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

net/publication/233001796

Development of 3‐D finite element model for flexible pavements

Article in Journal of the Chinese Institute of Engineers · July 2004

DOI: 10.1080/02533839.2004.9670918

CITATIONS READS

32 282

2 authors, including:

Chen-Ming Kuo
National Cheng Kung University
26 PUBLICATIONS 381 CITATIONS

SEE PROFILE

All content following this page was uploaded by Chen-Ming Kuo on 01 August 2024.

The user has requested enhancement of the downloaded file.

 Development of 3-D Finite Element Model for Flexible Pavements

∗
Chen-Ming Kuo Fang-Ju Chou
Department of Civil Engineering
National Cheng Kung University
Tainan, Taiwan 701, R.O.C.

Key words: flexible pavement; three-dimensional finite element; viscoelastic; creep; infinite

elements; geogrid; AASHO Road Test; ABAQUS

Abstract：This paper presents the development of a three-dimensional finite element model for

flexible pavements. The procedures of building a model and performing static analysis with the

ABAQUS package are introduced. The element shapes and the size of the finite element model

were studied to improve analysis efficiency. The model should at least consist of finite elements

up to three times the loading diameter if centered at the loading position. Infinite elements

should be used beyond the boundary of finite elements. The accuracy was justified with

Boussnisq solutions for semi-infinite elastic solids. The surface layer was modeled with creep

test data of asphalt concrete. Viscoelastic behavior was observed for the pavement model under

wheel loading. The model was also validated by comparing computed results and field test data.

Further calibrations of the model are underway by comparison with laboratory test results. The

application of the model is introduced with an example of placing a geogrid-reinforced asphalt

concrete overlay on rigid pavement. Significant reduction of tensile strains at the bottom of the

asphalt concrete was found with geogrid reinforcement.

∗
Correspondence addressee

1
INTRODUCTION

Detailed modeling of complicated pavement structures is always desired to enhance insight

into pavement behavior. With the improvement of analysis techniques, many interactions among

pavement components and failure mechanisms can be revealed with three-dimensional (3-D) finite

element analysis, such as, pavement sections with culverts in subgrade, multiple wheel loading,

nonlinear behavior of base materials, and distresses in pavements.

The objective of this research is to develop procedures for building a model of flexible

pavement and performing static analysis with a finite element analysis package. For finite

element analysis, fine mesh gives higher accuracy, but increases number of elements and

computation time. Furthermore, a large model will assure no boundary errors, but again demands

much computer memory and time. In order to obtain guidelines of model size and mesh fineness,

a semi-infinite elastic solid was modeled and checked by comparing the computed displacements

and stresses with Boussinesq’s solutions.

The viscoelastic layer was put on top of the semi-infinite elastic solid to form a one-layer

pavement structure. This system was validated by examining the viscoelastic behavior, and by

comparing with the deflections of the AASHO Road Test. The development of the 3-D model

was advantageous in many aspects. For example, the effects of geogrid reinforced AC overlay

can be explored with finite element analysis developed in this study. Based on the procedures of

meshing and validation, it is simple to build various multi-layer pavement systems.

It should be noted that several advanced features, e.g., nonlinear materials and dynamic

analysis are not in the scope of this research. These features are currently at the development

stage.

2
MAIN FEATURES OF THE 3-D FLEXIBLE PAVEMENTS MODEL

After the development of computer programs in multi-layer elastic theory, finite element

analysis has been adopted for pavement research since the 1980s. At first, two-dimensional (2-D)

finite element models were developed (Ioannides et al, 1985). Although they were widely

prevalent at that time, many inherent limitations of 2-D finite elements led to restrictions on

refining the models towards further implementation in realistic pavement structures (Kuo et al.,

1996). For example, the programs that based on axisymmetric revolution of solids could only

take care of a single load print. Besides, the existing finite element programs used for flexible

pavement are incapable of handling these situations: widened base, different materials in one layer,

and various interfacial bondings between layers (Hadi & Bodhinayake, 2003).

Three-dimensional finite element analysis being a robust alternative can incorporate the

complicated features mentioned above to be used for advanced flexible pavement analysis,

although revising the existing a 2-D model to overcome its limitations is possible.

