Você está na página 1de 15

MECHANICS

. OF
MATERlALS
ELSEVIER Mechanics of 24 (1996) 241-255
A viscoelastic continuum damage model and its application to
uniaxial behavior of asphalt concrete
Sun Woo Park a,*, Y. Richard Kim a, Richard A. Schapery b
a Deparlment ofCivil Engineering, North Caro lina State Uniuersity, Raleigh, NC 27695-7908, USA
b Department of Aerospace Engineering and Engineering Mechanies, The Universitj ofTexas at Auslin, Auslin, TX 78712, USA
Received 18 December 1995; revised version received 28 August 1996
Abstract
An existing viscoelastic constitutive model which accounts for the effects of rate-dependent damage growth is described
and appIied successfully to characterize the uniaxial stress, constant strain rate behavior of asphalt concrete. The special case
of an elastic continuum damage model with multiaxial loading, which is based upon thermodynamics of irreversible
processes with internal state variabIes, is first reviewed and then it is shown how this model has been extended to a
corresponding viscoelastic darnage model through the use of an elastic-viscoelastic correspondence principIe. The general
mathematical model is next specialized to uniaxial loading. A rate-type evolution law, similar in form to a crack growth law
for a viscoelastic medium, is adopted for deseribing the damage growth within the body. Results from laboratory tests of
uniaxial specimens under axial tension at different strain rates are then shown to be consistent with the theory. The
discussion of data analysis describes the specific procedure used here to obtain the material parameters in the constitutive
model for uniaxial loading and how the method may be generalized for multiaxial lo ad ing.
Keywords: Viscoelasticity; Damage; Constitutive equation; Correspondence principIe; IntemaI state variable; Asphalt concrete
1. Introduction
Mechanical behavior of a material, monolithic or
composite, is affected by changes in the microstruc-
ture that occur in the body due to imposed loading_
Asphalt concrete experiences a significant amount of
microcracking under service loading and we suppose
heI'e that it is the source of non1inear behavior.
However, the model used in this study is phe-
and therefore we shall use the more
term damage in discussing the source of
non1inear behavior.
, Corresponding author.
Most damage model s that have generality are
based on principles of thermodynamics, because the
inelastic mechanical behavior of engineering materi-
aIs is intrinsically linked to irreversible thermody-
namic processes accompanying energy dissipation
and physical changes of the microstructure. In gen-
eral, a continuum damage model consists typically of
three major components: (a) selection of damage
variabIes, (b) definition of strain energy density (as a
function of damage variabIes and other state vari-
ables) and (c) a damage evolution law. Of course, if
the model considers only fixed damage, the evolu-
tion law is not needed. The tensoriaI natureof dam-
age variabIes has been argued by differen.t re-
0167-6636/96/$15.00 1996 EIsevier Science B.V. All reserved.
Pll SOI67-6636(96)00042-7
242 S.W. Park et al./ Mechanics ofMaterials 24 (J996) 241-255
searchers (e.g., Krajcinovic, 1989; Ju, 1989); they
have' been claimed to be either scalars, vectors,
second-order, or fourth-order tensors, with the choice
being dependent upon the nature of the material and
particular class of problems of interest.
Developing a realistic mathematical model for
viscoelastic solids with growing damage is not an
easy problem. The complexity is attributed to the
viscoelastic hereditary effects of the material, com-
plex nature of describing the damage evolution and
the coupling between these two mechanisms, Le.,
viscoelasticity and damage. Therefore, without a rig-
orous theoretical treatment as weB as physically
realistic simplifications, the task would be insur-
mountable. There have been a number of works
which address both viscoelasticity and damage (e.g.,
Harper, 1986; Harper, 1989; Simo, 1987; Sjolind,
1987; Weitsman, 1988; Kim and Little, 1990; Peng,
1992; Gazonas, 1993). Schapery (1981,1982, 1990b,
1994) has developed a series of viscoelastic damage
models for particulate and fiber-reinforced compos-
ites. His models are based on his other works on
viscoelastic fracture mechanics (Schapery, 1975,
1984) and elastic-viscoelastic correspondence prin-
ciples in terms of pseudo variables (Schapery, 1984)
and have further evolved in combination with his
micromechanical model (Schapery, 1986) and work
potential theory (Schapery, 1990a). Park (1994) and
Park and Schapery (1996) have proposed an explicit
viscoelastic damage model based on the g e n e r a
theories and guidelines developed by Schapery
(1990a, 1990b), and applied the model to character-
ize the mechanical behavior of a particle-filled elas-
tomer under axial extension and hydrostatic pressure.
In Section 2, a constitutive model for elastic
solid s with growing damage developed by Schapery
(1990a) is briefly reviewed and then, based on this
and the elastic:-viscoelastic correspondence principie
(in terms of pseudo strain), the. corresponding vis-
coelasticity problem with time-dependent damage
growth is discussed. The three-dimensional vis-
coelastic damage model is ne;x:t. specialized to a
simple uniaxial model. The characterization of the
behavior of asphalt concrete, subjected to extension
at different strain rates under uniaxial stress serves in
Sections3, 4 and 5 to illustrate the theory when one
internal,state variable is sufficient tp account for the
damage growth. Descriptions of the material charac-
teristics and the laboratory tests performed are given
in Section 3 and the detailed procedure for determin-
ing the parameters employed in the constitutive model
using the test data is presented in Section 4. Also
discussed is an extension of the method to multiaxial
loading. In Section 5 the mechanical response and
the damage evolution are predicted using the model
and discussed.
It should be noted that the theoretical approach
taken in Section 2 is not new, but its application to
asphalt concrete with strong nonlinear behavior ap-
parently is new. A similar but more complex model
(with two internal state variabies) was needed by
Park and Schapery (1996) for particle-fil1ed rubber
(solid propellant) under multiaxial loading. In this
paper emphasis 1S placed on applicability of the
theoretical model to a composite with different con-
stituent characteristics. Asphalt concrete typically
used in construction of pavements consists of an
extremely high volume fraction (90-95%) of natu-
rally-occurring aggregates with sizes ranging from
0.005 to 2 inch in effective diameter and bituminous
cement which possesses very high viscoelasticity. A
detailed description of the procedure for determining
the model parameters for multiaxial loading in Sec-
tion 4 and the discussion on the effects of loading
rate and 10ading duration on the damage evolution
given in Section 5 are new.
2. Constitutive model for viscoelastic solids with
growing damage
2.1. Damage model for elastic solids
The mechanical behavior of an elastic medium
with consiant materia! properties (i.e., without dam-
age growth) can usually be described using an appro-
priate thermodynamic potential (e.g., the Helrnholtz
free energy when temperature is an independent
variable, or the internal energy if entropy is used in
place of temperature). These potentials are point
functions of thermodynamic state variabIes. When
temperature is constant, the Helrnholtz free energy
may be identified with the so-called strain energy
which is the work done on the system by external
loading. However, when damage occurs due to exter-
nal loading, the work done on the body is not
S.W. Park et al.j Mechanics oj Materials 24 (I996) 241-255 243
entirely stored as strain energy but part of it is
consumed in causing damage to the body. The
amount of work required to produce a given extent
of damage can be expressed as a function of internal
state variables within the context of thermodynam-
ics of irreversible process. The total work input to
the body during the processes in which damage
occurs depends, in general, on the path of loading.
However, it was observed that the applied work for
processes in which damage occurs is often not very
sensitive to the path of loading (Schapery, 1987a,
1987b; Lamborn and Schapery, 1988, 1993).
Schapery (1990a) applied the method of thermo-
dynamics of irreversible processes and the observed
phenomenon of path independence of work (for suit-
ably limited paths) to develop a theory applicable to
describing the mechanical behavior of elastic media
with growing damage and other structural changes.
The theory is general enough to allow for strong
nonlinearities and coupling between the internal state
variabIes and to account for a v ariet y of mechanisms
including micro- and macro-crack growth in mono-
lithic and composite materiaIs. Schapery and Sicking
(1995) applied the theory to model the damage-re-
