Practical Procedure for Developing

Dynamic Modulus Master Curves

for Pavement Structural Design
Ramon Bonaquist and Donald W. Christensen

A dynamic modulus master curve for asphalt concrete is a critical input It is desirable that equipment for performing dynamic modulus
for flexible pavement design in the mechanistic–empirical pavement master curve testing be available to highway agencies at a reasonable
design guide developed in NCHRP Project 1-37A. The recommended cost, and that the test procedure be appropriate for agency laborato-
testing to develop the modulus master curve is presented in AASHTO ries. In a recently completed FHWA pooled fund study, Dougan
Provisional Standard TP62-03, Standard Method of Test for Deter- et al. identified several issues associated with AASHTO TP62-03
mining Dynamic Modulus of Hot-Mix Asphalt Concrete Mixtures. It and concluded that the overall time required to perform the testing
includes testing at least two replicate specimens at five temperatures must be shortened if highway agencies are going to use it for routine
between 14°F and 130°F (−10°C and 54.4°C) and six loading rates testing (1). In NCHRP Project 9-19, Witczak, has studied the mini-
between 0.1 and 25 Hz. The master curve and shift factors are then mum testing required to develop the dynamic modulus master curves
developed from this database of 60 measured moduli using numerical and concluded that reasonable master curves can be developed using
optimization. The testing requires substantial effort, and there is much tests at three temperatures, 14°F, 70°F, and 130°F (−10.0°C, 21.1°C,
overlap in the measured data, which is not needed when numerical and 54.4°C), and loading frequencies of 33, 2.22, 0.15, and 0.01 Hz
methods are used to perform the time–temperature shifting for the mas- (2). This reduced sequence still requires testing at 14°F (−10°C),
ter curve. This paper presents an alternative to the testing sequence which Dougan et al. found difficult because of moisture condensa-
specified in AASHTO TP62-03. It requires testing at only three tem- tion and ice formation (1). Additionally, low-temperature testing sig-
peratures between 40°F and 115°F (4.4°C and 46.1°C) and four rates of nificantly increases the cost of the environmental chamber and
loading between 0.01 and 10 Hz. An analysis of data collected using the increases the loading capacity and cost of the testing equipment. If
two approaches shows that comparable master curves are obtained. testing at this temperature can be eliminated, the cost of the equip-
This alternative testing sequence can be used in conjunction with the ment, the complexity of the procedure, and the overall time required
simple performance test system developed in NCHRP Project 9-29 to to develop a master curve can be significantly reduced.
develop master curves for structural design. The approach described in this paper takes advantage of the fact
that, for engineering purposes, asphalt binders reach approximately
the same glassy modulus at very low temperatures. By using this
This paper describes an approach for reducing the amount of effort binder modulus and recently developed relationships to predict
required to develop dynamic modulus master curves for pavement mixture dynamic modulus from binder modulus and volumetric
structural design. The modulus of asphalt concrete is a critical input data, an estimate of the limiting maximum modulus of the mixture
for flexible pavement design in the mechanistic–empirical pavement can be made and used in the development of the dynamic modulus
design guide developed in NCHRP Project 1-37A. A dynamic mod- master curve.
ulus master curve is used to express the modulus of asphalt concrete
as a function of temperature and loading rate. The dynamic modulus
test protocol was developed in NCHRP Projects 9-19 and 1-37A and NCHRP PROJECT 1-37A DYNAMIC MODULUS
has been standardized as AASHTO Provisional Standard TP62-03, MASTER CURVE
Standard Method of Test for Determining Dynamic Modulus of Hot-
Mix Asphalt Concrete Mixtures. The recommended test sequence in To account for temperature and rate of loading effects on the modu-
AASHTO TP62-03 consists of testing a minimum of two replicate lus of asphalt concrete, the mechanistic–empirical pavement design
specimens at temperatures of 14°F, 40°F, 70°F, 100°F, and 130°F guide uses a modulus determined from a master curve constructed
(−10°C, 4.4°C, 21.1°C, 37.8°C, and 54.4°C) and loading frequencies from dynamic modulus measurements at multiple temperatures and
of 25, 10, 5, 1.0, 0.5, and 0.1 Hz. This testing provides a database of loading rates. Master curves are constructed using the principle of
60 dynamic modulus measurements from which the parameters of time–temperature superposition. First a standard reference tempera-
the master curve are determined by numerical optimization. ture is selected, in this case 70°F (21.1°C); then the measured data at
various temperatures are shifted with respect to loading frequency
R. Bonaquist, Advanced Asphalt Technologies, LLC, 108 Powers Court, Suite until the curves merge into a single smooth function. The master curve
100, Sterling, VA 20166-9321. D. W. Christensen, Advanced Asphalt Tech- of modulus as a function of frequency formed in this manner describes
nologies, LLC, 210 W. Hamilton Avenue, #304, State College, PA 16801. the loading rate dependency of the material. The amount of shifting
Transportation Research Record: Journal of the Transportation Research Board,
at each temperature required to form the master curve describes
No. 1929, Transportation Research Board of the National Academies, Washington, the temperature dependency of the material. Thus, both the master
D.C., 2005, pp. 208–217. curve and the shift factors are needed for a complete description of

