Professional Documents
Culture Documents
Since the linear elastic layer solution for the layered systems was developed in the 1940s, the
linear elastic layer analysis has been systemized and widely used for the design of roadway
pavements as a tool for evaluating their structural soundness. The primary assumption made in
the analysis is the layered system consisting of materials that are linear elastic, and hence, an
application of the elastic layer analysis to asphalt mixtures, which is a well-known viscoelastic
material, has been limited. Therefore, the intention of the study was to derive a viscoelastic
solution able to take into account the time- and rate-dependent nature of the viscoelastic
materials in the multilayered system. In this paper, a linear viscoelastic solution for the
multilayered system subjected to a cylindrical unit step (static) load was derived from the elastic
solution using the principle of elastic-viscoelastic correspondence and the numerical inversion of
Laplace transforms. The solution was then extended to simulating pavement responses subjected
output from the viscoelastic solution was confirmed by comparing them to those of the Finite
Element Analysis (FEA). Compared to the time and effort required in FEA, the analysis based on
the viscoelastic solution was faster and the results more reliable. Therefore, it is expected that the
viscoelastic solutions derived in this study will be an effective tool for the design of flexible
pavements.
solutions
INTRODUCTION
Designing pavement structures that are safe, reliable, and economical requires both efficient use
structures in terms of stress and/or strain have been evaluated by the use of linear elastic layer
analysis programs, such as Bisar (De Jong et al. 1973) and Kenlayer (Huang 1993). These were
developed based on the well-known linear elastic theory for layered systems (Burmister 1945).
With the advent of the modern electric digital computer, the elastic layer analysis has been
systemized and widely used for the analysis and design of flexible pavements.
For the past few decades, the stiffness of asphalt concrete mixtures used for roadway
design and construction has been commonly characterized by means of resilient modulus,
defined as the ratio of the applied stress to the recoverable strain (for example, Roque et al.
1997). The loading function for this test, typically consisting of a 0.1-second haversine curve
followed by a 0.9-second rest period, was aimed at simulating the pattern of traffic loading over
pavement structures. By treating the resilient modulus as an elastic modulus, the structural
soundness of pavement has been evaluated through the elastic layer analysis. The resilient
modulus, however, is not a fundamental material property, and hence, the concept of resilient
modulus has been subsequently diminished in the latest Mechanistic-Empirical Pavement Design
Guide (ARA., ERES Division 2004). Instead, the new design guide uses the complex modulus,
representing the creep behavior of asphalt mixtures, as the primary measure of asphalt mixtures
in the procedure for both new construction and rehabilitation projects. It is, however, noted that
the new design guide still uses the linear elastic layer analysis as the primary model for the
evaluation of pavement performance. It is obvious that if the model of materials is changed, then
the structural model that is able to take their behavior into account should be used.
2
Therefore, the study was undertaken to develop a structural analysis model that is able to
integrate the viscoelastic behavior of asphalt mixtures into the multilayered system. The paper
begins with an overview of the linear elastic layer solution. Then, the corresponding linear
viscoelastic solution for the multilayered system subjected to a cylindrical unit step load is
derived. Because the behavior of viscoelastic materials is rate and time dependent, meaning that
their response is significantly influenced by the speed of traffic loading, the viscoelastic solution
is extended to formulate the structural behavior of the multilayered pavement system subjected
to a moving load.
Figure 1 illustrates the geometry and loading conditions of a multilayered system. Burmister
(1945) derived the general solution for the two- and three-layered systems based on the three
the layered systems. By employing the theory of elasticity for the three-dimensional problem in
axisymmetric coordinates (Love 1927), his derivation begins with an assumption of a stress
function, F, with the Bessel function, J, which satisfies the equations of equilibrium and
F = J 0 (mr )[ Ai (m) exp(mz ) - Bi (m) exp(- mz ) + C i (m) × z × exp(mz ) - Di (m) × z × exp(- mz )] (1),
to which the subscripts i were added for the purpose of extending the elastic solution to the
multilayered system. The coefficients in Equation (1) were considered as functions of m because
Burmister essentially derived the elastic solution for the surface load of -mJ0(mr).
The following stress and displacement equations are obtained by substituting the stress
function into the general stress and displacement equations for the asymmetric conditions. Note
3
that the superscript * is an indication of the stress or displacement due to the surface load, -
mJ0(mr), not due to the uniform, cylindrical load q distributed over a circular area of radius a.