Material properties

Many special features including asphalt material, granular material, and subgrade soil have
been incorporated in other flexible pavement programs. ILLI-PAVE has incorporated stress

dependent resilient modulus and failure criteria for granular material and fine-grained soils

(University of Illinois at Urbana-Champaign, 1990). KENLAYER has considered linear elastic,

nonlinear elastic, linear viscoelastic, and combinations of nonlinear elastic and linear viscoelastic

behaviors of materials in different layers (Huang, 1993). Furthermore, ABAQUS, the finite

element package with broad applications, has provided a variety of options in analyzing material

properties such as viscoelasticity, porous material, clay plasticity, and Drucker-Prager models

(ABAQUS,1996a), gaining significant attention recently in pavement analysis. These options of

ABAQUS have not been properly applied for flexible pavements although viscoelastic behavior

was simulated, earlier, using a quasi-elastic approach.

The main feature of flexible pavement model is to characterize the behavior of asphalt

concrete. Asphalt acts as an elastic solid and obeys Hook's law at low temperature. It deforms

and recovers instantly upon loading and unloading. Yet, it softens as temperature rises and

behaves as a viscous fluid, which obeys Newton's law of viscosity holding constant creep rate at

3
high temperature (Huang, 1993; Rowe et al., 1995). At ambient temperature, the behavior of

asphalt material is between pure elastic and viscous model and it is called "viscoelastic." However,

simple one-dimensional models are not practical in predicting field responses of flexible

pavements because of complexity of appropriate combinations of damping and springs and

difficulties in calibrating damping and spring parameters for three-dimensional behavior (Haddad

1995).

Meshing Criteria

Meshing criteria are critical in a flexible pavement model in which the subgrade is

considered as semi-infinite elastic solid. Inappropriate mesh size may result in either poor

accuracy or enormous demand for computer resources. In the study proposed by Zaghloul and

White (1992), fine meshes were used under the wheel path. The mesh size in the vertical

direction was fitted by thickness of structural layer except that the subgrade was modeled with five

layers of finite elements.

The research conducted by Hjelmstad et al. (1997) concluded that the element shape can

significantly control the model analysis in terms of its run-time and accuracy. In addition, the

stress gradient is usually quite large in the vicinity of a loading area due to the fact that a pavement

system comprises relatively thin layers, leading to a highly localized loading. This justifies the

need for fine mesh around loading areas as illustrated in Fig.1. Infinite elements were also

helpful in reducing the number of elements in the subgrade model to improve the efficiency of

analysis (Chen et al. 1995).

4
 Fig. 1 Ideal mesh of interior loading (Chen et al. 1995)

Boundary conditions

Attention should be paid to determine the appropriate boundary conditions to assure a

realistic modeling. In the model developed by Zaghloul and White (1992), transverse

displacements were restricted along the longitudinal edge. The study further suggested that a

pavement section of 10m of length is appropriate to eliminate "edge effect error." Hjelmstad et al.

(1997) reported that a domain size equivalent to 150 times of the loading radius is appropriate to
avoid the effects caused by boundary conditions, either roller or hinge.

DEVELOPMENT OF 3-D ASPHALT PAVEMENT MODEL

A 3-D finite element model of asphalt pavement was developed in this research by taking the

initiatives in correcting major deficiencies of the existing models. This 3-D model employed

cubic finite elements for surface layer and parts of subgrade and infinite elements for the
boundaries of each layer to reduce the number of elements for the far field. The meshing criteria

were developed as described below.

5
Subgrade

Subgrade was modeled with both brick finite elements and infinite elements as shown in Fig.

2. It is desirable to minimize the number of brick element layers to save run-time and computer

resources. In Fig.2, C represents the side length of a square which has the same area as that of

the circular loading with radius of a. Eq. (3) shows the relationship between C and a. P and q

represent total load and uniform pressure of the load.

