CONVERSION OF TESTING FREQUENCY TO LOADING TIME APPLIED TO THE MECHANISTIC-EMPIRICAL PAVEMENT DESIGN GUIDE

Samer Katicha Ph.D. Candidate, The Charles Via, Jr. Department of Civil and Environmental Engineering Graduate Research Assistant, Center for Sustainable Transportation Infrastructure, VTTI 3500 Transportation Research Plaza Virginia Tech, Blacksburg, VA 24061 Tel: (540) 231-1586, fax: (540) 231-1555, email: skaticha@vt.edu Gerardo W. Flintsch, Ph.D., P.E.1 Director, Center for Sustainable Transportation Infrastructure, VTTI, Associate Professor, The Charles Via, Jr. Department of Civil and Environmental Engineering 3500 Transportation Research Plaza Virginia Tech, Blacksburg, VA 24061-0105 Phone: (540) 231-9748, fax: (540) 231-1555, email: flintsch@vt.edu Amara Loulizi, Ph.D., P.E. Assistant Professor Ecole Nationale d'ingnieur de Tunis, ENIT Dpartement Gnie Civil B.P. 37 Le Belvdre 1002 Tunis, Tunisia Phone: +(216 21) 271 701, email: amlouliz@vt.edu Linbing Wang, Ph.D., P.E. Associate Director, Center for Sustainable Transportation Infrastructure, VTTI, Associate Professor, Via Department of Civil and Environmental Engineering 3500 Transportation Research Plaza Virginia Polytechnic Institute and State University, Blacksburg, VA 24061-0105 Phone: (540) 231-5262: (540) 231-7532, email: wangl@vt.edu Keywords: hot-mix asphalt, creep compliance, dynamic modulus, load duration, lading frequency, pavement design Submission Date: August 1, 2007 Submitted for Presentation at the 2008 TRB Annual Meeting and Publication in the Transportation Research Record: Journal of the Transportation Research Board Word Count: Abstract : 180 Text : 4,220 Figures 12 x 250 = 3,000 Tables : 0 x 250 = 0 TOTAL : 7,400 1 Corresponding author

TRB 2008 Annual Meeting CD-ROM

Paper revised from original submittal.

Katicha, Flintsch, Loulizi and Wang

ABSTRACT In this paper, the unresolved issue of converting testing frequency in a dynamic modulus test of hot mix asphalt (HMA) to loading time for implementation in the proposed MechanisticEmpirical Pavement Design Guide (MEPDG) is investigated. Two methods have been proposed in the literature. The first is to convert the test frequency in Hz to loading time in seconds using t = 1/f. The second method is to convert the test angular frequency to loading time in seconds using t = 1/ = 1/2f. An exact interconversion based on the representation of dynamic modulus results using a generalized Kelvin model (GKM) and a generalized Maxwell model (GMM) is presented. The exact interconversion is compared to the two debated methods of converting the dynamic modulus or the storage modulus to stiffness (inverse of creep compliance) and relaxation modulus. It is shown that both methods result in error in determining either the relaxation modulus or stiffness. On the other hand, it is shown that the resilient modulus can be adequately approximated by the dynamic modulus taken at a frequency f = 1/t.

TRB 2008 Annual Meeting CD-ROM

Paper revised from original submittal.