lated, material nonlinearity in graphite-epoxy lami-
nates and Lamborn and Schapery (1988, 1993)
showed the existence of a work potential for limited
deformation paths using experimental data from ax-
ial and torsional deformation tests on angle-ply
fiber-reinforced plastic laminates. The elements of
Schapery's theory in terms of strain energy formula-
don may be represented as follows:
Strain energy density function:
Stress-strain relationships:
oW
(Tij= --
OBij
Damage evolution laws:
(1)
(2)
(3)
where (Tij: stress tensor, Bii strain tensor, Sm: inter-
nal state variabIes (or damage parameters) and W
s
=
W/Sm): dissipated energy due to damage growth.
The internal state variabIes Sm (m = l, 2, ... ,M)
account for the effects of damage; their tensorial
order depends on the model and application of inter-
esL Eq. (3) is similar to a crack growth equation
(e.g., P =Pc where p is the energy release rate and
Pc is the fracture toughness) and, in fact, Eq. (3) is
used to find the Sm as functions Bij' which is always
possible for stable behavior (Lambom and Schapery,
1993). The left hand side of Eq. (3) is the so-called
available thermodynamic force. for damage growth
whiJe the right hand side is the required force.
2.2. Elastic-viscoelastic correspondence principie
The theory of linear viscoelasticity, and some
nonlinear viscoelasticity model s have be en est,ab-
lished in the pasL However, solution of various kinds
of viscoelastic boundary value problems stiil remains
a complex problem. Fortunately, the nature of COrre-
spondence existing between the elastic and the inte-
gral-transformed (e.g., Laplace-transformed or
Fourier-transformed) viscoelastic field equations for
linear, nonaging materials enables one to obtain a
solution to a viscoelastic boundary value problem
from the solution to its corresponding elastic bound-
ary value problem. On the other hand. it has been
shown that the stress-strain relations for a broader
class of viscoelastic materiaIs can be represented by
elastic-like equations through the use of so-caIIed
pseudo variabIes. This simplifying feature enables
correspondence principles to be established and ap-
plied to linear and nonlinear viscoelastic boundary-
value problems, with or without agin:g (Schapery,
1981; Sthapery, 1984); the Laplace or Fourier trans-
form is not used. U sing these correspondence princi-
pIes one can obtain viscoelastic solutions from their
elastic counterparts through a simple substitution and
integration. These correspondence principles are ap-
plicable with stationary or time-dependent boundary
conditions and do not require a transform inversion
step, but rather require onIy the evaluation of a
convolution or hereditary integral.
In order to introduce pseudo strain, let us consider
a uniaxial stress-strain equation for linear, nonaging
viscoelastic materiaIs,
f
i' dB
d= E(t-r)-dr
o ,dr
( 4)
244 S.W. Park et al./ Mechanics ofMaterials 24 (1996) 241-255
which can be written as
er
er = E B R or B R = -
R E
R
if we define
1 1/ dB
B
R
= - E( t - T) - dr
, ER o dT
(5)
(6)
where ER is termed the referencemodulus, which is
a free constant and has the same dimension as the
relaxation modulus, E(t).
The usefulness of Eq. (5) is that it takes the form
of an elastlc stress-strain equation even though it is
actually a viscoelastic stress-strain equation. The
quantity B
R
is called the pseudo strain. The pseudo
strain accounts for a11 the hereditary effects of the
material through the convolutionintegral. The refer-
ence modulus ER is introduced because it is a useful
parameter in discussing special material behavior
and introducing dimensionless variabies. For exam-
ple, if we take E(t) = ER in Eq; (6) we obtain
B
R
= B and Eq. (5) reduces to the linear elastic
equation er = EB or B = er/E. The definition has
been extended to multiaxial stress-strain relations
with aging and transient temperature (Schapery,
1981).
2.3. Damage model for viscoelastic solids
In order to account for the effects of viscoelastic-
ity, we will take an approach of generaliiing the
elastic damage model described above such that it
may accommodate' the effects of viscoelasticity.
Schapery (1981) showed how this may be done
through the use of the micromechanics of compos-
ites, his elastic-viscoelastic correspondence princi-
pIe and rate-type dilmage evolution laws.
Based on these ideas, first we replace the
strains Bij that appear in the elastic constitutive
equation for, ,the undamaged matrix in a composite
with corresponding pseudo strains
R 1 1/ OBij
B.,,= - E(t- T)- dT.
'} ER o ' ' aT
(7)
Then, according to the correspondence principIe and
amicromechanical prediction of the effective elastic
composite constitutive equation, this equation written
in terms of pseudo strains now govems the viscoelas-
tic composite with damage. Again, if we take E(t)
= ER (constant modulus), from Eq. (7) we obtain
= Bij and the original elastic composite with
damage is recovered. However, the damage evolu-
tion laws cannot directly be translated into evolution
laws for viscoeIastic through the corre-
spondence principIe. It is to be understood that not
onIy the available force for growth of Sm is rate-de-
pendent (through pseudo strains), but the resistance
against the growth of Sm is rate-dependent for most
viscoelastic materials. The following evolution law,
which is similar in form to the well-known power-Iaw
crack growth laws for viscoelastic materiaIs (e.g.,
Schapery, 1975; Schapery, 1984), will be adopted as
it can reasonably represent the actual damage evolu-
tion processes of many viscoelastic materials:
Sm = (- OW
R
) "'m (m not summed)
oSm
(8)
where W
R
= Sm) is the pseudo strain energy
density function, a
m
is a material-dependent con-
stant and the overdot denotes a time note
that W(Bij' Sm) is the elastic strain energy density.
This form of evolution law was used successfully in
deseribing the multiaxial behavior of a fi11ed elas-
tomer with growing damage (Park and Schapery,
1996). The right hand si de is usual1y written with a
factor that may depend on Sm' but it can be elimi-
nated through a change of the damage param-
eter. Obviously, the simpler Eq. (8) uses an Sm
with complex units. The units for Sm are
[stress]a/(a+ 1)[time]I/(a+ I).
2.4. Uniaxial viscoelastic damage model for asphalt
concrete
Based on discussions in the preceding sections, a
simple constitutive model for uniaxial stress-strain
behavior of asphalt concrete (asphalt-aggregate mix-
ture) with time-dependent damage growth is pro-
posed as fol1ows:
First, for the pseudo strain energy density func-
tion of the material, the simplest possible form is
proposed:
(9)
S.W. Park et al.j Mechanics ol Materials 24 (J996) 241-255 245
where the coefficient C is a function of a single
damage parameter S. Then,
aw
R
er=: --= C(S)e
R
aeR
( 10)
which, for fixed damage, is an equation for linear
viscoelastic behavior. The damage evolution law Eq.
(8) is specialized to the following single equation for
S: .
s=(_a:SRr (11)
(9)-(11) constitute a model for the uniaxial
stress-strain behavior of a viscoelastic solid (asphalt
concrete, in our case) with time-dependent damage
growth. The form of (10) is consistent with the
behavior of the stress versus pseudo strain curves
(see Fig. 7 in Section 3.4). The figure indicates that
stress is linearly proportional to pseudo strain at low
strain levels where practically no damage has oc-
curred. This situation corresponds to a constant value
of the constitutive function C in Eq. (10) and thus
results in a linear relationship between er and ER.
As the pseudo strain level increases, each curve
deviates from the linear relationship indicating a
reduction of C, so that dC/dS < O.
Interestingly, each curve in Fig. 7 bends at a
different rate for a given pseudo strain level, suggest-
ing that the damage growth is not only rate-depen-
dent through ER, but also exhibits additional rate
dependence. Lower strain-rate curves bend at lower
eR, indicating that more damage occurs at longer
loading durations for the same pseudo strain. This
behavior is consistent with (11). The evolution
law Eq. (11) ref1ects rate-effects through e R in W R
and explicit time-effects through the time-derivative
of S on the left hand side.
The only damage-related' constitutive parameters
to be determined for a given material are the func-
tion C(S) and the constant ex. Here, we have as-
sumed that the state of damage can be represented by
a single parameter S. Thevalidity of this assumption
will be judged by alater comparison of the model
with the experimental observation of response of the
specimens. A detailed procedure for determination of
C(S) and ex from the experimental data is given in a
later section.
For this uniaxial loading problem, the material
characterization may be simplified by changing the
damage parameter. Specifically, upon substituting
Eq. (9) into (11), we may integrate the equation
to obtain an implicit one-to-one functional relation-
ship between S and a new damage parameter S ,
[
sd S ] 1/(2 a) .