P
= C2 = π a2 (3)
q

A typical problem of semi-infinite elastic solid under circular uniform load was used as an

example and modeled in different approaches. These results were compared with the theoretical

solution by Boussinesq (ABAQUS,1995). Using the model of Fig. 2, the analyzed stress was

conformable to the theoretical solution with the horizontal extent of a brick finite element being

equal to C. Yet, significant discrepancy existed in displacements. Later, a model using depth of

2C for brick elements did not improve the discrepancy of displacement. The model was then

expanded horizontally, while using a fixed depth of 1C for brick elements. Fig.3 shows the
convergence of displacement as the horizontal extent of brick elements increases to 3C. Hence, it

is suggested that the semi-infinite solid is meshed with brick elements being extended to 3C

horizontally and 1C vertically away from the loading, and then connected with infinite elements

beyond the brick elements.

Meanwhile, it was observed that good accuracy could be obtained with fine mesh at loading

areas. Namely, it is suggested that the length of elements under the loading area be assigned to

become one half of the length of the loading area. A similar conclusion was presented by

Ioannides (1984) based on 2-D finite element analysis.

6
 x 2C
1C
z 0.5C

1C
brick element Infinite element

Infinite element

Fig. 2 Side view of elastic solid model for subgrade

5
600 3-D model, x=1C
3-D model, x=1C
3-D model, x=2C
3-D model, x=2C 4 3-D model, x=3C
3-D model, x=3C 3-D model, x=4C
500 3-D model, x=4C Bousinessq semi-
3
Vertical stresses (kPa)

Deflection (mm)

Bousinessq infinite elastic sol.

semi-infinite elastic
400 l
2

300 1

200 0 25 50 75
0 100 200 300 400 Z Depth at X=Y=0 (mm)
Z depth (mm) at X=Y=21.5mm

Fig. 3 Results of 3-D models with various horizontal extent of finite elements

Surface Layer

The surface layer is composed of finite elements. The meshing criteria for finite elements,

which were rarely noticed before in pavement analysis, are included in this study to assure

accuracy of analysis.

(1) The aspect ratios of all elements were kept below 5.

7
 (2) The ratio of height to width of an element was less than 2.

(3) The vicinity of loading was composed of cuboids elements with similar edge length.

(4) The mesh was fitted to the loading area to avoid partially loaded elements.

(5) The transition area was set between fine mesh and coarse mesh for gradual change of

element size to ensure good accuracy.

A special feature of the flexible pavement model is its viscoelastic property for asphalt

material. ABAQUS provides a set of options to simulate viscoelastic behavior. The properties of

viscoelastic material can be described with different approaches (ABAQUS 1996a).

In this research, the 3-D viscoelastic models expressed in the form of equation (1) and

equation (2) are adopted by the finite element package (ABAQUS,1996b). The equations are a

superposition of the generalized Kelvin-Voigt model for three-dimensional relationships (Malvern,

1969).

1 ∞ dG (τ )
S(t ) = S 0 (t ) +
G0
∫ 0
dτ
S 0 (t − τ )dτ (1)

1 ∞ dK (τ )
p(t ) = p 0 (t ) +
K0 ∫ 0 dτ
p 0 (t − τ )dτ (2)

where
S(t) = deviatoric stress tensor
p(t) = material pressure
S0(t) = deviatoric stress state which would exist for the current state of strain
p0(t)= pressure state which would exist for the current state of strain if the material
should behave purely elastically.
G(t) = shear relaxation modulus
K(t) = bulk relaxation modulus
G0(t) = instantaneous shear relaxation modulus
K0(t) = instantaneous bulk relaxation modulus
t = creep time
τ = Reduced time

To validate the model, a creep test is required to define the relationship between material
modulus and elapsed time of loading. Bulk modulus at any moment during the creep test can be
obtained with a creep test curve as shown in Fig. 4. Then, shear modulus corresponding to bulk

8
modulus at an arbitrary moment, can be calculated indirectly and used as inputs for viscoelastic
analysis by ABAQUS ("Analysis", 1996). To assure correct simulation for the designated asphalt
behavior, it is important to know that accurate description of the creep curve should be available;
and it can be retrieved as shown in Table 1.