strain e z* i ( z , r ) , which will be used in the subsequent sections, are also given as:
(2-5)
4
ì Ai (m) × m + Bi (m) × m × exp(-2m × z ) ü
* (1 + m i ) 2 ï ï
e z i ( z, r ) = - J 0 (mr )m exp(mz )í- C i (m) × (1 - 4 m i - mz ) ý (2-7)
Ei ï+ D (m) × (1 - 4 m + mz ) × exp(-2m × z )ï
î i i þ
The equations above are the stress, strain, and displacement equations for the elastic multilayered
system in which Ai(m) and Ci(m) are equal to zero at the lowest layer of n to satisfy the boundary
In order to determine the coefficients, Ai(m), Bi(m), Ci(m), and Di(m), the sets of
simultaneous equations satisfying all the boundary conditions of the multilayered system need to
be solved. For the efficiency of computation analysis, the simultaneous equations are commonly
expressed in matrix form. For the continuity conditions between the layers, the coefficients
· At the surface, z =0, sz* = -mJ0(mr) and the shear stress trz* is zero. By substituting these
conditions into Equations (2-1) and (2-2), a set of the simultaneous equations satisfying those
é A1 (m) ù
ê B ( m) ú 1
[M 1 (m)] × ê 1 ú = éê ùú (3-1)
ê C1 (m) ú ë0û
ê ú
ë D1 (m)û
where,
m m - (1 - 2 m1 ) (1 - 2 m1 )ù
[M 1 (m)] = éê (3-2)
ëm -m ( 2 m1 ) (2 m1 ) úû
· For the continuous boundary conditions at the interfaces at z = h1, h2…hn, the vertical stress,
shear stress, vertical displacement, and horizontal displacement of the upper layer must be
the same as those of the lower layer. A series of simultaneous equations satisfying those
5
boundary conditions at interfaces between the upper and lower layers can be also expressed
é Ai (m) ù é Ai +1 (m) ù
ê B ( m) ú ê B ( m) ú
[M i (m)] × ê i ú = [M i +1 (m)] × ê i +1 ú (4-1)
ê C i ( m) ú ê C i +1 (m) ú
ê ú ê ú
ë Di (m)û ë Di +1 (m)û
where,
[M i (m)] =
ém exp(m × hi ) m exp(- m × hi ) - (1 - 2 m i - mhi ) exp(m × hi ) (1 - 2 m i - mhi ) exp(- m × hi ) ù
êm exp(m × h ) - m exp( - m × h ) ( 2 m + mh ) exp( m × h ) ( 2 m - mh ) exp( - m × h ) ú
ê i i i i i i i i ú
ê (1 + m i ) (1 + m i ) (1 + m i ) (1 + m i ) ú
ê- m exp(m × hi ) m exp(- m × hi ) (2 - 4 m i - mhi ) exp(m × hi ) (2 - 4 m i + mhi ) exp(- m × hi )ú
ê Ei Ei Ei Ei ú
ê (1 + m i ) (1 + m i ) (1 + m i ) (1 + m i ) ú
ê- m exp(m × hi ) - m exp(- m × hi ) - (1 + mz ) exp(m × hi ) (1 - mz ) exp(- m × hi ) ú
ë Ei Ei Ei Ei û
(4-2)
· Because the stresses and displacements must disappear as z approaches infinity, ¥, the above
é An -1 (m) ù
ê B ( m) ú B ( m)
[M n-1 (m)] × ê n-1 ú = [M n (m)] × éê n ùú (5)
ê C n -1 (m) ú ë Dn ( m) û
ê ú
ë Dn -1 (m)û
· By successive multiplications, the coefficients for the lowest layer can be related to Equation
(3-1):
é Bn ( m) ù -1 é1 ù
[
ê D (m)ú = [M 1 (m)] × [M 2 (m)] × [M 3 (m)] × [M 4 (m)] L[M n (m)] × ê0ú
-1 -1
] (6)
ë n û ë û
6
· Once Bn(m) and Dn(m) at the lowest layer are obtained from the above, they are substituted
into Equation (5). Then, the coefficients of the upper layers are successively determined by
é An -1 (m) ù
ê B ( m) ú
ê n -1 ú = [M (m)]-1 × [M (m)] × é Bn (m) ù (7-1)
ê C n -1 (m) ú n -1 n ê D ( m) ú
ë n û
ê ú
ë Dn -1 (m)û
é A1 (m) ù
ê B ( m) ú
ê 1 ú = [M (m)]-1 L[M (m)] × [M (m)]-1 × [M (m)] × é Bn (m) ù (7-2)
ê C1 (m) ú 2 n-2 n -1 n ê D ( m) ú
ë n û
ê ú
ë D1 (m)û
· To find the stresses, strains, and displacements caused by the cylindrical load of q distributed
over the circular area of radius a, the Hankel transform method is often employed (Huang
1993). If R* is the stress, strain, or displacement in Equation (2), the stress, strain, or
displacement R corresponding to the loading condition can be obtained through the Hankel
¥
R( z, r ) *
R( z, r ) = q × a × ò J 1 (m × a )dm (8)
0
m
In order to derive a viscoelastic solution for the multilayered system shown in Figure 2, the
elastic-viscoelastic correspondence principle was employed for materials that are homogeneous,
isotropic, and linearly viscoelastic. The elastic-viscoelastic correspondence principle states that
the form of a relationship between stress and strain for a linear elastic material is the same as the
Laplace transform of the relationship for the corresponding linear viscoelastic material. The
7
procedure of applying the elastic-viscoelastic correspondence principle to known elastic
