 1, it can be seen that the information required for determining
fracture properties (A and n) includes 1) the SIF corresponding to any specific crack length (c)
and 2) crack length (c) corresponding to a specific number of load repetitions (N). The proposed
approach for determining the SIF and crack length (c) is discussed as follows:
• Determination of SIF
A two dimensional (2D) finite element (FE) program named 2D-CrackPro was
developed to analyze the SIF under the OT testing. In the 2D-CrackPro program, the desired
1 stress singularity in the crack tip region was met by placing the mid-side nodes of two
r
adjacent sides of an 8-node isoparametric element at the one-fourth distance mark from the
common corner node (2). The accuracy of this program has been verified by comparing the
computed SIFs of an infinite slab with a center crack with those given in “the stress analysis of
cracks handbook” (3).
Figure A1 shows the 2D FE mesh plus the singularity elements used. Since Poisson’s
ratio has minor influence on SIF, a constant Poisson’s ratio (υ=0.35) was used for all the
analyses. With the above quasi-elastic assumption, it has been found that the SIF is proportional
to dynamic modulus (E) of the overlay tester (OT) specimen and the specified maximum opening
displacement (MOD). Therefore, the SIFs corresponding to variable crack length (c) were
calculated at an assumed condition: 1) dynamic modulus of the OT specimen: E =1 MPa, and 2)
MOD = 1 mm. The results are presented in Figure A2. To facilitate implementation, a
regression equation shown in Figure A2 was developed for the SIF versus crack length at the
condition of E=1 MPa and MOD = 1 mm.
91
For any other E and MOD combination, the corresponding SIF can be determined by the
following equation:
SIF = 0.2911 ∗ E ∗ MOD ∗ c −0.4590 (2)
where:
SIF = stress intensity factor, MPa*mm0.5,
E = dynamic modulus, MPa,
MOD = maximum opening displacement, mm, and
c = crack length, mm.
Figure A1. A 2-D FE Mesh of the OT System.
E=1 MPa and MOD=1 mm
0.4
0.3
SIF-K (MPa*mm^0.5)
0.2
y = 0.2911x -0.4590
R2 = 0.9996
0.1
0
0 5 10 15 20 25
Crack Length (mm)
Figure A2. Calculated SIF vs. Crack Length.
Additionally, it can be seen that the SIF shown in Figure A2 decreases rapidly at the
beginning, and its decreasing rate becomes smaller and smaller with crack length growth. This
observation indicates that the initial crack propagation stage is very important to determine
reasonable fracture properties of HMA mixtures, which means that the required fracture
properties can be determined from the initial stage of the OT testing (perhaps within 15 minutes).
This feature separates the OT from other types of fracture tests (such as, direct tension test [4, 5,
6], indirect tension test [7]), because the other tests often focused on the late crack propagation
stage where the SIF increased rapidly so that these tests generally take a very long time (say
hours).
92
• Determination of crack length (c)
To monitor crack length growth, researchers have used several different techniques such
as crack foil (5). Recently, Seo et al. applied a Digital Image Correlation (DIC) technique to
monitor crack propagation and crack length (6). The DIC is a non-contact, full-field
displacement (or strain) measurement system that analyzes the displacement (or strain) by
comparing digital images of a deformed specimen with that of an initial undeformed specimen.
Compared with other techniques, the DIC is one of most advanced techniques for monitoring
crack propagation. However, using the DIC system will definitely increase the difficulty and cost
of running the OT. Fortunately, there is an alternative method used for estimating crack length,
namely the backcalculation approach, which has been successfully used by Jacobs (5) and Roque
et al. (7) to backcalculate the crack length from the recorded load or displacements. However,
this backcalculation approach needs to be calibrated.
Three assumptions listed below were made for establishing the theoretical relationship
between an equivalent crack length and the maximum load required to reach a specified MOD.
1) An equivalent (or ideal) crack starts from the bottom at the center of the OT
specimen and propagates vertically to the top surface of the specimen.
2) The reduction of maximum load from the first cycle is attributed to crack growth.
3) As assumed previously, HMA mixtures are quasi-elastic and are represented by a
dynamic modulus and Poisson’s ratio (υ=0.35).
With the above three assumptions, the maximum load required to reach a MOD is
proportional to the dynamic modulus of the OT specimen and decreases with crack length
growth, provided that the MOD is constant. To exclude the influence of the dynamic modulus
and the MOD, the maximum load corresponding to any crack length was normalized to the
maximum load corresponding to “zero” crack length which is determined through extrapolation.
Figure A3 shows the relationship between the normalized maximum load (y-axis) and crack
length (x-axis). A corresponding regression equation is also presented in Figure A3.
Since the maximum load at each cycle is automatically recorded during the OT testing, it
is easy to develop the relationship between the normalized maximum load at each cycle and the
number of cycles. Finally, combining with Figure A3, crack growth rate (dc/dN) can be
calculated.
• Determination of fracture properties: A and n
With known SIF (K) and crack growth rate (dc/dN), the fracture properties (A and n) can
be readily determined. Figure A4 shows the five steps of determining fracture properties (A and
n) of HMA mixtures. Currently, a Microsoft Excel® macro named TTI-OT is under
development to automatically analyze the OT test results and determine fracture properties (A
and n).
93
Normalized Maximum Load vs. Crack Length
0.8
Normalized Max. Load 4 3 2
y = 3E-05x - 0.0012x + 0.0189x - 0.155x + 1.0043
2
R = 0.9993
0.6
0.4
0.2
0
0 2 4 6 8 10 12 14 16
Crack Length (mm)
Figure A3. Normalized Maximum Load vs. Crack Length.
In summary, this section focused on developing the methodology of determining fracture

properties (A and n) using the OT. As listed below, this OT based methodology for fracture
properties has several desirable features:
1. Specimen size (150 mm long by 75 mm wide by 38 mm high): This size of specimen

can be easily cut from samples compacted by the SGC or from field cores.
2. Specimen preparation: Neither a hole in the center nor a notch at the bottom of the
specimen is required, since a crack is always initiated in the first cycle due to large
opening displacement.
3. Testing time: In contrast to other fracture types of tests (i.e., IDT or repeated direct
tension test), which generally take long testing time, the OT for fracture properties
(A and n) can generally be done within 15 minutes.
94
Normalized Max. Load vs. No. of Cycles Normalized Maximum Load vs. Crack Length
1 1
2
1
0.8 0.8
4 3 2
y = 3E-05x - 0.0012x + 0.0189x - 0.155x + 1.0043
Norm alized M ax. Load

