Download as pdf or txt
Download as pdf or txt
You are on page 1of 24

This article was downloaded by: [University of Illinois at Urbana-Champaign]

On: 01 August 2013, At: 21:41


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer
House, 37-41 Mortimer Street, London W1T 3JH, UK

Road Materials and Pavement Design


Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/trmp20

A Simplified Overlay Design Model against Reflective


Cracking Utilizing Service Life Prediction
a b
Mostafa A. Elseifi & Imad L. Al-Qadi
a
Virginia Tech Transportation Institute, 3500 Transportation Plaza, Blacksburg, VA, 24061
E-mail:
b
Virginia Polytechnic and State University, 200 Patton Hall, Blacksburg, VA, 24061 E-
mail:
Published online: 19 Sep 2011.

To cite this article: Mostafa A. Elseifi & Imad L. Al-Qadi (2004) A Simplified Overlay Design Model against Reflective
Cracking Utilizing Service Life Prediction, Road Materials and Pavement Design, 5:2, 169-191

To link to this article: http://dx.doi.org/10.1080/14680629.2004.9689968

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of
the Content. Any opinions and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied
upon and should be independently verified with primary sources of information. Taylor and Francis shall
not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other
liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or
arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
A Simplified Overlay Design Model against
Reflective Cracking Utilizing Service Life
Prediction
Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

Mostafa A. Elseifi* — Imad L. Al-Qadi**


* Virginia Tech Transportation Institute
3500 Transportation Plaza
Blacksburg, VA 24061
melseifi@vt.edu
** Virginia Polytechnic and State University
200 Patton Hall
Blacksburg VA 24061
alqadi@vt.edu

ABSTRACT. Although it is the major mode of failure in rehabilitated pavement structures,


reflection of cracking seldom has been considered in the overlay design process mainly due to
its complexity. This paper presents the development of an overlay design procedure to predict
the service life of rehabilitated flexible pavement structures against reflective cracking. A
simple equation was derived based on three-dimensional (3D) finite element (FE) models and
by utilizing linear elastic fracture mechanics (LEFM) principles. The FE models simulate a
variety of rehabilitated cracked pavement structures. A detailed sensitivity analysis was
performed to establish the accuracy of the FE models. Then, accurate simulation of the crack
singularity was achieved by modeling several contour integral evaluations along the crack
front. Both crack initiation and propagation phases were considered in the formulation. The
crack initiation phase is described using a traditional fatigue law developed by the Belgium
Road Research Center, and the crack propagation phase is described using Paris-Erdogan
phenomenological law. Three contour lines were used around the crack front to calculate the
path-independent J-integral. Calculations of the stress intensity factors based on the J-
integral are presented. An example is presented to demonstrate the use of the developed
design equation in a routine overlay design.
KEYWORDS: Reflective Cracking, Finite Element, Fracture Mechanics, Flexible Pavement,
Overlay Design, Mechanistic Design.

Road Materials and Pavement Design. Volume 5 – No. 2/2004, page 169 to 191
170 Road Materials and Pavement Design. Volume 5 – No. 2/2004

1. Introduction

Hot-mix asphalt (HMA) overlays are typically applied to existing flexible and
rigid pavements when the structural or functional conditions of the pavement have
reached an unacceptable level of service. Most of the overlay design approaches
propose an increase in structural capacity to improve fatigue and/or rutting
pavement service lives (Pierce et al., 1993; Bayomy et al., 1996). However,
adequately designed overlays against rutting and fatigue distresses may still show
Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

cracking patterns similar to the ones which existed in the old pavement after a short
period of time (Jacobs et al., 1992). This distress is known as ‘reflective cracking.’
Although reflective cracking is the most common failure mechanism in rehabilitated
pavements, it is sometimes neglected in the overlay design mainly due to its
complexity. Several methodologies were developed to predict the overlay life
against reflective cracking. The following model is one of these methods (Eltahan
and Lytton, 2001):

 β +1 
 
 β   β 
N fo = ρN fc   [1]
 β +1

where:
– Nfo = number of cycles for crack initiation (defined as the time cracks become
visible in the overlay surface); and
– β and ρNfc = parameters of the models empirically related to the stiffness and
structural design of the pavement system.
This regression equation was developed based on the computed number of cycles
for crack propagation assuming an elastic beam formulation. The model was
calibrated and validated against the field performance of 20 different test sites. It
was reported that Equation [1] can predict the number of cycles for crack initiation
with an absolute error of 18.6%.
There is a growing recognition among pavement engineers that, despite the
complexity of pavement systems (especially when reflective cracking is considered),
an ideal design tool should be based on well-established engineering theories.
However, the final design method should be presented in a simple approach free of
complicated engineering tools such as finite element method (FEM) or fracture
mechanics. This means that simplicity does not justify empiricism.
This paper presents the formulation of a design scheme that considers reflective
cracking as the controlling failure mechanism. Finite element technique and fracture
mechanics principles were used in combination with actual stress measurements at
the Virginia Smart Road to formulate the suggested design equation.
A Simplified Overlay Design Model 171

