Professional Documents
Culture Documents
(Asce) MT 1943-5533 0001078 PDF
(Asce) MT 1943-5533 0001078 PDF
Abstract: Several prediction criteria have been proposed to rank asphalt binders based on their rutting performance, and each one has been
Downloaded from ascelibrary.org by Politeknik Negeri Bandung on 02/08/18. Copyright ASCE. For personal use only; all rights reserved.
shown to be able to predict rutting behavior observed in either accelerated loading field tests or under highway traffic with some level of
success. In this study, an analysis of these rutting prediction criteria was conducted in order to identify the virtues and shortcomings of each.
Asphalt binders that were used in the accelerated loading facility (ALF) study conducted by the Federal Highway Administration (FHWA)
were obtained and studied. The experimental data obtained on these binders were fit with a nonlinear viscoelastic model so that it can be used
to predict their response under different loading conditions. By comparing predicted behavior based on the nonlinear viscoelastic model and
the various rutting prediction criteria, the part of the complex nonlinear mechanical behavior of these binders that each criterion addresses was
determined. Results show that each criterion represents the resistance offered by asphalt binders to permanent strain over a specific range of
stresses. Results also show that a simple viscosity test might be sufficient to characterize the rutting performance of asphalt binders accurately.
DOI: 10.1061/(ASCE)MT.1943-5533.0001078. © 2014 American Society of Civil Engineers.
Author keywords: Asphalts; Rutting; Viscoelasticity; Nonlinear response; Laboratory tests; Apparent viscosity.
where i represents the ith data point; M̄ and F̄ are the steady-state But the assumption that the shear stress varies linearly along the
torque and normal force from the steady-shear experiments, respec- radial direction is invalid when the mechanical response of the as-
tively; γ_ ss is the shear-rate level at which the steady-shear experi- phalt binder is nonlinear. Considering this issue, raw data of angle
ment is conducted; γ and F are the creep strain and the normal force of twist per unit length (ϕ) was back-calculated from the nonrecov-
from the creep-recovery test at the highest torque level, respec- ered creep compliance data J nr and was compared to the model
tively; and wN is the weight given to the errors in the predictions predictions. Model predictions were obtained for this purpose by
of normal force, which was set at 0.1. The minimization was con- simulating the MSCR test numerically; model parameters charac-
ducted by trial and error. terizing each binder were used in Eqs. (4)–(6), and the resulting
The model parameters characterizing each asphalt binder along differential equations were solved for appropriate loading conditions.
with the error measurements of the fits are presented in Table 2. In The ZSV exhibited by the model equals η þ η̄. This value for
addition, the best-fitting model solutions obtained for the CR-TB each binder was compared to the same experimental measurements
binder along with the experimental data are shown in Fig. 1. It obtained by Gibson et al. (2012). Although more experimental
illustrates the efficacy of the model in describing the response measurements, like the Superpave criterion and Shenoy’s rutting
of the binders. The creep strain from the creep-recovery tests parameter for different binders, were available in Gibson et al.
and the steady-state torque from the steady-shear experiments were (2012), they were not utilized for the verification simply because
described reasonably well by the fits. The model predictions of it is impossible to back-calculate the raw data from such information.
