(2004) Finite Elements For Functionally Graded Reissner-Mindlin Plates

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 21

Comput. Methods Appl. Mech. Engrg.

193 (2004) 705–725


www.elsevier.com/locate/cma

Finite elements for functionally graded


Reissner–Mindlin plates
Lucia Della Croce a, Paolo Venini b,*

a
Department of Mathematics, University of Pavia, Via Ferrata 1, I-27100 Pavia, Italy
b
Department of Structural Mechanics, University of Pavia, Via Ferrata 1, I-27100 Pavia, Italy
Received 18 March 2003; received in revised form 15 July 2003; accepted 12 September 2003

Abstract

Equations governing the behavior of rectangular plates made of functionally graded materials (FGMs) are deter-
mined in this paper using the variational approach. Derivation of such equations is based on Reissner–Mindlin plate
theory that is extended to handle two-constituent material distribution through the thickness. Material properties are
assumed to vary with the power law in terms of the volume fractions of the constituent. Within a static analysis
framework, the main focus of the paper is the proposal of a locking-free hierarchic family of finite elements that is
numerically tested on plates of different material grading. Convergence and stability properties of our approach are
assessed and comparisons with available solutions presented.
 2003 Elsevier B.V. All rights reserved.

Keywords: Functionally graded plates; Hierarchic finite elements; Reissner–Mindlin model

1. Introduction

Composite materials and structures are more and more frequently used in different engineering fields
mainly because of a stiffness-to-weight ratio that is particularly favorable [24]. Composite materials may be
designed at the microstructural level for example by means of topology optimization approaches leading to
extreme materials, see e.g. [30] for optimal materials enjoying extreme Poisson ratio values. Plates and shells
made of composite materials are usually designed with optimal design strategies that have ply thicknesses,
lamination angles and stacking sequences as design f3 variables. Refined models for the analysis of lami-
nated plates and shells have been proposed since the mid eighties and [22] appears to be a pioneering
contribution in this area. The main downside of laminated composites is however represented by the
weakness of interfaces between adjacent layers known as delamination phenomena that may lead to
structural failure. Likewise, debonding between matrix and fiber may occur in the presence of extreme

*
Corresponding author. Tel.: +39-382-528-450/451; fax: +39-382-528-422.
E-mail address: venini@vedipmec.unipv.it (P. Venini).

0045-7825/$ - see front matter  2003 Elsevier B.V. All rights reserved.
doi:10.1016/j.cma.2003.09.014
706 L. Della Croce, P. Venini / Comput. Methods Appl. Mech. Engrg. 193 (2004) 705–725

thermal loads due to the mismatch in their mechanical properties. To (partially) overcome these problems,
a new class of materials named functionally graded materials (FGMs) has recently been proposed whose
physical properties vary through the thickness in a continuous manner and are therefore free from interface
weaknesses typical of laminated composites. Grading of mechanical properties is usually obtained by
varying continuously volume fractions of two (or more) material constituents such as ceramic and metal.
After the concept of FGMs was set by the Japanese school of material science, see e.g. [15] for an earlier
contribution and [31] for a more recent treatise, and confirmation of their potentials found even in natural
materials [2], several branches of research originated and are still being broadened by research groups all
over the world:

(1) thermal stress analysis and applications. This represents the most typical and frequently investigated
area of research. Remarkable contributions in this field include [33] where an approach to design FGMs
with the goal of reducing thermal stresses is proposed, [23] that introduces a finite-element approxima-
tion scheme for the transient thermoelastic analysis of von Karman functionally graded plates, [25] that
extends the analysis of [23] to the case of heated cylinders and, last but not least, a series of pioneering
papers of the Japanese school for which reference is made to [26] among others;
(2) mechanical models with applications to FGMs plate instability. In [1] analytical exact solutions are de-
rived within the framework of 3D and 2D elasticity, the latter case calling for the introduction of an
‘‘equivalent plate’’. Buckling of functionally graded plates under in-plane compressive loading is inves-
tigated in [9] where a peculiar nonlinear plate model is adopted; the influence of side-to-thickness and
aspect ratios on the buckling load is therein assessed by means of extensive numerical simulations. Na-
vierÕs solutions and finite-element models based on third-order shear deformation theory are derived in
the recent contribution [26] where Lagrange interpolation functions are adopted for in-plane displace-
ments while Hermite ones are used for the approximation of the out-of-plane displacement and rota-
tions. Shallow shells made of FGMs are numerically investigated in [34];
(3) fracture mechanics analysis and models. Mixed-mode fracture for orthotropic FGMs is investigated
numerically in [13] via a finite-element approach whereas [14] presents an analysis of the stress intensity
factor in functionally graded media, yet by means of the finite-element method. Earlier basic and meth-
odological considerations on fracture mechanics for FGMs may be found in [7,12];
(4) vibration and control applications. A considerable part of the existing literature on this topic may be
attributed to the Singapore school directed by K.M. Liew. The parametric resonance of plates struc-
tures made of FGMs is investigated in [20] in the presence of harmonic in-plane loadings. Standard
Lagrangian dynamics represents the core of the approach that is substantiated by extensive numerical
simulations. The contribution [21] is concerned with frequency domain applications of piezoelectric
sensors and actuators for the mitigation of the vibrations of shells made of FGMs. Active control of
FGM plates using integrated piezoelectric sensing and actuating devices is investigated in [10] whereas
the effects of temperature gradients on piezo-controlled plates and shells are studied in [16,17] respec-
tively. In both cases, temperature variation is assumed to occur in the thickness direction only and dif-
ferent positions of sensing/actuating piezoelectric patches are analyzed in order to find the most efficient
ones. Further results are presented in [11] in which a systematic finite-element framework for the dy-
namic analysis of piezoelectric shells is also introduced. The problem of deflection control is also ana-
lyzed in [18] within the framework of the element-free Galerkin method that is presented as an
appealing alternative to the more classical finite-element schemes utilized in previous contributions.