2. Fundamentals of reflective cracking

Reflective cracks are caused by discontinuities (cracks or joints) in underlying


layers that propagate through an HMA overlay. This propagation is due to
continuous movement at the crack prompted by thermal expansion and traffic
loadings. If the new overlay is bonded to the distressed layer, cracks in the existing
pavement propagate to the surface after a period of time, which is generally function
of the thickness of the overlay.
Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

According to Lytton (1989), the passing of a wheel load over a crack in the
existing pavement causes three critical pulses (one maximum bending and two
maximum shear stresses). As the movement of the crack increases, the propagation
of the crack to the overlay accelerates. Changes in temperature can also contribute to
crack propagation. Contraction and curling of old rigid pavement caused by
temperature variation result in the opening of existing cracks, which induce
horizontal stresses in the HMA overlay. Although seasonal temperature variations
may affect the reflection cracking process, especially when dealing with
rehabilitation of rigid pavements, only reflective cracking due to traffic loadings is
simulated in this study.
To establish if fracture mechanics principles can be applied to discontinuous
crack growth in HMA, Jacobs et al. (1996) compared FEM simulation (continuous
crack growth) to experimental crack growth of HMA (discontinuous crack growth).
Although the level of agreement was highly influenced by the mix nominal
aggregate size (better agreement was found for mixes with small aggregate size), the
crack growth process in HMA might be accurately described using linear elastic
fracture mechanics. In general, a cracked pavement system can be loaded in any one
or a combination of the three fracture modes (see Figure 1):
– Mode I loading (opening mode, KI) results from loads that are applied
normally to the crack plane (thermal and traffic loading).
– Mode II loading (sliding mode, KII) results from in-plane shear loading, which
leads to crack faces sliding against each other, normal to the leading edge of the
crack (traffic loading).
– Mode III loading (tearing mode, KIII) results from out-of-plane shear loading,
which causes sliding of the crack faces parallel to the crack leading edge. Although
this mode may occur if the crack plane is not normal to the direction of traffic, this
mode of loading was neglected in this study for simplicity.
Two distinct phases are considered in the cracking process in pavement systems;
neglecting the ultimate failure stage, in which the crack growth rate increases rapidly
as global instability is approached: the crack initiation and the crack propagation
phases. The crack initiation phase is composed of two distinct phases of
microcracking and formation of macrocracks, and is defined by the necessary
number of load applications to form a visible damaged zone at the bottom of the
overlay (Uzan, 1997). The number of cycles of a specific load a pavement can
172 Road Materials and Pavement Design. Volume 5 – No. 2/2004

withstand before it cracks may be related to the critical strain using a fatigue law.
The direction of the strain depends on the failure mechanism under consideration.
For example, in the case of fatigue cracking, the horizontal tensile strain at the
bottom of the HMA is used. In case of reflective cracking induced by Mode II
loading, the number of cycles for crack initiation may be determined as follows
(BRRC, 1998):

N = 4.856 x10 −14 γ zx −4.76 [2]


Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

where
N = Number of cycles before crack initiation, and
γzx = Shear strains 10mm above the existing crack.

Mode I Mode II Mode III

Figure 1. Modes of fractures in flexible pavements

Equation [2] was developed based on the results of laboratory two-point bending
tests. It assumes that the controlling mode for crack initiation in HMA is Mode II
loading. It is based on the hypothesis that Mode I would only initially propagate the
crack to a short distance, and then becomes ineffective (Uzan, 1997). In contrast,
Mode II loading may propagate the crack to the surface. It is important also to
mention that when the reflection of cracks is considered, the pavement service life
against crack initiation may be much shorter than that resulting from regular
distresses (such as fatigue cracking) since the crack is already well established in the
existing pavement.
A Simplified Overlay Design Model 173

The crack propagation phase represents the stage where the crack propagates to
the surface through the entire thickness of the HMA overlay. A description of the
crack propagation phase in flexible pavements can be based on the empirical power
law developed by Paris and Erdogan (1963):

dc
= A(∆K ) n [3]
dN
where:
Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

c = crack length,
N = number of loading cycles,
A and n = fracture parameters of the material, and
∆K = stress intensity factor amplitude.
It is generally recognized that Paris’ law adequately describes the rate of crack
growth in HMA overlay (Lytton, 1989; Carpenter and Lytton, 1978). Since no exact
definition of the stress intensity factor for a multi-layer pavement system is
available, a regression analysis was performed to define the stress intensity factor as
a function of the crack length (c) for each considered pavement design. The
developed regression models are dependent on the assumed stiffness for the
pavement layers and the crack propagation resistance of the mix.
The computed number of cycles is highly sensitive to the assumed values of the
fracture parameters (A and n). The appropriate way to determine the fracture
parameters of a material (A and n) is to examine the stable crack growth of HMA
beam samples under repeated loading conditions, which is a difficult and expensive
process (Francken, 1993). Since no direct measurements of the fracture parameters
(A and n) were feasible in this study, and since such testing is not expected to be
conducted in a routine overlay design, theoretical relations between the fracture
parameters of the material and its creep properties were used (Schapery, 1982):
2
n= [4]
m

where:
m = slope of the log creep compliance versus log time curve.
As an illustrative example, Figure 2 shows the variation of the creep compliance
with time for a typical mix (SM-9.5D, which is a surface mix with an aggregate
maximum nominal size of 9.5mm and a PG 70-22 binder) at three different
temperatures. As shown in this figure, a linear regression model may accurately
describe the variation of the creep compliance with time. Using the linear regression
equations shown in this figure, one can determine the first fracture parameter (n)
from the slope of the line based on Equation [4]. It may be noticed from this figure
that the first fracture parameter (n) gradually increases with the increase in
temperature indicating a faster crack propagation rate at higher temperatures.
174 Road Materials and Pavement Design. Volume 5 – No. 2/2004