According to the suggested model extent for a semi-infinite solid, the criteria of model
extents for a layered system were investigated by using a model composed of infinite elements in
traffic direction (x-dir.) and foundation (z-dir.). Finite boundary was set to be equal to the width
of a traffic lane in the direction transverse to the centerline (y-dir.). The parameters of pavement
section and loading are shown in Fig. 5.
Stiffness, S (psi or kg/cm2) (log scale)

Elastic behavior
Delayed or retarded
elastic behavior

Viscous
(1) At short loading times S=E
behavior
(2) At intermediate loading times S = E(t)
(3) At long loading times S = λ/t
Where λ = viscous traction ≒ 3 η
η : a measure of viscous deformation

Time, t (log scale)

Fig. 4 Stiffness-time curve of asphalt under uniaxial loading (Roberts et al., 1991)

9
 Table 1 Measurements of uniaxial creep test (Matthews and Schmitt 1995)

Loading time Creep stiffness ( psi ) (1 psi = 6.89 kPa)

( sec ) 25℃ 38℃ 49℃ 60℃
AR 8000 asphalt concrete
0.13 366227 81173 43736 44561
1.05 134646 35243 26313 33935
11.09 60980 24386 21131 29622
116.7 31252 20100 17769 26215
3500 17466 15236 10421 18214
AR 4000 asphalt concrete
0.13 148147 61363 35052 23990
1.05 76752 25433 21136 19860
11.09 37379 18309 16776 15818
116.7 22412 14940 14181 13418
3500 12002 10910 8734 8278

C = 0.4 m
q = 484 kPa (70psi) Loading area = 0.4m x 0.4m

AC surface layer (AR4000) 101.6 mm

C = 0.4 m
4 layers of finite elements
Subgrade (E = 68.9 Mpa)
1 layer of infinite elements

Fig. 5 Model parameters for pavement section and loading with AC surface layer

Based on the conclusions mentioned in the previous analysis, it turns out from this work that
the vertical extent of finite element mesh dominates convergence of stress. On the other hand,
extent of finite element mesh in x-direction and y-direction can control accuracy of displacements.
Fig.6 shows the convergence of stresses and displacements with respect to models of different
horizontal extents. The model with horizontal extent of 3C delivered the same stress accuracy as
the one meshed up to 6C. It can be seen from Fig.6 that displacement does not converge very
well as opposed to the vertical compressive stress. Because of the limitation of computer resource,
meshing finite element to 3C horizontally and followed by infinite elements is still recommended
as a compromise, since the stress is the major concern in pavement analysis. Vertically, finite

10
element should be used down to 1C below the bottom of surface layer and followed by infinite
element. It should be noted that the meshing recommendations mentioned above are justified in
the flexible pavement model only with one asphalt surface layer on top of the semi-infinite elastic
solid subgrade. For pavement systems with multiple layers, further study is needed to set the
meshing criteria.
0 0.10

-140 0.08

(in)
Vertical stress (kPa)

Vertical displacement
-280 0.06
z = 1C
z = 1C
x = 3C
x = 3C
x = 4C
-420 0.04 x = 4C
x = 5C
x = 5C
x = 6C
x = 6C
Theory (BISAR)
Theory (BISAR)
-560
0.02
0 4 8 12 16 20 0 4 8 12 16 20
Z Depth (mm) at x = y =21.5mm Depth at x = y =0
(a) (b)
0.002 0.002

z = 1C
x = 3C
x = 4C
Vertical strain

0 x = 5C
0.001
x = 6C
Theory (BISAR)
Horizontal strain

z = 1C
x = 3C
-0.002
0 x = 4C
x = 5C
x = 6C
Theory (BISAR)

-0.004
-0.001
0 5 10 15 20
0 4 8 12 16 20
Depth at x = y = 0.8453" Depth at X=Y=0.8453" (inch)

(c) (d)
Fig. 6 Results of various horizontal extents of pavement system in reference to Fig.5

11
MODEL VALIDATION

The 3-D finite element model for flexible pavement was qualitatively validated in three
aspects: theoretical solution of semi-infinite elastic solid with circular loading by Boussinesq,
creep of viscoelastic material, and comparison with the estimated static deflections of AASHO
Road Test.

