Comparing Resilient Modulus and Dynamic

Modulus of Hot-Mix Asphalt as Material

Properties for Flexible Pavement Design
Amara Loulizi, Gerardo W. Flintsch, Imad L. Al-Qadi, and David Mokarem

With the current trend toward developing mechanistic flexible pave- direction. Usually the loading consists of a haversine pulse with
ment design and the need for more reliable design procedures, accurate duration of 0.1 s followed by a rest period of 0.9 s. However, a pulse
characterization of hot-mix asphalt (HMA) properties is needed. Resilient duration of 0.03 s followed by a rest period of 0.97 s was used in this
and dynamic modulus tests were performed at five temperatures on two study to simulate the pulse load induced from moving trucks and
typical mixes used in the Commonwealth of Virginia to compare the test from falling-weight deflectometer testing (3). The vertical and hor-
results. The dynamic modulus was measured at six frequencies at each izontal deformations at the center of the specimen that result from
of the testing temperatures, and the resilient modulus test was per- the applied pulse load are recorded as the output of the test.
formed at one loading time. The study found that the size of the speci- In this study, to obtain the resilient modulus value from the mea-
men statistically affected the measured resilient modulus value. Resilient sured vertical and horizontal deformations, the Roque and Buttlar
modulus values obtained in the 100-mm-diameter specimens were higher (4) procedure was used. This procedure corrects for the effect of
than those obtained in the 150-mm-diameter specimens at all testing specimen bulging, which causes the externally mounted exten-
temperatures. No statistical differences were observed in the resilient someters to rotate and results in errors in the vertical and horizontal
modulus of the two mixes. However, statistical differences were found deflection readings. With this procedure, the resilient modulus (Mr)
in the dynamic modulus of the two mixes. A strong relation between the is computed as follows:
dynamic modulus test performed at 5 Hz and the resilient modulus was
found. Three different pavement structures were analyzed to estimate (σ − νσ ycorr )
Mr =
xcorr
the variation of their bottom-up fatigue life when different moduli were (1)
⑀ xcorr
used for the HMA layer. It was found that the measured dynamic mod-
uli resulted in the highest fatigue life estimates for the three considered where
pavements. It is concluded that the dynamic modulus test provides a
better characterization of HMA than the resilient modulus test because σxcorr = corrected horizontal point stress,
it provides full characterization of the mix over temperature and loading σycorr = corrected vertical point stress,
frequencies. ν = Poisson’s ratio, and
⑀xcorr = corrected horizontal strain.
The dynamic modulus test, known also as the complex modulus
Although several testing procedures are accepted to characterize
test, is performed by the application of sinusoidal vertical loads to
hot-mix asphalt (HMA), only resilient modulus test results are incor-
cylindrical specimens and the measurement of the corresponding
porated in the current accepted AASHTO pavement design guide- vertical deformation (Figure 1b). The test is usually performed at
lines (1). However, in the mechanistic–empirical (M-E) design different temperatures and at different frequencies. The applied
guide proposed by the recently completed NCHRP Project 1-37A stress and corresponding measured strain are represented as follows:
and currently under evaluation, this test has been replaced by the
σ = σ 0 sin ( ωt )
complex dynamic modulus test to characterize HMA (2).
(2)
The HMA resilient modulus test, ASTM D 4123, is relatively
simple and can be conducted on field cores. The test is performed
by loading a cylindrical specimen along and parallel to its vertical ⑀ = ⑀ 0 sin ( ωt − φ ) (3)
diametral plane (Figure 1a). This loading configuration develops a
relatively uniform state of tensile stresses perpendicular to the load where
σ0 = applied stress amplitude,
⑀0 = measured strain amplitude,
A. Loulizi and G. W. Flintsch, Virginia Tech Transportation Institute, 3500 Trans-
portation Research Plaza, Blacksburg, VA 24061. I. L. Al-Qadi, Illinois Center for ω = 2πf = angular frequency,
Transportation, Civil and Environmental Engineering Department, University of t = time,
Illinois, Urbana–Champaign, 1611 Titan Drive, Rantoul, IL 61866. D. Mokarem, f = 1/T = frequency,
Virginia Transportation Research Council, Virginia Department of Transportation, T = period, and
530 Edgemont Road, Charlottesville, VA 22903-2454.
φ = phase angle, computed as shown by Equation 4.
Transportation Research Record: Journal of the Transportation Research Board,
Δt
No. 1970, Transportation Research Board of the National Academies, Washington, φ= 360⬚ (4)
D.C., 2006, pp. 161–170. T

