Home

Search

Collections

Journals

About

Contact us

My IOPscience

Integrated design optimization of voltage channel distribution and control voltages for tracking
the dynamic shapes of smart plates

This article has been downloaded from IOPscience. Please scroll down to see the full text article.
2010 Smart Mater. Struct. 19 125013
(http://iopscience.iop.org/0964-1726/19/12/125013)
View the table of contents for this issue, or go to the journal homepage for more

Download details:
IP Address: 202.118.74.168
The article was downloaded on 13/11/2010 at 06:04

Please note that terms and conditions apply.

IOP PUBLISHING

SMART MATERIALS AND STRUCTURES

Smart Mater. Struct. 19 (2010) 125013 (8pp)

doi:10.1088/0964-1726/19/12/125013

Integrated design optimization of voltage

channel distribution and control voltages
for tracking the dynamic shapes of smart
plates
Shutian Liu and Zheqi Lin
State Key Laboratory of Structural Analysis for Industrial Equipment, Department of
Engineering Mechanics, Dalian University of Technology, Dalian 116024,
Peoples Republic of China
E-mail: [email protected]

Received 28 March 2010, in final form 10 October 2010

Published 12 November 2010
Online at stacks.iop.org/SMS/19/125013
Abstract
This paper investigates a control scheme for tracking the dynamic shapes of structures with
limited numbers of voltage channels. Integrated design optimization of voltage channel
distribution and control parameters for structural dynamic shape control is formulated as an
optimization problem with discrete variables and continuous variables coexisting. A two-level
optimization method based on a simulated annealing algorithm is proposed. In the first level,
the optimum channel distribution is determined by optimizing the objective function which is
the optimal value obtained in the second level. The optimum control parameters are obtained by
using a sequential linear least-squares algorithm in the second level. The effectiveness of the
present design methodology and optimization scheme is then demonstrated through numerical
examples for tracking the dynamic shapes of composite plates.
(Some figures in this article are in colour only in the electronic version)

needed to achieve a specified desired shape. Hsu et al [8]

applied the gradient projection algorithm to find the optimal
values of design variables in the shape control of plates. Chee
et al [9] employed the buildup voltage distribution (BVD),
linear least-squares (LLS) and simplex algorithms for finding
the optimum voltages in all actuators to achieve the desired
structural shapes. Liew et al [10] presented an optimization
algorithm based on computational intelligence (CI) to derive
the optimal voltage distribution and gain control matrix for
the shape control of FGM plates that contain piezoelectric
sensor and actuator patches. Chen et al [11] proposed an
approach employing displacementstress dual criteria for static
shape control. This approach is based on normal displacement
control and stress modification is considered in the whole
optimization process to control high stress in the local domain.
The control input voltage is obtained by using an iteration
algorithm.
Besides optimizing the applied voltages for shape
control of structures, another important issue that needs

1. Introduction
Smart structures have attracted a great deal of attention in
the last few decades. The development of smart structures
has created new avenues of research, particularly in the
fields of health monitoring, vibration control and shape
control [14]. With the emergence of high-performance
piezoelectric materials which undergo electromechanical
effects, a smart structure with embedded or surface mounted
piezoelectric actuators can be capable of varying or morphing
its geometric shape during its operational life to meet the
missions requirement. Such a structure usually requires
a suitable combination of active control methods for high
performance. Shape control of structures using piezoelectric
materials has gained wide attention, and many works have
already been reported in this area [5, 6].
Koconis et al [7] first investigated the shape control of
composite plates and shells with embedded actuators. They
used the RaleighRitz method to determine the voltages
0964-1726/10/125013+08$30.00