208
Bonaquist and Christensen 209

the rate and temperature effects. Figure 1 presents an example of a tion have been used with asphalt materials (4). In the mechanistic–
master curve constructed in this manner and the resulting shift factors. empirical pavement design guide, the shift factors were expressed
In the mechanistic–empirical pavement design guide, the sig- as a function of the binder viscosity to allow aging over the life of
moidal function in Equation 1 is used to describe the rate dependency the pavement to be considered using the global aging model devel-
of the modulus master curve (3). oped by Mirza and Witczak (5). Equation 3 presents the shift factor
relationship used in the mechanistic–empirical pavement design
α
log ( E* ) = δ + (1) guide (3).
1 + eβ+ γ ( log ωr )
log[ a(T )] = c[log( η) − log( η70 RTFOT )] (3)
where
where
E* = dynamic modulus,
ωr = reduced frequency, a(T) = shift factor as a function of temperature and age,
δ = minimum value of E*, η = viscosity at the age and temperature of interest,
δ+α = maximum value of E*, and η70RTFOT = viscosity at reference temperature of 70°F (21.1°C) and
β, γ = parameters describing the shape of the sigmoidal function. rolling thin-film oven (RTFO) aging, and
c = fitting parameter.
The temperature dependency of the modulus is incorporated in the
reduced frequency parameter, ωr, in Equation 1. Equation 2 defines The viscosity as a function of temperature can be expressed using
the reduced frequency as the actual loading frequency multiplied by the ASTM viscosity–temperature relationship given in Equation 4 (6 ).
the time–temperature shift factor, a(T).
log log η = A + VTS log TR ( 4)
ω r = ω × a( T ) (2 a)
where
log(ω r ) = log(ω ) + log[ a(T )] (2 b)
η = viscosity, cP;
TR = temperature, Rankine;
where
A = regression intercept; and
ω = loading frequency, VTS = regression slope of viscosity–temperature susceptibility.
a(T) = shift factor as a function of temperature, and
Combining Equations 3 and 4 yields the shift factor as a function
T = temperature.
of the temperature relationship used in the mechanistic–empirical
The shift factors are a function of temperature. Various equations pavement design guide for the construction of dynamic modulus
such as the Arrhenius function and the Williams–Landel–Ferry equa- master curves from laboratory test data.

10,000,000

1,000,000

15.8°F
40°F
E*, psi

70°F
100,000
100°F
6
130°F
4
Log Shift Factor

Fit
2
0
10,000 -2
-4
-6
0 50 100 150
Temperature, °F

1,000
1.0E-07 1.0E-05 1.0E-03 1.0E-01 1.0E+01 1.0E+03 1.0E+05 1.0E+07
Reduced Frequency, Hz

FIGURE 1 Schematic of master curve and shift factors.

210 Transportation Research Record 1929

log[ a(T )] = c[10 A + VTS log TR − log( η70 RTFOT )] (5) VMA = voids in mineral aggregates, %;
VFA = voids filled with asphalt, %; and
where A and VTS are viscosity–temperature susceptibility parameters ⎟ G*⎟ binder = dynamic shear modulus of binder, psi.
for RTFOT aging and c is a fitting parameter. On the basis of research conducted by Christensen and Andersen
Substituting Equation 5 into Equation 2b and the result into Equa- during the Strategic Highway Research Program (SHRP), a rea-
tion 1 yields the form of the dynamic modulus master curve rela- sonable engineering estimate of the maximum dynamic shear mod-
tionship used in the mechanistic–empirical pavement design guide ulus for asphalt binders is 145,000 psi (1 GPa) (4 ). This estimate is
for the development of master curves from laboratory test data. based on data collected on neat binders, and care should be used
α when it is applied to modified binders. Substituting this value into
log ( E* ) = δ + (6) Equation 8 yields the recommended equation for estimating the
1+ e
{ (
β + γ log( ω )+ c ⎡⎣10 A + VTS log TR − log η70
RTFOT )⎤⎦}
limiting maximum modulus of asphalt concrete mixtures from
volumetric data.
where δ is the limiting minimum value of E* and δ + α is the limiting
maximum value of E*. ⎡ ⎛ VMA ⎞
The fitting parameters (α, β, δ, γ, and c) are determined through E* max = Pc ⎢ 4, 200, 000 ⎜⎝1 − ⎟
⎣ 100 ⎠
numerical optimization of Equation 6 using mixture test data col-
lected in accordance with AASHTO TP62-03. Because of equip- FA × VMA ⎞ ⎤
⎛ VF
+ 435, 000 ⎜
ment limitations, neither the limiting maximum nor limiting minimum ⎝ 10, 000 ⎟⎠ ⎥⎦
modulus can be measured directly; therefore, these parameters are 1 − Pc
estimated through the curve-fitting process. +
⎡⎛ VMA ⎞ ⎤
⎢ ⎝1 − 100 ⎠ VMA ⎥
⎢ + ⎥
⎣ 4, 200, 000 ( )
435, 000 VFA ⎦ (9a)
PROPOSED MODIFICATION
where
The modification proposed in this paper is to estimate the limiting
435,000 ( VFA ) ⎞
0.58
maximum modulus on the basis of binder stiffness and mixture vol- ⎛
⎝ 20 + ⎠
umetric data using the Hirsch model developed by Chirstensen et al. Pc = VMA
in NCHRP Projects 9-25 and 9-31 (7 ). For a specified limiting max- (
⎛ 435,000 VFA ⎞ ) 0.58

