Sunteți pe pagina 1din 22

COMMUNICATIONS IN NUMERICAL METHODS IN ENGINEERING

Commun. Numer. Meth. Engng 2007; 23:733–754


Published online 19 September 2006 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/cnm.923

Analysis of thick functionally graded plates by local


integral equation method

J. Sladek1, ∗, † , V. Sladek1 , Ch. Hellmich2 and J. Eberhardsteiner2


1 Institute of Construction and Architecture, Slovak Academy of Sciences, 845 03 Bratislava, Slovakia
2 Institute for Mechanics of Materials and Structures, Vienna University of Technology,
Karlsplatz 13/202, 1040 Wien, Austria

SUMMARY
Analysis of functionally graded plates under static and dynamic loads is presented by the meshless local
Petrov–Galerkin (MLPG) method. Plate bending problem is described by Reissner–Mindlin theory. Both
isotropic and orthotropic material properties are considered in the analysis. A weak formulation for the
set of governing equations in the Reissner–Mindlin theory with a unit test function is transformed into
local integral equations considered on local subdomains in the mean surface of the plate. Nodal points
are randomly spread on this surface and each node is surrounded by a circular subdomain, rendering
integrals which can be simply evaluated. The meshless approximation based on the moving least-squares
(MLS) method is employed in the numerical implementation. Numerical results for simply supported and
clamped plates are presented. Copyright q 2006 John Wiley & Sons, Ltd.

Received 4 May 2006; Revised 24 July 2006; Accepted 9 August 2006

KEY WORDS: functionally graded material; plate bending problem; orthotropic material; local boundary
integral equations; static and impact load; Laplace-transform; meshless approximation

1. INTRODUCTION

Functionally graded materials are multi-phase materials with the phase volume fractions varying
gradually in space, in a pre-determined profile. This results in continuously graded thermal and
mechanical properties at the (macroscopic) structural scale. Often, these spatial gradients in material
behaviour render FGMs as superior to conventional composites. FGMs posses some advantages

∗ Correspondence to: J. Sladek, Institute of Construction and Architecture, Slovak Academy of Sciences, 845 03
Bratislava, Slovakia.

E-mail: sladek@savba.sk

Contract/grant sponsor: Slovak Research and Development Support Agency; contract/grant number: APVV-51-021205
Contract/grant sponsor: Slovak Grant Agency; contract/grant number: VEGA-2/6109/6
Contract/grant sponsor: EU Network of Excellence; contract/grant number: NMP3-CT-2004-502243

Copyright q 2006 John Wiley & Sons, Ltd.


734 J. SLADEK ET AL.

over conventional composites because of their continuously graded structures and properties [1, 2].
FGMs may exhibit isotropic or anisotropic material properties, depending on the processing tech-
nique and the practical engineering requirements. Due to the high mathematical complexity of
the initial-boundary value problems, analytical approaches for the FGMs are restricted to simple
geometry and boundary conditions. Thus, analyses in FGM demand accurate and efficient numer-
ical methods. The finite element method can be successfully applied to problems with an arbitrary
variation of material properties by using special graded elements [3]. In commercial computer
codes, however, material properties are considered to be uniform on each element. The boundary
element method (BEM) is a suitable numerical tool for this purpose, too. However, for a general
continuously non-homogeneous body the fundamental solution for many governing equations are
not yet known in literature. One possibility to obtain a BEM formulation is based on the use of
fundamental solutions for a fictitious homogeneous medium. Such an approach, which is the basis
of the global BEM, however, leads to a boundary-domain integral formulation. The price which
is to be paid in that approach is the loss of a pure boundary integral character of the formulation.
Recently, meshless methods are becoming popular, and they have been successfully applied to
2-D and 3-D axisymmetric transient heat conduction analyses for isotropic and anisotropic FGMs
[4–8] and in elasticity [9–11].
During the last several decades, laminated composite plates have been widely used in engineering
structures. Previous research results show that the transverse shear effects are more significant for
orthotropic plates than for isotropic ones [12]. The analysis of thin elastic plates by the BEM is
a research subject since many years. Much previous research works have been done for static and
dynamic analysis of isotropic thin plates by the BEM [13, 14]. It is well known that the classical
thin plate theory of Kirchhoff gives rise to certain non-physical simplifications mainly related to
the omission of the shear deformation and the rotary inertia, which become more significant for
increasing thickness of the plate. The effects of shear deformation and rotary inertia are taken
into account in the Reissner–Mindlin plate bending theory [15, 16]. The first application of the
boundary integral equation method to Reissner’s plate model was given by Van der Ween [17].
Dynamic analysis of elastic Reissner–Mindlin plates has been performed by the direct BEM in the
frequency domain [18, 19]. All previous BEM applications are dealing with isotropic Reissner–
Mindlin plates. Wang and Huang [12] were the first who applied BEM to orthotropic thick plates.
Number of applications to analyse functionally graded plates is very limited. Reddy [20] applied
the FEM to FGM plates using the third-order shear deformation plate bending theory. A higher-
order theory for plates deformed in shear and normal mode was used by Qian et al. [21]: There,
the plate material is made of two isotropic constituents and exhibits macroscopically isotropic
material properties which vary in the thickness direction only. Results of the static analysis and
computed natural frequencies of a simply supported square plate match well with corresponding
analytical values.
In the present paper, the authors have extended a meshless method based on the local Petrov–
Galerkin weak form to dynamic problems for orthotropic thick plates with material proper-
ties continuously varying through the plate thickness. The Reissner–Mindlin theory reduces
the original 3-D thick plate problem to a 2-D problem. Meshless approaches for problems of
continuum mechanics have attracted much attention during the past decade [22–26]. Focusing
only on nodes or points instead of elements as in the conventional FEM or BEM, meshless
approaches have certain advantages. In our meshless method, nodal points are randomly dis-
tributed over the mean surface of the considered plate. Each node is the centre of a circle
surrounding this node. A similar approach has been successfully applied to a thin Kirchhoff

Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:733–754
DOI: 10.1002/cnm
ANALYSIS OF THICK FUNCTIONALLY GRADED PLATES 735

plate [27–29]. In this paper, the Laplace-transform technique is applied to the set of governing
partial differential equations for the elastodynamic bending of Reissner–Mindlin plates. Then,
the unknown Laplace-transforms of the relevant physical quantities on the local boundaries are
constrained by the local boundary integral equations. A unit test function is used in the local
weak form. By applying Gauss divergence theorem to the weak form of the partial differen-
tial equations, the local boundary-domain integral equations are derived. Meshless approxima-
tion of the generalized displacements on a simple domain, based on the moving least-squares
(MLS) method, allows for elegant and efficient numerical integration of the domain-integral.
Several quasi-static boundary value problems have to be solved for various values of the Laplace-
transform parameter. The Stehfest’s inversion method [30] is applied to obtain the time-dependent
solution.
Numerical results for isotropic and orthotropic square plates with various boundary con-
ditions and subjected to static and impulse loadings are presented to illustrate the applica-
bility of the proposed method. Linear and quadratic variations of material properties along
a Cartesian coordinate perpendicular to the plate plane are considered in the numerical
analyses.

2. LOCAL INTEGRAL EQUATIONS FOR AN ORTHOTROPIC THICK FGM PLATE

Consider an elastic orthotropic plate of constant thickness h, with the mean surface occupying
the domain  in the plane (x1 , x2 ). The plate is subjected to a transverse dynamic load q(x, t).
The Reissner–Mindlin plate bending theory [15, 16] is used to describe the plate deformation. The
transverse shear strains are represented as constant throughout the plate thickness and some cor-
rection coefficients are required for computation of transverse shear forces in that theory. Then,
the spatial displacement field, expressed in terms of displacement components u 1 , u 2 , and u 3 , has
the following form [31]:

u 1 (x, t) = x3 w1 (x, t)

u 2 (x, t) = x3 w2 (x, t) (1)

u 3 (x, t) = w3 (x, t)

where w and w3 represent the rotations around the x -direction and the out-of-plane deflection,
respectively (Figure 1). The linear strains are given by

11 (x, t) = x3 w1,1 (x, t)

22 (x, t) = x3 w2,2 (x, t)

12 (x, t) = x3 (w1,2 (x, t) + w2,1 (x, t))/2 (2)

13 (x, t) = (w1 (x, t) + w3,1 (x, t))/2

23 (x, t) = (w2 (x, t) + w3,2 (x, t))/2

Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:733–754
DOI: 10.1002/cnm
736 J. SLADEK ET AL.

q(x,t)

x3 x2
Q1
h M11
x1
M22
Pt M12
M21 Q2
Pb

Figure 1. Sign convention of bending moments and forces for the FGM plate.

For a plane stress problem in orthotropic materials one can write


⎡ ⎤ ⎡ ⎤
11 11
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
⎢ 22 ⎥ ⎢ 22 ⎥
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
⎢ 12 ⎥ = D(x) ⎢ 212 ⎥ (3)
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
⎢ 13 ⎥ ⎢ 213 ⎥
⎣ ⎦ ⎣ ⎦
23 223

where
⎡ ⎤
E 1 /e E 2 12 /e 0 0 0
⎢ ⎥
⎢ E  /e E 2 /e 0 ⎥
⎢ 2 12 0 0 ⎥
⎢ ⎥
⎢ ⎥
D(x) = ⎢ 0 0 G 12 0 0 ⎥ with e = 1 − 12 21
⎢ ⎥
⎢ ⎥
⎢ 0 0 0 G 13 0 ⎥
⎣ ⎦
0 0 0 0 G 23

E k are the Young’s moduli refering to the axes x k , k = 1, 2, G 12 , G 13 and G 23 are shear moduli,
i j are Poisson’s ratios.
Next, we assume that the material properties are graded along the plate thickness, and we
represent the profile for volume fraction variation by
 
x3 1 n
P(x3 ) = Pb + (Pt − Pb )V with V = + (4)
h 2
where P denotes a generic property like modulus, Pt and Pb denote the property of the top and
bottom faces of the plate, respectively, and n is a parameter that dictates the material variation
profile (Figure 2). Poisson ratios are assumed to be uniform.

Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:733–754
DOI: 10.1002/cnm
ANALYSIS OF THICK FUNCTIONALLY GRADED PLATES 737

n =1
0.8
n =2

0.6
Vf

0.4

0.2

0
-0.5 -0.3 -0.1 0.1 0.3 0.5
z/h

Figure 2. Variation of volume fraction over the plate thickness, for linear and quadratic power-law index.

The bending moments M and the shear forces Q  are defined as
⎡ ⎤ ⎡ ⎤
M11 11


13
⎢ ⎥ h/2 ⎢ ⎥ Q1 h/2
⎢ M22 ⎥ = ⎢ 22 ⎥ x3 dx3 and = dx3 (5)
⎣ ⎦ ⎣ ⎦
−h/2 Q2 −h/2 23
M12 12

where  = 5/6 in the Reissner plate theory.


Substituting Equations (3) and (2) into moment and force resultants (5) allows for expression
of the bending moments M and of the shear forces Q  , ,  = 1, 2, are expressed in terms of
rotations and lateral displacements of the orthotropic plate. For uniform material properties (n = 0)
one obtains
M = D (w, + w, ) + C w,
(6)
Q  = C (w + w3, )

In Equation (6), repeated indices ,  do not imply summation, and the material parameters D
and C are given as

D1 D2 G 12t h 3
D11 = (1 − 21 ), D22 = (1 − 12 ), D12 = D21 =
2 2 12
C11 = D1 21 , C22 = D2 12 , C12 = C21 = 0
(7)
E 1t h 3 E 2t h 3
D1 = , D2 = , D1 21 = D2 12
12(1 − 12 21 ) 12(1 − 12 21 )
C1 = G 13t h, C2 = G 23t h

Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:733–754
DOI: 10.1002/cnm
738 J. SLADEK ET AL.

