You are on page 1of 15

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/245563347

Accuracy of Current Complex Modulus Selection Procedure from Vehicular Load


Pulse: NCHRP Project 1-37A Mechanistic-Empirical Pavement Design Guide

Article  in  Transportation Research Record Journal of the Transportation Research Board · December 2008
DOI: 10.3141/2087-09

CITATIONS READS

51 665

8 authors, including:

Imad L. Al-Qadi Mostafa A. Elseifi


University of Illinois, Urbana-Champaign Louisiana State University
452 PUBLICATIONS   10,868 CITATIONS    163 PUBLICATIONS   3,128 CITATIONS   

SEE PROFILE SEE PROFILE

Jun P. Yoo Samer Dessouky


Korea Institute of Civil Engineering and Building Technology University of Texas at San Antonio
53 PUBLICATIONS   1,121 CITATIONS    113 PUBLICATIONS   2,034 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Simplified Approach for Structural Evaluation of Flexible Pavements at the Network Level View project

Special Issue in SENSORS (ISSN 1424-8220) "Sensing Advancement and Health Monitoring of Transport Structures" View project

All content following this page was uploaded by Mostafa A. Elseifi on 24 June 2015.

The user has requested enhancement of the downloaded file.


Accuracy of Current Complex Modulus Selection
Procedure from Vehicular Load Pulse in NCHRP 1-
37A Mechanistic-Empirical Pavement Design Guide
Imad L. Al-Qadi
Founder Professor of Engineering
Illinois Center for Transportation, Director
Department of Civil and Environmental Engineering
University of Illinois at Urbana-Champaign
205 N. Mathews Ave., MC-250, Urbana, IL 61801
Tel: (217) 265-0427
e-mail: alqadi@uiuc.edu

Mostafa A. Elseifi
Visiting Assistant Professor
Department of Civil and Environmental Engineering
University of Illinois at Urbana-Champaign
e-mail: elseifi@uiuc.edu

Pyeong Jun Yoo


Graduate Research Assistant
Department of Civil and Environmental Engineering
University of Illinois at Urbana-Champaign
e-mail: pyoo@uiuc.edu

Samer H. Dessouky
Research Scientist
Department of Civil and Environmental Engineering
University of Illinois at Urbana-Champaign
e-mail: dsamer@uiuc.edu

Nelson Gibson
Highway Research Engineer
Federal Highway Administration
e-mail: nelson.gibson@fhwa.dot.gov

Tom Harman
Team Leader
Pavement & Materials Technical Service Team
Federal Highway Administration
e-mail: Tom.Harman@dot.gov

John D’Angelo, P. E.
Asphalt Team Leader
Federal Highway Administration
Office of Pavement Technology
e-mail: John.D’Angelo@dot.gov

Katherine Petros
Pavement Design & Performance Modeling
Team Leader
Federal Highway Administration
e-mail: katherine.petros@fhwa.dot.gov
ABSTRACT
The Mechanistic-Empirical Pavement Design Guide (MEPDG) makes use of the complex modulus to simulate the
time and temperature dependency of Hot-Mix Asphalt (HMA). To account for the time dependency of HMA, the
Guide recommends calculating the frequency of the applied load as a function of the vehicle speed and the pavement
structure. In this approach, the Odemark method of thickness equivalency is first used to transform the pavement
structure into a single layer system and it is then assumed that the stress distribution occurs at a constant slope of 45°
in the equivalent pavement structure. Concerns were raised that the current MEPDG methodology may be over-
estimating the frequency which would result in under-conservative distress predictions. Therefore, to evaluate the
MEPDG methodology for loading time calculation, results of the MEPDG procedure were compared to an advanced
three-dimensional (3D) Finite Element (FE) approach, which simulates the approaching-leaving rolling wheel at a
specific speed. The developed model accurately simulates actual tire rib sizes and the applicable contact pressure
for each rib. In addition, laboratory-measured viscoelastic properties were incorporated into the FE model to
describe the constitutive behavior of HMA. Comparison between these two methods shows that the calculated
frequencies based on the MEPDG procedure are greater than the ones determined from the 3D FE method, which
indicates that the loading time determined from the MEPDG is not conservative. Ultimately, this would result in
under-estimation of the pavement response to load and therefore greater errors in calibrations of the pavement
response to field distress. To address this issue, this paper presents correction factors to ensure the correctness of
loading time calculation in the MEPDG. Adoption of the proposed factors within the MEPDG software does
necessitate a recalibration of the performance models.