2
R = 0.9993
0.6 0.6
0.4 0.4
0.2 0.2
0 0
20 18 16 14 12 10 8 6 4 2 0 0 2 4 6 8 10 12 14 16
No. of Cycles Crack Length (mm)
C rack Length (mm) vs. N o. of C ycles
20
16
3
Crack Length (mm)
12
0
0 2 4 6 8 10 12 14 16 18 20
No. of Cycle s
E=1 MPa and MOD=1 mm
0.4
4
0.3
SIF-K (MPa*mm^0.5)
0.2
y = 0.2911x -0.4590
R2 = 0.9996
0.1
0
0 5 10 15 20 25
Crack Length (mm)
F H WA-AL F L ane 6-T e rp o lyme r: 19C and 0.48 mm O p e n ing
10
y = 2.6220E -07x 4 . 0 79 0 E +0 0
For this R 2 = 9.8241E -01

dc/dN
1
case:
A =2.6220E-7
n = 4.0790
0.1
1 10 100
S tre ss In te n sity F a cto r (M P a *m m ^ 0.5)
Figure A4. Determination of Fracture Properties: A and n.
95
APPENDIX A REFERENCES
1. Paris, P.C., and Erdogan, F., A Critical Analysis of Crack Propagation Laws, Transactions
of the ASME, Journal of Basic Engineering, Series D, 85, No. 3, 1963.
2. Barsoum, R.S., On the Use of Isoparametric Elements in Linear Fracture Mechanics,
International Journal of Numerical Methods in Engineering, Vol. 10, pp. 25, 1976.
3. Tada, H., Paris, P.C., and Irwin, G.R., The Stress Analysis of Cracks Handbook, 3rd Ed.,
New York, ASME Press, 2000.
4. Molenaar, A.A.A., “Structural Performance and Design of Flexible Road Constructions
and Asphalt Concrete Overlays,” Ph.D. Dissertation, Delft University of Technology, 1983.
5. Jacobs, M.M. J., Crack Growth in Asphaltic Mixes, Ph.D. Dissertation, Delft University of
Technology, Road Railraod Research Laboratory, The Netherlands, 1995.
6. Seo, Y., Kim, Y.R., Schapery, R.A., Witczak, M.W., and Bonaquist, R., A Study of Crack-
Tip Deformation and Crack Growth in Asphalt Concrete Using Fracture Mechanics,
Journal of the Association of Asphalt Paving Technologists, Vol. 73, 2004, pp. 200-228.
7. Roque, R., Zhang, Z., and B. Sankar, B., Determination of Crack Growth Rate Parameters
of Asphalt Mixtures Using the Superpave IDT, Journal of the Association of Asphalt
96
APPENDIX B
VESYS TEST PROTOCOL FOR ASPHALT MIXES
1. Test Samples
1.1 Size
Testing shall be performed on 100 mm (4 inch) diameter by 150 mm (6 inch) high or
more test samples from laboratory or cores from field.
1.2 Aging
For laboratory compacted samples, mixture shall be aged in accordance with the short-
term oven aging procedure in AASHTO PP2.
1.3 Gyratory Specimens

For laboratory compacted samples, prepare 150 mm (6 inch) high samples to the required
air void content in accordance with AASHTO TP-4. The gyratory compactor is shown in
Figure B1.
1.4 End Preparation
The ends of all test samples shall be smooth and perpendicular to the axis of the
specimen. Prepare the ends of the samples by milling with a single- or double-bladed
saw. To ensure that the sawed samples have parallel ends, the sample ends shall have a
cut surface waviness height within a tolerance of ±0.05 mm across any diameter.
1.5 Air Void Content

Determine the air void content of the final test sample in accordance with AASHTO
T269. Reject samples with air voids that differ by more than 0.5 percent from the target
air voids.
1.6 Replicates
The number of test samples required depends on the number of axial strain measurements
made per sample and the desired accuracy of the average permanent deformation.
Normally, two replicates are acceptable for each sample with two LVDTs.
97
2. Test Sample Instrumentation
2.1 Attach mounting studs for the axial LVDTs to both sides of the sample with 180 degree
intervals (in plan view) using epoxy cement (shown in Figure B2). Make sure the studs
are aligned.
Figure B1. Superpave Gyratory Compactor.
98
4 inch
Figure B2. Samples with Studs.
2.2 The gauge length for measuring axial deformations shall be 100 mm ± 1 mm (4 inch ±
0.04 inch). The gauge length is normally measured between the stud centers.
3. Test Procedures
3.1 The recommended test protocol for Alpha and Mu used in the VESYS program consists
of testing the asphalt mix at two temperatures with specified stress level. Table B1
shows the recommended test temperatures and associated stress level.
Table B1. Recommended Test Temperatures and Associated Stress Level.
Test Temperature (°F [°C]) Test Stress Level (psi [kPa])
77 (25) 30 (207)
104 (40) 20 (138)
99
3.2 Place the test sample in the environmental chamber and allow it to equilibrate to the
specified testing temperature. A dummy specimen with a temperature sensor mounted at
the center can be monitored to determine when the specimen reaches the specified test
temperature. In the absence of the dummy specimen, Table B2 provides a recommended
temperature equilibrium time for samples starting from room temperature (77°F).
Table B2. Recommended Equilibrium Times.
Test Temperature (°F [°C]) Time (min.)
77 (25) 10
104 (40) 30
3.3 After temperature equilibrium is reached, place one of the friction-reducing end
treatments on top of the platen at the bottom of the loading frame. Place the sample on
top of the lower end treatment, and mount the axial LVDTs to the studs glued to the
sample. Adjust the LVDT to near the end of its linear range to allow the full range to be
available for the accumulation of compressive permanent deformation.
3.4 Place the upper friction-reducing end treatment and platen on top of the sample. Center
the specimen with the load actuator visually in order to avoid eccentric loading.
3.5 Apply a contact load equal to 5 percent of the total load level that will be applied to the
specimen, while monitoring the proper response of the LVDTs (i.e., check for proper
direction sensing for all LVDTs).
3.6 Close the environmental chamber and allow sufficient time (normally 10 to 15 minutes)
for the temperature to stabilize within the specimen and the chamber.
3.7 After the time required for the sample to reach the testing temperature, apply the
haversine load that yields the desired stress on the specimen. The procedure uses a
loading cycle of 1.0 Hz frequency and consists of applying 0.1 second haversine load
followed by a 0.9 second rest period. The maximum applied load (Pmax) is the maximum
100
total load applied to the sample, including the contact and cyclic load: Pmax = Pcontact +
Pcyclic.
3.8 The contact load (Pcontact) is the vertical load placed on the sample to maintain a positive
contact between loading strip and the sample: Pcontact = 0.05 x Pmax.
3.9 The cyclic load (Pcyclic) is the load applied to the test sample that is used to calculate the
permanent deformation parameters: Pcyclic = Pmax - Pcontact.
3.10 Apply the haversine load (Pcyclic) and continue until 5000 cycles or until the sample fails
and results in excessive tertiary deformation, whichever comes first.
3.11 During the load applications, record the load applied and the axial deflection measured
from all LVDTs through the data acquisition system. All data should be collected in real
time and collected so as to minimize phase errors due to sequential channel sampling. It
is recommended to use the data acquisition of the cycles shown in Table B3.
101
Table B3. Suggested Data Collection for VESYS Rutting Test.
Data Collected during Data Collected during Data Collected during

