You are on page 1of 11

Applied Mathematical Modelling 31 (2007) 381–391

www.elsevier.com/locate/apm

On time-dependent behavior of cross-ply laminated strips


with viscoelastic interfaces
Wei Yan a, W.Q. Chen a,b,*
, B.S. Wang a

a
Department of Civil Engineering, Zhejiang University, Zheda Road 38, Hangzhou 310027, PR China
b
State Key Lab of CAD&CG, Zhejiang University, Hangzhou 310027, PR China

Received 1 October 2004; received in revised form 1 August 2005; accepted 1 November 2005
Available online 27 December 2005

Abstract

The two-dimensional problem of a simply supported laminated orthotropic strip with viscoelastic interfaces under static
loading is studied. State-space formulations are developed based on the exact elasticity equations governing orthotropic
media and the Kelvin–Voigt constitutive relation of interfaces. Since the response of the strip is time-dependent, the power
series expansion technique is adopted to model the variations of elastic fields with time. Results show that the response of
the laminated strip with viscoelastic interfaces changes remarkably with time, which is also significantly different from that
of a plate with perfect interfaces or with viscous interfaces. Note that from the present analysis, the response for a lam-
inated plate with spring-like interfaces or with viscous interfaces can be easily obtained because they are just two particular
cases of the present Kelvin–Voigt model.
Ó 2005 Elsevier Inc. All rights reserved.

Keywords: Kelvin–Voigt model; Viscoelastic interfaces; Orthotropic laminate; State-space method

1. Introduction

The interface region is a subject of particular concern in composites. Interfacial properties can strongly
influence the behavior of the composite under thermal, mechanical, and environmental conditions arising
in service. In the traditional theory, the interfaces are treated as perfectly bonded in the most of the existing
research works, where the traction and displacement vectors are continuous across the interfaces [1–3]. How-
ever, from a strictly physical point of view, the existence of a perfect interfacial bond in a real laminated com-
posite seems impossible. Various flaws, such as microcracks, inhomogeneities, and cavities, can be introduced
into the bond in the process of fabrication. During the service lifetime, the structure will be subjected to var-
ious loads and exposed to corrosive environment. Consequently, these tiny flaws can get significant and finally

*
Corresponding author. Address: Department of Civil Engineering, Zhejiang University, Zheda Road 38, Hangzhou 310027, PR China.
Tel.: +86 571 879 52284; fax: +86 571 879 52165.
E-mail address: chenwq@ccea.zju.edu.cn (W.Q. Chen).

0307-904X/$ - see front matter Ó 2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.apm.2005.11.014
382 W. Yan et al. / Applied Mathematical Modelling 31 (2006) 381–391