Katicha, Flintsch, Loulizi and Wang

INTRODUCTION In the NCHRP 1-37A Mechanistic-Empirical Pavement Design Guide (MEPDG) (1) the dynamic modulus master curve is constructed as a function of loading time t. Since dynamic modulus results are obtained as a function of frequency, a conversion from frequency to loading time is needed. Two methods have been debated among pavement researchers. The first method suggested is to directly convert frequency to loading time using t = 1/f where f is the frequency. The second method suggested is to first convert the frequency to angular frequency = 2f and then determine the loading time t = 1/. It has been argued by Dongre and his collaborators, that the second method is the appropriate method since it is the one widely used in the field of rheology (2, 3). The issue is still not resolved among researchers and is a source of controversy. This paper addresses this issue and presents a correct method to determine the material response under any type of loading using material properties obtained from a dynamic test. The use of the term loading time is, in itself, ambiguous and confusing and can be interpreted differently among researchers and engineers. Dynamic modulus results for a typical surface mix used in the state of Virginia are used to compare the two debated conversion methods. The exact conversion is obtained by fitting a generalized Kelvin model (GKM) or generalized Maxwell model (GMM) to the dynamic modulus data. At this point it is stressed that the interconversion is an exact interconversion of the models fitted to the experimental data under the assumption that linear viscoelastic theory is applicable to hot-mix asphalt (HMA). In addition, a method to obtain the resilient modulus from the dynamic modulus data is also presented and the resilient modulus at loading time t is compared to the dynamic modulus at f = 1/t and = 1/t. All calculations are performed using an Excel spreadsheet in order that the approach presented can be implemented practically by transportation agencies. BACKGROUND The response of a linear viscoelastic material is time and loading path dependent. Therefore, specification of a given loading time should be accompanied by specifications of the loading path during this time (instantaneous loading and unloading, ramp loading and unloading, etc). It seems that most pavement engineers talk about loading time meaning one of the following three cases: 1. A constant instantaneous load/stress is applied for a given time period and then instantaneously removed as in a creep test. 2. A constant instantaneous deflection/strain is applied for a given time period and then instantaneously removed as in a stress relaxation test. 3. A haversine loading pulse is applied for a certain time period followed by a rest period as in the resilient modulus test. The relationship between frequency and loading time for bituminous materials started with the definition of stiffness for bitumen by Van der Poel (4) who indicated it can be treated as either the inverse of the creep compliance at loading time t or the dynamic modulus at an angular frequency = 1/t. The conversion was adopted by Anderson and his group in the development of the Strategic Highway Research Program (SHRP) binder specifications (5-8). The relationship is mathematically expressed as:

TRB 2008 Annual Meeting CD-ROM

Paper revised from original submittal.

Katicha, Flintsch, Loulizi and Wang

E * ( ) 1 D(t ) t 1 Where, E * ( ) = dynamic modulus D (t ) = creep compliance

(1)

The maximum error using Van der Poels approximate method was calculated to be 18% (5). Van der Poels conversion for bitumen was adopted by most researchers for hot-mix asphalt HMA (9-12). In some cases, researchers used Van Der Poels conversion ignoring the fact it is a conversion between dynamic modulus and stiffness modulus (inverse of creep compliance). For example, Brown and Gibb (11) used a loading such that the frequency of the load pulse was 0.5 Hz which results in a period of 2 sec. However, the load was applied for a period of 1 sec followed by a rest period of 1 sec. This is clearly different from the sinusoidal loading applied in the dynamic modulus test for which Van der Poel proposed his approximate conversion. Jacobs et al. (13) used a loading frequency of 8 Hz to correspond to a vehicular speed of 60 km/h and used t = 0.1/f to convert from dynamic modulus as a function of frequency to creep compliance as function of time. On the other hand, Kim and Lee (14) compared the uniaxial dynamic modulus results to IDT resilient modulus results at a frequency of 10 Hz, assuming the loading time in the resilient modulus test is inversely related to the dynamic modulus test frequency in Hertz such that t = 1/f. The same relationship was used to convert a falling weight deflectometer (FWD) loading time of 0.03 sec to a dynamic modulus test frequency of 33 Hz. Overall, the researchers found the resilient modulus values to fall between the dynamic moduli at 5 Hz and 10 Hz. To convert the dynamic modulus as a function of frequency to dynamic modulus as a function of loading time for input in the MEPDG, Bonaquist and Christensen (15), and Witzack et al. (16) suggested using a frequency of 10 Hz to represent highway speeds and therefore use the dynamic modulus result at 10 Hz. Since the loading pulse time for highway speeds is 0.1 sec, their conversion from frequency to loading time is f = 1/t. The issue of converting frequency to loading time for the MEPDG was discussed by Dongre et al. (2) who disagreed with the method proposed in (15, 16) and suggested that the correct conversion is t = 1/. They investigated the difference between using the two different methods in (3). Contrary to Van der Poels suggestion that the conversion is an approximate conversion between the inverse of the creep compliance and dynamic modulus, the authors suggest that the modulus obtained from the conversion (t=1/) is the exact relaxation modulus. VISCOELASTIC MATERIAL RESPONSE The response of a viscoelastic material is generally characterized by either a transient function or a dynamic function. The transient responses are either the relaxation modulus or the creep compliance. The dynamic response determines the dynamic complex modulus or compliance. The dynamic complex modulus (compliance) is determined by applying a sinusoidal loading (either stress or strain in principle) and measuring the corresponding steady-state response as

TRB 2008 Annual Meeting CD-ROM

Paper revised from original submittal.