Keyword: MEPDG, complex modulus, HMA, finite element

Word counts: 4670 + 10 Figures and tables


Al-Qadi, Elseifi, Yoo, Dessouky, Gibson, Harman, D’Angelo, and Petros 1

INTRODUCTION
The National Cooperative Highway Research Program Project 1-37A, Mechanistic-Empirical Pavement Design
Guide (MEPDG), introduces concepts and methodologies to analyze pavement performance that address
shortcomings in previously employed empirical models (1). In contrast to existing versions of the AASHTO
pavement design guide, which relied heavily on the results of AASHO Road Test conducted in Ottawa, Illinois in
the early 1960s, the new Guide is based on a mechanistic-empirical design procedure. Although this method still
requires local calibration based on field and/or laboratory measurements, major advantages are expected such as the
improved level of design reliability, the understanding of and ability to predict various pavement distresses, and
better allocation of available funds based on the predicted level of performance.
A major milestone of the MEPDG is that hot-mix asphalt (HMA) is no longer described as a purely linear
elastic material. Instead, both the effects of temperature and rate of loading on the behavior of HMA are considered
through the complex modulus E*. The enhanced integrated climatic model (EICM) was adopted to predict the
temperature distribution in the HMA layers. A simplified procedure based on the Odemark and 45o influence zone
method was also adopted to calculate the frequency of the applied load as a function of the vehicle speed and the
cross-section of the pavement structure. Through the use of the E* mastercurve, the MEPDG software then
internally selects the applicable complex modulus depending on the pertinent temperature and loading frequency.
This approach is a significant improvement compared to the classical elastic method; however, concerns have been
raised about the accuracy of the frequency calculation methodology. These concerns are worthy of study because
using inaccurate frequency would result in an unreliable prediction of the complex modulus and may impact the
prediction of pavement distress.
An independent review of the MEPDG indicated the selected complex moduli E* are inaccurate and too
stiff; and therefore over estimate the influence of the HMA layer in the modeled response used in performance
prediction (2). The current frequency calculation procedure was also reported to override the effect of temperature
as compared to other mechanistic-empirical design procedures. This was thought to be related to two sources of
error. The first source of error is the Odemark approach adopted in the MEPDG and its inability to account for the
far-field effect of an approaching-leaving rolling wheel. The second source of error is thought to be related to the
conversion from the time domain to the frequency domain. To quantify the level of inaccuracy in the MEPDG from
the first source of error, the presented analysis utilized an advanced three-dimensional Finite Element (FE) approach
to simulate the approaching-leaving rolling wheel at a specific speed. The inaccuracy associated with the
conversion from the time domain to the frequency domain is presented briefly in this paper and is covered in greater
depth in another paper (3).

BACKGROUND
The MEPDG Loading Time Calculation
As presented in NCHRP 1-37A and currently employed in Version 1.0 of the MEPDG software (1), the frequency of
loading in HMA due to a moving load is determined based on the following equation (see Figure 1):
L eff
t (1)
17.6 v s

where,
t = time of loading (sec);
Leff = effective length (in); and
vs = vehicle speed (mph).
Al-Qadi, Elseifi, Yoo, Dessouky, Gibson, Harman, D’Angelo, and Petros 2

FIGURE 1 Schematic of load pulse frequency determination by MEPDG.

Using Equation (1), the effective length can be related to the duration of the vertical stress pulse at any depth within
the pavement system. To calculate the effective length at a specified depth, the MEPDG uses the Odemark’s
method of thickness equivalency to transform the pavement structure into a single subgrade layer system. In
addition, the stress distribution for a given subgrade soil is assumed to be at 45° as shown in Figure 1. For the
transformed section, the depth of frequency calculation changes and is calculated as follows:

n 1  
Ei   hn 3 En
Zeff    h
i 1
i3
ESG 
 ESG
(2)

where,
Zeff = effective depth;
hn = thickness of the layer of interest (layer n);
ESG = subgrade modulus of elasticity; and
En = modulus of elasticity of the layer of interest.

To determine the effective length from the effective depth Zeff, three cases are considered depending on the level of
overlap (if any) between adjacent tires. No overlap occurs between adjacent tires at shallow depth and in the case of
a single tire. More details about the calculation of the effective depth have been presented elsewhere (1). This
methodology assumes the load is distributed uniformly and over an equivalent circular area. However in reality, the
tire-pavement contact area is not circular. With regular tires, the actual contact area assumes a generally rectangular
shape with a constant ratio between the width and the length (4). Previous research results have shown that this
assumption of circular loading and equivalency of the tire pressure to the tire-surface contact stress leads to
erroneous prediction of pavement responses (5; 6).
Conversion from the Time Domain to the Frequency Domain
In the MEPDG, the conversion from the time (t) domain to the frequency (f) domain in Hz is based on the following
equation:
1 (3)
f
t

Recent research has questioned the ability of Equation (3) to accurately determine the cyclic frequency in the
complex modulus test (7). With the exception of the MEPDG, Dongre and co-workers have not found a reference
that supports the direct conversion of the loading time to frequency in Hz. An alternative approach widely used in
the field of rheology is to convert the loading time to cyclic frequency in Hz using the angular frequency in rad/s as
follows:
Al-Qadi, Elseifi, Yoo, Dessouky, Gibson, Harman, D’Angelo, and Petros 3

