Luo, Xue Luo, Rong Lytton, Robert L. - Energy-Based Mechanistic Approach To Charac

Journal of Materials in Civil Engineering.
Submitted March 7, 2012; accepted August 21, 2012;

posted ahead of print August 30, 2012. doi:10.1061/(ASCE)MT.1943-5533.0000666
An Energy-Based Mechanistic Approach to Characterize Crack

Growth of Asphalt Mixtures
Downloaded from ascelibrary.org by University of Nebraska-Lincoln on 04/02/13. Copyright ASCE. For personal use only; all rights reserved.
Xue Luo, Ph.D., A.M.ASCE
t
Postdoctoral Research Associate
ip
Texas Transportation Institute
Texas A&M University System
d cr
3135 TAMU, CE/TTI Bldg. 508B
College Station, Texas 77843
Phone: (979) 458-8535
te s
Email: xueluo@tamu.edu
di nu
Rong Luo, Ph.D., P.E., M.ASCE
Associate Research Engineer
ye a
Texas Transportation Institute
op M
Texas A&M University System
3135 TAMU, CE/TTI Bldg. 503C
Phone: (979) 845-9897
C ted
Fax: (979) 845-0278

Email: rongluo@tamu.edu
ot p
Robert L. Lytton, Ph.D., P.E., F.ASCE

Professor, Fred J. Benson Chair
N ce
Zachry Department of Civil Engineering

Texas A&M University
3136 TAMU, CE/TTI Bldg. 503A
Ac

Phone: (979) 845-9964
Fax: (979) 845-0278
Email: r-lytton@civil.tamu.edu
Copyright 2012 by the American Society of Civil Engineers

J. Mater. Civ. Eng.
Journal of Materials in Civil Engineering. Submitted March 7, 2012; accepted August 21, 2012;
Luo et al. 2
Abstract
Fatigue cracking is one of the most common distresses of asphalt pavements. In order to reduce
and control such distress, the fatigue crack growth of asphalt mixtures must be predicted.
However, most existing fatigue crack growth models are generally empirical or
t
ip
phenomenological in nature. To fill this gap, this paper aims at developing an energy-based
mechanistic approach to model the fatigue crack growth in asphalt mixtures. The core of this
d cr
approach is establishing the energy balance equations between the apparent energy of the bulk
te s
specimen and the true energy of the intact material.
di nu
A controlled-strain repeated direct tension (RDT) test is used to generate fatigue cracking
ye a
damage in an asphalt mixture specimen. The true stress/true strain/true pseudo strain of both the
op M
nondestructive and destructive RDT tests are simulated through the force equilibrium and energy
balance equations. The true stress in a damaged asphalt mixture is found to be the driving force
C ted
for the crack growth; the crack growth in turn aggravates the localization of the true stress/true
strain. After obtaining the true stress, the true material properties are determined to obtain the
ot p
true strain/true pseudo strain. The true material properties include nondestructive and destructive
N ce
true material properties, which are obtained from the nondestructive and destructive tests,
respectively, based on the energy balance equations.

Ac
The ratio of the true stress and the apparent stress in a damaged asphalt mixture is used to
calculate the damage density for every loading cycle of the destructive RDT test. The evolution
of the damage density with repeated loading demonstrates the development of fatigue cracking in
the asphalt mixture.

J. Mater. Civ. Eng.
Luo et al. 3
In addition to modeling fatigue cracking in asphalt mixtures, the energy-based
mechanistic approach developed in this paper can be used with a wide range of tests to predict
crack growth of different types of materials because of its mechanical nature.

t
ip
d cr
te s
di nu
ye a
op M
C ted
ot p
N ce
Ac

J. Mater. Civ. Eng.
Luo et al. 4
Introduction
Fatigue cracking is one of the most common distresses of asphalt pavements. The asphalt layer
of the pavement does not crack immediately after the pavement is opened to traffic, but after
many load applications cracks start to appear. As the loading application continues, the cracks
t
ip
develop and connect to form a series of interconnected pieces, which is usually called “alligator
cracking” that resembles the back of an alligator. In more serious cases, the alligator cracking
d cr
will further deteriorate to potholes, allowing considerable moisture infiltration into the pavement
te s
and significantly reducing the strength and durability of the pavement. To avoid such serious
di nu
problems, fatigue cracking must be reduced and controlled in asphalt pavements. As a result,
ye a
asphalt mixtures which have high resistance to fatigue cracking should be selected in pavement
op M
design. The resistance of an asphalt mixture to fatigue cracking is usually examined using
laboratory tests, and the test results are then interpreted using a model to represent the fatigue
C ted
resistance of the material. Models for fatigue cracking used in asphalt pavement engineering
generally can be divided into four categories:

ot p
1) The strain approach (Monismith et al. 1971; The Asphalt Institute 1981; Tayebali 1994);
N ce
2) The dissipated energy approach (Van Dijk et al. 1975; Tayebali et al. 1992; Ghuzlan and
Carpenter 2000);
Ac
3) The fracture mechanics approach; and
4) The continuum damage mechanics approach.
In the strain approach, the fatigue resistance is expressed as the number of load
applications to failure, which is related to the tensile strain by a regression function developed
based on the test data. In the dissipated energy approach, similarly, the fatigue resistance is
represented using a regression function between the dissipated energy and the number of load

J. Mater. Civ. Eng.
Luo et al. 5
applications to failure. The common feature for the first and second categories of models is that
the test data are analyzed using a statistical regression method so the regression coefficients
depend on the experimental conditions, the materials used, and the geometry of the specimen, etc.
This means that a different set of regression coefficients are required if any of these parameters
t
ip
changes. The empirical nature of this statistical analysis method and the resulting models limits
their applications since it is not prudent to describe such phenomena outside the range of the
d cr
original test data. In addition, these models are more dependent on experience rather than on
te s
specifying the cause and effect of the particular phenomenon. In contrast, the mechanics-based
di nu
approach and models do not have such imitations. The need to develop mechanics-based
ye a
pavement design and its benefits are now recognized by more and more people (NCHRP 1-37A,
op M
2002). From this perspective, the third and fourth categories, the fracture mechanics approach
and continuum damage mechanics approach, seem more promising to model the fatigue cracking
C ted
of asphalt mixtures.
In fracture mechanics, the most widely used model for fatigue cracking is a power
ot p
function in the form of the Paris’ Law (Paris and Erdogan 1963), which relates the crack growth
N ce
per cycle to the stress intensity factor or the J integral using two coefficients A and n , as shown
in Equation 1.
Ac
dc dc
AK n or AJ n (1)
dN dN
where dc dN is the crack growth per cycle; K is the stress intensity factor; J is the J integral;
and A and n are fracture coefficients determined from experiments. Schapery (1975, 1978)
studied the crack growth in viscoelastic materials and derived A and n as a function of
fundamental material properties. Based on Schapery’s work, more modifications have been made

J. Mater. Civ. Eng.
Luo et al. 6
on A and n so that the fatigue crack growth of asphalt mixtures can be predicted from a series
of simple tests such as the creep test, the tensile strength test, and surface energy measurement
instead of extensive fatigue tests (Lytton et al. 1993; Little et al. 1997 and 2001; Walubita et al.
2006; Masad et al. 2007). The major problem of these Paris’ Law based models is that they are
t
ip
derived from the analysis of the growth of one crack, which is not suitable for simulating the
situation of numerous cracks growing in the asphalt mixture specimen under repeated loading.
d cr
Another important mechanics approach that studies cracking damage is continuum
te s
damage mechanics. Compared to fracture mechanics, it has the advantage of considering all
di nu
cracks as damage and measures the damage by the effects of all cracks on the macroscopic
ye a
response of the material (nonlinear stress-strain behavior and degradation of material stiffness).
op M
Kachanov (1958) pioneered this area by differentiating the initial area of an undamaged section
of the material and the actual area of this section after a certain part of the section is “lost” as a
C ted
result of crack growth. He defined a damage parameter as follows:
A A
Z (2)
ot p
A
N ce
where Z is the damage parameter; A is the initial area of the undamaged section; and A is the
actual area of the damaged section. The value A A represents the lost area which is the total
Ac
area of all cracks. Then through the force equilibrium VA V A , the damage parameter can be
expressed by the stress ratio as follows:
V
Z 1 (3)
V
where V is nominal stress on the initial area of the undamaged section; and V is effective stress
on the actual area of the damaged section. In this way, the damage parameter can be formulated