However, it should be emphasized that the second fracture parameter (A) also varies
with the change in temperature. It was also shown that the second fracture
parameters (A) may be determined by means of the mix properties as follows
(Lytton et al., 1993):

log A = −2.605104 + 0.184408AV − 4.704209 log AC − 0.00000066E [5]

where:
AV = air voids (%),
Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

AC = asphalt content (%), and


E = resilient modulus of the mixture (in psi).
Based on this concept, and using results of conventional creep and resilient
modulus testing performed on different types of HMA mixtures, three levels, with
respect to the overlay fracture characteristics, were established:
Type I: n = 3.40 and A = 6.54x10-6 m/cycle.MPa.m0.5
Type II: n = 3.85 and A = 6.88x10-6 m/cycle.MPa.m0.5
Type III: n = 4.50 and A = 8.12x10-6 m/cycle.MPa.m0.5
As indicated in previous studies (Jacob et al., 1996, Mobasher et al., 1997), a
weak or a high air void mixture would typically lie in a Type III category, while a
strong mixture would be better represented through Type I. To accurately evaluate
the number of cycles for the crack to propagate from a location c1 to a location c2,
the stress intensity factor was determined using FEM for different locations of the
crack. The following section explains the stress intensity factor concept, and how the
stress intensity factor was determined using FEM.

1.0E-07

log (D)=0.6936log(t)-8.452
R2=0.96
1.0E-08
Creep Compliance (1/Pa)

log (D)=0.4038log(t)-9.1706
1.0E-09
R2=0.98

log (D)=0.4906log(t)-10.868
1.0E-10
R2=0.93

5°C 25C
40C
1.0E-11
1 10 100
Time (sec)
Figure 2. Variation of the creep compliance for a surface mix
A Simplified Overlay Design Model 175

Fracture mechanics analysis using the finite element method

The fracture behavior of a pavement layer depends on the presence and geometry
of a flaw, the stress field at the flaw, the properties of the layer, and the loading
mechanism around the flaw. To measure the severity and stability of a crack, the
stress intensity factor (K) was utilized. The stress intensity factor is a scale factor
that is used to define the magnitude of the crack tip stress field (Kannine and
Popelar, 1985):
Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

K = σ0 Y a [6]

where
σ0 = applied stress,
a = crack length, and
Y = geometry depending function.
Crack propagation will occur when the stress intensity factor reaches a critical
value KIc, known as the fracture toughness of the material. In contrast to the stress
intensity factor, the fracture toughness is a quantity that is independent of the crack
geometry and the loading imposed on the structure (Kanninen and Popelar, 1985).
To date, fracture toughness for HMA has not been well defined. The plane strain
fracture toughness (KIc), however, was reported for regular HMA at around 0.77
MPa.m0.5 at a temperature of –1°C (Mobashet et al., 1997).
The commercial software ABAQUS 5.8-1 was used to compute the stress
intensity factor (ABAQUS, 1998). This software indirectly calculates the stress
intensity factor using the path independent integral, called J-integral. The J-integral
is defined as the change in mechanical energy per unit area of new crack surface
(Kanninen and Popelar, 1985):

 ∂u 

J =  Udn − T ds 
Γ
 ∂x 
[7]

where:
Γ = a curve that surrounds the crack tip,
U = strain energy density,
n = direction normal to the crack line,
T = traction vector,
u = displacement vector, and
ds = differential element of arc Γ.
For linear elastic materials, and for elastic-plastic materials (when unloading
does not occur), the J-integral should be path independent. This allows the
characterization of stress and strain discontinuities (at the singularity) from results
176 Road Materials and Pavement Design. Volume 5 – No. 2/2004

obtained at some distance from the discontinuities. The accuracy of the results may
then be checked by calculating the J-integral for several contour (path) lines.
ABAQUS uses a domain integral technique to calculate the J-integral. This method
consists of first creating the mesh singularity at the crack tip. This is achieved by
collapsing one side of a continuum element on the crack tip to form triangular
elements. To create a singularity of order r1/2, the mid-side nodes are then moved on
the sides connected to the crack tip to the ¼ point nearest the crack tip. This type of
singularity is suitable for linear elastic problems.
Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

The major advantage of this technique is that the singularity at the location of the
crack is accurately simulated, which allows to calculate the stresses at the vicinity of
the crack. This may not be achieved with ordinary FE models with fine meshes.
After the calculation of the J-integral, the stress intensity factor may be determined
as follows (assuming plane strain condition and omitting the effect of Mode III
loading for the considered pavement structure, Kanninen and Popelar, 1985):