As one may realize looking at the above picture, that is however by no means exhaustive, FGMs rep-
resent a rapidly growing area of research that encompasses material science, mechanics, dynamics and
numerics. Within this last framework, though a few finite-element models have been presented, an inves-
tigation concerning classical issues such as stability, convergence and (shear) locking seems to be missing.
L. Della Croce, P. Venini / Comput. Methods Appl. Mech. Engrg. 193 (2004) 705–725 707

The objective of this contribution is therefore to derive a hierarchic family of finite elements of which
convergence, stability and locking properties are assessed. As a byproduct, a variational formulation of the
FGM plate problem is derived. Results of physical interest are also proposed concerning thermoelastic
analysis of relatively thick to thin FGM plates made of two graded constituents.
The paper is organized as follows: Sections 2 and 3 presents basic mechanical and thermal equations
needed to arrive at the variational formulation that is derived in Section 4. Section 5 presents the hierarchic
family of finite elements that is used within the numerical investigations presented in Section 6. Attention is
paid to the capability of the proposed approximation scheme to handle the shear-locking phenomenon as
well as to the assessment of convergence rates and stability properties. Concluding remarks and need for
future investigations are presented in Section 7.

2. Governing equations

A plate of thickness t is considered made of a material functionally graded through the thickness. An
orthogonal Cartesian frame is introduced, the ðx; yÞ plane being the undeformed midplane of the plate.
A linear elastic material behavior is considered and the temperature field is assumed to be known.

2.1. Kinematics

Plate structures made of graded materials are characterized by non-negligible shear deformation so that
the extension of the classical model suggested by Reissner and Mindlin [19,27] to the case of graded
material plates provides a good compromise between numerical accuracy and computational burden. In-
plane displacements u and v and the normal displacement w are therefore given the form:
uðx; y; zÞ ¼ zhx ðx; yÞ; vðx; y; zÞ ¼ zhy ðx; yÞ; wðx; y; zÞ ¼ wðx; yÞ; ð1Þ
where hx and hy are the rotations of the normal to the undeformed middle surface in the x–z and y–z planes,
respectively. In-plane strain measures are classically denoted as xx , yy , xy and may classically be written
within the Reissner–Mindlin framework as
2 3 2 3
xx hx;x
4 yy 5 ¼ z4 hy;y 5: ð2Þ
xy hx;y þ hy;x
Transverse shear strains cx and cy are conversely constant through the thickness and are defined as
   
cx w;x þ hx
¼ : ð3Þ
cy w;y þ hy

2.2. Functionally graded constitutive law and statics

A single layer FGM plate of uniform thickness is considered. To obtain the effective properties of the
graded plate a simple rule of mixture is adopted, i.e. a power law of type
 n
2z þ t
V ðzÞ ¼ ðVt  Vb Þ þ Vb ; ð4Þ
2t
where V denotes a generic material property, Vt and Vb are the corresponding values at the top and bottom
faces of the plate, z is the thickness coordinate (t=2 6 z 6 t=2) and n is the volume fraction exponent which
takes on non-negative values.
708 L. Della Croce, P. Venini / Comput. Methods Appl. Mech. Engrg. 193 (2004) 705–725

In the recent literature, see e.g. [8,28,29], a few more refined mixture models have been proposed that
shall be object of future investigations, even for comparison purposes with the simpler one adopted herein.
A homogeneous plate made of the top-material only (ceramic) is retrieved for n ¼ 0, while the content of
bottom-material (metal) in the plate increases as the value of n increases. Here we assume the modulus of
elasticity E and the thermal coefficient a to change through-the-thickness z according to Eq. (4), while
PoissonÕs ratio m is assumed to be constant. In-plane and out-of-plane stress–strain constitutive law is (only
formally) classical and one may write
2 3 2 3
rx xx    
4 ry 5 ¼ zB4 yy 5; sx c
¼S x ; ð5Þ
sy cy
rxy xy
T T
where ½rx ; ry ; rxy  and ½sx ; sy  are the in-plane and shear stresses whereas grading matrices B and S are
defined as:
2 3 2 3
1 m 0 1m
EðzÞ 4 m 1 EðzÞ 6 2 0 7
B¼ 0 5; S¼ 4 ð6Þ
1m 0 02 1  m 1m 2 1  m 5:
0
2 2
Moments and shear forces are obtained via standard integration over the thickness, i.e.
2 3 2 3 2 3
Mx Z t=2 rx hx;x Z t=2
4 My 5 ¼ 4 ry 5z dz ¼ 4 hy;y 5 Bz2 dz ð7Þ
Mxy t=2 rxy hx;y þ hy;x t=2

and
  Z t=2     Z t=2
Qx sx w;x þ hx
¼ dz ¼ S dz: ð8Þ
Qy t=2 sy w;y þ hy t=2
In the presence of material grading, Eqs. (7) and (8) cease to be standard in that the resulting plate
coefficients depend on the selected grading profile. Eventually one may define constitutive matrices D and Q
as
Z t=2 Z t=2
2
Dij ¼ Bij z dz; i; j ¼ 1; 2; 3; Qkl ¼ Skl dz k; l ¼ 1; 2; ð9Þ
t=2 t=2

where
Z  n Z t=2
ðEt  Eb Þ t=2 2z þ t 2 Eb 2
D11 ¼ z dz þ z dz
1  m2 t=2 2t t=2 1  m2
ðEt  Eb Þt3 n2 þ n þ 2 Eb t 3
¼ 2
þ ;
1m 4ðn þ 1Þðn þ 2Þðn þ 3Þ 12ð1  m2 Þ
Z   n  ð10Þ
ðEt  Eb Þk t=2 2z þ t
Q11 ¼ þ Eb dz;
2ð1 þ mÞ t=2 2t
 
kt Et  Eb
¼ þ Eb ;
2ð1 þ mÞ nþ1

1m
D12 ¼ D21 ¼ mD11 ; D13 ¼ D31 ¼ 0; D33 ¼ D11 ; ð11Þ
2
L. Della Croce, P. Venini / Comput. Methods Appl. Mech. Engrg. 193 (2004) 705–725 709

and
Q22 ¼ Q11 ; Q12 ¼ Q21 ¼ 0: ð12Þ

3. Thermal analysis

FGMs are primarily used in situations where large temperature gradients are encountered. This section
is therefore devoted to the assessment of the influence of a temperature field on functionally graded plates.
Within our analysis, constant temperatures are imposed at the ceramic and metal surfaces and the scalar
temperature field is assumed to depend on the z coordinate only. Furthermore, a stress free reference
temperature T0 ¼ 0 is considered. The temperature distribution along the thickness can be obtained by
solving the one-dimensional steady state heat transfer equation:
 
d dT
 kðzÞ ¼0 ð13Þ
dz dz

with the thermal boundary conditions

T ¼ Tt at z ¼ t=2;
T ¼ Tb at z ¼ t=2:

Notice that Eq. (13) may be integrated once allowing to write


dT
kðzÞ ¼ constant: ð14Þ
dz
By imposing one of the previous boundary conditions, the constant in Eq. (14) may be evaluated and one is
left with an initial value problem that may be solved numerically e.g. with a Runge–Kutta scheme for any
distribution kðzÞ. In the simple case thermal conductivity kðzÞ varies according the power-law distribution
 n
2z þ t
kðzÞ ¼ ðkt  kb Þ þ kb ; ð15Þ
2t

the solution of Eq. (13) may be evaluated exactly as we did in our computations. One may write
Z t=2
Tt  Tb dn
T ðzÞ ¼ Tt  Z t=2 : ð16Þ
dz z kðnÞ
t=2 kðzÞ

In the case of a thermally homogeneous plate, i.e. when k does not depend on z, the temperature distri-
bution is linear through the thickness. Excursions from the linear distribution are obtained by changing the
volume fraction index n. Fig. 1 shows the variation of temperature distributions versus plate thickness for
various values of n. It is observed that temperature within the FGM plates with both ceramic and metal
constituents is always smaller than that corresponding to a purely ceramic or metallic plate. Thermal effects
are classically introduced at the constitutive level, the novelty being the variation of the thermal coefficient a
(and of the Young modulus E) with respect to z. One may write
2 3 02 3 2 31
rx xx aðzÞ DT
4 ry 5 ¼ B@4 yy 5  4 aðzÞ DT 5A ð17Þ
rxy xy 0
710 L. Della Croce, P. Venini / Comput. Methods Appl. Mech. Engrg. 193 (2004) 705–725

Distribution of temperature versus the thickness of the plate


300
n=0
n=1
n=2
n=0.5
250

200
temperature in ˚C

150

100

50

0
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
non-dimensional thickness

Fig. 1. Temperature versus plate thickness for various n values.

or, including out-of-plane shear stresses,


2 3 2 3 2 3
rx xx DT
6 ry 7 6 yy 7 6 DT 7
6 7 6 7 6 7
6 rxy 7 ¼ BS 6 xy 7  aðzÞEðzÞ 6 0 7; ð18Þ
6 7 6 7 1  m 6 7
4 sx 5 4 cx 5 4 0 5
sy c y 0

where BS is a 5 · 5 matrix which collects in an obvious way matrices B and S. Thermal moment and shear
resultants are eventually obtained as
2 3 2 R t=2 3
Mxt aðzÞEðzÞT ðzÞz dz
6 Mt 7 6 Rt=2 7
6 y7 6 t=2 7
6 t 7 1 6 aðzÞEðzÞT ðzÞz dz 7
6 Mxy 7 ¼ 6 t=2 7
6 7 1  m6 0 7: ð19Þ
6 Qt 7 6 7
4 x 5 6 7
4 0 5
t
Qy 0

4. Energy and variational formulations

A three-dimensional plate-like elastic body V is introduced as


V ¼ fðx; y; zÞ 2 R3 : ðx; yÞ 2 X  R2 ; z 2 ðt=2; t=2Þg; ð20Þ
L. Della Croce, P. Venini / Comput. Methods Appl. Mech. Engrg. 193 (2004) 705–725 711

where X is a bounded smooth domain. In the Reissner–Mindlin model the potential energy P ¼
Pðhx ; hy ; wÞ can be additively decomposed in its bending, shear and load terms as
P ¼ Pb þ Ps  Pl ; ð21Þ
where
2 3
Z rx
1
Pb ¼ ½ xx yy xy 4 ry 5dV ; ð22Þ
2 V rxy
Z  
k c
Ps ¼ ½ sx sy  x dV ; ð23Þ
2 V cy
Z
Pl ¼ qw dV ð24Þ
V

in which k is the shear correction factor and q is the external load. Notice that primary variables of our
formulations are hx , hy , w and therefore functional above should be re-written thanks to compatibility and
constitutive relations in terms of these only independent fields. One may then integrate through the
thickness and observe that the bending term Pb may be in turn decomposed into its elastic and thermal
contributions Pb ¼ P fb  Pt that respectively read
b
2 3
Z Z t=2 hx;x
fb ¼ 1
P ½ hx;x ; hy;y ; hx;y þ hy;x  Bz dz4 hy;y
2 5dx dy ð25Þ
2 X t=2 h þh x;y y;x

and
2 3
Z Z t=2 aðzÞT ðzÞ
1
Ptb ¼ ½ hx;x ; hy;y ; hx;y þ hy;x  Bz4 aðzÞT ðzÞ 5dz dx dy: ð26Þ
2 X t=2 0
As to the shear contribution one obtains:
Z   Z t=2  
k w;x þ hx w þ hx
Ps ¼ S dz ;x dx dy ð27Þ
2 X w;y þ hy t=2 w;y þ hy

allowing to determine the total potential energy in the form


Z  
1 1  m 2
Pðhx ; hy ; wÞ ¼ D11 h2x;x þ h2y;y þ 2mhx;x hy;y þ hx;y þ hy;x dx dy
2 X 2
Z
k  2
þ Q44 ½ðw;x þ hx Þ2 þ w;y þ hy  dx dy
2 X
Z Z
1 t t
 ðhx;x Mx þ hy;y My Þ dx dy  qw dx dy: ð28Þ
2 X X