Semi-infinite elastic solid

The analysis results shown in Fig.3 were further examined in horizontal direction. Fig.7
and Fig.8 compare the 3-D model results and the theoretical solutions of vertical stress and
displacement along the surface of semi-infinite elastic solid. It is clear that the results of the
proposed 3-D model match the theoretical solutions very well. Fig.7 and Fig.8 were plotted with
the computed results at integration points of each finite element to avoid “numerical singularity”
resulting from extrapolations to the loading point.

Fig. 7 Profile of vertical stress along surface of semi-infinite solid

12
 0
S/C=4 Theoretical
S/C=3 Sol.
S/C=2
0.04

Vertical deflection (in)

0.08 S/C=1

0.12

0.16
S = extent of horizontal finite element

0.20
0 2 4 6
x/C (x = dist. from loading center)

Fig. 8 Profile of vertical displacement along surface of semi-infinite solid

Viscoelastic behavior

To justify the visoelastic behavior of the finite element model, a simple uniaxial creep test
was simulated with an axial load of 689 kPa (100 psi) for 3600 seconds. The creep stiffness data
of AR4000 at 49oC in Table 1 were used and the stiffness at 0.1 second was considered as initial
elastic modulus. Fig. 9 shows the comparison between strains of elastic and their viscoelastic
analysis. As expected, the strains calculated in elastic analysis match the theoretical solution and
are independent of loading duration. In the viscoelastic analysis, the strains increased with
loading time and a typical creep curve was obtained. The viscoelastic strain became steady after
approximately t = 2000 seconds.

Although rigorous model validations are desirable, it is beyond the scope of this work to
provide much more detailed strain data from laboratory creep tests at this time. A series of model
validations with lab data and field data will be accomplished at a time when the next generation
model development is proposed.

13
 --0.002

Elastic
--0.003

Horizontal strain in y-direction

-0.004

-0.005

-0.006 Viscoelastic

-0.004
0.00 1000 2000 3000 4000
Creep Time (sec)

Fig. 9 Comparison between viscoelastic and elastic analysis of the 3-D model

Validation with field tests

The third step of validation was to justify the computed results with field measured data.
However, we were not successful in obtaining the static deflection measurements of the road test.
The validation data for static loading were then indirectly obtained by manipulating the field
deflections measured at 48 km/hr (30 mph). The static deflections shown in Table 2 were derived
from Table 3 using equation (3) which was presented in the AASHO Road Test Report (1962).
The correlation coefficients of equation (4) for the selected cases in the AASHO Report range
from 0.79 to 0.94 in various seasons.

δ 2 = δ v × 10( v − 2) A1 (4)

Where
δ2 : static deflection at creep speed of 2 mph
δv : dynamic deflection at vehicle speed of v in mile/hr
A1 : speed coefficient

14
 Table 2 Estimated static deflections for comparison
Loop Axle Axle Load Structural
no. type (ksi) thickness (in) Date A1 R2 δ30 (in) δ2 (in)
4 Single 18 19 1959,8,26 0.0058 0.79 0.027 0.0392
6 Single 30 27 1959,8,26 0.0055 0.86 0.030 0.04277
1 kips = 4.45kN, 1 in = 25.4 mm
R2 : square correlation coefficient

Table 3 Measured deflections at vehicle speed of 30 mph ("The AASHTO" 1962)

Axle load Structural Surface
Loop no. Axle type (ksi) thickness (in) Measured time deflections (in)
4 Single 18 19 Summer, 1959 0.027
4 Single 18 19 Summer, 1960 -
6 Single 30 27 Summer, 1959 0.029
6 Single 30 27 Summer, 1960 0.030
1 mph = 1.6 km/hr, 1 in = 25.4 mm, 1 kips = 4.45kN

The pavement sections of AASHO Road Test shown in Table 4 were analyzed with the 3-D
model. The detailed information on the associated pavement sections in Loop 4 and Loop 6 is
listed in Table 4. Note that there were 6 loops of test roads built for AASHO Road Tests in the
1950’s. The validation was made by comparing the output of simulated road test with the static
loading and the estimated field deflections.