1 − ν2 2
J= (K I + K 2II ) [8]
E

where:
υ = Poisson’s ratio,
E = Elastic modulus, and
KI and KII = Stress intensity factor associated with Mode I and Mode II loading,
respectively.
For simplicity, an equivalent critical stress intensity factor (the maximum value
calculated after passage of the load on top of the crack) was assumed in this study.
As the load travels on top of the cracked area, the variation of the stress intensity
factor represents a combination of Mode I and Mode II. When the load position is at
the edge of the crack, the calculated stress intensity factor represents a combination
of modes I and II. When the load is centered with respect to the crack, the calculated
stress intensity factor is only associated with the opening Mode I loading. The stress
intensity factor (K) was then calculated from the following relation:

1 − ν2 2
J= (K ) [9]
E

where:
K = Stress intensity factor.

3. Finite element formulation

The developed FE codes represent a variety of three-layer systems that are


encountered regularly in typical flexible pavement overlay applications (see
A Simplified Overlay Design Model 177

Table 1). A total of 216 different pavement designs were analyzed to develop the
suggested design equation. An existing pavement structure consists of a cracked
HMA layer, a base layer, on top of a subgrade. A HMA overlay is applied to the
cracked HMA with variable thicknesses. To investigate the crack initiation and
propagation phases, dynamic three-dimensional (3D) models were developed for
different location of the cracks.
Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

Table 1. Overview of investigated cases

Design Parameters Level


Overlay Thickness (mm) 50 100 150
Overlay Modulus (MPa) 3450 (0.25) 4480 (0.25) 5510 (0.25)
HMA Thickness (mm) 100 150 200
HMA Modulus (MPa) 1725 (0.30) 2065 (0.30) 2415 (0.30)
Base Thickness (mm) 150 300 600
Base Modulus (MPa) 205 (0.35) 410 (0.35)
Subgrade Modulus (MPa) 40 (0.40) 135 (0.40)
* Poisson’s ratios are presented in parenthesis.

The dimensions of the modeled portion are 560mm x 38000mm. These


dimensions were selected to reduce any edge effect errors, while keeping the
elements’ sizes within acceptable limits (modeling constraints). Due to the
symmetry in loading and geometry, only half the pavement was considered (see
Figure 3).

(a)
178 Road Materials and Pavement Design. Volume 5 – No. 2/2004
Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

(b)

Figure 3. General layout of the developed finite element model

The generated mesh was designed to give an optimal accuracy (small elements
around the crack and the load, and large elements far from the crack). To improve
the rate of convergence, all elements were 8-node linear brick reduced integration
elements (C3D8R) with the exemption of the crack front, in which 20-node brick
elements (C3D20R) were used. All layers were simulated with the same shape to
preserve the continuity of nodes between consecutive layers.
Infinite elements (CIN3D8) were used to simulate the far field region of the
model in the horizontal directions. Infinite elements can capture the decay of field
variables with respect to the distance from the pole (the center of the loading). The
formulation of infinite element behavior is exactly the same as that of ordinary
elements.

3.1. Sensitivity analysis and element dimensions

When evaluating the results of an FE model, two criteria need to be considered:


– The FE solution has to converge to the continuum model solution. To satisfy
this criterion, a regular mesh refinement process may be used as long as the finest
mesh contains all previous meshes. The FE solution may then be checked against a
simplified solution (the layered theory solution was used in this study for a static
loading case).
A Simplified Overlay Design Model 179

– The accuracy of the FE model has to be acceptable within the contest of the
application. Bathe’s criterion states that an FE mesh is sufficiently fine when jumps
in stresses across inter-element boundaries become negligible (Bathe, 1982). The
jump in stresses may be considered within the same plane or at the interfaces
between different layers.
To ensure the accuracy of the results, several aspects of the FE model were
analyzed and refined until specific criteria were met. Though the in-plane
dimensions (width = 17.5mm and length = 25mm) were selected to reduce the jumps
Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

across inter-element boundaries within the same XY plane, only the results of the
sensitivity analysis for the element thickness (depth of the element) are presented in
this paper as an illustrative example for the reader.
Each layer of elements represents an additional 3360 degrees of freedom to the
model, which characterizes a significant increase in computational time and data
storage space requirements. However, the continuity of stresses at the interface
between the layers is highly affected by the selected element thickness. Therefore,
different element thicknesses were investigated ranging from 3.175mm to layer
thickness.
For a continuum model, no jumps in vertical stresses should occur at the
interface between the layers. Hence, the first criterion used in the evaluation of the
different cases is to determine the jump in vertical stresses that may occur at the
critical interfaces. Figure 4a illustrates the difference in vertical stresses at different
interfaces within the pavement model. As shown in the figure, this problem can be
significantly minimized by appropriate refinement of the mesh. It appears also that
only an element thickness ranging between 3.175mm and 6.35mm would provide an
acceptable level of accuracy.

