Professional Documents
Culture Documents
(Asce) MT 1943-5533 0002707
(Asce) MT 1943-5533 0002707
Abstract: This paper presents a methodology to obtain creep and recovery response of asphalt binders merely by using frequency sweep
tests, which are relatively easier to conduct. This proposed methodology would be useful for practitioners and researchers in modeling time-
dependent rutting performance of asphalt binders using simple linear viscoelastic measurements obtained from dynamic shear rheometer
(DSR). Frequency sweep tests were conducted on four different asphalt binders at temperatures ranging from 10°C to 70°C. Complex modulus
master curves were constructed for 40°C, 50°C, and 60°C followed by presmoothing using a power law and modeling using a Prony series. The
estimated complex modulus, which is in frequency domain, was translated into the relaxation modulus, a time domain entity. The relaxation
modulus was converted to creep compliance using Laplace transform. The resultant creep compliance was modeled using a generalized Burgers
model (GBM), which was used for predicting creep and recovery response of asphalt binders for different loading and unloading periods. The
predicted creep and recovery profiles were successfully validated with the multiple stress creep and recovery (MSCR) measurements obtained at
40°C, 50°C, and 60°C for all binders. DOI: 10.1061/(ASCE)MT.1943-5533.0002707. © 2019 American Society of Civil Engineers.
where
Frequency Sweep Test
A frequency sweep test was performed, using a DSR, on all the ðε1 − ε10 Þ · 100
asphalt binders. In the initial part of the study, a strain sweep test εr ðσ; NÞ ¼ ð4Þ
ε1
was conducted to measure the LVE strain limit for all the asphalt
binders at different frequencies and temperatures. The magnitude
Similarly, the average percent nonrecoverable creep compliance
of strain in the frequency sweep test was chosen to ensure that
was calculated as
the asphalt binders remained in the LVE domain. The frequency
was varied from 0.1 to 20 Hz. The test was conducted from
J nr ðσÞ ¼ sumðJ nr ðσ; NÞÞ=10 ð5Þ
10°C to 70°C, with a 10°C increment in temperature. The spindle
was 8 mm in diameter with a 2-mm plate gap for temperatures
where
between 10°C and 40°C, and was 25 mm in diameter with a
1-mm plate gap for the remaining temperatures.
ðε10 Þ
J nr ðσ; NÞ ¼ ð6Þ
σ
MSCR Test
An MSCR test was performed to evaluate the performance of the The average nonrecoverable creep compliance (J nr ) (up to three
binders in rutting at 40°C, 50°C, and 60°C. These temperatures significant digits) and average percent recovery (R) at the three test
correspond to the high pavement temperatures at which rutting temperatures are given in Table 2 and Fig. 1. Because the main
could be the predominant failure criteria. The MSCR test was con- objective of the paper is to evaluate the suitability of analytical
ducted on asphalt binder specimens in accordance with AASHTO techniques for predicting the rutting response of asphalt binders,
TP70 (AASHTO 2012). DSR using 25-mm spindle and 1-mm gap a detailed comparison of the performance of these binders is not
was used for conducting the test. Despite the fact that a standard presented.
MSCR test is performed at stress levels of 1 and 3.2 kPa, several
studies have correlated the rutting behavior of asphalt mixes cor-
responding to rut depths measured with a wheel rut tester with the
creep and recovery response at 3.2 kPa. In this study also, the re- Table 2. Average J nr (kPa−1 ) values at different temperatures
sponse measurement for the binders was restricted to a stress level Stress level 3.2 kPa
of 3.2 kPa. Ten cycles of loading were given at the stress level.
Binder 40°C 50°C 60°C
The rutting susceptibility of the binders was evaluated by calcu-
lating the nonrecoverable creep compliance (J nr ) and percent VG-10 0.081 0.588 2.760
recovery (R) from the measured data averaged over the 10 cycles. VG-30 0.055 0.414 1.868
PMB-1 0.002 0.023 0.176
The creep strain at the beginning and end of each creep cycle was
PMB-2 0.000 0.008 0.032
denoted as ε0 and εc . The strain at the end of each recovery cycle
100 50˚C
Constructing the Master Curves
Average % Recovery
Fig. 1. Average percent recovery (R) for different binders. Presmoothing and Modeling Complex Modulus
Master Curves
Different engineering analysis of LVE materials requires developing
explicit analytical expressions for laboratory test data. One such
Derivation of Creep Compliance from Frequency analytical function is the complex modulus spectra as a function
Sweep Data of reduced frequency acquired after construction of the master curve.