Katicha, Flintsch, Loulizi and Wang

well as the phase angle between the stress and strain (strain is always lagging). The dynamic complex modulus and dynamic complex compliance are respectively defined as:
E * ( ) = * ( ) = e = cos( ) + i sin ( ) = E ( ) + iE ( ) ( ) * ( ) i = e = cos( ) i sin ( ) = D( ) iD( ) D * ( ) = * ( )
i *

(2) (3)

Where,

* ( ) = ei (t + ) * ( ) = eit
i E ( ) E ( ) D( ) D( )
= 1 = angular frequency = phase angle = stress = strain amplitude = storage modulus = the loss modulus = storage compliance, and = loss compliance.

The dynamic complex modulus and dynamic complex compliance are related through the following equations (17):

D( ) =

E ( ) [E ( )]2 + [E ( )]2 E ( ) D ( ) = [E ( )]2 + [E ( )]2

(4-a) (5-a)

E ( ) = E ( ) =

D( ) [D( )]2 + [D( )]2 D ( )

(4-b) (5-b)

[D ( )]2 + [D ( )]2

The norm of the dynamic complex modulus (compliance) is calculated as the norm of any complex number. This norm is often simply referred to as the dynamic modulus (compliance):

E * ( ) = E 2 ( ) + E 2 ( ) D * ( ) = D 2 ( ) + D 2 ( )
For the transient response, the relaxation modulus and creep compliance are defined as:

(6) (7)

E (t ) =

(t ) (t ) D(t ) =

(8)
(9)

TRB 2008 Annual Meeting CD-ROM

Paper revised from original submittal.

Katicha, Flintsch, Loulizi and Wang

Where, E (t ) = relaxation modulus (t ) = measured stress response = applied constant strain D(t ) = creep compliance (t ) = measured strain response = applied constant stress
Mechanical Analogues The time dependence of viscoelastic response can be described by ordinary differential equations in time. A convenient way of developing these relations is to employ a combination of mechanical analogues consisting of springs and dashpots. The combinations used are either the GMM or the GKM. The dynamic modulus and relaxation modulus of a GMM are expressed as:
n 2 i2 i + i Ei E ( ) = E e + Ei 2 2 1+ i 1 + 2 i2 i =1 i =1

(10) (11)

E (t ) = E e + Ei e t i
i =1

Where, Ee = equilibrium modulus Ei = spectral intensities i = relaxation times The dynamic and creep compliance of a GKM are expressed as:
D * ( ) = Dg + Di
i =1 n n

1 1+
2 2 i

+ i Di
i =1

i 1 + 2 i2

(12)
(13)

D(t ) = Dg + Di 1 e t i
i =1

Where, Dg = initial compliance

Di = spectral intensities i = retardation times

Equations 11 and 13 are referred to as a Dirichlet-Prony series. Note that the dynamic modulus can be expressed in terms of the GKM using equations 4 and 5. Similarly, the dynamic compliance can be expressed in terms of the GMM. The advantage of using the GKM or GMM is that once the model parameters are determined, both transient and dynamic responses can be exactly obtained.

TRB 2008 Annual Meeting CD-ROM

Paper revised from original submittal.

Katicha, Flintsch, Loulizi and Wang

COMPARISON OF EXACT AND APPROXIMATE INTERCONVERSION METHODS

The time temperature superposition principle was used to construct a dynamic modulus master curve for a typical surface mix used in the state of Virginia. Data were obtained by performing dynamic modulus tests on three replicates at temperatures of -12.5, 4.4, 21.1, 37.8, and 54.4C and frequencies 0.1, 0.5, 1, 10, and 25 Hz. To obtain the relaxation modulus as well as the creep compliance, a GMM and a GKM were fitted to the dynamic modulus master curve as shown in Figure 1. The number of parameters used to determine each model was 12 with the relaxation times and retardation times being equally spaced on a log scale. The fit was performed using an Excel spreadsheet and the Solver. The agreement between both models and the experimental data is excellent. The exact creep compliance is directly obtained from the GKM while the exact relaxation modulus is directly obtained from the GMM.
100 fitted GKM measured E* Dynamic modulus (GPa) fitted GMM 10