250
Difference in Stress at the Interface (kPa)

D=38.10mm
200 D=138.10mm
D=188.10mm

150

100

50

0
L 50.8 25.4 12.7 6.35 3.175
Element Thickness (mm)

(a)
180 Road Materials and Pavement Design. Volume 5 – No. 2/2004

Vertical Stress (kPa) at depth=138.1mm 500


3D FE

450 Bisar

400

350
Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

300

250
L 50.8 25.4 12.7 6.35 3.175
Element Thickness (mm)

(b)
Figure 4. (a) Jump in vertical stresses at the critical interfaces and (b) convergence
of the vertical stresses with mesh refinement

To further evaluate the accuracy of each case, similar models were developed for
the same loading and material conditions using BISAR 3.0 (De Jong et al., 1973).
Results presented in Figure 4b show convergence of the vertical stresses with mesh
refinement. However, results of these models do not appear to converge to the
BISAR’s solution (though assumed close to the exact solution) as the mesh is
refined. It is worth noting that the assumptions made in both analysis methods are
slightly different. For instance, the support of the natural soil is simulated
differently: in the layered theory, it is assumed as semi-infinite while in the FEM,
linear springs were used.
Figure 5a illustrates the calculated vertical stresses using 3D FE (element
thickness = 6.35mm) and the BISAR program. The percentage difference between
the calculated vertical stresses using FE and the calculated vertical stresses using the
BISAR program was always less than ± 5%. This good agreement between the FE
and the layered theory solutions for this simplified static uncracked case established
the adequacy of the geometry, mesh, and boundary conditions of the FE model.
Given that the two FE models (element thickness 6.35 and 3.175mm) provide a
comparable level of accuracy, it was decided to use an element thickness of 6.35mm
for all 3D models in this study to save in computational time. Figure 5b illustrates
the variation of the vertical deflections with the distance from the load for the two
approaches (i.e., BISAR and 3DFE). A useful observation as related to the vertical
deflections is that the calculated displacements did not significantly change with
mesh refinement. This fact allows using a coarse mesh in a regular backcalculation
process without jeopardizing the level of accuracy. This is very convenient since a
A Simplified Overlay Design Model 181

regular backcalculation process requires a significant amount of iterations to obtain


an acceptable match between measured and calculated deflections.
Vertical Stress (kPa)
0 100 200 300 400 500 600 700
0
20 3D FE

40 Bisar
Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

60
Depth (mm)

80
100
120
140
160
180
200

(a)
80.0

70.0
Bisar 3D FE
60.0
Deflection (microns)

50.0

40.0

30.0

20.0

10.0

0.0
0 100 200 300 400 500
Distance (mm)

(b)
Figure 5. Calculated (a) vertical stresses and (b) surface deflections based on the
3D FE model and the BISAR’s solution

3.2. Model constraints and contact modeling

Elastic element foundations were used to simulate the support provided by the
subgrade to the pavement structure. These elements, which act as springs to the
ground, provide a simple way of including the stiffness effects of the subgrade
182 Road Materials and Pavement Design. Volume 5 – No. 2/2004

without unnecessary fixation of nodes at the bottom of the model. A foundation


stiffness of 200N/cm3 was assumed for the support provided by the subgrade. This
represents a medium level of resistance.
Contact between the overlay and the existing HMA layer was assumed to be
fully bonded. This assumption was based on the expected good bonding between the
existing HMA layer and the installed overlay due to tack coat application and the
high temperature during HMA placement. Poor bonding between the overlay and the
existing HMA layer may significantly alter the hypothesized mechanism. Under
Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

poor bonding conditions, it is expected that the crack propagates at a slower rate, but
other failure mechanisms such as fatigue of the overlay may accelerate. Bonding
between the HMA and the base layers was assumed to be a friction-type contact
(Mohr-Coulomb theory) with a friction coefficient of 45°.

3.3. Crack simulation

A crack was induced in the existing HMA layer (see Figure 3). To create a
singularity of order r1/2, the mid-side nodes of 20-node brick elements (C3D20R)
along the sides of the cracks were moved to the quarter positions next to the crack
tip. To validate the singularity, all elements around the crack tip were cubic
elements, and were ‘collapsed’ to form pyramidal elements. This focused mesh
allows evaluation of the J-integral through different contour lines (three contour
lines are shown in Figure 3).

3.4. Loading model

A quasi-static approach was adopted in this study. As compared to regular


dynamic methods, the major advantage of this approach is the faster converging rate
and the significant reduction in computational time. Movement of the load at a speed
of 8km/hr was achieved by gradually shifting the loading area over the refined
element path (shown in gray in Figure 3). Vehicular loading was simulated as an
approximate rectangular shape (250mm long by 160mm wide). The load amplitude
function was obtained based on actual vertical stress measurements in the flexible
pavement at the Virginia Smart Road. The measured vertical stress was first
normalized with respect to the maximum-recorded value. The normalized vertical
stress was then multiplied by the average tire pressure expected during movement
(724kPa). This configuration results in a total load of 49.2-kN over a single axle
with single tire. In total, 13 different steps (locations of the load) were needed to
achieve one full passage of the tire over the entire model, which is 0.56m (see
Figure 6). The equivalent axle load factor (EALF) for this configuration relative to
the standard 80-kN single axle load is 1.3378 (Huang, 1993).
A Simplified Overlay Design Model 183

4. Development of design equations

As previously mentioned, the reflection of a crack to the surface is divided into


two distinct phases (neglecting the ultimate failure stage): crack initiation and crack
propagation. The total number of cycles before a crack reflects to the pavement
surface is defined as follows (assuming that global instability is reached when the
crack front is at 12.7mm from the pavement surface):

Ntotal = Ninitiation + Npropagation [10]


Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

Where:
Ntotal = total number of cycles before the crack reach 12.7mm from the surface of
the overlay,
Ninitiation = number of cycles for crack initiation at the bottom of the overlay, and
Npropagation = number of cycles for the crack to propagate from the bottom of the
overlay to 12.7mm from the surface of the overlay.
The following sections explain how each component of Equation [10] was
formulated and incorporated into the design equations.

