Available online at www.sciencedirect.

com
Construction
and Building

Construction and Building Materials 22 (2008) 2170–2179

MATERIALS
www.elsevier.com/locate/conbuildmat

Nonlinear viscoelastic analysis of unaged and aged asphalt binders

a,* a,1 b,2 c,3
Eyad Masad , Chien-Wei Huang , Gordon Airey , Anastasia Muliana
a
Department of Civil Engineering, Texas A&M University, College Station, TX 77843-3136, USA
b
School of Civil Engineering, University of Nottingham, University Park, Nottingham NG7 2RD, UK
c
Department of Mechanical Engineering, Texas A&M University, College Station, TX 77843-3123, USA

Received 25 November 2006; received in revised form 25 August 2007; accepted 26 August 2007
Available online 10 October 2007

Abstract

This paper presents analyses of the nonlinear viscoelastic behavior of unaged and aged asphalt binders tested using a dynamic shear
rheometer (DSR) at several temperatures and frequencies. It was not possible to conduct all DSR tests at the same range of stresses,
which is necessary for establishing the master curve for nonlinear viscoelastic materials. Therefore, the stress levels for each test, at a
given temperature and frequency, were normalized by the ultimate stress level of that test. Consequently, all test results were transformed
to a common range of normalized stresses that were used in establishing the master curve.
A phenomenological model was used to obtain the creep response of the binders in the time domain from the normalized frequency
domain measurements. Then, the Schapery single integral equation was used to model the binder nonlinear creep response. A master
curve at a reference temperature of 30 °C was formed using the time–temperature superposition principle (TTSP) at selected normalized
stress levels. The Schapery’s nonlinear stress dependent parameters (g1g2) were determined by vertical shifting the master curves at the
different normalized stress levels. An aging shift factor was used to obtain the aged binder response from the properties of the unaged
binder. The aging-time shift factor was found to be a function of temperature, but independent of stress level. The nonlinear viscoelastic
model was implemented in the ABAQUS finite element (FE) software and used to back calculate the creep response of the unaged and
aged binders. The FE results were in very good agreements with the experimental measurements.
Ó 2007 Elsevier Ltd. All rights reserved.

Keywords: Asphalt binder; Nonlinear viscoelasticity; The Schapery model; Finite element

1. Introduction binders at different temperatures. The results showed that

Asphalt binders exhibit both linear and nonlinear vis-
coelastic behavior. The nonlinear viscoelastic properties
depend on the stress or strain levels [1]. Cheung and
Cebon [2,3] indicate that asphalt binders behave linearly
at low stress levels and nonlinearly at higher stress levels.
Airey et al. [4,5] conducted stress sweep tests using a
DSR to obtain the linearity limits of various asphalt
*
Corresponding author. Tel.: +1 979 845 8308; fax: +1 979 845 0278.
E-mail addresses: emasad@civil.tamu.edu (E. Masad), stanhuang917@
neo.tamu.edu (C.-W. Huang), gordon.airey@nottingham.ac.uk (G. Airey),
amuliana@neo.tamu.edu (A. Muliana).
1
Tel.: +1 979 458 0893; fax: +1 979 845 0278.
2
Tel: +44 (0) 115 951 3913; fax: +44 (0) 115 951 3909.
3
Tel.: +1 979 458 3579.
the strain dependent linear viscoelastic (LVE) limit is
between 2% and 6% at low temperatures and the stress
dependent LVE limit is between 1.5 and 7 kPa at high
temperatures.
Kose et al. [6] and Masad and Somadevan [7] conducted
finite element analysis of asphalt mix microstructure, which
was modeled as a composite of two phases (asphalt binder
and aggregates). The analysis aimed at calculating the
strain distribution within the aggregate and binder phases
at different macroscopic strain levels applied to the mix-
ture. It was found that the orders of magnitude difference
in stiffness between the aggregate and binder phases caused
high strain levels to localize within the asphalt binder. The
average binder strain was to be about eight times larger
than the macroscopic bulk strain of the mixture, and some

0950-0618/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.conbuildmat.2007.08.012
 E. Masad et al. / Construction and Building Materials 22 (2008) 2170–2179 2171