lead to the local failure of bond or even the whole collapse of structure. For example, a radial tire is a very
complex structure made from rubber elastomers and fiber–rubber composite materials. During its use, exten-
sion propagation of interface crack between belts can occur, which obviously affects its durability and life [4].
Therefore, an appropriate model accounting for the effect of imperfect bonding should be regarded to be more
realistic in the study of composite laminates. The influence of interlaminar bonding imperfections on the glo-
bal mechanical response of laminated structures has been recognized. It has also been possible to show the
impact of interfacial relaxation on the interlaminar stress distribution. Bui et al. [5] discussed the effect of
weakened interfaces on the distribution of interlaminar stresses and the change in strain-energy release rates.
Shu and Soldatos [6] proposed a highly accurate plate theory for analyzing laminates with weakly bonded lay-
ers. Wang and Zhong [7] derived an exact solution of smart laminated anisotropic infinite circular cylindrical
shells with imperfect bonding. Fan and Wang [8] formulated mathematically the interaction of a screw dislo-
cation with an imperfect interface. Hashin [9] studied that the imperfect interface conditions involving dis-
placement and traction discontinuities or jumps. It turns out that for very compliant interphase there are
significant displacement jumps and negligible traction jumps, while for very stiff interphase the situation is
reversed; the displacement jumps are insignificant and the traction jumps may be considerable. Note that
in available works related to composites with imperfect interfaces, most researchers paid their attention to
the interface conditions involving displacement jumps [10–19].
For a practical laminate, it is generally very difficult to predict the exact behavior of the interlaminar bond-
ing either theoretically or experimentally. Some simplified interfacial models thus have been introduced. The
most popular one is the linear spring-like model. For theoretical analysis, a thin layer of interphase material is
introduced between two adjacent layers, which should be representative for interlaminar adhesives. When the
thickness of this layer vanishes, the interfacial tractions become continuous, while the displacements at either
side of the interface layer become discontinuous. Micromechanics analysis shows that the jump in displace-
ment is linearly proportional to the interfacial traction [11] and the proportional constants are the effective
interface parameters (spring-like constant) [10–17]. Recently, the dislocation-like model has been proposed
to describe mathematically the effect of an imperfect interface on the load transfer. The boundary conditions
for this new model are similar to the linear spring-like model except that the jump in displacement at the inter-
face is assumed to be linearly proportional to the displacement at the interface of the constituent where the
load is applied [18,19].
In the related works mentioned above, the responses of laminates under static loading are independent of
the time variable. Recently, He and Jiang [20] derived an exact two-dimensional solution for isotropic lami-
nates with viscous interfaces and showed that the response of laminates varies with time remarkably, especially
at the initial stage. Chen and Lee [21] proposed an efficient and accurate semi-analytical method for the anal-
ysis of angle-ply laminates in cylindrical bending with viscous interfaces. Both studies revealed that, when time
approaches infinity, the viscous interfaces would lose the ability of transferring shear stress totally. However,
this seems unsuitable for certain types of practical composites, especially within the framework of small defor-
mation. In fact, a viscoelastic interface will be more appropriate for characterizing the creep and relaxation
behavior of interlaminar bonding material under high temperature circumstance [22]. According to Fan
and Wang [23], the viscoelastic behavior should be considered when the working temperature of a solid is
above 1/3 to 1/2 of its melting temperature (Kelvin scale). There are plenty of cases where the interface should
be considered as viscoelastic. As an example, let us consider a case where two pieces of metal (e.g. aluminum)
are jointed by a ‘‘glue’’ (e.g. epoxy) of a lower melting temperature. The melting temperature for aluminum is
about 933 K, while the melting temperature for epoxy is about 340–380 K. If this joint piece is working at
room temperature (300 K), the materials are considered as linear elastic, while the interface should be consid-
ered as viscoelastic.
More recently, Yan and Chen [24] derived an analytical solution of a layered isotropic plate with viscoelas-
tic interfaces in cylindrical bending. The most significant difference from the viscous interfaces is that the vis-
coelastic ones will not lose the ability of transferring shear stress when t ! 1, which should be more realistic
for certain practical situations. However, it should be pointed out that the exact elasticity analysis becomes
computationally expensive when the number of layers in the laminate increases, because of a large number
of integration constants that are involved. Moreover in the case of an anisotropic laminate with viscoelastic
interfaces, the analysis presented by Yan and Chen [24] becomes very complicated because it is also necessary
W. Yan et al. / Applied Mathematical Modelling 31 (2006) 381–391 383

to study different cases of material eigenvalues to obtain real form solutions, as shown by Pagano [1–3] in the
analysis of perfect laminates.
In this paper, we investigate the responses of a simply supported laminated orthotropic plate in cylindrical
bending with viscoelastic interfaces, subjected to a sinusoidal static loading. The layer materials are assumed
to be linearly elastic, while the interfaces are characterized by a Kelvin–Voigt viscoelastic relation. The anal-
ysis is presented based on the state-space formulations. For simplicity, we assume that the deformation is rel-
atively slow so that the inertia terms can be neglected in the analysis. A special technique using power series
expansion is adopted to approximate the variations of elastic fields with time, which was originally suggested
by Chen and Lee for laminates with viscous interfaces [21]. Finally, numerical results are presented and
discussed.

2. State-space equations

Consider a laminate composed of n orthotropic layers such that the various axes of material symmetry are
parallel to the plate axes x, y, z as shown in Fig. 1. The body is in a state of plane strain with respect to the xz
plane and is simply supported on the ends x = 0 and x = l. The constitutive relations then read as Pagano [1]
 
ou ow ou ow ou ow
rx ¼ c11 þ c13 ; rz ¼ c13 þ c33 ; sxz ¼ c55 þ ; ð1Þ
ox oz ox oz oz ox
where cij are elastic constants, ri and sij are the normal and shear stress components, respectively. The equi-
librium equations, in absence of body forces and with the neglect of inertia terms are
orx osxz osxz orz
þ ¼ 0; þ ¼ 0. ð2Þ
ox oz ox oz
With a routine derivation, the following state equation can be obtained from Eqs. (1) and (2):
2 3
o
6 0 0 0  7
8 9 6 ox 78 9
> rz > 6 o 1 7> rz >
> >
> > 6  7> >
o < u = 6 0 0
ox c < u >
7> =
6 55 7
¼6 1 c13 o 7 ; ð3Þ
oz >
> w> > 6
0 0
7>
> w> >
> >
: ; 6 c33 6 7> >
sxz c33 ox 7: s ;
6 7 xz
4 c13 o o 2 5
b 2 0 0
c33 ox ox
where rz, u, w and sxz are termed as state variables, from which rx can be determined as
c13 ou
rx ¼ rz  b ð4Þ
c33 ox
in which
c2
b ¼ 13  c11 . ð5Þ
c33

Fig. 1. Sketch of a plate in cylindrical bending.