2010 IOP Publishing Ltd Printed in the UK & the USA

Smart Mater. Struct. 19 (2010) 125013

S Liu and Z Lin

to be addressed is the determination of the optimal

configuration parameters of the piezoelectric actuators.
Different optimization methods are proposed for design
optimization of piezoelectric actuator patterns. Merkhujee
and Joshi [12] presented a heuristic iterative procedure to
find the optimal shape of piezoelectric actuators based on the
residual voltages of elements in order to minimize the error
between the desired and the current structural configuration.
Dawood et al [13] used a three-dimensional brick element for
the shape control analysis of composite plates integrated with
distributed piezoelectric actuators. A weighted shape control
method is used to determine the required input voltages to
achieve a specified structural shape. They also examined the
effects of different lamination angles, boundary conditions,
plate length-to-thickness ratios and actuator dimensions on the
control voltages. Liu et al [14] presented an investigation
into optimization of PZT-based actuators by simultaneously
optimizing two sets of design variables, i.e. controlling and
geometrical parameters for morphing structural shapes. In
the algorithm, the controlling parameters are optimized using
the linear least-squares method and their determination is
embedded in the optimization of geometrical parameters.
In contrast to static shape control, less work has been
carried out on dynamic shape control of structures. Forster
and Livne [15] presented an integrated multidisciplinary
optimization to solve the dynamic shape control problem in
which the structure is forced to vibrate at given frequencies and
a given shape of motion. The host structure and its actuators,
as well as electric fields required, are used as design variables.
Luo and Tong [16] formulated a sequential linear least-squares
algorithm (SLLS) for tracking the dynamic shapes of PZT
smart structures. In their later work [17], a segment-based
sequential least-squares algorithm (SSLLS) is proposed to
track the dynamic shapes of smart structures considering both
tracking precision and electrical energy consumption.
In implementing shape control of structures using
piezoelectric materials, extensive studies have been devoted to
the associated design optimization problems. In most of these
optimization models, the voltage of each actuator is treated as
an independent variable, which means the number of voltage
channels must be the same as that of the actuators. To obtain
high-precision control of structure shape, a large number of
actuators may be used. In many practical cases, only limited
numbers of voltage channels may be available when taking
control cost and operational efficiency into account. Therefore,
it would be ideal to find a way to carry out precise shape control
with limited voltage channels.
To address this issue, a shape control design problem
considering the limitation on the number of voltage channels
is formulated as a design optimization of voltage channel
distribution and voltage of each channel. A parameterized
formulation of voltage channel distribution which shows the
relationship between the actuators and voltage channels is
presented in this paper. The optimal channel distribution and
optimal voltage of each channel for tracking dynamic shapes
are determined by a two-level optimization method based
on a simulated annealing algorithm. Numerical examples
of composite plate dynamic shape control are presented to
validate the developed method.

Figure 1. Geometry and distribution of the PZT stiffeners of a

composite plate.

2. Description of the optimization model

For simplicity in presentation and as a representative example,
a thin composite plate with PZT beam actuators bonded on
the surface is investigated here, as shown in figure 1. The
shape of the composite plate is controlled by Na actuators with
Nchannel voltage channels and Nchannel  Na . In order to give
a parameterized formulation for voltage channel distribution,
a distribution indicator vector = (1 , 2 , . . . , Na )T and
a distribution matrix D = [Di j ] are introduced. i (i =
1, 2, . . . , Na ) is an integer ranging from 1 to Nchannel . The
element of the distribution matrix D is defined as

Di j = ( j j )

(1)

where (x) is the Kronecker function:


1
j = j
( j j ) =
0
j = j .

(2)

Then, in the static shape control problem, the relationship between applied voltages of PZT actuators
V = (V1 , V2 , . . . , VNa )T and voltages of channels =
(1 , 2 , . . . ,  Nchannel )T can be written as
V = D.

(3)

In the dynamic shape control problem, an applied voltage

history for PZT actuators over a period of time V(t) =
{{V(ti+1 )}, {V(ti+2 )}, . . . , {V(ti+n )}} can be formulated as

(t)
V(t) = D
where

=
D

(4)

D
D

[ 0]

..

[ 0]

(5)

D
and (t) = {{(ti+1 )}, {(ti+2 )}, . . . , {(ti+n )}} denotes the
voltage history of the channels.
The motion equations of PZT smart structures can be
expressed in the following finite element formulation:

(t)} + [C]{U
(t)} + [Kuu ]{U(t)}
[M]{U
]{(t)}
= {F(t)} [Ku ]{V(t)} = {F(t)} [Ku ][D
2

(6)

Smart Mater. Struct. 19 (2010) 125013

S Liu and Z Lin

where [M], [C], [Kuu ] are mass, damping and stiffness

(t)], [U
(t)] are nodal displacement,
matrices; [U(t)], [U
velocity and acceleration vectors; [Ku ] is the coupled
electrical/mechanical stiffness matrix and {F(t)} is the
mechanical loading.
The transverse displacements {w(ti+1 )} of the nodes
considered at time ti+1 can be obtained from the global
displacement vector by using a weighting matrix [Ru(i+1) ]:

{w(ti+1 )} = [Ru(i+1) ]{U(ti+1 )}.