Stationarity of the total potential energy P is used hereafter to derive the variational formulation of
the functionally graded plate needed for the development of our finite-element approach. One formally
writes
 
dPðhx ; hy ; wÞ ¼ d Pfb þ Ps  Pl  Pt ¼ 0; ð29Þ
712 L. Della Croce, P. Venini / Comput. Methods Appl. Mech. Engrg. 193 (2004) 705–725

where variations are of course with respect to the independent variables hx , hy , w. Upon introduction of a
scalar parameter s and of three independent fields wx , wy and v, a variation of the independent variables hx ,
hy , w is now written as
hx hx þ swx ; hy hy þ swy ; w w þ sv: ð30Þ
In order to enforce the minimum potential energy condition, the expressions in (30) are plugged into the
total potential energy (28) and the derivative with respect to s computed. Setting finally s ¼ 0 in the
resulting expression, i.e. requiring hx , hy , w to be the solution, one gets
Z  
dP 1m
¼ D11 hx;x wx;x þ hy;y wy;y þ mðhx;x wy;y þ hy;y wx;x Þ þ ðhx;y þ hy;x Þðwx;y þ wy;x Þ dx dy
ds s¼0 X 2
Z
þ S44 ðhx þ w;x Þðwx þ v;x Þ þ ðhy þ w;y Þðwy þ v;y Þ dx dy
Z X Z
 ðwx;x Mxt þ wy;y Myt Þ dx dy  qv dx dy; ð31Þ
X X

that is the milestone for the variational formulation to be formalized next. The following functional spaces
are introduced on the purpose of deriving the variational formulation

• L2 ðXÞ the space of square integrable functions in X:



Z
L2 ðXÞ ¼ w j 2 2
jwj dx ¼ kwkL2 ðXÞ < þ 1 ; ð32Þ
X

• H01 ðXÞ, the space of distributions which together with their first derivatives are in L2 ðXÞ, and whose traces
on C vanish:

 
ow ow
1
j 2
H0 ðXÞ ¼ w w 2 L ðXÞ : ;
ox oy
2
2 L ðXÞ; wjC ¼ 0 : ð33Þ

Both L2 ðXÞ and H01 ðXÞ are Hilbert spaces endowed with the classical inner products. Then the solution
2
ðhx ; hy ; wÞ of the plate problem is the minimizer over the space H  W ¼ ðH01 ðXÞÞ  H01 ðXÞ of the scaled
3
total energy functional obtained from (32) after dividing by t . The variational formulation we arrive at is
the following:
8
< Find ð~
2
> h; wÞ 2 ðH01 ðXÞÞ  H01 ðXÞ such that
DAð~ h; ~
wÞ þ Qt kkrw  ~ hk0 ¼ t3 ðr~
2 2
w; M t Þ þ ðp; vÞ ð34Þ
>
: ~ 1 2 1
8ðw; vÞ 2 ðH0 ðXÞÞ  H0 ðXÞ;

where ~h ¼ ðhx ; hy Þ, D ¼ D11 , Q ¼ Q44 , ð; Þ denotes the L2 ðXÞ inner product, k  k0 denotes the norm in
L ðXÞ, pðx; yÞ is the scaled load function and Að~
2
h; ~
wÞ is the bilinear form
Z  
~ ~ 1m
Aðh; wÞ ¼ hx;x wx;x þ hy;y wy;y þ mðhx;x wy;y þ hy;y wx;x Þ þ ðhx;y þ hy;x Þðwx;y þ wy;x Þ dx dy : ð35Þ
X 2
The variational formulation of the continuous problem leads naturally to the one associated to the discrete
problem to be adopted in a finite-element framework. Having in mind a conformal approximation,
introduced the finite dimensional subspace Hh  Wh  H  W , one may write the discrete variational
problem as
L. Della Croce, P. Venini / Comput. Methods Appl. Mech. Engrg. 193 (2004) 705–725 713

8
>
< Find ð~ ~ h  Wh such that
hh ; wh Þ 2 H
DAðhh ; wh Þ þ Qt2 kkrwh  ~
~ ~ hh k20 ¼ t3 ðr~
wh ; M t Þ þ ðp; vh Þ ð36Þ
>
: ~ ~
8ðwh ; vh Þ 2 Hh  Wh :
The discretization of the problem with standard low degree polynomials typically locks, that is the
approximation they afford is not uniform with respect to the thickness t, but deteriorates as t tends to 0.
High order elements have shown to be robust and free of locking phenomena. Along this path, we shall
investigate the locking behavior of various finite-element techniques for the approximation of functionally
graded plates. An efficient hierarchic family of finite elements will be proposed and utilized for the com-
putation of the response of plates with material properties varying between those of the ceramic and metal.

5. Hierarchic finite elements for Mindlin functionally graded plates

In order to give a finite-element approximation of Problem (36) a family of hierarchic finite elements is
introduced (see [6] for the details). We refer to [3–5] for an introduction of the so-called p-version of the
finite-element method. If P denotes the current degree of polynomial shape functions, we recall that a
finite-element approach is said to be hierarchic when upon mesh refinement, say from P to P þ 1, all shape
functions of ‘‘level’’ P are retained and new ones added in such a way that the P-level stiffness matrix is
embedded into the new one P þ 1 as a block. A dramatic reduction of computation time follows when this
property is enjoyed by the adopted approximation scheme. Polynomial functions are used to construct the
approximation spaces in the present study and normal deflection and rotations are approximated by means
of the same space. This allows a further reduction of the computational burden since repeated blocks are
present in the stiffness matrix. The shape functions used are based on the family of Legendre polynomials.
The orthogonality properties of Legendre polynomials are inherited by the elementary stiffness matrix
which has a minimum number of non-vanishing entries. Let Pp be the spaces of polynomials defined on the
square [)1,1] · [)1,1] of degree less than or equal to p in two variables. We take all monomials of the space
Pp and we add the following monomial terms:

• the monomial fxyg in the case p ¼ 1;


• the monomials fxp y; xy p g in the case p P 2.

The basis functions obtained in this way can be named as nodal, side and internal functions. More
precisely, there are four nodal shape functions which are non-zero at one vertex while vanishing at all other
vertices. There are 4ðp  1Þ side functions associated to the sides which are defined by using the Legendre
polynomials. Each side function is non-zero on one side while vanishing on the remaining sides. Internal
shape functions appear only for p P 4. These functions vanish along the whole boundary of the quadri-
lateral and are therefore often referred to as bubble functions. The first internal function is found for p ¼ 4
whereas other functions may easily be obtained from it in term of Legendre polynomials. The number of
internal functions is 12 ðp  2Þðp  3Þ. In Table 1 the number of shape functions of each type is given for
general values of degree p.
Let Sp be the space generated by the basis functions obtained in this way. The finite-element discreti-
~ h for the approximate rotations
zation of Problem (36) calls for the choice of suitable finite-element spaces H
and Wh for the approximate normal displacement. The following choice is made:
~ h ¼ f~
H gh jK 2 ðSp Þ2 8K 2 Th g;
gh 2 H01 ðXÞ  H01 ðXÞ : ~
ð37Þ
Wh ¼ fwh 2 H01 ðXÞ : wh jK 2 Sp 8K 2 Th g:
714 L. Della Croce, P. Venini / Comput. Methods Appl. Mech. Engrg. 193 (2004) 705–725