J. Mater. Civ. Eng.
Luo et al. 7
in constitutive equations, and further in the damage evolution law to analytically describe the
effect of crack growth. The application of continuum damage mechanics to fatigue yields many
different models (Hashin and Rotem 1978; Lemaitre and Chaboche 1978; Ostergren and Krempl
1979; Hashin and Laird 1980; Altus 1991; Suresh 1991; Lemaitre 1992). In these models, the
t
ip
incremental damage per loading cycle is postulated to be a function of stress, plastic strain and
damage already accumulated, and the damage evolution law is usually manipulated into the form
d cr
of well-established empirical rules so as to provide legitimacy to the model. Most of these
te s
models require adjustable parameters to fit the test data. As a result, these models are still
di nu
phenomenological in nature (Krajcinovic 1996).
ye a
In order to address the disadvantages associated with the aforementioned methods, the
op M
paper aims at developing an energy-based mechanistic approach to model the fatigue crack
growth in asphalt mixtures. A controlled-strain repeated direct tension (RDT) is used in this
C ted
paper to generate fatigue cracking damage in an asphalt mixture specimen. The RDT test has
been used primarily to evaluate tensile properties of asphalt mixtures and assess cracking
ot p
damage for load related fatigue (Bolzan and Huber 1993). The RDT test is conducted in the
N ce
controlled-strain mode rather than the controlled-stress mode. This is because under the
controlled-strain mode plastic deformation resulting from yielding of the bulk material is small
Ac
and the localized plastic deformation around cracks is significant, whereas a large amount of
yielding of the bulk material occurs under the controlled-stress mode. It is the localized plastic
deformation that is associated with the cracking process, so the controlled-strain test is believed
to better characterize the fatigue cracking behavior.
The paper is organized as follows. First, the materials and procedure of the controlled-
strain RDT test are detailed in the next section. In addition, the analysis method and mechanical

J. Mater. Civ. Eng.
Luo et al. 8
background regarding damage characterization using pseudo strain energy are presented after
introducing the testing method. The following section introduces the method to determine the
material responses and material properties required in the energy-based mechanistic approach.
Based on material responses and material properties determined from the test, the evolution of
t
ip
damage density in a damaged asphalt mixture will be derived in order to predict the fatigue crack
growth. The final section summarizes the major contributions of this paper and presents the
d cr
direction of the ongoing research.
te s
di nu
Configurations and Procedures of Controlled-Strain RDT Test
ye a
Test Configuration
op M
The asphalt mixture used in this study is a dense-graded Texas Department of Transportation
C ted
(TxDOT) Type C mixture (TxDOT 2004). The asphalt binder is labeled as AAD in the Strategic
Highway Research Program (SHRP) Materials Reference Library (MRL) (Jones 1993). The
ot p
aggregate is a crushed limestone obtained from a quarry in San Marcos, Texas and satisfies the
N ce
TxDOT specification requirements. Fabrication of an asphalt mixture specimen involves
aggregate batching, asphalt binder-aggregate mixing, compaction, coring and cutting. The
Ac
Superpave gyratory compactor (SGC) is used in compaction to form cylindrical specimens with
a 152 mm diameter and a 178 mm height. Then the specimen from the SGC is cored and cut to a
specimen with a diameter of 102 mm and a height of 150 mm as recommended by the Superpave
Models team (1999). Volumetric analysis was conducted to ensure the air void content of each
specimen within 4.0±0.5%.
The controlled-strain RDT tests are conducted using the Material Test System (MTS)
machine. The environment chamber of the MTS controls the temperature at 20°C. The test

J. Mater. Civ. Eng.
Luo et al. 9
specimen is glued to a pair of end-caps through which the load is applied. The ends of the test
specimen should be smooth and parallel so that the load is concentric with the axis of the
specimen to avoid possible bending moment imposed on the specimen. Three axial linear
variable differential transformers (LVDTs) are mounted on the middle part of the test specimen.
t
ip
They are placed at 120° apart from each other around the specimen surface and the average of
the three LVDTs is used as the axial deformation of the specimen. Two replicate specimens are
d cr
subjected to the same test procedure introduced as follows.
te s
di nu
Test Procedure
ye a
The test procedure of the controlled-strain RDT test is designed to determine the threshold
op M
between the undamaged state and damaged state of an asphalt mixture, and the destructive
C ted
behavior of the asphalt mixture, which characterizes the cracking damage generated in the
material. The threshold between the undamaged state and damage state is used as the reference
based on which the fatigue damage can be quantified and is examined by performing a series of
ot p
N ce
consecutive controlled-strain RDT tests with different strain levels on the same specimen. From
these tests, the material properties of the asphalt mixture specimen under different strain levels
Ac
are measured. As a result, the behavior of the asphalt mixture under different strain levels can be
identified, which in sequence includes the linear viscoelastic, nonlinear viscoelastic, and
damaged behavior, as shown in Figure 1. The stress-strain diagrams and material properties of an
asphalt mixture at these three different material states have the following characteristics:
x in the linear viscoelastic state, the loading path (OA) and the unloading path (AO) are the
same; the material properties do not change with the increase of the loading cycles at a
given strain level, and they do not change with the change of the strain levels;

J. Mater. Civ. Eng.
Luo et al. 10
x in the nonlinear viscoelastic state, the loading path (OBC) and the unloading path (CO)
are different, and the deformation can fully recover after unloading; the material
properties do not change with the increase of the loading cycles at a given strain level,
but they change with the change of the strain levels
t
ip
x in the damaged state, the loading path (OCE) and the unloading path (EF) are different,
and the deformation cannot be fully recovered after unloading; the material properties
d cr
change with the increase of the loading cycles at a given strain level, and they change
te s
with the change of the strain levels.
di nu
The linear and nonlinear viscoelastic behaviors belong to the undamaged region, separated by
ye a
Point B in Figure 1. The dividing point between the nonlinear viscoelastic state and the damaged
op M
state (Point D) is the threshold between the undamaged and damaged state. Such a threshold is
defined as the critical nonlinear viscoelastic state, the strain level of which is the threshold strain
C ted
level between the nondestructive loading and the destructive loading. The material properties
measured at the threshold strain level are defined as the critical nonlinear viscoelastic properties.
ot p
The method of measuring and determining the critical nonlinear viscoelastic properties is
N ce
detailed in a previous companion paper (Luo et al. 2011b).

Ac
In the particular case of this paper, five consecutive controlled-strain RDT tests are
conducted on the same asphalt mixture specimen. All the tests are performed at a frequency of 1
Hz using a haversine shaped wave. There is a 15 min rest period between two consecutive RDT
tests in order to completely recover any possible deformation. The first four RDT tests are
performed at the maximum axial strain of 40, 50, 60, and 70 με, respectively, with 200 cycles of
load repetition. The material properties are measured and analyzed to identify the state of the test
specimen at each strain level as described above. Detailed results are presented in Luo et al.

J. Mater. Civ. Eng.
Luo et al. 11
(2011b), which indicates that the test specimen is in the undamaged region at the strain levels 40,
50, and 60 με, while it is in the damaged region at 70 με. Therefore, the critical nonlinear
viscoelastic point (Point D in Figure 1) corresponds to a strain level between 60 με and 70 με.
For convenience, the strain level of 60 με is regarded as the threshold strain level corresponding
t
ip
to the critical nonlinear viscoelastic state. Once the threshold is identified, a destructive RDT test
with 200 με is conducted on the same test specimen for 1,000 cycles to introduce significant
d cr
damage to the specimen.
te s
di nu
Analysis of Test Data
ye a
After measuring the critical nonlinear viscoelastic properties and destructive properties of an
op M
asphalt mixture using the controlled-strain RDT test, the next step is to analyze the test data
using an energy-based mechanistic approach in order to model fatigue crack growth in the
C ted
material. Three steps are involved in the analysis of test data, which is introduced as follows.
Firstly, the material responses of the asphalt mixture specimen in the controlled-strain
ot p
RDT test are simulated using mathematical models, so the strain energy can be calculated using
N ce
the simulated stress and strain. In a typical controlled-strain RDT test, the axial strain is
Ac
controlled to be a standard haversine shape (the axial strain is always positive). In order to
control the axial strain in such a profile, the applied stress must be a shifted haversine shape,
which consists of a tensile (positive) portion and a compressive (negative) portion. Within a
loading cycle of the controlled-strain RDT test, the asphalt mixture specimen is considered to be
in “tension” when it has a tensile strain and is subjected to a tensile stress; when the specimen
has a tensile strain but is under a compressive stress, it is regarded to be in “quasi-compression”.
In order to distinguish the tension from the quasi-compression, the stress and strain are simulated

J. Mater. Civ. Eng.
Luo et al. 12
using Equations 4 and 5 for the tensile stress portion, and Equations 6 and 7 for the compressive
stress portion. The formulation of Equations 4, 5, 6, and 7 is elaborated in a previous companion
paper (Luo et al. 2011a).