The second optimization level is done to determine the

optimum voltage history for a given channel distribution and
to obtain the objective function considered as a measure
of the voltage channel distribution for the first optimization
level. Since it is an optimization problem with discrete design
variables in the first optimization level, the mathematicalprogramming-based algorithms lose their power.
Some
iterative algorithms, such as the genetic algorithm, simulated
annealing and tabu search, are believed to be applicable to this
kind of optimization problem [18]. In this paper a simulated
annealing algorithm is employed to find the optimum voltage
channel distribution. A simulated annealing algorithm is an
iterative search method for optimization problems inspired
by an analogy to the statistical mechanics of annealing in
solids [19]. It performs a stochastic search in the state space,
gradually adjusting a parameter called temperature. Unlike
other downhill-type optimization algorithms, annealing allows
perturbations to move uphill in a controlled fashion, which
makes it possible to jump out of local minima and potentially
fall into a more promising downhill path.

(7)

The square error between the actuated and desired shape

at time t is defined as

e = {{wc (t)} {wd (t)}}T {{wc (t)} {wd (t)}}

(8)

where {wc (t)} and {wd (t)} are the computed shapes and desired
shapes at time t , respectively.
Considering the shape error function over a period of
time [ti+1 , ti+n ], the design optimization of voltage channel
distribution and control voltages for tracking dynamic shapes
can be formulated as
Find: X = ( , (t))T
ti+n
Min: E =
{{wc (t)} {wd (t)}}T {{wc (t)}
(9)

4. Simulated annealing algorithm-based two-level

optimization method
Starting with an initial solution and armed with adequate
perturbation and evaluation functions, the simulated annealing
algorithm does a random walk in the space of possible
solutions. During the search process, whether a new solution
is accepted or not is determined by the Metropolis criterion
which simulates the annealing process at a given temperature
T . Temperature is initialized to a value T0 at the beginning
and is slowly reduced. The annealing procedure halts when the
temperature exceeds the terminating temperature.
The core of the simulated annealing algorithm is the
Metropolis criterion. In the physical problem to coerce some
material into a low energy state, we heat it and then cool it
very slowly, allowing it to come to thermal equilibrium at each
temperature. Metropolis devised a similar scheme to simulate
how the system reaches thermodynamic equilibrium at each
fixed temperature in the schedule of decreasing temperatures
used to anneal it. The idea is implemented as follows: generate
a local neighbor Snew of any given state S , such as moving
a particle to a new location in the physical system, and then
evaluate the resulting change in objective function (energy)
F . Let F(S) return the objective value of a given state
S . If the objective value of the new state Snew is lower than
that of the current state Scur , that is F < 0, then the new
state is accepted as the starting point for the next move by
setting Scur = Snew . If F > 0, the new state is accepted on
a probabilistic basis, which means uphill moves will happen
occasionally. The probability of an uphill move at temperature
T is
F(Snew )F(Scur )
T
PT = e
.
(12)

ti+1

{wd (t)}} dt.

3. Optimization scheme
For the structural shape control problem defined in equation (9), both continuous and discrete design variables are
included in the optimization model. No existing optimization
algorithm is effective in solving these problems directly. A new
solution strategy is proposed as follows.
We analyze problem (9) again and find that it can be reformulated as follows:
Find: = (1 , 2 , . . . , Na )T
ti+n
Min: g( ) = min
{{wc (t)} {wd (t)}}T {{wc (t)}
(10)

ti+1

{wd (t)}} dt.