Notation

g0, g1, g2 stress or strain viscoelastic nonlinear parame- Deij incremental shear strain
ters ekk volumetric strain
D0 instantaneous elastic compliance Dekk incremental bulk strain
wt reduced time J0 instantaneous elastic shear compliance
DD transient compliance DJ transient shear compliance
aT temperature shift factor B0 instantaneous elastic bulk compliance
as strain or stress shift factor DB transient bulk compliance
ag aging shift factor Jn the nth shear compliance of Prony series
Dn the nth coefficient of Prony series Bn the nth bulk compliance of Prony series
kn the nth retardation time of Prony series qij,n shear hereditary integrals
J shear compliance qkk,n volumetric hereditary integrals
Sij deviatoric stress Rij residual strain
B bulk compliance J 0Model the storage compliance of model
rkk volumetric stress J 00Model the loss compliance of model
dij Kronecker delta J 0Exp the storage compliance of experimental data and
eij deviatoric strain J 00Exp the loss compliance of experimental data

parts of the binder phase experienced strain levels within
the nonlinear viscoelastic range.
The Schapery’s single integral model has been widely
used to characterize the nonlinear viscoelastic behavior of
engineering materials [8–10]. Lou and Schapery [11]
extended the Schapery’s integral model to characterize the
nonlinear time dependent behavior of glass fiber reinforced
epoxy, while Sheilds et al. [12] used Schapery’s model to
analyze the nonlinear behavior of asphalt mixtures.
Several numerical algorithms that are compatible with
finite element analysis have been developed for solving
the integral form in the Schapery’s viscoelastic model
[13–15]. Taouti and Cederbaum [13,14] presented a numer-
ical scheme and used it to predict the nonlinear stress relax-
ation of the orthotropic laminated plate. Sadd et al. [16]
implemented Schapery’s theory in a recursive finite element
scheme to represent the micromechanical behavior of
asphalt mixtures. Haj-Ali and Muliana [15] developed a
recursive-iterative integration algorithm to analyze the
nonlinear viscoelastic behavior of polymeric materials.
2. Objectives and scope of the study
Previous work has shown that asphalt binders exhibit
nonlinear viscoelastic behavior especially under high strain
levels that binders may experience in the mix. However,
most current binder tests, specifications and mathematical
models are developed assuming linear viscoelastic behav-
ior. The objective of this paper is to develop a framework
for nonlinear viscoelastic analysis of asphalt binders that
take into account the possible interactions between stress
level, temperature, time of loading (or frequency) and
aging. This framework will be useful for researchers and
practitioners in describing and comparing the behavior of
asphalt binders under various temperatures, aging and
loading conditions. It can also be useful to mathematically
quantify the influence of binder modification on the mod-
el’s parameters and binder performance. This manuscript
is organized as follows:
1. Brief presentation of the Schapery’s nonlinear viscoelas-
tic model.
2. Description of experimental measurements using DSR
on unaged and aged asphalt binders at different stress
levels, temperatures, and frequencies.
3. Determination of the parameters of the nonlinear visco-
elastic model from the experimental measurements.
These parameters include the coefficients of the Prony
series that describes the LVE behavior of the binder,
nonlinear stress-dependent parameters, and stress, tem-
perature and aging shift factors.
4. Use of a finite element model to simulate the nonlinear
viscoelastic behavior of the binders at different tempera-
tures and stress levels.
3. The Schapery’s nonlinear viscoelastic model
The strain response of the Schapery’s integral form
(1969) due to an applied stress rs can be expressed as
Z t
dðg2 rs Þ
eðtÞ ¼ g0 D0 rt þ g1 DDðwt  ws Þ ds ð1Þ
0 ds
in which, D0 is the instantaneous elastic compliance, DD is
the transient compliance. g0, g1 and g2 are the stress or
strain nonlinear parameters related to the nonlinear instan-
taneous compliance, the nonlinear transient compliance,
and the loading rate effect on nonlinear response, respec-
tively. wt is the reduced time given by
Z t
dn
wt ¼ ð2Þ
0 aT a s ag
2172 E. Masad et al. / Construction and Building Materials 22 (2008) 2170–2179

aT is the temperature shift factor, as is the strain or stress where eij is the deviatoric strain and ekk is the volumetric
shift factor, and ag is the aging shift factor. The Prony ser- strain. J0 and B0 are instantaneous elastic shear compliance
ies is used to represent the transient compliance DD as fol- and instantaneous elastic bulk compliance, respectively. DJ
lows: and DB are transient shear compliance and transient bulk
X
N compliance, respectively.
T
DDw ¼ Dn ð1  expðkn wt ÞÞ ð3Þ Assuming the Poisson’s ratio t to be time independent
n¼1 and using the recursive method, the deviatoric and volu-
metric strain components can be written in terms of hered-
where Dn is the nth coefficient of Prony series and kn is the itary integral formulation as follows [15]:
nth retardation time. " #
1 t XN X N
1expðkn Dwt Þ t
The three dimensional isotropic constitutive relations t t t
eij ¼ g0 J 0 þg1 g2 t t
J n g1 g2 Jn S ij
can be decoupled into deviatoric and volumetric parts 2 n¼1 n¼1
kn Dwt
and it can be presented as  
1 X N t
tDt ð1expðkn Dw ÞÞ tDt
 gt1 J n expðkn Dwt ÞqtDt g 2 S
1 1 2 n¼1 ij;n
kn Dwt ij