When the stress is tensile:
t
ip
Vt V 0t >1 cosZt @ V st (4)
d cr
H H 0 >1 cosZt Mt @ (5)
te s
When the stress is compressive:
di nu
Vc
ye a V 0c >1 cosZt @ V cm (6)
H H 0 ª¬1 cos Zt M qc º¼
op M
(7)
where Z is the loading frequency in rad/sec; t is the loading time; V 0t is the tensile stress
C ted
amplitude; H0 is the strain amplitude. The ratio of V 0t to H 0 is defined as the magnitude of the
tensile complex modulus, E t* . Mt is phase angle of the tensile complex modulus; V st is the
ot p
N ce
absolute value of the downward shift of a standard haversine wave; V 0c is the compressive stress
amplitude. The ratio of V 0c to H 0 is defined as magnitude of the quasi-compressive complex

Ac
modulus, E qc* . M qc is phase angle of the quasi-compressive complex modulus; V cm is the
magnitude of the minimum stress. The strain energy is calculated by integrating the stress and
strain using Equation 8 for the tensile stress portion and the compressive stress portion separately.
t2 dH t
W ³ V t dt dt
t1
(8)

J. Mater. Civ. Eng.
Luo et al. 13
where W is the strain energy density (energy per unit volume) in a loading period >t1 , t2 @ . There
are two types of strain energy in a complete loading cycle of the RDT test: dissipated strain
energy (DSE) and recoverable strain energy (RSE). The DSE represents the energy dissipated in
a loading cycle due to the viscoelastic effects of the material or the damage, and the RSE
t
ip
represents the stored energy which can be recovered after removing the load. The procedure of
d cr
calculating the DSE and RSE can be found in the companion paper (Luo et al. 2011a).
Secondly, pseudo strain is employed to separate the viscoelastic effect from the strain
te s
di nu
energy that is available to cause damage, so the energy calculated using the pseudo strain,
defined as pseudo strain energy, can be used to characterize the fatigue damage that is generated
ye a
in the asphalt mixture. Equations 9 and 10 give the expressions of the pseudo strain, which are
op M
formulated based on the critical nonlinear viscoelastic properties. The process of deriving
Equations 9 and 10 is presented in detail in the previous paper (Luo et al. 2011b).
C ted
When the stress is tensile:

ot p
H0 V st ,NLVE
H R H 0 ª¬1 cos Zt Mt MtNLVE º¼ (9)
H *
N ce
0, NLVE EtNLVE
When the stress is compressive:

Ac
H0 V cm,NLVE
H R H 0 ª¬1 cos Zt M qc M qcNLVE º¼ (10)
H 0, NLVE
*
EqcNLVE
where H R is the pseudo strain; H 0,NLVE is the strain amplitude in the RDT test at the threshold
strain level; V st , NLVE is the absolute value of the downward shift of a standard haversine wave at
the threshold strain level; V cm,NLVE is the magnitude of the minimum stress at the threshold strain

J. Mater. Civ. Eng.
Luo et al. 14
level; MtNLVE and E tNLVE

*
are the phase angle and magnitude of the tensile complex modulus at
the critical nonlinear viscoelastic state, respectively; M qctNLVE and E qcNLVE

*
are the phase angle
and magnitude of the quasi-compressive complex modulus at the critical nonlinear viscoelastic
t
state, respectively. The pseudo strain energy is calculated using the stress and pseudo strain
ip
following the same method and procedure as those for the strain energy in Equation 8. Similarly,
d cr
there are two types of pseudo strain energy, one of which is the energy dissipated to develop
te s
damage in the specimen, defined as dissipated pseudo strain energy (DPSE), the other of which
di nu
is the recoverable pseudo strain energy (RPSE) that is stored and recovered corresponding to the
purely elastic effect of the material. Expressions of the DPSE and RPSE can be found in the
ye a
companion paper (Luo et al. 2011b).
op M
Thirdly, the energy balance equations, which are the core of the energy-based
C ted
mechanistic approach, are established based on the measurement and determination of different
types of energy from the controlled-strain RDT test. The energy balance equations are
ot p
established between the apparent measurements of the bulk asphalt mixture specimen and the
N ce
true measure in the intact material based on the principle of energy balance. The apparent
measurements refer to the stress and strain (Equations 4 to 7), pseudo strain (Equations 9 and 10),
Ac
the DSE and RSE, and the DPSE and RPSE obtained from the controlled-strain RDT tests as
described above. The object of the apparent measurement is the bulk specimen, which is
composed of the intact material and the air voids and possible cracks that grow from the air voids.
Because of the presence of air voids/cracks, the stress, strain, and energy of the intact material
are certainly different from those of the bulk specimen measured from the test. There is stress
localization (stress concentration and nonlinear distribution), strain magnification, and energy
redistribution (energy release and restoration) in the intact material in the vicinity of cracks. As a

J. Mater. Civ. Eng.
Luo et al. 15
result, the stress, strain, and energy of the intact material are defined as “true stress”, “true strain”,
and “true energy”. Correspondingly, the stress, strain, and energy of the bulk specimen are
defined as “apparent stress”, “apparent strain”, and “apparent energy”. The true measure is then
related to the apparent measurements based on the principle of energy balance, which states that
t
ip
any kind of true energy within the intact material equals its counterpoint from the apparent
measurement. Specifically, the principle of energy balance is presented in the form of four
d cr
energy balance equations:
te s
1) DSE balance equation:
di nu
DSE A
ye a DSE T (11)
2) RSE balance equation:

op M
RSE A RSE T (12)

C ted
3) DPSE balance equation:
DPSE A DPSE T (13)

ot p
N ce
4) RPSE balance equation:
§2 2 ·
RPSE AVm RPSE T Vm RPSE T ¨ M N S 2 c N 3 M I S 2 cI 3 ¸ 2J M N S c N 2 M I S cI 2 (14)
Ac
©3 3 ¹
where the superscript “A” indicates “apparent”, and the superscript “T” indicates “true”; Vm is
the volume of the asphalt mastic in one layer of the asphalt mixture specimen, whose thickness
equals the mean film thickness. The formulation of Vm is:
Vm Am t (15)

J. Mater. Civ. Eng.
Luo et al. 16
where Am is the area of the asphalt mastic on a cross section of the specimen; and t is the mean
film thickness. cI and c N are average crack size of the initial crack before crack growth and
average crack size of the new crack after crack growth; M I and M N are the initial number of
t
cracks and the number of new cracks; and J is the surface energy density (energy per unit area).
ip
The theoretical background regarding the true and apparent measures and these four energy
d cr
balance equations is elaborated in the previous companion paper (Luo et al. 2011b).
te s
In this paper, these four energy balance equations will be used to determine the damage
di nu
density that indicates the extent of cracking damage in the asphalt mixture. All apparent
measurements on the left side of the equations are provided by the controlled-strain RDT test. On
ye a
the right side of the balance equations, the true strain energy, DSE T and RSET, is calculated by
op M
integrating the true stress and true strain; the true pseudo strain energy, DPSE T and RPSET, is
C ted
calculated by integrating the true stress and true pseudo strain. The following section details the
simulation of the true stress, true strain and true pseudo strain in the controlled-strain RDT test.
ot p
N ce
Simulation of True Stress, True Strain, and True Pseudo Strain

Ac
In a typical controlled-strain RDT test, the stress, strain and pseudo strain in the intact material,
or the true stress, true strain and true pseudo strain, should have the same pattern as those in the
bulk specimen. In other words, the true stress has the same formulation as the apparent stress
shown in Equation 4 or 6; the true strain has the same formulation as the apparent strain shown
in Equation 5 or 7; the true pseudo strain has the same formulation as the apparent pseudo strain
shown in Equation 9 or 10. However, being different from the apparent stress/strain/pseudo
strain that have tension and quasi-compression due to crack opening and closure (Luo et al.