the normal force data from the creep-recovery tests were also The comparison of model predictions of angle of twist per unit
reasonably accurate. However, the model was found to underpre- length in MSCR tests with experimental results of the same is pre-
dict the steady-state normal force in steady-shear tests. sented in Fig. 2. Similarly, model predictions of ZSV of the asphalt
(c) (d)
Fig. 1. Comparison of model predictions with experimental results on CR-TB, the crumb-rubber modified asphalt binder at 64°C: (a) strain
from creep-recovery tests; (b) normal force from creep-recovery tests; (c) steady-state torque from steady-shear tests; (d) steady-state normal force
from steady-shear tests. ∘ ¼ 0∶1 mN∶m; × ¼ 0∶316 mN∶m; ⋄ ¼ 1 mN∶m; þ ¼ 3∶16 mN∶m; □ ¼ 10 mN∶m
(a) (b)
(c)
Fig. 2. Comparison of model predictions with experimental measurements of the nonrecovered compliance at 0.10, 3.20, and 25.6 kPa stress levels
from MSCR tests (data from D’Angelo et al. 2007): (a) J nr at 0.1 kPa; (b) J nr at 3.20 kPa; (c) J nr at 25.6 kPa
repeated creep tests is never greater than the permanent strain in a Here, ηc∞ ðτ Þ is bounded above only by the ZSV if there are no
single creep-recovery test at the same stress level whose creeping restrictions on tc and tr . If the ratio tr =tc is fixed, however, a smaller
time equals Ntc . Thus, the apparent viscosity [η̂ðτ Þ] represents the upper bound can be established. The upper bound for ηc∞ ðτ Þ when
least resistance shown by the model to permanent strain in RCRTs. tr =tc is fixed is the apparent viscosity corresponding to the average
stress over the repeated creep test; i.e., η̂½tc =ðtc þ tr Þτ . This was
again found through numerical simulations. This upper bound
was obtained by observing that the permanent strain in infinitely
long RCRTs at a particular stress level (τ ) is never less than the
permanent strain in a single creep-recovery test at a stress
level equal to the average of the stress over a cycle time
[ðτ tc Þ=ðtc þ tr Þ] applied over the total duration of the repeated
creep test (Fig. 4).
Here, ηc∞ ðτ Þ was found to be reasonably close to the ap-
parent viscosity [η̂ðτ Þ] when the creep time (tc ) is sufficiently
large. When tc is small, ηc∞ ðτ Þ is closer to the upper bound, which
is illustrated in Fig. 5. It should be noted that the lower bound is
independent of tr and tc , but the upper bound depends on tr =tc .
When the stress τ is fixed, the upper bound increases as tr =tc
increases.
The bounds established for the resistance to permanent strain
in RCRTs are both based on apparent viscosity. Because
of this relationship, the apparent viscosity can be used in place of
RCRTs to perform the analysis. Rutting parameters can be com-
Fig. 3. Comparison of model predictions with experimental measure-
pared to the apparent viscosity instead of the permanent strain
ments of ZSV (data from Gibson et al. 2012)
in RCRTs.
Fig. 4. A motivation for the upper bound and lower bound for the con- curve characterizes the permanent strain shown by asphalt binders
vergence limit of average permanent strain in RCRTs as the number of in repeated creep tests. The apparent viscosity at a particular stress
cycles tends toward infinity represents a lower bound to resistance offered by the material to
permanent strain in RCRTs. In what follows, each of the various
rutting prediction criteria is handled separately, and it will be shown
that each rutting parameter is related to the apparent viscosity over a
Analysis
different range of stress.
The raw data of rut depth from the ALF test lanes was statistically
analyzed using a Hotelling-T2 multivariate equivalence test to
determine whether the rutting observed in different test lanes were The Superpave Rutting Parameter—jG j= sin δ
statistically equivalent. In that process, the following ranking The main shortcoming of the Superpave rutting parameter, as
with respect to performance against rutting was established: shown by many previous investigations, is that it overpredicts rut-
SBS 64-40 < Control-PG 70-22 < SBS-LG < Elvaloy < CR-TB. ting of polymer-modified asphalt binders. This is evident in the
This ranking, along with the measured permanent deformation in FHWA ALF study as well. The jG j= sin δ of the control asphalt
the field, are given in Table 3. binder, the unmodified PG 70-22 binder, was higher than that of
The objective in this study is not to determine the rutting param- SBS-LG and Elvaloy, while the rutting observed in the ALF facil-
eter that best correlates with the rutting results of the FHWA field ilty of the two modified binders were discernibly less than that of
study. D’Angelo et al. (2007) has already found that the ranking of the unmodified binder (Fig. 2).
the nonrecovered creep compliance at the stress level of 25,600 Pa In order to analyze the Superpave parameter and determine the
from MSCR tests to match the ranking from the FHWA rutting re- reason for its failure in predicting rutting performance of polymer
sults exactly. Instead, this study is focused on determining the con- modified bitumens, it is necessary to determine if there is any stress
ditions in which a given rutting parameter is useful in predicting the level at which the permanent strain in RCRTs is characterized by
level or rank order of asphalt binders with respect to their rutting this rutting parameter. Alternatively, it can be determined if the
performance. This includes determining the reason for the failure or Superpave rutting parameter can be related to the apparent viscosity
success of a rutting parameter in ranking the binders. For this pur- at some stress level, in view of the relationship between apparent
pose, the nonlinear viscoelastic model developed by Narayan et al. viscosity and permanent strain in RCRTs.