1
0.9
0.8
Normalized Pressure

0.7
0.6 Measured Vertical Stress
(Normalized)
0.5
0.4
0.3
0.2
0.1
0
0 0.5 1 1.5 2
Time (sec)

Figure 6. Load amplitude function

4.1. Crack initiation

After placement of the overlay, an existing crack will continuously move due to
traffic and thermal stresses until it is able to break in the bottom of the overlay. The
presence of a crack in the pavement system was found to substantially affect the
184 Road Materials and Pavement Design. Volume 5 – No. 2/2004

stress field in the vicinity of the crack tip. In this case, severe straining actions
developed around the crack tip as compared to the rest of the pavement structure.
The number of cycles for crack initiation may be determined using Equation [2],
which requires the calculation of the maximum shear strain in the vicinity of the
crack tip. To adjust the number of cycles for the difference between laboratory and
field conditions (e.g., load distribution in the wander area, state of stresses, and rest
periods), a shift factor of nine was used in all the calculations (Al-Qadi and Nassar,
2003; Al-Qadi, 2004). Figure 7 illustrates the variation of the shear strain with time
Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

as computed using FE for different positions of the load. In this figure, the simulated
speed was 8km/h. As shown in this figure, the maximum and minimum shear strain
will occur when the load position is at the edge of the crack.

200

150

100

50
Shear Strain

0
0 0.05 0.1 0.15 0.2 0.25 0.3
-50

-100

-150

-200
Time (sec)

Figure 7. Variation of the shear strain with time

For each considered pavement structure and at the original position of the crack,
the maximum shear strain was calculated. This allowed calculating the number of
cycles for crack initiation for all pavement structures considered in this study.
Figure 8 illustrates the calculated number of cycles for crack initiation for different
overlay thicknesses and subgrade moduli. As shown in the figure 8, the overlay
thickness is the major factor dictating the pavement performance against reflective
cracking (all other parameters were kept constant in this comparison).
A Simplified Overlay Design Model 185

1.2E+05
Esubgrade=40MPa
1.0E+05 Esubgrade=135MPa
Number of Cycles

8.0E+04

6.0E+04
Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

4.0E+04

2.0E+04

0.0E+00
50 100 150
Overlay Thickness (mm)

Figure 8. Number of cycles for crack initiation for different overlay thickness

4.2. Crack propagation

As previously indicated, the J-integral was obtained for different locations of the
crack in the overlay. An equivalent critical stress intensity factor (the maximum
value calculated after passage of the load on top of the crack) was then calculated
using Equation [9]. Figure 9a illustrates the variation of the stress intensity factor for
a 100mm overlay thickness for different overlay moduli. As presented in this figure,
the stress intensity factor gradually increases until reaching its maximum at the
surface. It is interesting to notice that the stress intensity factor increases with the
increase of the overlay modulus. The same trend was previously reported by Uzan
(Uzan, 1997). This may explain why stone mastic asphalt (SMA) provides superior
performance although its resilient modulus is always reported lower than that of
conventional mixes (Brown, 1992). However, it should be emphasized that the
fracture parameters (A and n) also vary with the material stiffness and crack
resistance, and therefore a lower resilience modulus does not necessarily mean better
performance.
To evaluate the number of cycles for crack propagation using Equation [3],
polynomial regression models were fitted for the variation of the stress intensity
factor with crack length within the overlay for all investigated cases; see Figure 9b.
The coefficient of determination (R2) for all cases was greater than 0.90.
186 Road Materials and Pavement Design. Volume 5 – No. 2/2004

4.3. Service life prediction for pavements with potential reflective cracking

It is clear that although the presented analysis is capable of effectively evaluating


the overlay service life against reflective cracking, it is time consuming and requires
the designer to be familiar with advanced engineering theories (e.g., FEM, and
fracture mechanics). To overcome this major drawback, one objective of this study
was to develop design equations that might be easily used to predict overlay
performance against reflective cracking.
Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

Based on the results of all the considered cases in this study, a regression model
was developed to predict the number of cycles as a function of the significant
variables. The total number of cycles, which varies over several orders of
magnitude, often results in an excessive accumulation of error components and an
overflow in computations. To avoid these problems, the model was dealt with in the
logarithmic domain. The suggested simple regression model is the following:

Log Wt80 = aHoverlay + bEoverlay + cHHMA + dEHMA + eHbase + fEbase + gEsubgrade [11]

where
Wt80 = total number of 80-kN single-axle load applications, a, b, c, d, e, f, and
g = regression constants,
Hoverlay = thickness of HMA overlay (mm),
Eoverlay = modulus of resilience of HMA overlay (MPa),
HHMA = thickness of existing HMA layer (mm),
EHMA = modulus of resilience of existing HMA layer (MPa),
Hbase = thickness of base layer (mm),
Ebase = modulus of resilience of base layer (MPa), and
Esubgrade = modulus of resilience of subgrade (MPa).