J. Mater. Civ. Eng.
Luo et al. 17
2011a), the true stress/true strain/true pseudo strain do not need to distinguish the tension and
quasi-compression since the cracks are excluded from the intact material. As a result, the true
stress/true strain/true pseudo strain can be formulated as Equation 16, 17, and 18, respectively.
VT V 0T ¬ª1 cos Z t ¼º V cm
T
(16)
t
ip
HT H 0T ª¬1 cos Zt M T º¼ (17)
d cr
H 0T V cm
T
te s
H T
R H ª¬1 cos Zt M M
T
0
T T
NLVE º¼ H T
*T
, NLVE
(18)
ENLVE
di nu
0, NLVE
where V T is the true stress; V 0T is the true stress amplitude; V cm

T
is the magnitude of the true
ye a
minimum stress; H T is the true strain; H 0T is the true strain amplitude, which is related to V 0T by
op M
the magnitude of the true complex modulus E *T as follows:

C ted
V 0T
H 0T (19)
E *T
ot p
N ce
M T is the phase angle of the true complex modulus; H RT is the true pseudo strain; M NLVE
T
is the
phase angle of the true complex modulus at the critical nonlinear viscoelastic state; E NLVE
*T
is the
Ac
magnitude of the true complex modulus at the critical nonlinear viscoelastic state; H 0,T NLVE is the
true strain amplitude in the RDT test at the threshold strain level; and V cm
T
, NLVE is the magnitude
of the true minimum stress at the threshold strain level. The simulation parameters, such as V 0T ,
V cm
T
, and H 0T , and the true material properties ( E *T and M T ) in Equations 16 to 19 are not
directly obtainable since it is impossible to extract the intact material from the bulk specimen and
test it separately. Instead, they must be obtained using the apparent measurements provided by

J. Mater. Civ. Eng.
Luo et al. 18
the controlled-strain RDT test data. The method of determining the simulation parameters and
the true material properties are introduced in the following subsections.

Determination of Simulation Parameters in Undamaged Asphalt Mixtures
t
ip
There are five simulation parameters in Equations 16 to 19: V 0T , V cm
T
, H 0T , H 0,T NLVE , and V cm
T
, NLVE .
d cr
As a matter of fact, only two of these five parameters are independent simulation parameters
te s
including V 0T and V cm
T
, because:
di nu
x H 0T is related to V 0T by Equation 19;
ye a
x H 0,T NLVE is the value of H 0T at the threshold strain level; and
op M
x V cm
T
, NLVE is the value of V cm at the threshold strain level.
T
C ted
These two independent simulation parameters, V 0T and V cm

T
, are determined using the principle
ot p
of force equilibrium proposed as follows.

N ce
The principle of force equilibrium for the undamaged asphalt mixtures states that the true
force carried by the intact material equates the apparent force assumed to be carried by the entire
Ac
cross section of the specimen. According to this principle, a force equilibrium equation can be
constructed between the true stress and the apparent stress. Specifically, for a complete loading
cycle of the RDT test, force equilibrium equations are constructed:
x between V t and V T for the tensile stress portion as shown in Equation 20, and
x between V c and V T for the compressive stress portion as shown in Equation 21.

J. Mater. Civ. Eng.
Luo et al. 19
A^V 0t >1 cosZt @ V st ` A Sct ^V 0T >1 cosZt @ V cmT ` (20)
A^V 0c >1 cosZt @ V cm` A Scc ^V 0T >1 cosZt @ V cmT ` (21)

where A is the entire cross-sectional area of the asphalt mixture specimen; Sct is the area of the
t
ip
cracks on a cross section in the tensile stress portion. The value of A Sct represents the area
d cr
of the intact material on a cross section in the tensile stress portion; S cc is the area of the cracks
te s
on a cross section in the compressive stress portion, which is different from Sct because the
di nu
extent of the crack opening and closure is different. The value of A Scc represents the area of
ye a
the intact material on a cross section in the compressive stress portion.
op M
Arranging Equation 20 yields two equations shown in Equations 22 and 23:
AV 0t A Sct V 0T (22)
C ted
AV st A Sct V cmT (23)

ot p
Similarly, Equation 21 yields two equations as follows:

N ce
AV 0c A Scc V 0T (24)
Ac
AV cm A Scc V cmT (25)
The sum of the left sides of Equations 22 and 24 equals the sum of their right sides:
AV 0t V 0c A Sct A Scc V 0T (26)
Combine Equations 23 and 25 in the same way:
AV st V cm A Sct A Scc V cmT (27)

J. Mater. Civ. Eng.
Luo et al. 20
Define the average of Sct and S cc as the average area of the cracks on a cross section in a
loading cycle of the RDT test, denoted as S c , which is calculated using Equation 28:
Sct Scc
Sc (28)
2
t
ip
Substituting Equation 28 into Equations 26 and 27, gives:
d cr
AV 0t V 0c 2A Sc V 0T (29)
te s
AV st V cm 2A Sc V cm
di nu
T
(30)
Solve for V 0T by Equation 29 and for V cm

T
by Equation 30 to obtainn:
ye a
op M
V 0t V 0c 2
V 0T (31)
1 Sc A
C ted
V st V cm 2
V cm
T
(32)
1 Sc A
ot p
Equations 31 and 32 are the expressions of the simulation parameters for the true stress
N ce
calculated using the apparent measurements from the RDT test. They are only applicable to
Ac
undamaged asphalt mixture in nondestructive RDT test. In the nondestructive test, the value of
Sc A equals the air void content since the air voids are the initial cracks and the area of the air
voids does not change in the nondestructive test. Consequently, Equations 31 and 32 are used to
determine the true stress in an undamaged asphalt mixture.
Determination of True Material Properties in both Undamaged and Damaged Asphalt Mixtures

J. Mater. Civ. Eng.
Luo et al. 21
The true material properties are the material properties of the intact material in an asphalt
mixture specimen. In a controlled-strain RDT test, the true complex modulus E *T ( E *T and M T )
has the following characteristics:

t
x Both E *T and M T remains the same with the increase of the number of loading cycles in
ip
an RDT test at any given strain level as long as the strain level stays unchanged. This is
d cr
because the air voids and cracks are excluded from the intact material, so the crack
te s
growth within an RDT test does not influence the behavior of the intact material.
di nu
x Both E *T and M T change ( E *T decreases and M T increases) from one strain level to a
ye a
higher strain level if the asphalt mixture specimen is beyond the linear viscoelastic region
op M
(in the nonlinear viscoelastic region or the damaged region). This is because of the
nonlinear viscoelastic nature of the intact material when an asphalt mixture specimen is
C ted
beyond the linear viscoelastic region.
According to these characteristics, the true material properties are classified into two categories:
ot p
N ce
1) The true material properties of an undamaged asphalt mixture, which are defined as
nondestructive true material properties; and

Ac
2) The true material properties of a damaged asphalt mixture, which are defined as
destructive true material properties.
The nondestructive true material properties at the threshold strain level between the
nondestructive and destructive regions are the true critical nonlinear viscoelastic properties
( E NLVE
*T
and M NLVE
T
). The nondestructive true material properties are obtained from the
nondestructive RDT tests; the destructive true material properties are obtained from the

J. Mater. Civ. Eng.
Luo et al. 22
destructive RDT test. They cannot be measured directly from the test. Instead, both
nondestructive and destructive true material properties are determined using the DSE and RSE
balance equations.
In order to use the DSE balance equation (Equation 11) and the RSE balance equation
t
ip
(Equation 12), DSET and RSET must be determined first. They are calculated by integrating V T
d cr
(Equation 16) and H T (Equation 17) using the same method and procedure as those for the DSE A
and RSEA, as detailed in the companion paper (Luo et al. 2011a). The formulation of the DSE T
te s
di nu
and RSET are:
DSET SV0T H 0T sinM T (33)

ye a
op M
sin M T S 2M T cos M T ª¬cos TtZ 1º¼
RSE T
V H
T
0
T
0
2 (34)
H 0T cos M T cos §¨© Tt2Z ·¸¹
C ted
2 V 0T V cm
T
where Tt is the period corresponding to the tensile stress portion. Substituting Equations 33 and
ot p
N ce
34 into Equations 11 and 12, respectively, gives:
DSE A SV 0T H0T sin M T