O (-O.SdC/dS) Ci
(12)
where this is an explicitfunction of
history,
(13)
in which I Idenotes absolute value. Apart from the
numerkal factor k, S * is a Lebesgue norm (Reddy
and Rasmussen, 1982) of the pseudo strain; k is a
free constant that will be selected so that the maxi-
mum values 6f S and S* in our study are numeri-
cally equal, for ease of plotting results. Thus,
(10) may be replaced by
(14)
The mbdulus functi9n C in (lO) and tbat in
(14) are defined In terms of different arguments (S
and S', respectively) and thus their forms differ, in
general; however, both functions will be denoted by
C for notational simplicitY. The functional forin of C
in (14) may be found from experimental uniaxial
data more easily than that in (IO). However, for
non-proportional, multiaxialloading, (13) cannot
be used, but the form of (I1) is applicable.
Anticipating such applications, later we will describe
methods for determining both functions C in Eqs.
(10) and (14) directly from data;
Schapery (1982) showed that S * is valid damage
parameter for proportional loading and used' it SllC-
cessfully in model ing particle-filled rubber under
uniaxial stres s and spe"cified complex strain histories.
Kim and Little (l990) demonstrated that the uniaxial
behavior of asphalt concrete with fine particles. can
be characterized by using Lebesgue norms of the
pseudo strain as damage parameters; but their inves-
tigation was limited to a low range of strain 'and a
moderate extent of damage compared to the.
study.
246 S. W. Park et al./ Mechanics ofMaterials 24 (J996) 241 -255
3. and test program
3.1.
The material used in the research is asphalt con-
crete, which is a particulate composite composed of
aggregate particles embedded in a bituminous ma-
trix. Asphalt concrete is primarily used in the con-
struction of flexible pavements for roads and air-
fields. Accurate characterization of mechanical be-
havior of asphalt concrete has long been sought by
many engineers with a view to improving the guide-
lines for designing new or rehabilitating the existing
pavement structures. The bituminous matrix (or as-
phalt binder) is a viscoelastic' polymeric material
normally obtained through various distillation pro-
cesses from crude petroleum. The asphalt binder
strongly adheres to the aggregate particIe s and binds
them to form asphalt concrete. The resulting compos-
ite . possesses high viscoelasticity imparted by the
binder matrix. Also, in most cases, when the com-
posite is subjected to mechanical loading, distributed
microstructural damage occurs in it in the form of
microcrack nuc1eation and growth due to high stress
concentrations along the partic1e-matrix interfaces
and within the matrix (Kim and Little, 1989). There-
fore, any mathematical model which describes
fue cOIlstitutive oehavior of asphalt concrete mus t be
capable of for the effects of viscoelastic-
ity and damage growth.
Table l
Gradation of aggregates used in the asphalt concrete
Sieve size (mm)' Passing (%)b
25.0 (l") 100
19.0 (3/4') 95
12.50,/2") 79
9.5 (3/8") 67
4.75 (#4) 47,'"
2.36 (#8) 33
1.18(#16) 22
0.6(#30) 14
.0:3 (#50) 9
0.15 (#100) 5
0.Q75 (#200) O
'Side dimension of a square hole offue sieve.
bpercent of aggregales by weight thal pass the sieve.
3.2. Specimen labrication
Specimens were made of granite aggregate and an
asphaIt binder (Type AAM-l) specified by the
Strategie Highway Research Program. The aggregate
gradation is given in Table l. Aggregates were mixed
with 5.0% of asphalt binder by weight of dry aggre-
gate at 140C for 4 minutes. The mixture was then
compacted at 116C to a rectangular slab and cured
at 60C for 15 hours before being cored into cylin-
drical specimens with a diameter of 102 mm (4 inch)
and a height of 203 mm (8 inch). Compaction of the
mixture was performed using the rolling wheel com-
paction method.
3.3. Uniaxial creep tests and determination ol ther-
momechanical properties
A serie s of tensile creep tests were performed on
the cylindrical specimens described above at five
different temperatures (- 5, 5, 15, 25 and 33C) in
order to obtain the basic thermoviscoelastic proper-
ties of the materia!, including the linear viscoe!astic
relaxation modulus and the time-temperature shift
factor. Relative!y low stress levels depending on the
test temperature (192 kPa for - 5C to 33 kPa for
33C) were applied to the specimen to ensure mini-
mum damage during the testing; !ower-level stresses
were applied to the specimen at high temperatures
because of the reduced modulus of the specimen.
The creep curves obtained at five different tempera-
ture leve!s are given in Fig. l. The individual creep
curves in Fig. 1 were then used to construct a master
creep curve shown in Fig. 2. The master creep curve
(or master creep function) was generated by shifting
each individual curve obtained at different tempera-
ture in the horizonta! direction on a log-log plot.
This feature has a significant consequence in that the
dependence of the material property on both time
and temperature can be represented by dependence
on a single variable, the reduced time, and the
feature is often referred to as time-temperature su-
perposition; the c1ass of materials to which this
behavior applies is described as thermorheologically
simple (Leaderman, 1943). It already had been ob-
served that time-temperature superposition applies
to asphalt concrete (e.g., Pagen, 1965; Kim and Lee,
1995). In mathemati9al notation,
D(t,T)=DM(g) (15)
S.W. Park et al./ Mechanics oj Materials 24 (J996) 241-255 247
0.001:;;-----------------,
----