For linear gradation of material properties (n = 1), the previous expressions still hold, with the
exception of

(E t + E b )h 3 (G 12t + G 12b )h 3
D = , D12 = D21 =
24(1 − 12 21 ) 24
(8)
(G 3t + G 3b )
C = h
2
Finally, for a quadratic gradation:

E b h 3 (E t − E b )h 3 G 12b h 3 (G 12t − G 12b )h 3


D = + , D12 = D21 = +
12(1 − 12 21 ) 30(1 − 12 21 ) 12 30
(9)
(G 3t − G 3b )
C = G 3b h + h
3
In the Reissner–Mindlin plate bending theory the governing equations (equations of motion) have
the form [15, 16]
h 3
M, (x, t) − Q  (x, t) = ẅ (x, t)
12 (10)
Q , (x, t) + q(x, t) = h ẅ3 (x, t), x∈
where  is the mass density. The Greek indices vary from 1 to 2. The dots indicate differentiations
with respect to time t.
Applying the Laplace-transform

L[y(x, t)] = y(x, s) = y(x, t)e−st d
0

to the a general quantity y(x, t) occurring in system of governing equations (10), one obtains

h 3 2
M , (x, s) − Q  (x, s) = s w (x, s) − R  (x, s) (11)
12
Q , (x, s) = hs 2 w3 (x, s) − R 3 (x, s) (12)

where s is the Laplace-transform parameter, while R  and R 3 are given by

h 3
R  (x, s) = [sw (x) + ẇ (x)]
12 (13)
R 3 (x, s) = q(x, s) + hsw3 (x) + h ẇ3 (x)
with w (x), w3 (x), ẇ (x) and ẇ3 (x) being the initial values and the initial velocities of the
generalized displacement field.
According to the meshless local Petrov–Galerkin (MLPG) method, we construct a weak form
of (11) and (12) over the local subdomains is around each node xi -inside the global domain 
(see Figure 3) [32]. The subdomains may overlap each other, while covering the whole global

Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:733–754
DOI: 10.1002/cnm
ANALYSIS OF THICK FUNCTIONALLY GRADED PLATES 739

Figure 3. Local boundaries for weak formulation, the domain x for MLS approximation of the trial
function, and support area of weight function around node xi .

domain  (see Figure 3). The local subdomains could be of any geometrical shape and size.
For the sake of simplicity, the local subdomains are taken here to be of a circular shape. For this
case, the evaluation of domain-integrals is quite easy. In mathematical terms, the local weak form
of the governing equations (11) and (12) for xi ∈ is can be written as

h 3 2
M , (x, s) − Q  (x, s) − s w (x, s) + R  (x, s) u ∗ (x) d = 0 (14)
is 12
 
Q , (x, s) − hs 2 w 3 (x, s) + R 3 (x, s) u ∗ (x) d = 0 (15)
is

where u ∗ (x) and u ∗ (x) are weight or test functions.


Applying the Gauss divergence theorem to Equations (14) and (15), one can write

M  (x, s)u ∗ (x) d − M  (x, s)u ∗, (x) d − Q  (x, s)u ∗ (x) d
*is is is

h 3 2
− s w (x, s)u ∗ (x) d + R  (x, s)u ∗ (x) d = 0 (16)
is 12 is



Q  (x, s)n  (x)u (x) d − Q  (x, s)u ∗, (x) d
*is is


− hs w3 (x, s)u (x) d +
2
R 3 (x, s)u ∗ (x) d = 0 (17)
is is

Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:733–754
DOI: 10.1002/cnm
740 J. SLADEK ET AL.

where *is is the boundary of the local subdomain and

M  (x, s) = M  (x, s)n  (x)


is the Laplace-transform of the normal bending moment and n  is a unit normal vector to the
boundary *is . The local weak forms (16) and (17) are the starting point for deriving local boundary
integral equations on the basis of appropriate test functions. Unit step functions are chosen for the
test functions u ∗ (x) and u ∗ (x) in each subdomain
 

 at x ∈ (s ∪ *s ) 1 at x ∈ (s ∪ *s )
u ∗ (x) = , ∗
u (x) = (18)
0 at x ∈
/ s 0 at x ∈
/ s
Then, the local weak forms (16) and (17) are transformed into the following local boundary integral
equations:

h 3 2
M  (x, s) d− Q  (x, s) d − s w (x, s) d + R  (x, s) d = 0 (19)
*is is is 12 is

Q  (x, s)n  (x) d− hs w 3 (x, s) d+
2
R 3 (x, s) d = 0 (20)
*is is is

In the above local integral equations, the trial functions w  (x, s), related to rotations, and w 3 (x, s),
related to displacements, are chosen as the MLS approximation over a number of nodes randomly
spread within the domain of influence.

3. NUMERICAL SOLUTION

The MLS approximation [33, 34] is used in the present analysis for the approximation of the
physical quantities throughout the plate domain. Let us consider a subdomain x of the problem
domain , in the neighbourhood of a point x, for the definition of the MLS approximation of the
trial function around x (Figure 3). To approximate the distribution of the Laplace-transform of
the generalized displacements (rotations and deflection) in x over a number of randomly located
nodes {xa }, a = 1, 2, . . . , n, the MLS approximant wih (x, s) of wi , i = 1, 2, 3, is defined by

wh (x, s) = pT (x)ã(x, s) ∀x ∈ x (21)


In (21), wh = [w1h , w2h , w3h ]T collects Laplace-transformed rotations and displacements, pT (x) =
[ p 1 (x), p 2 (x), . . . , p m (x)] is a complete monomial basis of order m, and ã(x, s) = [a1 (x, s),
j j j
a2 (x, s), . . . , am (x, s)]T is composed of the vectors a j (x, s) = [a1 (x, s), a2 (x, s), a3 (x, s)]T which
are functions of the space coordinates x = [x 1 , x2 ]T and the Laplace-transform parameter s. For
example, for a 2-D problem

pT (x) = [1, x1 , x2 ] for linear basis m = 3 (22a)

pT (x) = [1, x1 , x2 , (x1 )2 , x1 x2 , (x2 )2 ] for quadratic basis m = 6 (22b)

Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:733–754
DOI: 10.1002/cnm
ANALYSIS OF THICK FUNCTIONALLY GRADED PLATES 741