Ac
(35)
sin M T S 2M T cos M T ª¬cos TtZ 1º¼

RSE V H
A T T
0 0
2
(36)
2 V 0T V cm
T
H 0T cos M T cos §¨© T2tZ ·¸¹
The ratio of Equations 35 and 36 is formed by dividing the left side and right side of two
equations, respectively, to give:

J. Mater. Civ. Eng.
Luo et al. 23
DSE A SV 0T sin M T
(37)
RSE A sin M T S 2M T cos M T ª¬cos TtZ 1º¼ ½
°V 0T °
° 2 °
® ¾
°2 V T V T cos M T cos § TtZ ·
°¯ 0 cm
°
¨ 2 ¸
© ¹ °¿
t
ip
Equation 37 contains the simulation parameters V 0T and V cm
T
. They are eliminated by
d cr
substituting Equations 31 and 32 into Equation 37:
te s
DSE A S V 0t V 0c sin M T
(38)
di nu
RSE A sin M T S 2M T cos M T ª¬cos TtZ 1º¼ ½
°V 0t V 0c °
° 2 °
® ¾
°2 V V V V cos M T cos § TtZ · °
ye a
°¯ 0t 0c st cm ¨ 2 ¸ °¿
© ¹
op M
In Equation 38, DSEA, RSEA, V 0t , V 0c , V st , V cm , and Tt are all apparent measurements from
C ted
the RDT test, so M T is the only unknown variable in this equation. The Solver function in the
software Excel is used to solve Equation 38 for M T .

ot p
N ce
After determining M , Equation 35 is used to determine E *T . Based on Equation 19,

T
V 0T
Ac
H 0T , which is substituted into Equation 35 to obtain:

E *T
V 0T
DSE A
SV T
0 *T
sin M T (39)
E
Solving Equation 39 for E *T gives:
S V 0T sin M T
2
*T
E (40)
DSE A

J. Mater. Civ. Eng.
Luo et al. 24
In Equation 40, V 0T is calculated using Equation 31 and M T is determined using Equation 38.
The above approach is applied to the threshold strain level and the destructive RDT test to
determine the true critical nonlinear viscoelastic properties and destructive true material
properties, respectively.
t
ip
The true critical nonlinear viscoelastic properties, E NLVE
*T
and M NLVE
T
, are determined
d cr
based on the test data of the controlled-strain RDT test at 60 με. The procedure is as follows:
te s
1) Calculate M NLVE
T
of the test specimen using Equation 38;
di nu
2) Calculate V 0T using Equation 31, in which Sc A equals the air void content; and
ye a
3) Calculate E NLVE
*T
using Equation 40 with M NLVE
T
and V 0T determined in the above two
op M
steps.
C ted
The calculated values of M NLVE

T
and E NLVE
*T
are presented in Table 1, which also lists the apparent
critical nonlinear viscoelastic properties. It shows that the true material properties are
ot p
substantially different from the apparent properties.

N ce
The same procedure is applied to the controlled-strain RDT test at 200 με to determine
Ac
the destructive true material properties. However, the determination of V 0T is different since S c
is unknown in the destructive test. The above procedure of calculating V 0T is modified in the
following way:
x Calculate the simulation parameters, V 0t and V 0c , of the apparent stress as shown in
Equations 4 and 6 for every loading cycle;

J. Mater. Civ. Eng.
Luo et al. 25
x Plot V 0t and V 0c versus the number of loading cycles through the entire destructive test,
respectively. Fit the plotted curve using a best-fit function; then extrapolate the fitted
curve to intersect with the vertical axis to determine the value when the number of
loading cycles is zero. Such values are defined as the simulation parameters for the
t
ip
apparent stress at the beginning of the test, denoted V 0t0 and V 0c
0
. An example of V 0t in
d cr
the destructive test is shown in Figure 2. A power function is fitted to the curve of V 0t
te s
versus the number of loading cycles N using the software MATLAB:
di nu
V 0t 8.988 N 0.244 304.6, R2 0.9945 (41)
ye a
If the fitted curve is extrapolated as shown by the dotted curve, it will intersect with the
op M
vertical axis at V 0t0 , which is 304.6 kPa.
C ted
Substitute V 0t and V 0c into Equation 31 to calculate V0 at the beginning of the test.

0 0 T
x
Since V 0t and V 0c are determined at the beginning of the test, the specimen is not
0 0
ot p
damaged so the value of Sc A still equals the air void content.

N ce
The calculated values of M T and E *T in the destructive test are presented in Table 1, and the
Ac
apparent material properties at the 1st loading cycle of the destructive test are shown as well for
comparison. The true destructive material properties are also different from the true
nondestructive material properties.
Based on the determined simulation parameters of the apparent stress in the
nondestructive test and the determined true nondestructive material properties, the hysteresis
loops can be produced to illustrate the difference between the true stress/strain and the apparent
stress/strain, as shown in Figure 3. The true stress-true strain hysteresis loop is slightly larger

J. Mater. Civ. Eng.
Luo et al. 26
than the apparent stress-apparent strain hysteresis loop. As the loading cycle increases, for
instance from the 1st to the 100th cycle, the shape and area of both hysteresis loops do not change.
Determination of Simulation Parameters in Damaged Asphalt Mixtures
t
ip
After determining the true stress/strain in nondestructive tests and the true material properties in
d cr
nondestructive/destructive tests, this subsection further determines the true stress/true strain in a
te s
damaged asphalt mixture specimen in destructive tests. The DPSE balance equation (Equation 13)
di nu
is employed to solve for the true stress in the destructive RDT test.
First, the DPSET is calculated by integrating V T (Equation 16) and H RT (Equation 18),
ye a
op M
and the result is shown in Equation 42:
DPSET SV 0T H0T sin M T MNLVE

T
(42)
C ted
Substitute for H 0 by Equation 19 to obtain:

T
ot p
V
T 2

N ce
DPSE T
S 0
*T
sin M T M NLVE
T
(43)
E
Ac
in which MNLVE and E NLVE

T *T
are the nondestructive true material properties determined in the
above subsection and shown in Table 1; M T and E *T are the destructive true material properties
also determined above and shown in Table 1. Substituting Equation 43 into Equation 13, gives:
DPSE A
S
V
T 2
0

sin M T M NLVE
T
(44)
*T
E
Solving for V0 from Equation 44 gives:

T

J. Mater. Civ. Eng.
Luo et al. 27
DPSE A E *T
V T
(45)
S sin M T M NLVE
0 T
Second, the principle of the force equilibrium is established for the damaged asphalt
mixture in the destructive tests. This principle states that the true force carried by the intact
t
ip
asphalt mastic equals the apparent force assumed to be carried by the asphalt mastic with the
d cr
increased crack area without counting the initial air void area. According to this principle, a set
te s
of force equilibrium equations are developed that are similar to Equations 20 to 32 with the
di nu
following modifications:
x Replace A with Am , which is the area of the asphalt mastic including the area of the new
ye a
op M
cracks increased from the air voids (but excluding the air void area);
x Replace Sct with S ctm , which is the new crack area excluding the original air voids in the
C ted
tensile stress portion;
x Replace S cc with Sccm , which is the new crack area excluding the original air voids in the
ot p
N ce
compressive stress portion; and
x Replace S c with Scm , which is the average area of the new cracks excluding the original
Ac
air voids in a loading cycle.
With the above modifications, simulation parameters for the true stress can be determined based
on the apparent measurements from the destructive RDT tests using the Equations 46 and 47:
V 0T
V 0t V 0c 2
(46)
1 Scm Am

J. Mater. Civ. Eng.
Luo et al. 28
V cm
T V st V cm 2
(47)
1 Scm Am
Based on Equations 46 and 47, the ratio of V cm

T
to V 0T is calculated to be:
t
V cm
T
V st V cm
ip
(48)
V 0T V 0t V 0 c
d cr
The ratio of V cm
T
to V 0T is then used together with Equation 45 to determine V cm
T
.
te s
Third, based on the determined V 0T and V cm
T
for the true stress and the destructive true
di nu
material properties, the hysteresis loops in the destructive test can be produced as shown in
ye a
Figure 4. The true stress-true strain hysteresis loop is larger than the apparent stress-apparent
op M
strain hysteresis loop; this difference becomes much more distinct as the number of loading cycle
increases to 250. Furthermore, the apparent stress-apparent strain hysteresis loop becomes
C ted
smaller from the 1st to the 250th cycle. In contrast, the true stress-true strain hysteresis loop
becomes larger. To better explain this phenomenon, the true and apparent stress amplitudes are
ot p
plotted versus the number of loading cycles for the entire destructive test, which is shown in
N ce
Figure 5.
Ac
Two observations are made from Figure 5:
1) V 0T is larger than V 0t and V 0c at the same loading cycle; and
2) V 0T increases whereas V 0t and V 0c decrease with the increase of loading cycles.
The first observation is explained by the fact that the area of the intact material sustaining the
true stress is smaller than the entire cross sectional area that is used to calculate the apparent
stress. To satisfy the force equilibrium, the true stress must be larger than the apparent stress. In