Two optimization levels are included in equation (10). The first
optimization level is a combinatorial optimization problem, in
which we seek to find some channel distribution to minimize
the objective function, which can be obtained through solving
the second optimization level. The second optimization level
means to find the optimum voltage history over a period of time
for each channel under a given voltage channel distribution.
For the tracking dynamic shape problem in (10), the
Houbolt numerical integration method is employed to solve
the structure motion equation. The optimum voltage history for
each channel can be obtained by using a sequential linear leastsquares algorithm. The appendix presents the details of the
calculation and the optimum solution can be found by solving
the linear algebraic equation expressed as

]T [RkV ]T [RkV ][D

]{kn } = [RkV ]T ({wkd } {wkc }).
[D
n
n
n
n
n

The inferior state is accepted only if a uniform random number

which is generated in the range 01 is smaller than PT .
This criterion for accepting the new state is known as the
Metropolis criterion. For simulating the amount of time for
which annealing must be applied at temperature T , a value of

(11)
3

Smart Mater. Struct. 19 (2010) 125013

S Liu and Z Lin

Table 1. The mechanical properties of the smart plate.

Composite plate

PZT actuator

Adhesive layer

E L = 143 GPa
E T = 9.7 GPa
G LT = 6.0 GPa
LT = 0.3
= 2300 kg m3

E s = 70 GPa
s = 0.25
s = 7500 kg m3
e31 = 5.2 N mV1
e32 = 5.2 N mV1

E a = 3.0 GPa
G a = 1.07 GPa

M needs to be provided. That means M states are generated

and examined by the Metropolis criterion at each temperature.
Simulated annealing is applicable for combinatorial
optimization problems, and when it is used for the task in this
paper several base components are needed.

Figure 2. The desired time functions of the shape variations.

Step 6: temperature decreases, Tnew = Told , set

= 0.95. If the current temperature reaches the given
terminating temperature, the algorithm ends; else, turn
back to step 4.

(a) States: distribution indicator vector = (1 , 2 , . . . , Na )T

represents the solution space over which a good answer is
to be searched for.
(b) Move set: a local neighbor  of any given state
is generated by probabilistic changing of i (i =
1, 2, . . . , Na ).
(c) Evaluation function: the objective function of the second
level g( ), which is defined by equation (10), is chosen as
the evaluation function measuring any given state. g( )
is obtained by solving an optimization problem which
considers control voltages as variables.
(d) Initial temperature: the simulated annealing algorithm
needs to start from a high temperature which permits
an aggressive random search of the state space and
most uphill moves are allowed. However, too high a
temperature causes a waste of processing time. It is
initialized by using the procedure described in [20], T0 =
(Fmax Fmin )/ ln pr , where Fmax and Fmin are the
maximum and minimum objective function values in a set
of random solutions. pr is the initial accepted probability,
pr 1.
(e) Cooling schedule: to anneal the problem from a random
state to a good, frozen one, the temperature decreases in
the form of Tnew = Told , 0 < < 1.

5. Numerical simulations
To validate the present formulation and the proposed solution
strategy, the design optimization examples of a composite
laminated plate with orthogonally placed PZT stiffeners are
employed. Consider a laminated cantilever rectangular plate
of 240 mm 200 mm with orthogonally distributed PZT beam
actuators as shown in figure 1. The mechanical properties of
each part of the smart plate are shown in table 1. The laminated
plate has the following ply sequence: (90/0/90)s. The plate
thickness is taken as h = 0.6 mm. The cross section of the
PZT beam is assumed to be in rectangular shape with a width
of 2 mm and a height of 0.5 mm. The PZT beam actuators are
bonded to the plates with a bondline thickness of ta = 0.1 mm.
A finite element model for the PZT stiffened plate has been
depicted in [21].
The PZT stiffened plate is controlled by the given numbers
of voltage channels to achieve the desired shapes. Consider the
following twisting deformation as the desired structural shape
of the rectangular plate:

According to the base components for the simulated

annealing algorithm, the whole procedure of the optimization
scheme for problem (10) is defined as shown below.

wd (x, y) =

(cosh x 1) sin y
G

(13)

where G is a factor (assumed to be 30 here). The maximum

deflection is 9.63 105 (m), which is 16.1% of the plate
thickness. We consider two types of time functions for
deflections of the plate as shown in figure 2, namely sine and
triangular functions of time. The present optimization scheme
is implemented to find the optimum solutions according to
different numbers of voltage channels. As a representative
example, the optimum solution of six voltage channels is
depicted in detail.
Case 1. Tracking twisting shape with sine function of time.
Figure 3 shows the process of optimization iteration, from
which we can see the resemblance between the optimization
process and the annealing process of solid substances in
physics. At high temperature, the search is almost random and
the probability of an uphill move is large. With the decline of