Usually quadratic monomials are sufficient, and they have been applied also in the numerical
computations presented in this paper.
The coefficient vector ã(x, s) is determined by minimizing a weighted discrete L 2 -norm defined
as

na
J (x) = v a (x)[pT (xa )ã(x, s) − ŵa (s)]2 (23)
a=1

where v a (x)>0 is the weight function associated with the node a and the square power is understood
in the sense of scalar product. In (23), n a is the number of nodes in x for which the weight
function v a (x)>0, and ŵa (s) are the fictitious nodal values at node a, but in general not the values
of the unknown trial function at the nodes, wh (xi , s). The stationarity of J in Equation (23) with
respect to a(x, s)
*J/*a = 0
leads to the following linear relation between the coefficient vector ã(x, s) and the vector collecting
the fictitious nodal values of all nodes, ŵ(s)
A(x)ã(x, s) − B(x)ŵ(s) = 0 (24)
where
⎡ ⎤
ŵ1 (s)
⎢ ⎥
⎢ 2 ⎥
⎢ ŵ (s) ⎥
ŵ(s) = ⎢



⎢ ··· ⎥
⎣ ⎦
ŵn (s) (25)


n
A(x) = v a (x)p(xa )pT (xa )
a=1

B(x) = [v 1 (x)p(x1 ), v 2 (x)p(x2 ), . . . , v n (x)p(xn )]


The MLS approximation is well defined only when the matrix A in Equation (24) is non-singular.
This requires at least m weight functions to be non-zero (i.e. n a m) for each sample point x ∈ ,
and the nodes in x to be not arranged in a special pattern, such as on a straight line.
Substituting the solution of Equation (24) for ã(x, s) into Equation (21) leads to the following
relation:

na
wh (x, s) = UT (x) · ŵ(s) = a (x)ŵa (s) (26)
a=1

where
UT (x) = pT (x)A−1 (x)B(x) (27)
In Equation (26), a (x) is usually referred to as the shape function of the MLS approximation
corresponding to the nodal point xa . From Equations (25) and (27), it can be seen that a (x) = 0

Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:733–754
DOI: 10.1002/cnm
742 J. SLADEK ET AL.

when v a (x) = 0. In practical applications, v a (x) is often chosen such that it is non-zero over the
support of the nodal point xa . The support of the nodal point xa is usually taken to be a circle of
the radius r a centred at xa (see Figure 3). The radius r a is an important parameter of the MLS
approximation because it determines the range of the interaction (coupling) between the degrees
of freedom defined at the considered nodes.
A fourth-order spline-type weight function is applied in the present work
⎧  a 2  a 3  a 4

⎨1 − 6 d d d
a
+ 8 a
− 3 , 0d a r a
v (x) =
a r r ra (28)


0, d ra a

where d a = x − xa  and r a is the radius of the circular support domain. With Equation (28), the
C 1 -continuity of the weight function is ensured over the entire domain, therefore the continuity
condition of the bending moments and the shear forces is satisfied. The size of the support r a
should so large that the number of nodes covered in the domain of definition is sufficient for
ensuring the regularity of the matrix A. In fact, the value of n a is determined by this number of
nodes lying in the support domain with radius r a .
The directional derivatives of w(x, s) are approximated in terms of the same nodal values as
w(x, s) itself


n
w,k (x, s) = ŵa (s) a,k (x) (29)
a=1

whereby the partial derivatives of the MLS shape functions are obtained as [32]


m
[ p,k (A−1 B) ja + p j (A−1 B,k + A−1
j
a,k = ,k B) ]
ja
(30)
j=1

with A−1 −1
,k = (A ),k representing the derivative of the inverse of A with respect to x k , which is
given by

A−1 −1
,k = −A A,k A
−1

Substituting the approximation (29) into the definition of the normal bending, (5) and (6) one
obtains for M(x, s) = [M 1 (x, s), M 2 (x, s)]T


n 
n 
n
M(x, s) = N1 Ba1 (x)w∗a (s) + N2 Ba2 (x)w∗a (s) = N (x) Ba (x)w∗a (s) (31)
a=1 a=1 a=1

where the vector w∗a (s) is defined as a column vector w∗a (s) = [ŵ1a (s), ŵ2a (s)]T , the matrices
N (x) are related to the normal vector n(x) on *s by




n1 0 n2 C11 0 n1 n1
N1 (x) = and N2 (x) =
0 n2 n1 0 C22 n2 n2

Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:733–754
DOI: 10.1002/cnm
ANALYSIS OF THICK FUNCTIONALLY GRADED PLATES 743

and the matrices Ba are represented by the gradients of the shape functions as
⎡ ⎤
2D11 a,1 0
a
⎢ ⎥ ,1 0
Ba1 (x) = ⎢
⎣ 0 2D22 ,2 ⎥
a
⎦ , B2 (x) =
a
0 a,2
a a
D12 ,2 D12 ,1
Dependence on the type of gradation is expressed through C and D defined in
Equations (7)–(9).
Similarly one can obtain the approximation for the shear forces

n
Q(x, s) = C(x) [ a (x)w∗a (s) + Fa (x)ŵ3a (s)] (32)
a=1

where Q(x, s) = [Q 1 (x, s), Q 2 (x, s)]T and





C1 (x) 0 a,1
C(x) = , Fa (x) =
0 C2 (x) a,2
Then, insertion of the MLS-discretized force fields (31) and (32) into the local boundary integral
equations (19) and (20) yields the discretized local integral equations (LIEs)


n h 3 (x) 
N (x)Ba (x) d − C(x) + E s 2 a (x) d w∗a (s)
a=1 L is +isw is 12


n
˜
− ŵ3a (s) C(x)Fa (x) d = − M(x, s) d − R(x, s) d (33)
a=1 is is M is

 

n 
n
Cn (x) (x) d w∗a (s) +
a
ŵ3a (s)
a=1 *is a=1
 
a
× Cn (x)F (x) d −
a
(s h(x) (x) d
2
*is is

=− R 3 (x, s) d (34)
is

in which
   
1 0 C1 0
E= , Cn (x) =(n 1 , n 2 ) = (C1 n 1 , C2 n 2 )
0 1 0 C2

Equations (33) and (34) are considered on the subdomains adjacent to the interior nodes xi as well
as to the boundary nodes on is M .

Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:733–754
DOI: 10.1002/cnm
744 J. SLADEK ET AL.

For the source point xi located on the global boundary  the boundary of the subdomain *is
is decomposed into L is and is M (part of the global boundary with prescribed bending moment)
according to Figure 3.
It should be noted here that there are neither Lagrange multipliers nor penalty parameters
introduced into the local weak forms (7) and (8) because the essential boundary conditions on isw
(part of the global boundary with prescribed rotations or displacements) can be imposed directly,
using the interpolation approximation (26)

n
a (x)ŵa (s) = w̃(xi , s) for xi ∈ isw (35)
a=1

where w̃(xi , s) is the Laplace-transform of the generalized displacement vector prescribed on


the boundary isw . For a clamped plate all three vector components (rotations and deflection)
are vanishing at the fixed edge, and Equation (35) is used at all the boundary nodes in such a
case. However, for a simply supported plate only the third component of the displacement vector
(deflection) is prescribed, while the rotations are unknown. Then, the entire Equation (33) and
the third component of Equation (35) are applied to the nodes lying on the global boundary. On
parts of the global boundary where no displacement boundary conditions are prescribed both local
integral equations (33) and (34), are applied.
The time-dependent values of rotations, displacements, moments, and shear forces are obtained
from an inverse transform of the corresponding Laplace-transformed quantities. Thereby, great
attention is paid to the numerical inversion of the Laplace transformation, since the inverse Laplace-
transform is an ill-posed problem and small truncation errors can be greatly magnified in the
inversion process with yielding poor numerical results. In the present analysis, the sophisticated
Stehfest’s algorithm [30] is used for the numerical inversion.

4. NUMERICAL EXAMPLES

In this section, numerical results are presented for plates under static loads and under impact
loads with the Heaviside time dependence. In order to test the accuracy, the numerical results
obtained by the present method are compared with the results provided by the FEM-NASTRAN
code using a very fine mesh. Clamped and simply supported square plates are analysed. In all
numerical calculations, FGM plates with isotropic and orthotropic material properties are subjected
to a uniformly distributed transverse load.

4.1. Static load


Consider a clamped square plate with a side-length a = 0.254 m and plate thicknesses h/a = 0.05.
The plate is subjected to a uniformly distributed static load. Firstly, isotropic material properties
of the FGM plate are considered. The following material parameters on top side of the plate are
used in numerical analysis: Young’s moduli E 2t = 0.6895 × 1010 N/m2 , E 1t = E 2t , Poisson’s ratios
21 = 0.3, 12 = 0.3. Linear and quadratic variations of volume fraction V defined in Equation (4)
are considered here, and Young’s moduli on the bottom side are: E 1b = E 1t /2 and E 1b = E 2b .
In our numerical calculations, 441 nodes with a regular distribution were used for the approxi-
mation of the rotations and the deflection (Figure 4). If s is the distance of two neighbouring nodes,
then the radius of the circular subdomain is chosen as rloc = 0.4 s and the radius of the support

Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:733–754
DOI: 10.1002/cnm
ANALYSIS OF THICK FUNCTIONALLY GRADED PLATES 745

x2,E2
110 wi=0 121

wi=0 M11

wi=0

x1,E1
1 11
wi=0
M22

Figure 4. Node distribution for numerical analyses of a clamped square plate.

1.6

1.4
w3(x 2=a/2) / w3ho m(x 1=x2=a/2)

1.2

0.8

0.6
FEM: n = 0
0.4 MLPG: n = 0
n=1
0.2
n=2
0
0 0.1 0.2 0.3 0.4 0.5
x1 / a

Figure 5. Isotropic clamped square plates under a uniform static load: variation of the deflection with
the x1 -coordinate at x2 = a/2 = const.

domain for node a is r a = 4rloc . Smaller values of the support domain lead to lower approximation
accuracy, and larger values of the support domain prolong the computational time for the evaluation
of the shape functions given by Equations (26) and (29). More detailed analyses on the influence
of the support domain on computational accuracy can be found in Reference [32]. The variation of
the deflection with the x1 -coordinate at x2 = a/2 of homogeneous and FGM plates are presented in
Figure 5. The deflection value is normalized by the corresponding central deflection of an isotropic
plate with homogeneous material properties identical to the ones on the top of the FGM plate. For a
uniformly distributed load q0 = 300 psi (2.07 × 106 Nm−2 ) we have w3hom (a/2) = 8.842 × 10−3 m.
Since Young’s modulus on bottom side is considered to be smaller than on the top one, deflec-
tion for FGM plate is larger than for homogenenous plate with material properties corresponding
to the top side, E 2t = 0.6895 × 1010 N/m2 . The numerical results are compared with the results
obtained by the FEM-NASTRAN code with a very fine mesh of 400 quadrilateral eight-node

Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:733–754
DOI: 10.1002/cnm
746 J. SLADEK ET AL.

1.6

1.4

(x 1=x2=a/2)
1.2

ho m
1

0.8
w3(x 2=a/2) / w3

0.6 FEM: n= 1
MLPG: n= 1
0.4
FEM: n=2
0.2 MLPG: n =2

0
0 0.1 0.2 0.3 0.4 0.5
x1 / a

Figure 6. Comparison of FEM and MLPG deflections of isotropic clamped square FGM plates.

1.5

1
M11(x2=a/2)/M11hom(x1=x2=a/2)

0.5

-0.5

-1

-1.5

-2
FEM: n=0
MLPG: n=0
-2.5
n=2
-3
0 0.1 0.2 0.3 0.4 0.5
X1/a

Figure 7. Isotropic clamped square plates under a uniform static load: variation of the bending moment
with the x1 -coordinate at x2 = a/2 = const.

shell elements for a quarter of the plate. Our numerical results are in a very good agreement
with those obtained by the FEM for the FGM plate (Figure 6). The variation of the bending
moment M11 is presented in Figure 7. Here, the bending moments are normalized by the central
bending moment value corresponding to a homogeneous isotropic plate M11 hom (a/2) = 3064 Nm.