eij ¼ JS ij þ Brkk dij ð4Þ

2 3 ð7Þ
" #
J and B are shear compliance and bulk compliance, respec- 1 X
N
1expðkn Dw Þ t X
N t
etkk ¼ gt0 B0 þgt1 gt2 Bn gt1 gt2 Bn rkk
tively. Sij is the deviatoric stress and rkk is the volumetric 3 n¼1 n¼1
kn Dwt
stress. The deviatoric and volumetric viscoelastic strain  
1 tX
N
t tDt
t
tDt ð1expðkn Dw ÞÞ tDt
components can be expressed as  g1 Bn expðkn Dw Þqkk;n g2 rkk
3 n¼1 kn Dwt
Z t s s
1 1 t s dðg 2 S ij Þ ð8Þ
etij ¼ gt0 J 0 S tij þ gt1 DJ ðw w Þ ds ð5Þ
2 2 0 ds The incremental shear and bulk strains are also formulated
Z t s s
1 1 t s dðg r Þ and used in the finite element implementation, which are
etkk ¼ gt0 B0 rtkk þ gt1 DBðw w Þ 2 kk
ds ð6Þ
3 3 0 ds shown in Eqs. (9) and (10):

1.00E+08 1.00E+08

1.00E+07 1.00E+07
Complex Modulus (Pa)

Complex Modulus (Pa)

1.00E+06 1.00E+06

1.00E+05 1.00E+05

10 ºC 0.1Hz 20 ºC 0.1Hz
1.00E+04 10 ºC 1Hz 1.00E+04 20 ºC 1Hz
10 ºC 5Hz 20 ºC 5Hz
10 ºC 10Hz 20 ºC 10Hz
1.00E+03 1.00E+03
0.00E+00 1.00E+05 2.00E+05 3.00E+05 4.00E+05 5.00E+05 0.00E+00 1.00E+05 2.00E+05 3.00E+05 4.00E+05 5.00E+05
Stress (Pa) Stress (Pa)
(a) 10 oC (b) 20 oC

1.00E+08 1.00E+08
30 ºC 0.1Hz 40 ºC 0.1Hz
30 ºC 1Hz 40 ºC 1Hz
30 ºC 5Hz 40 ºC 5Hz
1.00E+07 1.00E+07
30 ºC 10Hz 40 ºC 10Hz
Complex Modulus (Pa)

Complex Modulus (Pa)

1.00E+06 1.00E+06

1.00E+05 1.00E+05

1.00E+04 1.00E+04

1.00E+03 1.00E+03
0.00E+00 1.00E+05 2.00E+05 3.00E+05 4.00E+05 5.00E+05 0.00E+00 1.00E+05 2.00E+05 3.00E+05 4.00E+05 5.00E+05
Stress (Pa) Stress (Pa)

(c) 30 oC (d) 40 oC

Fig. 1. The stress sweep test data of unaged asphalt binder.

 E. Masad et al. / Construction and Building Materials 22 (2008) 2170–2179 2173

De0t t tDt
ij ¼ eij  eij

1X N
  tDt
J n gt1 expðkn Dwt Þ  gtDt
¼ J t S tij  J tDt S ijtDt  1 qij;n ð9Þ
2 n¼1
  tDt   t 
1 tDt XN
tDt 1  expðkn Dw Þ t 1  expðkn Dw Þ
 g2 J n g1  g1 S ijtDt
2 n¼1 kn DwtDt kn Dwt
Detkk ¼ etkk  ekk
tDt

1X N
  tDt
¼ Bt rtkk  BtDt rkk
tDt
 Bn gt1 expðkn Dwt Þ  g1tDt qkk;n ð10Þ
2 n¼1
  tDt   t 
1 tDt XN
tDt 1  expðkn Dw Þ t 1  expðkn Dw Þ
 g2 Bn g 1 tDt
 g1 t rtDt
kk
3 n¼1 k n Dw k n Dw