Table 1
Number of shape functions for different degree p
Degree Nodal s.f. Side s.f. Internal s.f. Total no. s.f.
1 4 – – 4
2 4 4 – 8
3 4 8 – 12
4 4 12 1 17
5 4 16 3 23
6
.. 4
.. 20
.. 6. 30
. . . ..
1
p 4 4ðp  1Þ 2
ðp  2Þðp  3Þ 4p þ 12 ðp  2Þðp  3Þ

Upon denoting with F ðx; yÞ the shape functions previously introduced, the three fields ðhx ; hy ; wÞ are
approximated as follows:
X
hx ðx; yÞ ¼ ai Fi ðx; yÞ ¼ F  a;
i
X
hy ðx; yÞ ¼ bi Fi ðx; yÞ ¼ F  b; ð38Þ
i
X
wðx; yÞ ¼ ci Fi ðx; yÞ ¼ F  c;
i

where the subscripts vary over the range of the approximation space. In Fig. 2 we show the set of the finite
elements together with their degrees-of-freedom. To each node the degrees-of-freedom ðhx ; hy ; wÞ are
associated. Unknown nodal displacements and thermomechanical loads are formally defined as
2 3 2 3
a Tx
d ¼ 4 b 5; p ¼ 4 T y 5;
c pz

where
Z Z Z
Tx ¼ FT;x Mxt dx dy; Ty ¼ FT;y Myt dx dy; pz ¼ FT w dx dy: ð39Þ
E E E

Then the discrete problem at the element level is given the matrix form
Kd ¼ p; ð40Þ
where K is the stiffness matrix. The expression of the elementary stiffness matrix for a generic element E of
the mesh is the following:

Fig. 2. Proposed finite elements and degrees-of-freedom.


L. Della Croce, P. Venini / Comput. Methods Appl. Mech. Engrg. 193 (2004) 705–725 715

Table 2
Number of integration points
p Max Np K
1 2 2·2 3
2 4 3·3 5
3 6 4·4 7
4 8 5·5 9

2 3
1m 1m
6 DBxx þ 2 Byy þ QB DmBxy þ
2
Byx QBx 7
6 1m 7
K¼6 7; ð41Þ
4 DByy þ Bxx þ QB QBy 5
2
symmetric QðBxx þ Byy Þ
where
Z
B¼ FT F dx dy;
E
Z Z
Bx ¼ FT F;x dx dy; By ¼ FT F;y dx dy;
ZE E
Z ð42Þ
Bxx ¼ FT;x F;x dx dy; Byy ¼ FT;y F;y dx dy;
ZE ZE
Bxy ¼ FT;x F;y dx dy; Byx ¼ FT;y F;x dx dy:
E E

Classical quadrature formulas of Gaussian type are used for the computation of the stiffnes matrix inte-
grals. Let us denote by Max the maximum degree of the polynomials appearing in the integrals, by Np the
number of integration points and by K the degree of exactness of the correspondent quadrature formula. In
Table 2 we report the quantities versus the degree p of the finite elements. Using the number of integration
points here indicated we compute exactly all the integrals defined in (42).

6. Numerical results

The analysis of the FGMs plates is performed for a combination of materials of type ceramic-metal. The
lower plate surface is assumed to be aluminum while the top surface is assumed to be zirconia. Material
properties vary with the power law and the values n ¼ 0, 0.5, 1, 2 are considered. Physical material
properties are given in Table 3. As to geometry and side constraints, a simply supported square plate is
considered of side a ¼ 0:2 m and thickness t ranging from 102 to 103 m (see Fig. 3). In addition to an
uniformly distributed mechanical normal load on the top surface, the plate is subject to a thermal field
where the ceramic rich top surface is held at 300 C and the metal rich bottom surface is held at 20 C.

Table 3
Material properties
Aluminum Zirconia
YoungÕs modulus 70 GPa 151 GPa
PoissonÕs ratio 0.3 0.3
Thermal conductivity 204 W/mK 2.09 W/mK
Thermal expansion 23 · 106 /C 10 · 106 /C
716 L. Della Croce, P. Venini / Comput. Methods Appl. Mech. Engrg. 193 (2004) 705–725

Fig. 3. Plate geometry, load and material distribution.

Table 4
Meshes and degrees-of-freedom for finite elements Q2 , Q3 and Q4
Mesh Q2 dofs Q3 dofs Q4 dofs
4·4 195 315 483
5·5 288 468 723
6·6 399 651 1011
8·8 675 864 1291

A stress free temperature of 0 C is assumed. Due to the symmetry of the domain, computations have been
performed on a quarter plate only. All the generated meshes were uniform ones and, for all the adopted
finite elements (Q2 , Q3 and Q4 ), Table 4 presents the consequent global number of degrees-of-freedom.
Objectives of the numerical investigations to be discussed in forthcoming subsections are of physical and
numerical nature and include the following ones:

(1) to determine the behavior of the plate with respect to the selected material grading using as expected
bounds the two homogeneous plates made of a single constituent (aluminum or zirconia). To this goal,
several simulations are performed for different values of the grading parameter n;
(2) to assess the behavior of our finite-element scheme with respect to locking. For given in-plane dimen-
sions (a ¼ b ¼ 0:2 m), the thickness h is progressively reduced from 0.01 to 0.001 m and the numerical
solution compared with available analytical results;
(3) to investigate on stability and convergence rates of the proposed approach.
L. Della Croce, P. Venini / Comput. Methods Appl. Mech. Engrg. 193 (2004) 705–725 717

The following non-dimensional parameters are introduced for use within the numerical simulations to be
presented next (here q denotes the intensity of the applied mechanical load and Eb is the YoungÕs modulus
at the bottom face (aluminum)):