Cycles Cycles Cycles
1 through 10 598 through 600 2723 through 2725
498 through 500 2498 through 2500
4. Calculations
4.1 Calculate the average axial deformation for each specimen by averaging the readings
from the two axial LVDTs. Convert the average deformation values to total axial strain
by dividing by the gauge length (100 mm [4 inch]).
4.2 Compute the cumulative axial permanent strain and resilient strain (εr) at the 100th load
repetition.
4.3 Plot the cumulative axial permanent strain versus number of loading cycles in log-log
space (Figure B3). Determine the permanent deformation parameters, intercept (a), and
slope (b) from the linear portion of the permanent strain curve.
4.4 Compute the rutting parameters: Alpha, Mu:
102
ab
μ=
εr
α = 1− b
5. Report
Report all sample information including mix identification, dates of manufacturing (or cored)
and testing, sample diameter and length, volumetric properties, stress levels used, and axial
permanent deformation parameters: α, μ (or εr, a, b).
10000
Permanent Microstrain
1000
a=67.41 b=0.3895
100
10
1 10 100 1000 10000
Load Repetitions
Figure B3. Plot of Regression Constants “a” and “b” from Log Permanent Strain – Log
Number of Loading Cycles.
Example: Alpha and Mu Calculation
εr = 88.1250
A = 67.4100
b = 0.3895
μ = a x b/εr =67.41 x 0.3895/88.125 = 0.2979
α = 1 – b = 1 – 0.3895 = 0.6105
103
APPENDIX C
RECOMMENDED PERMANENT DEFORMATION AND RESILIENT

MODULUS LABORATORY TEST PROTOCOLS FOR UNBOUND
GRANULAR BASE/SUBBASE MATERIALS AND SUBGRADE SOILS
1. Scope
1.1. This test method describes the laboratory preparation and testing procedures for the
determination of permanent deformation and resilient modulus (Mr) of unbound
granular base/subbase materials and subgrade soils for pavement performance
prediction. The stress conditions used in the test represent the ranges of stress states
likely to be developed beneath flexible pavements subjected to moving wheel loads.
This test procedure has been adapted from the standard test methods given in the
VESYS user manual, National Cooperative Highway Research Program (NCHRP)
1-28A Draft Report (unpublished), and AASHTO designation: T294-92, TP46, and
T292-91.
1.2. The methods described herein are applicable to laboratory-molded samples of
unbound granular base/subbase materials and subgrade soils.
1.3. In this test procedure, stress states used for permanent deformation and resilient
modulus testing are based upon whether the specimen is located in the base/subbase or
the subgrade. Specimen size for testing depends upon the maximum particle size of
the material.
1.4. The values of permanent deformation and resilient modulus determined from these
procedures are the measures of permanent deformation properties and the elastic
modulus of unbound granular base/subbase materials and subgrade soils with the
consideration of their stress-dependency.
1.5. Resilient modulus values can be used with structural response analysis models to
calculate the pavement structural response to wheel loads, and with the combination of
permanent deformation properties and pavement design procedures to predict rutting
performance.
105
1.6. This standard may involve hazardous materials, operations, and equipment. This
standard does not purport to address all of the safety concerns associated with its use.
It is the responsibility of the user of this standard to consult and establish appropriate
safety and health practices and determine the applicability of regulatory limitations
prior to use.
2. Referenced Documents
2.1. AASHTO Standards:

T88 Particle Size Analysis of Soils
T89 Determining the Liquid Limit of Soils
T90 Determining the Plastic Limit and the Plasticity Index of Soils
T100 Specific Gravity of Soils
T180 Moisture-Density Relations of Soils using a 454 kg (10 lb) Rammer and 457 mm
(18 inch) Drop
T233 Density of Soil-in-Place by Block, Chunk, or Core Sampling
T292-91 Resilient Modulus of Subgrade Soils and Untreated Base/Subbase Materials
T296 Strength Parameters of Soils by Triaxial Compression
T265 Laboratory Determination of Moisture Content of Soils
3. Terminology
3.1. Unbound Granular Base and Subbase Materials – These include soil-aggregate
mixtures and naturally occurring materials. No binding or stabilizing agent is used to
prepare unbound granular base or subbase layers. These materials are classified as
Type 1 and Type 2, as subsequently defined in 3.3 and 3.4.
3.2. Subgrade – Subgrade soils may be naturally occurring or prepared and compacted
before the placement of subbase and/or base layers. These materials are classified as
Type 1, Type 2, and Type 3, as subsequently defined in 3.3, 3.4, and 3.5.
3.3. Material Type 1 – Includes all unbound granular base and subbase materials and all
untreated subgrade soils with maximum particle sizes greater than 9.5 mm (3/8 inch).
All material greater than 25.4 mm (1.0 inch) shall be scalped off prior to testing.
106
Materials classified as Type 1 shall be molded in either a 152 mm (6 inch) diameter
mold or a 102 mm (4 inch) diameter mold. Materials classified as Type 1 shall be
compacted by impact or vibratory compaction.
3.4. Material Type 2 – Includes all unbound granular base and subbase materials and all
untreated subgrade soils that have a maximum particle size less than 9.5 mm (3/8 inch)
and that meet the criteria of less than 10 percent passing the 75 µm (No. 200) sieve.
Materials classified as Type 2 shall be molded in a 102 mm (4 inch) diameter mold
and compacted by vibratory compaction.
3.5. Material Type 3 – Includes all untreated subgrade soils that have a maximum particle
size less than 9.5 mm (3/8 inch) and that meet the criteria of more than 10 percent
passing the 75 mm (No. 200) sieve. Materials classified as Type 3 shall be molded in
a 102 mm (4 inch) diameter mold and compacted by impact compaction.
3.6. Permanent Deformation – Permanent deformation is determined by repeated load
compression tests on specimens of the unbound materials. Permanent deformation is
the unrecovered deformation during the testing.
3.7. Resilient Modulus – The resilient modulus is determined by repeated load
compression tests on test specimens of the unbound materials. Resilient modulus (Mr)
is the ratio of the peak axial repeated deviator stress to the peak recoverable axial
strain of the specimen.
3.8. Loading Wave Form – Test specimens are loaded using a haversine load pulse with
0.1-second loading and 0.9-second rest period.
3.9. Maximum Applied Axial Load (Pmax) – The load applied to the sample consisting of
the contact load and cyclic load (confining pressure is not included):
Pmax = Pcontact + Pcyclic
3.10. Contact Load (Pcontact) – Vertical load placed on the specimen to maintain a positive
contact between the loading ram and the specimen top cap. The contact load includes
the weight of the top cap and the static load applied by the ram of the loading system.
3.11. Cyclic Axial Load – Repetitive load applied to a test specimen:
Pcyclic = Pmax – Pcontact
107
3.12. Maximum Applied Axial Stress (Smax) – The axial stress applied to the sample
consisting of the contact stress and the cyclic stress (the confining stress is not
included):
Smax = Pmax/A
where: A = cross sectional area of the sample.
3.13. Cyclic Axial Stress – Cyclic (resilient) applied axial stress:
Scyclic = Pcyclic/A
3.14. Contact Stress (Scontact) – Axial stress applied to a test specimen to maintain a positive
contact between the specimen cap and the specimen:
Scontact = Pcontact /A
The contact stress shall be maintained so as to apply a constant anisotropic confining
stress ratio:
(Scontact + S3)/S3 = 1.2
where: S3 = confining pressure.
3.15. S3 is the applied confining pressure in the triaxial chamber (i.e., the minor principal
stress σ3).
3.16. er is the resilient (recoverable) axial deformation due to Scyclic.
3.17. εr is the resilient (recoverable) axial strain due to Scyclic:
εr = er/L
where: L = distance between measurement points for resilient axial deformation, er.
3.18. ep is the permanent (unrecoverable) axial deformation due to Scyclic.
3.19. εp is the permanent (unrecoverable) axial strain due to Scyclic:
εp = ep/L
where: L = distance between measurement points for permanent axial deformation, ep.
3.20. Resilient Modulus (Mr) is defined as:
Mr = Scyclic/εr
3.21. Load duration is the time interval the specimen is subjected to a cyclic stress pulse.
3.22. Cycle duration is the time interval between the successive applications of a cyclic
stress (usually 1.0 second).
108
4. Summary of Method
4.1. A repeated axial stress of fixed magnitude, load duration, and cycle duration is applied
to a cylindrical test specimen. The test is performed in a triaxial cell, and the
specimen is subjected to a repeated (cyclic) stress and a constant confining stress
provided by means of cell air pressure. Both total resilient (recoverable) and
permanent axial deformation responses of the specimen are recorded and used to
calculate the permanent deformation properties and the resilient modulus.
5. Significance and Use
5.1. The resilient modulus test results provide a basic constitutive relationship between
stiffness and stress state of pavement materials for use in the structural analysis of
layered pavement systems. Furthermore, permanent deformation properties of
pavement materials also can be determined from the first 10,000 cycles of the repeated
load test. The information is critical for pavement rutting performance prediction.
The permanent deformation and resilient modulus tests simulate the conditions in a
pavement with the application of moving wheel loadings.
6. Permanent Deformation and Resilient Modulus Test Apparatus
6.1. Triaxial Pressure Chamber – The pressure chamber is used to contain the test
specimen and the confining fluid during the test. A typical triaxial chamber suitable
for use in resilient modulus testing of soils is shown in Figure C1. The axial
deformation is measured internally, directly on the specimen, using normal gauges
with rubber bands (shown in Figure C2), an optical extensometer, non-contact sensors,
or clamps. For soft and very soft subgrade specimens (i.e., Su < 36 kPa or 750 psf,
where Su is the undrained shear strength of the soil), rubber bands or clamps should
not be used since they may damage the specimen. However, a pair of LVDTs
extending between the top and bottom platens can be used to measure axial
deformation of these weak soils.
109
Figure C1. Triaxial Cell and Test System.
Figure C2. Sample with Instruments.
6.1.1. Air shall be used in the triaxial chamber as the confining fluid for all testing.
6.1.2. The chamber shall be made of suitable transparent material (such as
polycarbonate).
110
6.2. Loading Device – The loading device shall be a top-loading, closed-loop electro-
hydraulic testing machine with a function generator that is capable of applying
repeated cycles of a haversine-shaped load pulse. Each pulse shall have a 0.1 second
duration followed by a rest period of 0.9 second duration for base/subbase materials
and 0.2 second duration followed by a rest period of 0.8 second duration for subgrade
materials. For non-plastic granular material, it is permissible, if desired, to reduce the
rest period to 0.4 second to shorten testing time; the loading time may be increased to
0.15 second if required.
6.2.1. All conditioning and testing shall be conducted using a haversine-shaped load
pulse. The electro-hydraulic system generated haversine waveform and the
response waveform shall be displayed to allow the operator to adjust the gains
to ensure they coincide during conditioning and testing.
6.3. Load and Specimen Response Measuring Equipment:
6.3.1. The axial load measuring device should be an electronic load cell, which is
preferred to be located inside the triaxial cell. The load cell should have the
capacities presented in Table C1.
Table C1. Load Cell Capacity.
Sample Diameter mm Max. Load Capacity kN (lb) Required Accuracy N (lb)