Step 1: obtain the influence coefficient matrix using finite

element analysis.
Step 2: compute initial temperature T0 .
Step 3: generate an initial state randomly, calculate the
objective function value F .
Step 4: in the current temperature T , generate a new state
 , which is a local neighbor of , evaluate the new state
and return the objective function value F  .
Step 5: accept the new state according to the Metropolis
criterion. If min(1, exp((F  F)/T ))  random(0, 1),
accept new state =  ; else, maintain state . Then,
turn back to step 4. Until M (set to be the amount
of actuators) states are generated and examined by the
Metropolis criterion, turn to step 6.
4

Smart Mater. Struct. 19 (2010) 125013

S Liu and Z Lin

Figure 3. The iterative process of twisting shape control with a sine

function of time.

Figure 6. The deflections of the plate twisting at point A for case 1.

Figure 4. The optimum time-dependent control voltage of each

channel for case 1.
Figure 7. The achieved shape of the plate twisting at time 250 ms for
case 1.

(optimum solution) is obtained when annealing terminates.

The optimum time-dependent control voltage of each channel
is found finally as showed in figure 4 and the optimum voltage
channel distribution is plotted as figure 5.
In the dynamic twisting tracking for the PZT stiffened
plate using the present algorithm, the time-dependent
deflections at point A are illustrated in figure 6, which shows
that the desired structural movements are tracked with high
precision. Figure 7 shows the computed plate shape at a
time of 250 ms. In order to show further information about
the effectiveness of shape control, the relative differences of
observation points between the actuated and desired ones are
illustrated in figure 8. It is worth noting that the relative error in
the region of y = 0 has no definition as the desired deformation
is equal to zero. For plotting convenience the relative errors
in this region are set to zero. It can be seen that the larger
relative errors occur near the supporting area. The maximum
relative error is 9.15%. Since the deflections at these areas are
very small, the relative errors are larger. Therefore, the desired
twisting shape motions can be achieved with high accuracy by
using the proposed method.
The effect of the different numbers of voltage channels
on the control performance is examined. The optimal control
schemes of 4, 6, 8, 10, 20, 30 and 60 channels are obtained
by using the proposed method. Figure 9 shows the squared

Figure 5. Optimum channel distribution to achieve twisting shape of

composite plate for case 1. The numbers labeled indicate the channel
distribution indicator.

temperature, fewer uphill moves are allowed and the energy of

the system decreases. At the coldest temperatures, very few
disruptive uphill moves are permitted and the state is close
to freezing into its final form. A state with lowest energy
5

Smart Mater. Struct. 19 (2010) 125013

S Liu and Z Lin

Figure 11. The optimum time-dependent control voltage of each

channel for case 2.
Figure 8. Comparisons of the achieved and desired shape at time
250 ms for case 1.

Figure 9. The square difference of the displacement between the

desired shape and the actual shape according to different numbers of
channels.

Figure 12. Optimum channel distribution to achieve twisting shape

of composite plate for case 2. The numbers labeled indicate the
channel distribution indicator.

triangular function of time, the iteration history is shown in

figure 10. The optimum time-dependent control voltage of
each channel and the optimum voltage channel distribution are
plotted in figures 11 and 12. The voltage variations illustrated
in figure 11 are almost the same as those defined by the
desired time functions except in the vicinities of the nonsmooth points. The time-dependent deflections at point A are
illustrated in figure 13, from which we can see that the desired
dynamic shape is achieved with high precision. Figure 14
shows the relative error between the actuated and desired shape
at 200 ms, which shows the desired twisting shape is matched
well at the specified moment.