The literature highlights different analytical expressions proposed
This paper proposes a methodology to obtain creep compliance us- for these LVE materials. One of the widely used expressions is
ing frequency sweep data. The methodology involves integration of the Prony or the Dirichlet series, which is derived from spring
various steps previously found in the literature that have not been and dashpots models. The complex modulus representation of the
used in a synergetic manner before. Fig. 2 presents the outline of Prony series can be mathematically written as (Ebrahimi et al. 2014)
Complex Modulus, Pa
Complex Modulus, Pa
1.E+06 1.E+06
1.E+05 1.E+05
1.E+04 1.E+04
VG 10 VG 30
1.E+03 VG 10 VG 30 1.E+03
PMB 1 PMB 2
PMB 1 PMB 2
1.E+02 1.E+02
Downloaded from ascelibrary.org by Indian Institute of Technology (BHU), Varanasi on 12/19/22. Copyright ASCE. For personal use only; all rights reserved.
1.E+01 1.E+01
1.E+00 1.E+00
1.E-05 1.E-02 1.E+01 1.E+04 1.E-04 1.E-01 1.E+02 1.E+05
(a) Reduced Frequency, Hz (b) Reduced Frequency, Hz
1.E+08
1.E+07
Complex Modulus, Pa
1.E+06
1.E+05
1.E+04
1.E+03
1.E+02 VG 10 VG 30
1.E+01 PMB 1 PMB 2
1.E+00
1.E-03 1.E+00 1.E+03 1.E+06
(c) Reduced Frequency, Hz
Fig. 3. Complex modulus master curves for various binders at the temperatures of (a) 40°C; (b) 50°C; and (c) 60°C.
X
n
ω2 ρ2j Xn
ωρj The symbol Ge denotes the equilibrium modulus as in the pre-
G ðωÞ ¼ Ge þ Gj 2 2
þi Gj ð7Þ ceding expression. The curve was fitted by minimizing the sum of
1 þ ω ρj 1 þ ω2 ρ2j
j¼1 j¼1 the squared difference between the measured and modeled data.
The master curves for all the binders at different reference temper-
where Ge = equilibrium modulus; Gj = relaxation strength; ρj = atures were fitted with different values of M (starting from 1) and
relaxation time; and ω = frequency (Hz). one such solution is presented in Fig. 4 for VG-30 at a reference
Different researchers have proposed simple and straightforward temperature of 40°C. The optimum solution for the least-square
methods for fitting Prony series and finding the model parameters minimization technique corresponded to M ¼ 6.
Ge , Gj , and ρj . However, these methods usually produce negative
values for model coefficients, which are not realistic. These neg-
ative values are associated with high variance in data, which is com-
monly seen in complex modulus master curves. The methodology
proposed in this paper utilized the modified power law (MPL) pre- 1.E+08
sented by Park and Kim (2001) for presmoothing the complex
modulus master curves and thus overcome the previous issue re- 1.E+07 R2 =0.992
lated to the fitting of a Prony series. The presmoothed series pre-
Complex Modulus, Pa
1.E+06
serves the representation of the data in a large domain without
being affected by any significant variance in the data at the local 1.E+05
level. The power law model, however, lacks computational effi-
1.E+04
ciency, specifically when Laplace transform is involved for inter- measured
conversion of material functions. Such cases require a Prony series 1.E+03 MPL presmooth
representation, whose Laplace transform can be computed with
1.E+02
greater ease. The MPL proposed for creep compliance by Park
and Kim has been used for complex modulus master curves in this 1.E+01
study using the following mathematical expression:
1.E+00
1.E-04 1.E-01 1.E+02 1.E+05
X
M
Ĝi
G ðωÞ ¼ Ge þ ð8Þ Reduced Frequency, Hz
τ̂ i n
j¼1 1 þ ω
Fig. 4. MPL presmoothing applied to complex modulus master curve
of VG-30 at a reference temperature of 40°C.
where Ge ; Ĝi ; τ̂ i ði ¼ 1; 2; 3; : : : ; MÞ; M; and n are all constants.
1.E+07 1.E+07
Relaxation Modulus, Pa
R2=0.99
Complex Modulus, Pa
1.E+06 1.E+06
1.E+05 1.E+05
1.E+04 1.E+04
measured, MPL
1.E+03 Prony fit 1.E+03
1.E+02 1.E+02
Downloaded from ascelibrary.org by Indian Institute of Technology (BHU), Varanasi on 12/19/22. Copyright ASCE. For personal use only; all rights reserved.