1 (4)
f
2t

It is the authors’ opinion that the conversion from the time domain to the frequency domain should be based on a
detailed Fourier series analysis, which has not been conducted to date. The loading pulse duration in the field is in
the time domain, while the complex modulus test is in the frequency domain. As the viscous behavior of HMA is
dictated by the phase angle (=2ft, f is in cycles/sec and t is the relaxation time lag between input stress and
output strain), which is related to the loss component of the HMA complex modulus in the angular frequency
domain (). Therefore, the inter-conversion between the dynamic modulus and time-dependent relaxation modulus
should be based on the relationship between angular frequency and time domain using an Inverse Fast Fourier
Transform (IFFT). This paper focuses on evaluating the accuracy of loading time calculation as presented by
Equation (1); the correct approach of converting the loading pulse into the frequency domain using an IFFT is
presented in another paper (3).
Need for Frequency Adjusting Factor
The difference between the calculated loading time and the “correct” values could be affected by three major
factors. First, the elastic solution does not consider the time-dependent response of HMA, which is significant;
especially in the unloading phase of the response. Delayed response of HMA may increase the loading time on the
material, and therefore, affect the actual loading time. Second, differences between modified and straight asphalt
binder, stiff and soft binders, may not be accurately captured by the methodology currently adopted in the MEPDG.
Third, the Odemark equivalency of thickness was used frequently in the past prior to the development of the multi-
layer elastic theory. The applicability of this approach to current pavement systems is uncertain and may require
validation. This investigation is especially needed when dealing with special purpose HMA such as stone-matrix
asphalt and SuperPaveTM mixes.
To address the aforementioned shortcomings without completely altering the recently introduced MEPDG
approach, an adjustment factor may be used to correct the calculated frequency. The presented analysis is a first
step in providing a methodology that may be used to determine the necessary correction for the MEPDG frequency
calculation. The presented analysis was based on a three-dimensional (3D) FE approach making use of viscoelastic
characterization of HMA, simulation of the moving wheel load, measurements of rib contact stresses, and
considering the effect of surface lateral stresses.
FINITE ELEMENT MODEL DESCRIPTION
FE Model Geometry
A previously validated 3D FE approach was used to simulate pavement responses to a moving load (5). The 3D FE
model input file includes the geometric properties (layer thicknesses), material properties, loading conditions in the
form of a moving tire imprint at a specific speed, contact conditions between various pavement layers, and boundary
conditions. Table 1 summarizes the FE cases that were considered in this study. As shown in this table, two
pavement structures were simulated representing an Interstate pavement design and a low volume road. In addition,
two binder types (stiff and soft) were evaluated at two temperatures (10° and 46°C).

TABLE 1 Summary of Simulation Conditions


Case Pavement HMA Pavement Target Binder
No. Temperature Thickness Thickness Speed
(C) (mm) (mm) (km/h)
1 46 152 660 13 Soft
2 46 304 813 13 Soft
3 10 152 660 80 Soft
4 10 304 813 80 Soft
5 46 152 660 13 Stiff
6 46 304 813 13 Stiff
7 10 152 660 80 Stiff
8 10 304 813 80 Stiff
Al-Qadi, Elseifi, Yoo, Dessouky, Gibson, Harman, D’Angelo, and Petros 4

The commercial software ABAQUS version 6.5 was used for FE modeling of the pavement structure (8). Figure 2
illustrates the general layout of the developed FE model. The in-plane dimensions of the modeled portion were
2,200mm x 2,850mm. An extended modeled portion was selected in the longitudinal direction to capture pavement
responses in the far field region, which is associated with the approaching-leaving wheel. The generated mesh was
designed to give optimal accuracy (fine mesh around the load and coarse mesh far from it). Eight-node linear brick
reduced-integration elements (C3D8R) were used in the simulation. In-plane dimension was selected between
16.0mm and 18.0mm in the X direction depending on the rib geometry, and 20mm in the Y direction. The
dimensions of the element thickness varied between 9.53mm for the upper HMA layers, between 15 and 50mm for
the bottom granular, and 200-500 mm for the subgrade while maintaining the aspect ratio for all elements, which is
the ratio of the shortest to longest edge of an element, greater than 0.04. These dimensions were selected based on
the results of a detailed sensitivity analysis reported elsewhere (5).

FIGURE 2 General layout of the 3D FE model.