/
----
--
1 E ~ . Q 1
0.1 1 10 100 1000
1ime(sec}
Fig. l. Creep curves at different temperatures.
where D
M
( g) is the master creep function corre-
sponding to a certain reference temperature and g is
reduced time defined as
l
ldT
g= -
o aT
(16)
where aT is the time-temperature shiftfactor, which
is a temperature-dependent material parameter that
reflects the influence of temperature on the internal
viscosity of the material (Ferry, 1980). For a con-
stan t temperature, Eq. (16) reduces to
t
g=-. (17)
aT
The shift factor was obtained by measuring the
horizontal distance by which each curve in Fig. 1
0.001
0.0001
1E.()7'
r--
.oc
5C
1SC
25C
oc
33C
~
,,"
"
"
<!iP
i!'
II'
1E1E.()6 1E.Q5 0.0001 0.001 m 0.1 1 10 100 ;000 10000
Reduced TIme {sec}
Fig. 2. The master creep curve at reference temperature of 25C.
1E..oS
1E+04
... 1E+03
Ol
a 1E
-g
u.
1E.Q1
1 E . Q ~ - - r - - - - - ~ - - - - ~ - ~
-10 O 10 20 30 40 50
Temperature (C)
Fig. 3. Time-temperature shift factor data and linear fit.
shifts to form a master curve and is presented in Fig.
3.
Now, the required master relaxation function can
be obtained by converting the master creep function
using their interrelationship based on the theory of
linear viscoelasticity (e.g., Ferry, 1980). Hereinafter,
we shalI take the room temperature (T = 25C) as
our reference temperature and the reduced time will
be denoted by t and the master creep function
DM(g) and the master relaxation function EM(g)
will be denoted by D(t) and E(t), respectively, for
notational brevity. The definition of pseudo strain,
Eqs. (6) and (7), may be written for thermorheologi-
cally simple behavior by replacing physical time
with reduced time. Use of reduced time in the rate
law Eq. (11) and Lebesgue norm Eq. (13) extends
these equations to thermorheologically simple behav-
ior in the nonlinear range.
FinalIy, an analytical expression of the relaxation
function may be obtained by fitting an appropriate
mathematical.representation, such as the one given
below (often referred to as Prany series representa-
tion), to the known relaxation data:
N
E(t) = E" + E Ele-t/Pi (18)
;= l
where E", El and Pi are all constants. The quantity
Eoo in Eq. (18) is the long-time equilibrium modulus
and Pi are relaxation times. Different versions of
power-Iaw representations are often used to describe
the relaxation or creep behavior of many polymeric
materials inc1uding asphalt binder and asphalt con-
248 S.W. Park et al./ Mechanics ofMaterials 24 (1996) 241-255
'd'
8

:;
""
o
:::;;
"
o
"g
i:d
"5

IE8
JE7,
IE6
lE5
IE4

IE-5 IE-4 IE-3 lE-2 lE-l lEO lEI IE3 IE4
Time (sec)
Fig. 4. The master relaxation curve at reference temperature of
25C.
crete. However, even though the power-law repre-
sentations have many advantages associated with
their simplicity, in some cases they fai! to adequately
describe the creep or relaxation curves over a long
range (many decades) of time, as may be seen in Fig.
4. AIso in some computational procedures, using an
exponential representation, su ch as Prony serie s, is
more efficient than using a power-Iaw representation.
3.4. Constant-strain-rate uniaxial extension tests
A series of uniaxial extension tests was perfonned
on the cylindrical specimens at 25C at different
constant strain rates ranging from 0.00005/s to
0.0016/s using a servo-hydraulic closed-loop testing
machine. Each specimen was glued to the end plates;
which were then connected to the loading frame
through a load cell. A schematic diagram of the
testing fixture is shown in Fig. 5.
Time was measured by the internal clock of a
computer. Axial elongation was obtained by multi-
plying the elapsed time by the constant speed of the
moving crosshead of the tester. Axial strain was
calculated by dividing the elongation by the initial
length of the undeformed specimen. Axial stress was
obtained by dividing the measured axial force by the
initial cross-sectionaI area of the undefonned speci-
men.
Test results are presented in Fig. 6. The theoreti-
cal predictions that are also included in Fig. 6 will be
discussed in Section 5. As can be expected for highly
Glued Surface
Bottom Plale
Fig. 5. Schematic diagram of uniaxial testing apparatus.
viscoelastic materials, the stress response for a given
strain level depends sensitively on the strain rate.
Test data onIy up to the maximum stress level are
presented and will be used in our later analysis.
Severe Iocalization in defonnation behavior appears
to occur in each specimen beyond the maximum
stress point unti! the global specimen rupture occurS.
The test results shown in Fig. 6 are re-presented in
tenns of pseudo strain in Fig. 7. The pseudo strain,