0.1 0.00001 0.001 0.1 10 1000 Reduced Frequency (Hz) 100000 10000000

FIGURE 1 Fit of GMM and GKM to dynamic modulus test results. Relaxation Modulus The exact relaxation modulus at a given loading time is obtained from Equation 11. The dynamic modulus at = 1 t and = 2 t ( f = 1 t , = 2f ) is calculated using equations 6 and 10. The percent difference between the exact relaxation modulus and the relaxation modulus determined by assuming E (t ) = E * ( ) or E (t ) = E * ( ) is presented in Figure 2. Both
=1 t = 2 t

methods produce considerable error. The error from using E (t ) = E * ( ) error from using E (t ) = E * ( ) times of around 1 sec.
= 2 t

=1 t

is less than the

however, this error can still be as much as 50% for loading

TRB 2008 Annual Meeting CD-ROM

Paper revised from original submittal.

Katicha, Flintsch, Loulizi and Wang

200

150

E * H w = 1 tL - E H t L E H tL

% Error

E * H w = 2 p tL - E H tL E H tL

100

50

0
-4 -2

FIGURE 2 Error in determining the relaxation modulus from the dynamic modulus using t = 1/f and t = 1/.

0 Log t HsecL

When the number of parameters in the GMM, is infinitely increased, the summation becomes an integral. In this case, the relaxation modulus and storage modulus are expressed as in equations 14 and 15 and an exact relationship between the storage modulus and relaxation modulus can be obtained as in Equation 16 using equations 14 and 15 as suggested by Ferry (18):
E (t ) = Ee + H ( )e t d ln

(14) (15) (16)

E ( ) = H ( )

2 2 d ln 1 + 2 2 2 t E ( )1 = t E (t ) = H ( ) 2 t + 2 e d ln

Where H ( ) = continuous spectrum of relaxation times. It should be noted that this is not the only relationship that can be written between the storage modulus and the relaxation modulus. For example, using equations 15 and 16 another exact relationship could have been written as:
t 2 2 E ( ) = t E (t ) = H ( ) e t d ln 2 2 1+ t

(17)

Or

TRB 2008 Annual Meeting CD-ROM

Paper revised from original submittal.

Katicha, Flintsch, Loulizi and Wang

4 2 2 t E ( )1 / f = t E (t ) = H ( ) 2 t + 4 2 2 e d ln

(18)

The reason for using Equation 16 is that the difference E ( )1 = t E (t ) is relatively small and, therefore, the integral represents a minor correction which need not be calculated with great precision (18). It seems that in the conversion from frequency to loading time, HMA researchers directly assume the integral to be zero and often replace the storage modulus by the dynamic modulus (3). The effect of considering this integral to be zero and therefore assuming E ( )1 = t = E (t ) is compared to the assumption that E * ( ) = E (t ) in Figure 3. The
1 =t

approximation E ( )1 = t = E (t ) is much better than E ( )

*

1 =t

= E (t ) , however the error can still

be greater than 20%.

50 E* H w = 1 t L - E H t L E H tL E' H w = 1 t L - E H t L E H tL

40

% Error

30

20

10

-4

-2

FIGURE 3 Error in determining the relaxation modulus from the dynamic modulus and the storage modulus using t = 1/.

0 Log t HsecL

In all this analysis it was assumed that the HMA is subjected to a constant strain that is applied instantaneously. If HMA is subjected to a constant stress that is applied instantaneously, it will undergo creep and, therefore, the creep compliance is more appropriate to analyze the stress-strain relationship.
Creep Compliance In this paper, under a creep type of loading (constant stress) the stiffness M (t ) is defined as:

TRB 2008 Annual Meeting CD-ROM

Paper revised from original submittal.

Katicha, Flintsch, Loulizi and Wang

10

M (t ) =

1 = D(t ) (t )

M (t ) E (t )

(19) are presented in Figure 4.