161
162 Transportation Research Record 1970

(a) (b)

FIGURE 1 Testing setups for (a) resilient modulus test and (b) dynamic modulus test.

where Δt is the time lag between the applied stress and the corre- 75 mm (3 in.). Two specimen sizes were used for the SM-9.5A mix:
sponding strain. 150 mm (6 in.) in diameter by 115 mm (4.5 in.) in thickness and
The dynamic modulus is calculated by using Equation 5. The in- cut to a final thickness of 75 mm (3 in.) and 100 mm (4 in.) in diam-
phase and out-of-phase components are obtained with Equation 6 eter by 125 mm (5 in.) in thickness and cut to a final thickness of
and Equation 7, respectively. 50 mm (2 in.). All specimens were cut from the middle of the
compacted specimens.
σ0 • Dynamic modulus specimens. Specimens 150 mm (6 in.) diam-
E* = (5)
⑀0 eter by 178 mm (7 in.) thick were prepared for both mixes, which
were later cored and cut to a final size of 100 mm (4 in.) in diameter
E ′ = E* cos ( φ ) (6) by 150 mm (6 in.) in thickness.

The gradation of the batched aggregate was checked by using

E ′′ = E ′ = E* sin ( φ ) (7) representative samples from each aggregate mixed according to
the JMF. For each aggregate batch, three sieve analysis tests were
Given that the two test methods have been widely used worldwide performed; the gradation curves are shown in Figure 2.
and may be used interchangeably, there is a need to compare the The volumetric properties for both mixes were obtained. For the
results obtained from both tests. Therefore, this study was initiated SM-9.5A mix, the average air voids of the prepared specimens were
to compare the resilient modulus and dynamic modulus tests on two 3.6% for the resilient modulus test and 4.2% for the dynamic mod-
typical mixes used in the Commonwealth of Virginia. ulus test. For the BM-25.0 mix, the average air voids of the prepared
specimens were 6.2% for the resilient modulus test and 5.0% for the
dynamic modulus test. For both types of tests, five testing tempera-
EXPERIMENTAL PROGRAM tures were selected: −15°C, 5°C, 20°C, 30°C, and 40°C. Two spec-
imens were tested at each temperature. In addition, six frequencies
The two mixes used in this study—a surface mix, SM-9.5A, and a were used at each temperature for the dynamic modulus test: 0.1,
base mix, BM-25.0—were designed in accordance with Superpave® 0.5, 1, 5, 10, and 25 Hz.
specifications. All specimens were prepared according to the job
mix formulas (JMFs) provided by the Virginia Department of Trans-
portation (VDOT). All specimens were prepared with a Troxler RESULTS
gyratory compactor. The specimen sizes were as follows:
Resilient Modulus
• Resilient modulus specimens. All BM-25.0 specimens were
compacted 150 mm (6 in.) in diameter by approximately 115 mm Table 1 shows the results for all the tested specimens for both mixes.
(4.5 in.) in thickness and were then cut to a final thickness of During testing, horizontal and vertical deformations were measured
Loulizi, Flintsch, Al-Qadi, and Mokarem 163

100 TABLE 1 Resilient Modulus Results

90
80 Mr Average
Percent Passing

70
ID Temp. (°C) (MPa) (ksi) (MPa) (ksi)
60
Gradation 1
50 SM-9.5A
Gradation 2
40 100-mm specimens (gauge length = 25.4 mm)
Gradation 3
30
Average S45 −15 14,541 2,109 16,572 2,404
20
S52 −15 18,602 2,698
10
0 S46 5 11,969 1,736 12,655 1,836
0.075 0.3 0.6 1.18 2.38 4.75 9.5 12.5 S53 5 13,341 1,935
Sieve Size Raised to 0.45 Power S47 20 8,336 1,209 7,829 1,136
(a) S54 20 7,322 1,062
S48 30 5,674 823 5,099 740
100 S55 30 4,523 656
90 S49 40 2,606 378 2,644 384
80 S56 40 2,682 389
Percent Passing

