Evaluation and verification of an HSDT-Layerwise

Generalized Finite Element formulation for adaptive

piezoelectric laminated plates
Diego Amadeu F. Torres1,∗, Paulo de Tarso R. Mendonça∗, Clovis S. de Barcellos∗
Group of Mechanical Analysis and Design - GRANTE, Department of Mechanical Engineering,
Federal University of Santa Catarina, 88040-900,
phone: (55)(48)3721-9899, Florianópolis, SC, Brazil.

Abstract

A formulation for the bending analysis of composite laminated plates with piezo-
electric layers is implemented using the Generalized Finite Element Method. This
formulation is derived from a mechanical description based on Higher-Order Shear
Deformation Theory which allows for the use of C 0 continuous approximation func-
tions on the domain. On the other hand, a Layerwise Theory is employed for in-
terpolation of electric potential across the thickness of piezoelectric layers, in such a
way that the kinematical hypotheses result in a mixed model. The paper presents
an analysis of the approximation capability of the proposed numerical model for
static analysis, using C 0 continuous Partition of Unity and polynomial enrichments
to span the approximation spaces, by assessment of convergence. Analytical solu-
tions obtained from the same kinematical hypotheses are used as references. Results
for relative error in the energy norm considering p- and h- refinements for regular and

∗
Corresponding author
Email addresses: diego.amadeu@gmail.com (Diego Amadeu F. Torres),
mendonca@grante.ufsc.br (Paulo de Tarso R. Mendonça)
1
Graduate student

Preprint submitted to Computer Methods in Applied Mechanics and EngineeringNovember 29, 2011
distorted meshes, in addition to a pointwise evaluation of the stresses and electric
field, are presented. The evaluations show that the numerical methodology is a very
effective tool for improving the solution through the enrichment, even for pointwise
values across the thickness, and is robust to mesh distortions. Moreover, the results
furnish insight about the physical modeling for both active and sensory modes, for
thick and thin plates.
Keywords: GFEM, Adaptive plates, Mixed HSDT-Layerwise model, Convergence
analysis
PACS:
2000 MSC:

1. Introduction

Smart/intelligent structures have received increasing attention from researchers

during recent decades due to several factors. The motivation behind the use of adap-
tive materials is to enable a structure to change its shape or its material/structural
properties, thereby improving performance and service life.
Electrostrictive materials, magnetostrictive materials, shape memory alloys, mag-
neto or electro-rheological fluids, polymer gels, and piezoelectric materials, for ex-
ample, can all be used to design and develop structures that can be called smart.
However, these materials themselves are not smart. “Smartness” refers to the ex-
ploitation of material properties to better serve a design function than would be
possible through conventional structural design [1].
Piezoelectric materials, among several applications, are easily integrated to struc-
tural components, when compared to adaptive materials of other categories. Because
they exhibit coupled mechanical-electrical behavior, they can be used as sensors, to
measure strains and acceleration, and as actuators, allowing the application of elec-

2
tric potential to generate a field of deformations in the structure. In this regard,
they can be applied in laminated composite structures as patches and films or in the
form of layers within a laminate.
During the last three decades a large variety of models have been developed to
predict the behavior of piezoelectric materials in smart structures. These models
may be classified into three different categories: induced strain models, coupled
electromechanical models and coupled thermo-electromechanical models.
The induced strain models use theoretical approaches to consider the piezoelectric
effect and they are usually limited to predicting only the active response of piezoelec-
tric materials. In this case, the strain in the piezoelectric materials due to the applied
electric potential is approximated by statically equivalent forces and moments. Even
though some models have been developed with the introduction of the piezoelectric
constitutive equations, allowing the sensitive behavior to be represented, the electric
potential is usually not included as a state variable and is not considered in the con-
servation of electric flux in the displacement equations, and thus the electric voltages
generated by the sensors are post-processed using the electric charge equation [2].
On the other hand, coupled electromechanical models provide a more consis-
tent representation of both active and sensitive responses of piezoelectric materials
through the incorporation of both mechanical displacements and electric potentials
as state variables in the formulation. The coupled electromechanical models are the
models most commonly implemented in finite element codes.
Prior to the 1990s, structures with piezoelectric devices were modeled with solid
elements, following the work of Allik and Hughes [3] who formulated the dynamic
equations through variational methods with the piezoelectric constitutive relations
and finite element models for piezoelectric solids. However, the hexahedron or brick
elements display excessive shear stiffness as the element thickness decreases. This

3
problem was circumvented by adding three incompatible internal degrees of freedom
to the element [4]. Another approach based on solid elements was presented in
[5]. Nevertheless, as a rule, the completely three-dimensional modeling of laminated
piezoelectric structures results in systems with a large number of degrees of freedom,
since they require one solid element per layer of the laminate.
In addition, most of the previous theories use simplifying approximations at-
tempting to replicate the induced strain or electric fields generated by a piezoelectric
layer under an external electric field or applied load.
Due the inconveniences related to fully three-dimensional modeling, the develop-
ing of two-dimensional models like plates and shells became a tendency. Hwang and
Park [6] presented a two-dimensional quadrilateral plate element considering actua-
tion by induced strain and kinematical description according the Classical Laminated
Plate Theory (CLPT). Since this finite element formulation was derived considering
one electrical degree of freedom, the output voltage is dependent on the integrated
strain rates over the element and was calculated from the direct piezoelectric equa-
tion. When used as an actuator, the piezoelectric layer induces the control moments
at the ends of the element.
Some models based on the First-Order Shear Deformation Theory (FSDT) were
tested, like the ones presented in Detwiler et al. [7], for the linear response of
coupled electrical-mechanical behavior, and [8], which considered the geometrical
non-linearity of the structure. But, remarkably, the simple CLPT offers poor kine-
matic approximation capability for complex laminate systems like those equipped
with piezoelectric patches.
One strategy to reduce the costs associated with full solid element models consists
of the use of a Layerwise Theory (LT) to describe both mechanical and electric
unknowns. Among others authors, Saravanos et al. [9] studied the complete dynamic

4
electromechanical response of smart piezoelectric plates under external mechanical
or electric loading. Lee [2] presents a complete family of finite elements for beams,
plates and shells based on LT for all state variables.
The Layerwise Theories still lead to a large number of degrees of freedom in the
model when they are used for both mechanical and electrical variables, to similar
extent as the full solid elements.
The particular configuration of a laminated structure with piezoelectric layers or
patches makes it very difficult to adequately model with a single equivalent layer
kinematic model, applied to both mechanical and electric unknowns. Therefore, a
very appropriate strategy consists of using a single equivalent layer model for the
mechanical unknowns and a Layerwise Theory for the electric potential. Some of the
most simple of these combinations are those using the FSDT for the mechanical dis-
placements and LT for the potential, like the shell element formulated by Saravanos
[10]. Cen et al. [11] have also employed a partially hybrid energy functional for
correcting the transverse shear deformation in their mixed FSDT-LT finite element
formulation. In contrast, Liew et al. [12] used the mixed FSDT-LT models in a
formulation based on the Element-Free Galerkin Method.
It is well known that the FSDT also results in some deficiencies in the approx-
imation of the mechanical response of anisotropic laminates, particularly in under-
estimating displacements and providing poor transverse shear stress results. The
Higher-Order Shear Deformation Theories (HSDT), for example [13] [14] [15], be-
have much better and several researches indicate that they represent a good choice
for the mechanical displacements in an piezoelectric laminate, combined with the LT
for the electric potential. This can be seen in the work of Reddy [16] [17], Chee [18]
and Faria [19], among others.
Firstly, [16] presented a hybrid kinematical hypothesis, and corresponding ana-

5
lytical solutions, based on an equivalent single layer theory, Reddy’s Higher-Order
Shear Deformation Theory (HSDT) [13], for the mechanical displacements and layer-
wise discretization of the electric potential. [17] also derived a theoretical formulation
of the Navier solutions of simply supported rectangular laminates and displacement
finite element models for laminated composite plates with integrated sensors and
actuators. The CLPT, Higher-Order Shear Deformation Theory [13] and thermo-
electric mechanical coupling were also considered using piecewise linear functions
across the thickness to approximate temperature and electric fields.
It should be noted that Reddy’s HSDT essentially generates a C 1 -continuous
formulation, but most of the finite element formulations are based on C 0 -continuous
shape functions and therefore requires some especial treatment.
Following this tendency, [18] and [19] also implemented finite elements based on
HSDT [15] and LT, using C 0 -continuous shape functions. A similar approach was
used in [20], which also presents a strong formulation for the problem, and performed
some optimization analysis of piezoelectric structural configurations.
A more elaborate kinematical hypothesis is proposed in [21], who presented a
finite element formulation based on a fourth-order expansion through the laminate
thickness, combined with a piecewise linear term to describe the mechanical variables
and a quadratic distribution of the electric potential inside each piezoelectric layer.
On the other hand, a fully layerwise description was implemented in [22] using
Lagrange polynomials within each layer. Although traditional layerwise theories are
able to accurately predict stress, electrical displacements, mechanical displacements
and electric potential, they are unable to guarantee the continuity of transverse
stresses and transverse electrical displacement (when there are stacked piezoelectric
layers). Following [23], who used a Reissner variational principle with a layerwise
description, [24] developed a layerwise formulation for piezolaminated plates based

6
on a mixed variational principle, in such a way that interfacial conditions of stress
and electrical displacement are directly imposed. [25] extended the formulation of
[24] to incorporate the magneto-electro-elastic phenomenon also using a partially hy-
brid functional with higher-order functions across the thickness based on a complete
layerwise description.
The present paper develops a procedure to numerically analyze the coupled
electro-structural response of laminated plates with orthotropic fiber reinforced lay-
ers and piezoelectric layers, using the Generalized Finite Element Method (GFEM).
The mechanical unknowns, the displacements, are modeled by a higher-order shear
deformation theory (HSDT) of the third-order [14] which allows for the use of C 0 -
continuous approximation functions. The electric unknowns, the potentials, are mod-
eled by a Layerwise Theory, utilizing piecewise linear functions across the thickness
of piezoelectric layers. All fields are enriched according to the Generalized Finite
Element Method, utilizing polynomial enrichment functions applied to a bilinear
Partition of Unity defined on each element and the main focus is to verify the perfor-
mance of the numerical method. The remainder of this paper is outlined as follow:
Section 2 provides a review of the GFEM basics. Section 3 deals with the lin-
ear electro-elasticity formulation. Section 4 details the proposed application of the
GFEM strategy to the problem of laminated plates with orthotropic piezoelectric lay-
ers, showing the discretization, the global enrichment of the approximating spaces,
stiffness and inertia matrices, mechanical-electrical coupling matrices and mechani-
cal and electric dynamic load vectors. Section 5 presents a strong formulation for the
problem under analysis, from which analytical solutions may be obtained. Section 6
shows numerical applications for static analysis of laminated plates with piezoelectric
layers considering both active and sensory modes, presents a convergence analysis
using a global measure of the relative error in energy norm and pointwise assess-

7
ment of the stress and electric field for comparison with the analytical solutions, and
Section 7 summarizes some conclusions.