Boundary and Contact Modeling
The proper choice of the lateral boundary conditions significantly impacts the model response. A pavement
structure is defined in unbounded domains—e.g., in the horizontal and vertical directions to some extent, if a
bedrock layer is far enough to be considered—where the region of interest is small compared with the surrounding
medium. Based on previous work conducted by Al-Qadi et al., infinite elements (CIN3D8, 8-node 3D linear infinite
element) were used to simulate the far-field region on the four lateral sides and at the bottom of the FE model (9).
Contact between the layers was assumed to be of a friction type. This assumption was previously shown to
reasonably represent real field conditions (5).
Loading Area
The exact footprint shape and dimensions of the dual-tire assembly were available for this research at the exact load
level and tire pressure considered in the simulation; see Figure 3a (10). The modeled loading area accurately
simulated the actual contact between the pavement and the tire under consideration. Figure 3b illustrates the
modeled contact area for the dual-tire assembly. The exact contact stress conditions on each tire rib were also
incorporated in the developed FE models. The measured maximum vertical contact stress on each tire rib is shown
in Figure 3b. Even though the nominal tire pressure was 724kPa, the measured contact stresses at the pavement
surface were significantly greater; the applied load was 36.7kN and the center-to-center spacing of the dual tires was
338mm. Assuming the nominal tire pressure is equal to the actual contact stress is a common source of error in the
multi-layer elastic theory, which may impact the accuracy of the calculations. This assumption is made in the
MEPDG.
It is also evident from field-measured contact stresses that the moving wheel load induces transverse and
longitudinal shear stresses on the pavement surface as well as vertical compressive stresses. These tangential
stresses are the results of the inward pressure of tires and the resistance to movement by surface contact. As a result,
these stresses may have significant impacts on pavement responses especially at the surface and at shallow depths
(11). Measurements of lateral contact stresses indicate the maximum longitudinal and transverse shear stresses are
about 10% to 15% and 30% to 52% of the vertical stress, respectively (Figure 3c). These values depend on the
applied axle load and may not be accurately simulated if the elastic layered theory is used.
Al-Qadi, Elseifi, Yoo, Dessouky, Gibson, Harman, D’Angelo, and Petros 5

(a)
Rib 2: 1,030 kPa 122 mm

Rib 3: 1,160 kPa

Rib 4: 1,060 kPa


Rib 1: 588 kPa

Rib 5: 570 kPa

1,070 kPa

1,170 kPa

1,040 kPa

160.0 mm

143.0 mm
570 kPa

588 kPa
34 mm 34 mm 34 mm 34 mm
31 mm 32 mm 31 mm 31 mm 32 mm 31 mm

(b)

1.00
Normalized Contact Stresses

0.70
Vertical Stress
Longitudinal Stress
0.40 Lateral Stress

0.10

0 100 200
-0.20
Data not available
Longitudinal Contact Points

(c)

2 E6 E7 E8 E9 E1 E2 E3 E4 E5
1.2
1.8
Contact Stress Amplitude

Contact Stress Amplitude

1.6 1
1.4
0.8
1.2
1 0.6
0.8
0.6 0.4
0.4
0.2
0.2
0 0
1 2 Time Step 3 4 1 2 Time Step 3 4

(d) (e)
FIGURE 3 (a) Tire footprint for the dual tire configuration; (b) Simulated tire contact area at 36.7kN load
and 724kPa nominal inflation pressure; (c) Normalized measured-contact stress distribution; and Contact
stress ramp-loading amplitude by time step: (d) leading half of tire imprint and (e) following half of tire
imprint.
Al-Qadi, Elseifi, Yoo, Dessouky, Gibson, Harman, D’Angelo, and Petros 6

To incorporate measured surface tangential stresses in the FE analysis, all transverse and longitudinal stresses
induced by a rolling tire were discretized into tangential forces according to the contact area of each individual tire
rib. Two vehicle speeds, 13 and 80km/h were considered in the FE analysis. For each speed, the tire imprint area is
maintained constant within its influential finite-elements area. To simulate the movement of the tire at a certain
speed, the tire imprint area is gradually shifted over the loaded area until a single wheel pass is completed. A linear
continuous loading amplitude, Figures 3d and 3e, was applied to accurately simulate the variation of contact stress
on the tire imprint during load movement. In this case, various loading amplitude paths are used to define the
entrance and the exit parts of the tire imprint, respectively. In this process, as the vehicle approaches a given
element in the loading path, the element is loaded with the amplitude that simulates the increase in loading with time
(Figure 3d). Similarly, as the tire moves away from a given element, the loading amplitude that simulates the
decrease in loading with time is used (Figure 3e). Figures 3d and 3e show that the entrance and the exit tire loading
amplitudes are asymmetric, which is in agreement with field observations (11).

Material Characterization
To accurately simulate pavement responses to vehicular loading, it is essential to characterize the properties of all
relevant materials. An elastic constitutive model was assumed for the granular layers and subgrade (Ebase=207MPa
and Esubgrade=41MPa). Additionally, a viscoelastic constitute model was used to simulate HMA responses to
vehicular loading. To describe a viscoelastic material, the initial instantaneous modulus and its decay with loading
time must be defined. To obtain these material properties, complex modulus experimental data were used to
develop an extended master curve in accordance with the procedure suggested by Bonaquist and Christensen (12).
This procedure employs the sigmoidal function to describe the rate dependency of the complex modulus master
curve:

log(E*)   
Max   (5)
 E a  1   1   
   log( t )      
 19.14714  T   295.25   
1 e
where,
E* = complex modulus;
t = loading time;
T = temperature in  Rankine;
,  and  = fitting parameters; and
Max = limiting maximum modulus.