Figure 10. The iterative process of twisting shape control with

triangular function of time.

difference of the displacement between the desired shape and

the actuated shape according to different numbers of channels.
We can see that control performance gets better with the
increase in channel number. However, the improvement gets
smaller. In fact, as we show in the previous example, high
accuracy has been achieved even with just six voltage channels.
It is worth noting that fewer numbers of voltage channels
should be used for designing efficient and cost-effective control
systems.
Case 2. Tracking twisting shape with triangular functions
of time. When the composite plate is desired to move in a

6. Conclusion
This paper presents an investigation into constrained dynamic
shape control of a composite plate with PZT actuators. The
PZT actuators are controlled by limited numbers of voltage
6

Smart Mater. Struct. 19 (2010) 125013

S Liu and Z Lin

voltage history of each channel in the dynamic shape tracking

of smart structures. The motion equation at time ti+1 in the
Houbolt numerical integration method is given by

[A0 ]{U(ti+1 )} + [A1 ]{U(ti )} + [A2 ]{U(ti1 )} + [A3 ]{U(ti2 )}

]{(t)}
= {F(ti+1 )} [Ku ][D
(A.1)
[M]
11[C]
5[M]
3[C]
where [A0 ] = 2t
2 + 6t + [Kuu ], [A1 ] = t 2 t ,
4[M]
3[C]
[M]
[C]
[A2 ] = t 2 + 2t , [A3 ] = t 2 3t .
By combining motion equations for the k th segment of
times (ti+1 , ti+2 , . . . , ti+n ), motion equations of the Houbolt
numerical integration method can be written as

]{kn }+({Fkn }[B3k1 ]{U3k1 }) (A.2)

[Akn ]{Ukn } = [Kkun ][D

Figure 13. The deflections of the plate twisting at point A for case 2.

where

A0
A1

A2
[Akn ] =
A3

[ 0]

A0
A1
A2
..
.

A0
A1
A3

A0
A2

..

[ 0]
,

A1

A0

k
[Kun ] =
[ 0]

Ku
Ku
Ku

..

[ 0]
,

Ku

..
A3
A2
A1 .

A3
A2 ..

.
.

A3 .
[ 0]
,
[B3k1 ] =

.
.. [0]

..

.
..
[ 0]
.

Figure 14. Comparisons of the achieved and desired shape at time

200 ms for case 2.

channels. The aim of the present design is to find the optimal

control parameters and the optimal distribution of voltage
channels for tracking a desired structural shape. A twolevel method composed of a simulated annealing algorithm
and a sequential linear least-squares algorithm is proposed
for solving the optimization model with discrete variables
and continuous variables coexisting. Numerical examples
verify that the proposed method is effective in solving the
design optimization problem for structural shape morphing
with limited numbers of voltage channels.

{Fkmn } = {{Fm (ti+1 )}, {Fm (ti+2 )}, . . . , {Fm (ti+n )}}T ,

Acknowledgments

{U3k1 } = {{U(ti2 )}, {U(ti1 )}, {U(ti )}, . . . , {0}}T .

{Ukn } = {{U(ti+1 )}, {U(ti+2 )}, . . . , {U(ti+n )}}T ,

{kn } = {{(ti+1 )}, {(ti+2 )}, . . . , {(ti+n )}}T ,

The computed shape for the k th segment of times

(ti+1 , ti+2 , . . . , ti+n ) can be expressed as

This work is supported by the National Basic Research

Program of China (no. 2006CB601205), the National Natural
Science Foundation of China through grant nos. (90816025
and 10721062) and the Research Fund for the Doctoral
Program of Higher Education of China (no. 20090041110023).
The financial contributions are gratefully acknowledged.

]{kn }
{wkcn } = [Rku n ][Akn ]1 ([Kkun ][D
+ ({Fkn } [B3k1 ]{U3k1 }))

(A.3)

where [Rku n ] = diag[Ru(i+1) , Ru(i+2) , . . . , Ru(i+n) ].

For convenient formulation, let [RkVn ] denote [Rku n ]
k 1
[An ] [Kkun ], which is an influence coefficient matrix and
{wkcn } denote [Rku n ][Akn ]1 ({Fkn } [B3k1 ]{U3k1 }). Then we
can express equation (A.3) as

Appendix. SLLS method for a given channel

distribution
In this appendix, for a given channel distribution, we derive a
sequential linear least-squares algorithm to obtain the optimum

kn } + {wkc }.
{wkcn } = [RkVn ][ D]{
n
7

(A.4)

Smart Mater. Struct. 19 (2010) 125013

S Liu and Z Lin

The desired shapes at times (ti+1 , ti+2 , . . . , ti+n ) can be written

as

{wkdn } = {{wd (ti+1 )}, {wd (ti+2 )}, . . . , {wd (ti+n )}}T .