J. Mater. Civ. Eng.
Luo et al. 29
the second observation, the decrease of V 0t and V 0c represents the reduction of the modulus in
the controlled-strain RDT test. This conforms to the common explanation for the effect of
cracking damage, which leads to the degradation of the apparent modulus of the asphalt mixture.
In fact, the material properties of the intact material do not change when cracks propagate. The
t
ip
degradation of the apparent modulus is actually a result of the loss of the cross-sectional area
d cr
with the crack growth, which also causes the variation of the true stress and true strain of the
intact material with the increase of loading cycles.
te s
di nu
The influence of the cracking damage on the true stress/true strain is explained by the fact
of the change of geometry of the cracks, which is schematically shown in Figure 5. The initial
ye a
cracks, or the air voids in the asphalt mixture specimen are approximately elliptical holes with
op M
much shorter minor axis than the major axis. As the air voids grow from the 1st cycle to the 250th
cycle, the major axis increases greatly to form the “new crack”, and the air voids begin to take on
C ted
the appearance of a real crack: two surfaces meeting along two sharp edges at the crack tips.
During the process of crack growth, the stress concentration in the intact material is aggravated
ot p
significantly at the crack tip due to the change of geometry of the crack according to Inglis’
N ce
stress concentration analysis (Inglis 1913). As a result, the true stress increases as the number of
Ac
loading cycles increases. In the meantime, the true strain of the intact material also increases
corresponding to the increasing true stress. Therefore, the actual effect of crack growth is the
aggravation of the localization of the true stress and true strain.
On the other hand, the aggravation of true stress localization promotes the crack growth
in asphalt mixtures. Paris and Sih (1965) pointed out that the stress concentration near the tip of a
crack most likely initiated the growth of the crack. Other researchers also proved that the high
local stress and strain concentration would dominate the fatigue failure process (Li and Metcalf

J. Mater. Civ. Eng.
Luo et al. 30
2004; Wang et al. 2007). Therefore, the complete description of the relationship between the
crack growth and the true stress/true strain is: the true stress/true strain localization is the cause
of the crack growth; then the growth of cracks further enhances the severity of the true stress/true
strain localization. Compared to a phenomenological description using the apparent
t
ip
stress/apparent strain, the true stress/true strain reveals the essence of crack growth and should be
used to model the cracking damage, which is described in the next section.
d cr
te s
di nu
Determination of Damage Density
Damage density is an index that describes the extent of cracking damage in an asphalt mixture
ye a
specimen. Since the cracks grow from the air voids in the asphalt mixture specimen, the damage
op M
density can be expressed in the following way:
C ted
I I0 I' (49)
where I is the damage density; I0 is the initial damage density, which equals the air void
ot p
content; and I ' is the increase of the damage density due to the crack growth, which is
N ce
calculated in Equation 48:

Ac
D% Am Vm
I' D% D% Vm % (50)
A V
in which D% is the percentage of cracking damage in the asphalt mastic; Am is the area of the
asphalt mastic on the cross section of the specimen. The value of D% Am represents the new
crack area excluding the initial air void area in the cross section of the specimen. Vm and V are
the volume of the asphalt mastic and asphalt specimen, respectively; and Vm % is the volumetric

J. Mater. Civ. Eng.
Luo et al. 31
percentage of the asphalt mastic that is determined from the volumetric analysis and aggregate
gradation of the mixture design.
The cracking damage in the asphalt mastic cannot be measured directly; instead, it is
inferred from the force equilibrium equation by solving for Scm Am from Equation 46:
t
ip
D%
Scm
1
V 0t V 0c 2
(51)
V
d cr
T
Am 0
te s
Substitute Equations 50 and 51 into 49 to obtain the damage density as follows:
di nu
ª
I I0 «1
V 0t V 0c 2º
»Vm % (52)
¬ V T
0 ¼
ye a
op M
Figure 6 shows the damage density of a number of representative loading cycles
calculated by Equation 52 in the destructive test. At the end of the 1,000 cycles, I' is about
C ted
4.4%; the total damage density is 8.2% by adding I' to I0 (3.8%). There is a rapid increase of
the damage density at the beginning of the test; then the rate of the change of the damage density
ot p
decreases as the loading cycle increases. The evolution of the damage density with the increase
N ce
of the number of loading cycles N is modeled by a power function using MATLAB:

Ac
I 0.536N 0.310 3.790, R2 0.989 (53)
Conclusions and Future Work
The paper models the damage density that is the index of the fatigue cracking damage of an
asphalt mixture specimen in the controlled-strain RDT test. An energy-based mechanistic
approach is developed to determine the evolution of the damage density with the increase of the

J. Mater. Civ. Eng.
Luo et al. 32
loading cycles in the destructive RDT test. The core of this approach is establishing the energy
balance equations between the apparent energy of the bulk asphalt mixture specimen and the true
energy of the intact material.

The true stress/true strain/true pseudo strain are first modeled to calculate the true strain
t
ip
energy/true pseudo strain energy. The true stress in the nondestructive test is solved from the
force equilibrium equation; the true stress in the destructive test is solved from the DPSE balance
d cr
equation. By studying the true stress in a damaged asphalt mixture specimen, it is found that the
te s
true stress is the driving force of the crack growth, and the crack growth in turn aggravates the
di nu
stress and strain localization in the intact material. After obtaining the true stress, the true
ye a
material properties are determined to obtain the true strain/true pseudo strain. The true material
op M
properties include nondestructive and destructive true material properties, which are obtained
from the nondestructive and destructive tests, respectively. They are calculated using the DSE
C ted
and RSE balance equations.
The ratio of the true stress to the apparent stress in a damaged asphalt mixture is used to
ot p
calculate the damage density for every loading cycle of the destructive RDT test. The evolution
N ce
of the damage density demonstrates the development of fatigue cracking in the asphalt mixture
specimen. The curve of the damage density versus the number of loading cycles can be simulated
Ac
by a power function.
The testing protocol and all the analysis introduced in this study can be accomplished
within one day. The testing only takes two hours for an asphalt mixture specimen. All the
analysis is automatic, which is embedded in the Excel spreadsheet with Visual Basic programs
and the link to the software Matlab. The analysis results provide comprehensive information
regarding the tensile behavior of asphalt mixtures, including nondestructive properties, the

J. Mater. Civ. Eng.
Luo et al. 33
critical nonlinear viscoelastic properties, the threshold between the undamaged state and
damaged state, destructive properties, and damage density, etc.
In addition to modeling fatigue cracking in asphalt mixtures, the energy-based

mechanistic approach developed in this paper can be used with a wide range of tests to predict
t
ip
crack growth of different types of materials because of its mechanical nature. Furthermore, it is
capable of modeling the healing of asphalt mixtures, a subject which will be introduced in the
d cr
following companion papers. The major advantage of this approach in addition to its shorter time
te s
and greater accuracy and repeatability is its mechanistic nature that reveals the causal
di nu
relationship between the phenomena of cracking and healing, which is distinct from the common
ye a
empirical or phenomenological methods.
op M
In a continuation of this study, the damage density determined in this paper will be used
to further calculate the fracture coefficients, average crack size and number of cracks in the
C ted
asphalt mixture under destructive loading. These fracture coefficients can be determined
uniquely using the coefficients in Equation 53. The fracture coefficients obtained from the
ot p
damage density address the limitation of the traditional Paris’ Law that only models the growth
N ce
of a single crack. Being able to predict the growth of a multitude of cracks using Paris’ Law is
another practical application of this study.