The obtained master curves using this approach are shown in Figures 4a through 4d for the soft and stiff binders at a
reference temperature of 46°C and 10°C as determined from laboratory test data.

10000 10000

1000 1000
E*, ksi
E*, ksi

4
100 100 19
4 31
19 46
31 58
10 46 10 Fit
58
Fit

1 1
1.E-03 1.E-01 1.E+01 1.E+03 1.E+05 1.E+07 1.E+09 1.E-07 1.E-05 1.E-03 1.E-01 1.E+01 1.E+03 1.E+05

Reduced Frequency, Hz Reduced Frequency, Hz

(a) (b)
Al-Qadi, Elseifi, Yoo, Dessouky, Gibson, Harman, D’Angelo, and Petros 7

10000 10000

1000 1000
E*, ksi

E*, ksi
100 100
4 4
19 19
31 31
10 46 10 46
58 58
Fit Fit

1 1
1.E-03 1.E-01 1.E+01 1.E+03 1.E+05 1.E+07 1.E+09 1.E-07 1.E-05 1.E-03 1.E-01 1.E+01 1.E+03 1.E+05

Reduced Frequency, Hz Reduced Frequency, Hz

(c) (d)

FIGURE 4 Master curve for HMA with soft binder at a) 46C and b) 10C reference temperatures; and
HMA with stiff binder at c) 46C and d) 10C reference temperatures (1ksi= 6.89 MPa; Numbers on the
series indicate the testing temperatures in oC).

After fitting the sigmoidal model (NCHRP 1-37A) to the measured data, this model was used to extend the master
curve to very low and high frequencies not attainable in the laboratory, which allowed for reaching the plateau
regions on both sides of the curve; Figures 5a and 5b. Based on the time-temperature superposition principle and as
shown in these figures, shifting of the curve to various temperatures does not affect the behavior at very high
frequencies, which correspond to the elastic glassy region. This is in agreement with the horizontal shifting
approach allowed by the time-temperature superposition principle. Based on this analysis, the elastic instantaneous
modulus (E0) was determined as corresponding to the modulus at very high frequencies (i.e., 1x1014Hz).
Although the elastic instantaneous modulus is independent of temperature, the decay of the modulus is
greater at 46°C than at 10°C indicating the softer behavior of the HMA at high temperatures. This is in agreement
with the elastic theory that is valid in the glassy region and describes a material independent of temperature and
loading frequency. This analysis reveals a major drawback of the resilient modulus test, which uses the modulus at
10Hz while assuming that it represents the elastic component of the material. To describe the isotropic, linear
viscoelastic behavior of HMA in the FE model, a generalized Maxwell solid model was used and fitted to the
sigmoidal model (Figures 5c and 5d). The model consists of a spring and n-Maxwell elements connected in parallel:

 g 1  e 
N
G t   G 0  i
 t/τ i (6)
i 1

where,
G(t) = shear relaxation modulus at time t;
G0 = glassy instantaneous shear relaxation modulus;
gi = material constants referred to as relaxation strengths; and
 i = relaxation times.

1.E+04 1.E+04

1.E+03 1.E+03
Complex Modulus, ksi

Complex Modulus, ksi

1.E+02 1.E+02

1.E+01 HMA - Soft 10C 1.E+01 HMA - Stiff 10C


HMA - Soft 46C HMA - Stiff 46C

1.E+00 1.E+00
1.E-10 1.E-06 1.E-02 1.E+02 1.E+06 1.E+10 1.E+14 1.E-12 1.E-08 1.E-04 1.E+00 1.E+04 1.E+08 1.E+12 1.E+16

Reduced Frequency (Hz) Reduced Frequency (Hz)

(a) (b)
Al-Qadi, Elseifi, Yoo, Dessouky, Gibson, Harman, D’Angelo, and Petros 8

1.E+00 1.E+00
Normalized Shear Modulus

Normalized Shear Modulus


1.E-01 1.E-01

1.E-02 1.E-02
HMA - Soft 10C - Prony HMA - Stiff 10C - Prony
HMA - Soft 10C - sigmoidal HMA - Stiff 46C - Prony
1.E-03 1.E-03
HMA - Soft 46C - Prony HMA - Stiff 10C - sigmoidal

HMA - Soft 46C - sigmoidal HMA - Stiff 46C - sigmoidal


1.E-04 1.E-04
1.E-10 1.E-06 1.E-02 1.E+02 1.E+06 1.E+10 1.E+14 1.E-10 1.E-06 1.E-02 1.E+02 1.E+06 1.E+10 1.E+14

Reduced Frequency (Hz) Reduced Frequency (Hz)

(c) (d)
o
FIGURE 5 Master curves at 10° and 46 C based on the sigmoidal model for (a) HMA with soft binder and
(b) HMA with stiff binder and representation of the viscoelastic material properties in Prony series form for
(c) soft and (d) stiff HMA.