(2013) was used to predict the behavior of asphalt binders in differ- To this end, it is first necessary to understand how the Superpave
ent loading conditions. criterion is obtained experimentally. As mentioned previously,
DSRs measure only the torque and angle of twist per unit length
in the parallel plate geometry. After assuming that the response of
the material is linear, which implies that the shear stress and
shear strain linearly vary in the radial direction, τ R and γ R can
Fig. 5. Typical variation of ηc∞ ðτ Þ with creep time tc , along with the
upper and lower limits for the same. The model parameters used for
simulation correspond to SBS 64–40 at 64°C; tr =tc was fixed at 10,
and RCRTs were simulated for 1,000 cycles to get an approximate
value for ηc∞ ðτ Þ Fig. 6. Apparent viscosity of the asphalt binders as a function of stress
Since, for many polymer-modified binders, the prescribed strain viscosity. However, for this binder, at high temperatures such as
amplitude for measuring the Superpave parameter may not be those considered in this study, η is small enough that η̄ is a good
within the linearity limit; the mechanical behavior of asphalt bind- approximation of the infinite shear viscosity. As shown previously,
ers at the prescribed strain level may be nonlinear. Shear stress may jg j= sin Δ=ω at 10 rad=s is a good measure of η̄; it characterizes
not vary linearly in the radial direction, nor the torque response to a the infinite shear viscosity, at least for the binders tested in
sinusoidally varying angular velocity be sinusoidal. Let jg j and Δ this study.
be defined as Since, for the polymer-modified binders tested in this study, the
infinite shear viscosity represents the resistance to permanent strain
τ̂ R at stress levels of 10 MPa or higher, the equivalence of
jg j :¼ ð15aÞ
γ̂ R jg j= sin Δ=ω and η̄ implies that the parameter jg j= sin Δ ranks
these binders based on their resistance to permanent strain at such
and Δ :¼ ts ω ð15bÞ high stress levels (Fig. 6). It is at these stress levels that the poly-
mer-modified binders SBS-LG and Elvaloy exhibit less resistance
to permanent strain than the Control PG 70-22 binder. However, the
These quantities, jg j and Δ, are the ones that are actually cal- stresses on the binder in the ALF test lanes should be much lower,
culated in a DSR. Note that jg j and Δ defined this way are not the considering that the effective contact stress between the loading
same as jG j and δ defined in the linearized viscoelastic theory; this equipment and the test lanes is only about 1 MPa. At such stress
is because of the nonlinearity of the response. It is important to levels, the resistance shown by the two polymer-modified binders is
make this distinction between the two sets of quantities. greater than that of the unmodified binder, as seen in Fig. 6.
For the model, when α is zero, the parameter jG j= sin δ can be This concludes the explanation for the discrepancy between the
shown to be equal to actual ALF rutting results and the prediction with the Superpave
parameter, jG j= sin δ. First, the parameter being measured for pol-
jG j η̄ω ymer-modified bitumen is a crudely defined jg j= sin Δ, not
¼ ð16Þ
sin δ ð β−1 þ 1Þ jG j= sin δ from linear viscoelasticity. Second, the parameter
1þλ2 β 2 ω2
appears to underestimate the performance of modified binders
where ω is the angular frequency of oscillation and β ¼ η̄=η þ η̄. because it ranks them based on the resistance that they offer to
This relation, for reasonably large ω, can be approximated as permanent strain at relatively high stress levels—stress levels that
are not expected to be experienced by asphalt binders in pavement.
jG j
¼ η̄ω ð17Þ One can also speculate why the parameter has been working well
sin δ for unmodified binders. Most unmodified binders do not show
For cases where α and n are nonzero, the parameter jG j= sin δ much shear thinning, and their infinite shear viscosity is generally
no longer makes sense as the response in oscillatory tests is non- close to the viscosity at other stress levels. Thus, the Superpave
linear. For these cases, the parameter jg j= sin δ, as defined in parameter is as good as any other rutting parameter for these
Eq. (15), was calculated by simulating the oscillatory experiment binders.