Experiment
Theory
200
0.002
Strain Rates
c c " = c .0016/sec
c"
_-=--...... ...,...,'-"'-""'-D .0004/sec
.0001/see
- - ........ -........
-------- .00005/5"C
0.004 0.006 0.008 0.01
Strain.E
Fig. 6. Uniaxial stress-strain behavior at different strain rates.
S.W. Park et al./ Mechanics ofMaterials 24 (1996) 241-255 249
Strain Rates
I
600
.OO005/sec
o 0.001 0.002 0.003 0.004 0.005
Pseudo S train, E"
Fig. 7. Uniaxial stress-pseudo strain behavior at different strain,
raIes.
according to Eq. (6) with the use of Eq. (18), is
given by
EE/p/(l-e-
I
/
Pi
)} (19)
ER /= I
where R is the constant strain rate and t is time that
can be computed by t = e / R in the case of a
constant-strain-rate test. The reference modulus was
taken to be equal to the initial modulus of 167000
kPa corresponding to S = t = O. Fig. 8 shows a
typical strain-pseudo strain relationship (for R =
0.0004/s). Remember that, for a linear viscoelastic
material without damage, the pseudo strain and the
stress are identical except for a constant multiplying
factor.
0.004 ,,-------------:-------,
"uj 0.003

'
cii
o 0.002
'O
"
"
::
0.001

o 0.002 0.004 0.006 0.008
Straih. e
Fig. 8. Strain-pseudo slrain relationship (for R = 0.0004/s).
4. Determination of model parameters
In order to find aU components of a modulus
array c..(S), say, some multiaxial loading is
IJ m .
needed. In this case a direct characterization ap-
proach using onI y the Lebesgue norm of pseudo
strain Eq. (13) does appear possible. However, a
method invoIving the original Sm may be used. Here
we shall first illustrate the method for uniaxial load-
ing and then compare the results with the direct, but
limited, method based on S * .
4.1. Function C( S) and constant o:
In view of the proposed model, Eqs. (9)-(11),
there are two material-dependent quantities to be
determined. They are the function C(S) and the
constant 0:.
In order to find the function C(S), first we obtain
the value of C by fitting Eq. (10) to an experimental
stres s-pseudo strain curve for each strain rate. This
process gives us a modulus C that depends on eR
and strain rate (c.f. Fig. 7), but not S directly. In
order to find its dependence on S alone, we need to
find a relationship between S, eR and rate (or time).
Finding this relationship requires the use of the
postulated evolution law. However, the rate-type
evolution law Eq. (11) is not adequate for finding S
directly for a given eR because the equation itself
requires a priori knowledge of the functional form of
C(S) before the equation can be solved for S; the
solution of S for the given eR would require an
extensive iteration process which would be highly
impractical. However, using an approximation proce-
dure developed by Schapery (1990b), it can be shown
that the rate-type evolution law Eq. (11) can be
converted to the following integrated form:
aw
R
---=r
l
/
a
as
where
1
S == S(l + l/aj.
(1+1/0:)
(20)
(21 )
In deriving Eq. (20) from Eq. (I1), it is assumed
that o: is much greater than unity. Therefore the
250 S.W. Park et al.j Mechanics ofMaterials 24 (J996) 241-255
evolution law Eq. (20) should be viewed as an
approximate version of Eq. (ll). If the evolution law
Eq. (20) is used, the characterization process will be
much simpler and straightforward. Therefore, we
first find the function C and the constant et using
the evolution law Eq. (20) and then refine them
using the original rate-type evolution law Eq. (11).
Rewriting Eqs. (9) and (10) in terms of the trans-
formed damage variable 8, we obtain, respectively,
w
R
= te( 8)( e
R
)2
and
(22)
(23)
where C in Eq. (22) or Eq. (23) is numerically the
same as C in Eq. (9) or Eq. (lO), but its functional
dependence on 8 is different from ces); however, as
before for notational brevity, we shall not employ
separate notations for C in Eqs. (22) and (23).
We shall first determine the function C(S) and
the constant et that appear in the transformed model
Eqs. (22), (23) and (20) and then obtain the function
C(S) used in the model by replacing the
independent variable 8 with S according to Eq. (21).
The constant et determined in the transformed model
may be carried over to the original model without
being changed. The detailed steps for characteriza-
tion of the model will now be illustrated.
First, the modulus C is found by fitting Eq. (23)
to a stress-pseudo strain curve. This step gives us
modulus C that depends on eR (and strain rate) as
shown in Fig. 9. There are four experimental stress-
180000 ,
13
160000
c
, c
'" 0.2 <CI.!.

C
c
c
B
u !
rn
o
'"
.; 140000 L
o

E
:l Cc

'3 :;;
"O
.,.
o =0.,
0.1
'" ::s
"
=
Ol)
120000
.,
C
c
E
'"
"
Ci
"
100000
1
I
O

0.001 0.002 0.003 0.004
Pseudo S train. c
R
Fig. 9. Modulus-pseudo strain-damage parameter relationship for
R = 0.0004/s.
strain curves available. Only one curve is required to
find the function C(S), assuming that the constant et
can be accurately estimated from the material's vis-
coelastic properties (as discussed later); a refined
value may be obtained using more than one rate.
Altematively, for uniaxial stress, the exponent may
be found from the use of a Lebesge norm, as dis-
cussed above. An intermediate experimental curve
with a strain rate of 0.0004(s was used for charac-
terizing the model in terms of C.
Next, we need to find the relationship between 8
and eR for the selected strain rate. The transformed
variable 8 may be obtained from the following nu-
merical scheme, in which it should be understood
that the functional forms of C(en and C(S) are
different,
R) "( R) "
S ei+ l = S e
i
+ AS (24)
where
" C( - C(en
A S =
C'(en
(25)
AIso, eF Ci = 1, 2, ... ,N) denote discrete pseudo
strain levels, with = O and = current pseudo
strain, C' == dC(d8 and 8(0) = O; Fig. 9 shows that
for 8> O, there is a one-to-one relationship between
8 and eR at the given strain rate. 'Eq. (25) was
obtained from the definition of the derivative of C
with respect to 8:
or
(27)
" "( R) ""( R) "( R)
where Si = S 8
i
and AS = S 8
i
+ l - S e
i