Hot-Mix Asphalt complex modulus and phase angle data were used as the base of inter-conversion to other
viscoelastic functions. The shear moduli [G(t)] variation with time was estimated using the following relation
assuming that Poisson’s ratio (  ) has a relatively small effect on pavement behavior and does not change with time
(assuming that the material is isotropic and homogeneous and that the resulting strain is small):
E( t ) (7)
G(t) 
2(1  )
The calculated shear moduli were then normalized with respect to the instantaneous shear moduli associated with
the instantaneous modulus of elasticity (E0) as follows:
G t 
 g 1  e 
K
1 i
 t i (8)
Go i 1

where,
G(t) = shear relaxation modulus at time t; and
Go = instantaneous shear modulus (e.g., for soft HMA, G0 = 9112MPa; for stiff HMA, G0 = 9710MPa).

Using a built-in module in the FE commercial software ABAQUS, variations of the normalized shear modulus with
time were fitted to Equation (8) through a nonlinear least squares curve fitting process. Eleven to thirteen Prony
series terms were used to obtain an accurate fit. Figures 5c and 5d present a comparison between the input data
(sigmoidal model) and the Prony series model. Based upon this analysis, characterization of the viscoelastic
properties of HMA was completed at two temperatures, and all required parameters for the constitutive model were
defined.

LOADING TIME CALCULATION AND FREQUENCY ANALYSIS


Results of the FE model were used to calculate the loading time at various pavement depths using vertical stress
responses. Vertical stress responses were selected in this analysis over other stress or strain components (e.g.,
longitudinal strain) for two main reasons: First, the vertical stress response is always compressive and is not
confounded with a change in loading directions. Second, strain responses in a viscoelastic analysis tend to exhibit
delayed responses, which may also confound the end of the loading process.
Variation of the vertical stress at the desired depth was determined from the FE model and plotted versus
time (Figure 6a). In this figure, the variation of vertical stress is shown under three tire ribs: T1, T3, and T5. As
expected, the maximum vertical stress was underneath the center rib (T3). Therefore, results under the center rib
were used in the loading time calculation. As shown in Figure 6a, it was difficult to visually locate the exact start
and end of loading on the vertical stress-pulse plot. Therefore, a procedure was developed to address this problem
and to establish a sound comparison between the various cases. In this approach, it is considered that actual loading
Al-Qadi, Elseifi, Yoo, Dessouky, Gibson, Harman, D’Angelo, and Petros 9

and unloading of the pavement will be associated with a change in slope direction of the wave (see Figure 6b).
Therefore, the span of loading may be estimated by considering the time at which the slope changes from negative
to positive at the start and end of the wave. The slope of the wave at any loading time was calculated as follows:
1   2 (9)
Slope 
t1  t 2
where,
1 and 2 = vertical stress at two consecutive points on the pulse; and
t1 and t2 = associated loading times with 1 and 2.

500 3.E+04
450
400 2.E+04
Vertical stress (kPa)

350
1.E+04 Start of Loading End of Loading
300
250

Slope
T1 0.E+00
200
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
150 T3
-1.E+04
100 T5
Loading Time
50 -2.E+04
0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 -3.E+04
Time (sec) Time (s)

(a) (b)
FIGURE 6 (a) Calculated vertical stress under different tire ribs and (b) Loading time calculation based on
the results of the FE model.

RESULTS AND DISCUSSION


Based on the various simulation cases, the influences of several design and operation parameters were investigated.
Results of the FE model were also compared to the methodology adopted in the MEPDG. The frequency correction
factor can be defined at any depth as follows:
f (10)
CF  FE
f DG

where,
CF = correction factor applicable under selected design and operation conditions;
fDG = frequency calculated from the design guide; and
fFE = frequency calculated from the 3D FE analysis.