Ac

J. Mater. Civ. Eng.
Luo et al. 34
References
Altus, E. (1991). “Fatigue, Fractals, and a Modified Miner’s Rule.” Journal of Applied
Mechanics, Vol. 58, 99. 37-42.
Bolzan, P. E. and Huber, G. (1993). “Direct Tension Test Experiments.” SHRP-A-641,

Strategic Highway Research Program, National Research Council. Washington, D.C
t
ip
Ghuzlan, K. A., and Carpenter, S. H. (2000). “Energy-Derived, Damage-Based Failure
Criterion for Fatigue Testing.” Transportation Research Record 1723, Transportation
d cr
Research Board, National Research Council, Washington, D.C. pp. 141–149.
te s
Hashin, Z. and Rotem, A. (1978). “A Cumulative Damage Theory of Fatigue Failure.”
di nu
Material Science of Engineering, Vol. 34, pp. 147-160.
Hashin, Z. and Laird, C. (1980). “Cumulative Damage under Two Level Cyclic Loading:
ye a
Some Theoretical Predictions and Test Data.” Fatigue of Engineering Materials and
op M
Structures, Vol. 2, pp. 345-350.
Inglis, C. E. (1913). “Stresses in a Plate Due to the Presence of Cracks and Sharp Corners.”
C ted
Transactions of the Institute of Naval Architects, Vol. 55, pp. 219-241.
Jones, D. R. (1993). “SHRP Materials Reference Library: Asphalt Cements: A Concise Data
Compilation.” Strategic Highway Research Program Rep. No.SHRP-A-645, National
ot p
Research Council, Washington, D.C.

N ce
Kachanov, L. M. (1958). Introduction to Continuum Damage Mechanics. Kluwer Academic

Poblishers, Dordrecht.
Ac
Krajcinovic, D. (1996). Damage Mechanics, Elsevier, Amsterdam, The Netherlands.
Lemaitre, J. and Chaboche, J. L. (1978). “Aspect Phenomenologique de la Rupture par

Endommagement.” J. Mech. Applique, Vol. 2, pp. 317-365.
Lemaitre, J. (1992). A Course on Damage Mechanics, Springer-Verlag, Berlin, Germany.
Li, Y. and Metcalf, J. B. (2004). “Fatigue Characteristics of Asphalt Concrete from Asphalt
Slab Tests.” Journal of Materials in Civil Engineering, Vol. 16, No. 4, pp. 306-314.

J. Mater. Civ. Eng.
Luo et al. 35
Little, D. N., Lytton, R. L., Williams, D., and Kim, Y. R. (1997). “Propagation and Healing
of Microcracks in Asphalt Concrete the Their Contributions to Fatigue.” Asphalt Science
and Technology, pp. 149-195.
Little, D. N., Lytton, R. L., Williams, D., and Chen, C.W. (2001). “Fundamental Properties
of Asphalts and Modified Asphalts – Volume 1: Microdamage and Microdamage
t
ip
Healing.” Federal Highway Administration Final Report, No. FHWA-RD-98-141,
Washington D.C.
d cr
Luo, X., Luo, R., and Lytton, R. L. (2011a). “Characterization of Asphalt Mixtures Using
te s
Controlled-Strain Repeated Direct Tension Test.” Journal of Materials in Civil
di nu
Engineering, ASCE, accepted for publication.
Luo, X., Luo, R., and Lytton, R. L. (2011b). “Characterization of Fatigue Damage in Asphalt
ye a
Mixtures Using Pseudo Strain Energy.” Journal of Materials in Civil Engineering, ASCE,
accepted for publication.
op M
Lytton, R. L., Roque, R. L., Uzan, J., Hiltunen, D. R., Fernando, E., and Stoffels, S. M.,
(19993). Performance Models and Validation of Test Results. Final Report to Strategic
C ted
Highway Research Program; Asphalt Project A-005, SHRP Report A-357.
Masad, E., Castelo Branco, V., Little, D., and Lytton, R. L. (2007). “A Unified Method for
ot p
the Analysis of Controlled-Strain and Controlled-Stress Fatigue Testing,” International

N ce
Journal of Pavement Engineering, Vol. 9, No. 4, pp.1-14.
Monismith, C. L., Epps, J. A., Kasianchuk, D. A., and McLean, D. B. (1971). Asphalt
Ac
Mixture Behavior in Repeated Flexure. Report TE 70-5, University of California,

Berkeley.
NCHRP 1-37A (2002). 2002 Design Guide, Design of New and Rehabilitated Pavement
Structure, Draft Final Report prepared for National Cooperative Highway Research
Program, Transportation Research Board.
Ostergren, W. J. and Krempl, E. (1979). “A Uniaxial Damage Accumulation Law for Time-
Varying Loading Including Creep-Fatigue Interaction.” Journal of Pressure Vessel
Technology, Vol. 101, pp. 118-124.

J. Mater. Civ. Eng.
Luo et al. 36
Paris, P. C. and Erdogan, F. (1963). “A Critical Analysis of Crack Propagation Laws.”

Journal of Basic Engineering, Vol. 85, pp. 528-534.
Paris, P. C. and Sih, G. C. (1965). “Stress Analysis of Cracks.” Fracture Toughness Testing
and Its Applications: A Symposium Presented at the Sixty-Seventh Annual Meeting,
American Society for Testing and Materials, Chicago, III., June 21-26.
t
ip
Schapery, R. A. (1975). “A Theory of Crack Initiation and Growth in Visco-Elastic Media; I:
Theoretical Development, II: Approximate Methods of Analysis, III: Analysis of
d cr
Continuous Growth.” International Journal of Fracture, Vol. 11, No. 1, pp. 141-159, Vol.
te s
11 No. 3, pp. 369-388, and Vol.11, No. 4, pp. 549-562.
di nu
Schapery, R. A. (1978). “A Method for Predicting Crack Growth in Non-homogeneous
Visco-Elastic Media.” International Journal of Fracture, Vol. 14, No. 3, pp. 293-309.
ye a
Superpave Models Team. (1999). Task F – Advanced Mixture Characterization, Team
op M
Report SLS -3. SUPERPAVE Support and Performance models Management, NCHRP 9-
19, Department of Civil and Environmental Engineering, Arizona State University,
Tempe, AZ
C ted
Suresh, S. (1991). Fatigue of Materials, Cambridge University Press, Cambridge, UK.
Tayebali, A. A., Rowe, G. M., and Sousa, J. B. (1992). “Fatigue Response of Asphalt–
ot p
Aggregate Mixtures.” Journal of the Association of Asphalt Paving Technologists, Vol.

N ce
61, pp. 333–360.
Tayebali, A. A. (1994). Fatigue Response of Asphalt-Aggregate Mixtures. Report A404.

Ac
SHRP A003A.
Texas Department of Transportation (2004). Standard Specifications for Construction and

Maintenance of Highways, Streets, and Bridges. Available at
ftp://ftp.dot.state.tx.us/pub/txdot-info/des/specs/specbook.pdf.
Texas Department of Transportation (2005). Design of Bituminous Mixtures. Available at

http://www.txdot.gov/business/contractors_consultants/test_procedures/tms_forms_and_a
rchives.htm?tms=Tex-204-F&qtype=archive.

J. Mater. Civ. Eng.
Luo et al. 37
The Asphalt Institute. (1981). Thickness Design Manual Series No. 1 (MS-1). Ninth Edition,
College Park, Maryland.
Van Dijk, W. (1975). “Practical Fatigue Characterization of Bituminous Mixes”. Journal of

the Association of Asphalt Paving Technologists, Vol. 44, pp.38–74.
Walubita, L. F., Epps Martin, A., Glover, C. J., A, and Lytton, R. L. (2006). “Computation of
t
ip
Pseudo Strain Energy and Paris Law Fracture Coefficients from Surface Energy and
Uniaxial Strain-Controlled Tension Test Data.” International Journal of Pavement
d cr
Engineering, Vol. 7, No. 3, pp. 167-178.
te s
Wang, L, Wang, Y., Li, Q., and Flintsch, G. (2007). “Stress Concentration Factor as a
di nu
Performance Indicator for Asphalt Mixes.” Proceedings of the Symposium on the
Mechanics of Flexible Pavements, ASCE, June 25-30, Boulder, Colorado.
ye a
op M
C ted
ot p
N ce
Ac

J. Mater. Civ. Eng.
)LJXUH&DSWLRQ/LVW
List of Figures
Figure 1. Different Material States of an Asphalt Mixture
Figure 2. Apparent Tensile Stress Amplitude at Various Numbers of Loading Cycles in