The results of using M (t ) = E * ( ) For both cases

=1 t

or M (t ) = E * ( )
= 2 t

= 2 t

M (t ) = E * ( )
or E (t ) = E * ( )

=1 t = 2 t

or M (t ) = E * ( )

the error is less than using

=1 t

E (t ) = E * ( )

=1 t

. Also the error in the approximation M (t ) = E * ( )

*

is much smaller than the error in the approximation M (t ) = E ( )

= 2 t

. It should be noted that

this approach is also incorrect; however, it produces much less error and, therefore, the approximation is better under creep loading. = M (t ) , In Figure 5, the error from using E ( )1 =t = M (t ) is compared with E * ( )
1 =t

which is the approximation suggested by Van der Poel. The minimum maximum absolute error (~14%) results from the approximation E ( )1 =t = M (t ) . It is also noted that the maximum error from using E * ( )
1 =t

= M (t ) is 18% as reported in (4).

140 120 100

E * H w = 1 tL - M H tL M H tL

% Error

80 60 40 20 0
-4

E * H w = 2 p tL - M H tL M Ht L

-2

FIGURE 4 Error in determining the creep compliance from the dynamic modulus using t = 1/f and t = 1/.

0 Log t HsecL

TRB 2008 Annual Meeting CD-ROM

Paper revised from original submittal.

Katicha, Flintsch, Loulizi and Wang

11

15

10

% Error

0 E* H w = 1 tL - M H tL M H tL E' H w = 1 tL - M H tL M H tL
-2

-5

- 10

-4

FIGURE 5 Error in determining the creep compliance from the dynamic modulus and the storage modulus using t = 1/.

0 Log t HsecL

Both the creep compliance and relaxation modulus are transient responses. In real pavements, a more realistic load application is the haversine load used in determining the resilient modulus of HMA. Therefore, in the following section, the resilient modulus response of HMA is determined using the dynamic modulus master curve.
VISCOELASTIC MATERIAL RESPONSE UNDER DIFFERENT LOADINGS

For a linear viscoelastic material, the strain resulting from an applied stress can be determined using the Boltzmann superposition principle:

(t ) = D(t )
0

d d d

(20)

The stress resulting from an applied strain can be determined from:

(t ) = E (t )
0

d d d

(21)

When the relaxation modulus and creep compliance are expressed as in equations 11 and 13, the integrals in equations 20 and 21 can often be analytically evaluated for various loadings.

TRB 2008 Annual Meeting CD-ROM

Paper revised from original submittal.

Katicha, Flintsch, Loulizi and Wang

12

Another approach which is used in this paper is to perform the integration numerically using finite differences. The purpose of using this numerical integration scheme is that it is very simple and can be implemented in a spreadsheet which makes it very attractive for a transportation agency. Moreover, the numerical integration gives very accurate results if the time step is appropriately chosen. A recursive relationship for a Kelvin element can be obtained as:

i +1 = i e

+ D i 1 e

(22)

For the GKM, the total response is determined by adding the responses of the individual elements. The accuracy of Equation 22 depends on the time step used for the analysis. This accuracy is analyzed by determining the dynamic modulus using the strain calculated using Equation 20 for a sinusoidal applied stress. The calculation is performed at frequencies of 0.0001, 0.001, 0.01, 0.1, 1, 10, 100, 1000, 10000, 100000, and 1000000 Hz. The time interval was chosen so as to generate 100 points per cycle. The calculated dynamic modulus is compared to the exact dynamic modulus in Figure 6. The agreement between the exact method and the numerical method using Equation 22 is excellent with the maximum error calculated as 0.56%. Equation 20 is therefore used to determine the modulus of HMA under different types of loading (e.g. the resilient modulus).

100 dynamic modulus from finite difference Dynamic Modulus (GPa) exact dynamic modulus 10

0.1 0.0001

0.001

0.01

0.1

10 Frequency (Hz)

100

1000

10000

100000 1000000

FIGURE 6 Comparison between calculated dynamic modulus using finite differences and exact dynamic modulus.

TRB 2008 Annual Meeting CD-ROM

Paper revised from original submittal.