(in)
102 (4.0) 8.9 (2000) ±17.8 (±4)
152 (6.0) 22.24 (5000) ±22.24 (±5)
Note 1 – During periods of permanent deformation and resilient modulus testing, the load
cell shall be monitored and checked once every two weeks or after every 50 permanent
deformation and resilient modulus tests with a calibrated proving ring to assure that the load
cell is operating properly. An alternative to using a proving ring is to inset an additional
calibrated load cell and independently measure the load applied by the original cell.
Additionally, the load cell shall be checked at any time there is a suspicion of a load cell
problem. The testing shall not be conducted if the testing system is found to be out of
calibration.
111
6.3.2. The chamber pressures shall be monitored with conventional pressure gauges,
manometers, or pressure transducers accurate to 0.69 kPa (0.1 psi).
6.3.3. Axial Deformation: Axial deformation is to be measured with displacement
transducers referenced to gauge points contacting the specimen with rubber
bands as shown in Figure C2. Deformation shall be measured over
approximately the middle half of the specimen. Axial deformations shall be
measured at a minimum of two locations 180 degrees apart (in a plain view),
and a pair of spring-loaded LVDTs are placed on the specimen at quarter
points. Spring-loaded LVDTs shall be used to maintain a positive contact
between the LVDTs and the surface on which the tips of the transducers rest.
Note 2 – Table C2 summarizes the specifications for spring-loaded LVDTs.
Table C2. Specifications for Axial LVDTs.
Min. Approximate Resilient

Material/specimen Diameter (inch) Range Specimen Displacement
(inch) (inch)
6 ±0.25 0.001
Aggregate Base
4 ±0.10 0.00065
Subgrade Soil
4 ±0.25 0.0014
(sand and cohesive)
Note: For soft subgrade soil, permanent and resilient displacement shall be measured
over entire specimen height.
Note 3 – Misalignment or dirt on the shaft of the transducer can cause the shafts of the
LVDTs to stick. The laboratory technician shall depress and release each LVDT back and
forth a number of times prior to each test to assure that they move freely and are not sticking.
A cleaner/lubricant specified by the manufacturer shall be applied to the transducer shafts on
a regular basis.
6.3.4. Data Acquisition: An analog-to-digital (A/D) data acquisition system is