Relaxation Modulus
1.E+01 1.E+01
1.E+00 1.E+00
1.E-05 1.E-02 1.E+01 1.E+04 1.E+07 1.E-08 1.E-05 1.E-02 1.E+01
(a) Reduced Frequency, Hz (b) Reduced Time, seconds
Fig. 5. (a) Prony fit complex modulus master curve for PMB-1 at 60°C; and (b) relaxation modulus master curve obtained for PMB-1 at 60°C.
Converting Complex Modulus G ω to Relaxation Converting Relaxation Modulus Et to Creep
Modulus Et Compliance Dt
Since the objective in the first stage of the study was to estimate the The creep compliance and relaxation modulus are not reciprocal
creep compliance from complex modulus values, it was important but are related by a convolution integral (Ferry 1980) as follows:
to convert the complex modulus, which is in the frequency domain, Z t
to the relaxation modulus, which is in time domain. Hence, as an Eðt − τ ÞDðτ Þdτ ¼ t; for t > 0 ð10Þ
initial step, the complex modulus master curve obtained from 0
laboratory measurements was fitted using the Prony series as pre-
sented in Eq. (1). The coefficients obtained by fitting the master where E and D = time-dependent relaxation modulus and creep
curve were applied for evaluating the relaxation modulus using compliance, respectively. Laplace transform of the previous equa-
the following expression: tion further yields
X
M 1
EðtÞ ¼ Ee þ Ej e−t=ρj ð9Þ ĒðsÞD̄ðsÞ ¼ ð11Þ
s2
j¼1
where s = transform parameter; and the Laplace transform of fðtÞ is
where Ee = equilibrium modulus (also called the long-term relax- represented as fðsÞ ≡ ∫ ∞ st
0 fðtÞe dt.
ation modulus); Ej = relaxation strength analogous to Gj obtained Different numerical integration schemes may be adopted to
by fitting the MPL curve with Eq. (7); and ρj = relaxation time. solve the preceding equation. In this paper, the power-law-based
Model parameters were obtained by minimizing the sum of the interrelationship was used, assuming that relaxation modulus and
squared difference between the MPL and modeled values. The creep compliance values are approximately represented using a
generalized reduced gradient (GRG) nonlinear technique was used simple power law over their transition zones. Often this technique
for solving the minimization problem. Fig. 5 illustrates a 10-term provides good approximation for relationship between EðtÞ
Prony coefficient fitted to the complex modulus master curve for and DðtÞ for LVE materials. Moreover, the low computational
PMB-1 at a temperature of 60°C. The Prony series was fitted to the requirement makes it a robust tool for efficient modeling and
presmoothed complex modulus data as obtained using MPL. This analysis of material properties (Park and Kim 2001). The power
is denoted as “measured, MPL” in Fig. 5(a). The resulting relax- laws for relaxation modulus and creep compliance can be written
ation modulus curve is depicted in Fig. 5(b). The Prony series as follows:
coefficients obtained for this data set are presented in Table 3.
EðtÞ ¼ E1 t−n ð12Þ
Table 3. Prony series coefficients obtained for PMB-1 at 60°C DðtÞ ¼ D1 tn ð13Þ
Gj (Pa) ρj (s)
1.05 × 107 1.90 × 10−7 where E1 , D1 , and n = positive constants.
1.41 × 107 6.60 × 10−7 Taking the s-multiplied Laplace transform of Eq. (6), we get
3.93 × 106 5.56 × 10−6
1.08 × 106 8.59 × 10−5 E1 Γð1 − nÞ
sĒ ¼ ð14Þ
4.08 × 105 6.88 × 10−4 s−n
6.38 × 104 8.07 × 10−3
9.68 × 103 7.84 × 10−2 where Ē = Laplace transform; and Γ is the gamma function.
1.43 × 103 9.47 × 10−1 From Eq. (5)
1.60 × 102 1.62 × 10−1
1.99 × 102 1.34 × 101 1
D̄ ¼ ð15Þ
Note: Gg ¼ 1.00 × 10 −1
Pa. E1 Γð1 − nÞs1þn
1.E-01 1.E-01
R2=0.979 R2=0.966
Compliance, Pa -1
Compliance, Pa-1
1.E-02 1.E-02
1.E-03 1.E-03
1.E-05 1.E-05
1.E-06 1.E-06
0.E+00 5.E+00 1.E+01 2.E+01 2.E+01 3.E+01 0.E+00 5.E-01 1.E+00 2.E+00
(a) Time, Seconds (b) Time, Seconds
Fig. 6. (a) Interconverted and measured compliance for VG-30 at 40°C; and (b) interconverted and measured compliance for PMB-1 at 60°C.