solutions is well documented in Findley et al. (1976). The steps for applying the elastic-
viscoelastic correspondence principle to the elastic solution derived in the previous section are
described below:
· Take the Laplace transforms of the elastic stress, strain, and displacement equations shown
above in which the coefficient Ai(m), Bi(m), Ci(m), and Di(m) are dependent on time because
they include the time-dependent material properties E(t) and m(t), and replace E and m with
sEˆ ( s ) and smˆ ( s ) where a caret (Ù) over the symbols means that the quantity has been
transformed and is a function of the Laplace variables, s. The Laplace transforms of the
equations are:
8
é ì Aˆ i (m, s ) × m 2 exp(m × z ) + Bˆ i (m, s ) × m 2 exp(- m × z )ü ù
ê ïï ïï ú
êmJ 0 (mr )í+ Cˆ i (m, s ) × m(1 + 2 smˆ i ( s ) + mz ) exp(m × z ) ý ú
ê ï ˆ ï ú
ê ïî- Di (m, s ) × m(1 + 2 smˆ i ( s ) - mz ) exp(- m × z ) ïþ ú
sˆ r* ( z, r , s) = ê ú (9-5)
i
ê ì Aˆ i (m, s ) × m 2 exp(m × z ) + Bˆ i (m, s ) × m 2 exp(- m × z )üú
ê J (mr ) ïï ˆ ïïú
ê- m 1 í+ C i (m, s ) × m(1 + mz ) exp(m × z ) ýú
ê mr ï ïú
êë ïî- Dˆ i (m, s ) × m(1 - mz ) exp(- m × z ) ïþúû
ì Aˆ i (m, s ) × m 2 exp(m × z ) ü
ï ï
(1 + smˆ i ( s )) æ J 0 (mr ) - J 2 (mr ) öï+ Bˆ i (m, s ) × m exp(- m × z )
2
ï
eˆr* ( z, r , s) = - mç ÷í ý (9-7)
i
sEˆ i ( s ) è 2 øï+ Cˆ i (m, s ) × m(1 + mz ) exp(m × z ) ï
ï ˆ ï
î- Di (m, s ) × m(1 - mz ) exp(- m × z )þ
(9-6)
· The same scheme outlined above should be applied to the simultaneous equations expressed
in the matrix form and also to the time-dependent material properties inside the matrices such
that:
é Aˆ1 (m, s ) ù
êˆ ú
B1 (m, s ) ú é1ù
[
ˆ ]
M 1 (m, s ) × ê =
ê Cˆ (m, s ) ú êë0úû
(10-1)
1
ê ú
êë Dˆ 1 (m, s )úû
where,
and
9
é Aˆ i (m, s ) ù é Aˆ i +1 (m, s ) ù
êˆ ú êˆ ú
Bi (m, s ) ú Bi +1 (m, s ) ú
[ˆ ]
M i (m, s ) × ê
ê Cˆ (m, s ) ú
[
= Mˆ i +1 (m, s ) ]
× ê
ê Cˆ (m, s ) ú
(11)
ê i ú ê i +1 ú
êë Dˆ i (m, s )úû êë Dˆ i +1 (m, s )úû
· As shown in the elastic case, the coefficients for the lowest layer can be related to Equation
(10-1) as:
êˆ =[ ˆ
M
ú ê 1 ( m, s ) ][
× ˆ
M 2 ( m, s ) ] [
-1
× ˆ
M 3 ( m, s ) × ][
ˆ
M 4 ( m, s )
-1
L ˆ
M n] [
( m, s )
úû ê0ú
]
ë Dn (m, s )û ë ë û
· After Bn(m) and Dn(m) are obtained from the above equation, the coefficients of the layers
é Aˆ n -1 (m, s ) ù
êˆ ú
ê Bn -1 (m, s ) ú = Mˆ (m, s ) -1 × Mˆ (m, s )
ê Cˆ (m, s ) ú
[ n -1 ] [n ]× éêDBˆˆ ((mm,, ss))ùú
n
(13-1)
ê n -1
ú ë n û
êë Dˆ n -1 (m, s )úû
é Aˆ1 (m, s ) ù
êˆ ú
ê B1 (m, s ) ú = Mˆ (m, s ) -1 L Mˆ (m, s ) × Mˆ (m, s ) -1 × Mˆ (m, s )
ê Cˆ (m, s ) ú
[2 ] [ n-2 n -1 ][ n ] [ ]× éêDBˆˆ ((mm,, ss))ùú
n
(13-2)
ê 1
ú ë n û
êë Dˆ 1 (m, s )úû
· All the coefficients of each layer are now functions of m and s. To find the stresses, strains,
and displacements caused by a cylindrical unit step load of q(t) = qH(t) distributed over the
circular area of radius a, where H(t) is the Heaviside step function, the Hankel transform is
10
Rˆ ( z , r , s ) *
¥
q
Rˆ ( z , r , s ) = × a × ò J 1 (m × a )dm (14)
s 0
m
Taking the inversion of the Laplace transforms is the final step in deriving a viscoelastic solution
for the multilayered system. One of the obstacles in deriving a viscoelastic solution by means of
the typical Laplace transform approach has been the complexity of inverting the Laplace
transforms, and hence, this study considered the numerical inversion technique to be used for
deriving the viscoelastic solution for the multilayered system. The numerical inversion of
Laplace transforms has been continuously improved with the advent of the modern electric
computer. There are nearly over 100 algorithms available for the numerical inversion. Among
them, this study used the inversion process of Laplace transforms, namely a Fixed Talbot
Algorithm (FTA), which was proposed by Abate and Valko (2004), because of its accuracy,
g + i×¥
1
f (t ) = × ò exp( s × t ) fˆ ( s )ds (15)
2p × i g-i ×¥
Many studies (for example, De Hoog et al. 1982) change Equation (15) to a form of Fourier
series consisting of the real and imaginary parts in order to make a better shape for numerical
¥ ¥
exp(gt ) exp(gt )
f (t ) =
p ò [Re{fˆ ( p)}cos wt - Im{fˆ ( p)}sin wt ]dw = p ò [Re{fˆ ( p) exp(iwt )}]dw