imum modulus, the master curve relationship given in Equation 6 650 + ⎝ ⎠ (9b)
VMA
reduces to
⎟ E*⎟ max = limiting maximum mixture dynamic modulus.
( Max − δ )
log ( E* ) = δ + (7) The limiting maximum modulus computed from Equation 9 varies
1+ e
{ (
β + γ log( ω )+ c ⎡⎣10 A + VTS log TR − log η70
RTFOT )⎤⎦}
from about 3,000,000 (20.6 GPa) for mixtures with very high VMA
and low VFA to about 3,800,000 (26.2 GPa) for mixtures with low
where VMA and high VFA. These limiting maximum modulus values appear
t = loading frequency; very rational. For conditions with low VMA and high VFA, the lim-
Max = specified limiting maximum modulus; and iting maximum modulus approaches the 4,000,000 psi (28 GPa) often
δ, β, γ, and c = fitting parameters. assumed for portland cement concrete.
The four unknown fitting parameters are still estimated using
numerical optimization of the test data, but because the limiting EFFECT ON MASTER CURVES OF ELIMINATING
maximum modulus is specified, data at low test temperatures are no
LOW TEMPERATURE TEST DATA
longer needed.
Equation 8 presents the Hirsch model, which allows estimation of This section compares master curves developed by fitting Equation 6
the modulus of asphalt concrete mixtures from binder stiffness data using the complete AASHTO TP62-03 data set and Equation 7
and volumetric properties. using a reduced data set obtained by eliminating the lowest temper-
ature data from the AASHTO TP62-03 test series. For the reduced
⎡ ⎛ VMA ⎞ ⎛ VFA × VMA ⎞ ⎤
E* mix = Pc ⎢ 4,200,000 ⎜⎝1− *
⎟⎠ + 3 G binder ⎜
⎝ 10,000 ⎟⎠ ⎥⎦
data set, the limiting maximum modulus was estimated using Equa-
⎣ 100
tion 9. For the comparisons, the database of dynamic modulus mea-
1 − Pc surements assembled by Pellinen for NCHRP Project 9-19 was
+ used (8). This database includes test data from replicate samples tested
⎡⎛ VMA ⎞ ⎤
⎢ ⎝1 − 100 ⎠ VMA ⎥ at temperatures of 15.8°F, 40°F, 70°F, 100°F, and 130°F (−9°C,
⎢ + ⎥ 4.4°C, 21.1°C, 37.8°C, and 54.4°C) and frequencies of 25, 10, 5,
⎢⎣ 4 ,2
2 00 , 000 *
3VFA G binder ⎥⎦ (8a)
1.0, 0.5, and 0.1 Hz. Pertinent properties of the mixtures included
in the evaluation are summarized in Table 1. The mixtures include
where
five mixtures from the MNRoad project, 11 mixtures from the FHWA
0.58 Pavement Testing Facility, and six mixtures from the WesTrack
⎛ VFA × 3 G* binder ⎞
⎜ 20 + ⎟ project. This combination of mixtures has a range of nominal max-
⎝ VMA ⎠
Pc = 0.58 imum aggregate sizes, binder type including two modified binders,
⎛ VFA × 3 G* binder ⎞ and volumetric properties. Master curves for each mixture in Table 1
650 + ⎜⎝ ⎟
⎠
VMA (8b) were fit using the two methods. Goodness of fit statistics for the
Bonaquist and Christensen 211

TABLE 1 Summary of Master Curve Comparison Study

Master Curve Goodness of

Fit Statistics

Mix Volumetric Properties RTFOT Binder Properties Complete Reduced

Description Data Set Data Set
AC Va VMA VFA η70RTFOT
Project ID Binder Mix Type (%) (%) (%) (%) A VTS (cP) R2 Se/Sy R2 Se/Sy