• central deflection w ¼ w=t


a4 q
• load parameter P ¼
Eb t 4
• thickness t ¼ t=a

6.1. Physical behavior of functionally graded plates

Several simulations were conducted to assess the behavior of functionally graded plates to mechanical
and thermal loads. Focusing on the midpoint deflection w  , the objective of the analysis is to compare the
response of functionally graded plates with the one of homogeneous single-constituent plates, i.e. those
made of aluminum or zirconia only. Toward this goal, several numerical studies are conducted with
increasing mechanical and thermal loads. Fig. 4 shows the central deflection w  ¼ w=t due to a sequence of
mechanical loads for different values of n. The non-dimensional load P takes values in the interval ()13,0)
and one may see that all the plates with intermediate material properties experience intermediate values of
deflection. This was expected since the metallic plate is the one with the lowest stiffness and the ceramic
plate is the one with the highest stiffness. To further consider the influence of a temperature field on FGM
response, in Fig. 5 we report the non-dimensional deflection w=t of a plate under uniform mechanical load
and thermal field (for two consecutive hierarchic finite-element families, i.e. p ¼ 3 and 4). The temperature
distribution through the thickness is linear in the case of homogeneous plates. Since we assume that the
thermal conductivity varies according the power-law distribution, by changing the exponent parameter n
we can study the effects of temperature deviation from a linear distribution. One can observe that the

Deflection with mechanical load


0

-0.1

-0.2
Non-dimensional deflection

-0.3

-0.4

-0.5

ceramic
-0.6 n=0.5
n=1
n=2
metal
-0.7
-14 -12 -10 -8 -6 -4 -2 0
Load parameter P

Fig. 4. Midpoint deflection versus load intensity P (mechanical load).


718 L. Della Croce, P. Venini / Comput. Methods Appl. Mech. Engrg. 193 (2004) 705–725

Square plate under uniform load and temperature field (p=3)


0.8
metal
Non-dimensional deflectio n

ceramic
0.6 n=0.5
n=1
n=2
0.4

0.2

-5 -4.5 -4 -3.5 -3 -2.5 -2 -1.5 -1 -0.5 0


Non-dimensional load parameter

Square plate under uniform load and temperature field (p=4)


0.8
Non-dimensional deflection w/h

ceramic
metal
0.6 n=0.5
n=1
n=2
0.4

0.2

-5 -4.5 -4 -3.5 -3 -2.5 -2 -1.5 -1 -0.5 0


Non-dimensional load parameter

Fig. 5. Midpoint deflection versus load intensity P (thermomechanical load).

response of graded plates is not intermediate to the metal and ceramic plates. The center deflections of both
the metallic and the ceramic plates are higher than those of the graded plates. Furthermore, the deflection
becomes negative as the mechanical load increases. It is worth noting that Figs. 4 and 5 are in excellent
agreement with the results presented in [23] (see Figs. 4 and 5 therein), even though the adopted mechanical
models are slightly different since von K arman nonlinearities are included in [23].
A further issue that deserves attention is the one of through-the-thickness stress distribution that is
obviously influenced by the temperature profile and the consequent moduli profile. At this regard, for the
thermomechanical load introduced previously, Fig. 6 shows the variation of rxx with respect to the
thickness space variable for different values of the mixture parameter n. The results are obtained with Q3
elements and quite similar ones were obtained using Q2 and Q4 elements. Fig. 7 represents the counterpart
of Fig. 6 in the case of a purely mechanical load. The results are in excellent agreement with those presented
in [23], except for a single stress profile in thermomechanical case that is to be attributed to the different
kinematic assumptions used (von K arman in [23] and Reissner–Mindlin herein).

6.2. Convergence, stability and locking analysis of the proposed finite-element scheme

In Fig. 8 the effect of constituent volume fraction on the center deflection is presented by varying the
volume fraction exponent n. The plate is subject to purely mechanical normal load of intensity q ¼ 106 Pa
L. Della Croce, P. Venini / Comput. Methods Appl. Mech. Engrg. 193 (2004) 705–725 719

0.5
ceramic
n=1
0.4 n=2
n=0.5
metal
0.3

0.2
Non-dimensional thickness

0.1

-0.1

-0.2

-0.3

-0.4

-0.5
-1.5 -1 -0.5 0
Non-dimensional axial stress

Fig. 6. Stress rxx versus thickness space variable (thermomechanical load).

ceramic
n=1
0.5
n=2
n=0.5
0.4 metal

0.3
Non-dimensional thickness

0.2

0.1

-0.1

-0.2

-0.3

-0.4

-0.5

-0.3 -0.25 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2
Non-dimensional axial stress

Fig. 7. Stress rxx versus thickness space variable (mechanical load).


720 L. Della Croce, P. Venini / Comput. Methods Appl. Mech. Engrg. 193 (2004) 705–725

Deflection with power - law index– uniformly distributed load


0.54

0.52

0.5
Non-dimensional deflection w* Et * h3/ (q *a4)

0.48

0.46

0.44

0.42

0.4

0.38

0.36 element Q4
element Q3
element Q2
0.34
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Power-law index n

Fig. 8. Midpoint deflection versus power-law index n for various finite elements (mechanical load).

Deflection with power-law index-mechanical and thermal load


0.5

0.45
Non-dimensional deflection w* Et * h3/ (q *a4)

0.4

0.35

0.3

0.25

0.2

0.15
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Power-law index n

Fig. 9. Midpoint deflection versus power-law index n (thermomechanical load).

and the thickness is t ¼ 0:01 m. Comparing the behavior of the elements Q2 , Q3 and Q4 one can see that
insignificant differences occur when Q3 and Q4 are used, suggesting that Q3 elements suffice for the scope of
L. Della Croce, P. Venini / Comput. Methods Appl. Mech. Engrg. 193 (2004) 705–725 721

Displacement error vs. d.o.f. for p=2, 3, 4


0.2
p=2
p=3
0.18 p=4

0.16

0.14
displacement error

0.12

0.1

0.08

0.06

0.04

0.02

0
0 100 200 300 400 500 600 700 800
degrees of freedom

Fig. 10. Displacement error versus FEM degrees-of-freedom.