Therefore, if we know the value of C and its deriva-
tive C' at each discrete pseudo strain level, we may
obtain values at each corresponding pseudo strain
level from Eqs. (24) and (25) through a summation
process. As the interval between the successive
pseudo strain levels becomes smaller, Eqs. (24) and
(25) render a more accurate estimation of 8.
We have already obtained the C value for each
eR, as explained before and we can also obtain C'
S.W. Park et al.; Mechanics oj Materials 24 (J996) 241-255 251
value for each eR from Eq. (20); rewriting Eq. (20)
with W R replaced by Eq. (22) one obtains
_C(e
R
)2=t-
I
/
Ct
(28)
The time t in Eq. (28) is computed by using t = e/R.
The constant a needs to be known a priori; how-
ever, we can make a reasonable initial guess for a
on the basis of the linear viscoelastic fracture me-
chanics. In many viscoelastic crack growth prob-
lems, the crack speed is governed by the a til power
in pseudo energy release rate, in which a is related
to the material's creep or relaxation properties. For
example, a = 1 + l/n, in which n =
log D(r)/log(t) or n = -log E(t)/log(t) at times
depending on crack speed, which was shown by
Schapery CI 975, part III) for rubber. As we may
reasonably assume that the damage which occurs in
the specimen is closely related to the growth of
microcracks, we will use the relationship, a = 1 +
l/n, for an initial trial value for the parameter a. An
initial trial value of 3 was adopted for a based on
n = 0.5, which is the appropriate value of slope for
the near power-law regime in Fig. 2. The resulting
versus eR curve is shown in Fig. 9. Now, cross-plot-
ting C against at all pseudo strain levels consid-
ered, as show n in Fig. 9, we obtain the function
C(S) in Fig. 10. Finally, the function C(S) can be
obtained from C(S) by replacing the transformed
variable with the original variable S, according to
Eq. (21).
Recall that the evolution laws Eqs. (1 i) and (20)
are not exactly equivalent because Eq. (20) was
180000 r----------------,
~ 160000
~ c ~
u
< 140000
:::l
:;
'"C
o
::E 120000
" " "
100000 '-------'--------'--------'
o 0.1 0.2 0.3
Damage Parameter, S
Fig. 10. Modulus-transformed damage parameter relationship (for
R = 0.0004; s).
180000 I
.-. 160000
~
U
",- 140000
"
:;
-g
::::E 120000
100000 '------'-----'-----'-----'-------'
O 0.1 0.2 OJ 0.4 0.5
Damage Param eter, S
Fig. II. Graphical representation of the constitutive function
C(S).
obtained from Eq. (11) based on an assumption that
a 1. Therefore, the function C(S) and the con-
stant a determined so far are approximate to the
degree of approximation involved in the derivation
of Eq. (20) from Eq. (11). The true C(S) and a can
be obtained by improving those obtained from the
approximate evolution law Eq. (20) through a method
of successiue approximations, which is a refinement
process that utilizes the original evolution law Eq.
(11). That is, the tentative C(S) and a can be
substituted into Eq. (9) and then Eq. (11) and S for
each given eR is obtained by summing the incre-
mental values of S over time with an initial condi-
tion S[t=o = O. Since we already know C values for
different eR levels from Eq. (10), by cross-plotting
C against the new S for aU eR levels considered, we
obtain the refined function C(S). The same step s Can
be repeated until the new C(S) and a are not
practically different from their preceding versions. In
our case, after three cycles of refinement, a reason-
able convergence was reached. The initial (first esti-
mate) and the final versions of the relationship be-
tween C and S for the material are graphically
shown in Fig. 11. A functional form of CeS) =
167000(1 - 0.29So.
5
- 0.19S) was obtained from
curve-fitting. Finally, it was found that with a = 2.5,
the model predicts the experimental observations
most closely. In principle, a can be determined
using a minimum of two experimental curves with
different rates, but in our analysis all four available
curves were considered in finding the optimum value
of er. More details concerning the characterization of
the model are available in Park (t 994).
252 S.W. Park et al./ Mechanics oj Materials 24 (J996) 241-255
180000 r-------------------,
"" 160000
~
U
"," 140000
::l
:g
o
::E 120000
From Test Data
Curv.Fit
...
'" ...
... ...
.........
~ ...
.........
...
100000 L-_.1..-_-'--_--'--_-'-_-'-_-'--_--'
O 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Damage Parameter, S'
Fig. 12. Graphical representation of the constituti ve function
C(S').
4.2. Function C(S' ) and constant a in the alterna-
tive model
In Section 4.1 we discussed a procedure for deter-
mination of the function C(S) and the constant a
employed in the model, Eqs. (9)-(11). This model is
based on principles of thermodynamics and equipped
with essential components of a continuum damage
model (a damage variable, a strain energy density
function and a damage evolution law) and may
easily be extended to a multiaxial model. However;
as noted above, for the case of uniaxial state of
stress, the model may be degenerated to a simpler
version, Eqs. (13) and (14). They do not require a
separate damage evolution law because the damage
parameter is directly evaluated in terms of the ap-
plied input history.
The procedure for finding C(S) and associated
constant a is much simpler than that for finding
C(S) and a. First, the parameter S* is obtained for
given eR according to Eq. (13). At the same time,
the value of C is obtained for given eR from Eq.
(14). Then the functional form C(S') is found by
cross-plotting and curve-fitting C against S" for all
eR levels considered. The constant a is obtained
through a trial-and-error process su ch that C versus
S" curves from different strain-rate data match best
with each other. Like in Subsection 4.1, a = l +
l/n = 3 was used as the initial tri al value and
a ~ 2.5 was found to be the optimum value. The
constant k = 124 was selected so that the maximum
values of S and S' are equal within the range of the
experiment. The resulting C versus S (with a =
2.5) curves for al1 four straill rates considered and a
cubic polynomial fit are given in Fig. 12. A func-
tional form of C(S') = 167000[1 - O.13S -
0.34S
2
- 0.33S 3] was obtained. It should be noted
that the data in Fig. 12 for all strain rates would fall
along a single curve if the model were exact; the
closeness of the data indicates the model is a good
approximation.
5. Prediction of mechanical response and damage
evolution
Having found the constitutive parameters em-
ployed in the model, one can now predict the stress
for a given strain history. First, the pseudo strain eR
is computed according to Eq. (6). The damage pa-
rameter S is obtained from Eqs. CI l) and (9) through
a summation of the increments of S (where !:::.. S =
SM) over the time elapsed, with an initial condition
of S = O at t = O. Finally, the stress is predicted by
Eq. (10). The continuous lines in Fig. 6 are the
theoretically predicted stresses. Overall, the predic-
tions are in good agreement with the experimental
data, which demonstrates the validity of the proposed
model (in terms of S) for the test histories used. The
curve for R = 0.0004/s is in best agreement among
the curves as its experimental data were used in the
characterization of the model. The smali discrepancy
observed in the R = 0.0004/s curve is primarily due
to the smoothing effects associated with the curve
0.8 r----------------,
CI) 0.6
g ~