2. Generalized Finite Element Method

A hybrid method combining the hp Clouds Method and the conventional form of
the Finite Element Method was presented by [26]. This strategy takes into account
the idea of adding hierarchical refinements to a set of shape functions associated with
finite elements, such as the Lagrangian interpolations, for example, which satisfy the
requirement of a Partition of Unity (PoU). This procedure allows the construction
of richer ansatz spaces for approximate solutions of partial differential equations
(PDEs).
The resultant methodology, established in [27], [28], and [29], called Generalized
Finite Element Method (GFEM) could be mentioned in the context of the Meshfree
Methods.
In Meshfree methods, the approximation of field variables is constructed in terms
of nodes without the aid of a mesh. Due to the flexibility in constructing conforming
shape functions to meet specific needs for different applications, Meshfree methods
are particularly suitable for hp-adaptivity, simulation of crack propagation and large
deformation problems, among others [30]. However, they require the partition of the
domain through the use of a “background grid” for domain integration. Hence, one of
the main drawbacks of Meshfree methods is that the computational cost is too high in
some applications due to the fact that one has to use a great number of integration
points in order to perform the domain integration accurately [31]. Moreover, it
must be guaranteed that the weighted moment matrix derived in the construction of
approximation functions by the Moving Least Square Method is invertible, which is
related to the nodal density [32].

8
 In the GFEM context, discretization spaces for a Galerkin method are defined
using the concept of the Partition of Unit Finite Element Method (PUFEM), [33],
[34], [35], [36], and of the hp Clouds Method, [37] and [38].
The GFEM is one of the instances of the Partition of Unity Method and allows
a simple and effective scheme for tailoring the approximation functions for each
kind of problem. In this context, a mesh of elements is created to perform the
following operations: (a) to facilitate the numerical integration; and (b) to define
the PoU locally within elements by intrinsic coordinates. Finally, the PoU functions
are externally enriched by functions defined in global coordinates. This last aspect
is responsible for the high efficiency of the method.
Usually the enrichment functions are polynomials, but special enrichment func-
tions can be used to provide more accurate and robust simulations since, through
this framework, it is possible to use compactly supported non-polynomial bases in
the solution space. These functions can be built based on a-priori knowledge regard-
ing the analytical expressions which reflect the nature of the solution of a problem,
like crack modeling [39], as well as based on the solution of local boundary value
problems, for instance, in global-local strategies [40] or the mesh-based handbook
function generated from canonical domains containing microstructural features [41].
In recent years, the GFEM has been applied to a diversity of phenomena such as
the analysis of dynamic crack propagation [42] and materials with internal cracks [29]
and voids [41]. It has also being used to build arbitrarily smooth approximations for
handling higher-order distributional boundary conditions [30]. [43] uses it to perform
a posteriori error estimation and p-adaptive analysis using C k base functions. [44]
address the a posteriori error estimation using mesh-based handbook functions as
enrichments for materials with many voids. [45] and [40] use a global-local strategy
in GFEM for heat transfer with high gradients. [46] develops a C k continuous finite

9
element formulation for the Kirchhoff laminate model. A procedure to define richer
approximate subspaces for shell structures and the treatment for boundary layer
phenomenon is addressed by [47].
Partition of Unity (PoU) is a set of functions for which the sum of their values
is equal to unity on every point of the support. Besides this aspect, three other
characteristics are used to verify the applicability of this definition to a set of basis
functions. According to [48], over an open bounded domain Ω in Rn , with a countable
open covering constructed with a countable collection of open sets {Gj } such that
Ω⊂ N
S
j=1 Gj , a collection of functions ϕj (x) forms a Partition of Unity if it has the

following properties:

• ϕj (x) ∈ C0∞ (Gj );

• ΣN
j=1 ϕj (x) = 1;

• ϕj (x) ≥ 0 em Ω;

• every compact subset of Ω intersects only a finite number of the supports of

the ϕj (x).

The first one is usually relaxed to C0k (Gj ) for some k ≥ 0.

It should be noted that the standard finite element shape functions form a Par-
tition of Unity. In general, the GFEM can be briefly defined as a strategy to enlarge
the FEM approximation space by adding special functions to the conventional ap-
proximation basis. This basis now takes the role of a Partition of Unity and allows
inter-element continuity and creates conforming approximations. The enrichment
allows the application of any information that reflect previous knowledge on the
boundary value problem solution, such as a singular function resulting from local

10
asymptotic expansion of the exact solution close to a point. The approximating
capabilities of the enrichment functions are included in the function space of the
method while keeping the same standard structure of an FEM code.
For instance, the GFEM local approximation capability can be described as fol-
eij
lows. Let a function u be defined over the domain Ω covered by the clouds. Let u
be a local approximation of u which belongs to a local subspace Xj (ωj ) defined over
the cloud suport ωj , such that Xj (ωj ) = span{Lji }i∈(j) , where (j), j = 1, 2, . . . , N ,
is the index set which refers to the enrichment functions associated with each node,
Lji denotes an enrichment function i related to the node xj and N is the number of
nodes of the connected domain.
Basically, GFEM proposes that each subspace Xj (ωj ) may be chosen in such
a way that Lji ∈ Xj (ωj ) can closely approximate u|ωj over the cloud ωj , without
compromising the conforming requirement.
The cloud functions family FNk,p is composed of the PoU, which generates the
polynomial space of degree k, Pk , and is endowed with the capability of representing
exactly the polynomials of the space of degree p, Pp . It is defined by

n [ o
FNk,p = {ϕj (x)}N
j=1 {ϕ j (x)L ji (x)}N
j=1 |i ∈ (j) (1)

where ϕj (x) are PoU functions and Lji (x) are the enrichment functions, both related
to the node j, and (j) is an index set which refers to the enrichment functions
associated with each node. Presently
 


 ϕ1 1 ϕ2 1 ··· ϕN 1 



 

ϕ1 L11 ϕ2 L21 · · · ϕN LN 1

 

FNk,p = .. .. .. .. (2)


 . . . . 



 

 ϕL
1 1M ϕ2 L2M · · · ϕN LN M 
 

11
where N is the number of clouds and p is the dimension of the highest complete
polynomial space spanned by FNk,p and M is the number of enrichment functions
related to the nodes.
This cloud family is used to build the following approximation

qj
N
( )
X X
u
e(x) = ϕj (x) uj + Lji (x)bji = ΦT U (3)
j=1 i=1

with

h
UT (x) = u1 b11 · · · b1qj · · ·
i (4)
··· u N bN 1 · · · bN qj

where uj and bji are the nodal parameters associated with the standard finite ele-
ment shape functions ϕj (x), and to the enriched functions ϕj (x)Lji (x), respectively,
grouped in the vector

h
ΦT = ϕ1 L11 ϕ1 · · · L1qj ϕ1 · · ·
i (5)
ϕN LN 1 ϕN · · · LN qj ϕN

where qj is the number of enrichment functions of each node.

Let U and V ∈ H1 (Ω) be Hilbert spaces of degree 1, which are standard Sobolev
spaces of square integrable functions whose first derivatives are also square integrable
(in the sense of Lebesgue), and which are defined on the domain Ω. Consider the
boundary value problem of finding u ∈ U such that B(u, v) = L (v), ∀ v ∈ V .
Now let Uh be the subspace spanned by a set of kinematically admissible func-
tions and Vh the subspace spanned by a set of kinematically admissible variations.
We define the following Galerkin approximation, in the GFEM approach, for the

12
 e ∈ Uh such that B(e
boundary value problem, which consists of: find u u, ve) = L (e
v ),
∀ ve ∈ Vh , where u
e and ve ∈ Uh = Vh ⊂ H1 , B(•, •) is a bilinear form of H1 ×H1 −→ R
and L (•) a linear form of H1 −→ R, leading to the discrete form B(ΦT U, ΦT V) =
L (ΦT V) where

h
VT = v1 c11 · · · c1qj · · ·
i (6)
vN cN 1 · · · cN qj

are nodal parameters related to the test function, such that ΦT V ∈ Vh .

In the present work, the PoU is defined by bilinear Lagrangian shape functions
and, without loss of generality, only polynomial enrichment up to the third degree is
used, according to the following linear combinations

(  2
x − xj y − y j x − xj
ϕj × 1, , , ,
hxj hyj hxj
    3 ) (7)
x − xj y − yj y − yj
,...,
hxj hyj hxj

where ϕj , j = 1, 2, . . . , N , are standard FEM bilinear shape functions, xj = (xj , yj )

are nodal coordinates of an arbitrary node j, hxj and hyj are the cloud charac-
teristic dimensions of the node in the directions x and y, respectively, and N is
the number of nodes of the finite element mesh. Consequently, k = 1, p = 4 and
(j) = {1, 2, · · · , 10}.
It can be noted that by using this enrichment procedure only the PoU satisfies
the Kronecker delta condition, i.e., ϕj (xi ) = δij , since the enriched functions asso-
ciated with a given node are zero at this node. Therefore, the Dirichlet boundary
conditions cannot be directly imposed, a special procedure being required to impose

13
them. For instance, one efficient procedure to impose essential boundary conditions
is through the so-called boundary functions, as proposed in [47], which allows an
adequate enrichment in the boundary vicinity while preserving the completeness of
the polynomials that span the approximation subspace.
In spite of this, once a displacement component u is defined over a cloud j as

u = uj ϕj + bj1 (ϕj Lj1 ) + bj2 (ϕj Lj2 ) + · · · + bj9 (ϕj Lj9 ) (8)

it is possible, for boundaries parallel to the axis of the global problem coordinates,
to use a simplified method to enforce essential boundary conditions. For instance,
for nodes on a boundary line x = const. one may enforce the condition u = 0 by just
making the following parameters null

uj = bj2 = bj5 = bj9 = 0 (9)

On the other hand, for a boundary line y = const. and the same condition u = 0,
one must impose

uj = bj1 = bj3 = bj6 = 0 (10)

For non-homogeneous Dirichlet boundary conditions u = u one can impose uj = u

in strong form making all the other nodal parameters null.