Table 4 Pavement condition of the AASHO Road Test (1960, summer) ("The AASHTO" 1962)

Loop Thick. design (in) Base layer Subbase Subgrade

no. (surface-base-subbase) unit wt. CBR unit wt. CBR unit wt. CBR k
(pcf) (pcf) (pcf) (psi/in)
5 - 6 - 12 145.0 200 129.0 49.8 114.7 6.8 169

4 5-6-8 133.6 109 130.4 54.4 112.8 5.9 138

5 - 3 - 12 142.0 124 132.9 45.6 113.8 2.4 154
Mean 140.2 144.3 130.8 49.9 113.8 5.0 154
6 - 6 - 16 142.2 82 138.6 10.3 115.3 8.2 196
6 - 9 - 12 142.0 104 130.3 85.4 114.7 9.7 159
6
6 - 9 - 16 141.9 116 142.5 60.3 114.6 5.2 240
Mean 142.0 100.7 137.2 52.0 114.9 7.7 198
3
1 pcf = 157 N/m , 1 in = 25.4 mm, 1 psi/in = 27.1 kPa/m

Fig.10 and Fig.11 demonstrate the comparisons of Loop 4 and Loop 6 between estimations

15
of field measurements and computed results from the 3-D finite element model. As for the
sections of Loop 4, the deflection differences between the model and estimated field data was less
than 3 percent. As to the case of Loop 6, 13% and 18% of deflection differences were found.

The proposed 3-D model has been examined from three viewpoints, i.e., theoretical solution,
viscoelastic behavior, and field test data. Although further justifications are necessary to assure
accuracy, the capabilities and reliability of the proposed 3-D finite element model for flexible
pavements were qualitatively validated.

* Estimated speed from Table 3

0.05
structural thickness = 19
0.0399 0.0403
0.0392
0.04

Computed for Section 4-3-12

Computed for sect. 5-6-8
AASHO measurements*
Deflection (in)

0.03

0.02

0.01

0.00

computed 0.0399
= = 1.0179 Sect. 5-6-8 Loop 4
estimated 0.0392

computed 0.0403
= = 1.0281 Sect. 4-3-12 Loop 4
estimated 0.0392

Fig. 10 Comparison of estimated and computed deflection for sections of Loop 4

16
 * Estimations at creep speed from Table 3
0.06
structural number = 27
0.0505
0.0485
0.05

Computed for Section 5-6-16

0.04277

Computed for Section 6-9-12

Deflection (in)
0.04

Measurement
0.03

0.02

0.01

0.00

computed 0.05045
= = 1.18 Sect. 5-6-16 Loop 6
estimated 0.04277

computed 0.04849
= = 1.13 Sect. 6-9-12 Loop 6
estimated 0.04277

Fig. 11 Comparison of estimated and computed deflection for sections of Loop 6

APPLICATIONS OF THE MODEL

A preliminary model of asphalt overlay on an existing PCC surface layer was constructed
and analyzed to demonstrate the capacity of the 3-D finite element model. Generally, geosynthetic
material, normally called “geogrid”, is used to provide the reinforcement to prevent reflection
cracks. Modeling geogrid is not as straightforward as modeling surface layer and subgrade because
geogrid is like a strong and coarse-spaced net. It is desirable to model with interconnected truss
elements or beam elements (Kuo and Hsu, 2003). However, it is beyond the scope of this paper
which focuses on the development and validation of 3-D viscoelastic finite element model of
flexible pavements. A simplified modeling of gregrid was employed to serve the preliminary study
of the 3-D composite structure model. A layer of membrane elements was placed between asphalt
overlay and existing concrete slabs to simulate the reinforcement in the interface. Fig. 12 shows
the equivalency between grid and membrane.

17
 E

A … E: Young’s modulus of geogrid material

F F F F F F A: Cross sectional area of a single grid
F: Force carried by a single grid
nF E’: Young’s modulus of the equivalent membrane
w: width of the equivalency extent
h’: membrane thickness (use the diameter of grids)
nA
E' =
wh'
w
h’
q

qwh’

Fig. 12 Transformed geogrid for FEM model

Many designs of overlay with geogrid have been tested by highway agencies. Fig. 13
shows the analytical model loaded with an 80 kN axle directly on the location of an old joint.
The geogrid modeled in the example was made of self-adhesive glass fiber having a Young's
modulus of 76 GPa (11.02×106 psi) and a thickness of 1.3 mm (0.05 in).