[8] Hsu C Y, Lin C C and Gaul L 1997 Shape control of composite

plates by bonded actuators with high performance
configuration J. Reinf. Plast. Compos. 16 1692710
[9] Chee C, Tong L and Steven G 2000 A buildup voltage
distribution (BVD) algorithm for shape control of smart
plate structures Comput. Mech. 26 11528
[10] Liew K M, He X Q and Ray T 2004 On the use of
computational intelligence in the optimal shape control of
functionally graded smart plates Comput. Methods Appl.
Mech. Eng. 193 447592
[11] Chen W, Wang D and Li M 2004 Static shape control
employing displacementstress dual criteria Smart Mater.
Struct. 13 46872
[12] Merkhujee A and Joshi S 2002 Piezoelectric sensors and
actuators spatial design for shape control of piezolaminated
plates AIAA J. 40 120410
[13] Shaik Dawood M S I, Iannucci L and Greenhalgh E S 2008
Three-dimensional static shape control analysis of composite
plates using distributed piezoelectric actuators Smart Mater.
Struct. 17 025002
[14] Liu S, Tong L and Lin Z 2008 Simultaneous optimization of
control parameters and configurations of PZT actuators for
morphing structural shapes Finite Elem. Anal. Des.
44 41724
[15] Forster E and Livne E 2000 Integrated design optimization of
strain actuated structures for dynamic shape control
AIAA-2000-1366
[16] Luo Q and Tong L 2006 A sequential linear least square
algorithm for tracking dynamic shapes of smart structures
Int. J. Numer. Methods Eng. 67 6688
[17] Luo Q and Tong L 2007 A segment based sequential least
squares algorithm with optimum energy control for tracking
the dynamic shapes of smart structures Smart Mater. Struct.
16 151726
[18] Youssef H, Sait S M and Adiche H 2001 Evolutionary
algorithms, simulated annealing and tabu search: a
comparative study Eng. Appl. Artif. Intell. 14 16781
[19] Kirkpatrick S, Gelatt C D Jr and Vecchi M P 1983 Optimization
by simulated annealing Science 220 67180
[20] Wang L 2001 Intelligent Optimization Algorithms with
Applications (Beijing: Tsinghua University Press)
[21] Luo Q and Tong L 2006 High precision shape control of plates
using orthotropic piezoelectric actuators Finite Elem. Anal.
Des. 42 100920

(A.5)

The error function over a time period can be defined by

E({kn }) = {{wkcn } {wkdn }}T {{wkcn } {wkdn }}

]{kn } + {wkc } {wkd }}T {[RkV ][D
]{kn }
= {[RkV ][D
n

+ {wkcn } {wkdn }}.

(A.6)

The optimal voltages are then obtained by imposing

E
=0
{kn }

(A.7)

which gives

]T [RkV ]T [RkV ][D

]{kn } = [RkV ]T ({wkd } {wkc }). (A.8)
[D
n
n
n
n
n

References
[1] Alkhatib R and Golnaraghi M F 2003 Active structural
vibration control: a review Shock Vib. Dig. 35 36783
[2] Chopra I 2002 Review of state of art of smart structures and
integrated systems AIAA J. 40 214587
[3] Lin M, Qing X, Kumar A and Beard S J 2001 Smart layer and
smart suitcase for structural health monitoring applications
Proc. SPIE 4332 98106
[4] Hurlebausa H and Gaul L 2006 Smart structure dynamics
Mech. Syst. Signal Process. 20 25581
[5] Irschiik H 2002 A review on static and dynamic shape control
of structures by piezoelectric actuation Eng. Struct. 24 511
[6] Frecker M I 2003 Recent advances in optimization of smart
structures J. Intell. Mater. Syst. Struct. 14 20716
[7] Koconis D B, Kollar L P and Springer G S 1994 Shape control
of composite plates and shells with embedded actuators II:
desired shape specified J. Compos. Mater. 28 26285