required. The overall system should include automatic data reduction to
minimize production. Suitable signal excitation, conditioning, and recording
equipment is required for simultaneous recording of axial load and
deformations. The system should meet or exceed the following additional
requirements: (1) 25 μs A/D conversion time; (2) 12-bit resolution; (3) single-
112
or multiple-channel throughput (gain = 1), 30 kHz; (4) software selectable
gains; (5) measurement accuracy of full scale (gain = 1) of ±0.02 percent; and
(6) non-linearity of ±0.5 percent. The signal shall be clean and free of noise.
Filtering the output signal during or after data acquisition is discouraged. If a
filter is used, it should have a frequency higher than 10 to 20 Hz. A
supplemental study should be made to ensure correct peak readings are
obtained from filtered data compared to unfiltered data. A minimum of 200
data points from each LVDT shall be recorded per load cycle.
6.4. Specimen Preparation Equipment–A variety of equipment is required to prepare
compacted specimens that are representative of field conditions. Use of different
materials and different methods of compaction in the field requires the use of varying
compaction techniques in the laboratory.
6.5. Miscellaneous Apparatus–This includes calipers, micrometer gauge, steel rule
(calibrated to 0.5 mm [0.02 inch]), rubber membranes from 0.25 to 0.79 mm (0.02 to
0.031 inch) thickness, rubber O-rings, vacuum source with bubble chamber and
regulator, membrane expander, porous stones (subgrade), 6.4 mm (0.25 inch) thick
porous stones or bronze discs (base/subbase), scales, moisture content cans, and data
sheets.
6.6. Periodic System Calibration–The entire system (transducers, signal conditioning, and
recording devices) shall be calibrated every two weeks or after every 50 tests. Daily
and other periodic checks of the system may also be performed as necessary. No
permanent deformation and resilient modulus testing will be conducted unless the
entire system meets the established calibration requirements.
7. Preparation of Test Specimens
7.1. The following guidelines, based on the sieve analysis test results, shall be used to
determine the test specimen size:
7.1.1. Use 152 mm (6.0 inch) diameter and 305 mm (12 inch) high specimens for all
materials with maximum particle sizes greater than 19 mm (0.75 inch). All
113
material of particle size greater than 25.4 mm (1.0 inch) shall be scalped off
prior to testing.
7.1.2. Use 102 mm (4.0 inch) diameter and 204 mm (8.0 inch) high specimens for all
materials with maximum particle sizes less than 19 mm (0.75 inch).
7.2. Laboratory Compacted Specimens–Reconstituted test specimens of all types shall be
prepared to the specified or in situ dry density (γd) and moisture content (w).
Laboratory compacted specimens shall be prepared for all unbound granular base and
subbase material and for all subgrade soils.
7.2.1. Moisture Content–For in situ materials, the moisture content of the laboratory
compacted specimen shall be the in situ moisture content for that layer
obtained in the field using T238. If data are not available on in situ moisture
content, refer to Section 7.2.3.
7.2.1.1. The moisture content of the laboratory compacted specimen should
not vary from the required value by more than ±0.5 percent for all
materials.
7.2.2. Compacted Density–The density of a compacted specimen shall be the in-
place dry density obtained in the field for that layer using T239 or other
suitable methods. If these data are not available on in situ density, then refer
to Section 7.2.3.
7.2.2.1. The dry density of a laboratory compacted specimen should not
vary more than ±1.0 percent from the target dry density for that
layer.
7.2.3. If either the in situ moisture content or the in-place dry density is not
available, then use the optimum moisture content and 95 percent of the
maximum dry density by using T180 for the base/subbase and 95 percent of
T99 for the subgrade.
7.2.3.1. The moisture content of the laboratory compacted specimen should
not vary from the required value by more than ±0.5 percent for all
materials. The dry density of a laboratory compacted specimen
114
should not vary more than ±1.0 percent from the target dry density
for that layer.
7.2.4. Sample Reconstitution – Reconstitute the specimen for all materials. The
target moisture content and density to be used in determining needed material
qualities are given in Section 7.2. After this step is completed, specimen
compaction can begin.
7.3. Compaction Methods and Equipment for Reconstituting Specimens:
7.3.1. Specimens of Type 1 materials shall be compacted by vibratory or impact
compaction. The general method of vibratory compaction is given in T292-
91. The general method of impact compaction is given in T292.
7.3.2. Specimens of Type 2 materials shall be compacted by vibratory compaction.
The general method of vibratory compaction is presented in T292-92.
7.3.3. Specimens of Type 3 materials shall be compacted by impact compaction.
The general method of impact compaction is given in T292-91.
8. Test Procedure
Following this test procedure, a permanent deformation and resilient modulus test is
performed on all materials using a triaxial cell (confined).
8.1. Base/Subbase Materials

The procedure described in this section applies to all unbound granular base and
subbase materials.
Apparatus and Sample Preparation

8.1.1. Assembly of the triaxial cell: If not already in place, place the specimen with
end platens into position on the pedestal of the triaxial cell. Proper
positioning of the specimen is extremely critical in applying a concentric load
to the specimen. Couple the loading device to the specimen using a smooth
steel ball. To center the specimen, slowly rotate the ball as the clearance
between the load piston ball decreases and a small amount of load is applied
to the specimen. Be sure the ball is concentric with the piston that applies the
115
load (watch the gap around the ball). Shift the specimen laterally to achieve a
concentric loading.
8.1.2. Check and adjust the axial displacement measurement system, load cell, and
data acquisition system, and make sure they are working properly.
8.1.3. If not already connected, connect the confining air pressure supply line to the
triaxial chamber.
8.1.4. Open all valves on drainage lines leading to the inside of the specimen. This is
necessary to develop confining pressure on the specimen.
8.1.5. Apply the specified conditioning confining pressure of 103.5 kPa (15.0 psi) to
the test specimen. A contact stress equal to 20 percent of the confining
pressure shall be applied to the specimen so that the load piston stays in
contact with the top platen at all times.
8.1.6. Preconditioning: Apply 100 repetitions of a load equivalent to a maximum
axial stress of 41.4 kPa (6 psi) and a corresponding cyclic stress of 20.7 kPa
(3 psi) using a haversine-shaped, 0.1 second load pulse followed by 0.9
second rest period.
Permanent Deformation Test

8.1.7. Apply the haversine loading (Pcyclic) equivalent to a maximum axial stress of
227.7 kPa (33 psi) and a corresponding cyclic stress of 207 kPa (30 psi) using
a haversine-shaped, 0.1 second load pulse followed by 0.9 second rest period,
and continue until 10,000 cycles (2.8 hours) or until the specimen fails and the
vertical permanent strain reaches 5 percent during the testing, whichever
comes first. The total number of cycles or the testing time will depend on the
stress levels applied.
8.1.8. During the load applications, record the load applied and the axial
deformation measured from two LVDTs through the data acquisition system.
Signal-to-noise ratio should be at least 10. All data should be collected in real
time and collected/processed so as to minimize phase errors due to sequential
channel sampling. In order to save storage space during data acquisition for
116
10,000 cycles, researchers recommend using the data acquisition of the cycles
shown in Table C3.
Table C3. Suggested Data Collection for Triaxial Repeated Load Permanent
Deformation Test for Granular Base and Subbase.
Data Collection Data Collection Data Collection Data Collection

During Cycles During Cycles During Cycles During Cycles
1-15 450 1300 4000
20 500 1400 4500
30 550 1500 5000
40 600 1600 5500
60 650 1700 6000
80 700 1800 6500
100 750 1900 7000
130 800 2000 7500
160 850 2200 8000
200 900 2400 8500
250 950 2600 9000
300 1000 2800 9500
350 1100 3000 10000
400 1200 3500
Resilient Modulus Test
Specimen Testing–If the vertical permanent strain has neither reached 5

percent nor failed during permanent deformation test, use the same specimen
to perform the resilient modulus test following the load sequence shown in
Table C4. Begin by decreasing the maximum axial stress to 14.5 kPa (2.1 psi)
(Sequence No. 1 Table C4), and set the confining pressure to 20.7 kPa (3
psi).
If the vertical permanent strain has reached 5 percent or failed during
permanent deformation test, mold a new specimen, and then go back to
section 8.1.1. In addition, reduce the load repetitions from 10,000 to 5000
117
during the repeated load permanent deformation test. If the sample again
reaches 5 percent total vertical permanent strain during repeated load test,
then the test shall be terminated. No further testing of this material is
necessary. If not, perform the resilient modulus test following the load
sequence shown in Table C4. Begin by decreasing the maximum axial stress
to 14.5 kPa (2.1 psi) (Sequence No. 1 Table C4) and set the confining pressure
to 20.7 kPa (3 psi).
8.1.10. Apply 100 repetitions of the corresponding cyclic axial stress using a
haversine-shaped load pulse consisting of a 0.1 second load followed by a 0.9
second rest period. Record the average recovered deformations from each
LVDT separately for the last five cycles.
8.1.11. Increase the maximum axial stress to 30 kPa (4.2 psi) and the confining
pressure to 41.4 kPa (6 psi) (Sequence No. 2 Table C4), and repeat the
previous step at this new stress level.
8.1.12. Continue the test for the remaining stress sequences in Table C4 (3 to 30)
recording the vertical recovered deformation. If at any time the total
permanent strain of the sample exceeds 5 percent, stop the test and report the
result on the appropriate worksheet.
118
Table C4. Permanent Deformation and Resilient Modulus Test Sequence for Granular
Base and Subbase.
Confining Pressure Contact Stress Cyclic Stress Maximum Stress