This study employs the iterative scheme at material level 4. Experimental measurements and data analysis
to correct stress state from the current strain increment.
In the iterative scheme algorithm, the residual strain The DSR test was used to characterize the linear and
should be defined and the residual strain equation can nonlinear viscoelastic parameters of asphalt binders. The
be shown as test applies a sinusoidal, oscillatory stress to a thin disc
of asphalt between two parallel plates. The plate geome-
1
Rtij ¼ Detij þ Detkk dij  Detij ð11Þ tries used in this study were 8 mm in diameter with 2 mm
3 testing gap at low temperatures (10, 20 and 30 °C) and
where Detij is the mechanical strain provided from the struc- 25 mm in diameter with 1 mm testing gap at a high temper-
tural level. The Newton–Raphson typed iterative algorithm ature (40 °C). The binder was from a Venezuelan crude
is used to minimize the strain residual in Eq. (11). source and it was designated as 50 penetration grade

1.00E+08 1.00E+08
30 ºC 0.1Hz
30 ºC 1Hz
1.00E+07 1.00E+07 30 ºC 5Hz
30 ºC 10Hz
Complex Modulus (Pa)

Complex Modulus (Pa)

1.00E+06 1.00E+06

1.00E+05 1.00E+05

20 ºC 0.1Hz
1.00E+04 20 ºC 1Hz 1.00E+04
20 ºC 5Hz
20 ºC 10Hz
1.00E+03 1.00E+03
0.00E+00 1.00E+05 2.00E+05 3.00E+05 4.00E+05 5.00E+05 0.00E+00 1.00E+05 2.00E+05 3.00E+05 4.00E+05 5.00E+05
Stress (Pa) Stress (Pa)
(a) 20 oC (b) 30 oC

1.00E+08
40 ºC 0.1Hz
40 ºC 1Hz
40 ºC 5Hz
1.00E+07
40 ºC 10Hz
Complex Modulus (Pa)

1.00E+06

1.00E+05

1.00E+04

1.00E+03
0.00E+00 1.00E+05 2.00E+05 3.00E+05 4.00E+05 5.00E+05
Stress (Pa)

(c) 40 oC

Fig. 2. The stress sweep test data of aged asphalt binder.

2174 E. Masad et al. / Construction and Building Materials 22 (2008) 2170–2179

Ultimate Stress Levels The magnitude of the dynamic complex compliance jJ*j
25
Unaged 40C 10Hz and phase angle d were obtained from the test at each tem-
Unaged 40C 1 Hz perature and frequency. Then, the storage compliance J 0
20
and loss compliance J00 were calculated (J 0 = jJ*jcos d and
J00 = jJ*jsin d). The coefficients of the Prony series of the
N G* / Gini

15
shear compliance (Jn and kn) were determined by minimiz-
*

ing the difference between the shear compliance in the con-

10
5 !2
0 J 0Exp  1
0.00E+00 1.00E+04 2.00E+04 3.00E+04
Stress (Pa)
Fig. 3. The relationship between N jG j=jGini j and stress.
according to the British standard BS3690 (penetration of
49 dmm and softening point of 52 °C). Short-term aging
was conducted using the standard rolling thin film oven
(RTFO) according to the ASTM D 2872 procedure. The
unaged binder was tested at temperatures of 10, 20, 30
and 40 °C with frequencies of 0.1, 1, 5 and 10 Hz. The aged
binder was tested using temperatures 20, 30, 40 °C and fre-
quencies of 0.1, 1, 5 and 10 Hz. Stress sweeps at each tem-
perature/frequency combination were performed from the
minimum torque limit of the DSR to either the maximum
torque value or a 30% reduction in complex modulus.
stitutive model and the experimental measurements using
the error function in Eq. (12).
!2
J 0Model J 00Model
ERR ¼ þ ð12Þ
J 00Exp  1
4.00E+04 5.00E+04 6.00E+04
The stress sweep test data for the unaged and aged bind-
ers are shown in Figs. 1 and 2, respectively. These two fig-
ures show that the range of stress that can be applied on
the binder is a function of temperature and frequency.
This is considered a limitation for the modeling efforts
as the same range of stress levels is needed to establish
the master curve of a nonlinear viscoelastic material. In
order to overcome this limitation, the stress levels of each
test at a given temperature and frequency were normal-
ized by the ultimate stress level of that test. This ultimate
stress was determined following the method proposed in a
number of studies to determine the maximum stress that
an asphalt mix can sustain [17–19]. This method relies
on plotting the function N jG j=jGini j versus stress level,