70
150-mm specimens (gauge length = 38.1 mm)
60
50 Gradation 1 S7 −15 17,864 2,591 16,292 2,363
40 Gradation 2 S10 −15 14,720 2,135
30 Gradation 3 S8 5 10,949 1,588 11,669 1,693
20 Average S13 5 12,390 1,797
10
S9 20 5,688 825 6,660 966
0
0.075 0.6 1.18 2.36 4.75 9.5 12.5 19 25 37.5 S19 20 7,632 1,107
Sieve Size Raised to 0.45 Power S11 30 4,447 645 4,144 601
(b) S14 30 3,840 557
S12 40 2,220 322 2,096 304
FIGURE 2 Aggregate gradation for (a) SM-9.5A mix and S18 40 1,972 286
(b) BM-25.0 mix.
BM-25.0
B7 −15 15,182 2,202 17,047 2,473
from both sides of the specimen and hence the resilient modulus was B10 −15 18,912 2,743
calculated accordingly. B8 5 11,852 1,719 11,187 1,623
The variability between the two sides was found to be high for B13 5 10,521 1,526
most of the tested specimens. This variability has been attributed to B9 20 8,556 1,241 7,564 1,097
imperfect centering of the specimen, different aggregate orientation B16 20 6,571 953
from one side to the other, or the air distribution throughout the B11 30 4,164 604 4,302 624
specimen depth, or all three factors. The results shown in Table 1 B14 30 4,440 644
represent the average values of the two sides. Averaging the two B15 40 2,565 372 2,275 330
sides gave reasonable results for the tested temperatures. As would B18 40 1,986 288
be expected, the resilient modulus decreases with an increase in
temperature.
Figure 3 shows a comparison of the resilient modulus values mea-
sured for the two specimen sizes for the SM-9.5A mix; the resilient domized design with effect type (SM-9.5A or BM-25.0), size within
moduli obtained for the 100-mm specimens were always higher type (100 mm and 150 mm), and temperature as covariates. It was
than those obtained for the 150-mm specimens. As the temperature found that the temperature and the specimen size were significant
increases, the difference between the two sizes becomes more for the resilient modulus (p-values < 0.05). However, the mix type
notable. On average, the ratios of the resilient modulus obtained for was found not to have a significant effect ( p-value of 0.637 > 0.05)
the 100-mm specimens to those obtained for the 150-mm specimens even though there are major differences between the two mixes,
were 1.02, 1.08, 1.18, 1.23, and 1.26 at −15°C, 5°C, 20°C, 30°C, and especially in the void content (2.6% difference in the void content
40°C, respectively. between the two mixes).
The resilient moduli of the 150-mm specimens of the SM-9.5A
mix and the BM-25.0 mix are compared in Figure 4. It can be
observed that the variation within the same mix specimens (at the Dynamic Modulus
same temperature) is high. Furthermore, no clear trend in the rela-
tionship between the two mixes was noted. A statistical analysis was Figure 5 shows the measured dynamic modulus results for the SM-
conducted to evaluate the effect of the mix type and specimen size. 9.5A and BM-25.0 mixes as a function of frequency for each test-
The SAS software was used to analyze the data as a completely ran- ing temperature. As expected, under a constant loading frequency,
164 Transportation Research Record 1970

18000
16000 150 mm
100 mm

Resilient Modulus (MPa)

14000
12000
10000
8000
6000
4000
2000
0
-15 5 20 30 40
Temperature (°C)

FIGURE 3 Comparison between resilient modulus of 100-mm and 150-mm

specimens.