384 W. Yan et al. / Applied Mathematical Modelling 31 (2006) 381–391

The simple support conditions are [1]


rx ¼ w ¼ 0 at x ¼ 0 and x ¼ l. ð6Þ
To satisfy these conditions, we assume
8 9 8 ð1Þ 9
>
> rz >> >
> c11 r z ðf; tÞ sinðnpnÞ >
>
> >
> > > > >
>
< u >
> = > < h >
uðf; tÞ cosðnpnÞ =
¼ ; ð7Þ
>
> w> > >
> wðf; tÞ sinðnpnÞ >
h >
>
> > >
> > > >
>
>
: ; > : ð1Þ >
;
sxz c11 sxz ðf; tÞ cosðnpnÞ
ð1Þ
where f = z/h and n = x/l are dimensionless coordinates, c11 represents the elastic constant of the first layer
(the bottom layer), and a quantity with overbar indicates the dimensionless one. The substitution of Eq.
(7) into Eq. (3) yields
o
Vðf; tÞ ¼ AVðf; tÞ; ð8Þ
of
where Vðf; tÞ ¼ ½ rz ðf; tÞ u ðf; tÞ w ðf; tÞ sxz ðf; tÞ T , and
2 3
0 0 0 k
6 ð1Þ 7
6 c11 7
6 0 0 k 7
6 c55 7
6 7
6 7
A¼6 6 c11
ð1Þ
c13
7;
7 ð9Þ
6 k 0 0 7
6 c33 c 33 7
6 7
6 7
4 c13 b 2 5
k  ð1Þ k 0 0
c33 c11
where k = nps and s = h/l is the thickness-to-span ratio. The solution to Eq. (8) can be obtained as
Vðf; tÞ ¼ exp½Aðf  fk1 ÞVðfk1 ; tÞ ðfk1 6 f 6 fk ; k ¼ 1; 2; . . . ; nÞ; ð10Þ
Pk
where f0 = 0, fk ¼ zk =h ¼ j¼1 hj =h and hk is the thickness of the kth layer.
Setting f = fk in Eq. (10), the relationship between the state variables at the upper and lower surfaces of the
kth layer can be established
ðkÞ ðkÞ
V 1 ¼ Mk V 0 ; ð11Þ
ðkÞ ðkÞ
where V1and are the state vectors at the upper and lower surfaces, respectively, of the kth layer, and
V0
Mk = exp[A(fkfk1)] is the transfer matrix. Similarly, we get
ðkþ1Þ ðkþ1Þ
V1 ¼ Mkþ1 V0 . ð12Þ

3. Kelvin–Voigt viscoelastic interfacial model

As shown in Fig. 1, the following boundary conditions at the uppermost and lowermost surfaces must be
satisfied:
rzðnÞ ¼ p0 sinðnpx=lÞ; ðnÞ
sxz ¼0 at z ¼ h;
ð13Þ
rzð1Þ ¼ 0; ð1Þ
sxz ¼ 0 at z ¼ 0.
On the other hand, we have at the viscoelastic interface
rzðkþ1Þ ¼ rðkÞ
z ; sðkþ1Þ
xz ¼ sðkÞ
xz ;
ð14Þ
uðkþ1Þ ¼ uðkÞ þ dðkÞ ; wðkþ1Þ ¼ wðkÞ at z ¼ zk ;
W. Yan et al. / Applied Mathematical Modelling 31 (2006) 381–391 385

where d(k) is the relative sliding displacement at the kth interface. In this paper, we assume that the shear stress
and the sliding obey the Kelvin–Voigt viscoelastic law [23]
ðkÞ ðkÞ ðkÞ ðkÞ
sðkÞ
xz ¼ g0 d þ g1 d_ at z ¼ zk ; ð15Þ
ðkÞ ðkÞ
where d_ is the sliding velocity (the dot over a quantity denotes differentiation with respect to time), and g0
ðkÞ
and g1 are the elastic constant and viscous coefficient, respectively, of the viscoelastic interface. Setting
ðkÞ ðkÞ
g0 ¼ 0, we get the viscous model studied by He and Jiang [20]. Setting g1 ¼ 0, the constitutive relation given
in Eq. (15) degenerates to the linear spring model [6,10,25], for which there is no time effect in the solution
since the viscous-effect vanishes.
We assume
ðkÞ
dðkÞ ¼ h
d ðtÞ cosðnpnÞ. ð16Þ
Thus, Eq. (14) can be written in a matrix form:
ðkþ1Þ ðkÞ
V0 ¼ V1 þ QðkÞ ; ð17Þ
ðkÞ
¼ ½0; 
ðkÞ T
where Q d ; 0; 0 , and Eq. (15) can be rewritten as follows
ðkÞ
ðkÞ ðkÞ ðkÞ dd
sðkÞ g0 
xz ¼  d þ g1 at z ¼ zk ; ð18Þ
ds
where gðkÞ ðkÞ ð1Þ ð1Þ ð1Þ
0 ¼ g0 h=c11 , are the dimensionless elastic constants, s ¼ c11 t=ðg1 hÞ is the dimensionless time, and
ðkÞ
g
ðkÞ ð1Þ
1 ¼ g1 =g1 are the viscosity ratios.