Axle load (80 kN) geotextile

0.5 m points of analysis output

Asphalt overlay (102 mm)
Existing PCC slab (203 mm) Existing PCC slab

joint
Subgrade (k = 27.1 MPa/m)

Fig. 13 The 3-D finite element model for reinforced overlay pavement

Basically, the critical stresses or strains that occur along the dashed line in Fig. 13 represent
the points at the bottom of the overlay and at the joint. The computed strains at the bottom of the
asphalt overlay along the joint are shown in Fig. 14. The figure clearly shows that geogrid

18
reinforcement substantially reduces tensile strain at the bottom of the asphalt overlay. To
illustrate the significance of strain reduction, allowable load repetitions for two different
pavements were estimated and compared using the fatigue model shown in equation (5) (ERES,
1987).
log10N = -15.74 - 5.964 log10εt (5)

Where
N : the allowable equivalent 80 kN (18 kips) single axle applications for Pt = 2.5
εt : the tensile strain at the bottom of the asphalt overlay

Remarkable reduction of tensile strain and the increase of estimated ESAL are found in Fig.
15, after placing geogrid between the existing layer and the overlay. However, the advantages of
geogrid should not be overestimated from the example of this preliminary analysis because of
other contributing effects such as: structural conditions of the existing slabs, quality of bonding
among geogrid, slab, asphalt overlay, and temperature cycling. All of these are crucial in
evaluating the performance of reinforced asphalt overlay. Comprehensive analysis is needed for
precise conclusions. Generally, this example demonstrates that the proposed 3-D finite element
model is capable of performing efficient analysis of reinforced overlay designs and exploring the
interactions between pavement layers and geogrid.

0.004
With Geogrid
Without Geogrid
Principal strain at the bottom of overlay

0.003

0.002

0.001

0
0 40 80 120 160
Distance in Y direction (in)

19
 Fig. 14 Computed principal strain at bottom of asphalt overlay along the joint

Max. tensile strain at bottom of Estimated allowable ESAL

asphalt overlay 10000000
0.0009
with geogrid with geogrid
0.0008 1000000 without geogrid
without geogrid

No. of ESAL
0.0007
Tensile strain

100000
0.0006
0.0005 10000

0.0004 1000
0.0003
100
0.0002
10
0.0001
0 1

Fig. 15 The effect of geogrid on overlayed pavement

20
CONCLUSIONS

1. It is efficient to build a 3-D finite element model of flexible pavement by using finite
elements and infinite elements simultaneously. It is suggested that finite elements be used to
the extent of 3C away from the loading center horizontally, and to 1C vertically below the
bottom of the surface layer. The finite elements boundary is connected with infinite
elements to handle an infinite boundary such as uniform subgrade without rock bed or
continuous pavement in the traffic direction.

2. The accuracy of computed stresses of the flexible pavement model was dominated by the
extent of the finite element mesh in the vertical direction. Horizontal mesh refinement was
not effective to improve accuracy of stress computation.

3. The accuracy of computed critical deflections of flexible pavement model was dominated by
the extent of finite element mesh in horizontal direction. Vertical mesh refinement was not
effective to improve accuracy of deflection.

4. In this model, time histories of strains or creep stiffness of creep tests are needed to simulate
viscoelastic responses of asphalt material.

5. The model was qualitatively validated in the aspects of analytical adequacy, material
characteristics of viscoelasticity, and field test data. The results show that the model
properly simulates the behavior of flexible pavements and predicts reasonable responses.
Advanced justifications obtained by directly using comparisons from measured strains via lab
and field tests are still needed to verify the accuracy of model outputs.

6. Significant improvement of estimated service life of asphalt overlays on concrete slabs were
found after placing geogrid: between overlay and slabs, and under loading above the old
joints. Further study is under way to characterize the behavior of an asphalt overlay with
geogrid.

21
REFERENCES

ABAQUS Users Manual, Version 5.6, 1996a, Hibbitt, Karlsson and Sorensen, Inc., Rhode Island,
USA.

ABAQUS/Standard Example Problems Manual, Version5.5, Vol.1, 1995, Hibbitt, Karlsson and
Sorensen, Inc., Rhode Island, USA.