conducted in the DSR, and it was found that for most parameter
combinations, jg j= sin Δ tends to η̄ω when ω is sufficiently large. Zero-Shear Viscosity
Thus, it can be stated with reasonable confidence that jg j= sin δ=ω
for large ω is a good measure of the model parameter η̄ whether α In the FHWA study, the ZSV was not able to accurately rank the
equals zero or not. asphalt binders based on their resistance to rutting. The main differ-
For the binders tested in this study, the quantity λβ is large ence between the two rankings was the SBS 64-40 binder. While
enough for the relation in Eq. (17) to hold true when the angular the rutting resistance of this binder was found to be the lowest in the
frequency (ω) equals 10 rad=s [from Eq. (16)]. A comparison of the ALF rutting tests, the ZSV ranks it near the highest among the
experimentally observed jg j= sin Δ with the model prediction of tested binders.
the same and the quantity η̄ω is presented in Table 4. It can be seen The reason for this difference appears to be straightforward. In
from this comparison that the experimentally observed value for Fig. 6, it can be observed that the apparent viscosity of SBS 64-40
each binder correlates well with the model prediction, which in turn at stress levels less than ∼1,000 Pa is amongst the highest. But
equals η̄ω. above 1,000 Pa, the apparent viscosity rapidly decreases and
The model parameter η̄ is the infinite shear viscosity of the becomes the smallest among all asphalt binders for stress levels
model when α and n are nonzero (i.e., when the model shows shear higher than ∼20 kPa. The single tire load on the test lanes in this
thinning). For binders that do not show shear thinning, like the FHWA study is relatively high (contact stress of 1 MPa) and the
unmodified binder tested in this study, η̄ þ η is the infinite shear speed of the loading is relatively low (18 km=h). For this type
recovery time is reasonably high, the resistance that the material bounds for nonrecovered creep compliance. With the help of these
exhibits against permanent strain can be significantly higher than bounds, the nonrecovered creep compliance for any creep time and
that represented by the apparent viscosity curve (Fig. 5). Under recovery time can be estimated if the nonrecovered creep compli-
such conditions, the ZSV would be representative of the resistance ance for 1-s creep time and 9-s recovery time is known. In this way,
shown by the binders, even at high stress levels. For example, when the current MSCR test with 1-s creeping time and 9-s recovery time
the creeping time is 0.01 s and the recovery time is 10 s, ZSV char- can be used to simulate any traffic loading.
acterizes resistance to permanent strain at stress levels as high as
100 kPa.
Effective Stress Range of Rutting Parameters
In the previous sections, each rutting parameter was attributed a
Multiple Stress Creep Recovery Test
stress range—i.e., a range of stress levels at which, if a RCRT
The MSCR test has already been shown to predict the rutting re- is conducted, the resistance to permanent strain observed in the test
sults of the ALF study by D’Angelo et al. (2007). The nonrecov- will be accurately characterized by the corresponding rutting
ered creep compliance at 25.6 kPa was shown to rank the asphalt parameter. The ranking of binders based on a rutting parameter
binders as observed in the rutting tests. In this study, the model would match the ranking of binders based on RCRTs that are con-
prediction of the nonrecovered creep compliance of each asphalt ducted at stress levels within the corresponding range. For instance,
binder at 25.6 kPa was found to agree with the experimental mea- if one needs to know the ranking of asphalt binders with respect to
surements of the same [Fig. 2(c)], which in turn was found to rank the resistance to permanent strain in RCRTs at a stress level less
the asphalt binders accurately. than about 300 Pa, one can perform the ZSV test on the binders
The range of stress levels at which the MSCR characterizes the and find the ranking based on ZSV. Similarly, if the ranking with
resistance to permanent strain in RCRTs is essentially the range of respect to resistance to permanent strain at stresses greater than
stress in which the nonrecoverable creep compliance is measured, about 7 MPa is needed, one can measure G = sin δ and rank the
namely stress levels ranging from 50 Pa to 25.6 kPa. Moreover, binders accordingly. Considering the relationship to permanent
because of the nature of the test, this range can be extended on strain in RCRTs and the apparent viscosity at any given stress level,
either side, provided that the same experiment procedure can be each stress range can alternatively be viewed as the range of stress
used at higher or lower stress levels. However, it must be ques- levels at which the ranking of the apparent viscosity of the binders
tioned whether the binders must be tested at multiple stress levels matches the ranking based on the corresponding rutting parameter.
if the unnonrecovered creep compliance at only one particular These stress ranges of rutting parameters are presented in Fig. 7.