The variation of material properties through the plate thickness has practically no influence on
the bending moment. Bending moments in homogeneous and FGM plates are almost the same.
Minimal differences between them are caused by numerical inaccuracies.

Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:733–754
DOI: 10.1002/cnm
ANALYSIS OF THICK FUNCTIONALLY GRADED PLATES 747

1.6

1.4 n = 0: E1=E2

w3(x 2=a/2)/w3ho m(x 1=x2=a/2)


E1=2*E2
1.2 n = 2: E1=E2
E1=2*E2
1

0.8

0.6

0.4

0.2

0
0 0.1 0.2 0.3 0.4 0.5
X1/a

Figure 8. Orthotropic clamped square plates under a uniform static load: variation of the deflection with
the x1 -coordinate at x2 = a/2 = const.

Next, orthotropic FGM clamped plate is analysed. The following material parameters are
used in numerical analysis: Young’s moduli E 2t = 0.6895 × 1010 N/m2 , E 1t = 2E 2t , Poisson’s
ratios 21 = 0.15, 12 = 0.3. The used shear moduli correspond to Young’s modulus E 2 , namely,
G 12 = G 13 = G 23 = E 2 /2(1 + 12 ). A quadratic variation of the volume fraction V is considered
with Young’s moduli on the bottom side are: E 1b = E 1t /2 and E 1b = 2E 2b . Variation of the deflec-
tion with the x1 -coordinate at x2 = a/2 of homogeneous (dashed line) and FGM (solid line) plates
are presented in Figure 8. The deflection value is normalized by the corresponding deflection at
the centre of an isotropic homogeneous plate with material constants given above and E 1t = E 2t .
One can observe that the deflection value is reduced for an orthotropic plate if one of the Young’s
moduli is increased. However, the deflection value of FGM plate is larger than that of the homoge-
neous plate with material properties identical to those on the top of the FGM plate, characterized
by larger Young’s moduli. Variation of the bending moment along (x 1 , a/2) for an orthotropic
FGM plate is presented in Figure 9. In an orthotropic plate, the absolute values of the bending
moment at the plate centre and at the centre of the clamped side are higher than those occurring
in an isotropic plate, for both homogeneous and FGM plates.
Let us consider a simply supported square plate with the same geometrical and material para-
meters as in the above analysed clamped plate. Also the same nodal distribution is used
in the numerical analysis. The plate is subjected to a uniform static loading. The variations of
the deflection with the x1 -coordinate at x2 = a/2 of the plate with a uniform, linear and quadratic
volume fraction variation are presented in Figure 10. The deflection value is normalized
by the central deflection of a homogeneous isotropic plate (E 1t = E 2t ) under a static load. For a
uniformly distributed load q0 = 2.07 × 106 Nm−2 we have obtained w3hom (a/2) = 28.29 × 10−3 m.
The structural effects of the orthotropic material properties are similar to those encountered pre-
viously with clamped plates.

Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:733–754
DOI: 10.1002/cnm
748 J. SLADEK ET AL.

2
1.5

(x 1=x2=a/2)
1
0.5
0
ho m
M11(x 2=a/2)/M11 -0.5
-1
-1.5
n=0: E1=E2
-2
E1=2*E2
-2.5
n=2: E1=E2
-3
E1=2*E2
-3.5
0 0.1 0.2 0.3 0.4 0.5
X1/a

Figure 9. Orthotropic clamped square plates under a uniform static load: variation of the bending moment
with the x1 -coordinate at x2 = a/2 = const.

1.6

1.4
w3(x 2=a/2) / w3ho m(x 1=x2=a/2)

1.2

0.8

0.6 FEM: n= 0
MLPG: n= 0
0.4
n=1
0.2 n=2

0
0 0.1 0.2 0.3 0.4 0.5
x1 / a

Figure 10. Isotropic simply supported square plates under a uniform static load: variation of the deflection
with the x1 -coordinate at x2 = a/2 = const.

4.2. Impact load


A simply supported isotropic thick square plate under an impact load with Heaviside time vari-
ation is analysed. The used geometrical and material parameters are the same as in the static
case. For the numerical modelling we have used again 441 nodes with a regular distribution.
The time variation of the central deflection at x 1 = x2 = a/2 for both homogeneous and FGM
plates are presented in Figure 11. The deflection value is normalized by the central deflection of

Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:733–754
DOI: 10.1002/cnm
ANALYSIS OF THICK FUNCTIONALLY GRADED PLATES 749

3.5

3 n = 0: FEM
n = 0: MLPG

w3/w3hom(x 1=x2=a/ 2)
2.5
n = 2: MLPG
2

1.5

0.5

-0.5
0 0.5 1 1.5 2
t / t0

Figure 11. Time variation of the deflection at the centre of isotropic simply supported square plates
subjected to a suddenly applied load.

a homogeneous plate under a static load. This value has been√ given in Section 4.1. The time vari-
able is normalized by a characteristic time [35] t0 = a 2 /4 h/D = 1.35 × 10−2 s corresponding
to homogeneous plate, too. The MLPG results for a homogeneous plate are compared with those
obtained by FEM-NASTRAN computer code, where 400 quadrilateral eight-node shell elements
with 1000 time increments have been used. A good agreement between both the numerical results
is achieved. A quadratic variation of the volume fraction V is considered for FGM plate with
Young’s moduli on the bottom side as: E 1b = E 1t /2 and E 1b = E 2b . The oscillation of the plate
deflection for FGM plate is moderated relatively to a homogeneous plate with higher flexural
rigidity. The absolute value of the central deflection of FGM plate is larger than that of the ho-
mogeneous plate. The time variation of the central bending moment M11 is given in Figure 12.
The bending moment at the centre of the plate is normalized by the quantity corresponding to a
homogeneous plate under a static load, M11 hom (a/2) = 6480 Nm. In this case, the amplitudes of the

bending moments for both FGM and homogeneous plates are almost the same. However, in the
FGM plate with smaller flexural rigidity, the peaks of the moment amplitudes occur at later time
instants.
In the next example, let us consider a clamped isotropic square plate with the same geometrical
and material parameters as in the above analysed simply supported plate. The time variation of the
central deflection for both the homogeneous plate and the FGM plate are presented in Figure 13.
The deflection value is normalized by the central deflection of an isotropic plate with homogeneous
material properties identical to those prevailing on the top of the FGM plate, under a static load.
The time variations of the bending moments M11 for homogeneous and FGM plates are given
in Figure 14. The structural effects of functional gradation are similar those previously discussed
for simply supported plates. The central deflection is larger and oscillations slower for the FGM
plate than for a homogeneous plate with higher flexural rigidity. The amplitudes of the bending
moments for both plates are almost the same. A good agreement of the present results for the
deflections and the bending moments at the central point and the FEM results is observed in both
figures for a homogeneous plate.

Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:733–754
DOI: 10.1002/cnm
750 J. SLADEK ET AL.

3.5

3 n = 0: FEM
n = 0: MLPG

M11/M11hom(x 1=x2=a/2)
2.5
n = 2: MLPG
2

1.5

0.5

-0.5
0 0.5 1 1.5 2
t / t0

Figure 12. Time variation of the bending moment at the centre of isotropic simply supported square plates
subjected to a suddenly applied load.

3.5
n = 0: FEM
3
n = 0: MLPG
w3 /w3hom(x1 =x2=a/ 2)

2.5 n = 2: MLPG

1.5

0.5

-0.5
0 0.2 0.4 0.6 0.8 1 1.2
t/t0

Figure 13. Time variation of the deflection at the centre of isotropic clamped square plates
subjected to a suddenly applied load.

Next, the orthotropic FGM clamped plate under a uniform impact load with a Heaviside time
dependence is analysed. The material properties and geometry parameters are the same as in the
static case analysed above. Again a quadratic variation of the volume fraction V is considered for
the FGM plate. The time variation of the deflection at x 1 = x2 = a/2 for both the isotropic and
orthotropic FGM plates are presented in Figure 15. The deflections are normalized by the static
equivalent for the isotropic and homogeneous plate, w3hom (a/2) = 8.842 × 10−3 m. The amplitudes

Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:733–754
DOI: 10.1002/cnm
ANALYSIS OF THICK FUNCTIONALLY GRADED PLATES 751

3.5
n = 0: FEM
3
n = 0: MLPG

M11/M11hom(x 1=x2=a/2)
2.5
n = 2: MLPG
2

1.5

0.5

-0.5

-1
0 0.2 0.4 0.6 0.8 1 1.2
t/t0

Figure 14. Time variation of the bending moment at the centre of isotropic clamped square plates
subjected to a suddenly applied load.

3
isotropic FGM
2.5
orthotropic FGM
(x1=x2=a/2)

1.5
hom

1
w3 /w3

0.5

-0.5
0 0.2 0.4 0.6 0.8 1 1.2
t/t0

Figure 15. Influence of orthotropic material properties on time variation of the deflection for a clamped
square plates under a uniform impact load.

of the deflections for the orthotropic plate are reduced with respect to the isotropic one, due to
higher Young’s modulus E 1 for the orthotropic plate. However, the oscillations are more rapid for
the orthotropic FGM plate. The comparison of the time variations of bending moments for the
isotropic and orthotropic plates is given in Figure 16. Here, the bending moments are normalized
by the bending moment at the centre of the homogeneous isotropic plate under a static load,
M11hom (a/2) = 3064 Nm. One can observe that the amplitudes of the bending moments in the

orthotropic plate are larger than those in the isotropic plate. Again, more rapid oscillations are
encountered in the orthotropic plate with higher flexural rigidity.

Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:733–754
DOI: 10.1002/cnm
752 J. SLADEK ET AL.

3.5
isotropic FGM
3
orthotropic FGM

M11/M11hom(x 1=x2=a/2)
2.5

1.5

0.5

-0.5

-1
0 0.2 0.4 0.6 0.8 1 1.2
t/t0

Figure 16. Influence of orthotropic material properties on time variation of the bending moment for
a clamped square plates under a uniform impact load.

5. CONCLUSIONS

A MLPG is presented for solving bending problems of thick FGM plates described by the
Reissner–Mindlin theory. Both isotropic and orthotropic material properties are considered in
the analysis. The Laplace-transform technique is applied to eliminate the time variable in the
coupled governing partial differential equations for plates under an impulsive load. The use of
the Laplace-transform in forced vibration analysis converts the transient dynamic problem to
a quasi static problem. The analysed domain is divided into small overlapping circular sub-
domains. A unit step function is used as the test functions in the local weak form. The de-
rived local boundary-domain integral equations are non-singular. The MLS scheme is adopted
for approximating the physical quantities. The proposed method is a truly meshless method,
which requires neither domain elements nor background cells, in neither the interpolation nor the
integration.
The proposed method is an alternative numerical tool to many existing computational methods.
The main advantage is its simplicity. Contrary to the conventional BEM, all integrands in the
present formulation are regular. No special computational techniques are required to evaluate the
integrals.

ACKNOWLEDGEMENTS
The authors acknowledge the supports by the Slovak Research and Development Support Agency registered
under the project number APVV-51-021205, by the Slovak Grant Agency under the project number
VEGA-2/6109/6. This work was supported in part by the EU Network of Excellence project Knowledge-
based Multicomponent Materials for Durable and Safe Performance (KMM-NoE) under the contract no.
NMP3-CT-2004-502243.

Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:733–754
DOI: 10.1002/cnm
ANALYSIS OF THICK FUNCTIONALLY GRADED PLATES 753

REFERENCES
1. Suresh S, Mortensen A. Fundamentals of Functionally Graded Materials. Institute of Materials: London, 1998.
2. Miyamoto Y, Kaysser WA, Rabin BH, Kawasaki A, Ford RG. Functionally Graded Materials; Design, Processing
and Applications. Kluwer Academic Publishers: Dordrecht, 1999.
3. Kim JH, Paulino GH. Isoparametric graded finite elements for nonhomogeneous isotropic and orthotropic materials.
Journal of Applied Mechanics (ASME) 2002; 69:502–514.
4. Sladek J, Sladek V, Zhang Ch. Transient heat conduction analysis in functionally graded materials by the meshless
local boundary integral equation method. Computational Material Science 2003; 28:494–504.
5. Sladek J, Sladek V, Krivacek J, Zhang Ch. Local BIEM for transient heat conduction analysis in 3-D axisymmetric
functionally graded solids. Computational Mechanics 2003; 32:169–176.
6. Sladek J, Sladek V, Atluri SN. Meshless local Petrov–Galerkin method for heat conduction problem in an
anisotropic medium. Computer Modeling in Engineering and Sciences (CMES) 2004; 6:309–318.
7. Sladek V, Sladek J, Zhang Ch. Local integro-differential equations with domain elements for the numerical
solution of partial differential equations with variable coefficients. Journal of Engineering Mathematics 2005;
51:261–282.
8. Sladek J, Sladek V, Hellmich Ch, Eberhardsteiner J. Heat conduction analysis of 3-D axisymmetric and anisotropic
FGM bodies by meshless local Petrov–Galerkin method. Computational Mechanics 2005, in press.
9. Sladek J, Sladek V, Atluri SN. Meshless local Petrov–Galerkin method in anisotropic elasticity. Computer
Modeling in Engineering and Sciences (CMES) 2004; 6:477–489.
10. Sladek J, Sladek V, Zhang Ch. Stress analysis in anisotropic functionally graded materials by the MLPG method.
Engineering Analysis by Boundary Elements 2005; 29:597–609.
11. Sladek J, Sladek V, Krivacek J, Zhang Ch. Meshless local Petrov–Galerkin method for stress and crack analysis
in 3-D axisymmteric FGM bodies. Computer Modeling in Engineering and Sciences (CMES) 2005; 8:259–270.
12. Wang J, Huang M. Boundary element method for orthotropic thick plates. Acta Mechanica Sinica 1991; 7:258–266.
13. Beskos DE. Dynamic analysis of plates. In Boundary Element Analysis of Plates and Shells, Beskos DE (ed.).
Springer: Berlin, 1991; 35–92.
14. Providakis CP, Beskos DE. Dynamic analysis of plates by boundary elements. Applied Mechanics Reviews
(ASME) 1999; 52:213–236.
15. Reissner E. The effect of transverse shear deformation on the bending of elastic plates. Journal of Applied
Mechanics (ASME) 1945; 12:A69–A77.
16. Mindlin RD. Influence of rotary inertia and shear on flexural motions of isotropic, elastic plates. Journal of
Applied Mechanics (ASME) 1951; 18:31–38.
17. Van der Ween F. Application of the boundary integral equation method to Reissner’s plate model. International
Journal for Numerical Methods in Engineering 1982; 18:1–10.
18. Antes H. Static and dynamic analysis of Reissner–Mindlin plates. In Boundary Element Analysis of Plates and
Shells, Beskos DE (ed.). Springer: Berlin, 1991; 312–340.
19. Wen PH, Aliabadi MH. Boundary element frequency domain formulation for dynamic analysis of Mindlin plates.
In Boundary Element Techniques, Aliabadi MH, Selvadurai APS, Tan CL (eds). 2005; 35–42.
20. Reddy JN. Analysis of functionally graded plates. International Journal for Numerical Methods in Engineering
2000; 47:663–684.
21. Qian LF, Batra RC, Chen LM. Static and dynamic deformations of thick functionally graded elastic plates
by using higher-order shear and normal deformable plate theory and meshless local Petrov–Galerkin method.
Composites Part B—Engineering 2004; 35:685–697.
22. Belytschko T, Krongauz Y, Organ D, Fleming M, Krysl P. Meshless methods: an overview and recent developments.
Computer Methods in Applied Mechanics and Engineering 1996; 139:3–47.
23. Long SY, Atluri SN. A meshless local Petrov–Galerkin method for solving the bending problem of a thin plate.
Computer Modeling in Engineering and Sciences 2002; 3:11–51.
24. Krysl P, Belytschko T. Analysis of thin plates by the element-free Galerkin method. Computational Mechanics
1985; 16:1–10.
25. Noguchi H, Kawashima T, Miyamura T. Element free analyses of shell and spatial structures. International
Journal for Numerical Methods in Engineering 2000; 47:1215–1240.
26. Liew KM, Chen XL, Reddy JN. Mesh free radial basis function method for buckling analysis of non-uniformly
loaded arbitrarily shaped shear deformable plates. Computer Methods in Applied Mechanics and Engineering
2004; 193:205–224.

Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:733–754
DOI: 10.1002/cnm
754 J. SLADEK ET AL.

27. Sladek J, Sladek V, Mang HA. Meshless formulations for simply supported and clamped plate problems.
International Journal for Numerical Methods in Engineering 2002; 55:359–375.
28. Sladek J, Sladek V, Mang HA. Meshless LBIE formulations for simply supported and clamped plates under
dynamic load. Computers and Structures 2003; 81:1643–1651.
29. Sladek J, Sladek V. Local boundary integral equation method for thin plate bending problems. Building Research
Journal 2004; 52:89–120.
30. Stehfest H. Algorithm 368: numerical inversion of Laplace transform. Communications of Association for
Computing Machinery 1970; 13:47–49.
31. Reddy JN. Mechanics of Laminated Composite Plates: Theory and Analysis. CRC Press: Boca Raton, FL, 1997.
32. Atluri SN. The Meshless Method (MLPG) For Domain and BIE Discretizations. Tech Science Press: Encino,
2004.
33. Lancaster P, Salkauskas K. Surfaces generated by moving least square methods. Mathematics of Computation
1981; 37:141–158.
34. Nayroles B, Touzot G, Villon P. Generalizing the finite element method. Computational Mechanics 1992;
10:307–318.
35. Providakis CP, Beskos DE. Inelastic transient dynamic analysis of Reissner–Mindlin plates by the D/BEM.
International Journal for Numerical Methods in Engineering 2000; 49:383–397.

Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:733–754
DOI: 10.1002/cnm

S-ar putea să vă placă și