ABAQUS - Analysis of Viscoelastic Problems, 1996b, Hibbitt, Karlsson and Sorensen, Inc., Rhode
Island, USA

Chen, D. H. , Zaman, M., Laguros, J., and Soltani, A., 1995, "Assessment of Computer Programs
for Analysis of Flexible Pavement Structure," Transportation Research Record 1482,
Transportation Research Board, Washington, D. C., pp.123-133.

ERES Consultants, Inc., 1987, “Pavement Analysis: Principles and Practice,” Training Manual
prepared for Federal Highway Administration, Illinois, USA, pp.97.

Haddad, Y. M., 1995, Viscoelasticity of Engineering Materials, Chapman & Hall, New York.

Hadi, M. N. S., and Bodhinayake, B. C., 2003, “Non-linear Finite Element Analysis of Flexible
Pavements,” Advances in Engineering Software, Vol.34, No.11-12, pp.657-662.

Hjelmstad, K. D., Kim, J., and Zuo, Q. H., 1997, "Finite Element Procedures for
Three-dimensional Pavement Analysis," Aircraft Pavement/Technology - In the Midst of
Change, ASCE , Seattle ,Washington, pp.125-137.

Huang, Y.H., 1993, Pavement Analysis and Design, Prentice-Hall Inc., Englewood Cliffs, N.J.

Matthews, M. J. and Schmitt, F. C., 1995, "Creep Tests for The Rutting Design of Field
Highways," Proceedings of the International Conference of the Instrument Society of
America, New Orleans, p 837-847.

Rowe, G. M. , Brown, S. F., Sharrock, M. J., and Bouldin, M. G., 1995, "Viscoelastic Analysis of
Hot Mix Asphalt Pavement Structures, " Transportation Research Record 1482,
Transportation Research Board, Washington, D. C., pp.44-51.

Ioannides, A. M., 1984, “Analysis of Slab-On-Grade for a Variety of Loading and Support
Conditions,” Ph.D.Thesis, University of Illinois at Urbana-Champaign, Champaign, Illinois,

22
 USA..

Ioannides, A.M., Thompson, M. R., and Barenberg, E. J., 1985, “Finite Element Analysis of
Slab-On-Grade Using a Variety of Support Models,” Proceedings of the 3rd International
Conference on Concrete Pavement Design and Rehabilitation, pp. 309-324, Purdue University,
W. Lafayette, Indiana, USA.

Kuo, C.-M., Hall, K. T., and Darter, M. I., 1996, “Three-Dimensional Finite Element Model for
Analysis of Concrete Pavement Support,” Transportation Research Record, No.1505,
Transportation Research Board, Washington, D.C., pp.119-127.

Kuo, C.-M., and Hsu, T.-R., 2003, “Traffic Induced Reflective Cracking on Pavements with
Geogrid-Reinforced Asphalt Concrete Overlay,” Proceedings of the 11th Conference of
Pavement Engineering, Kaoshiung, Taiwan, pp.323-331.

Malvern, L. E., 1969, Introduction to the Mechanics of A Continuous Medium, Prentice-Hall,

Englewood Cliffs, N.J.

HRB, 1962, The AASHTO Road Test Report 5 Pavement Research, Highway Research Board
(HRB), Washington, D. C.

Roberts, F. L., Kandhal, P. S., Brown, E. R., Lee, D. Y., and Kennedy, T. W., 1991, Hot Mix
Asphalt Materials, Mix Design, and Construction, 1st ed., NAPA Educational Foundation,
Lanham, Maryland, USA.

NCHRP, 1990, Calibrated Mechanistic Structural Analysis Procedures for Pavements, Vol. 1 -
Final Report, NCHRP 1-26, National Cooperative Highway Research Program (NCHRP),
Transportation Research Board, University of Illinois at Urbana-Champaign, Construction
Technology Lab., and The Asphalt Institute.

Zaghloul, S., and White, T., 1992, "Use of Three-dimensional Dynamic Finite Element Program
for Analysis of Flexible Pavement," Transportation Research Record, No.1388,
Transportation Research Board, Washington, D. C., pp.60-69.

23

View publication stats