Fig. 7. The effective stress range of rutting parameters marked on the apparent viscosity versus stress plots
Conclusions
Fig. 8. Comparison of rutting measured in ALF tests to the viscosity
predicted for each binder at a stress level of 200 kPa Prediction criteria that have been proposed with regard to rutting
susceptibility were analyzed with the help of a nonlinear viscoelas-
tic model. The results of this analysis essentially show that each
rutting parameter ranks asphalt binders based on their resistance
It should be noted that the stress ranges established for the ZSV
to permanent strain over a different range of stress levels. While
and the Superpave criterion in this study are not concrete, as they
there appears to be an equivalent relation between these rutting
are both based on binders tested in this study. If it is discovered that
parameters for the unmodified binder tested in this study, each
for some asphalt binder, the apparent viscosity changes greatly
rutting parameter represents a markedly different level of resistance
even at stress levels less than 300 Pa, then the higher end of the
to permanent strain for the modified binders. The Superpave cri-
stress range corresponding to the ZSV should be reduced accord-
terion was found to be closely related to the resistance to permanent
ingly. Similarly, if for an asphalt binder, the apparent viscosity
strain offered by asphalt binders at relatively high stress levels,
changes greatly, even at stress levels greater than 7 MPa, the lower
while the ZSV was found to be related to the resistance at relatively
end of the stress range coresponding to the Superpave criterion
low stress levels. The intermediate range of stress levels was found
must be increased accordingly.
to be covered by MSCR tests, with the nonunrecovered compliance
Also shown in Fig. 7 is the range of stress levels at which the
as the rutting parameter.
ranking of the apparent viscosity exactly matches the ranking of the
The apparent viscosity at a chosen stress level is suggested as a
binders with respect to the rutting behavior observed in the ALF
new parameter to predict the rutting susceptibility of asphalt bind-
study. This is labeled the ALF range (which may or may not be
related to the actual stresses experienced by the binder in the ers. A relationship between the viscosity and permanent strain from
ALF pavement). That is, if one were to conduct RCRTs at some repeated creep tests already has been shown to exist at low stress
stress level within this range, one can accurately rank the rutting levels for several asphalt binders in previous research (e.g., Merusi
behavior in the ALF facility. It can be seen that the ALF range and Giuliani 2011). This relationship was partially extended to
is overlapped only by the higher end of the stress range correspond- higher stress levels; bounds were established for permanent strain
ing to the MSCR test. Barring that, the ALF range is not covered by in RCRTs using the viscosity determined at different stress levels.
the stress range of any rutting parameter. Moreover, it was also shown that the viscosity at a judiciously
chosen stress level can accurately characterize the rutting behavior
observed in a field study. Considering that experiments for deter-
Viscosity as an Alternative Rutting Parameter mining viscosity are much more straightforward and do not involve
Considering the relation between the viscosity and permanent strain issues with delayed elasticity, as is the case with MSCR tests, it is a
established in this study, an alternative rutting prediction parameter good alternative to these tests for determining rutting susceptibility.
to the nonrecovered compliance from MSCR testing could be the There are, however, certain issues regarding experimental meas-
viscosity at a given stress level. It can be seen from Fig. 7 that the urement that need to be addressed. Measuring viscosity at relatively
viscosity predicted by the model for the stress levels in the ALF high stress levels can be difficult because most viscometric flows
range accurately ranks the asphalt binders with respect to their are susceptible to several flow instabilities; flow in parallel-plate
performance against rutting. In particular, a good correlation geometry, in particular, is susceptible to edge fracture and inertial
was observed between the rutting measured in the field study instabilities. While conducting steady-shear experiments on asphalt
and the viscosity predicted at a stress level of 200 kPa (Fig. 8). binders at 64°C, edge fracture was observed at shear rates greater
The R2 goodness of the fit was 0.9961. than 3−1 for most of the asphalt binders. These instabilities can be
Using viscosity as a rutting prediciton parameter is not entirely avoided by decreasing the separation between the plates in the par-
new. In NCHRP Report 459, Bahia et al. (2001a) suggested using a allel plate geometry or by using a cone-and-plate geometry with a
viscosity parameter, determined from RCRT test results at 100 Pa low cone-angle. This is discussed in more detail by Motamed and
and 3,200 Pa, as a rutting parameter. They also used it to predict the Bahia (2011).