0 0
(16)
where p = g+iw. The appearance of the above inversion formula looks familiar; however, it is
still not an effective function for numerical integration because of the infinite upper boundary. In
11
order to facilitate the numerical integration process, Abate and Valko (2004) introduced the
following form in which they converted the inversion formula to the complex domain based on
f (t ) =
p
{ }
a (t ) éRe exp(t × a (t ) × q × (cot q + i )) × fˆ (a (t ) × q × (cot q + i )) ù
údq (17)
p ò0 ë´ {1 + i (q + (q cot q - 1) cot q )}
ê
û
2M
where a (t ) = and q is the angle in radians. As seen in the above, the upper limit is bounded
5t
as p, indicating that numerical integration would converge onto the expected value at a much
faster rate than that of the Fourier series approach (Equation (16)). The integral in Equation (17)
é éì kp kp ü ùù
ê êïïexp(t × a (t ) × M × (cot M + i ))ïï úú
ê êí ý úú
a (t ) ê exp(a (t ) × t ) ˆ M -1
êï´ fˆ (a (t ) × kp × (cot kp + i )) ï úú
f (t ) = ê f (a (t )) + å Re êï ï úú (18)
M ê 2 î M M þ
k =1
ê úú
ê ê´ ì1 + i ( kp + æ kp cot kp - 1) cot kp öüú ú
ê ê í ç ÷ý
ë ë î M èM M M øþúû úû
where M, the number of terms to be summed, means the number of precision decimal digits and
is to control the round-off error in the computation of Equation (18). In other words, the higher
the number used, the more precision can be achieved, but the longer computation time is
required; therefore, the selection of a suitable number M is important in order to accomplish the
efficient computation analysis. The estimation of the computation error regarding the value of M
f (t ) - f (t , M )
» 10 -0.6 M (19)
f (t )
The equation above indicates that the selection of M in the range between 5 and 10 would
provide a reasonable estimation of stress, strain, or displacement for most engineering problems.
12
Last, R( z , r , t ) is obtained by substituting Rˆ ( z , r , s ) in fˆ (s ) shown above. A complete form of the
é é ì kp kp ü ùù
ê ê ïïexp(t × a (t ) × M × (cot M + i )) ïï ú ú
ê êReí ý úú
a (t ) ê exp(a (t ) × t ) ˆ M -1
ê ï´ Rˆ ( z , r , a (t ) × kp × (cot kp + i ))ï ú ú (20)
R( z, r , t ) = ê R ( z , r , a (t )) + å ê ï ïþ ú ú
M ê 2 î M M
k =1
ê úú
ê ê´ ì1 + i ( kp + æ kp cot kp - 1) cot kp öüú ú
ê ê í ç ÷ý
ë ë î M èM M M øþúû úû
By replacing R ( z , r , t ) with stress, strain, or displacement, any of these can be calculated at any
Pavement Geometry
For the axisymmetric condition, which was a primary basis of derivation of the viscoelastic
solution for the multilayered system, a pavement system consisting of four layers including the
rigid layer at its bottom and subjected to the circular unit step load with a uniform stress of 690
kPa and a contact radius of 152.4 mm was considered to be analyzed, thicknesses were assumed
to be typical values used for the design of pavements, and all layers were assumed to be fully
bonded, which is identical to the assumption made in the derivation above (Figure 3a). To
confirm the suitability of output from the viscoelastic solution derived in the previous section,
Finite Element Analysis (FEA) was chosen and performed for the same pavement structure but
with a finite width of 2.54 m. The width was estimated based on the previous work done by
Thompson (1982). The study recommended that the minimum distance of the width should be at
least 12 times the radius of the applied load. By introducing the fixed boundary conditions
13
underneath the subgrade layer (that is, this would act as the rigid layer), the three-layered
From the literature, several studies (Al-Qadi et al. 2005, Elseifi et al. 2006, Kim et al.
2007, and Kim et al. 2009) used viscoelastic analysis, which considered the asphalt layer as
viscoelastic and the other layers as elastic using the Finite Element Method (FEM) based on a
flexible pavements. For the pavement structures (Figure 3), the same method was employed both
in the FEA and the analysis based on the viscoelastic solution. Except for the asphalt layer, the
linear elastic material properties for the base, subgrade, and rigid layers of the given pavement
structure were assumed to be the typical values used for the design of pavements (Table 1).