4. Analysis procedure

It would be difficult to obtain an exact solution based on the above state-space formulations. Note that
although an exact solution can be obtained in a manner similar to that suggested by Yan and Chen [24],
the analysis is computationally expensive as mentioned earlier. In this paper, to use the state-space formula-
tions, we adopt an approximate treatment of interfacial conditions using power series expansions [21]. To this
end, we divide the time domain into a series of equal intervals [0,Ds], [Ds, 2Ds], [2Ds, 3Ds], . . . , each with a small
length of Ds. For a typical interval [mDs, (m + 1)Ds] (m = 0, 1, 2, . . .), it is assumed that
ðkÞ ðkÞ ðkÞ 2 ðkÞ 3 ðkÞ

d ¼ dm;1 þ ðs  mDsÞ 
dm;0 þ ðs  mDsÞ dm;2 þ ðs  mDsÞ dm;3 þ   
ðkÞ ðkÞ ðkÞ ðkÞ
ðkÞ
rz ¼ r rm;1 þ ðs  mDsÞ2 r
m;0 þ ðs  mDsÞ m;2 þ ðs  mDsÞ3 r
m;3 þ   
ðkÞ 2 ðkÞ 3 ðkÞ ðkÞ
uðkÞ ¼ 
 um;0 þ ðs  mDsÞ
um;1 þ ðs  mDsÞ um;2 þ ðs  mDsÞ um;3 þ    ð19Þ
ðkÞ ðkÞ ðkÞ ðkÞ
 ðkÞ ¼ w
w wm;1 þ ðs  mDsÞ2 w
 m;0 þ ðs  mDsÞ  m;2 þ ðs  mDsÞ3 w
 m;3 þ   
ðkÞ 2 ðkÞ 3 ðkÞ ðkÞ
sðkÞ
xz ¼ 
sm;0 þ ðs  mDsÞsm;1 þ ðs  mDsÞ sm;2 þ ðs  mDsÞ sm;3 þ   
ðkÞ
Obviously, we have  d0;0 ¼ 0 because of the zero initial condition at t = s = 0; this corresponds to the case of
a plate which is already loaded before the interfaces exhibit the viscoelastic behavior. In view of Eq. (19), by
equating coefficients of the same order of s  mDs at the two sides of Eq. (17), we obtain
ðkÞ ðkÞ
Vkþ1
0;m;i ¼ V1;m;i þ Qm;i ðm; i ¼ 0; 1; 2; . . .Þ; ð20Þ
ðkÞ ðkÞ
where Qm;i ¼ ½0; 
T
dm;i ; 0; 0 , which becomes zero for a perfect interface. By virtue of Eq. (19), we obtain from
Eq. (18) by equating coefficients of the same order of s  mDs at the two sides
ðkÞ ðkÞ ðkÞ ðkÞ ðkÞ
 g0 
dm;i ¼ ðsm;i1   dm;i1 Þ=ð
g1  iÞ; ði ¼ 1; 2; 3; . . .Þ. ð21Þ
From Eqs. (11), (12), (19) and (20), we get the relations
ðkþ1Þ ðkÞ ðkÞ
V1;m;i ¼ Mkþ1 V1;m;i þ Mkþ1 Qm;i ðm; i ¼ 0; 1; 2; . . .Þ. ð22Þ
386 W. Yan et al. / Applied Mathematical Modelling 31 (2006) 381–391

Continuing the above procedure layer by layer, we finally obtain


ðnÞ ð1Þ
V1;m;i ¼ TV0;m;i þ Sm;i ðm; i ¼ 0; 1; 2; . . .Þ; ð23Þ
Q1
where T ¼ j¼n Mj is the global transfer matrix and