MNRoad Cell 16 AC-20 Fine 9.5 mm 5.08 8.2 18.0 54.4 10.7826 −3.6065 1.22E+09 0.9965 0.002 0.9959 0.004
MNRoad Cell 17 AC-20 Fine 9.5 mm 5.45 7.7 18.2 57.6 10.7826 −3.6065 1.22E+09 0.9938 0.004 0.9938 0.006
MNRoad Cell 18 AC-20 Fine 9.5 mm 5.83 5.6 17.1 67.2 10.7826 −3.6065 1.22E+09 0.9974 0.002 0.9974 0.002
MNRoad Cell 20 120/150 Pen Fine 9.5 mm 6.06 6.3 18.3 65.6 10.8101 −3.6254 3.96E+08 0.9925 0.006 0.9924 0.008
MNRoad Cell 22 120/150 Pen Fine 9.5 mm 5.35 6.5 16.9 61.5 10.8101 −3.6254 3.96E+08 0.9717 0.020 0.9692 0.028
ALF Lane 1 AC-5 Fine 19 mm 4.75 6.1 16.9 63.9 10.6766 −3.5740 5.35E+08 0.9801 0.013 0.9789 0.020
ALF Lane 2 AC-20 Fine 19 mm 4.85 6.5 17.3 62.5 10.6569 −3.5594 1.38E+09 0.9765 0.017 0.9610 0.040
ALF Lane 3 AC-5 Fine 19 mm 4.75 7.7 18.3 57.9 10.6766 −3.5740 5.35E+08 0.9856 0.011 0.9793 0.024
ALF Lane 4 AC-20 Fine 19 mm 4.9 9.7 20.3 52.1 10.6569 −3.5594 1.38E+09 0.9803 0.013 0.9800 0.019
ALF Lane 5 AC-10 Fine 19 mm 4.8 8.6 19.0 54.7 10.7805 −3.6116 5.72E+08 0.9922 0.005 0.9921 0.008
ALF Lane 7 Styrelf Fine 19 mm 4.9 11.9 22.1 46.2 8.9064 −2.9089 4.02E+09 0.9726 0.016 0.9723 0.023
ALF Lane 8 Novophalt Fine 19 mm 4.7 11.9 21.6 45.0 8.8136 −2.8817 1.58E+09 0.9929 0.004 0.9902 0.006
ALF Lane 9 AC-5 Fine 19 mm 4.9 7.7 18.4 58.1 10.6766 −3.5740 5.35E+08 0.9902 0.007 0.9902 0.010
ALF Lane 10 AC-20 Fine 19 mm 4.9 9.3 19.8 53.0 10.6569 −3.5594 1.38E+09 0.9960 0.003 0.9930 0.007
ALF Lane 11 AC-5 Fine 37.5 mm 4.05 6.0 14.2 57.9 10.6766 −3.5740 5.35E+08 0.9913 0.006 0.9839 0.017
ALF Lane 12 AC-20 Fine 37.5 mm 4.05 7.4 15.5 52.3 10.6569 −3.5594 1.38E+09 0.9844 0.009 0.9811 0.016
WesTrack Sec 2 PG 64-22 Fine 19 mm 5.02 10.4 17.3 39.9 11.0757 −3.7119 1.63E+09 0.9933 0.004 0.9893 0.008
WesTrack Sec 4 PG 64-22 Fine 19 mm 5.24 6.6 14.3 53.8 11.0757 −3.7119 1.63E+09 0.9899 0.006 0.9899 0.008
WesTrack Sec 7 PG 64-22 Coarse 19 mm 6.28 6.9 15.9 56.6 11.0757 −3.7119 1.63E+09 0.9547 0.029 0.9536 0.040
WesTrack Sec 15 PG 64-22 Fine 19 mm 5.55 8.7 16.9 48.4 11.0757 −3.7119 1.63E+09 0.9883 0.008 0.9877 0.012
WesTrack Sec 23 PG 64-22 Coarse 19 mm 5.78 4.9 13.0 62.3 11.0757 −3.7119 1.63E+09 0.9949 0.003 0.9949 0.004
WesTrack Sec 24 PG 64-22 Coarse 19 mm 5.91 7.2 15.4 53.2 11.0757 −3.7119 1.63E+09 0.9742 0.015 0.9735 0.022

AC = asphalt content; Va = air voids.