8.0

7.0 E=3450MPa
E=4480MPa
6.0
E=5510MPa
K (MPamm0.5)

5.0

4.0

3.0

2.0

1.0

0.0
0 10 20 30 40 50 60 70
Crack Length (mm)

(a)
A Simplified Overlay Design Model 187

4.5
3 2
y = 3E-06x + 0.0002x + 0.0145x + 1.3838
4 2
R = 0.9994
3.5
K (MPa mm )
0.5

2.5

1.5 150mm Overlay (E=3450MPa)


Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

1
Poly. (150mm Overlay
0.5 (E=3450MPa))

0
0 10 20 30 40 50 60 70 80 90
Crack Length (mm)

(b)
Figure 9. Variation of the stress intensity factor with the crack length for (a)
different overlay moduli and (b) a 150mm-overlay design

The interaction between the different variables was also considered, but was
found statistically insignificant. Values of the regression coefficients and goodness
of the fit are as follows:

1
Log Wt80 = [255Hoverlay + 2.08Eoverlay + 45.3HHMA + 8.73EHMA + 1.34Hbase +
10 4
6.93Ebase + 1.49Esubgrade] R2 = 0.98, MSE=0.03 [12]

It is interesting to notice from this equation that the overlay thickness is


undoubtedly the major factor in dictating the overlay performance against reflective
cracking failure, followed by the thickness of the existing HMA layer. Additionally,
it appears that the base thickness and subgrade modulus has the least effect on the
overlay performance. Figure 10 illustrates the comparison between the calculated
design lives using FE and those predicted using Equation [12]. To further investigate
the model accuracy, Figure 11 illustrates a comparison between the relative
differences in lifetime between different structures as predicted by the FE method
and Equation [12]. It is worth noting that each pair of pavement structures was
picked randomly among the 216 investigated cases. In total, 113 pairs are shown in
this figure.
Although the developed performance model was developed considering a single
load, a significant level of interaction is expected between the different loads, their
location with respect to the crack, and the characteristic of the pavement structure.
This interaction was not considered in this study, and an adjustment factor may be
needed to improve the accuracy of the model prediction. In addition, although this
model was exclusively developed based on a variety of three-layer pavement
188 Road Materials and Pavement Design. Volume 5 – No. 2/2004

structures, it may be easily extended to cover a broader number of cases using the
Odemark’s method.

10.0
Log Predicted Number of Cycles

8.0
Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

6.0

4.0

2.0
R2 = 0.98

0.0
0.0 2.0 4.0 6.0 8.0 10.0
Log Calculated Number of Cycles

Figure 10. Comparison between the calculated number of cycles using FEM and the
predicted ones using the developed model

4.0

3.5

3.0
LogNmodeli-LogNmodelj

2.5

2.0

1.5

1.0

0.5

0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
LogNFEi - LogNFEj

Figure 11. Comparison between the calculated relative differences in number


of cycles using FEM and the predicted ones using the developed model for different
pair of investigated cases
A Simplified Overlay Design Model 189

For example, if a subbase was used, it may be converted to an equivalent base


thickness as follows (Ullidtz, 1987):

1/ 3
 E (1 − υ2base ) 
h e = h sub  sub 2  [13]
 E base (1 − υsub ) 

where
Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

he = equivalent base thickness,


hsub = original subbase thickness,
υbase = base material Poisson’s ratio, and
υsubbase = base material Poisson’s ratio.

5. Design example

Assume a pavement structure consists of a 150mm HMA layer


(EHMA= 2500MPa, ν=0.25), 150mm base layer (Ebase = 275MPa, ν = 0.35), 300mm
subbase layer (Esubbase=135MPa, ν=0.35), on top of a subgrade (Esubgrade = 68MPa,
ν = 0.40). Using the Odemark’s method, the subbase layer may be transformed to an
equivalent base layer resulting in a total base thickness of 385mm. The designer
estimates a total accumulative traffic (80-kN repetitions) of 9.0x105 over the service
life of the overlay. Using Equation [12], the required overlay thickness is 75mm
(E = 4480MPa, ν = 0.25) to sustain the estimated traffic.

6. Summary

Although it is the major mode of failure in rehabilitated pavement structures,


reflection of cracking seldom has been considered in the overlay design process
mainly due to its complexity. This study presents a developed model to predict the
service life of an overlay against reflective cracking. The model is based on three-
dimensional (3D) finite element (FE) models and linear elastic fracture mechanics
(LEFM) principles. The developed model utilized a large number of 3D FE cases
that account for both the crack initiation and propagation phases in the overlay on
top of three-layer flexible pavement systems. Using the Odemark method, the
developed model may be expanded to cover flexible pavement systems with more
than three layers. The developed model can be used effectively to predict the
number of cycles to failure against reflective cracking for rehabilitated flexible
pavement structures. Validation and long-term field evaluation of the proposed
model are recommended.
190 Road Materials and Pavement Design. Volume 5 – No. 2/2004

Acknowledgments
The authors would like to thank the Virginia Transportation Research Council of
VDOT, Bekaert Inc., and Maccaferri Inc. for their financial support. The assistance
of A. Loulizi, G. Flintsch, D. Leonard, S. Case, A. Appea, S. Lahouar and
S. Reubush is acknowledged.