ðn1Þ ðn2Þ
Y
2
ð1Þ
Sm;i ¼ Mn Qm;i þ Mn Mn1 Qm;i þ  þ Mj Qm;i ; ð24Þ
j¼n

are the inhomogeneous terms associated with viscoelastic interfaces, which vanish in the case of a perfectly
bonded laminate. The boundary conditions shown in Eq. (13) are expressed by
ðnÞ ð1Þ ðnÞ ðnÞ T
V1 ¼ ½p0 =c11 ;   1 ; 0 ;
u1 ; w
ð1Þ ð1Þ ð1Þ T
ð25Þ
V0 ¼ ½0;   0 ; 0 .
u0 ; w
By applying the above boundary conditions in Eq. (23), we obtain
8 ð1Þ
9ðnÞ 8 9ð1Þ
>
> p0 =c11 > > > 0 >
>
> > > >
< ðnÞ > = >
< ð1Þ >
u0 =

u1
¼T ð1Þ þ Sm;0 ðm ¼ 0; 1; 2; . . .Þ; ð26Þ
> 1 >
ðnÞ > 0 >
>
> w
>
>
>
> >w
>
:
>
>
;
: ;
0 m;0
0 m;0
8 9ðnÞ 8 9ð1Þ
> 0 > > 0 >
> ðnÞ >
> > > ð1Þ >
> >
< 
u1 = < u0 =

ðnÞ ¼ T ð1Þ
þ Sm;i ðm ¼ 0; 1; 2; . . . ; i ¼ 1; 2; 3; . . .Þ. ð27Þ
>
> 1 >
w > >
> 0 >
w >
>
: >
; >
: >
;
0 m;i 0 m;i

Since S0,0 = 0, the corresponding state variables at the top and bottom surfaces of the plate at s = 0 can be
solved from Eq. (26), from which all physical variables in the plate can be obtained (see Eqs. (28) and (4)).
These in turn determine the inhomogeneous term S0,1 by virtue of Eq. (21). Then from Eq. (27), all physical
variables for i = 1 can be computed. Continuing this procedure, we can eventually obtain all the coefficients in
Eq. (19) for m = 0. Now by setting s = Ds in Eq. (19), we can further get all the physical variables at s = Ds.
This gives the value of S1,0. By applying the same procedure described above for m = 0 to m = 1 and so on, all
physical variables at any time can be determined.
It is noted that when the state variables at the bottom surface are determined from Eq. (26) or (27), the ones
at any interior point can be calculated by the following equation:

ðkÞ
Y
1
ð1Þ ðk1Þ ðk2Þ ðk3Þ
Vm;i ðfÞ ¼ exp½Aðf  fk1 Þ Mj V0;m;i þ Qm;i þ Mk1 Qm;i þ Mk1 Mk2 Qm;i þ 
j¼k1
!
Y
2
ð1Þ
þ Mj Qm;i ðfk1 6 f 6 fk ; m; i ¼ 0; 1; 2; . . .Þ. ð28Þ
j¼k1

5. Numerical computation

In this paper, we assume that the normal sinusoidal pressure p ¼ p0 sinðpnÞ is applied on the top surface
of the plate. The following non-dimensional quantities are introduced:
ð1Þ
rz ðl=2; z; tÞ sxz ð0; z; tÞ c11 wðl=2; z; tÞ
r¼ ; s0 ¼ ; w0 ¼ ;
p0 p0 p0 h
ð1Þ ð1Þ
ð29Þ
c uð0; z; tÞ c dð0; tÞ
u0 ¼ 11 ; d0 ¼ 11 .
p0 h p0 h
W. Yan et al. / Applied Mathematical Modelling 31 (2006) 381–391 387

where the superscript is omitted for brevity, which indicates the sequence number of the layer, just as that done
earlier in this paper.
First, we consider a symmetric three-layered isotropic strip, of which the Poisson’s ratio of each layer (of
same thickness) is the same and is denoted by l. The exact solution derived by Yan and Chen [24] for this
problem enables us to check the method proposed here through numerical comparison. We take l = 0.3,
ð1Þ ð2Þ ð1Þ ð2Þ
E2 = 3E1 = 3E3,  g0 ¼ g0 ¼ 0:01; g1 ¼ g1 , and s = 0.1 in the numerical calculation. For isotropic materi-
als, the elastic stiffness coefficients cij are determined by
Eð1  lÞ El E
c11 ¼ c33 ¼ ; c13 ¼ ; c55 ¼ . ð30Þ
ð1 þ lÞð1  2lÞ ð1 þ lÞð1  2lÞ 2ð1 þ lÞ
The parameter M in Table 1 (and also in Table 2) represents the number of total terms adopted in the series
expansions in Eq. (19) in the calculation. As shown in Table 1, the present method converges rapidly to the
exact solution. However, as shown in our calculation, the exact solution is very sensitive to the finite length of
decimal digits retained in formulations, certain numerical deviation is observed from these results.
Next, we consider a symmetric five-layered orthotropic strip with viscoelastic interfaces. Each layer in the
strip is of the same thickness and the following parameters are adopted:
EL =ET ¼ 40; GTT =ET ¼ 0:2; GTL =ET ¼ 0:5; lLT ¼ lTT ¼ 0:25; s ¼ 0:1;
ð2Þ ð4Þ ð1Þ ð3Þ ð2Þ ð4Þ ð1Þ ð3Þ
ð31Þ
g0
 ¼ g0
 ¼ 2
g0 ¼ 2
g0 ¼ 0:02; g1 ¼ g1 ¼ 2g1 ¼ 2g1 ;
where E is Young’s modulus, G the shear modulus, L signifies the direction parallel to the fibers, T the trans-
verse direction, and lLT is Poisson’s ratio measuring strain in the transverse direction under uniaxial normal
stress in the L direction. The results are given in Table 2; the good convergence character of the proposed
method is also evident for the orthotropic strip.
From Tables 1 and 2, it is obvious that the numerical results are of high precision when M = 4, Ds = 0.05.
So in the following numerical computations, we will always take M = 4 and Ds = 0.05.
Again, the above five-layered orthotropic strip will be analyzed, but with different viscoelastic interfaces.
Here, we consider the following two cases: (1) For Case 1, the second and fourth interfaces are assumed to
ð1Þ ð3Þ ð3Þ ð1Þ
be perfect, while the first and third ones are viscoelastic with g0 ¼ 2g0 ¼ 0:02; g1 ¼ 2g1 , and (2) for Case
ð2Þ ð4Þ ð1Þ ð3Þ
2, the second and fourth interfaces are viscous with g1 ¼ g1 ¼ g1 ¼ 0:5g1 , while the first and third ones are
ð1Þ ð3Þ
g0 ¼ 2
viscoelastic with  g0 ¼ 0:02.