our analysis. To gain insight into the capabilities of FGMs, a thermomechanical load is then applied to the
plate for various exponents n. Fig. 9 shows that n  0:7 is in a certain sense an optimal choice in that it
minimizes the deflection. These considerations open the way to the issue of optimal design of FGM plates
that shall be tackled in forthcoming contributions.
To further assess the performance of the presented elements, Fig. 10 presents the relative displacement
error versus the number of degrees-of-freedom in the presence of mechanical and thermal loads as well. We
report the case in which n ¼ 1 and t=a ¼ 1=20. Numerical results indicate that for p ¼ 4 a good accuracy
can be reached even for very coarse meshes. Conversely note that in the case of coarse meshes the Q2
element outperforms the Q3 one. This does not come as a (full) surprise if one recalls that standard dis-
placement error estimates provide a convergence ratio for given finite elements and allow to compare
different finite elements only asymptotically. This is fully confirmed by results showcased in Fig. 7 that
enhance a faster convergence rate of Q3 elements with respect to Q2 ones, even though the error itself may
be smaller for Q2 elements in the case of coarse meshes.
To test the ability of the proposed finite elements to overcome shear-locking, we present in Fig. 11 the
non-dimensional center deflection for various side-to-thickness ratios. First a purely mechanical uniform
normal load is applied ()106 N/m2 ). It is known [32] that for a homogeneous simply supported uniformly
loaded plate the central displacement is given as w ¼ 0:00406  qa4 =D. For the ceramic plate this
amounts to w ¼ 0:00046978  106 =t3 and for the metal plate one has w ¼ 0:000101341  106 =t3 . Solu-
tions computed with locking-free approximation schemes do not deteriorate for decreasing thicknesses
3
and wðt2 Þ=wðt1 Þ / ðt1 =t2 Þ . Unfortunately the numerical approximation of Reissner–Mindlin model leads
often to solutions that are quite sensitive to the plate thickness. Our numerical results indicate that the
presented finite elements are able to handle the locking phenomenon, and in fact the ratio ðt1 =t2 Þ3 is
preserved down to thin plates (t=a ¼ 1=200). The very thin plate behavior is numerically shown to ap-
proach the homogeneous plate one as the intuit seems to suggest. Similar considerations hold true when
722 L. Della Croce, P. Venini / Comput. Methods Appl. Mech. Engrg. 193 (2004) 705–725

Deflection versus side–to–thickness ratio–Purely mechanical load


0

-0.2

-0.4
deflection w

-0.6

-0.8

-1

ceramic
-1.2 n=0.5
n=1
n=2
metal
-1.4
0 20 40 60 80 100 120 140 160 180 200
relative thickness a/ t

Fig. 11. Deflection versus side-to-thickness ratio (purely mechanical loads).

Deflection with side to thickness ratio − Thermal and mechanical load


700
ceramic
n=0.5
n=1
600 n=2

500
Non-dimensional deflection w

400

300

200

100

–100
0 20 40 60 80 100 120 140 160 180 200
relative thickness a/t

Fig. 12. Deflection versus side-to-thickness ratio (thermal and mechanical loads).
L. Della Croce, P. Venini / Comput. Methods Appl. Mech. Engrg. 193 (2004) 705–725 723

-4 displacement for decreasing t – Q1 -4 displacement for decreasing t – Q2


x 10 x 10
0 1
non-dimensional displacement

non-dimensional displacement
h=1/3 h=1/3
h=1/4 0 h=1/4
h=1/8 h=1/8
-2 exact solution -1 exact solution

-2

-4 -3

-4

-6 -5
101 102 103 104 10
1
10
2
10
3
10
4

a/t a/t

-4 displacement for decreasing t – Q3 -4 displacement for decreasing t – Q4


x 10 x 10
0 -4.6
non-dimensional displacement

h=1/3 non-dimensional displacement


h=1/4 -4.65
h=1/8
-2 exact solution -4.7

-4.75

-4 -4.8 h=1/3
h=1/4
-4.85 h=1/8
exact solution
-6 1 2 3 4 -4.9 1 2 3 4
10 10 10 10 10 10 10 10
a/t a/t

Fig. 13. Midpoint displacement versus thickness for various mesh sizes and finite-elements order.

the temperature field is imposed as well, as reported in Fig. 12 that shows the non-dimensional deflection
w=t as well.
To show the stability of the proposed finite-element scheme, the midpoint displacement w  versus the
side-to-thickness ratio a=t for different values of mesh size h is showcased in Fig. 13 in the case of
homogeneous plate. The figure allows to understand the stable behavior of the proposed finite-element
solution with respect to the (decreasing) thickness. For increasing element-degree, one may see that the
stability improves substantially. Eventually, in Fig. 14 the non-dimensional deflection w  Et t3 =106 is
presented with respect to the aspect ratio a=t for different values of volume fraction.

7. Conclusions

A hierarchic family of finite elements has been presented for the analysis of Reissner–Mindlin func-
tionally graded plates based on an ad hoc variational formulation that has been derived on the purpose.
From a physical point of view, the properties of plates with various gradings have been assessed and
compared to the one of single-constituent homogeneous plates. The response of functionally graded plates
724 L. Della Croce, P. Venini / Comput. Methods Appl. Mech. Engrg. 193 (2004) 705–725

x 10
-4 Deflection versus side–to–thickness ratio–Purely mechanical load
-3

Non-dimensional deflection w * (10 6 *t 3) -4

-5

-6

-7

-8

-9

ceramic
-10 n=0.5
n=1
n=2
metal
-11
0 50 100 150
relative thickness a/t

Fig. 14. Deflection versus side-to-thickness ratio (purely mechanical loads).

to mechanical loads is shown to lie in between the one of single-constituent homogeneous plates whereas
this does not hold true when a thermal load is applied as well.
From a numerical viewpoint, convergence and stability of the proposed family of hierarchic elements
have been assessed by means of extensive numerical simulations. Furthermore, the absence of shear-locking
for (almost) vanishing thicknesses has been demonstrated as a peculiar feature of the presented approxi-
mation scheme.
Ongoing extensions on this topic include applications to higher-order plate models and possibly non-
linear and cracked plates. From a modelling viewpoint, we also plan to adopt more accurate descriptions
of the mixture following e.g. [8,28].