rutting performance of asphalt binders with a reasonable level of
success. ZSV is also a viscosity parameter that has been shown
to predict the rutting performance of binders reasonably. In this Acknowledgments
study, the usage of viscosity as a rutting parameter is merely
extended to a wider range of stress levels. The authors thank Dr. Nelson Gibson and the FHWA for materials
In order to implement this as a full specification, it is first nec- and information, and Dr. Eyad Masad for his comments on a
essary to determine the critical stress level to measure the viscosity. previous draft.
ments for use in dynamic testing.” Highway Res. Rec., 345, 32–44. the context of a Gibbs-potential based thermodynamic framework.” J.
Bouldin, M., Dongré, R., and D’Angelo, J. (2001). “Proposed refinement of Eng. Mech., 10.1061/(ASCE)EM.1943-7889.0000682, 04014116.
Superpave high-temperature specification parameter for performance- Phillips, M., and Robertus, C. (1996). “Binder rheology and asphaltic
graded binders.” Transp. Res. Rec.: J. Transp. Res. Board, 1766, 40–47. pavement permanent deformation: The zero-shear-viscosity.” Eura-
D’Angelo, J., Kluttz, R., Dongre, R., Stephens, K., and Zanzotto, L. (2007). sphalt Eurobitume Cong., European Asphalt Pavement Association,
“Revision of the Superpave high-temperature binder specification: The
Brussels, Belgium.
multiple stress creep recovery test.” J. Assoc. Asphalt Paving Technol.,
Rowe, G., D’Angelo, J., and Sharrock, M. (2002). “Use of zero shear
76, 123–162.
D’Angelo, J. A. (2009). “The relationship of the MSCR test to rutting.” viscosity as a parameter for the high-temperature binder specification
Road Mater. Pavement Des., 10(1), 61–80. parameter.” J. Appl. Asphalt Binder Technol., 2(2).
Delgadillo, R., Cho, D., and Bahia, H. (2006). “Nonlinearity of repeated Shenoy, A. (2001). “Refinement of the Superpave specification parameter
creep and recovery binder test and relationship with mixture permanent for performance grading of asphalt.” J. Transp. Eng., 10.1061/(ASCE)
deformation.” Transp. Res. Rec.: J. Transp. Res. Board, 1962, 3–11. 0733-947X(2001)127:5(357), 357–362.
Dongré, R., Button, J., Kluttz, R., and Anderson, D. (1997). “Evaluation of Shenoy, A. (2004a). “A comprehensive treatise of the high-temperature
Superpave binder specification with performance of polymer-modified specification parameter—G /(1-(1/tan d sin d)) for performance
asphalt pavements.” ASTM Special Technical Publication, Vol. 1322, grading of asphalts.” Appl. Rheol., 14(6), 303–314.
American Technical Publisher, Orland Park, IL, 80–100. Shenoy, A. (2004b). “High temperature performance grading of asphalts
Dongré, R., and D’Angelo, J. (2003). “Refinement of Superpave high- through a specification criterion that could capture field performance.”
temperature binder specification based on pavement performance in
J. Transp. Eng., 130(1), 132–137.
the accelerated loading facility.” Transp. Res. Rec.: J. Transp. Res.
Stuart, K., Mogawer, W., and Romero, P. (1999). “Validation of asphalt
Board, 1829, 39–46.
Gibson, N., et al. (2012). “Full-scale accelerated performance testing for binder and mixture tests that measure rutting susceptibility using the
superpave and structural validation.” Rep. No. FHWA-HRT-11-045, accelerated loading facility.” Rep. No. FHWA-RD-99-204, Federal
Federal Highway Administration, Washington, DC. Highway Administration, Washington, DC.
Golalipour, A. (2011). “Modification of multiple stress creep and recovery Sybilski, D. (1996). “Zero-shear viscosity of bituminous binder and its
test procedure and usage in specification.” M.S. thesis, Univ. of relation to bituminous mixture’s rutting resistance.” Transp. Res. Rec.:
Wisconsin, Madison, WI. J. Transp. Res. Board, 1535, 15–21.