Sequence Nrep.
kPa psi kPa psi kPa psi kPa psi
Preconditioning 103.5 15.0 20.7 3.0 20.7 3.0 41.4 6.0 100
Permanent
103.5 15.0 20.7 3.0 207.0 30.0 227.7 33.0 10000
Deformation
1 20.7 3.0 4.1 0.6 10.4 1.5 14.5 2.1 100
2 41.4 6.0 8.3 1.2 20.7 3.0 29.0 4.2 100
3 69.0 10.0 13.8 2.0 34.5 5.0 48.3 7.0 100
4 103.5 15.0 20.7 3.0 51.8 7.5 72.5 10.5 100
5 138.0 20.0 27.6 4.0 69.0 10.0 96.6 14.0 100
6 20.7 3.0 4.1 0.6 20.7 3.0 24.8 3.6 100
7 41.4 6.0 8.3 1.2 41.4 6.0 49.7 7.2 100
8 69.0 10.0 13.8 2.0 69.0 10.0 82.8 12.0 100
9 103.5 15.0 20.7 3.0 103.5 15.0 124.2 18.0 100
10 138.0 20.0 27.6 4.0 138.0 20.0 165.6 24.0 100
11 20.7 3.0 4.1 0.6 41.4 6.0 45.5 6.6 100
12 41.4 6.0 8.3 1.2 82.8 12.0 91.1 13.2 100
13 69.0 10.0 13.8 2.0 138.0 20.0 151.8 22.0 100
14 103.5 15.0 20.7 3.0 207.0 30.0 227.7 33.0 100
15 138.0 20.0 27.6 4.0 276.0 40.0 303.6 44.0 100
16 20.7 3.0 4.1 0.6 62.1 9.0 66.2 9.6 100
17 41.4 6.0 8.3 1.2 124.2 18.0 132.5 19.2 100
18 69.0 10.0 13.8 2.0 207.0 30.0 220.8 32.0 100
19 103.5 15.0 20.7 3.0 310.5 45.0 331.2 48.0 100
20 138.0 20.0 27.6 4.0 414.0 60.0 441.6 64.0 100
21 20.7 3.0 4.1 0.6 103.5 15.0 107.6 15.6 100
22 41.4 6.0 8.3 1.2 207.0 30.0 215.3 31.2 100
23 69.0 10.0 13.8 2.0 345.0 50.0 358.8 52.0 100
24 103.5 15.0 20.7 3.0 517.5 75.0 538.2 78.0 100
25 138.0 20.0 27.6 4.0 690.0 100.0 717.6 104.0 100
26 20.7 3.0 4.1 0.6 144.9 21.0 149.0 21.6 100
27 41.4 6.0 8.3 1.2 289.8 42.0 298.1 43.2 100
28 69.0 10.0 13.8 2.0 483.0 70.0 496.8 72.0 100
29 103.5 15.0 20.7 3.0 724.5 105.0 745.2 108.0 100
30 138.0 20.0 27.6 4.0 966.0 140.0 993.6 144.0 100
119
8.1.13. At the completion of this test, reduce the confining pressure to zero, and
remove the sample from the triaxial chamber.
8.1.14. Remove the membrane from the specimen, and use the entire specimen to
determine moisture content in accordance with T265.
8.2. Coarse-Grained Subgrade Soils
This procedure is used for all laboratory compacted specimens of subgrade soils for
which the percent passing 75 µm (No. 200) sieve is less than 35 percent.
Reconstructed specimens will usually be compacted directly on the pedestal of the
triaxial cell.
8.2.1. Assembly of the triaxial cell: Refer to section 8.1.1.
8.2.2. Set up the axial displacement measurement system, and verify it is working
properly.
8.2.3. If not already connected, connect the confining air pressure supply line to the
triaxial chamber.
8.2.4. Open all valves on drainage lines leading to the inside of the specimen. This
is necessary to develop confining pressure on the specimen.
8.2.5. Apply the specified conditioning confining pressure of 27.6 kPa (4.0 psi) to
the test specimen. Apply a contact stress equal to 20 percent of the confining
pressure to the specimen so that the load piston stays in contact with the top
platen at all times.
8.2.6. Preconditioning: Apply 100 repetitions of a load equivalent to a maximum
axial stress of 12.4 kPa (1.8 psi) and a corresponding cyclic stress of 6.9 kPa
second rest period.
120
8.2.7. Apply the haversine loading (Pcyclic) equivalent to a maximum axial stress of
60.7 kPa (8.8 psi) and a corresponding cyclic stress of 55.2 kPa (8 psi) using a
haversine-shaped, 0.2 second load pulse followed by 0.8 second rest period,
and continue until 10,000 cycles (2.8 hours) or until the specimen fails and/or
the vertical permanent strain reaches 5 percent during the testing, whichever
deformation measured from two LVDTs through the data acquisition system.
All data should be collected in real time and collected/processed so as to
minimize phase errors due to sequential channel sampling. In order to save
storage space during data acquisition for 10,000 cycles, it is recommended to
use the data acquisition of the cycles shown in Table C5.
Deformation Test for Granular Subgrades.
Data Collection During Data Collection During Data Collection Data Collection
Cycles Cycles During Cycles During Cycles
1-15 450 1300 4000
20 500 1400 4500
30 550 1500 5000
40 600 1600 5500
60 650 1700 6000
80 700 1800 6500
100 750 1900 7000
130 800 2000 7500
160 850 2200 8000
200 900 2400 8500
250 950 2600 9000
300 1000 2800 9500
350 1100 3000 10000
400 1200 3500
121
8.2.9. Specimen Testing–If the vertical permanent strain has neither reached 5
(Sequence No. 1 Table C6), and set the confining pressure to 13.8 kPa (2 psi).
If the vertical permanent strain has reached 5 percent or failed during

section 8.2.1. In addition, reduce the load repetitions from 10,000 to 5,000
reaches 5 percent total vertical permanent strain during the repeated load test,
then terminate the test. No further testing of this material is necessary. If not,
perform the resilient modulus test following the load sequence shown in Table
C6. Begin by decreasing the maximum axial stress to 9.66 kPa (1.4 psi)
122
Table C6. Permanent Deformation and Resilient Modulus Test Sequence for Granular
Subgrades.