Katicha, Flintsch, Loulizi and Wang

13

Resilient Modulus Calculation from Dynamic Modulus Results

Resilient Modulus Test (ASTM-D4123) The first step is to determine, using dynamic modulus results, the resilient modulus obtained from a 0.1 sec load pulse duration followed by a 0.9 sec rest period. The applied load is a haversine shown in Figure 7. The resilient modulus is calculated as the maximum applied stress divided by the recovered strain. The strain is measured immediately after the load is removed and at 0.05 sec intervals until the next load application. The resulting strain from the applied load is shown in Figure 8 (3 cycles shown).
1.2 Haversine 1 Applied stress (GPa) 0.8 0.6 0.4 0.2 0 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 Loading Time (sec) Sine Haversine squared Sine^0.5 Haversine^4

FIGURE 7 Different load pulses used to simulate the resilient modulus.

0.2 0.18 0.16 0.14 0.12 Strain 0.1 0.08 0.06 0.04 0.02 0 0 0.5 1 1.5 Time (sec) 2 2.5 3

FIGURE 8 Strain in a resilient modulus test.

TRB 2008 Annual Meeting CD-ROM

Paper revised from original submittal.

Katicha, Flintsch, Loulizi and Wang

14

The calculated resilient modulus at different recovery times is shown in Figure 9. As expected, the calculated resilient modulus decreases as the material is allowed to recover. The resilient modulus after 0.05 sec recovery is measured at 7.89 GPa. The dynamic modulus measured at a frequency of 10 Hz (f=1/t, for 0.1 sec loading time) is 7.90 GPa. The dynamic modulus measured at an angular frequency of 10 rad/sec (=1/t, for 0.1 sec loading time) is 5.55 GPa. Using the dynamic modulus at f=1/t resulted in a very good estimate of the actual resilient modulus calculated after 0.05 sec recovery time (error = 0.16%). Using the dynamic modulus at =1/t resulted in a less accurate estimate (error = 29.62%). The absolute value of the error in using the dynamic modulus as the resilient modulus for the different recovery times is presented in Figure 10. In all cases, using f = 1/t resulted in a better estimate of the resilient modulus.
9

8.5 Resilient Modulus (GPa)

7.5

6.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Recovery Time (sec)

FIGURE 9 Resilient modulus at different recovery times for a 0.1 sec haversine loading pulse.
45 40 35 Absolute Error (%) 30 25 20 15 10 5 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Recovery Time (sec) f=1/t w=1/t

FIGURE 10 Absolute errors using the different approximations to the resilient modulus.

TRB 2008 Annual Meeting CD-ROM

Paper revised from original submittal.

Katicha, Flintsch, Loulizi and Wang

15

Effect of Load Pulse Shape In this section, the resilient modulus determined from the different load pulses shown in Figure 7 is determined. The resilient modulus determined at different recovery times along with the dynamic modulus at f = 10 Hz and = 10 rad/sec are presented in Figure 11. In all loading cases, the dynamic modulus measured at f = 1/t is a better estimate of the resilient modulus measured at 0.05 sec recovery time. Although the loading time for all pulses shown in Figure 11 is 0.1 sec, the pulses that are obtained by squaring or raising the haversine pulse to the fourth power can be well approximated by a haversine pulse of duration 0.08 sec and 0.06 sec, respectively. For a haversine load duration of 0.06 sec, the dynamic modulus at a frequency f = 1/0.06 = 16.7 Hz is 8.4 GPa which agrees very well with the resilient modulus obtained after 0.05 sec recovery. This agreement between the resilient modulus and the dynamic modulus at f = 1/t is expected since the haversine load pulse is the sum of a sinusoidal load pulse with a constant load equal to half the sinusoidal load pulse amplitude.
11 10 Resilient Modulus (GPa) 9 8 7 6 5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Recovery Time (sec) Haversine Sine Haversine squared Sine^0.5 Haversine^4 E* at f=1/t E* at w=1/t

FIGURE 11 Comparison between dynamic modulus at 10 Hz and resilient modulus determined using different loading pulses.

Effect of Loading Time In this section, various loading haversine pulses are investigated for approximating the resilient modulus using the dynamic modulus results. In all cases the total recovery period is 9 times the loading time. Since the dynamic modulus and resilient modulus are different for each frequency/loading time, the results are normalized with respect to the dynamic modulus at f = 1/t; for example, for a loading pulse of 10 sec, the results are normalized with respect to the dynamic modulus at f = 0.1 Hz. Also, the recovery time is normalized with respect to the loading time for each test. The results are shown in Figure 12. For loading times of 0.1 and 0.01 sec, the dynamic modulus at 10 and 100 Hz is a very good approximation to the resilient modulus calculated after 0.5 normalized recovery. For the other loading time cases, the agreement between the resilient modulus and the dynamic modulus occurs at normalized

TRB 2008 Annual Meeting CD-ROM

Paper revised from original submittal.