~ 0.4
~

CI 0.2
0.002 0.004
S tra in Rates
.0016lsee
0.006
Strain, e
.0004/seo
.000l/see
0.008 0.01
Fig. 13. Damage evolution (S) with strain at different strain rates.
S.W. Park et al./ Mechanics ofMaterials 24 (]996) 241-255 253
fitting in obtaining the analytical form of C(S) given
earlier.
The stress-strain response was also predieted us-
ing the altemative simplified model, Eqs. (13) and
(I 4). The resulting predietions were praetieally the
same as those presented earlier (Fig. 6) exeept for
negligible discrepancies due to errors involved in
curve-fitting.
Evolution of the damage parameter S, as the
speeimen is stretched at different strain rates, is
represented in Fig. 13. It is seen that for a given
strain level, the higher strain-rate deformations in-
duee more damage than the lower-rate deformations.
The relationship between the evolution of damage
and pseudo strain is given in Fig. 14. Here, for a
given pseudo strain level, the lower strain-rate defor-
mations induce more damage. This can be explained
from Eq. (I1) in that for a given eR, S is an
inereasing function of the loading duration. The
lower-rate deformations take more time to achieve
the given eR. When the effects eR and loading
duration eompete with each other, Fig. 13 indieates
that the effect eR on damage growth is more pro-
nouneed than that of loading duration; higher strain-
rate deformations take less time to achieve the given
e, but yield higher eR for the same e and result in
more damage. The evolution of damage is dependent
up on both strain rate (through eR) and loading
duration.
Finally, it should be noted that the direet method
of characterization using S' may be employed to
derive C(S) (for subsequent use in multiaxial load-
0.8 r----------------_
Strain Rates
~ . 0.6
8
'"
"
ci: 0.4
"
fal