Sequence Nrep
kPa Psi kPa psi kPa Psi kPa psi
Permanent
27.6 4.0 5.5 0.8 55.2 8.0 60.7 8.8 10,000
Deformation
1 13.8 2.0 2.8 0.4 6.9 1.0 9.7 1.4 100
2 27.6 4.0 5.5 0.8 13.8 2.0 19.3 2.8 100
3 41.4 6.0 8.3 1.2 20.7 3.0 29.0 4.2 100
4 55.2 8.0 11.0 1.6 27.6 4.0 38.6 5.6 100
5 82.8 12.0 16.6 2.4 41.4 6.0 58.0 8.4 100
6 13.8 2.0 2.8 0.4 13.8 2.0 16.6 2.4 100
7 27.6 4.0 5.5 0.8 27.6 4.0 33.1 4.8 100
8 41.4 6.0 8.3 1.2 41.4 6.0 49.7 7.2 100
9 55.2 8.0 11.0 1.6 55.2 8.0 66.2 9.6 100
10 82.8 12.0 16.6 2.4 82.8 12.0 99.4 14.4 100
11 13.8 2.0 2.8 0.4 27.6 4.0 30.4 4.4 100
12 27.6 4.0 5.5 0.8 55.2 8.0 60.7 8.8 100
13 41.4 6.0 8.3 1.2 82.8 12.0 91.1 13.2 100
14 55.2 8.0 11.0 1.6 110.4 16.0 121.4 17.6 100
15 82.8 12.0 16.6 2.4 165.6 24.0 182.2 26.4 100
16 13.8 2.0 2.8 0.4 41.4 6.0 44.2 6.4 100
17 27.6 4.0 5.5 0.8 82.8 12.0 88.3 12.8 100
18 41.4 6.0 8.3 1.2 124.2 18.0 132.5 19.2 100
19 55.2 8.0 11.0 1.6 165.6 24.0 176.6 25.6 100
20 82.8 12.0 16.6 2.4 248.4 36.0 265.0 38.4 100
8.2.11. Increase the maximum axial stress to 19.32 kPa (2.8 psi), and set the
123
confining pressure to 27.6 kPa (4 psi) (Sequence No. 2 Table C6), and repeat
the previous step at this new stress level.
8.3. Cohesive Subgrade Soils

This procedure is used for all laboratory compacted specimens of subgrade soils for
which the percent passing 75 µm (No. 200) sieve is greater than 35 percent.
Reconstructed specimens will usually be compacted directly on the pedestal of the
triaxial cell.

8.3.1. Assembly of the Triaxial Cell: Refer to section 8.1.1.
8.3.2. Stiff to Very Stiff Specimens–For stiff and very stiff cohesive specimens (Su >
36 kPa (750 psf), here Su designates the undrained shear strength of the soil),
axial deformation should preferably be measured either directly on the
specimen or between the solid end platens using grouted specimen ends.
8.3.3. Soft Specimens–The axial deformation of soft subgrade soils (Su < 36 kPa
[750 psf]) should not be measured using a rubber band circled on the
specimen. If the measured resilient modulus is less than 69,000 kPa (10,000
psi), axial deformation can be measured between top and bottom platens. An
empirical correction is not required for irregular specimen end contacts for
these low modulus soils. If the resilient modulus is greater than 69,000 kPa
(10,000 psi), follow the procedure in section 8.3.2.
124
8.3.4. Install Axial Displacement Device: Carefully install the axial displacement
instrumentation selected under 8.3.2 or 8.3.3. For top to bottom displacement
measurement, attach the LVDTs or proximity gauges on steel or aluminum
bars extending between the top and bottom platens. If the rubber band or
clamps are used, place the rubber band or clamps at the quarter points of the
specimen using two height gauges to ensure that clamps are positioned
horizontally at the correct height. Each height gauge can consist of two
circular aluminum rods machined to the correct length. These rods are placed
on each side of the clamp to ensure proper location. Then ensure the
displacement instrumentations are working properly by displacing each device
and observing the resulting voltage output as shown by the data acquisition
system.
8.3.5. Refer to section 8.1.1.
8.3.6. Set up the axial displacement measurement system, and verify it is working
properly.
8.3.7. Open all valves on drainage lines leading to the inside of the specimen. This
is necessary to develop confining pressure on the specimen.
8.3.8 If not already connected, connect the confining air pressure supply line to the
triaxial chamber.
8.3.9 Apply the specified conditioning confining pressure of 27.6 kPa (4.0 psi) to
the test specimen. Apply a contact stress equal to 20 percent of the confining
pressure to the specimen so that the load piston stays in contact with the top
platen at all times.
8.3.10. Preconditioning–Apply 100 repetitions of a load equivalent to a maximum
axial stress of 12.4 kPa (1.8 psi) and a corresponding cyclic stress of 6.9 kPa
second rest period.
8.3.11. Apply the haversine-loading (Pcyclic) equivalent to a maximum axial stress of
125
53.8 kPa (7.8 psi) and a corresponding cyclic stress of 48.3 kPa (7 psi) using a
haversine-shaped, 0.2 second load pulse followed by 0.8 second rest period
and continue until 10,000 cycles (2.8 hours) or until the specimen fails and the
vertical permanent strain reaches 5 percent during the testing, whichever
deformation measured from all LVDTs through the data acquisition system.
Signal-to-noise ratio should be at least 10. All data should be collected in real
time and collected/processed so as to minimize phase errors due to sequential
channel sampling. In order to save storage space during data acquisition for
10,000 cycles, it is recommended to use the data acquisition cycles shown in
Table C7.
Deformation Test for Fine-Grained Subgrades.
Data Collection Data Collection Data Collection Data Collection

During Cycles During Cycles During Cycles During Cycles
1-15 450 1300 4000
20 500 1400 4500
30 550 1500 5000
40 600 1600 5500
60 650 1700 6000
80 700 1800 6500
100 750 1900 7000
130 800 2000 7500
160 850 2200 8000
200 900 2400 8500
250 950 2600 9000
300 1000 2800 9500
350 1100 3000 10000
400 1200 3500
8.3.13. Specimen Testing–If the vertical permanent strain has neither reached 5
126
(Sequence No. 1 Table C8), and set the confining pressure to 55.2 kPa (8
psi).
If the vertical permanent strain has reached 5 percent or failed during the
section 8.3.1. In addition, reduce the load repetitions from 10,000 to 5000
reaches 5 percent total vertical permanent strain during the repeated load test,
then terminate the test. No further testing of this material is necessary. If not,
perform the resilient modulus test following the load sequence shown in Table
C4. Begin by decreasing the maximum axial stress to 38.6 kPa (5.6 psi)
8.3.15. Decrease the maximum axial stress to 35.9 kPa (5.2 psi), set the confining
pressure to 41.4 kPa (6 psi) (Sequence No. 2 Table C8), and repeat the
previous step at this new stress level.
127
Table C8. Permanent Deformation and Resilient Modulus Test Sequence for Fine-Grained
Subgrades.