This section describes the viscoelastic material properties used for the asphalt layer. Data
obtained from the previous studies was used as material input in both viscoelastic analyses. A
full description regarding laboratory testing can be found in the references cited below.
Lee and Kim (2009) and Kim and West (2010) proposed a methodology for determining
the viscoelastic creep compliance D(t) and Poisson’s ratio m(t) from the complex modulus test
using the indirect tension testing (IDT) mode. For the complex modulus test, field cores with a
diameter of 150 mm were cut to a thickness of 38 mm. Complex modulus tests were conducted
at various frequencies—0.01 Hz, 0.1 Hz, 0.5 Hz, 1 Hz, 5 Hz, 10 Hz, and 25 Hz—at a
temperature of 10ºC. In the studies, the authors concluded that the creep compliance and
Poisson’s ratio of asphalt mixtures can be represented by the well-known power functions as:
14
where D0, D1, m, m0, m1, and n are regression coefficients. Taking the Laplace transformations
D0 D1 × G(1 + m)
D( s) = + (22-1)
s s 1+ m
m0 m1 × G(1 + n)
m (s) = + (22-2)
s s 1+ n
Because creep compliance D(t) and relaxation modulus E(t) are two aspects of the same
viscoelastic behavior of materials, they are interchangeable using the following relationship
1 1 s m -1
Eˆ ( s ) = 2 ˆ
= = m (23)
s D( s) D G(1 + m) s D0 + D1G(1 + m)
sD0 + 1 m -1
s
Equations (22-2) and (23) were used as input of the viscoelastic solution in Equations (9) and
(10), mˆ1 ( s ) and Eˆ 1 ( s ) . The values of the power model parameters, D0, D1, m, m0, m1, and n,
viscoelastic shear and bulk moduli of given viscoelastic materials in forms of Prony series as
their input (Kim et al. 2010). The Prony series representation of creep compliance is of the
N
ì t ü
D(t ) = D0 + å Di í1 - exp(- )ý (25)
i =1 î ti þ
where D0, Di, and h are Prony series parameters, and ti are retardation times. Since the Prony
series also can be used as an analytical representation of the time-dependent Poisson’s ratio of
viscoelastic materials (Kim and West 2010 and Kim et al. 2010), the time-dependent Poisson’s
15
ratio is, therefore, expressed as the identical form used in the creep compliance but with different
coefficients, m i:
N
ì t ü
m (t ) = m 0 + å m i í1 - exp(- )ý (26)
i =1 î ti þ
Lee and Kim (2009) and Kim and West (2010) also proposed a fitting technique for determining
the above Prony series from the power functions (Equation (21)). In fitting the creep compliance
and Poisson’s ratio using the Prony series shown in Equations (25) and (26), Prony series with
one spring element and six Kelvin elements (N=6) and retardation times with one decade interval,
ti = 10(i-4) (i = 1,¼ 6) were used, and then the unknown coefficients were determined by solving
the linear system of equations. According to the theory of linear viscoelasticity, the relationships
among the relaxation, D(t), shear, G(t), and bulk, K(t), moduli are expressed as algebraic
Ù 1
G (s) = Ù
(27-1)
2 s (1 + s m ( s )) Dˆ ( s )
2
Ù 1
K (s) = Ù
(27-2)
3s (1 - 2 s m ( s )) Dˆ ( s )
2
After substituting Laplace-transformed Equations (25) and (26) into Equation (27), taking the
2 N -2
G (t ) = G0 + å G (exp(-r t ))
i =1
i i (28-1)
2 N -2
K (t ) = K 0 + å K (exp(-r t ))
i =1
i i (28-2)
The shear and bulk moduli determined through Equation (28) were used in FEA. The
16
Discretization of the FE model
In this study, the commercial FE software, ADINA 2008, was exclusively used. For the
axisymmetric stress condition, the rectangular element with nine nodes that has been
successfully used in two-dimensional FE analyses was used as the element type of model. A
typical meshing technique that achieves a finer mesh in the targeting area while keeping a
relatively coarser mesh in the remaining area was used as shown in Figure 3b. The numbers of
elements used for the asphalt, base, and subgrade layers were 696, 1044, and 2088, respectively.
The development of the time-dependent viscoelastic response depends not only on the
current state of stress or strain but also on the full history of its development. Considering the
nature of the numerical method (FE method), this indicates that time intervals between
computation points should be small enough to make an accurate prediction of the strain or stress
response. In order for the FE model to achieve such a condition, a loading time of 10 seconds
with an equal time interval of 0.01 seconds was applied. As a result, a total of 100 computation
Analysis Results
Because the radial stress or strain at the bottom of the asphalt layer and the vertical stress or
strain on the top of the subgrade layer have been considered critical responses and been used for
evaluation of cracking and rutting performance of flexible pavement systems, the stresses and
strains at each computation point compared to those from the viscoelastic solution at the same
time are presented in Figures 4 and 5 in which positive numbers means tension. Although a
slight gap between the values was observed, as we expected, because of the different nature of
FEM and the finite boundary conditions used for the FE model, not only the values but also the
trends were very close to each other. These observations appear to be sufficient for verifying the
17
soundness of the viscoelastic solution derived for the multilayered system; therefore, these lead
to the conclusion that the viscoelastic solution is a closed-form solution of the viscoelastic
As mentioned earlier, the value of M in Equation (18) could be determined in the range
between 5 and 10. Based on the estimation, a minimum value of 5 was used in the viscoelastic
analysis performed in order to expedite the computation process. It is, however, of interest to
know to what degree the assumption of the M value affects the stress and strain predictions.