3. Constitutive equations

The coupling between the mechanical, thermal and electric fields can be estab-
lished using thermodynamic principles and the Maxwell relations [49]. Analogously
to the deformation energy functional U0 , in the linear elasticity theory, and the

14
Helmholtz free energy functional Ψ0 , in thermoelasticity, the existence of a func-
tional Φ0 it is assumed such that

Φ0 (εij , Ei , φ) = U0 − E · D − ηφ

= 12 Cijkl εij εkl − eijk εij Ek − βij εij φ (11)

ρcv 2
− 21 χkl Ek El − pk Ek φ − 2φ0
φ
denoted the Gibbs free energy functional. Here, η is the enthalpy, Cijkl are the
elastic moduli, eijk are the piezoelectric moduli or, more precisely, the constants of
piezoelectric deformation, χij are the dielectric constants, pk are the pyroelectric
constants, βij are stress-temperature expansion coefficients, cv is the specific heat at
constant volume, per unit mass, φ is the temperature, and φ0 is the absolute reference
temperature. Differentiation of this functional with respect to the fields ε, E and
φ results in the coupled constitutive relations for a deformable pyro-piezoelectric
material, as can be seen in [49].
The formulation derived in the present paper ignores variations in temperature
such that the coupled equations become

σi = CijE εj − eik Ek

Dk = ekj εj + χεkl El (12)

η = βj εj + pk Ek

where σij are the components of the Cauchy stress tensor, Di are the components of
the electric displacement vector and η is the enthalpy. In (12) the contracted notation
was used, taking into account the symmetry of stress and strain tensors, such that
the stress components are organized in vector form as σ = { σx σy σz τyz τxz τxy }T .
The same operation is performed for the deformation components. The enthalpy

15
becomes uncoupled from the other fields, and the solution is obtained from the first
two equations. These relations can be reordered into one single matrix relation,
linear and electromechanically coupled, in the orthotropic material directions, along
axes 1, 2 and 3

 k  k  k
 σ1  1 1T 1 
C −e  ε
=  (13)
 D1  e1 χ1  E1 

where the superscript 1 indicates the material coordinate system and k is the number
of an arbitrary piezoelectric layer. For the piezoelectric extensional mode of actua-
1 1 1 1 1 1
tion, only the following coefficients in (13) are non-zero: C11 , C12 , C13 , C22 , C23 , C33 ,
1 1 1
C44 , C55 and C66 , for the 6 × 6 stiffness matrix C1 ; e115 , e124 , e131 , e132 and e133 , for the
3 × 6 piezoelectric matrix e1 ; and χ111 , χ122 and χ133 for the 3 × 3 dielectric matrix χ1 .
In this formulation, each composite layer is considered to be orthotropic, whether
it is piezoelectric or not. Therefore, the constitutive relation (12) must be rotated to
the global laminate coordinate system, according to the layer orientation angle, and
then it is combined into the laminate constitutive relation. The stress vector after
axes rotation σ x is given by

 T

σ x = T C1 RT−1 R−1 εx − e1 LEx (14)

where the superscript T indicates a transposed matrix. Through an analogous pro-

cedure one obtains the rotated electric displacement vector Dx

 
Dx = LT e1 RT−1 R−1 εx + χ1 LEx (15)

where T, R and L are defined as follows

16
  
c2 s2 0 0 0 −2sc
 
2 2
s c 0 0 0 2sc
 
 
 
 
 0 0 1 0 0 0 
T=


 (16)
 0 0 0 c s 0 
 
−s
 
 0 0 0 c 0 
 
sc −sc 0 0 0 c2 − s 2
where s = sin θ, c = cos θ, with θ being the rotation angle between principal axis 1
and global axis x, in a counter-clockwise rotation around the transverse axis 3.
l k
Moreover, R = diag 1 1 1 2 2 2 . Finally,
 
cos θ sin θ 0
 
L =  − sin θ cos θ 0  (17)
 
 
0 0 1
In addition, for simple monolithic piezoelectric materials polarized along the prin-
cipal transverse direction 3, the piezoelectric properties would be the same in both
the 1 and 2 in-plane directions. The two piezoelectric constants that are usually
tabulated are d31 and d33 (in the strain formulation) - the first subscript indicates
the direction of the electric field and the second subscript the direction of the strain.
It can be shown that the material parameters are interrelated by dmi = eml Cli .

4. GFEM formulation for the problem

In this section, a generalized finite element is implemented to model laminated

plates with piezoelectric sensors and actuators. The domain is partitioned into
quadrilateral elements defined by 8 nodes and corresponding standard biquadratic
Serendipity functions. The Partition of Unity is defined by the four vertice nodes and

17
the corresponding Lagrangian bilinear functions which are, in turn, enriched accord-
ing to the GFEM procedure in order to generate the enriched space of approximation
for the unknown electric and mechanical fields.
The formulation used here is derived from the Hamilton’s Principle, where the
Hamiltonian is built in such a way to be variationally equivalent to the differential
governing equations of the mechanical and electrical responses in an electromechan-
ically coupled continuum [2]. These equations are the Cauchy’s equations of motion,

∂σij
+ fi = ρüi (18)
∂xj
(where the summation convention is used) and the Maxwell’s equations of conserva-
tion of electric flux

∂Di
=Q (19)
∂xi
where σij are Cartesian components of stress tensor, fi are components of body
force per unit volume, ρ is the specific mass per unit volume of the material, ui are
components of mechanical displacement, Di are components of electric displacement
(electric flux) and q is the electric charge.
The functional of the Hamilton’s Principle is [18]

Z t1
(δK − δP + δW ) dt = 0 (20)
t0

which must hold for any t1 > t0 , where K, P and W are the total kinetic energy, the
total deformation energy and the total work of external forces applied to the system,
respectively, and δ is the variation operator. The expression can be expanded as

18
 ( Z ( )T ( )
t1 σx δx
Z Z
T
− ρδu ü(x, t) dV − dV +
t0 V V Dx −δEx
Z Z
+ T
δu b dV +V
δuT f S dS + δuT f P (21)
V S
Z Z )
+ δϕT Q dV − δϕT q dS dt = 0
V S

where u is the vector of mechanical displacement, σ is the Cauchy stress tensor, ε

is the linear strain tensor, D is the electric displacement, E is the electric field, f S is
surface force, f V body force, f P are concentrated forces, ϕ is the electric potential, Q
is a free electric charge and q is a free electric charge at the surface. The components
of the stress and strain tensors appear in the equation organized in vector form.
The mechanical behavior of the plate undergoing bending is modeled by the
Equivalent Single Layer (ESL) methodology, using the kinematical hypothesis fol-
lowing Levinson’s Higher-Order Shear Deformation Theory [14]. Hence, the present
formulation is based on following assumed displacement field

u(x, t) = u0 (x, y, t) + zψx (x, y, t) + z 3 ψ3x (s, y, t)

v(x, t) = v 0 (x, y, t) + zψy (x, y, t) + z 3 ψ3y (x, y, t) (22)

w(x, t) = w0 (x, y, t)

where (u, v, w) are the components, in the Cartesian directions, of the displacement of
an arbitrary point in the laminated plate. This theory is chosen here due its relatively
lower computational cost, since only 7 generalized displacements are required, u0 , v 0 ,
w0 , ψx , ψy , ψ3x and ψ3y . u0 , v 0 and w0 are the displacements of a point in the reference
plane, ψx and ψy are rotations of a segment normal to the middle plane around the

19
y-axis and x-axis, respectively, and ψ3x and ψ3y are higher-order warping variables
in the x − z and y − z-planes, respectively. These are the mechanical unknown fields
that can be approximated over the bi-dimensional (x, y) domain through function
spaces with C 0 continuity.
The choosing of a thin or thick plate model is a decision which cannot be made
a priori because it depends on the solution and the goals of the computation, for
example, the calculation of displacements or stresses. Moreover, according to [50],
hierarchic plate/shell models (HM |i), with order i, have the property

(HM |i)
lim ku3D
EX − uEX kE(Ω) ≤ Cdαi (23)
i→∞

when u3D
EX , the fully three-dimensional solution, is smooth. In (23), C is a constant,

independent of i, αi is a constant which depends on i, and αi+1 > αi . Thus, HSDT

has a higher rate of convergence than FSDT or CLPT.
Using the linear strain-displacement relations it is possible to obtain the strain
field, which is separated into coplanar strains εmf (x, t)

∂u0
 
  
 

ε (x, t) ∂x
  
 

 x
   
∂v 0
 
 
 

εmf (x, t) = εy (x, t) =

 
 
 ∂y 

0 0 
∂u ∂v

 γ (x, t) 
 
 
xy  

 + 

∂y ∂x
    (24)

 ∂ψx 
 
 ∂ψ 3x 

∂x ∂x

 
 
 


 
 
 

 ∂ψy 
3
 ∂ψ 3y

+z +z

 ∂y 
 
 ∂y 


 ∂ψx ∂ψy  
 
 ∂ψ 3x ∂ψ 3y


+ +

  
 

∂y ∂x ∂y ∂x
   

where it is possible to identify the generalized extensional strains, ε0 , the generalized

20
flexural rotations, κ1 , and the generalized warp rotations, κ3 , such that

εmf (x, t) = ε0 (x, y, t) + zκ1 (x, y, t) + z 3 κ3 (x, y, t) (25)

The transverse shear strains, γ c (x, t), are given by

 
 γ (x, t) 
yz
γ c (x, t) =
 γ (x, t) 
xz

∂w0 
 
  (26)
 ψy + ∂y 

   ψ 
3y
= 0 + 3z 2
∂w   ψ 
 ψx +

  3x
∂x


where it is possible identify the generalized shear strains, γ 0 , and the generalized
shear-warp strains, κ2 , such that

γ c (x, t) = γ 0 (x, y, t) + 3z 2 κ2 (x, y, t) (27)

In this stage, it is necessary to define how the electric degrees of freedom are
incorporated. Reddy’s Layerwise Theory [49] is adopted here for interpolation of the
electric potential field. The electric potential in the element, ϕ(x, t), is approximated
by piecewise functions across the piezoelectric layer thicknesses. This hypothesis is
acceptable since the voltage is generally applied parallel to the active layer, assuming
homogeneous material.
According to these laminate theories, each layer is modeled using independent
approximations for the in-plane displacement components and the electrostatic po-
tential, in a unified representation, in accordance with the linear theory of piezoelec-
tricity.