Table 1
Comparison for a three-layered isotropic stripa
(M, Ds) s=0 s=5
r s0 w0 u0 r s0 w0 u0
(2, 0.1) 0.240746 3.94040 1427.82 73.9176 0.239818 3.84868 1585.52 62.2259
(3, 0.1) 0.240746 3.94039 1427.83 73.9168 0.239815 3.84839 1586.03 62.1881
(4, 0.1) 0.240746 3.94039 1427.83 73.9167 0.239815 3.84837 1586.07 62.1853
(2, 0.05) 0.240756 3.94137 1426.14 74.0417 0.239850 3.85178 1580.20 62.6203
(3, 0.05) 0.240756 3.94137 1426.14 74.0416 0.239850 3.85171 1580.32 62.6110
(4, 0.05) 0.240756 3.94137 1426.14 74.0416 0.239850 3.85170 1580.33 62.6106
[24] 0.2408 3.9423 1424.5 74.160 0.2399 3.8539 1576.6 62.885
s = 10 s = 15
(2, 0.1) 0.238954 3.76340 1732.10 51.3572 0.238150 3.68410 1868.35 41.2539
(3, 0.1) 0.238949 3.76284 1733.07 51.2853 0.238141 3.68329 1869.75 41.1501
(4, 0.1) 0.238948 3.76280 1733.14 51.2799 0.238141 3.68322 1869.85 41.1423
(2, 0.05) 0.239005 3.76847 1723.39 52.0030 0.238219 3.69100 1856.49 42.1333
(3, 0.05) 0.239004 3.76833 1723.63 51.9851 0.238217 3.69080 1856.83 42.1074
(4, 0.05) 0.239004 3.76832 1723.64 51.9844 0.238217 3.69079 1856.85 42.1065
[24] 0.2390 3.7697 1721.19 52.166 0.2382 3.6898 1858.60 41.976
a
All data are evaluated for f = 1/3.
388 W. Yan et al. / Applied Mathematical Modelling 31 (2006) 381–391

Table 2
Convergence study of the present method for a five-layered orthotropic stripa
(M, Ds) s = 10 s = 50
r s0 w0 u0 r s0 w0 u0
(3, 0.1) 0.353282 4.52492 1833.09 26.5576 0.355548 4.36808 3179.48 16.3370
(3, 0.05) 0.353250 4.52651 1819.13 27.0064 0.355460 4.37369 3131.14 14.7938
(4, 0.05) 0.353250 4.52651 1819.14 27.0060 0.355460 4.37368 3131.19 14.7955
a
All data are evaluated for f = 0.4.

The through-thickness distributions of the four non-dimensional state variables defined in Eq. (29) of these
two cases are shown in Figs. 2 and 3, respectively. The response of the strip at t = 0 is the same as a strip with
perfectly bonded interfaces, as stated earlier in the paper. With the occurrence of interfacial sliding, the sig-
nificant stress relaxation occurs, and the four non-dimensional state variables change remarkably with time.
It can be seen from Figs. 2(a) and 3(a) that the distribution lines of the transverse displacement seem vertically
straight, which indicates that the variation of w0 is relatively smaller along the thickness. Although w0 for Case
1 is nearly equivalent to that for Case 2 at s = 0, this situation changes at a later instance, when w0 for Case 2

Fig. 2. Distributions of field variables along the thickness in the five-layered strip (Case 1).
W. Yan et al. / Applied Mathematical Modelling 31 (2006) 381–391 389