1.00E+08 1.00E+08

1.00E+07 1.00E+07
Complex Modulus (Pa)

Complex Modulus (Pa)

1.00E+06 1.00E+06

1.00E+05 1.00E+05

10 ºC 0.1Hz 20 ºC 0.1Hz
1.00E+04 10 ºC 1Hz 1.00E+04 20 ºC 1Hz
10 ºC 5Hz 20 ºC 5Hz
10 ºC 10Hz 20 ºC 10Hz
1.00E+03 1.00E+03
0 0.2 0.4 0.6 0.8 1 1.2 0 0.2 0.4 0.6 0.8 1 1.2
Normalized Stress Normalized Stress
(a) 10 ºC (b) 20 ºC

1.00E+08 1.00E+08
30 ºC 0.1Hz 40 ºC 0.1Hz
30 ºC 1Hz 40 ºC 1Hz
30 ºC 5Hz 40 ºC 5Hz
1.00E+07 1.00E+07
30 ºC 10Hz 40 ºC 10Hz
Complex Modulus (Pa)

Complex Modulus (Pa)

1.00E+06 1.00E+06

1.00E+05 1.00E+05

1.00E+04 1.00E+04

1.00E+03 1.00E+03
0 0.2 0.4 0.6 0.8 1 1.2 0.00 0.20 0.40 0.60 0.80 1.00 1.20
Normalized Stress Normalized Stress
(c) 30 ºC (d) 40 ºC

Fig. 4. The stress sweep test data of unaged asphalt binder after normalizing by the ultimate stress.
 E. Masad et al. / Construction and Building Materials 22 (2008) 2170–2179 2175

1.00E+08 1.00E+08
30 ºC 0.1Hz
30 ºC 1Hz
30 ºC 5Hz
1.00E+07 1.00E+07
30 ºC 10Hz
Complex Modulus (Pa)

Complex Modulus (Pa)

1.00E+06 1.00E+06

1.00E+05 1.00E+05

20 ºC 0.1Hz
1.00E+04 20 ºC 1Hz 1.00E+04
20 ºC 5Hz
20 ºC 10Hz
1.00E+03 1.00E+03
0 0.2 0.4 0.6 0.8 1 1.2 0 0.2 0.4 0.6 0.8 1 1.2
Normalized Stress Normalized Stress
(a) 20 ºC (b) 30 ºC

1.00E+08
40 ºC 0.1Hz
40 ºC 1Hz
40 ºC 5Hz
1.00E+07
40 ºC 10Hz
Complex Modulus (Pa)

1.00E+06

1.00E+05

1.00E+04

1.00E+03
0 0.2 0.4 0.6 0.8 1 1.2
Normalized Stress
(c) 40 ºC

Fig. 5. The stress sweep test data of aged asphalt binder after normalizing by the ultimate stress.

where N is the data point number, jG*j is the magnitude Fig. 3. This normalization causes points at the same
of the dynamic complex modulus, and jGini j is the magni- distance from the maximum stress for the different test
tude of the initial dynamic complex modulus for each temperatures and frequencies to be represented by the
combination of temperature and frequency. Examples of same normalized stress level. The normalized data of
the data are shown in Fig. 3. The ultimate stress is se- the unaged and aged binders are shown as Figs. 4 and
lected at the end data point for the cross-annotation data 5, respectively.
type in Fig. 3; while the ultimate stress is selected at the The analysis was conducted at four different normal-
peak of the data for the square-annotation data type in ized stress levels (0.01, 0.6, 0.8 and 1.0) for both the
unaged and aged asphalt binders. In this study, the linear
viscoelastic behavior is selected at the lowest normalized
stress level of 0.01. The analysis was determined for the
1000 unaged binder, and then the behavior of the aged binder
0.01
0.6 was represented using an aging shift factor.
0.8
100
1.0
4.1. Time–temperature shift
10
A temperature 30 °C was used as the reference tempera-
aT

ture for the unaged binder. The TTSP was used to conduct
1
0 5 10 15 20 25 30 35 40 45
time–temperature shifting at each normalized stress level.
Nonlinear regression using least squares analysis was used
0.1 in the time–temperature shifting. The relationship between
the shift factor aT and temperature for different stress levels
0.01 is shown in Fig. 6. These results show that aT is indepen-
Temperature (ºC) dent of stress level; consequently, the aT values were aver-
Fig. 6. The relationship between temperature shift factor (aT) and aged for the different stress levels at each temperature as
temperature. presented in Table 1.
2176 E. Masad et al. / Construction and Building Materials 22 (2008) 2170–2179

Table 1 1.40E-04
0.01
The temperature shift factor for each temperature
0.6
1.20E-04
Temperature aT 0.8
1.0
10 225.00 1.00E-04
20 26.50
30 1.00

J(t) (1/Pa)
8.00E-05
40 0.10
6.00E-05

1.3 4.00E-05

2.00E-05
Nonlinear Parameter g1g2

1.2
0.00E+00
0.00E+00 2.00E+01 4.00E+01 6.00E+01 8.00E+01 1.00E+02 1.20E+02
Time (Sec)
1.1

Fig. 8. The master curve after stress vertical shifting.