21
 Consider a laminate with npiez piezoelectric layers. To each arbitrary node no
on the laminate middle plane corresponds N = npiez + 1 electric potential nodal
values, ϕ1no to ϕN
no , if such piezoelectric plies are contiguous or N = 2npiez electric

potential nodal values if there is inert material between them. Hence the potential
ekno in an intermediary position z within an arbitrary piezoelectric
approximation ϕ
layer k in the instant t is given by the expression

ekno (z, t) = ϕk−1

ϕ k
no (t) ζk−1 + ϕno (t) ζk (28)

with

   
zk − z z − zk−1
ζk−1 = and ζk = (29)
hk hk
Other important aspect, which is an inherent limitation related to the HSDT
model used here, is that both σ3x and εx3 are assumed to vanish. Hence, similarly to
the correction proposed by [51] for their model based on the FSDT, the imposition
of the condition σ3x = 0, with the elimination of εx3 , leads to a reduced version of the
constitutive relations, where the new coefficients are

22
 (C 13 )2 C 13 C 23
C11 = C 11 − C12 = C 12 −
C 33 C 33
C 13 C 36 (C 23 )2
C16 = C 16 − C22 = C 22 −
C 33 C 33
C 23 C 36 (C 36 )2
C26 = C 26 − C66 = C 66 −
C 33 C 33
C44 = C 44 C45 = C 45 C55 = C 55
(30)

e14 = e14 e15 = e15 e24 = e24 e25 = e25

C 13 e33 C 23 e33
e31 = e31 + e32 = e32 +
C 33 C 33
e33 C 36
e36 = e36 +
C 33

χ11 = χ11 χ12 = χ12 χ22 = χ22

2
(e33 )
χ33 = χ33 +
C 33
where the barred quantities • are the coefficients obtained from the property matrices
after axis rotation.
The GFEM formulation is implemented beginning with the definition of the Par-
tition of Unity (PoU) functions over the element domain. The enrichment is made
adding new parameters linked to unknown nodal values which are associated with the
functions that multiply the original PoU functions. In this way, the generalized me-
chanical displacements over the laminate middle plane, (u0 , v 0 , w0 , ψx , ψy , ψ3x , ψ3y ),
can be approximated, for instance, as

23
  
N ne nf (u0no )
j
X X
e0 =
u Neno (x, y) u0no (t) + u0no (t)fuj0no (x, y)
no=1 j=1
 0 )

N ne nf (vno
j
X X
ve0 = Neno (x, y) vno
0
(t) + 0
vno (t)fvjno
0 (x, y)

no=1 j=1
 0 )

N ne nf (wno
X X
0j
(31)
e0 =
w Neno (x, y) wno
0
(t) + wno (t)fwj no
0 (x, y)

no=1 j=1
..
.
 
N ne nf (ψ3yno )
X X j
ψe3y = Neno (x, y) ψ3yno (t) + ψ3y no
(t)fψj 3yno (x, y)
no=1 j=1

where Neno (x, y) is the portion of the PoU functions matrix related to the node no,
N ne stands for the number of nodes and nf (•no ) denotes the number of enrichment
functions associated with the unknown • at node no. Hence, by assembling all the
functions in a single matrix of displacement approximations, one has a symbolic
representation for the discretized unknown fields, whose matrix Ne has dimensions
7 × 7(N ne + npar), with npar equal to the number of enrichment parameters of the
element.
The electric potential approximation in the coplanar directions (x, y), in any po-
sition within a piezoelectric ply k, according to the GFEM methodology, is built with
the same PoU functions, Neno (x, y), used to approximate the mechanical displacement
e
ek (x, t) is expressed by
fields. The approximation ϕ

 
N ne nf (ϕkno )
e j
X X
ek (x, t) =
ϕ Neno (x, y) ϕkno (z, t) + ϕkno (z, t)fϕj k (x, y) (32)
no
no=1 j=1

Substituting (28) into (32) one obtains

24
 N ne
(
h i
ke
X
ϕ
e (x, t) = Neno (x, y) ϕk−1
no (t) ζk−1 + ϕk
no (t) ζk
no=1
nf (ϕkno ) h
) (33)
X j j
i
j
+ ϕk−1 k
no (t) ζk−1 + ϕno (t) ζk fϕk (x, y) no
j=1

All degrees of freedom involving electric and mechanical unknowns are grouped
in a vector of nodal parameters for an element e, associated with each node no

n 0 0
T 1 0nf (uno ) 01 0nf (vno )
Ueno = . . . u0no , u0no , . . . , uno 0
, vno , vno , . . . , vno ,
0 )
nf (wno
0 0 1 0 1 nf (ψ )
wno , wno , . . . , wno , . . . , ψ3yno , ψ3y no
, . . . , ψ3yno 3yno ,
1 2
(34)
1 1nf (ϕno ) 1 2nf (ϕno )
ϕ1no , ϕ1no , . . . , ϕno , ϕ2no , ϕ2no , . . . , ϕno ,...
nf (ϕN
o
1 no )
. . . , ϕN N N
no , ϕno , . . . , ϕno ...

The extensional, flexural and warp strains can be grouped in a vector εmf such
that
 
0


 ε 

 
εmf = κ1 (35)

 

 κ 
 
3

In this way, the approximation of the coplanar strains over an arbitrary element
e is accomplished by substituting (31) into (24). Thus we obtain

N ne
X
e
ε(x, t) =
e Bemfno (x, y)Ueno (t)
no=1 (36)
= Bemf Ue

25
This provides the in-plane strains approximation matrix, Bemf , of dimensions 9 ×
7(N ne + npar).
Similarly, the transverse shear strains can be grouped as
 
 γ0 
γc = (37)
 3κ 
2

These strains are approximated over an element e by substituting (31) into (26)
as follows

N ne
X
e
γ
e c (x, t) = Becno (x, y)Ueno (t)
no=1 (38)
= Bec Ue

leading to the transverse shear approximation matrix, Bec , of dimensions 4 × 7(N ne +

npar).
The electric field vector E(x, t) is defined as the negative gradient of the electric
potential function, such that

   ∂ϕ(x,t) 
Ex   − ∂x
 
 


 
 
    

E(x, t) = −∇ϕ(x, t) = Ey = − ∂ϕ(x,t)
∂y
(39)

 
 
 

 
 E    
z  ∂ϕ(x,t) 
− ∂z

e
e t)k in (33), it is possible to express the electric
Then, using the definition of ϕ(x,
field approximation over an element e, within a piezoelectric ply k, by

26
  ( 
 h 
∂
Nno (x, y) ϕk−1
e
no (t) ζk−1

 



 ∂x 



 


 ) 


 i 

+ ϕkno (t) ζk

 


 


 


 


 ( 


 h 

∂
Neno (x, y) ϕk−1
no (t) ζk−1

 

∂y
 
N
Xne  

e
E(x,
e t)k = − ) (40)
 i 
no=1  



 + ϕkno (t) ζk 




 


 ( 


 h 

 
∂
Neno (x, y) ϕno
k−1
 


 ∂z
(t) ζk−1 



 


 


 ) 


 i 

+ ϕkno (t) ζk

 


 


Finally, the electric field vector within a piezoelectric ply k can be approximated
in the form

N ne
e e
X
E(x,
e t)k = − Bkno Ueno
no=1 (41)
ke
= −B Ue
This equation provides the electric field approximation matrix within a piezoelectric
e
ply k, Bk , of dimensions 3 × 7(N ne + npar).
Implementing each portion of the Hamilton’s Principle functional and inserting
the variable discretizations we obtain the expression for the element contributions as
follows. From the variation in the potential energy it is possible to identify the purely
mechanical stiffness, composed of a membrane-bending matrix, Kemf , and transverse
shear Kec matrix, whose integrations over the midplane domain lead to the following
representations

27
  
Z A B L
 
T 
Kemf = Bemf  B D F  Bemf dΩe (42)

Ωe  
L F H
 
A D
Z
T c c
Kec = Bec   Bec dΩe (43)
Ωe Dc Fc
The sub-matrices A, B, D, F, H and L are components of the purely mechanical
constitutive matrix of membrane and bending of the laminate, of dimensions 9×9, and
the sub-matrices Ac , Dc , Fc , are the components of the purely mechanical constitutive
matrix of transverse shear of the laminate. These matrix components are obtained
by integrating across the thickness in the following way

{Aij , Bij , Dij , Lij , Fij , Hij } =

XN Z zk (44)
= Cijk {1, z, z 2 , z 3 , z 4 , z 6 }dz , i, j = 1, 2, 6.
k=1 zk−1

{Acij , Dcij , Fcij } =

XN Z zk (45)
= Cijk {1, z 2 , z 4 }dz , i, j = 4, 5.
k=1 zk−1

At this point it is convenient to decompose the electric field in each piezoelectric

layer k into a constant and a linearly varying part across its thickness, in the form

N ne
( )
e ke
X
E (x, t) = − E0k
no + zE1k
no
no=1
( ) (46)
0k 1k
=− E + zE

28
 Similarly to the forms used for the deformation equations, (35) or (37), it is useful
to group the electric field E0k and E1k , in definition (46), in the following way
 
 E0k 
e ke
E =− (47)
 E1k 

From the expression of the variation of strain energy, it is also possible to identify
the coupled mechanical-electrical stiffness, composed of a membrane-bending coupled
stiffness part Kemf −ϕ and a transverse shear coupled stiffness Kec−ϕ , whose integration
on the in-plane domain results in

 lam  
e
Z O P E1
..