individual master curves are summarized in the last four columns To compare master curves for all of the mixtures, dynamic mod-
of Table 1. Two statistics are provided: the explained variance, uli were calculated for temperatures ranging from −30°F to 150°F
R2, and ratio of the standard error of estimate to the standard devi- (−34.4°C to 65.6°C) using loading rates of 25, 10, 1, 0.1, and 0.01 Hz.
ation of the measured values, Se/Sy. High R2 and low Se/Sy val- Results are shown in Figure 4. As can be seen in Figure 4, the two
ues indicate good fitting of the master curve to the measured data. approaches sometimes yield differences at high and low moduli
As shown in Table 1, both methods provide excellent fit to the primarily because of differences caused by the limiting maximum
measured data. modulus.
Although the two master curves provide excellent fit to the mea- Figure 5 compares limiting maximum moduli from the two pro-
sured data, there are differences in some cases. Figures 2 and 3 pre- cedures. The master curves developed using the complete AASHTO
sent examples of the master curves generated. Figure 2, which is for TP 62-03 data set yield unrealistically high moduli in four cases,
Lane 2 at the FHWA Pavement Testing Facility, is an example of the ALF 2, ALF 3, ALF 10, and ALF 11. Master curves developed in this
worst agreement between the two methods. The limiting maximum manner also yield unrealistically low values in two cases, ALF 8 and
modulus computed with the Hirsch model at 3,237,000 psi (223. GPa) WSTR 2. Table 2 summarizes limiting modulus values averaged
is much lower than the 6,714,000 psi (46.4 GPa) limiting maximum over similar mixtures. The master curves developed using the two
modulus obtained from the complete AASHTO TP62-03 data set. As approaches produce reasonably similar average limiting maximum
shown, the difference in the limiting maximum modulus also affects modulus values except for the ALF mixtures, which had the four
the limiting minimum modulus. The limiting minimum modulus unrealistically high values when the complete AASHTO TP62-03
from the reduced data set is higher, 17,000 psi (117 MPa) compared data set was used. The quality of the data for the low-temperature
with 2,000 psi (14 MPa) for the master curve developed using the test condition has a major impact on the limiting maximum modu-
complete AASHTO TP62-03 set. Both methods fit the measured data lus determined from the complete AASHTO TP62-03 data set. Pelli-
well over the temperature range from 40°F to 130°F (4.4°C to 54.4°C), nen reported higher coefficients of variation for data collected at
and the shift factors for the two methods are essentially the same. 15.8°F (−9°C) and 130°F (54.4°C) (8). She also reported that strain
Figure 3, which is for Cell 17 at the MNRoad project, is an example levels for the 15.8°F data were significantly lower than those at other
of best agreement between the two methods. In this case, the two temperatures. This coupled with potential friction in the LVDT guide
approaches yield the same master curves. rod used in the AASHTO TP62-03 recommended instrumentation
212 Transportation Research Record 1929

10,000,000

1,000,000

15.8°F
E*, psi

40°F
100,000
70°F
6
AASHTO TP 62 100°F
4

Log Shift Factor

REDUCED SET
2 130°F
0
AASHTO TP 62
10,000 -2
-4 REDUCED SET
-6
0 50 100 150
Temperature, °F

1,000
1.0E-07 1.0E-05 1.0E-03 1.0E-01 1.0E+01 1.0E+03 1.0E+05 1.0E+07
Reduced Frequency, Hz

FIGURE 2 Comparison of fitted master curves using complete AASHTO TP62-03 data and reduced data set for Lane 2 from FHWA
pavement testing facility.

10,000,000

1,000,000

15.8°F
40°F
70°F
E*, psi

100,000 100°F
130°F
6 AASHTO TP 62 AASHTO TP 62
4
Log Shift Factor

REDUCED SET REDUCED SET

2
0
10,000 -2
-4
-6
0 50 100 150
Temperature, °F

1,000
1.00E-07 1.00E-05 1.00E-03 1.00E-01 1.00E+01 1.00E+03 1.00E+05 1.00E+07
Reduced Frequency, Hz

FIGURE 3 Comparison of fitted master curves using complete AASHTO TP62-03 data and the reduced data set for Cell 17
from the MNRoad.
Bonaquist and Christensen 213

10,000,000

1,000,000
Reduced Data Set E*, psi

100,000

10,000

1,000
1,000 10,000 100,000 1,000,000 10,000,000
AASHTO TP62-03 E*, psi

FIGURE 4 Comparison of dynamic moduli from fitted master curves.

8,000,000

7,000,000
AASHTO TP62-03
REDUCED DATA SET
6,000,000
Maximum Modulus, psi

5,000,000

4,000,000

3,000,000

2,000,000

1,000,000

0
AL 0

AL 1

M 2
F1

W R15

W R23

24
W R2

W R4
F2

F3

F4

F5

F8

AL 9

M 6

18

20

W 22

W R7
F7

17
F1

F1

F1
F

R
N

N
AL

AL

AL

AL

AL

AL

AL

N
AL

ST

ST

ST
ST

ST

ST
M

Section

FIGURE 5 Comparison of limiting maximum moduli from two procedures.

214 Transportation Research Record 1929

TABLE 2 Limiting Maximum and Minimum Modulus Values Averaged over Mixture Type

Limiting Maximum Limiting Minimum

Modulus (psi) Modulus (psi)

Full Full
AASHTO Hirsch AASHTO
TP62-03 Model TP62-03 Reduced
Mixture Number Data Set Estimate Data Set Data Set

MNRoad 5 3,109,668 3,206,382 4,383 3,873

ALF 19 mm 9 3,867,636 3,077,348 11,247 11,044
ALF 25 mm 2 4,735,721 3,324,313 14,092 13,426
WesTrack fine 19 mm 3 2,985,757 3,187,770 16,589 14,077
WesTrack coarse 19 mm 3 2,934,664 3,345,151 19,101 15,569
All mixtures 22 3,526,808 3,180,702 11,745 10,662