Fig. 3. Distributions of field variables along the thickness in the five-layered strip (Case 2).

increases more greatly than that for Case 1 since the plate rigidity will reduced more significantly with the
appearance of viscous interfaces. It also can be seen from Fig. 2(b) that the variation of r with time is nearly
negligible. However, when there are viscous interfaces, the distribution of r will change with time more sig-
nificantly, as shown in Fig. 3(b).
Fig. 2(c) displays the time dependency of the shear stress. At the perfect interfaces, i.e. f = 0.4 and f = 0.8,
the curves always keep smooth. However, the variation of the curve is abrupt across the viscoelastic interfaces,
i.e. f = 0.2 and f = 0.6. This indicates that a relaxation occurs in the shear stress, and the two viscoelastic
interfaces lose their part capability of transferring shear stress. With time increasing, the distribution of shear
stress eventually arrives at a final state (s ! 1). From Fig. 3(c), we can see that shear stresses are nearly zeros
at the two viscous interfaces when s = 1000. In fact, they shall eventually vanish when s ! 1, because the
interfaces will lose the capability of transferring shear stress completely. Conversely, the viscoelastic interfaces
always holds the function of transferring shear stress, which should be more realistic for certain practical sit-
uations. Furthermore, with time increasing, the shear stress at f = 0.2 (viscoelastic interface) becomes larger as
shown in Fig. 3(c). This is also crucial since interfacial high shear stress is usually a main cause of failure of
composite laminates.
Besides, discontinuities of u0 at viscoelastic interfaces caused by the sliding are shown in Fig. 2(d). The dif-
ference between u0 at the two sides of an interface is just the sliding displacement, i.e. d(k). The sliding at the
third interface is more significant than that at the first interface. That is true because the elastic constant of the
ð1Þ ð3Þ
former is smaller than that of the latter ð g0 ¼ 2 g0 Þ. However, the viscosity of the former is bigger than that
390 W. Yan et al. / Applied Mathematical Modelling 31 (2006) 381–391

Fig. 4. Variation of sliding displacement d(1) with time (Case 2).

ð3Þ ð1Þ
of the latter ð
g1 ¼ 2g1 Þ. So, the influence of the elastic constant on the sliding displacement is more signif-
icant than that of the viscosity. The similar characteristic can also be seen from Fig. 3(d).
As a primary assumption of the analysis, the inertia terms in the equations of motion are neglected. This
corresponds to the well-known quasi-static analysis [26]. To check this assumption, we take the following typ-
ð1Þ
ical material constants while keeping all the dimensionless constants unchanged: g1 ¼ 1:611  1016 Pa s=m
ð1Þ 10
[27], and c11 ¼ 6:65  10 Pa. In addition, we take h = 0.1 m and p0 = 6.65 MPa. The variation of sliding dis-
placement at the first interface with time is shown in Fig. 4, for Case 2. It is seen that the variation is very
gentle, so that the inertia effect can be ignored in our analysis.

6. Conclusions

The response of a simply supported cross-ply laminated strip with viscoelastic interfaces is investigated
based on state-space formulations. A special technique using power series expansion is adopted to approxi-
mate the variations of elastic fields with time, and the numerical computation shows that this method con-
verges rapidly. The Kelvin–Voigt constitutive relation is employed to describe the viscoelastic behavior of
interfaces. This constitutive law covers two particular cases, i.e. the spring-like model and the viscous model.
Thus the present analysis is more universal than those available in literature.
Results indicate that the response of a laminated strip with viscoelastic interfaces changes remarkably with
time. Thus, if a laminated plate with viscoelastic interfaces is treated by an effective elastic laminate for the sim-
plification of design or initial analysis, then the elastic moduli of the effective plate should vary with time to
capture this time-dependent feature. The response of a laminated strip with viscoelastic interfaces is also found
to be greatly different from that of a plate with perfect interfaces or with viscous interfaces. The prominent fea-
ture of the viscoelastic interfaces is that they always hold the function as interlaminar bonds, although weak-
ened. This should be more realistic for composites under an environment of relatively higher temperature.
In this paper, we consider only a two-dimensional problem (plane strain or cylindrical bending), and each
layer in the strip is elastically orthotropic. Nevertheless, the present work is representative since the analysis is
based on state-space formulations and can be readily applied to more complicated configurations [17]. More
importantly, the results obtained here may provide a useful means of comparison for future research on more
complicated problems.