T 
  
Kemf −ϕ = Bemf  P Q   dΩe (48)
  
 .
Ωe    
e
R S Enpiez
 
1e
 lam E
T U
Z
..
 
eT
Kec−ϕ = Bc  dΩe (49)
 
   .
Ωe V W 
e

Enpiez
The sub-matrices Ok , Pk , Qk , Rk and Sk are independents for each piezoelectric
layer k, defining a mechanically-electrically coupled constitutive matrix for membrane-
bending of the laminate, of order 9 × 6npiez , and the sub-matrices Tk , Uk , Vk and
Wk , defining a mechanical-electrically coupled constitutive matrix for shear bending
of the laminate, of order 4 × 6npiez . Their components are given by

k
{Oij , Pijk , Qkij , Rij
k
, Sijk } =
Z zk (50)
= ekij {1, z, z 2 , z 3 , z 4 }dz , i, j = 1, 2, 6.
zk−1

29
 {Tijk , Uijk , Vijk , Wijk } =
Z zk (51)
= ekij {1, z, z 2 , z 3 }dz , i, j = 4, 5.
zk−1

The electriclly-mechanically coupled stiffness matrix, Keϕu , is obtained by a similar

procedure to that used for the mechanically-electrically coupled stiffness matrix, Keuϕ .
T
It should be noted that Keϕu = Keuϕ .
Finally, the purely electrical stiffness matrix Keϕϕ can be obtained by in-plane
integration by

 T  
e e
E1  lam E1
  X Y 
Z
.. ..
h i    
Keϕϕ =  dΩe (52)
  
 .   .
Ωe 
e
 Y Z 
e

Enpiez Enpiez
The purely electrical constitutive matrix of the laminate is of order 6npiez × 6npiez
and it is formed by sub-matrices Xk , Yk and Zk , whose components are given by

{Xij , Yij , Zij } =

Z zk (53)
= χkij {1, z, z 2 }dz , i, j = 1, 2, 3.
zk−1

The total element stiffness matrix is obtained adding all contributions

Ke = Keuu + Keuϕ + Keϕu − Keϕϕ (54)

The element inertia matrix Me is obtained by inserting the mechanical displace-

ment approximation into the variation in the kinetic energy and performing integra-
tion on the element midplane domain Ωe . This results in

30
  
Z P0 P1 P3
 
e T 
M = Ne  P1 P2 P4  Ne dΩe (55)

Ωe  
P3 P4 P6
where P0 = ρ0 I3x3 , P1 = ρ1 I3x2 , P2 = ρ2 I2x2 , P3 = ρ3 I3x2 , P4 = ρ4 I2x2 and
P6 = ρ6 I2x2 . The I3×2 = δij , 1 ≤ i ≤ 3, 1 ≤ j ≤ 2 and the I2×2 = δij , 1 ≤ i ≤ 2,
1 ≤ j ≤ 2. The generalized masses are defined by:

{ρ0 , ρ1 , ρ2 , ρ3 , ρ4 , ρ6 } =
X N Z zk (56)
= ρk {1, z, z 2 , z 3 , z 4 , z 6 } dz
k=1 zk−1

where ρk is an equivalent mass per unit area of the layer k.

Similarly, the vectors of equivalent nodal forces are obtained from the variation
in the external work, such that

Z
eV T e
F = Neuu Nf F V dΩe
ZΩe
S T e
Fe = Neuu Nf F S dΩe
Ωe (57)
eP T
F = Neuu F P
Z
eQ T e
F = Neϕϕ Nq QS dΩe
Ωe

Neuu and Neϕϕ are approximation arrays for the mechanical and electrical unknowns,
respectively. All the contributions are assembled into the element nodal force vector
Fe (t).
The semi-discrete algebraic equations of motion for the element are

31
 Me Üe (t) + Ke Ue (t) = Fe (t) (58)

For the particular case of a quasi-static problem, the system becomes

Ke Ue = Fe (t) (59)

It is known that, even after the correct imposition of essential boundary condi-
tions, the stiffness matrix shows a very high condition number, or becomes singular.
This linear dependence between the system equations occurs because the PoU and
the enrichment functions are both polynomials [27]. The algebraic solution must be
obtained using a procedure suitable for positive-semi-definite matrices. The present
work utilizes the iterative K- method of Babuška [28].

5. Strong formulation for the problem

The equilibrium equations, for the piezoelectric plate model used here, in terms
of generalized forces, are

∂Nx ∂Nxy
+ + qx = fx
∂x ∂y
∂Nxy ∂Ny
+ + qy = fy
∂x ∂y
∂Qx ∂Qy
+ + qz = fz
∂x ∂y
∂Mx ∂Mxy
+ − Qx + my = fmx
∂x ∂y

32
 ∂Mxy ∂My
+ − Qy − mx = fmy
∂x ∂y
∂M3x ∂M3xy
+ − 3Q2x + m3y = f3mx
∂x ∂y
∂M3xy ∂M3y (60)
+ − 3Q2y − m3x = f3my
∂x ∂y
(k) (k)
∂Lx ∂Ly
+ − Jk + q ek = 0
∂x ∂y
The generalized resultant internal forces are defined as

N Z
X zk
σ (k) 1, z, z 3 dz
 
N, M, M3 = (61)
k=1 zk−1

N Z
X zk
τ (k) 1, z 2 dz
 
Q, Q2 = (62)
k=1 zk−1

 (k)  (k)
 L  Z zk 
x z − zk   Dx

L(k) = = dz (63)
 L 
y zk−1 hk  D 
y

Z zk (k)
Dz
Jk = dz (64)
zk−1 hk
where N = {Nx Ny Nxy }T , M = {Mx My Mxy }T and M3 = {M3x M3y M3xy }T are in-
plane stress resultants, whereas Q = {Qy Qx }T and Q2 = {Q2y Q2x }T are transverse
shear stress resultants, σ = {σx σy τxy }T and τ = {τyz τxz }T . It is important to note
that both L(k) , the in-plane electrical displacement resultant, and Jk , the transversal
electrical displacement resultant, are defined for each piezoelectric layer.
In the derivation of the virtual work done by the external forces, the body is
considered to be subjected to distributed forces on the upper surface {q sx q sy q sz },
on the lower surface {q ix q iy q iz } and on the boundary {T n T ns T nz }. T n , T ns and T nz
are the normal, in-plane and transverse shear forces per unity area, respectively. It

33
is assumed the volumetric free charge is zero, that is, Qe = 0, since the piezoelectric
material is dielectric and does not contain free electric charge. Also, in this work the
application of free electric charge on the surface, denoted by q ek , is assumed only
on the lower surface of a piezoelectric layer k. The upper surface of each layer is
assumed to be free of charge. Additionally, it is important to note that δϕ is assumed
to be null on the upper surface of each piezoelectric layer.
In (60) the following generalized external forces on the bi-dimensional domain
are defined
     
 q    q sx + q ix   F 
 x   x 

  
Z h/2  
   
 
qy = q sy + q iy + Fy dz

 
 
 
 −h/2 
 

  
 q   q +q 
 
 F 
z sz iz z
     
 m  h q −q Z h/2 
x iy sy
 −Fy 
= + z dz
 m  2  q −q  −h/2  F 
y sx ix x
     
 m  h3  q − q  Z h/2  −F 
3x iy sy y
= + z3 dz (65)
 m 
3y
8  q −q  −h/2  F
x

sx ix

Moreover, the generalized external forces applied on the boundary are

   
N Mn T zT n
 n   n 
  Z h/2  
 N ns M ns   T ns zT ns 
 =   dz (66)
z3T n
   
 Qn M 3n  −h/2  T nz 
   
M 3ns z 3 T ns
and the generalized inertial forces are

34
 fx = ü0 ρ0 + ψ̈x ρ1 + ψ̈3x ρ3 ,

fy = v̈ 0 ρ0 + ψ̈y ρ1 + ψ̈3y ρ3 ,

fz = ẅ0 ρ0 ,

fmx = ü0 ρ1 + ψ̈x ρ2 + ψ̈3x ρ4 , (67)

fmy = v̈ 0 ρ1 + ψ̈y ρ2 + ψ̈3y ρ4 ,

f3mx = ü0 ρ3 + ψ̈x ρ4 + ψ̈3x ρ6 ,

f3my = v̈ 0 ρ3 + ψ̈y ρ4 + ψ̈3y ρ6

where the generalized mass moments are defined as

N Z
X zk
ρj = ρ(k) z j dz, j = 0, 1, 2, 3, 4, 6. (68)
k=1 zk−1

The following variational consistent boundary conditions are also obtained

Nn = N n Mn = M n M3n = M 3n
Nns = N ns Mns = M ns M3ns = M 3ns (69)
Qn = Qn
(k) (k)
on Γσ of the reference surface and Ln = 0 on Γq of the lower surface of each
piezoelectric layer k. If there is a free electric charge distributed on the piezoelectric
(k) (k)
surface, one has the boundary condition Ln = Lm .

6. Numerical results

The present GFEM formulation for piezoelectric laminated composite plates was
incorporated into a finite element code written with FORTRAN 90 from which nu-
merical solutions for some study cases were obtained. The verification was conducted

35
by comparison with the analytical solutions for the proposed problems, obtained from
the formulation shown in Section 5 using the Navier’s Method. Next, the results for
convergence in a global measure of the relative error in energy norm and a pointwise
assessment of stresses as well as of electric field are displayed. This assessment is
carried out through the static analysis of square composite laminated plates, thick
and thin, with continuous piezoelectric layers, in active and sensor modes.

6.1. Case 1 - Static bending analysis of a plate with piezoelectric sensors

The first problem consists of a thick square laminated composite plate with
continuous piezoelectric sensors bonded on its faces. The plate dimensions are
1000 mm × 1000 mm and a thickness of 100 mm, and the plate is simply sup-
ported at all boundaries and subjected to a uniformly distributed mechanical force.
The piezoelectric layers have only the bottom surface under the free potential condi-
tion, i.e., the top surface of the piezoelectric layers are grounded, with null potential
imposed. The stacking sequence is [0/0/90/90/0/0], as shown in Figure 1(a). The
piezoelectric layers have 10 mm of thickness each and the host laminate has layers
with thicknesses of 26.66 mm, 13.33 mm, 13.33 mm, 26.66 mm, respectively.