References

[1] M. Abid Mian, A.J.M. Spencer, Exact solutions for functionally graded and laminated elastic materials, J. Mech. Phys. Solids 46
(12) (1998) 2283–2295.
[2] S. Amada, T. Munekata, Y. Nagase, Y. Ichikawa, A. Kirigai, Z.F. Yang, The mechanical structures of bamboo in viewpoint of
functionally gradient and composite materials, J. Compos. Mater. 30 (1996) 800–819.
[3] I. Babuska, The p and h-p versions of the finite element method. The state of the art, in: D.L. Dwoyer, M.Y. Hussaini, R.G. Voigt
(Eds.), Finite Elements, Theory and Application, Springer-Verlag, New York, 1988, pp. 199–239.
[4] K.-J. Bathe, Finite Element Procedures, Prentice Hall, Englewood Cliffs, New Jersey, 1996.
[5] T. Belytschko, W.K. Liu, B. Moran, Nonlinear Finite Elements for Continua and Structures, Wiley, Chichester, 2000.
[6] L. Della Croce, T. Scapolla, Hierarchic finite elements for moderately thick to very thin plates, Comput. Mech. 10 (1992) 263–279.
[7] F. Erdogan, Fracture mechanics of functionally graded materials, Compos. Engrg. 5 (1995) 753–770.
[8] M. Grujicic, Y. Zhang, Derivation of effective elastic properties of two-phase materials using Voronoi cell finite element method,
Mater. Sci. Engrg. A 251 (1998) 64–76.
[9] R. Javaheri, M.R. Eslami, Buckling of functionally graded plates under in-plane compressive loading, Z. Angew. Math. Mech. 82
(2002) 77–283.
L. Della Croce, P. Venini / Comput. Methods Appl. Mech. Engrg. 193 (2004) 705–725 725

[10] X.Q. He, T.Y. Ng, S. Sivashanker, K.M. Liew, Active control of FGM plates with integrated piezoelectric sensors and actuators,
Int. J. Solids Struct. 38 (2001) 1641–1655.
[11] X.Q. He, K.M. Liew, T.Y. Ng, S. Sivashanker, A FEM model for the active control of curved FGM shells using piezoelectric
sensor/actuator layers, Int. J. Numer. Methods Engrg. 54 (2002) 853–870.
[12] J.H. Jin, R.C. Batra, Some basic fracture mechanics concepts in functionally graded materials, J. Mech. Phys. Solids 44 (1996)
1221–1235.
[13] J.H. Kim, G.H. Paulino, Mixed-mode fracture of orthotropic functionally graded materials using the finite element method,
Engrg. Fract. Mech. 69 (14–16) (2002) 1769–1790.
[14] J.H. Kim, G.H. Paulino, Finite element evaluation of mixed-mode stress intensity factor in functionally graded materials, Int.
J. Numer. Methods Engrg. 53 (8) (2002) 1903–1935.
[15] M. Koizumi, The concept of FGM, Ceram. Trans., Funct. Graded Mater. 34 (1993) 3–10.
[16] K.M. Liew, X.Q. He, T.Y. Ng, S. Sivashanker, Active control of FGM plates subjected to a temperature gradient: modelling via
finite element method based on FSDT, Int. J. Numer. Methods Engrg. 52 (2001) 1253–1271.
[17] K.M. Liew, X.Q. He, T.Y. Ng, S. Kitipornchai, Active control of FGM shells subjected to a temperature gradient via piezoelectric
sensor/actuator patches, Int. J. Numer. Methods Engrg. 55 (2002) 653–668.
[18] K.M. Liew, H.K. Lim, M.J. Tand, X.Q. He, Analysis of laminated composite beams with piezoelectric patches using the element-
free Galerkin method, Comput. Mech. 29 (2002) 486–497.
[19] R.D. Mindlin, Influence of rotary inertia and shear on flexural motion of isotropic, elastic plates, J. Appl. Mech. 18 (1951) 31–38.
[20] T.Y. Ng, K.Y. Lam, K.M. Liew, Effects of FGM materials on the parametric resonance of plate structures, Comput. Methods
Appl. Mech. Engrg. 190 (2000) 953–962.
[21] T.Y. Ng, X.Q. He, K.M. Liew, Finite element modeling of active control of functionally graded shells in frequency domain via
piezoelectric sensors and actuators, Comput. Mech. 288 (2002) 1–9.
[22] N.D. Phan, J.N. Reddy, Analysis of laminated composite plates using a higher-order shear deformation theory, Int. J. Numer.
Methods Engrg. 21 (1985) 2201–2219.
[23] G.N. Praveen, J.N. Reddy, Nonlinear transient thermoelastic analysis of functionally graded ceramic-metal plates, Int. J. Solids
Struct. 35 (1998) 4457–4476.
[24] J.N. Reddy, Mechanics of Laminated Composite Plates: Theory and Analysis, CRC Press, Boca Raton, 1997.
[25] J.N. Reddy, C.D. Chin, Thermomechanical analysis of functionally graded cylinders and plates, J. Thermal Stresses 26 (1) (1998)
93–126.
[26] J.N. Reddy, Analysis of functionally graded plates, Int. J. Numer. Methods Engrg. 47 (2000) 663–684.
[27] E. Reissner, The effect of transverse shear deformation on the bending of elastic plates, J. Appl. Mech. 23 (1945) 69–77.
[28] T. Reiter, G.J. Dvorak, V. Tvergaard, Micromechanical models for graded composite materials, J. Mech. Phys. Solids 45 (8)
(1997) 1281–1302.
[29] T. Reiter, G.J. Dvorak, Micromechanical models for graded composite materials: II. Thermomechanical loading, J. Mech. Phys.
Solids 46 (9) (1998) 1655–1673.
[30] O. Sigmund, A new class of extremal composites, J. Mech. Phys. Solids 48 (2) (1998) 397–428.
[31] S. Suresh, A. Mortensen, Fundamentals of Functionally Graded Materials, Barnes and Noble Publications, 1998.
[32] S. Timoshenko, S. Woinowski–Krieger, Theory of Plates and Shells, McGraw Hill, New York, 1970.
[33] K. Tanaka, Y. Tanaka, H. Watanabe, V.F. Poterasu, Y. Sugano, An improved solution to thermoelastic materials designed in
functionally graded materials: scheme to reduce thermal stresses, Comput. Methods Appl. Mech. Engrg. 106 (1993) 377–389.
[34] J. Woo, S.A. Meguid, Nonlinear analysis of functionally graded plates and shallow shells, Int. J. Solids Struct. 38 (2001) 7409–
7421.

You might also like