Cl 0.2
o
o
.0004/sec
.00 16/se: 1
o . . . . . ~ ~ .... """"'5-C--".=-" ---'-----'------'
O 0.001 0.002 0.003 0.004 0.005
Pseudo Strain, SR
Fig. 14. Damage evolution (S) with pseudo strain at different
strain rates.
0.7 ,------------------,
0.6
..
~ 0.5
8
~ 0.4
la
';; 0.3
~
0.2
Q
0.1
From Eq. (29)
...
From Eq. {12}
O ~ - ~ - ~ - - ~ - J - - ~ - ~ - ~
O 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Damage Parameter, S
Fig. 15. Relationship between damage parameters S' and S.
ing applications) without employing the previously
described method of suecessive approximations. This
approach to deriving C(S) comes from the inverse of
Eq. (12),
_ S'[2a
C
-dC/dS*)"CS *)2<>-ljl/(a+l)
s-Ia 2"k2a . dS .
(29)
Observe that from Eq. (29) we may Cimplicitly) find
S = S' (S) in terms of the known funetion dC / dS' .
The result is drawn in Fig. 15 and compared to that
obtained from Eq. (12) with the independent1y de-
rived C(S); the agreement serves as a direct check
on the method of successive approximations. Having
found C(S), the use of S = S"(S) then gives the
desired funetion C(S).
6. Concluding remarks
A viscoelastic constitutive model with effects of
distributed damage has been shown to provide an
accurate eharacterization of the mechanical behavior'
of asphalt conerete under uniaxial, monotonie load-
ing. The specific viscoelastic damage model devel-
oped by Park and Schapery (1996) for the meehani-
cal behavior of a filled elastomer under axial load
plus eonfining pressure has been, in this paper, spe-
cialized and applied to characterize the uniaxial
stress-strain behavior of asphalt concrete. In this
model, as well as that for multiaxial loading used by
Park and Schapery (1996), the material is linearly
254 S.W. Park et al.j Mechanics oj Materials 24 (J996) 241-255
viscoelastic except for the effect of damage. The
model in its general form for multiaxial loading is
based upon thermodynamics of irreversible process
with internal state variabies and is coupled with
phenomenological damage evolution laws.
It was shown that the specific forms of the pro-
posed stress-strain relationship Eq. (10) and the
damage evolution law Eq. (11) are consistent with
experimental observations (Fig. 7). The rate-type
evolution law Eq. (11), simi1ar to a crack growth law
for viscoelastic media, has been demonstrated to
successfully describe the state of time-dependent
damage growth in asphalt concrete . .Two different
systematic procedures for determining the material-
dependent parameters employed in the constitutive
model using experimental data were described and
found to be effective.
In spite of the accuracy demonstrated in this
paper, it should be emphasized that only uniaxial
stress, constant strain rates were used at one tempera-
ture and therefore the model should be considered as
tentative for its application to asphalt concrete. Fur-
ther study, both theoretical and experimental, is re-
quired to validate (and extend as necessary) the
model so that it may characterize the material's
behavior under more general types of loading condi-
tions; e.g., uniaxial cyclic loading and unloading,
simple shear, biaxial and general three-dimensional
loading. It is encouraging that the type of damage
model used in this paper (S * approach) has been
successfully applied to uniaxial cyclic stress-strain
behavior of asphalt concrete with fine aggregates by
Kim and Little (1990).
The mechanical response of asphalt concrete to
thermal or thermomechanical loading which induces
time- and temperature-dependent damage growth
could also be predicted using the model proposed in
this paper if time that appears in the model is
replaced by reduced time defined by Eq. (16). In
computing the reduced time according to Eq. (16,
the time-temperature shift factor aT given in Fig. 3
would be used. It may be observed that for ther-
morheologically simple viscoelastic materials the idea
of time-temperature superposition, normally appli-
cable to linear viscoelastic behavior without damage,
carries over to viscoelastic behavior with growing
damage if the damage evolution law can be ex-
pressed in the same reduced time; this was the case
for the particle-filIed elastomer studied by Park and
Schapery (1996).
Acknowledgements
The experimental data used in this paper were
generated from a research project sponsored by Texas
A&M Research FoundationjWestem Research In-
stitutejFederal Highway Administration. The sup-
port of these organizations and the assistance pro-
vided by H.J. Lee are gratefully acknowledged.
References
Ferry, J.D. (1980), Viscoelaslic Properties oj Polymers, John
Wiley & Sons, Inc., New York.
Gazonas, G.A. (J 993), A uniaxial nonlinear viscoelastic constitu-
tive model with damage for M30 gun propellant, Mech.
Mater., 15, 323-335.
Harper, B.D. (1986), A uniaxial nonlinear viscoelastic constitutive
relation for ice, J. Energy Resour. Techno/., 108, 156-160.
Harper, B.D. (! 989), Some implications of nonlinear viscoelastic
constitutive theory regarding interrelationships between creep
and strength behavior of ice at failure, J. Offshore Mech.
Arc/ic Eng., 111, 144-148.
Ju, J.W. (1989), On energy-based coupled elastoplastic damage
theories: Constitutive modeling and computational aspects,
Int. J. Solids Struct., 25(7), 803-833.
Kim, Y.R. and Y.C. Lee (1995), Interrelationships among stiff-
nesses of asphalt-aggregate mixtures, Proc. Assoc. Asphalt
Paving Technol., Vol. 64, pp. 575-609.
Kim, Y.R and D.N. Little (1989), Evaluation of healing in asphalt
concrete by means of the theory of nonlinear viscoelasticity,
Transportation Research Record, 1228, Transportation Re-
search Board, Washington, D.C., pp. 198-210.
Kim, Y.R and D.N. Uttle (1990), One-dimensional constitutive
modeling ofasphalt concrete, J. Eng. Mech., 116(4),751-772.
Krajcinovic, D. (1989), Damage mechanies, Mech. Mater., 8(2-3),
117-197.
Lambom, M.J. and RA. Schapery (1988), An investigation of
deforrnation path-independence of mechanical work in fiber-
reinforced plastics, Proc. 4th Japan-US. Conjerence on Com-
posite Materials, AS ME Special Technical Publication, Book
No. AMD, Vol. 92, pp. 991-997.
Lambom, M.J. and RA. Schapery (1993), An investigation of the
existence of a work potential for fiber-reinforced plastic, J.
Compos. Mater., 27(4), 352-382.
Leaderman, H. (J 943), E/as/ic and Creep Properties oj Filamen-
tOlIS Materials and O/her High Polymers, The Textile Founda-
lion, Washington, D.C.
Pagen, C.A. (J 965), Rheological response of bituminous concrete,
Highway Research Record, Vol. 67, Highway Research Board,
pp. 1-26.
S. W. Park et al.I Mechanics of Materia/s 24 (]996) 241-255 255
Park, S.W., (1994), Development of a Nonlinear Thenno-
Viscoelastic Constitutive Equation for Particulate Composites
with Growing Damage, Ph.D. Dissertation, The University of
Texas, Austin, TX.
Park, S.W. and R.A. Schapery (1996), A viscoe1astic constitutive
model for particulate composites with growing danlage. 1nt. J.
Solids Struct., in press.
Peng, S.TJ. (1992), Constitutive equations of solid propellant
with volume di1atation under multiaxial loading - Theory of
di1atation and dewetting criterion, JANNAF Propu/sion Meet-
ing, Indianapolis, IN.
Reddy, J.N. and M.L. Rasmussen (1982), Advanced Engineering
Analysis, John Wiley & Sons, pp. 203-205.
Schapery, R.A. (1975), A theory of crack initiation and growth in
viscoelastic media, Part I: Theoretical development, Part II:
Approximate methods of ana!ysis, Part III: Analysis of contin-
uous growth, int. J. Fract., 11, 141-159, 369-388, 549-562.
Schapery, R.A. (1981), On viscoeJastic deformation and failure
behavior of composite materials with distributed flaws, Ad-
uances in Aerospace Structures and Materia/s - ADOl,
ASME, New York, pp. 5-20.
Schapery, R.A. (1982), Models for damage growth and fracture in
nonlinear viscoelastic particulate composites, Proc. NinetJl
U.S. National Congress of Applied Mechanics, Book No.
H00228, ASME, New York, pp. 237-245.
Schapery, R.A. (1984), Correspondence principles and a general-
ized J integral for Iarge deformation and fracture analysis of
viscoelastic media, Int. J. Fract., 25, 195-223.
Schapery, R.A. (1986), A micromechanical model for nonlinear
viscoelastic behavior of particle-reinforced rubber with dis-
tributed damage, Eng. Fract. Mech., 25, 845-867.
Schapery, R.A. (1987a), Deformation and fracture characterization
of ine\astic composite materiais using potentials, Po/ym. Eng.
Sd., 27, 63-76.
Schapery, R.A. (1987b), Nonlinear constitutive equations for solid
propellant based on a work potential and micromechanical
model, Proc. 1987 JANNAF Struct. and Mech. Behauior
Meet., CPIA, Huntsville, AL.
Schapery, R.A. (1990a), A theory of mechanical behavior of
elastic media with growing damage and other changes in
structure, J. Mech. Phys. Solids, 38, 215-253.
Schapery, R.A. (1990b), Simplifications in the behavior of vis-
coelastic composites with growing damage, Proc. iUTAM
Symp. on Inelastic Deform. of Compos. Mater., Troy, NY,
Springer, New York-Wien, pp. 193-214.
Schapery, R.A. (1994), N onlinear viscoelastic constitutive equa-
tions for composites based on work potentiaIs, Mechanics
USA 1994 Proc. 12th US Nat. Congress of Appl. Mech., App!.
Mech. Re!}., Vol. 47, pp. 269-275.
Schapery, R.A. and Sicking, D.L. (1995), On nonlinear constitu-
tive equations for elastic and viscoelastic composites with
growing damage, in: Mechanica/ Behavior of Materia/s, ed.
A. Bakker, Delft University Press, Delft, the Netherlands, pp.
45-76.
Simo, J.C. CI 987), On a fully three-dimensional finite strain
viscoelastic damage model: Formulation and computational
aspects, Comput. Methods App!. Mech. Eng., 60, 153-173.
Sjolind, S.G. (1987), A constitutive model for ice as adamaging
viscoelastic material, Co/d Reg. Sd. Technol., 41, 247-262.
Weitsman, Y. (1988), A continuum damage model for viscoelastic
materiais, J. App!. Mech., 55, 773-780.

Você também pode gostar