1.60E-04
0.01
1.40E-04 0.6
0.9
0 0.2 0.4 0.6 0.8 1 1.2 0.8
Normalized Stress 1.20E-04 1.0

Fig. 7. The relationship between nonlinear parameter (g1g2) and normal-

1.00E-04
ized stress level.
J(t) (1/Pa)

8.00E-05

4.2. Nonlinearity stress shift 6.00E-05

4.00E-05
The nonlinear parameters g1g2 were determined by ver-
tically shifting the higher stress master curves to the linear 2.00E-05
stress level. The results in Fig. 7 show increasing nonlinear
0.00E+00
parameters with an increase in stress level; the nonlinear 0.00E+00 2.00E+01 4.00E+01 6.00E+01 8.00E+01 1.00E+02 1.20E+02 1.40E+02
parameters are given in Table 2. The experimental data Time (Sec)
after vertical shifting are shown in Fig. 8. Fig. 9. The master curve after stress horizontal shifting.
The master curves at different stress levels were shifted
horizontally, shown as Fig. 9, in order to predict the long
term behavior of the asphalt binder, and the Prony series Table 3
was calibrated to fit the long term response of the binder. Linear viscoelastic coefficients of the Prony series
The stress horizontal shift factors are shown as Table 2, n Jn kn
and the Prony coefficients are shown in Table 3. The Prony 0 3.57E09 –
series (similar to that in Eq. (3)) fitted to the long term 1 1.03E07 18.23
response was used in representing the linear viscoelastic 2 2.13E06 0.25
3 2.71E08 146.01
response of the binder. The stress horizontal shifting is 4 4.56E05 0.01
advantageous as it allows prediction of the linear binder 5 8.21E07 0.88
response at longer time periods by performing experiments 6 3.16E07 3.54
at stress levels higher than the linear range but at shorter 7 6.61E06 0.07
time intervals. The g1g2 can be further used to predict the 8 4.90E01 1.71E06
nonlinear response of the asphalt binders at the long term
intervals.
4.3. Aging shift
Table 2
The nonlinear parameters g1g2 and stress shift factors at different This study obtained the aging shift factor by horizontal
normalized stress levels shifting the aged asphalt binder data to the unaged asphalt
Normalized stress g1g2 as binder data. The aging shift factor was calculated for each
0.01 1.00 1.00 temperature and normalized stress levels. Examples of the
0.6 1.08 0.96 experimental data before and after aging shifting are shown
0.8 1.13 0.93 as Fig. 10. By evaluating the average and variance in Table
1.0 1.21 0.84
4 of the aging shift factor for the different temperatures and
 E. Masad et al. / Construction and Building Materials 22 (2008) 2170–2179 2177

2.50E-05 5. Finite element analysis of binder response

Unaged
Aged
2.00E-05 The material parameters of the unaged binder
(aT, as, g1g2, Prony series coefficients of long term response),
and the aging shift factors, were used as the input to the
1.50E-05
J(t) (1/Pa)

finite element subroutine. The finite element analysis was

used to calculate the creep response at each of the temper-
1.00E-05
ature and normalized stress levels. The model consisted of
a three-dimensional element (C3D8R) representing the
5.00E-06

2.50E-06
0.00E+00
0 2 4 6 8 10 12
Time (Sec)
(a) Before applying the aging shift factor. 2.00E-06

2.50E-05
1.50E-06

J(t) (1/Pa)
Unaged
Aged
0.01 FEM
2.00E-05
1.00E-06 0.01 Experiment
0.6 FEM
0.6 Experiment
1.50E-05 0.8 FEM
J(t) (1/Pa)

5.00E-07
0.8 Experiment
1.0 FEM
1.00E-05 1.0 Experiment
0.00E+00
0.0E+00 2.0E+00 4.0E+00 6.0E+00 8.0E+00 1.0E+01 1.2E+01 1.4E+01
Time (Sec)
5.00E-06
(a) Unaged binder