is the most likely cause of the high variability in the low tempera- modulus represents the stiffness of the aggregate structure in the
ture measurements. The combined effects of low strains and friction absence of binder. The two procedures provide the same rankings for
probably also explain the unrealistically high moduli measured for the mixtures compared in this evaluation.
four of the ALF mixtures.
The limiting maximum modulus also affects the limiting mini-
mum modulus because the sigmoidal master curve function is sym- TESTING TEMPERATURES AND RATES FOR
metric. Figure 6 compares limiting minimum modulus values for PROPOSED ALTERNATIVE APPROACH
individual mixtures from the two procedures. The largest differences
between the two methods occur for the same mixtures that have the The previous section showed that reasonable dynamic modulus
largest differences in the limiting maximum modulus. Limiting min- master curves can be obtained without low-temperature test data by
imum modulus values averaged over similar mixtures are summa- estimating the limiting maximum modulus using the Hirsch model.
rized in Table 2. The two approaches produce reasonably similar The measured data that were fitted were collected at temperatures of
average limiting minimum modulus values. The limiting minimum 40°F, 70°F, 100°F, and 130°F (4.4°C, 21.1°C, 37.8°C, and 54.4°C)

35,000
AASHTO TP62-03
REDUCED DATA SET
30,000

25,000
Maximum Modulus, psi

20,000

15,000

10,000

5,000

0
0

AL 1
2
F1

W 15

W 23

24
W 2

W 4
F2

F3

F4

F5

F8

AL 9

16

18

20

W 22

W R7
F7

17
F1

F1

F1

R
F

R
N

N
AL

AL

AL

AL

AL

AL

AL

N
AL

ST

ST

ST
AL

ST

ST

ST
M

M
M

Section

FIGURE 6 Comparison of limiting minimum moduli from two procedures.

Bonaquist and Christensen 215

and frequencies of 25, 10, 5, 1.0, 0.5, and 0.1 Hz. However, these Equation 2, that results in loading frequencies at 115°F (46.1°C) of
temperatures and loading rates may not be optimal. This section pre- approximately 0.02 Hz to obtain reduced frequencies at 70°F (21.1°C)
sents an analysis of the temperatures and loading rates that should between 10−4 and 10−5 Hz, respectively. Thus, the use of a loading rate
be used in combination with the proposed approach to develop of 0.01 Hz at 115°F (46.1°C) will provide approximately the same
dynamic modulus master curves efficiently. lower range as the AASHTO TP62-03 approach.
In the proposed approach, a reasonable, rational estimate of the Because the shift factor relationship is not linear, a minimum of
limiting maximum modulus is provided by the Hirsch model. This three temperatures spaced as widely as possible should be used in
eliminates the need for testing at low temperatures and the potential the testing program. That will provide a reasonable estimate of the
inaccuracies caused by these difficult testing conditions. To provide coefficient, c, in the shift factor relationship, Equation 3. A low test-
an accurate estimate of the limiting minimum modulus, data with this ing temperature of 40°F (4.4°C) would allow reasonably priced
approach should be collected to the smallest reduced frequency pos- environmental chambers to be used and would eliminate the icing
sible. From Equation 2, the reduced frequency is a function of both problems that occur during testing at temperatures below freezing.
temperature and frequency of loading. High-temperature, slow- Also as shown in Figure 7, the log of the shift factors at this tem-
frequency dynamic modulus tests result in the smallest reduced fre- perature have approximately the same magnitude, but opposite sign
quency values. For the testing conditions specified in AASHTO TP as those at the maximum temperature of 115°F (46.1°C).
62-03 the minimum reduced frequency for the mixtures included in Thus the recommended testing temperatures for the proposed
this study ranged from 10−4 to 10−5 Hz. These appear to have been suf- approach are 40°F, 70°F, and 115°F (4.4°C, 21.1°C, and 46.1°C).
ficient to define the limiting minimum modulus through fitting with On the basis of the performance of typical LVDT deformation sys-
the sigmoidal function. The most efficient way to decrease the mini- tems, the maximum frequency of loading should be limited to 10 Hz.
mum reduced frequency in the testing program is to increase temper- Using loading frequencies of 10, 1, 0.1, and 0.01 Hz at each of the
ature; however, for the glued gauge point instrumentation used in the temperatures results in well-spaced data in reduced frequency with
dynamic modulus test, the maximum testing temperature appears to a minimum of overlap.
be approximately 115°F (46.1°C). Above that temperature, the gauge
points may loosen, particularly when they are attached to the matrix
of fine aggregate and binder. Higher temperatures may be possible EXAMPLE USING PROPOSED PROCEDURE
when the gauge points are attached to the coarse aggregate, but that
cannot be ensured in most mixtures. Figure 7 presents experimentally This section illustrates the development of a master curve using
determined shift factors from the complete AASHTO TP62-03 data the proposed procedure. The mixture that was tested was a coarse-
set for the mixtures included in this evaluation. As shown, the shift graded, 9.5-mm limestone mixture made with a PG 70-22 binder.
factor for the maximum recommended testing temperature of 115°F The viscosity–temperature susceptibility parameters for the binder
(46.1°C) ranges from about 10−2.3 = 0.005 to 10−3.2 = 0.0006. From are A = 10.299 and VTS = −3.426. The test specimens were compacted

2
Log Shift Factor

-1

-2

-3

-4

-5

-6
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140
Temperature, °F

FIGURE 7 Shift factors as a function of temperature for mixtures in Table 1.