the magnitude of the dynamic modulus decreases with an increase overwhelms the response of the specimen at high temperatures and
in temperature; under a constant testing temperature, the magnitude low frequencies (5).
of the dynamic modulus increases with an increase in the frequency. A master curve of the dynamic modulus at the reference temper-
Figure 6 shows the calculated phase angle results for both mixes. It ature of 20°C was constructed to complete the characterization of
can be seen that the phase angle for both mixes decreases as the fre- the material. The method developed by Pellinen and Witczak (6)
quency increases at testing temperatures of −15°C, 5°C, and 20°C. was used in this study to construct the master curve. The method
However, at testing temperatures of 30°C and 40°C, the behavior of consists of fitting a sigmoidal curve to the measured dynamic mod-
the phase angle as a function of frequency is more complex. ulus test data using nonlinear least-squares regression techniques.
At 30°C, the phase angle seems to increase up to frequencies of The shift factors at each temperature are determined simultaneously
0.5 Hz, and then it starts to decrease with an increase in frequency. with the other coefficients of the sigmoidal function. The function
At 40°C, the behavior of the phase angle with frequency is even is given by
more complex. At this temperature, the phase angle seems to
increase from 0.1 Hz to 0.5 Hz, starts to decrease from 0.5 Hz to 1 Hz, α
log E * = δ + (8)
and then starts to increase again with frequency. At constant high 1 + eβ− γ log fr
frequencies (1 Hz to 25 Hz), the phase angle increases with an
increase in temperature; at lower frequencies, the behavior of the where δ, α, β, and γ are sigmoidal function coefficients (fit parameters),
phase angle with temperature is more complex. The complex behav- and fr is the reduced frequency, which is given by
ior of the phase angle at higher temperatures or lower frequencies
could be attributed to the increasing effect of the aggregate inter- log fr = log f + log aT (9)
lock. This finding is in agreement with the findings from other
researchers, who reported that the elastic behavior of the aggregate where aT is the shift factor at temperature T.

20000 BM-25.0
18000 SM-9.5A
16000
Resilient Modulus (MPa)

14000
12000
10000
8000
6000
4000
2000
0
-15 5 20 30 40
Temperature (°C)

FIGURE 4 Resilient modulus comparison between SM-9.5A and BM-25.0 mixes.

Loulizi, Flintsch, Al-Qadi, and Mokarem 165

100000 -15°C 5°C 20°C 30°C 40°C

10000

|E*| (MPa) 1000

100
0.01 0.1 1 10 100
Frequency (Hz)
(a)

100000 -15°C 5°C 20°C 30°C 40°C

10000
|E*| (MPa)

1000

100
0.01 0.1 1 10 100
Frequency (Hz)
(b)

FIGURE 5 Dynamic modulus results for (a) SM-9.5A mix and (b) BM-25.0 mix.

The SAS statistical software package was used for the nonlin- .log E * = 3.750063 + 0.02932ρ − 0.001767 (ρ )2
200 200
ear regression analysis; the determined coefficients for both mixes
⎛ Vbeff ⎞
are shown in Table 2. The best-fit master curves for both mixes are − 0.058097Va − 0.802208 ⎜
presented in Figure 7. ⎝ Vbeff + Va ⎠⎟
In order to compare the dynamic moduli of the two mixes, a plot [ + 3.871977 − 0.0021ρ + 0.003958ρ 4 38
of the dynamic modulus ratio of the BM-25.0 mix to that of the
SM-9.5A mix was computed (Figure 8). The plot shows that BM-25.0 − 0.000017 (ρ ) + 0.005470ρ ⎤⎦ ÷
38
2
34