0.00E+00 2.50E-06
0 2 4 6 8 10 12 0.01 FEM
Time (Sec) 0.01 Experiment
0.6 FEM
(a) After applying the aging shift factor. 2.00E-06
0.6 Experiment
0.8 FEM
Fig. 10. Data before and after applying the aging shift factor.
0.8 Experiment
1.50E-06
J(t) (1/Pa)

1.0 FEM
1.0 Experiment

stress levels, it can be concluded that the aging shift factor 1.00E-06

is mostly a function of temperature, while it is almost inde-

pendent of normalized stress levels. Hence, the average 5.00E-07
aging shift factor for all stress levels at each given temper-
ature was calculated and used to represent the response of
0.00E+00
the aged binder. This finding indicates that the behavior of 0.0E+00 2.0E+00 4.0E+00 6.0E+00 8.0E+00 1.0E+01 1.2E+01 1.4E+01
Time (Sec)
the aged asphalt binder can be obtained by using unaged
(b) Aged binder
binder parameters (temperature shift factor, nonlinear
parameters, Prony series coefficients) and the aging shift Fig. 11. Verification of the finite element analysis at a temperature of
factor for each temperature. 20 °C.

Table 4
The aging shift factor for each combination of temperature and normalized stress levels
Temperature °C Aging shift factors
Normalized stress Averaged at each temperature Variance between stress levels
0.01 0.6 0.8 1.0
20 1.3 1.15 1.05 1.05 1.14 0.0140
30 2.9 2.6 2.5 2.4 2.60 0.0467
40 2.7 2.7 2.65 2.6 2.66 0.0023
Averaged at each stress level 2.3 2.15 2.07 2.02
Variance between temperatures 0.7600 0.7525 0.7808 0.7108
2178 E. Masad et al. / Construction and Building Materials 22 (2008) 2170–2179

4.50E-06 asphalt specimen subjected to shear creep loading. The

Converted FE result
4.00E-06
Exp. data
deformation and rotation of the nodes in the bottom of
Storage Compliance J' (1/Pa)

3.50E-06 the element were fixed, while the different stress levels were
3.00E-06 applied at the upper face of the element. For brevity, only
2.50E-06 the results of unaged and aged binders at a temperature of
2.00E-06 20 °C are shown in Fig. 11. The results show that the model
1.50E-06
gives a very good prediction of the experimental measure-
1.00E-06
ments. The results in Fig. 11b indicate that the response
5.00E-07
of the aged binder can be obtained by simply shifting the
data of the unaged binder by the aging factor ag.
0.00E+00
0 2 4 6 8 10 12 The Prony series was fitted to the numerical results, and
Frequency (rad/sec)
(a) Storage compliance (J') the coefficients of this series were used to determine the bin-
der response in the frequency domain. Consequently, the
2.00E-05 numerical results in the frequency domain were compared
Converted FE result
1.80E-05
Exp. data
with the DSR experimental data. Examples of the unaged
1.60E-05 and aged comparisons of J 0 and J00 for linear and nonlinear
Loss Compliance J" (1/Pa)

1.40E-05 responses are shown in Figs. 12 and 13. In general, the

1.20E-05 results show that the numerical results are in very good
1.00E-05
agreement with the experimental measurements for the lin-
8.00E-06
ear and nonlinear stresses (see Figs. 12 and 13).
6.00E-06

4.00E-06
6. Conclusions
2.00E-06

0.00E+00
0 2 4 6 8 10 12 This paper presents a framework for the analysis of the
Frequency (rad/sec)
nonlinear viscoelastic behavior of unaged and aged asphalt
(b) Loss compliance (J")
binders at different temperatures and stresses. Due to the
Fig. 12. Viscoelastic properties (J 0 and J00 ) for unaged binder at 30 °C and influence of test temperatures on the stress levels that the
linear stress level (normalized stress of 0.01). binder can sustain prior to failure, the experimental mea-
surements do not have a common range of stress levels that
6.00E-06 can be used in developing the master curve of nonlinear vis-
Converted FE result
coelastic materials. This limitation was overcome by intro-
5.00E-06 Exp. datal
ducing the normalized stress concept in which the stress
Storage Compliance J' (1/Pa)