Further evaluation was performed by varying the M values. Two extreme levels of M—5 and
10—were selected, and the same analysis with the M value of 10 was performed. Figures 6 and 7
show the stresses and strains at the bottom of the asphalt layer and at the top of the subgrade
layer in which the correlation in each case was evaluated by means of R2. For all cases, the R2
values were 1.00000. No difference between the values was observed visually or statically,
indicating that the M value of 5 is the suitable number of the summation in Equation 20.
The viscoelastic solution derived in the preceding was extended to simulation of a moving load.
To formulate the responses of pavements to the moving load, the study used the Boltzmann’s
superposition principle (Findley et al. 1976 and Tschoegl 1989). For uniaxial loading conditions
shown in Figure 8, the Boltzmann’s superposition principle states that the sum of the strain
output resulting from each stress input is the same as the strain output resulting from the
combined stress input. Therefore, the strain output at any time tn corresponding to any arbitrary
n
e (t n ) = s i × D(t n - t 0 ) + å [× s i D(t n - t i ) - s i -1 D(t n - t i -1 )], for t n = t1 Kt N (29)
i =1
18
Note that the same argument applies when step changes or arbitrary changes in strain are applied
extended to the derivation of a viscoelastic solution for the multilayered system subjected to a
moving load. The concept is illustrated in Figure 9. First, let us consider the moving load,
beginning at time t0 and distance r0 apart from the point P shown in Figure 9a. Suppose we
measure the responses of the given structure at the point P. At time tn, one of the structural
responses, defined as R ( z , r0 , t ) in the previous section, because of the load of qH(t) becomes
R( z , r0 , t n - t 0 ) similar to the first term in Equation (29). Figure 9b shows that the load was
shifted to the second location r1. At time tn, the responses at point P resulting from the combined
load input—one at time t0 and others at time t1—are summed similar to the strain at time t1 in
R ( z , rn , t n ) = R ( z , r0 , t n - t 0 ) + R ( z , r1 , t n - t1 ) - R ( z , r0 , t n - t1 ) (30)
If the above superposition principle is successively applied to multiple loading positions, ri, at
n
R ( z , rn , t n ) = R ( z , r0 , t n - t 0 ) + å R ( z , ri , t n - t i ) - R ( z , ri -1 , t n - t i ) (31)
i =1
Given that time is related to the distance and speed of the moving load, the time ti is expressed as
follows:
xi -1 - xi
ti = + t i -1 (32)
V
where V is the speed of the moving load, and xi is the distance of the moving load from the point
P¢ in the loading axis, x, shown in Figure 10 (i.e. an independent loading axis should be used if
19
the point P is not placed in the loading axis). Assuming that time t0 at the beginning is zero, the
i x j -1 - x j
ti = å (33)
j =1 V
Substituting the above into Equation (31), the time variables can be removed and are replaced by
é x0 - x n ù
ê R ( z , r0 , )+ ú
x0 - x n ê V ú
R ( z , rn , )=ê n ì üïú
V x - xn i x j -1 - x j x0 - x n i x j -1 - x j
êå ïí R ( z , ri , 0 -å ) - R ( z , ri -1 , -å ) ýú
ê i =1 ï V j =1 V V j =1 V ïþúû
ë î
(34)
Note that because the distance from point P¢ could be a negative or positive value, the absolute
value of the difference between loading positions needs to be evaluated in Equation (34).
Finally, any structural response, such as stress, strain, or displacement, expressed in terms
of R at any loading position loaded at any speed can be obtained through Equation (34).
For the pavement structure used above, a moving load starting from -2.54 m (r0 = x0 = -2.54 m)
apart from point P (=P¢ ) was shifted every 50.8 mm (ri-ri-1 = xi-xi-1 = 50.8 mm) to the loading
direction. A total 100 shifts was made to simulate one pass of the moving load. Figure 11a shows
a graphical description of the technique used in this analysis. Figure 11b shows the change in
tensile (radial) stress monitored at the bottom of the asphalt layer in the loading axis over one
pass of the moving load traveling at 60 km/h. As seen in the plot, the stress distribution is similar
to what we observe from the elastic analysis, whereas the stress distribution is not symmetric
around the peak stress because of the viscoelastic effect. Table 3 shows that the change in the
20
radial stress and strain at the bottom and the change in the vertical stress and strain on the top of
the subgrade layer at different speeds varied from 10 to 100 km/h. The stresses and strains were
monitored in the loading axis when the moving load passed point P (ri=xi=0). For all cases, the
tensile stress at the bottom of the asphalt layer increases, whereas the others decease as the speed
of loading increases. For the bottom stress and strain, the observations agree well with the
general behavior of viscoelastic materials where strain decreases under a fixed level of loading as
the applied loading frequency increases. For the top stress and strain at the subgrade layer, the
results also agree well with the general trend that the higher the asphalt layer modulus, the less
vertical strain and stress occur because of the reduction in deflection caused by the increase in
modulus.