The material properties for the substrate layers are

E1 = 172.25 GPa E2 = E3 = 6.89 GPa

G12 = G13 = 3.45 GPa G23 = 2.75 GPa
ν23 = 0.25 ν12 = ν13 = 0.25

and the material properties for PZT-4 are

36
Figure 1: Two laminated composite plates, one thick (a) and the other thin (b), for the passive and
active cases, respectively. The mechanical and electrical boundary conditions are shown, as well as
the thickness of the layers. All dimensions are in millimeters.

E1 = E2 = 94.95 GPa E3 = 81.89 GPa

G12 = 35.90 GPa G13 = G23 = 25.40 GPa
ν12 = 0.32 ν13 = ν23 = 0.38
e15 = e24 = 9.20 C/m2 e31 = e32 = −2.10 C/m2
e33 = 9.50 C/m2
χ11 = χ22 = 4.07 × 10−9 F/m χ33 = 2.08 × 10−9 F/m

Three regular meshes were considered, as shown in Figure 2, over which isotropic
enrichment up to third polynomial degree was applied, for all generalized unknowns.

In sequence, the meshes were modified by a distortion of the elements, as shown

in Figure 3. For the mesh with the poorest h-refinement, the pattern of distortion
is shown by the dimensions and corresponds to the mesh shown in Figure 3(a). The

37
Figure 2: Three regular meshes of generalized finite elements. All meshes are used for the passive
case but only the last two, (b) and (c), are used for the active case.

Figure 3: Three distorted meshes used for the passive case. All dimensions are in millimeters. The
dimensional proportions are respected in the enriched h-refinements.

same pattern was used, respecting the proportions, and it was repeated four and
sixteen times for the meshes (b) and (c), respectively, both in Figure 3.
As a global measure of convergence, the relative error in energy norm was con-
sidered, following (70), evaluated from the numerical energy of deformation, E, and
the analytical energy of deformation, Eref .
s
E − Eref
Errorrel = (70)
Eref

38
 2 x 2 mesh 4 x 4 mesh 8 x 8 mesh

1.0E+00

Relative error
1.0
1.0E-01
2.03

1.0E-02
1.0E+01 1.0E+02 1.0E+03 1.0E+04
Degrees of freedon

Figure 4: Relative error in energy norm considering p-refinement for different regular meshes.

Figure 4 shows the convergence rates for the p-enrichment over the fixed regu-
lar meshes. Figure 5 shows the convergence rates when h-refinement is applied for
different polynomial degrees. It can be observed that there is a tendency to form en-
velopes which exhibit a rate of convergence for p-refinement which is approximately
twice that for h-refinement, as it to be expected.
The convergence rates for p-enrichment on distorted meshes are plotted in Fig-
ure 6. Since the enrichment is built with functions defined in global coordinates,
even with distorted meshes the characteristic rate of convergence for p-refinement is
practically maintained.
Nevertheless, it should be noted that as the number of degrees of freedom becomes
very large the ill-conditioning of the stiffness matrix does not allow effective error
reductions. In this case, this is occurring due to the linear dependence between the
PoU and enrichment functions and also because the problem under analysis has many
physical unknowns and mainly because the material properties have very different
magnitudes.

39
 PoU 1st degree 2nd degree 3rd degree

1.0E+00

Relative error

1.0E-01
1.0
1.33

1.0E-02
1.0E+01 1.0E+02 1.0E+03 1.0E+04
Degrees of freedon

Figure 5: Relative error in energy norm considering h-refinement with four polynomial degrees.

2 x 2 mesh 4 x 4 mesh 8 x 8 mesh

1.0E+00
Relative error

1.0
1.0E-01 1.98

1.0E-02
1.0E+01 1.0E+02 1.0E+03 1.0E+04
Degrees of freedon

Figure 6: Relative error in energy norm considering p-refinement for different distorted meshes.

40
 The quality of the numerical solution may be showed by pointwise evaluation of
stresses and electric field. For this reason, the selected points for stress measurements
were defined as displayed in Figure 7. These points correspond to the 3 × 3 Gauss
quadrature rule for the elements 1, 2, 5 and 6 of the mesh (b) in Figure 2. Their
global coordinates are listed in Table 1.

Figure 7: Lines A to E, over which the points where the stresses and electric field are calculated
are located.

For the stress computation the following nondimensionalized stress definitions,

Sxx and Syz , were considered

2
k k h E1k
Sxx = σxx 2
b q0 E1comp
(71)
k k h
Syz = τyz
bq0
where b is the plate dimension, h is the total thickness and q0 is the load value.
The parameter E1k /E1comp , where comp stands for the base materials (layers 2 to 5)

41
 Table 1: Global coordinates of the points selected for the stress calculations.

line coordinates 1 2 3 4 5 6

A x 28.175 125.000 221.825 278.175 375.000 471.825

y 28.175 28.175 28.175 28.175 28.175 28.175
C x 471.825 471.825 471.825 471.825 471.825 471.825
y 28.175 125.000 221.825 278.175 375.000 471.825
E x 28.175 125.000 221.825 278.175 375.000 471.825
y 28.175 125.000 221.825 278.175 375.000 471.825

and k stands for a given layer, is used to allow a better comparison between the
stress values in different materials. This scaling is not used for the transverse shear
stress because it is piecewise continuous from the integration of the local equilibrium
differential equations.
Figure 8 shows the nondimensionalized in-plane stress Sxx across the plate thick-
ness, computed at the points along line C of Figure 7, for the mesh (b) in Figure
2. The PoU can not approximate the stress and for the first degree enrichment the
computed values are very close the stresses provided by the analytical solution. It
is important to note in Figure 8(a) that the first degree enrichment is not able to
predict the stress in the outermost layers, which is a consequence of the way in which
the boundary conditions are imposed.

Figure 9 shows the distribution of the nondimensionalized transverse shear stresses

Syz evaluated through integration of the local equilibrium differential equations, at
the points over line A in Figure 7. In relation to 9(a), (b) and (c), for the points in
the element closest to the plate corner, the approximation functions are poorer than

42
at the inner elements due to the essential boundary conditions. But the third-degree
enrichment allows a better approximation of the analytical solution for all points,
even that closest to the middle point of the boundary.

Figure 10 shows the electric fields computed in the piezoelectric sensors, at the
points over line E of Figure 7.

6.2. Case 2 - Static bending analysis of a plate with piezoelectric actuators

A second problem was considered in order to verify the applicability of the present
formulation in the analysis of thin plates and its performance for the active case.
In order to identify possible singularities, an electric potential of 200 V was con-
sidered, applied to the actuators only on a region at the plate center, as shown
in Figure 1(b), negative on the top and positive on the bottom piezoelectric layer.
Since the plate is thin, these electrical boundary conditions lead to a solution whose
higher-order generalized rotation only diverges from zero over the boundary of the
region where the potentials are applied, with a sharp gradient, as can be seen in
Figure 11. The first one thousand terms were considered in the computation of the
solution using trigonometric series.

The material properties for the substrate layer are

E1 = 132.38 GPa E2 = E3 = 10.76 GPa

G12 = 3.61 GPa G13 = G23 = 5.65 GPa
ν12 = ν13 = 0.24 ν23 = 0.49

43
and the material properties for PVDF are

E1 = E2 = E3 = 62.97 GPa
G12 = G13 = G23 = 24.20 GPa ν12 = ν13 = ν23 = 0.30
e15 = e24 = 14.13 C/m2 e31 = e32 = 18.41 C/m2
e33 = 12.51 C/m2
χ11 = χ22 = χ33 = 15.30 × 10−7 F/m

In this case, two meshes were considered, as shown in Figures 2(b) and 2(c), on
which enrichment was applied up to the third-degree polynomial.
All the mechanical stresses were evaluated at the points over line E shown in
Figure 7, but this time only at the grey points, since the black points lie on the
interelement boundaries of mesh (c), in Figure 6, and thus one avoids auxiliary
procedures to regularize the stress fields.
In this case, the stress results shown were obtained using the following nondi-
mensional form

k k b2 E1k
Sxx = σxx
h2 E1piez E1comp
(72)
k k b
Syz = τyz
hE1piez

where b is the plate dimension, h is the total thickness, piez stands for piezoelectric
material and the parameter E1k /E1comp is also used to scale the values. Such scaling
is not applied to the transverse shear stress because it is piecewise continuous.
In Figure 12 it can be noted that the approximated nondimensionalized in-plane
stress Sxx is more accurately reproduced within the region where the potential is
applied than in the complementary portion of the domain. Since a higher polynomial
degree is used in these calculations, the results shown in Figures 12(a) and (b)

44
indicate the need for h-refinement to improve the approximate solution. Figure 13
shows good accuracy for the in-plane shear stresses. This is not seen for the transverse
shear stresses (Figure 14) because, since the plate is thin, the contribution of the
higher generalized rotations in the computation of stress derivatives is very small.
Also, in this problem, the transverse shear stresses are very small compared to the
in-plane stresses.

Concluding remarks

The generalized finite element formulation presented here was implemented and
verified by comparison with the respective analytical solution. This formulation
is based on a mixed HSDT-Layerwise model for laminated composite plates with
piezoelectric layers where both the active and sensory behaviors are modeled. This
mixed model represents an efficient tool for modeling adaptive plate structures and
the resulting requirement of only C 0 continuity allows for the implementation of
triangular or quadrilateral elements. The GFEM methodology for enriching approx-
imation subspaces is very versatile and may be easily understood as a extension to
the conventional FEM formulation if a Lagrangian finite element based Partition of
Unity is used. In this procedure, the enrichment degree may vary over the connected
domain without loss of continuity and it is also possible to enrich only some general-
ized unknowns of the model. The performance on application to the static behavior
of piezoelectric plates was verified and good quality approximation can be obtained

45
using either global error measurements or pointwise stress and potential computa-
tions. The reference results were obtained from the analytical solutions developed
and computed by the authors. The p-refinement leads to results that agree very
well with the reference values. The formulation is applicable to thin or thick plates
without locking, since the p-refinement behaves as a classical hierarchical FEM, even
though no specific strategies to deal with locking were included. For thick plates, the
results are better than for thin plates because even for the transverse shear stresses,
computed from the local equilibrium differential equations, the analytical values are
achieved for the third-degree enrichment in both materials. The influence of the way
in which the boundary conditions are enforced was also apparent from these results,
because some errors occur near the boundary. Nevertheless, these error are more no-
ticeable in the case of the stress derivatives. The enrichment with monomials defined
on global coordinates over the plane domain also improves the solutions across the
thickness and makes the procedure very robust to mesh distortions. Even though the
formulation was derived considering inertial effects, the results for dynamic analysis
will be presented in a forthcoming paper. For tailoring toward high gradients, it is
possible to use special enrichment functions to capture localized effects. Moreover, an
aspect which merits attention is the ill-conditioned stiffness matrices generated due
to the multiphysical phenomenon, since the material properties have very different
magnitudes, and methods to deal with ill-conditioned matrices need to be developed.