Further solving the preceding equation yields Like the complex modulus, the long-term creep behavior of asphalt
binders was evaluated using creep compliance ½JðtÞ master curves
sin nπ 1 that were generated for 40°C, 50°C, and 60°C temperatures using
DðtÞ ¼ ð17Þ MSCR test data. Figs. 7(a–d) depict the creep compliance master
nπ E1 t−n
curves for all binders at 3.2 kPa and a reference temperature of
50°C. The comparison of modeled and measured values of creep
Eq. (11) can be simplified as compliance as shown in Fig. 6(a and b) was made using these
master curves.
sin nπ
DðtÞEðtÞ ¼ ð18Þ
nπ
Modeling Creep Compliance Master Curve
In this equation, n is the local log-log slope of the source func-
As a viscoelastic material, bitumen can be characterized using vari-
tion, Eðτ Þ or Dðτ Þ, depending on which parameter should be
ous mechanical models including a combination of springs and
converted; n can be mathematically written as
dashpots placed in varying combinations. Many studies have dem-
onstrated the use of the four-element Burgers model (FEBM) for
d log Fðτ Þ
n¼ ð19Þ characterizing the viscoelastic response of asphalt binders obtained
d log τ atτ ¼t through creep and recovery tests (Hajikarimi et al. 2015; Yang et al.
2006; Zhou et al. 2009). FEBM is a combination of the Maxwell
where Fðτ Þ is the source function. The preceding solution gives model (spring and dashpot placed in series) and Kelvin-Voigt
accurate results in regions where EðtÞ and DðtÞ are approximated model (spring and dashpot placed in parallel) placed in series
using straight lines in log-log scale. As can be seen in Fig. 5(b), (Fig. 8). The mathematical form of a Burgers model can be derived
this is not the actual case. Park and Schapery (1999) proposed a by considering the strain response under constant stress of each
different form of interconversion for cases where the relationship coupled element in series as depicted in Fig. 8. The total strain
deviates from straight line. The mathematical relationship can be εB at time t is a sum of the strains in these three elements, where
written as the spring and dashpot in the Maxwell model are considered as two
elements. The subscripts B, M, and K indicate Burgers model,
1 Maxwell elements, and Kelvin-Voigt elements, respectively, while
DðtÞ ¼ ð20Þ
EðαtÞ εM1 , εM2 , and εK are the strains of the Maxwell spring, Maxwell
dashpot, and Kelvin-Voigt unit, respectively. The total strain of the
where α ¼ ½ðsin nπÞ=ðnπÞ1=n , with n being the same as in Eq. (13). material has three components, namely, an instantaneous elastic
After getting the Prony fitted relaxation modulus as depicted in strain, a viscous strain, and a retarded strain. Under constant stress,
the Eq. (3), Eq. (14) was used for converting relaxation modulus the creep and recovery response of the asphalt binder can be written
EðtÞ to creep compliance DðtÞ. The preceding interconversion in terms of compliance as
method gives fairly accurate results for both unmodified and modi-
fied binders. For illustration, the interconverted values of DðtÞ for 1 t 1
DðtÞ ¼ þ þ ð1 − e−t=ψ Þ ð21Þ
VG-30 at a reference temperature of 40°C and for PMB-1 at a tem- EM η M EK
perature of 60°C are shown in Figs. 6(a and b) along with the mea-
sured responses. The details of the measured value of creep
τ 1
compliance will be discussed subsequently. As can be seen, the in- DðtÞ ¼ þ ð1 − e−t=ψ Þeðt−τ Þ=ψ ð22Þ
terconverted values are in agreement with the measured response. η M EK
1.E-04 1.E-04
Downloaded from ascelibrary.org by Indian Institute of Technology (BHU), Varanasi on 12/19/22. Copyright ASCE. For personal use only; all rights reserved.