Sequence Nrep.
kPa psi kPa psi KPa psi kPa Psi
Permanent
27.6 4.0 5.5 0.8 48.3 7.0 53.8 7.8 10,000
Deformation
1 55.2 8.0 11.0 1.6 27.6 4.0 38.6 5.6 100
2 41.4 6.0 8.3 1.2 27.6 4.0 35.9 5.2 100
3 27.6 4.0 5.5 0.8 27.6 4.0 33.1 4.8 100
4 13.8 2.0 2.8 0.4 27.6 4.0 30.4 4.4 100
5 55.2 8.0 11.0 1.6 48.3 7.0 59.3 8.6 100
6 41.4 6.0 8.3 1.2 48.3 7.0 56.6 8.2 100
7 27.6 4.0 5.5 0.8 48.3 7.0 53.8 7.8 100
8 13.8 2.0 2.8 0.4 48.3 7.0 51.1 7.4 100
9 55.2 8.0 11.0 1.6 69.0 10.0 80.0 11.6 100
10 41.4 6.0 8.3 1.2 69.0 10.0 77.3 11.2 100
11 27.6 4.0 5.5 0.8 69.0 10.0 74.5 10.8 100
12 13.8 2.0 2.8 0.4 69.0 10.0 71.8 10.4 100
13 55.2 8.0 11.0 1.6 96.0 14.0 107.6 15.6 100
14 41.4 6.0 8.3 1.2 96.0 14.0 104.9 15.2 100
15 27.6 4.0 5.5 0.8 96.0 14.0 102.1 14.8 100
16 13.8 2.0 2.8 0.4 96.0 14.0 99.4 14.4 100
9. Calculations
Calculation of Permanent Strain
9.1. Calculate the average axial deformation for each specimen by averaging the readings
from the two axial LVDTs. Convert the average deformation values to total axial
strain by dividing by the gauge length, L (152 mm [6 inch] for 152 mm diameter
sample; 102 mm (4 inch) for 102 mm diameter sample). Typical total axial strain
versus time is shown in Figure C3.
128
9.2. Compute the cumulative axial permanent strain and resilient strain (εr) at 200th load
repetition.
9.3. Plot the cumulative axial permanent strain versus the number of loading cycles in log
space (shown in Figure C4). Determine the permanent deformation parameters,
intercept (a) and slope (b), from the linear portion of the permanent strain curve (log-
log scale), which is also demonstrated on Figure C4.
9.4. Compute the rutting parameters: Alpha, Mu
ab
μ=
εr
α = 1− b
Figure C3. Triaxial Repeated Load Test Results: Strain vs. Number of Load Repetitions.
Figure C4. Permanent Strain vs. Number of Load Repetitions.
129
Calculation of Resilient Modulus
9.5. Perform the calculations to obtain resilient modulus values. The resilient modulus is
computed from each of the last five cycles of each load sequence and then averaged.
The data reduction processes should be fully automated to minimize the chance for
human error.
9.6. Using nonlinear regression techniques fit the following resilient modulus model to
the data obtained from the applied procedure. The equation for the normalized log-
log k1, k2, k3, k6, and k7 model is as follows:
k2 k3
⎛ θ − 3k 6 ⎞ ⎛ τ oct ⎞
M R = k 1 p a ⎜⎜ ⎟⎟ ⎜⎜ + k 7 ⎟⎟
⎝ pa ⎠ ⎝ pa ⎠
k1, k2 ≥ 0
k3, k6 ≤ 0
k7 ≥ 1
where:
MR = resilient modulus
θ = Bulk Stress, θ = σ1 + σ2 + σ3
τoct = Octahedral shear stress,
1
τ oct = (σ 1 − σ 2 )2 + (σ 1 − σ 3 )2 + (σ 2 − σ 3 )2
3
σ1, σ2, σ3 = Principal Stresses
k = Regression constants
pa = atmospheric pressure (14.7 psi)
Assign initial values of zero for k6 and one for k7; restrain all regression constants
130
according to the model. Report the constants k1, k2, k3, k6, and k7, the ratio of the
standard error of estimate, to the standard deviation and the square of the correlation
coefficient.
10. Report
10.1. Permanent Deformation Test:
10.1.1. Report all specimen basic information including specimen identification, dates
of manufacturing and testing, specimen diameter and length, confining
pressure, stress levels used, and axial permanent deformation parameters: α, μ
(or εr, a, and b).
10.2. Resilient Modulus Test:
10.2.1. Report all specimen basic information including specimen identification, dates
of manufacturing and testing, specimen diameter, and length.
10.2.2. Report the average peak stress (σo) and strain (εo) for each confining pressure–
cyclic stress combination tested.
10.2.3. For each confining pressure–cyclic stress combination tested, report the
resilient modulus for each replicate test specimen.
10.2.4. Report nonlinear resilient modulus model and the model parameters: k1, k2,
k3, k6, and k7.
131

Technical Report Documentation Page

Uploaded by

Copyright:

Available Formats

Technical Report Documentation Page

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Technical Report Documentation Page

Uploaded by

Copyright:

Available Formats

Technical Report Documentation Page

1. Report No. 2. Government Accession No. 3. Recipient's Catalog No.

6. Performing Organization Code

7. Author(s) 8. Performing Organization Report No.

17. Key Words 18. Distribution Statement

Tom Scullion, P.E.

Performed in cooperation with the

TEXAS TRANSPORTATION INSTITUTE

The objective of Texas Department of Transportation (TxDOT) Project 0-5798 is to

MODELS FOR PREDICTING FATIGUE CRACKING OF HMA LAYERS

This chapter is divided into the following major sections:

• Section 2.1 provides a description of the concept of fatigue cracking.

2.1 BACKGROUND OF FATIGUE CRACKING

2.2 FATIGUE CRACKING MODELING

2.2.1 Crack Initiation Models

Constant stress: N f = A f [0.0252PI − 0.00126PI(Vb )+ 0.00673Vb − 0.0167]5 ε −t 5 E − 1.4 (4)

Nf = number of repetitions to fatigue cracking,

• Asphalt Institute (MS-1) model

Nf = number of repetitions to fatigue cracking,

• SHRP A-003A-Berkeley fatigue model

Nf = number of repetitions to fatigue cracking,

• MEPDG fatigue cracking model

• The number of the load repetitions fatigue model

Nf = number of repetitions to fatigue cracking,

• Fatigue damage model

FC = percentage of fatigue cracking of total lane area,

Energy-Based Fatigue Models

• Initial dissipated energy approach

N f = 6.72e 0.049VFB (w0 )−2.047 (12)

WN = cumulative dissipated energy to failure,

• Work ratio approach

N1 = 205Vb 6.44 w0−2.01ψ 1N.64

• Dissipated energy ratio approach

PV = 61.336ε 5.052 S 2.749 (VP )1.643 (GP )−0.094 (19)

ε = tensile strain, and

Schapery defined uniaxial pseudo-strain as follows:

From Equations 20 and 21, under LVE conditions, we have:

where S is a variable characteristic of the amount of internal damage in a material, and α is a

C = C10 − C11 (S )C12 (28)

Then, substitute Equation 30 into Equation 26 and integrate to solve for t:

Now, if the reference modulus ER = 1, then, ε R (t ) = σ (t ) . In addition, for sinusoidal

• CalME damage-based fatigue cracking model

α = exp(αA + αB × t + αC × ln (w) + αD × t × ln (w))

The use of SR damage-based approach has several advantages: 1) stiffness is easy to

2.2.2 Crack Propagation Models

Classical Fracture Mechanics Model

(a) (b) (c)

Figure 3. 8-node Quadrilateral and Quarter-point Triangular Elements.

• HMA fracture properties: A, n

In summary, the classical fracture mechanics-based fatigue crack propagation model is

Cohesive Crack/Zone Model

Non-Local continuum Damage Mechanics Model for Crack Propagation (74)

Non-local CDM is essentially an “enhancement” of local-CDM. Thus, the local-CDM is

Equation 40 is complemented by the damage evolution law:

& = g (ω, ~ε ) ~ε&

Implicit gradient formulation is proposed by Peerlings et al., in which a non-local strain

The introduction of Equation 44 leads to a coupled problem between the displacement

2.2.3 Integrated Crack Initiation and Crack Propagation Model

The key issue of estimating Ni is to establish a “bridge” between fracture properties (A