4.00E-06
values of each test at a given temperature and frequency
were normalized by the ultimate stress of that test. The
3.00E-06 response at the actual stress is obtained by dividing the
actual stress by the ultimate stress and using the normal-
2.00E-06
ized stress in the model.
1.00E-06 The data at the different stress levels were shifted hori-
zontally to obtain the long term response of the binder.
0.00E+00
0 2 4 6 8 10 12
This means that the long term response of the binder can
Frequency (rad/sec) be obtained by conducting short term tests at multiple
(a) Storage compliance (J') stress levels. The nonlinear response of the binder is deter-
2.50E-05
mined by vertical shifting between the nonlinear stresses
Converted FE result and the linear stress. The advantages of the analysis
Exp. data approach can be realized in providing a mathematical
Loss Compliance J" (1/Pa)

2.00E-05
framework for describing the nonlinear response of asphalt
1.50E-05 binders, and in the possibility of describing the behavior of
aged binders by using the unaged binder parameters
1.00E-05 (aT, as, g1g2, Prony series coefficients of long term response)
and aging shift factors.
5.00E-06 The analysis conducted on this paper was limited to only
one asphalt binder source. It is necessary to evaluate the
0.00E+00
0 2 4 6 8 10 12
applicability of the analysis method for more asphalt bind-
Frequency (rad/sec) ers that exhibit different properties from different sources.
(b) Loss compliance (J") Specifically, chemically modified binders need to be exam-
Fig. 13. Viscoelastic properties (J 0 and J00 ) for unaged binder at 30 °C and ined in order to determine the applicability of the aging
nonlinear stress level (normalized stress of 1). shifting factors to these binders.
 E. Masad et al. / Construction and Building Materials 22 (2008) 2170–2179
References
[1] Ferry JD. Viscoelastic properties of polymers. New York: John Wiley
and Sons, Inc.; 1961.
[2] Cheung CY, Cebon D. Deformation mechanisms of pure bitumen. J
Mater Civil Eng 1997:117–29.
[3] Cheung CY, Cebon D. Thin film deformation behavior of power-law
creeping materials. J Eng Mech 1997:1138–52.
[4] Airey GD, Rahimzadeh B, Collop AC. Linear viscoelastic limits
of bituminous binders. J Assoc Asphalt Paving Technol 2002:
160–96.
[5] Airey GD, Rahimzadeh B, Collop AC. Linear rheological
behavior of bituminous paving materials. J Mater Civil Eng
2004:212–20.
[6] Kose S, Guler M, Bahia H, Masad E. Distribution of strains within
hot-mix asphalt binders. J Transport Res Board 2000;1728:21–7.
[7] Masad E, Somadevan N. Microstructural finite-element analysis of
influence of localized strain distribution of asphalt mix properties. J
Eng Mech 2002:1105–14.
[8] Christensen RM. On obtaining solutions in nonlinear viscoelasticity.
J Appl Mech 1968;35:129–33.
[9] Schapery RA. On the characterization of nonlinear viscoelastic
materials. Polym Eng Sci 1969;9:295–310.
[10] Schapery RA. Nonlinear viscoelastic solids. Int J Solids Struct
2000:359–66.
2179
[11] Lou YC, Schapery RA. Viscoelastic characterization of a nonlinear
fiber-reinforced plastic. J Compos Mater 1971;5:208–34.
[12] Shields DH, Zeng M, Kwok R. Nonlinear viscoelastic behavior of
asphalt concrete in stress relaxation. J Assoc Asphalt Paving Technol
1998;67:358–400.
[13] Touati D, Cederbaum G. On the prediction of stress relaxation from
known creep of nonlinear materials. J Eng Mater Technol
1997;119:121–4.
[14] Touati D, Cederbaum G. Post buckling of nonlinear viscoelastic
imperfect laminated plates part I: material considerations. Compos
Struct 1998;42:33–41.
[15] Haj-Ali RM, Muliana AH. Numerical finite element formulation of
the Schapery non-linear viscoelastic material model. Int J Numer
Methods Eng 2004;59:25–45.
[16] Sadd MH, Parameswaran DQV, Shukla A. Simulation of asphalt
materials using finite element micromechanical model with damage
mechanics. J Transport Res Board 2004:86–95.
[17] Reese R. Properties of aged asphalt binder related to asphalt concrete
life. J Assoc Asphalt Paving Technol 1997;66:604–32.
[18] Rowe GM, Bouldin MG. Improved techniques to evaluate fatigue
resistance of asphaltic mixes. In: Proceeding of the second enraphalt
and eurobitume congress, Barcelona, 2000.
[19] Kim YR. Mechanistic fatigue characterization and damage modeling
of asphalt mixtures. PhD dissertation, Texas A&M University,
College Station, TX, 2003.