216 Transportation Research Record 1929

TABLE 3 Example Dynamic Modulus Test Data Collected with 15.8% and a VFA of 76.2%, the limiting maximum modulus from
Proposed Approach Equation 9 is 3,376,743 psi. By using this value of the limiting max-
imum modulus, the viscosity–temperature susceptibility parameters,
Specimen 1 Specimen 2 and the measured data, the master curve parameters can be obtained
Temp. Frequency Modulus Phase Modulus Phase
through numerical optimization of Equation 7. The optimization can
(F) (Hz) (ksi) (degree) (ksi) (degree) be performed using the Solver function in Microsoft Excel. That is
done by setting up a spreadsheet to compute the sum of the squared
40 0.01 771.6 25.0 901.1 23.7 errors between the logarithm of the measured dynamic moduli and the
40 0.1 1274.9 19.0 1496.1 17.9 values predicted by Equation 7. The Solver function is used to mini-
40 1 1861.7 13.9 2164.1 12.8 mize the sum of the squared errors by varying the fitting parameters
40 10 2458.2 9.6 2811.0 8.6 in Equation 7. The following initial estimates are recommended:
70 0.01 161.0 30.0 174.6 29.6
δ = 0.5, β = −1.0, γ = 0.5, and c = 1.2. The master curve developed
from the example data is shown in Figure 8. The goodness of fit sta-
70 0.1 362.7 29.2 398.3 29.2
tistics show that Equation 7 provides an excellent fit to the measured
70 1 771.7 24.5 844.4 24.3 data with an R2 greater than 0.99 and an Se/Sy less than 0.04. With the
70 10 1332.1 18.0 1446.0 18.0 use of the proposed temperatures and loading rates, the measured data
115 0.01 23.3 24.4 28.3 22.9 cover approximately 80% of the range defined by the fitted limiting
115 0.1 50.8 27.8 53.4 27.9 minimum and computed limiting maximum moduli.
115 1 137.8 29.4 140.9 29.5
115 10 336.2 29.7 352.8 29.9
EQUIPMENT REQUIREMENTS
AND TESTING TIME

Table 4 compares the requirements for equipment to perform the

at the optimum asphalt content of 5.5% to 4.0% air voids. For this dynamic modulus master curve testing as specified in AASHTO
condition the percent voids in the mineral aggregate was 15.8% and TP62-03 and the proposed approach (9). The primary difference in
the percent voids filled with asphalt was 76.2%. Table 3 presents the devices is the temperature range and the required dynamic load-
dynamic modulus data measured on replicate samples using the ing capacity. The substantial difference in the low temperature range
combination of temperatures and loading rates recommended in the of the environmental chamber is responsible for the difference in
proposed approach. the estimated cost of the equipment. The increased dynamic load-
The first step for fitting the master curve is to estimate the limiting ing capacity results in a nominal increase in the estimated cost of
maximum modulus using Equation 9. For a mixture with a VMA of the equipment. The Simple Performance Test System developed in

10,000,000

E* max = 3,376,744 psi

1,000,000
2
R = 0.9978
Se/Sy = 0.038

40°F
E*, psi

100,000
3 70°F
Log Shift Factor

2
1
115°F
0
-1
10,000
-2 FIT
-3
E* min = 4,259 psi 0 50 100 150
Temperature, °F

1,000
1.0E-06 1.0E-04 1.0E-02 1.0E+00 1.0E+02 1.0E+04 1.0E+06
Reduced Frequency, Hz
FIGURE 8 Example master curve using proposed procedure.
Bonaquist and Christensen 217

TABLE 4 Comparison of Dynamic Modulus Testing Devices

Characteristic AASHTO TP62-30 Proposed Procedure

Temperature range −10°C to 60°C (14°F to 140°F) 4°C to 60°C (40°F to 140°F)
Temperature control ± 0.5°C (± 1.0°F) ± 0.5°C (± 1.0°F)
Dynamic load capacity 22.5 kN (5.0 kips) 12.5 kN (2.8 kips)
Loading rates 0.1 to 25 Hz 0.01 to 10 Hz
Estimated cost $125,000 $50,000