In this paper, all frequency calculations were based on the MEPDG method using Equation (3). A more detailed
analysis that compares frequency calculations based on Equations (3) and (4) has been presented elsewhere (3).
Table 2 compares the calculated frequency at various depths based on the MEPDG procedure to the 3D FE method.
Due to the assumed non-uniform contact stress distribution, the frequency of loading was not determined at the
pavement surface because of its high variation over the tire contact area. It is evident from these results that the
calculated frequencies based on the MEPDG method were predominantly greater than the ones determined from the
3D FE method. This indicates that the loading time determined from the MEPDG method is not conservative and
the estimated complex modulus under normal operating conditions is greater than the values predicted by the
viscoelastic FE analysis.
Al-Qadi, Elseifi, Yoo, Dessouky, Gibson, Harman, D’Angelo, and Petros 10

TABLE 2 Calculated Frequencies (Hz) Based on the MEPDG Method and 3D FE for (a) Low Volume and
(b) Interstate Roads
(a)
Case 1^ Case 3 Case 5 Case 7
Depth (mm) MEPDG 3D FE MEPDG 3D FE MEPDG 3D FE MEPDG 3D FE
25 10.53 3.64 46.36 12.16 8.56 2.68 37.94 12.56
50 6.77 3.63 22.77 12.09 4.49 2.51 17.88 12.47
75 5.25 3.62 17.66 12.07 3.48 2.48 13.86 12.02
100 4.38 3.56 14.74 12.06 2.91 2.47 11.57 11.12
125 3.81 3.49 12.81 12.03 2.53 2.47 10.06 11.10
150 3.40 3.43 11.42 11.98 2.25 2.46 8.97 11.08
(b)
Case 2 Case 4 Case 6 Case 8
Depth (mm) MEPDG 3D FE MEPDG 3D FE MEPDG 3D FE MEPDG 3D FE
50 8.43 3.33 31.50 11.86 6.16 2.00 24.21 21.87
100 4.38 3.08 14.74 11.61 2.91 2.01 11.57 20.83
150 3.40 2.91 11.43 11.12 2.25 1.94 8.97 19.95
200 2.84 2.81 9.54 10.20 1.88 1.81 7.49 12.02
250 2.46 2.76 8.29 8.92 1.64 1.48 6.51 10.58
300 2.20 2.72 7.40 6.40 1.40 1.39 5.81 6.71
^ See Table 1 for Definition of Cases

Correction Factors
Table 3 presents the frequency correction factors as determined by Equation (10) and shows that the MEPDG
method over-predicts the applicable frequency in the majority of the cases resulting in an unreliable calculation of
the complex modulus. This also suggests that the 45o distribution angle considered in the Odemark approach is not
representative of actual field conditions since it does not consider the far field loading region of a tire while the FE
approach does.

TABLE 3 Calculated Frequencies Correction Factors Based on (a) the MEPDG Method and (b) the
Rheological Approach
(a)
Case 1 Case 3 Case 5 Case 7 Averages
Depth (mm) CFMEPDG CFMEPDG CFMEPDG CFMEPDG ----
25 0.35 0.26 0.31 0.33 0.31
50 0.54 0.53 0.56 0.70 0.58
75 0.69 0.68 0.71 0.87 0.74
100 0.81 0.82 0.85 0.96 0.86
125 0.92 0.94 0.98 1.10 0.99
150 1.01 1.05 1.09 1.23 1.10
(b)
Case 2 Case 4 Case 6 Case 8 Averages
Depth (mm) CFMEPDG CFMEPDG CFMEPDG CFMEPDG ----
50 0.31 0.31 0.31 0.31 0.31
100 0.58 0.58 0.58 0.58 0.58
150 0.74 0.74 0.74 0.74 0.74
200 0.86 0.86 0.86 0.86 0.86
250 0.99 0.99 0.99 0.99 0.99
300 1.10 1.10 1.10 1.10 1.10
Al-Qadi, Elseifi, Yoo, Dessouky, Gibson, Harman, D’Angelo, and Petros 11

Use of Correction Factors


To quantify the level of inaccuracy associated with Equations (1) and (2), the presented analysis utilized an
advanced three-dimensional Finite Element (FE) approach to simulate the approaching-leaving rolling wheel at a
specific speed. The inaccuracy associated with the conversion from the time domain to the frequency domain is
presented in another paper (3). Two pavement designs were considered in this analysis, one representing a typical
interstate highway and another simulating a secondary road. Figure 7 presents the variation of the correction factors
with depth for these two pavement designs (i.e., averages presented in Table 3). The equations presented in this
figure may be used to correct the calculated frequency obtained using the simplified approach adopted in the
MEPDG. For instance, if the MEPDG simplified procedure (Equations 1 and 2) predicts a frequency of loading of
5Hz for a mix at a representative depth of 75mm in a primary road, the frequency to be used to determine the
complex modulus for this layer would be 2.3Hz (CF = 0.003x75mm + 0.2333). This would address un-conservative
characterization of the load pulse and subsequent selection of HMA complex modulus in the MEPDG.

1.20

1.00
CF = 0.006z + 0.2378
R2 = 0.97
0.80
CF = 0.003z + 0.2333
CF = fFE/fDG

R2 = 0.97
0.60

0.40

0.20
Secondary Road Interstate

0.00
0 50 100 150 200 250 300 350
Depth (mm)

FIGURE 7 Variation of correction factors with depth and pavement design.