has a higher dynamic modulus than that of SM-9.5A at all tested fre- −0.603313− 0.313351 log( f ) − 0.393532 log( η )
1+ e (10)
quencies and temperatures. To verify the statistical significance of
the obtained results, SAS was used to model the natural logarithm
where
of the dynamic modulus data as a dependent variable with effect
type (SM-9.5A or BM-25.0), natural logarithm of the frequency, E* = dynamic modulus (psi),
and temperatures as covariates. Results of the analysis showed that ρ200 = percent passing No. 200 sieve,
all considered effects were statistically significant; calculated p-values ρ4 = cumulative percent retained on No. 4 sieve,
are smaller than 0.05. This finding suggests that the dynamic mod- ρ34 = cumulative percent retained on No. 3⁄4 sieve,
ulus for both mixes is different, as would be expected. Although the ρ38 = cumulative percent retained on No. 3⁄8 sieve,
f = frequency (Hz),
BM-25.0 mix has higher air voids, the larger aggregate content and
Vbeff = effective bitumen content (% by volume),
lower asphalt content increased the aggregate interlock.
Va = air void content, and
Once the dynamic modulus master curve was established for
η = bitumen viscosity (106 poise).
both mixes on the basis of the measured values, the Witczak pre-
diction equation was used to generate the dynamic modulus master The bitumen viscosity varies with temperature according to
curves for both mixes (2). The Witczak prediction equation is as
follows: log [ log ( η)] = A + VTS log ( TR ) (11)
166 Transportation Research Record 1970

-15°C 5°C 20°C 30°C 40°C TABLE 2 Parameters for the Dynamic Modulus Master Curve
40
35 Parameter Value Parameter Value
30
SM-9.5A
Phase angle (°)

25
δ 1.87615 log(a−15) 4.70518
20
α 2.41534 log(a5) 1.21741
15
β −1.28301 log(a20) 0
10
γ 0.59499 log(a30) −1.15024
5
log(a40) −2.26248
0
BM-25.0
0.01 0.1 1 10 100
Frequency (Hz) δ 2.1358 log(a−15) 4.38421
(a) α 2.26117 log(a5) 1.60879
β −1.11630 log(a20) 0
-15°C 5°C 20°C 30°C 40°C γ 0.62793 log(a30) −1.15168
40.0
log(a40) −2.05835
35.0
30.0
Phase angle (°)

25.0 methods consist in finding A and VTS based on calculation of vis-

20.0 cosities at different temperatures with the measured complex shear
15.0 modulus on the original (unaged) binder and the rolling thin-film
10.0 oven (RTFO)–aged binder, respectively. At an angular frequency of
10 rad/s, the viscosity, η, is related to the complex shear modulus,
5.0
G*, and phase angle, δ, as follows (2):
0.0
0.01 0.1 1 10 100
4.8628
Frequency (Hz) G* ⎛ 1 ⎞
η= (12)
(b) 10 ⎜⎝ sin δ ⎟⎠
FIGURE 6 Phase angle results for (a) SM-9.5A mix and
(b) BM-25.0 mix. For the original binder, the calculated A- and VTS-values were
11.66 and −3.94, respectively, and for the RTFO-aged binder 1.89
and −4.01, respectively. Figure 9 shows the dynamic modulus master
where η is the binder viscosity expressed in centipoise, TR is the curves for both mixes obtained with Equation 10 and the three dif-
temperature in degrees Rankine, and A and VTS are regression pa- ferent values for A and VTS. The graphs also show the master curves
rameters. generated for the measured data for comparison purposes. From Fig-
For this study, three different ways were used to find A and VTS. ure 9, it is noted that although the predicted dynamic modulus val-
The first is to use the default values as suggested by the proposed ues are on the same order of magnitude as the measured ones,
AASHTO M-E design guide for a PG 64-22 binder. These default significant differences exist, especially in the frequency range of
values are 10.98 for A and −3.68 for VTS. The second and third 0.01 Hz to 1,000 Hz. In this frequency range, the predicted values

Measured SM-9.5A Fit SM-9.5A

100000
Measured BM-25.0 Fit BM-25.0

10000
|E*| (MPa)

1000

100
1.E-04 1.E-03 1.E-02 1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07
Reduced frequency (Hz)

FIGURE 7 Developed dynamic modulus master curves for SM-9.5A mix and BM-25.0 mix.
Loulizi, Flintsch, Al-Qadi, and Mokarem 167

1.6 -15°C 5°C 20°C 30°C 40°C

1.4

|E*|BM25.5 / |E*|SM-9.5A
1.2
1
0.8
0.6
0.4
0.2
0
0.1 0.5 1 5 10 25
Frequency (Hz)

FIGURE 8 Ratio between BM-25.0 mix dynamic modulus and that of SM-9.5A mix.