Acknowledgements

D A F Torres gratefully acknowledges the financial support provided by the

Brazilian government agencies CAPES (Coordenação de Aperfeiçoamento de Pessoal
de Nı́vel Superior) and CNPq (Conselho Nacional de Desenvolvimento Cientı́fico e
Tecnológico) during this research.

46
[1] S. Hurlebaus, L. Gaul, Smart structures dynamics: review. Mechanical Systems
and Signal Processing 20 (2006) 255-281.

[2] H. -J. Lee, Finite element analysis of active and sensory thermopiezoelectric
composite materials (Glenn Research Center - National Aeronautics and Space
Administrations) Report 210892 (2001).

[3] H. Allik, T. J. R. Hughes, Finite element method for piezoelectric vibration.

International Journal for Numerical Methods in Engineering 2 (1970) 151-158.

[4] S. K. Ha, C. Keilers, F. K. Chang, Finite element analysis of composite struc-

tures containing distributed piezoceramic sensors and actuators. AIAA Journal
30 (1992) 772-780.

[5] H. S. Tzou, C. I. Tseng, Distributed piezoelectric sensor/actuator design for

dynamic measurement/control of distributed parameter systems: a piezoelectric
finite element approach. Journal of Sound and Vibration 138 (1990) 17-34.

[6] W. S. Hwang, H. C. Park, Finite element modeling of piezoelectric sensors and

actuators. AIAA Journal 31 (1993) 930-937.

[7] D. T. Detwiler, M. -H. H. Shen, V. B. Venkayya, Finite element analysis of

laminated composite structures containing distributed piezoelectric actuators
and sensors. Finite Elements in Analysis and Design 20 (1995) 87-100.

[8] J. -X. Gao, Y. -P. Shen, Active control of geometrically nonlinear transient
vibration of composite plates with piezoelectric actuators. Journal of Sound
and Vibration 264 (2003) 911-928.

47
 [9] D. A. Saravanos, P. R. Heyliger, D. A. Hopkins, Layerwise mechanics and finite
element for the dynamic analysis of piezoelectric composite plates. International
Journal of Solids and Structures 34 (1997) 359-378.

[10] D. A. Saravanos, Mixed laminate theory and finite element for smart piezoelec-
tric composite shell structures. AIAA Journal 35 (1997) 1327-1333.

[11] S. Cen, A. -K. Soh, Y. -Q. Long, Z. -H. Yao, A new 4-node quadrilateral FE
model with variable electrical degrees of freedom for the analysis of piezoelectric
laminated composite plates. Computers and Structures 58 (2002) 583-599.

[12] K. M. Liew, X. Q. He, M. J. Tan, H. K. Lim, Dynamic analysis of laminated

composite plates with piezoelectric sensor/actuator patches using the FSDT
mesh-free method. International Journal of Mechanical Science 46 (2004) 411-
431.

[13] J. N. Reddy, A simple higher-order theory for laminated plates. Journal of Ap-
plied Mechanics 51 (1984) 745-752.

[14] M. Levinson, An accurate simple theory of the static and dynamics of elastic
plates. Mechanics Research Communications 7 (1980) 343-350.

[15] K. H. Lo, R. M. Christensen, E. M. Wu, A high order theory of plate deformation

2: laminated plates. Journal of Applied Mechanics 44 (1977) 669-676.

[16] J. A. Mitchell, J. N. Reddy, A refined hybrid plate theory for composite lami-
nates with piezoelectric laminae. International Journal of Solids and Structures
32 (1995) 2345-2367.

[17] J. N. Reddy, On laminated composite plates with integrated sensors and actu-
ators. Engineering Structures 21 (1999) 568-593.

48
[18] C. Y. K. Chee, Static shape control of laminated composite plate smart structure
using piezoelectric actuators. Ph.D. dissertation, University of Sydney, 2000.

[19] A. W. Faria, Finite element modeling of composite plates incorporating piezo-

electric sensors and actuators: implementation and numerical assessment. M.Sc.
dissertation. Federal University of Uberlandia (2006) (in Portuguese).

[20] V. M. F. Correia, M. A. A. Gomes, A. Suleman, C. M. M. Soares, C. A. M.

Soares, Modelling and design of adaptive composite structures. Computer Meth-
ods in Applied Mechanics and Engineering 185 (2000) 325-346.

[21] V. Cotoni, P. Masson, F. Côté, A finite element for piezoelectric multilayered

plates: combined higher-order and piecewise linear C 0 formulation. Journal of
Intelligent Material Systems and Structures 17 (2006) 155-166.

[22] J. E. S. Garo, C. M. M. Soares, C. A. M. Soares, J. N. Reddy, Analysis of lami-

nated adaptive plate structures using layerwise finite element models. Comput-
ers and Structures 82 (2004) 1939-1959.

[23] E. Carrera, Evaluation of layerwise mixed theories for laminated plates analysis.
American Institute of Aeronautics and Astronautics Journal 36 (1998) 830-839.

[24] R. G. Lage, C. M. M. Soares, C. A. M. Soares, J. N. Reddy, Modeling of piezo-

laminated plates using layerwise mixed finite elements. Computers and Struc-
tures 82 (2004) 1849-1863.

[25] R. G. Lage, C. M. M. Soares, C. A. M. Soares, J. N. Reddy, Layerwise partial

mixed finite element analysis of magneto-electro-elastic plates. Computers and
Structures 82 (2004) 1293-1301.

49
[26] J. T. Oden, C. A. Duarte, O. C. Zienkiewicz, A new cloud-based hp finite element
method, Computer Methods in Applied Mechanics and Engineering 153 (1998)
117-126.

[27] C. A. Duarte, I. Babuška, J. T. Oden, Generalized finite element method for

three-dimensional structural mechanics problems. Computers and Structures 77
(2000) 215-232.

[28] T. Strouboulis, I. Babuška, K. Copps, The design and analysis of the generalized
finite element method. Computer Methods in Applied Mechanics and Engineer-
ing 181 (2000) 43-69.

[29] T. Strouboulis, I. Babuška, K. Copps, The generalized finite element method.

Computer Methods in Applied Mechanics and Engineering 190 (2001) 4081-
4193.

[30] C. A. Duarte, D. -J. Kim, D. M. Quaresma, Arbitrarily smooth generalized

finite element approximations. Computer Methods in Applied Mechanics and
Engineering 196 (2006) 33-56.

[31] C. A. Duarte, D. Q. Migliano, E. B. Becker, A technique to combine meshfree-

and finite element-based partition of unity approximations (Department of Civil
and Environmental Engineering - University of Illinois at Urbana-Champaign)
Structural Research Series 638 (2005).

[32] G. R. Liu, Mesh free methods: moving beyond the finite element method, CRC
Press, Boca Raton, 2003.

[33] I. Babuška, G. Caloz, J. E. Osborn, Special finite element methods for a class

50
 of second order elliptic problems with rough coefficients. SIAM Journal of Nu-
merical Analysis 31 (1994) 945-981.

[34] J. M. Melenk, On generalized finite element method. Ph.D. dissertation, Uni-

versity of Maryland, 1995.

[35] J. M. Melenk, I. Babuška, The partition of unity finite element method: basic
theory and applications. Computer Methods in Applied Mechanics and Engi-
neering 139 (1996) 289-314.

[36] I. Babuška, J. M. Melenk, The partition of unity finite element method. Inter-
national Journal for Numerical Methods in Engineering 40 (1997) 727-758.

[37] C. A. Duarte, The hp-cloud method. Ph.D dissertation, The University of Texas
at Austin, 1996.

[38] C. A. Duarte, J. T. Oden, An h-p adaptive method using cloud. Computer

Methods in Applied Mechanics and Engineering 139 (1996) 237-262.

[39] T. Belytschko, T. Black, Elastic crack growth in finite elements with mini-
mal remeshing. International Journal for Numerical Methods in Engineering 45
(1999) 601-620.

[40] P. J. O’Hara, C. A. Duarte, T. Eason, Generalized finite element analysis of

three-dimensional heat transfer problems exhibiting sharp thermal gradients.
Computer Methods in Applied Mechanics and Engineering 198 (2009) 1857-
1871.

[41] T. Strouboulis, L. Zhang, I. Babuška, Generalized finite element method using

mesh-based handbooks: application to problems in domains with many voids.

51
 Computer Methods in Applied Mechanics and Engineering 192 (2003) 3109-
3161.

[42] C. A. Duarte, O. N. Hamzeh, T. J. Liszka, W. W. Tworzydlo, A generalized

finite element method for the simulation of three-dimensional crack propagation.
Computer Methods in Applied Mechanics and Engineering 190 (2001) 2227-
2262.

[43] F. B. Barros, C. S. Barcellos, C. A. Duarte, -p Adaptive C k generalized finite

element method for arbitrary polygonal clouds. Computational Mechanics 41
(2007) 175-187.

[44] T. Strouboulis, L. Zhang, D. Wang, I. Babuška, A posteriori error estimation for

generalized finite element methods. Computer Methods in Applied Mechanics
and Engineering 195 (2006) 852-879.

[45] P. J. O’Hara, Finite element analysis of three-dimensional heat transfer for

problems involving sharp thermal gradients. M.Sc. dissertation, University of
Illinois at Urbana-Champaign, 2007.

[46] C. S. Barcellos, P. T. R. Mendonça, C. A. Duarte, A C k continuous gener-

alized finite element formulations applied to laminated Kirchhoff plate model.
Computational Mechanics 44 (2009) 377-393.

[47] O. A. Garcia, E. A. Fancello, P. T. R. Mendonça, Developments in the applica-

tion of the generalized finite element method to thick shell problems. Compu-
tational Mechanics 44 (2009) 669-682.

[48] J. T. Oden, J. N. Reddy, An introduction to the mathematical theory of finite

elements, Wiley, New York, 1976.