CONCLUSIONS
In this paper, results of the MEPDG methodology for calculating loading time were compared to an advanced 3D FE
model, which considers the viscoelastic behavior of HMA and the effect of the moving load in three dimensions.
The following conclusions were drawn:
 Analysis based on the MEPDG for frequency calculation indicates that the calculated frequencies were greater
than the ones determined from the 3D FE method. This would correspond to greater complex moduli, and
hence over predict the influence of the HMA layer in the pavement system. This indicates that the loading time
determined from the MEPDG would not be conservative since the associated complex modulus will be greater
than what it actually is under normal operating conditions.
 This un-conservative characterization of the load pulse and subsequent selection of HMA complex modulus in
the MEPDG is offset by the empirical calibration of the performance models. In other words, the MEPDG is an
effective tool in analyzing and predicting pavement performance. However, the time dependency of
unmodified and modified asphalt binders may be lost under the current approach. To address this issue, this
paper presents correction factors to ensure the correctness of loading time calculation in the MEPDG. Adoption
of the proposed factors does necessitate a recalibration of the performance models.
 To address the conversion from the time domain to the frequency domain, analysis based on Fast Fourier
Transform (FTT) should be investigated. The principal author has developed an accurate approach to calculate
the frequency from the loading pulse (3). This recent study also evaluated the accuracy of Equations (3) and (4)
Al-Qadi, Elseifi, Yoo, Dessouky, Gibson, Harman, D’Angelo, and Petros 12

for frequency calculation. Future work should also analyze elasto-plastic behavior of granular layers and
subgrade and dynamic response of pavement systems.
ACKNOWLEDGEMENTS
Part of this research was funded by the Federal Highway Administration. The contents of this paper reflect the view
of the authors. The contents do not necessarily reflect the official views or policies of the Federal Highway
Administration. This paper does not constitute a standard, specification, or regulation. The authors would like to
acknowledge the input of Hao Wang, Marshall Thompson, Sam Carpenter, and Raj Dongre.
REFERENCES
1. National Cooperative Highway Research Program. (2004). “Guide for Mechanistic-Empirical Design of
New and Rehabilitated Pavement Structures.” NCHRP, Washington, D.C.
2. Thompson, M.R., Carpenter, S.H., Dempsey, B.J., and Elliot, R.B. (2006). “Independent Review of the
Recommended Mechanistic-Empirical Design Guide and Software” NCHRP 1-40A, Transportation
Research Board, Washington, D.C.
3. Al-Qadi, I.L., Xie, W., and Elseifi, M.A. Frequency Determination from Vehicular Loading Time Pulse to
Predict Appropriate Complex Modulus in MEPDG. Submitted to The Journal of the Association of Asphalt
Pavement Technologists.
4. Huang, Y.H., Pavement Analysis and Design, Prentice Hall, Inc., 2nd Ed, 2004.
5. Al-Qadi, I.L., Elseifi, M.A. and Yoo, P.J. (2004). “In-Situ Validation of Mechanistic Pavement Finite
Element Modeling.” Proc., 2nd International Conference on Accelerated Pavement Testing, B. Worel and K.
Faults, Eds., Minneapolis, MN.
6. Loulizi, A., Al-Qadi, I.L., Elseifi, M.A. (2006). Difference between in-situ Flexible Pavement Measured
and Calculated Stresses and Strains. Journal of Transportation Engineering, ASCE, Vol. 132, No. 7, pp.
574-579.
7. Dongre, R., Myers, L., and D’Angelo, J. (2006). Conversion of Testing Frequency to Loading Time:
Impact on Performance Predictions Obtained from the M-E Pavement Design Guide. Paper No. 06-2394
Presented at the 85th Transportation Research Board Annual Meeting.
8. ABAQUS Theory Manual, Version 6.5. Habbit, Karlsson & Sorenson, Inc., Pawtucket, RI, 2005.
9. Al-Qadi, I. L. and P. J. Yoo. Effect of Surface Tangential Contact Stress on Flexible Pavement Response.
Journal of the Association of Asphalt Pavement Technologists, Vol. 76, 2007 (In Press).
10. Al-Qadi, I.L., Yoo, P.J., Elseifi, M.A., and Janajreh, I. Effects of tire configurations on pavement damage”
Journal of the Association of Asphalt Pavement Technologists, Vol. 74, 2005, pp. 921-962.
11. Yoo, P.J., Al-Qadi, I.L., Elseifi, M.A., and Janajreh, I. Flexible Pavement Responses to Different Loading
Amplitudes Considering Layer Interface Conditions and Lateral Shear Forces. International Journal of
Pavement Engineering, Vol. 7, No. 1, 2006, pp. 73–86.
12. Bonaquist, R., and Christensen, D.W. A Practical Procedure for Developing Dynamic Modulus Master
curves for Pavement Structural Design. Journal of the Transportation Research Board, 1929, Washington,
D.C., 2005, pp. 208-217.

View publication stats

You might also like