100000

10000
|E*| (MPa)

Measured/fit
1000 Witczak/default
Witczak/original VTS
Witczak/RTFO VTS

100
1.E-04 1.E-03 1.E-02 1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07
Reduced frequency (Hz)
(a)

100000

10000
|E*| (MPa)

Measured/fit
1000 Witczak/default
Witczak/original VTS
Witczak/RTFO VTS

100
1.E-04 1.E-03 1.E-02 1.E-01 1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07
Reduced frequency (Hz)
(b)

FIGURE 9 Dynamic modulus predicted with Witczak equation: (a) SM-9.5A mix and (b) BM-25.0 mix.
168 Transportation Research Record 1970

underestimate the measured dynamic modulus. The predicted almost 1.4 times the resilient modulus. The average air void contents
dynamic modulus with the default A and VTS parameters and those for the SM-9.5A specimens prepared for both test types were close
obtained with the RTFO-aged binder were the closest to the measured (3.6% for the resilient modulus and 4.2% for the dynamic modulus),
values. whereas for the BM-25.0 mix, the average air void contents were
different for both tests (6.2% for the resilient modulus test and 5.0%
for the dynamic modulus test). The high air voids in the resilient
Comparison Between Dynamic Modulus modulus test specimens of the BM-25.0 mix could have significantly
and Resilient Modulus reduced the measured resilient modulus value.

To compare the dynamic modulus and the resilient modulus data,

the time of loading should be equivalent in both tests. Since in this Effect of Modulus Test Type Selection
study the loading time used in performing the resilient modulus test on HMA Fatigue Life
was 0.03 s, the equivalent angular frequency would be 33.3 rad/s
(ω = 1/t, where ω is the angular frequency and t is the loading time). To study the effect of choosing the HMA modulus value on its
The angular frequency is then converted to frequency by division of fatigue life, three different pavement cross sections were investi-
2π. This computation results in a frequency of 5.2 Hz. Hence, the gated (Pavements A, B, and C). All pavements consist of a subgrade
comparison was conducted between the dynamic modulus at 5 Hz with a resilient modulus of 69 MPa (10 ksi), 152 mm (6 in.) of gran-
and resilient modulus tests. ular base with a resilient modulus of 276 MPa (40 ksi) and 51 mm
Figure 10 shows a plot of the average dynamic modulus results (2 in.) of SM-9.5A HMA surface mix. However, Pavement A has a
versus the average resilient modulus per mix. It is clear that there is 76-mm (3-in.) BM-25.0 HMA base mix, Pavement B has a 152-mm
a strong correlation between the dynamic modulus and the resilient (6-in.) BM-25.0 HMA base mix, and Pavement C has a 229-mm
modulus and that the correlation is mix dependent. For the SM-9.5A (9-in.) BM-25.0 HMA base mix.
mix, the dynamic modulus is almost equal to 1.07 times the resilient Figure 11 shows the three different pavement designs. The load
modulus. For the BM-25.0 mix, the dynamic modulus is equal to considered in this study consists of a set of dual tires carrying a total

25000
Line of equality
20000
y = 1.07x
R2 = 0.97
|E*| (MPa)

15000

10000

5000

0
0 5000 10000 15000 20000 25000
MR (MPa)
(a)

25000
Line of equality

20000
y = 1.4x
R2 = 0.99
|E*| (MPa)

15000

10000

5000

0
0 5000 10000 15000 20000 25000
MR (MPa)
(b)

FIGURE 10 Comparison between dynamic modulus and resilient modulus:

(a) SM-9.5A mix and (b) BM-25.0 mix.
Loulizi, Flintsch, Al-Qadi, and Mokarem 169

40 kN 40 kN 40 kN load was determined for each pavement structure by using three dif-
724 kPa 724 kPa 724 kPa ferent inputs for the HMA material properties at five different tem-
330 mm 330 mm 330 mm peratures (−15°C, 5°C, 20°C, 30°C, and 40°C). The three different
considered modulus selections were based on the measured resilient
modulus for both HMA mixes, the measured dynamic modulus for
51 mm SM-9.5A 51 mm SM-9.5A 51 mm SM-9.5A both HMA mixes, and the predicted dynamic modulus for both
76 mm BM-25.0 mixes with Equation 10 and the default A- and VTS-values.
152 mm BM-25.0
229 mm BM-25.0 Figure 12 shows the allowable number of repetitions of the
152 mm Aggregate standard load at the five different temperatures for the three stud-
152 mm Aggregate ied pavements with the three different inputs for the HMA prop-
erties. For Pavement A, the fatigue life estimated with the measured
152 mm Aggregate
dynamic modulus values is always higher than those estimated with
Subgrade the measured resilient modulus or the predicted dynamic modulus.
Subgrade
This result is mainly because the measured dynamic modulus was
Subgrade the highest at all temperatures, which results in the smallest calcu-
lated horizontal strains under the HMA layer and therefore longer
fatigue life.
Pavement A Pavement B Pavement C
The number of repetitions to failure estimated with the predicted
FIGURE 11 Considered pavement designs. dynamic modulus is higher than that estimated with the measured
resilient modulus at temperatures of −15°C and 5°C; at higher tem-
peratures, the opposite finding is observed. This result is mainly
because the fatigue law is a function of the horizontal transverse
load of 40 kN (9,000 lb), with a tire inflation pressure of 724 kPa
(105 psi) and separated by a distance of 330 mm (13 in.) center to
center.
Only bottom-up fatigue was considered in this study and the fatigue 100 MR |E*| Measured |E*| Predicted
model as suggested by the proposed AASHTO mechanistic–empirical
Fatigue Life (Millions)

design guide was used. The fatigue law is given by

3.9492 1.281 10
⎛ 1⎞ ⎛ 1⎞
N f = 0.00432 ∗ k1′ ∗ C ⎜ ⎟ ⎜⎝ E ⎟⎠ (13)
⎝ ⑀t ⎠
1
where -15 5 20 30 40
Temperature (°C)
Nf = number of repetitions to fatigue cracking,
(a)
k′1 = correction parameter for different asphalt layer thickness
(hac) effects,
C = laboratory-to-field adjustment factor, 1000 MR |E*| Measured |E*| Predicted
Fatigue Life (Millions)

⑀t = tensile strain at critical location, and

E = stiffness of material.
100
The calibration parameters used in Equation 13 are defined as
follows:

1 10
k1′ = (14) -15 5 20 30 40
0.003602
0.000398 + Temperature (°C)
1 + e(11.02−3.49∗hac )
(b)

where hac is the total thickness of the asphalt layers in inches.

10000 MR |E*| Measured |E*| Predicted
Fatigue Life (Millions)

C = 10 M
(15)
1000
where
100
⎛ Vb ⎞
M = 4.84 ⎜ − 0.69⎟ (16)
⎝ Va + Vb ⎠ 10
-15 5 20 30 40
Temperature (°C)
where Vb is the percentage of effective binder content and Va is the
(c)
percentage of air voids.
The pavements were analyzed with the KenPave software (7 ). FIGURE 12 Predicted fatigue life: (a) Pavement A, (b) Pavement B,
The allowable number of repetitions for the considered standard and (c) Pavement C.
170 Transportation Research Record 1970