Destructive RDT Test
t
ip
Figure 3. Apparent Hysteresis Loop and True Hysteresis Loop in Nondestructive RDT Test
d cr
Figure 4. Apparent Hysteresis Loop and True Hysteresis Loop in Destructive RDT Test
Figure 5. Measured and Calculated Stress Amplitudes at Various Numbers of Loading Cycles in
te s
di nu
Figure 6. Calculated Damage Density at Various Numbers of Loading Cycles in Destructive

ye a
RDT Test
op M
C ted
ot p
N ce
Ac

J. Mater. Civ. Eng.
)LJXUH
σ
Undamaged region Damaged region
Nonlinear viscoelastic Critical nonlinear

region viscoelastic point
E
Linear D
viscoelastic C
region
O
F ε
Figure 1. Different Material States of an Asphalt Mixture
Accepted Manuscript
Not Copyedited

J. Mater. Civ. Eng.
)LJXUH
305
Apparent Tensile Stress Amplitude (kPa)
300
295 σ 0t0
Tensile stress amplitude
Fitted curve
290
285
280
275
270
265
260
255
0 100 200 300 400 500 600 700 800 900 1000
Number of Loading Cycle
Figure 2. Apparent Tensile Stress Amplitude at Various Numbers of Loading Cycles in

Accepted Manuscript
Not Copyedited

J. Mater. Civ. Eng.
)LJXUH
200 200
150 150
100 100
Axial Stress (kPa)

Axial Stress (kPa)
50 50
0 0
0.E+00 2.E-05 4.E-05 6.E-05 0.E+00 2.E-05 4.E-05 6.E-05
-50 -50
-100 -100
-150 -150
Axial Strain
Axial Strain
Apparent tress vs. apparent strain at 1st cycle Apparent stress vs. apparent strain at 100th cycle
True stress vs. true strain at 1st cycle True stress vs. true strain at 100th cycle
(a) 1st Loading Cycle (b) 100th Loading Cycle

Figure 3. Apparent Hysteresis Loop and True Hysteresis Loop in Nondestructive RDT Test
Accepted Manuscript
Not Copyedited

J. Mater. Civ. Eng.
)LJXUH
400 400
300 300
200 200
Axial Stress (kPa)
Axial Stress (kPa)

100 100
0 0
0.E+00 5.E-05 1.E-04 2.E-04 2.E-04 0.E+00 1.E-04 2.E-04
-100 -100
-200 -200
-300 -300
-400 -400
Axial Strain Axial Strain
Apparent stress vs. apparent strain at 1st cycle Apparent stress vs. apparent strain at 250th cycle
True stress vs. true strain at 1st cycle True stress vs. true strain at 250th cycle
(a) 1st Loading Cycle (b) 250th Loading Cycle

Figure 4. Apparent Hysteresis Loop and True Hysteresis Loop in Destructive RDT Test
Accepted Manuscript
Not Copyedited

J. Mater. Civ. Eng.
)LJXUH
500
Stress Amplitude (kPa) 450
400
350
300
250
200
0 200 400 600 800 1000
Number of Loading Cycles
True stress amplitude

Apparent tensile stress amplitude
Apparent compressive stress amplitude
Figure 5. Measured and Calculated Stress Amplitudes at Various Numbers of Loading

Cycles in Destructive RDT Test
Accepted Manuscript
Not Copyedited

J. Mater. Civ. Eng.
)LJXUH
9%
8%
7%
Damage Density (%)
6%
5%
4%
3%
2%
1%
0%
0 200 400 600 800 1000
Number of Loading Cycles
Figure 6. Calculated Damage Density at Various Numbers of Loading Cycles in Destructive

RDT Test
Accepted Manuscript
Not Copyedited

J. Mater. Civ. Eng.
7DEOHGRF[
Table 1. Material Properties in Nondestructive and Destructive RDT Tests
Nondestructive Test at Threshold Loading Destructive Test

Apparent
True Apparent True
(1st loading cycle)
MtNLVE (q) 27.59 M t ( q) 37.24
M NLVE
T
(q) 28.06 M T (q) 38.47
*
EtNLVE MPa 4912 Et* (MPa) 3161
M qcNLVE (q) 30.08 M qc (q) 42.47

*T
ENLVE (MPa) 5315 *
E *T (MPa) 3634 *
EqcNLVE (MPa) 4076 Eqc (MPa) 3025
Accepted Manuscript
Not Copyedited

J. Mater. Civ. Eng.

Luo, Xue Luo, Rong Lytton, Robert L. - Energy-Based Mechanistic Approach To Charac

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

Luo, Xue Luo, Rong Lytton, Robert L. - Energy-Based Mechanistic Approach To Charac

Uploaded by

Copyright:

Available Formats

Journal of Materials in Civil Engineering.

Submitted March 7, 2012; accepted August 21, 2012;

An Energy-Based Mechanistic Approach to Characterize Crack

Xue Luo, Ph.D., A.M.ASCE

Fax: (979) 845-0278

Robert L. Lytton, Ph.D., P.E., F.ASCE

Zachry Department of Civil Engineering

College Station, Texas 77843

Copyright 2012 by the American Society of Civil Engineers

respectively, based on the energy balance equations.

the asphalt mixture.

Copyright 2012 by the American Society of Civil Engineers

In addition to modeling fatigue cracking in asphalt mixtures, the energy-based

crack growth of different types of materials because of its mechanical nature.

Copyright 2012 by the American Society of Civil Engineers

generally can be divided into four categories:

3) The fracture mechanics approach; and

4) The continuum damage mechanics approach.

Copyright 2012 by the American Society of Civil Engineers

Copyright 2012 by the American Society of Civil Engineers

result of crack growth. He defined a damage parameter as follows:

expressed by the stress ratio as follows:

Copyright 2012 by the American Society of Civil Engineers

to better characterize the fatigue cracking behavior.

Copyright 2012 by the American Society of Civil Engineers

TxDOT specification requirements. Fabrication of an asphalt mixture specimen involves

specimen within 4.0±0.5%.

Copyright 2012 by the American Society of Civil Engineers

Copyright 2012 by the American Society of Civil Engineers

but they change with the change of the strain levels

detailed in a previous companion paper (Luo et al. 2011b).

Copyright 2012 by the American Society of Civil Engineers

has a tensile strain but is under a compressive stress, it is regarded to be in “quasi-compression”.

Copyright 2012 by the American Society of Civil Engineers

stress portion. The formulation of Equations 4, 5, 6, and 7 is elaborated in a previous companion

paper (Luo et al. 2011a).

When the stress is tensile:

amplitude. The ratio of V 0c to H 0 is defined as magnitude of the quasi-compressive complex

modulus, E qc* . M qc is phase angle of the quasi-compressive complex modulus; V cm is the

Copyright 2012 by the American Society of Civil Engineers

When the stress is tensile:

When the stress is compressive:

Copyright 2012 by the American Society of Civil Engineers

level; MtNLVE and E tNLVE

the critical nonlinear viscoelastic state, respectively; M qctNLVE and E qcNLVE

Copyright 2012 by the American Society of Civil Engineers

2) RSE balance equation:

RSE A RSE T (12)

3) DPSE balance equation:

DPSE A DPSE T (13)

4) RPSE balance equation:

equals the mean film thickness. The formulation of Vm is:

Copyright 2012 by the American Society of Civil Engineers

Simulation of True Stress, True Strain, and True Pseudo Strain

Copyright 2012 by the American Society of Civil Engineers

where V T is the true stress; V 0T is the true stress amplitude; V cm

the magnitude of the true complex modulus E *T as follows:

Copyright 2012 by the American Society of Civil Engineers

the true material properties are introduced in the following subsections.

Determination of Simulation Parameters in Undamaged Asphalt Mixtures

These two independent simulation parameters, V 0T and V cm

of force equilibrium proposed as follows.

cycle of the RDT test, force equilibrium equations are constructed:

A^V 0t >1 cosZt @ V st ` A Sct ^V 0T >1 cosZt @ V cmT ` (20)

A^V 0c >1 cosZt @ V cm` A Scc ^V 0T >1 cosZt @ V cmT ` (21)

AV st A Sct V cmT (23)

AV cm A Scc V cmT (25)

AV 0t V 0c A Sct A Scc V 0T (26)

AV st V cm A Sct A Scc V cmT (27)

sin M T S 2M T cos M T ª¬cos TtZ 1º¼

DPSET SV 0T H0T sin M T MNLVE

I 0.536N 0.310 3.790, R2 0.989 (53)