These observations lead to the conclusion that the viscoelastic solution derived for the
moving load has a capability of simulating the pavement responses subjected to the moving load.
The benefit of the viscoelastic solution is that it significantly saves time and effort in evaluating
the structural responses of a pavement in comparison to those typically required to perform FEA
for the same structure. Therefore, it would be a fast and reliable tool for the design of flexible
pavement structure.
A general viscoelastic solution for the multilayered system was derived from the known elastic
solution using the principle of the elastic-viscoelastic correspondence and the numerical
inversion process of Laplace transforms. In order to take the time- and rate-dependent nature of
viscoelastic materials into account, the solution was then extended to simulation of the pavement
responses subjected to a moving load. The findings of this study are summarized as follows:
21
· To confirm the suitability of output from the viscoelastic solution for the cylindrical unit step
(static) load, a pavement system consisting of four layers including the rigid layer was
chosen, and the Finite Element Analysis (FEA) was performed for the same pavement
structure but with finite boundary conditions. Although a slight gap between the values was
observed because of the different nature of the FEA and the analytical approaches, the result
was very close and sufficient for verifying the soundness of the viscoelastic solution derived
for the multilayered system; therefore, it could be concluded that the viscoelastic solution
· The viscoelastic solution derived for the unit step loading case was extended to simulation of
principle, the viscoelastic solution for static loading could be extended to the viscoelastic
multilayered system subjected to the moving load. From the simulation performed on the
same structure as that used in the verification study, it was found that the results reflected
well the general behaviors of viscoelastic materials, as well as those generally observed from
the elastic layer analysis. These observations lead to the conclusion that the viscoelastic
solution derived for the moving load has a capability of simulating the pavement responses
· Time and effort in evaluating the structural responses of a pavement through the viscoelastic
solutions derived in this study would not be comparable to those typically required to
perform FEA for the same structure. It is therefore expected that the viscoelastic solutions for
the multilayered system derived in this study will be a fast and reliable tool for the design of
flexible pavements.
22
REFERENCES
ADINA user’s manual-version 8.5. (2008). ADINA R & D, Inc., Watertown, MA.
Al-Qadi, I. L., Yoo, P. J., Elseifi, M. A., and Janajreh, I. (2005). “Effects of tire configurations
962.
ARA., ERES Division (2004). Guide for Mechanistic-Empirical Design of New and
Burmister, D. M. (1945). “The general theory of stresses and displacements in layered soil
De Hoog, F. R., Knight, J. H., and Stokes, A. M. (1982). “An improved method for numerical
inversion of Laplace transforms,” Society for Industrial and Applied Mathematics, 3(3),
357-366.
De Jong, D. L., Peatz, M. G. F., and Korswagen, A. R. (1973). Computer Program BISAR,
Layered Systems under Normal and Tangential Loads. External Report. Koninklijke/Shell-
Elseifi, M. A., Al-Qadi, I. L., and Yoo, P. J. (2006). ‘‘Viscoelastic modeling and field validation
Findley, W. N., Lai, J. S., and Onaran, K., (1976). Creep and relaxation of nonlinear viscoelastic
23
Kim, J. and West, R. C. (2010). “Application of the viscoelastic continuum damage model to the
Kim, J., Lee, H., and Kim, N., (2010). “Determination of Shear And Bulk Moduli of Viscoelastic
Solids from the Indirect Tension Creep Test.” Journal of Engineering Mechanics ASCE,
in press.
Kim, J., Roque, R., and Byron, T., (2009). “Viscoelastic Analysis of Flexible Pavements and Its
Kim, J., Sholar, G., and Kim, S. (2008). ‘‘Determination of accurate creep compliance and
Kim, J., T. Byron, Sholar, G. A., and Kim, S. (2007). “Comparison of a Three-Dimensional
Viscoelastic Pavement Model to Full-Scale Field Tests,” Proceedings of the 86th Annual
Lee, H. and Kim, J. (2009). “Determination of viscoelastic Poisson’s ratio and creep compliance
from the indirect tension test.” Journal of Materials in Civil Engineering ASCE, 21(8),
416-425.
Inc., NY.