strain under the HMA layer and the HMA stiffness, which are cor- Three pavement structures were analyzed with the modulus data
related. In fact, higher stiffness results in lower fatigue life, but it from the dynamic test and the resilient test. It was found that the
also results in smaller strains under the HMA layer, which means fatigue life of pavements is affected by the choice of the HMA mod-
higher fatigue life. Pavement B behaved in the same manner as ulus. In general, fatigue life estimates obtained with the predicted
Pavement A with higher fatigue lives in all cases since the pavement dynamic moduli were smaller than those obtained with the actual
is thicker. For the thickest pavement (Pavement C), the same result measured values. However, as expected, the HMA thickness was the
is observed, except that the fatigue life estimated from the predicted controlling factor in predicting fatigue life.
dynamic modulus is higher than that predicted with the measured
resilient modulus at all considered temperatures.
As expected, the pavement thickness is the controlling parameter ACKNOWLEDGMENTS
for fatigue life. As shown in Figure 12, the fatigue life of Pavement B,
which is 76 mm HMA thicker than Pavement A, is 10 times higher This research is sponsored by the Virginia Department of Trans-
portation. The help of Billy Hobbs, Samer Katicha, and Myunggoo
than that of Pavement A regardless of the HMA property input. Sim-
Jeong from Virginia Tech Transportation Institute; Kevin McGhee
ilarly, Pavement C, which is 152 mm HMA thicker than Pavement A,
and Troy Deeds from the Virginia Transportation Research Coun-
has on average 40 times higher fatigue life than Pavement A.
cil; and Susanne Aref from the Statistics Department at Virginia
Tech is greatly appreciated.

SUMMARY
REFERENCES
In this study, resilient modulus and dynamic modulus tests were
conducted on two typical mixes used in the commonwealth of Vir- 1. Guide for the Design of Pavement Structures. AASHTO, Washington,
ginia: SM-9.5A and BM-25.0. The size of the specimen was found D.C., 1993.
2. Guide for Mechanistic-Empirical Design of New and Rehabilitated Pave-
to statistically affect the measured resilient modulus value. Resilient
ment Structures. Final Report, NCHRP 1-37A. Transportation Research
modulus values obtained in the 100-mm-diameter specimens were Board of the National Academies, Washington, D.C., 2004.
higher than those obtained in the 150-mm-diameter specimens at all 3. Loulizi, A., I. L. Al-Qadi, S. Lahouar, and T. E. Freeman. Measurement
testing temperatures. For the dynamic modulus test, at temperatures of Vertical Compressive Stress Pulse in Flexible Pavements: Represen-
tation for Dynamic Loading Tests. In Transportation Research Record:
of −15°C, 5°C, and 20°C, the phase angle decreases as the frequency
Journal of the Transportation Research Board, No. 1816, Transportation
increases, whereas at 30°C and 40°C, the behavior of the phase Research Board of the National Academies, Washington, D.C., 2002, pp.
angle as a function of frequency becomes complex, possibly because 125–136.
of the increasing effect of the aggregate. A similar pattern was 4. Roque, R., and W. G. Buttlar. Development of a Measurement and
Analysis Method to Accurately Determine Asphalt Concrete Properties
observed with frequency (low frequency corresponds to high tem-
Using the Indirect Tensile Mode. Proc., Association of Asphalt Paving
perature). A sigmoidal function was used to fit the dynamic modulus Technologists, Vol. 61, 1992, pp. 304–332.
data, with very good results. The BM-25.0 mix had a higher dynamic 5. Clyne, T. R., X. Li, M. O. Marasteanu, and E. L. Skok. Dynamic and
modulus than the SM-9.5A mix at all frequencies and at all tested Resilient Modulus of MNDOT Asphalt Mixtures. Report MN/RC-2003-
09. Minnesota Department of Transportation, Minneapolis, 2003.
temperatures.
6. Pellinen, T. K., and M. W. Witczak. Stress Dependent Master Curve
A strong relation was observed between the dynamic modulus Construction for Dynamic Modulus. Journal of the Association of Asphalt
test performed at 5 Hz and the resilient modulus test performed at a Paving Technologists, Vol. 71, 2002, pp. 281–309.
loading time of 0.03 s. However, the dynamic modulus test provides 7. Huang, Y. H. Pavement Analysis and Design, 2nd ed. Prentice Hall,
Upper Saddle River, N.J., 2003.
a better characterization of HMA than the resilient modulus test
because it provides full characterization of the mix over temperature The Characteristics of Bituminous Paving Mixtures to Meet Structural Require-
and loading frequencies. ments Committee sponsored publication of this paper.