NCHRP Project 9-29 is capable of performing the dynamic modulus pavement design guide master curve equation is fit to these data
master curve testing described in the proposed approach. using the same numerical optimization techniques. Comparison of
The Simple Performance Test System software applies 10 cycles master curves using the reduced testing and the complete AASHTO
at each loading frequency. The first five cycles are used to adjust the TP62-03 data set for 22 mixtures yielded similar master curves except
load to produce strains in the specified range. The data from the sec- in cases in which extremely high or low moduli were measured at
ond five cycles are then collected and used to compute the dynamic 15.8°F (−9°C).
modulus and phase angle. The four loading frequencies recom- A testing machine capable of performing the testing required by
mended in the proposed approach will require approximately 30 min the proposed approach is estimated to cost $50,000 compared with
per specimen at each testing temperature, including time for specimen $125,000 for a machine to perform the testing in AASHTO TP62-03.
instrumentation and chamber temperature equilibrium. Thus, a test- When the proposed approach and the software automation already
ing program including three replicate specimens will require 1.5 h incorporated in the Simple Performance Test System are used, spec-
per temperature. Generally, specimen fabrication, including batching, imen preparation and dynamic modulus master curve testing for a
preparation of overheight gyratory samples, sawing and coring of test mixture using three replicate specimens are estimated to require a
specimens, specific gravity measurements for the final test specimens, total of 13.5 person-hours of technician time.
and gluing gauge points, requires 3.0 h per specimen. Thus, master
curve testing using three specimens and the proposed approach will
require approximately 13.5 person-hours of preparation and testing. ACKNOWLEDGMENT

The research reported here was performed as part of NCHRP Program

SUMMARY AND CONCLUSIONS Project 9-29, Simple Performance Tester for Superpave Mix Design.

This document presents an alternative to the testing sequence presented

in AASHTO TP62-03 for developing dynamic modulus master REFERENCES
curves. In this approach, the limiting maximum modulus for the mix-
ture is estimated from mixture volumetric data using the Hirsch model 1. Dougan, C., J. Stephens, J. Mahoney, and G. Hansen. E*-Dynamic Mod-
and a limiting binder shear modulus of 145,000 psi (1 GPa) (7 ). Esti- ulus Test Protocol—Problems and Solutions. Publication CT-SPR-
0003084-F03-3. Connecticut Department of Transportation and Federal
mated limiting maximum mixture modulus values using the Hirsch Highway Administration, 2003.
model range from about 3,000,000 (20.6 GPa) for mixtures with very 2. Witczak, M. W. NCHRP 9-19 Video Conference. Presented at the National
high VMA and low VFA to about 3,800,000 (26.2 GPa) for mixtures Cooperative Highway Research Program, Washington, D.C., Nov. 7, 2003.
with low VMA and high VFA. These are very rational limiting 3. ARA, Inc. ERES Consultants Division. Guide for Mechanistic–Empirical
Design of New and Rehabilitated Pavement Structures. Final Report to the
maximum modulus values considering that the modulus of portland National Cooperative Highway Research Program. NCHRP, Washington,
cement concrete is often estimated at 4,000,000 psi (27.6 GPa). D.C., March 2004.
By estimating the limiting maximum modulus, testing at extremely 4. Christensen, D. W., and D. A. Anderson. Interpretation of Dynamic
low temperatures is not needed to develop a master curve using the Mechanical Test Data for Paving Grade Asphalt Cements. Journal of the
master curve equation contained in the mechanistic–empirical pave- Association of Asphalt Paving Technologists, Vol. 61, 1992, pp. 67–116.
5. Mirza, M. W., and M. W. Witczak. Development of a Global Aging Sys-
ment design guide. Testing at the lowest temperature in the AASHTO tem for Short and Long Term Aging of Asphalt Cements. Journal of the
TP62-03 sequence, 14°F (−10°C), greatly increases the cost of the Association of Asphalt Paving Technologists, Vol. 64, 1995, pp. 393–430.
environmental chamber for the testing system and increases the com- 6. Designation: D 2493-01. Standard Viscosity–Temperature Chart for
plexity of testing. Moisture condensation and icing at temperatures Asphalts. 2003 Annual Book of ASTM Standards, Volume 04.03, ASTM
below 32°F (0°C) make testing extremely challenging even for highly International, West Conshohoken, Pa., 2003.
7. Christensen, D. W., T. K. Pellinen, and R. F. Bonaquist. Hirsch Model for
experienced technicians. Data published by Pellinen show substan- Estimating the Modulus of Asphalt Concrete. Journal of the Association
tially increased variability for testing at 15.8°F (−9°C), and some of Asphalt Paving Technologists, Vol. 72, 2003, pp. 97–121.
data exhibit unreasonably high moduli, approaching 9,000,000 psi 8. Pellinen, T. K. Investigation of the Use of Dynamic Modulus as an Indi-
(62 GPa) (8). Such high moduli also have a significant effect on the cator of Hot-Mix Asphalt Performance. Ph.D. dissertation. Arizona State
University, May 2001.
limiting minimum modulus extrapolated from the test data using the 9. Bonaquist, R. F., D. W. Christensen, and W. Stump. Simple Performance
mechanistic–empirical pavement design guide master curve equation. Tester for Superpave Mix Design: First Article Development and Evalu-
The alternative approach presented in this paper requires testing at ation. NCHRP Report 513. National Cooperative Highway Research
40°F, 70°F, and 115°F (4.4°C, 21.1°C, and 46.1°C) using frequencies Program, Washington, D.C., 2003.
of 10, 1, 0.1, and 0.01 Hz. After substituting the limiting maximum
modulus estimated from the Hirsch model, the mechanistic–empirical The Bituminous Materials Section sponsored publication of this paper.