Katicha, Flintsch, Loulizi and Wang

16

recovery times between 0.25 and 0.75. As a general estimate, the resilient modulus calculated after 0.5 sec recovery is well approximated by the dynamic modulus determined at f=1/t.
1.4 0.1 sec 0.01 sec Normalized Resilient Modulus 1 sec 10 sec 100 sec 1000 sec 1

0.6 0 1 2 3 4 5 6 7 8 9 Normalized Recovery Time

FIGURE 12 Comparison between dynamic modulus at =1/t and resilient modulus determined using different haversine loading pulses. FINDINGS

The findings of this paper are summarized below:

Both methods that are debated to convert frequency to loading time induce errors in estimating the relaxation modulus or creep compliance. In all cases, the approximation between the dynamic modulus and the transient responses using t = 1/ instead of t = 1/f results in significantly less error. However, a better approximation was found between the storage modulus and the transient responses using t = 1/. The maximum error in using Van der Poels approximation in interconversion between the stiffness and dynamic modulus was calculated to be 18% which agrees with the result of Christensen and Anderson (5). Once the results of dynamic testing are expressed in terms of the GKM and GMM, the exact determination of the relaxation modulus and creep compliance can be done directly using the theoretical relationships. Specification of loading time should always be accompanied with specifications about the loading path. The response of a viscoelastic material can be obtained for any type of loading using the convolution integral. When the transient responses are expressed in terms of a Dirichlet-

TRB 2008 Annual Meeting CD-ROM

Paper revised from original submittal.

Katicha, Flintsch, Loulizi and Wang

17

Prony series, the integral is greatly simplified. An alternative to analytically evaluating the convolution integral is to use the recursive relationship of Equation 22. A very good agreement was found between the calculated dynamic modulus using the recursive relationship and the exact dynamic modulus.

The dynamic modulus at f = 1/t is a good estimate of the resilient modulus for various load pulses and loading times t which agrees with the results in Kim and Lee (14). For most cases, the agreement between the dynamic modulus at f = 1/t and the resilient modulus occurs for resilient moduli calculated after a recovery period of 0.25 to 0.75 times the loading time. It should be kept in mind that taking the resilient modulus to be equal to the dynamic modulus at f = 1/t is not an exact interconversion. The dynamic modulus at = 1/t is a less accurate estimate of the resilient modulus and therefore is not recommended to be used as an estimate of Youngs modulus in an elastic analysis.

CONCLUSION

In this paper, the controversial issue of converting frequency to loading time is discussed. For this purpose, exact interconversion methods between the different viscoelastic functions are used. The presented approach to perform the interconversion and calculate the material response under any type of loading is simple and practical in order that it can be implemented by different transportation agencies. For this purpose all calculations are performed using an Excel spreadsheet. These involve the following:

Determining the GMM and GKM parameters Determining the relaxation modulus and creep compliance (stiffness) Determining the material response under different types of loading, mainly the resilient modulus

Determining the relaxation modulus or stiffness as the dynamic modulus at t = 1/f or t = 1/ results in considerable errors and is fundamentally wrong. On the other hand, the resilient modulus resulting from a haversine load application calculated from the resilient strain measured after a recovery period equal to half the loading duration is reasonably estimated by the dynamic modulus measured at f = 1/t. This was verified for different load pulse durations ranging from 0.01 sec to 1000 sec. This is, however, an estimate and not an exact relationship between the dynamic modulus and the resilient modulus.
REFERENCES 1. National Cooperative Highway Research Program. Guide for Mechanistic-Empirical Design of New and Rehabilitated Pavement Structures. Final Report. NCHRP 1-37A. TRB, National Research Council, Washington, D.C., 2004.