52
[49] J. N. Reddy, Mechanics of laminated composite plates and shells: theory and
analysis, CRC Press, Boca Raton, 2004.

[50] B. Szabó, I. Babuška, Finite element analysis, John Wiley and Sons, New York,
1991.

[51] S. Vidoli, R. C. Batra, Derivation of plate and rod equations for a piezoelectric
body from a mixed three-dimensional variational principle. Journal of Elasticity
59 (2000) 23-50.

53
 PoU 1st degree PoU 1st degree
2nd degree 3rd degree 2nd degree 3rd degree
Analytical solution Analytical solution
0.5 0.5

0.4 0.4

0.3 0.3
Normalized thickness, z/h

Normalized thickness, z/h

0.2 0.2

0.1 0.1

0.0 0.0

-0.1 -0.1

-0.2 -0.2

-0.3 -0.3

-0.4 -0.4

-0.5 -0.5
-0.04 -0.02 0.00 0.02 0.04 -0.20 -0.10 0.00 0.10 0.20
Stress, S xx Stress, S xx
a b
PoU 1st degree PoU 1st degree
2nd degree 3rd degree 2nd degree 3rd degree
Analytical solution Analytical solution
0.5 0.5

0.4 0.4

0.3 0.3
Normalized thickness, z/h

Normalized thickness, z/h

0.2 0.2

0.1 0.1

0.0 0.0

-0.1 -0.1

-0.2 -0.2

-0.3 -0.3

-0.4 -0.4

-0.5 -0.5
-0.30 -0.20 -0.10 0.00 0.10 0.20 0.30 -0.40 -0.20 0.00 0.20 0.40
Stress, S xx Stress, S xx
c d
PoU 1st degree PoU 1st degree
2nd degree 3rd degree 2nd degree 3rd degree
Analytical solution Analytical solution
0.5 0.5

0.4 0.4

0.3 0.3
Normalized thickness, z/h

Normalized thickness, z/h

0.2 0.2

0.1 0.1

0.0 0.0

-0.1 -0.1

-0.2 -0.2

-0.3 -0.3
54
-0.4 -0.4

-0.5 -0.5
-0.40 -0.20 0.00 0.20 0.40 -0.50 -0.25 0.00 0.25 0.50
Stress, S xx Stress, S xx
e f

Figure 8: Distribution of non-dimensionalized in-plane stress Sxx within the plate thickness, at the
points along the line C.
 1st degree 2nd degree 1st degree 2nd degree
3rd degree Analytical solution 3rd degree Analytical solution
0.5 0.5

0.4 0.4

0.3 0.3
Normalized thickness, z/h

Normalized thickness, z/h

0.2 0.2

0.1 0.1

0.0 0.0

-0.1 -0.1

-0.2 -0.2

-0.3 -0.3

-0.4 -0.4
-0.5 -0.5
-0.06 -0.03 0.00 0.03 0.06 -0.05 0.00 0.05 0.10 0.15 0.20
Stress, S yz Stress, S yz
a b
1st degree 2nd degree 1st degree 2nd degree
3rd degree Analytical solution 3rd degree Analytical solution
0.5 0.5

0.4 0.4

0.3 0.3
Normalized thickness, z/h

Normalized thickness, z/h

0.2 0.2

0.1 0.1

0.0 0.0

-0.1 -0.1

-0.2 -0.2

-0.3 -0.3

-0.4 -0.4

-0.5 -0.5
-0.10 0.00 0.10 0.20 0.30 -0.10 0.00 0.10 0.20 0.30 0.40
Stress, S yz Stress, S yz
c d
1st degree 2nd degree 1st degree 2nd degree
3rd degree Analytical solution 3rd degree Analytical solution
0.5 0.5

0.4 0.4

0.3 0.3
Normalized thickness, z/h

Normalized thickness, z/h

0.2 0.2

0.1 0.1

0.0 0.0

-0.1 -0.1

-0.2 -0.2

-0.3 -0.3
55
-0.4 -0.4

-0.5 -0.5
-0.10 0.00 0.10 0.20 0.30 0.40 -0.10 0.00 0.10 0.20 0.30 0.40
Stress, S yz Stress, S yz
e f

Figure 9: Distribution of non-dimensionalized transversal shear stress Syz within the plate thickness,
at the points along the line A.
 PoU 1st degree PoU 1st degree
2nd degree 3rd degree 2nd degree 3rd degree
Analytical solution Analytical solution
0.5 0.5

0.4 0.4

0.3 0.3
Normalized thickness, z/h

Normalized thickness, z/h

0.2 0.2

0.1 0.1

0.0 0.0

-0.1 -0.1

-0.2 -0.2

-0.3 -0.3

-0.4 -0.4

-0.5 -0.5
-50 0 50 100 -100 -50 0 50 100
Electric field, E z Electric field, E z
a b
PoU 1st degree PoU 1st degree
2nd degree 3rd degree 2nd degree 3rd degree
Analytical solution Analytical solution
0.5 0.5

0.4 0.4

0.3 0.3
Normalized thickness, z/h

Normalized thickness, z/h

0.2 0.2

0.1 0.1

0.0 0.0

-0.1 -0.1

-0.2 -0.2

-0.3 -0.3

-0.4 -0.4

-0.5 -0.5
-150 -100 -50 0 50 100 150 -150 -100 -50 0 50 100 150
Electric field, E z Electric field, E z
c d
PoU 1st degree PoU 1st degree
2nd degree 3rd degree 2nd degree 3rd degree
Analytical solution Analytical solution
0.5 0.5

0.4 0.4

0.3 0.3
Normalized thickness, z/h

Normalized thickness, z/h

0.2 0.2

0.1 0.1

0.0 0.0

-0.1 -0.1

-0.2 -0.2

-0.3 -0.3
56
-0.4 -0.4

-0.5 -0.5
-200 -100 0 100 200 -200 -100 0 100 200
Electric field, E z Electric field, E z
e f

Figure 10: Distribution of electric field within the plate thickness, at the points along the line E.
 a

Figure 11: Analytical solution for the ψ3x and ψ3y higher order warping variables, a and b respec-
tively, for the thin plate with actuators.

57
 4 x 4 mesh 8 x 8 mesh Analytical solution 4 x 4 mesh 8 x 8 mesh Analytical solution
0.5 0.5

0.4 0.4

0.3 0.3
Normalized thickness, z/h

Normalized thickness, z/h

0.2 0.2

0.1 0.1

0.0 0.0

-0.1 -0.1

-0.2 -0.2

-0.3 -0.3

-0.4 -0.4

-0.5 -0.5
-0.02 -0.01 0.00 0.01 0.02 -0.50 -0.25 0.00 0.25 0.50
4 4
Stress, S xx (x10 ) Stress, S xx (x10 )
a b
4 x 4 mesh 8 x 8 mesh Analytical solution 4 x 4 mesh 8 x 8 mesh Analytical solution
0.5 0.5

0.4 0.4

0.3 0.3
Normalized thickness, z/h

Normalized thickness, z/h

0.2 0.2

0.1 0.1

0.0 0.0

-0.1 -0.1

-0.2 -0.2

-0.3 -0.3

-0.4 -0.4

-0.5 -0.5
-2.00 -1.00 0.00 1.00 2.00 -3.00 -2.00 -1.00 0.00 1.00 2.00 3.00
Stress, S xx (x104) Stress, S xx (x104)
c d

Figure 12: Distribution of non-dimensionalized in-plane stress Sxx within the plate thickness, at
the points along the line E, considering third degree enrichment.

58
 4 x 4 mesh 8 x 8 mesh Analytical solution 4 x 4 mesh 8 x 8 mesh Analytical solution
0.5 0.5

0.4 0.4

0.3 0.3
Normalized thickness, z/h

Normalized thickness, z/h

0.2 0.2

0.1 0.1

0.0 0.0

-0.1 -0.1

-0.2 -0.2

-0.3 -0.3

-0.4 -0.4

-0.5 -0.5
-3.00 -2.00 -1.00 0.00 1.00 2.00 3.00 -5.00 -2.50 0.00 2.50 5.00
5 5
Stress, S xy (x10 ) Stress, S xy (x10 )
a b
4 x 4 mesh 8 x 8 mesh Analytical solution 4 x 4 mesh 8 x 8 mesh Analytical solution
0.5 0.5

0.4 0.4

0.3 0.3
Normalized thickness, z/h

Normalized thickness, z/h

0.2 0.2

0.1 0.1

0.0 0.0

-0.1 -0.1

-0.2 -0.2

-0.3 -0.3

-0.4 -0.4

-0.5 -0.5
-4.00 -2.00 0.00 2.00 4.00 -0.06 -0.04 -0.02 0.00 0.02 0.04 0.06
Stress, S xy (x105) Stress, S xy (x105)
c d

Figure 13: Distribution of non-dimensionalized in-plane shear stress Sxy within the plate thickness,
at the points along the line E, considering third degree enrichment.

59
 4 x 4 mesh 8 x 8 mesh Analytical solution 4 x 4 mesh 8 x 8 mesh Analytical solution
0.5 0.5

0.4 0.4

0.3 0.3
Normalized thickness, z/h

Normalized thickness, z/h

0.2 0.2

0.1 0.1

0.0 0.0

-0.1 -0.1

-0.2 -0.2

-0.3 -0.3

-0.4 -0.4

-0.5 -0.5
-0.80 -0.60 -0.40 -0.20 0.00 0.20 -1.20 -0.80 -0.40 0.00 0.40 0.80
8 8
Stress, S yz (x10 ) Stress, S yz (x10 )
a b
4 x 4 mesh 8 x 8 mesh Analytical solution 4 x 4 mesh 8 x 8 mesh Analytical solution
0.5 0.5

0.4 0.4

0.3 0.3
Normalized thickness, z/h

Normalized thickness, z/h

0.2 0.2

0.1 0.1

0.0 0.0

-0.1 -0.1

-0.2 -0.2

-0.3 -0.3

-0.4 -0.4

-0.5 -0.5
-4.00 -2.00 0.00 2.00 -0.10 0.00 0.10 0.20 0.30
8 8
Stress, S yz (x10 ) Stress, S yz (x10 )
c d

Figure 14: Distribution of non-dimensionalized transversal shear stress Syz within the plate thick-
ness, at the points along the line E, considering third degree enrichment.

60