1.E-05 1.E-05
1.E-02 1.E-01 1.E+00 1.E+01 1.E-02 1.E-01 1.E+00 1.E+01
(a) Reduced Time, Seconds (b) Reduced Time, Seconds
1.E-02 1.E-02
Creep Compliance, Pa-1
1.E-04
1.E-04
1.E-05
1.E-06 1.E-05
1.E-01 1.E+00 1.E+01 1.E+02 1.E-02 1.E-01 1.E+00 1.E+01
(c) Reduced Time, Seconds (d) Reduced Time, Seconds
Fig. 7. Creep compliance master curves at 50°C for (a) VG-10; (b) VG-30; (c) PMB-1; and (d) PMB-2.
where EM , ηM , EK , and ηK = model parameters pertaining to modu- Qualitatively, a Burgers model represents the behavior of a vis-
lus and viscosity of Maxwell and Kelvin elements, respectively. coelastic material. Quantitatively, a single Kelvin model is usually
The viscous flow behavior parameter ηM is considered an important not sufficient to incorporate the long duration over which the
parameter to describe the rutting potential of different binders. The retarded strain takes place, and several Kelvin models could be
retardation time taken to produce 63.2% of the total deformation in needed (Huang 2010). The creep and recovery compliance under
the Kelvin unit is Ψ ¼ ηK =EK (Yang et al. 2006). a constant stress for the generalized model can be written as
Strain
1.E-04 2.E-01
1.E-05 2.E-01
1.E-06 1.E-01 measured
Downloaded from ascelibrary.org by Indian Institute of Technology (BHU), Varanasi on 12/19/22. Copyright ASCE. For personal use only; all rights reserved.
R2=0.966
1.E-07 5.E-02 modelled
1.E-08 0.E+00
1.E-07 1.E-05 1.E-03 1.E-01 1.E+01 0.E+00 5.E+00 1.E+01 2.E+01
(a) Reduced Time, Seconds (b) Time, Seconds
Fig. 9. (a) GBM model fit for PMB-2 at 50°C; and (b) measured and modeled response of creep and recovery for PMB-2 at 50°C.
Table 4. Six-element GBM for PMB-2 at 50°C example, the 6-term GBM model fit for PMB-2 at a reference tem-
EKi (Pa) ψi perature of 50°C is shown in Fig. 9(a). The corresponding creep
and recovery profile and the measured response are depicted in
6.34 × 10−7 1.56 × 102 Fig. 9(b). The GBM model parameters obtained for this binder
7.24 × 10−7 1.38 × 103
are also presented in Table 4. It can be seen that the modeled
1.66 × 10−6 5.38 × 103
4.93 × 10−5 6.87 × 104 and measured response are in good agreement. Similar results were
1.00 × 10−4 7.80 × 105 obtained for other binders at different reference temperatures.
1.20 × 100 1.83 × 105 The curves are not all presented here for brevity. For example,
Figs. 10(a and b) present the modeled and measured creep and re-
Note: EM ¼ 6.64 × 10−7 Pa; and ηM ¼ 1.27 × 10−6 Pa · s.
covery response of VG-10 and VG-30 at a reference temperature of
40°C. The validity of the model is self-explanatory. The selection of
n depends on the domain of the master curve. This paper suggests
1 t X
n
1 that n should be equal to the number of decades covered on a
DðtÞ ¼ þ þ ð1 − e−t=ψi Þ ð23Þ log-scale by the master curve. This is corroborated by a study by
EM ηM i¼1 EKi
Park and Kim (2001).
τ X
n
1
Dðt 0 Þ ¼ þ ð1 − e−t=ψi Þeðt−τ Þ=ψi ð24Þ Conclusions
ηM i¼1 EKi
The study appraised the use of frequency sweep data, obtained
The interconverted values of DðtÞ as obtained previously were using a DSR, for prediction of creep and recovery response of as-
modeled using Eq. (23) and the same model was used for predicting phalt binders. The process included modeling and interconversion
the recovery profile of the binder using Eq. (24). The values of DðtÞ techniques at various stages. It was found that presmoothing of data
acquired from MSCR (for 3.2-kPa stress level) test were compared using the modified power law can be beneficially used for modeling
with the modeled values. The domain for checking the suitability complex modulus master curves using a Prony series. The intercon-
of a generalized Burgers model (GBM) was chosen based on the verted values of creep compliance using the method proposed
MSCR test, i.e., 0–1 s of creep and 1.1–9 s of recovery. As an by Park and Schapery (1999) were in good agreement with the
5.E-01 4.E-01
5.E-01
3.E-01
4.E-01 R2=0.991 R2=0.981
4.E-01 3.E-01
3.E-01 2.E-01
Strain
Strain
3.E-01
measured
2.E-01 2.E-01
modelled
2.E-01 1.E-01
1.E-01 measured
5.E-02 modelled
5.E-02
0.E+00 0.E+00
0.E+00 5.E+00 1.E+01 2.E+01 0.E+00 5.E+00 1.E+01 2.E+01
(a) Time,Seconds (b) Time, Seconds
Fig. 10. Validation of creep and recovery at 40°C for (a) VG-10; and (b) VG-30.