2. Dongre, T., L. Myers, J. DAngelo, C. Paugh, and J. Gudimettla. Field Evaluation of Witczak and Hirsch Models for Predicting Dynamic Modulus of Hot-Mix Asphalt. Journal of the Association of Asphalt Paving Technologists. Vol.73, 2005, pp. 381-442.

TRB 2008 Annual Meeting CD-ROM

Paper revised from original submittal.

Katicha, Flintsch, Loulizi and Wang

18

3. Dongre, R., L. Myers, and J. DAngelo. Conversion of Testing Frequency to Loading Time: Impact on Performance Predictions Obtained from the M-E Pavement Design Guide. Proceedings of the Transportation Research Board. (2006) CD-ROM. 4. Van der Poel, C. A General System Describing the Viscoelastic Properties of Bitumens and its Relation to Routine Test Data. Journal of Applied Chemistry. Vol. 4, 1954, p. 221. 5. Christensen, D.W., and D.A. Anderson. Interpretation of Dynamic Mechanical Test Data for Paving Grade Asphalt Cement. Journal of the Association of Asphalt Paving Technologists. Vol. 61, 1992, pp. 67-116. 6. Anderson, D.A., D.W. Christensen, and H. Bahia. Physical Properties of Asphalt Cement and the Development of Performance-Related Specifications. Journal of the Association of Asphalt Paving Technologists. Vol. 60, 1991, pp. 437-475. 7. Anderson, D.A., and T.W. Kennedy. Development of SHRP Binder Specification. Journal of the Association of Asphalt Paving Technologists. Vol. 62, 1993, pp. 481-507. 8. Marasteanu, M., and D.A. Anderson. Time-Temperature Dependency of Asphalt Bindersan Improved Model. Journal of the Association of Asphalt Paving Technologists. Vol. 65, 1996, pp. 408-435. 9. Maccarone, S., G. Holleran, and G.P. Gnanaseelan. Properties of Polymer Modified Binders and Relationship to Mix and Pavement Performance. Journal of the Association of Asphalt Paving Technologists. Vol. 64, 1995, pp. 209-231. 10. Buttlar, W.G., and R. Roque. Evaluation of Empirical and Theoretical Models to Determine Asphalt Mixture Stiffnesses at Low Temperatures. Journal of the Association of Asphalt Paving Technologists. Vol. 65, 1996, pp. 99-130. 11. Brown, S.F., and M.J. Gibb. Validation Experiments for Permanent Deformation Testing of Bituminous Mixtures. Journal of the Association of Asphalt Paving Technologists. Vol. 65, 1996, pp. 255-289. 12. Deacon, J.A., J.T. Harvey, A. Tayebli, and C.L. Monismith. Influence of Binder Loss Modulus on the Fatigue Performance of Asphalt Concrete Pavements. Journal of the Association of Asphalt Paving Technologists. Vol. 66, 1997, pp. 633-665. 13. Jacobs, M.M.J., P.C. Hopman, and A.A.A. Molenaar. Application of Fracture Mechanics Principles to Analyze Cracking in Asphalt Concrete. Journal of the Association of Asphalt Paving Technologists. Vol. 65, 1996, pp. 1-28. 14. Kim, Y.R., and Y.C. Lee. Interrelationships among Stiffnesses of Asphalt-Aggregate Mixtures. Journal of the Association of Asphalt Paving Technologists. Vol. 64, 1995, pp. 575-600. 15. Bonaquist, R.F., and D.W. Christensen. Simple Performance Tester for Superpave Mix Design: First Article Development and Evaluation. NCHRP Report 513. TRB, National Research Council, Washington, D.C. 2003. 16. Witczak, M.W., K. Kaloush, T. Peillinen, M. El-Basyouny, and H. Von Quintus. Simple Performance Test for Superpave Mix Design. NCHRP Report 465. TRB, National Research Council, Washington, D.C. 2002.

TRB 2008 Annual Meeting CD-ROM

Paper revised from original submittal.

Katicha, Flintsch, Loulizi and Wang

19

17. Mead, D.W. Numerical Interconversion of Linear Viscoelastic Material Functions. Journal of Rheology. Vo. 38, 1994, pp. 1769-1795. 18. Ferry, J.D. Viscoelastic Properties of Polymers. 3rd edition. Wiley, New York, 1980.

TRB 2008 Annual Meeting CD-ROM

Paper revised from original submittal.