Acknowledgement

This work was supported by the Natural Science Foundation of China (Nos. 10372088 and 10432030).
W. Yan et al. / Applied Mathematical Modelling 31 (2006) 381–391 391

References

[1] N.J. Pagano, Exact solutions for composite laminates in cylindrical bending, J. Compos. Mater. 3 (1969) 398–411.
[2] N.J. Pagano, Exact solutions for rectangular bi-directional composites and sandwich plates, J. Compos. Mater. 4 (1970) 20–35.
[3] N.J. Pagano, Influence of shear coupling in cylindrical bending of anisotropic laminates, J. Compos. Mater. 4 (1970) 330–343.
[4] X.J. Feng, X.Q. Yan, Y.T. Wei, X.G. Du, Analysis of extension propagation process of interface crack between belts of a radial tire
using a finite element method, Appl. Math. Model. 28 (2004) 145–162.
[5] V.Q. Bui, E. Marechal, H. Nguyen-Dang, Imperfect interlaminar interfaces in laminated composites: interlaminar stresses and strain-
energy release rates, Compos. Sci. Tech. 60 (2000) 131–143.
[6] X.P. Shu, K.P. Soldatos, An accurate de-lamination model for weakly bonded laminates subjected to different sets of edge boundary
conditions, Int. J. Mech. Sci. 43 (2001) 935–959.
[7] X. Wang, Z. Zhong, Three-dimensional solution of smart laminated anisotropic circular cylindrical shells with imperfect bonding, Int.
J. Solids Struct. 40 (2003) 5901–5921.
[8] H. Fan, G.F. Wang, Screw dislocation interacting with imperfect interface, Mech. Mater. 35 (2003) 943–953.
[9] Z. Hashin, Thin interphase/imperfect interface in elasticity with application to coated fiber composites, J. Mech. Phy. Solids 50 (2002)
2509–2537.
[10] Y. Benveniste, The effective mechanical behavior of composite materials with imperfect contact between the constituents, Mech.
Mater. 4 (1985) 197–208.
[11] Z. Hashin, The spherical inclusion with imperfect interface, J. Appl. Mech. 58 (1991) 444–449.
[12] Z. Zhong, S.A. Meguid, On the eigenstrain problem of a spherical inclusion with an imperfectly bonded interface, J. Appl. Mech. 63
(1996) 877–883.
[13] W.Q. Chen, J.B. Cai, G.R. Ye, Exact solution of cross-ply laminates with bonding imperfections, AIAA J. 41 (2003) 2244–2250.
[14] W.Q. Chen, J. Ying, J.B. Cai, G.R. Ye, Benchmark solution of laminated beams with bonding imperfections, AIAA J. 42 (2004) 426–
429.
[15] W.Q. Chen, K.Y. Lee, Three-dimensional exact analysis of angle-ply laminates in cylindrical bending with interfacial damage via
state-space method, Compos. Struct. 64 (2004) 275–283.
[16] J.B. Cai, W.Q. Chen, G.R. Ye, Effect of interlaminar bonding imperfections on the behavior of angle-ply laminated cylindrical panels,
Compos. Sci. Tech. 64 (2004) 1753–1762.
[17] W.Q. Chen, Y.F. Wang, J.B. Cai, G.R. Ye, Three-dimensional analysis of cross-ply laminated cylindrical panels with weak interfaces,
Int. J. Solids Struct. 41 (2004) 2429–2446.
[18] H.Y. Yu, A new dislocation-like model for imperfect interfaces and their effect on load transfer, Composites Part A 29A (1998) 1057–
1062.
[19] H.Y. Yu, Y.N. Wei, F.P. Chiang, Load transfer at imperfect interfaces — dislocation-like model, Int. J. Eng. Sci. 40 (2002) 1647–
1662.
[20] L.H. He, J. Jiang, Transient mechanical response of laminated elastic strips with viscous interfaces in cylindrical bending, Compos.
Sci. Tech. 63 (2003) 821–828.
[21] W.Q. Chen, K.Y. Lee, Time-dependent behaviors of angle-ply laminates with viscous interfaces in cylindrical bending, Euro. J. Mech.
A/Solids 23 (2004) 235–245.
[22] Z. Hashin, Composite materials with viscoelastic interphase: creep and relaxation, Mech. Mater. 11 (1991) 135–148.
[23] H. Fan, G.F. Wang, Interaction between a screw dislocation and viscoelastic interfaces, Int. J. Solids Struct. 40 (2003) 763–776.
[24] W. Yan, W.Q. Chen, Time-dependent response of laminated isotropic strips with viscoelastic interfaces, J. Zhejiang Univ.
(SCIENCE) 5 (2004) 1318–1321.
[25] T.O. Williams, F.L. Addessio, A general theory for laminated plates with delaminations, Int. J. Solids Struct. 34 (1997) 2003–2024.
[26] R.M. Christensen, Theory of Viscoelasticity. An Introduction, second ed., Academic Press, New York, 1982.
[27] J.V. Funn, I. Dutta, Creep behavior of interfaces in fiber reinforced metal–matrix composites, Acta Mater. 47 (1999) 149–164.

You might also like