7. Bibliography
Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

ABAQUS, Finite element computer program, theory manual, Version 5.8-1, Hibbit, Karlsson
and Sorensen, Inc., Pawtucket, USA, 1998.
Al-Qadi I.L., Nassar W., “Fatigue shift factors to predict HMA performance”, International
Journal of Pavement Engineering, Vol. 4, 2003, p. 69-76.
Al-Qadi I.L., Elseifi M., Leonard D., “Development of an overlay design model for reflective
cracking with and without steel reinforcement”, Journal of the Association of Asphalt
Pavement Technologists, Vol. 72, 2004, p. 388-423.
Bathe K.J., Finite element procedures in engineering analysis, Prentice-Hall, 1982.
Bayomy F.M., Al-Kandari F.A., Smith R., “Mechanically based flexible overlay design
system for Idaho”, Transportation Research Record 1543, TRB, National Research
Council, Washington, D.C., 1996, p. 10-19.
Brown E.R., Evaluation of SMA used in Michigan, National Center for Asphalt Technology,
NCAT Report No. 93-3, Auburn, AL, 1992.
BRRC, Belgian Road Research Center, Design of overlaid cement concrete pavements
reinforced with Bitufor traffic loading, Research report EP5035/3544, 1998.
Carpenter S.H., Lytton R.L., “Procedure for predicting occurrence and spacing of thermal-
susceptibility cracking in flexible pavements”, Transportation Research Record 671,
TRB, National Research Council, Washington, D.C., 1978, p. 39-46.
De Jong D.L., Peatz M.G.F., Korswagen A.R., Computer program BISAR layered systems
under normal and tangential loads, Konin Klijke Shell-Laboratorium, Amsterdam,
External Report AMSR.0006.73, 1973.
Eltahan A.A., Lytton R.L., “Mechanistic-empirical approach for modeling reflection
cracking”, Transportation Research Record 1730, TRB, National Research Council,
Washington, D.C., 2001, p. 132-138.
Francken L., “Laboratory simulation and modeling of overlay systems”, Proceedings of the
2nd International RILEM Conference – Reflective Cracking in Pavements, E & FN Spon,
1993, p. 75-99.
Huang Y.H., Pavement analysis and design, New Jersey, Prentice Hall, 1993.
Jacobs M.M.J., De Bondt A.H., Molenaar A.A.A., Hopman P.C., “Cracking in asphalt
concrete pavements”, 7th International Conference on Asphalt Pavements, International
Society for Asphalt Pavements, Nottingham University, U.K., 1992,
p. 89-105.
A Simplified Overlay Design Model 191

Jacobs M.M., Hopman J.P.C., Molenaar A.A.A., “Application of fracture mechanics


principles to analyze cracking in asphalt concrete”, Proc. Annual Meeting of the
Association of Asphalt Paving Technologists, Vol. 65, Baltimore, MD, 1996, p. 1-39.
Kanninen M.F., and Popelar C.H., Advanced fracture mechanics, Oxford University Press,
Inc., NY, 1985.
Lytton R.L., “Use of geotextiles for reinforcement and strain relief in asphalt concrete”,
Geotextiles and Geomembranes, Vol. 8, 1989, p. 217-237.
Lytton R.L., Uzan J., Fernando E.G., Roque R., Hiltumen D., Stoffels S.M., “Development
Downloaded by [University of Illinois at Urbana-Champaign] at 21:41 01 August 2013

and validation of performance prediction models and specifications for asphalt binders
and paving mixes”, SHRP A-357, TRB, National Research Council, Washington, D.C.,
1993.
Mobasher B., Mamlouk M.S., Lin H-M., “Evaluation of crack propagation properties of
asphalt mixtures”, Jal of Transportation Engineering, Vol. 123, No. 5, 1997, p. 405-413.
Paris P.C. Erdogan F.A., “Critical analysis of crack propagation laws”, Transactions of the
ASME, Journal of Basic Engineering, Series D, No. 3, 1963.
Pierce L.M., Jackson N.C., Mahoney J.P., “Development and implementation of a
mechanistic, empirically-based overlay design procedure for flexible pavements”,
Transportation Research Record 1388, National Research Council, 1993, p. 120-129.
Schapery R.A., “Models for damage growth and fracture in nonlinear particulate composites”,
Proceedings of the 9th US Congress of Applied Mechanics, ASME, 1982.
Ullidtz P., Pavement Analysis, New York, Elsevier Science, 1987.
Uzan J., “Evaluation of fatigue cracking”, Transportation Research Record 1570, TRB,
National Research Council, Washington, D.C., 1997, p. 89-95.

Received: 12 September 2003


Accepted: 2 April 2004

You might also like