Roque, R., Buttlar, W. G., Ruth, B. E., Tia, M., Dickison, S. W., and Reid, B. (1997). Evaluation
of SHRP indirect tension tester to mitigate cracking in asphalt concrete pavements and
24
Talbot A. (1979) “The accurate numerical inversion of Laplace transforms.” Journal of the
25
Table 1. Material input for the viscoelastic analysis
26
Table 2. Coefficients in Prony series
Index, i ri (1/sec) Gi (MPa) ri (1/sec) Ki (MPa)
0 1.429E+02 4.520E+02
1 1.429E+03 2.131E+03 1.429E+03 2.477E+03
2 1.485E+02 1.491E+03 1.485E+02 2.286E+03
3 1.757E+01 1.305E+03 1.757E+01 2.025E+03
4 2.080E+00 8.261E+02 2.080E+00 1.487E+03
5 2.015E-01 3.843E+02 2.015E-01 7.494E+02
6 2.615E-02 1.997E+02 2.615E-02 4.610E+02
7 1.020E+03 -6.908E+00 9.096E+02 -1.776E+02
8 1.008E+02 -6.298E-01 9.600E+01 -2.525E+01
9 1.016E+01 -1.097E+00 9.172E+00 -5.166E+01
10 1.016E+00 -4.076E-01 9.095E-01 -2.732E+01
11 1.018E-01 -2.773E-01 8.862E-02 -2.588E+01
12 1.026E-02 -1.450E-01 8.055E-03 -2.385E+01
27
Table 3. Maximum stresses and strains for one pass of loading
At the Bottom of Asphalt Layer At the Top of Subgrade Layer
Speed (km/h) Radial Stress (kPa) Radial Strain Vertical Stress (kPa) Vertical Strain
10 2176 2.46E-04 -56 -6.00E-04
20 2327 2.27E-04 -54 -5.72E-04
30 2411 2.17E-04 -53 -5.57E-04
40 2469 2.10E-04 -52 -5.46E-04
50 2513 2.06E-04 -51 -5.38E-04
60 2582 1.98E-04 -50 -5.24E-04
70 2576 1.99E-04 -50 -5.26E-04
80 2601 1.96E-04 -50 -5.21E-04
90 2622 1.94E-04 -50 -5.17E-04
100 2640 1.92E-04 -49 -5.14E-04
28
Axis of Symmetry
a
q
r ¥
E1, m1 h1
E2, m2 h2
¼
hn
En, mn z
29
Axis of Symmetry
a
q(t)
r¥
E1(t), m1(t) h1
E2(t), m2(t) h2
¼
hn
En(t), mn(t) z
30
Axis of Symmetry a = 152.4 mm
q(t) = 690×H(t) kPa
r ¥
Rigid Layer z ¥
(a)
Axis of Symmetry
(b)
Figure 3. Pavement structures for a) the viscoelastic solution analysis and b) the FEA
31
Viscoelastic Solution FEA
2500
2000
Stress (kPa)
1500
1000
500
0
0.0 2.0 4.0 6.0 8.0 10.0 12.0
Tim e (sec)
(a)
Viscoelastic Solution FEA
6.00E-04
5.00E-04
4.00E-04
Strain
3.00E-04
2.00E-04
1.00E-04
0.00E+00
0.0 2.0 4.0 6.0 8.0 10.0 12.0
Tim e (sec)
(b)
Figure 4. (a) Stresses and (b) strains at the bottom of the asphalt layer
32
Viscoelastic Solution FEA
-30
-40
-50
-60
-70
-80
-90
Tim e (sec)
(a)
Viscoelastic Solution FEA
-5.00E-04
-6.00E-04
-7.00E-04
-8.00E-04
-9.00E-04
-1.00E-03
Tim e (sec)
(b)
Figure 5. (a) Stresses and (b) strains at the top of the subgrade layer
33
M=5 M=10
2500
R2=1.00000
2000
Stress (kPa)
1500
1000
500
0
0.0 2.0 4.0 6.0 8.0 10.0 12.0
Tim e (sec)
(a)
M=5 M=10
6.00E-04
5.00E-04
4.00E-04
Strain
3.00E-04
2.00E-04
1.00E-04
R2=1.00000
0.00E+00
0.0 2.0 4.0 6.0 8.0 10.0 12.0
Tim e (sec)
(b)
Figure 6. (a) Stresses and (b) strains with different M values at the bottom of the asphalt layer
34
M=5 M=10
-30
-40
-50
-60
-70
-80
Tim e (sec)
(a)
M=5 M=10
-4.00E-04
-5.00E-04
-6.00E-04
-7.00E-04
-8.00E-04
-9.00E-04
Tim e (sec)
(b)
Figure 7. (a) Stresses and (b) strains with different M values at the top of the subgrade layer
35
si
sn
¼
s3
s2
s1
s0
t0 t1 t2 t3 ¼ tn
36
a
q(t)
r, x
z
r0, t0
P
= R ( z , r0 , tn - t0 )
Loading Direction
¥
(a)
a a
q(t)
r, x
-q(t) z
r0, t0 R ( z , r0 , tn - t0 )
P =
- R ( z , r0 , tn - t1 )
r1, t1
+ R ( z , r1 , tn - t1 )
Loading Direction
z z
¥ ¥
(b)
37
Top View
r (Radial Axis)
Center of Loading P
ri
38
Top View
Axis of Symmetry
50.8 mm
P (=P¢)
r, x
Loading Path
(a)
3000
2500
Direction of Loading
2000
Stress (kPa)
1500
1000
500
0
-2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5
-500
-1000
Distance from Loading (m )
(b)
Figure 11. (a) Illustration of load shift; (b) stress distribution over one pass of loading
39