Universit

e Libre de Bruxelles
F a c u l t y o f A p p l i e d S c i e n c e s
Finite Element Modelling of Piezoelectric
Active Structures
Vincent Piefort
Thesis submitted in candidature for the
degree of Doctor in Applied Sciences
Academic Year 2000-2001
Active Structures Laboratory
Department of Mechanical Engineering and Robotics
Finite Element Modelling of Piezoelectric Active Structures i
To my son Makhno
ii
Finite Element Modelling of Piezoelectric Active Structures iii
Acknowledgements
I would like rst to express my gratitude to my supervisor Prof. Andre Preumont, head of the
Active Structures Laboratory for revealing me as a researcher ten years ago and for his guidance
and encouragements throughout this work.
I would also like to thank Philippe Jetteur from Samtech s.a. without the kind assistance of
whom the achievements of this work would not have been possible.
My gratitude also goes to my colleagues for their moral support and specially to Nicolas, who was
the rst member of the teamto be faced to piezoelectric shells modelling during the CFIE project,
and to Arnaud and Pierre who brought to my attention some other problems in vibroacoustics
control systems modelling.
Many thanks also to Kris Henrioulle (KUL-PMA) for submitting to my attention the ASAC
panel modelling problem.
I am also most grateful to my parents, to Valerie and to all my friends for helping me not to be
too much of an engineer.
This thesis nalizes a ve year research in the Active Structures Laboratory of Universite Libre de
Bruxelles. This work was funded by Region Wallonne - Division Generale des Techniques, de la
Recherche et de lEnergie (Conventions 95/2792 and 97/13549). The support of the IUAP-4/24
on Intelligent Mechatronic Systems (Belgian programme on InterUniversity Attraction Poles
initiated by the Belgian State, Prime Ministers Oce, Science Policy Programming) is also
acknowledged.
iv
Finite Element Modelling of Piezoelectric Active Structures v
Foreword: motivation of this work
In 1993, at the request of ESA, ULB developed a laboratory demonstration model of a very
exible active plate (35 cm60 cm) controlled by PZT piezoceramics; it was later transformed
into a ight model (to be own in a canister) by our industrial partner in this project Spacebel
Instrumentation and the experiment (named CFIE: Control-Flexibility Interaction Experiment),
was successfully own by NASA in the space shuttle ENDEAVOUR in September 1995 (Loix,
1998). According to the specications, the experiment had to t into a GAS (Get Away Special)
Figure 1: Laboratory demonstration model of the CFIE experiment.
canister (cylinder of 50 cm diameter and 80 cm high), demonstrate signicant gravity eects,
and use the piezoelectric technology. We settled on a very exible steel plate of 0.5 mm thickness
hanging from a support as shown on Fig.1; two additional masses were mounted, as indicated
in the gure, to lower the natural frequencies of the system. The rst mode was in bending and
the second one was in torsion. Because of the additional masses, the structure had a signicant
geometric stiness due to the gravity loads (responsible for a rise of the rst natural frequency
from 0.5 Hz in zero gravity to 0.9 Hz with gravity). The nite element model of the structure
in the gravity eld could be updated to match the experimental results on the ground, but the
in-orbit natural frequencies could only be predicted numerically and were therefore subject to
uncertainties. At the time, the model used by Loix (1998) was based on the simplied approach
using the equivalent piezoelectric loads. The CFIE experiment motivated the research work on
the nite element modelling of piezoelectric shells and, as a consequence, initiated the present
thesis.
vi
The need for an accurate modelling of local eects in nearly colocated control systems can
be illustrated by the experiment of Fig.2 (Loix & Preumont, 1995). In this experiment, a
cantilever beam is tted with four piezoelectric patches, indicated as p
1
to p
4
on the gure. p
1
is
taken as actuator and the frequency response functions are compared when the sensor is taken
successively as p
2
, p
3
and p
4
. Although the three situations are equivalent if a beam model is
considered in conjunction with equivalent piezoelectric loads, we observe substantial dierences
in the location of the zeros which in turn, produces substantial dierences in the performance
of the closed-loop control system. The objective of this work was to develop a general tool able
to predict accurately these situations.
Frequency response functions (p
1
taken as actuator)
Figure 2: Cantilever plate with nearly colocated actuator/sensor piezoceramics: experiment
In order to resolve the technological issues related to the practical realisation of a nearly colo-
cated actuator/sensor system using bonded piezoceramics, a simple cantilever plate was used
(Loix, 1998). Important dierences in the experimental frequency response functions for similar
actuator/sensor positions appeared (Fig.2) that can not be explained by a classical beam model.
More specically, the experiment of Fig.2 pointed out the importance of local membrane eects
that can only be predicted by a rigorous shell nite elements model; tis was the third motivation
of this research work.
Finite Element Modelling of Piezoelectric Active Structures vii
Summary
This thesis is organized in ve chapters.
Chapter 1 (Introduction) is an introduction to piezoelectric materials. It contains an historical
overview of piezoelectricity, some general considerations for a better understanding of piezoelec-
tic materials, and a brief overview of other smart materials.
Chapter 2 (Constitutive Equations) is dedicated to the constitutive equations of piezoelectricity.
It contains a unidimensional presentation of piezoelectric constitutive equations using a elec-
tromagnetic approach. The general thermopiezoelectric constitutive equations are established
starting from thermodynamics principles, and simplied for a Kirchho laminate embedding
piezoelectric layers.
Chapter 3 (Actuation and Sensing) introduces the various designs used in piezoelectric actuation
and sensing devices. It contains a review of the dierent piezoelectric modes of actuation/sensing.
Simplied laminar design modelling approaches are presented and their limitations stressed.
Some examples of electrode shaping to achieve targeted control devices are analyzed.
Chapter 4 (Finite Element Approach) presents an electromechanically coupled nite element
formulation. It contains a review of nite element modelling of piezoelectric structures and
a general electromechanically coupled nite element formulation starting from the constitutive
equations and the Hamilton variational principle. A Kirchho piezoelectric multilayered element
formulation is developed and its extension to a Mindlin formulation is described. Its actual im-
plementation into the commercial nite element package Samcef (Samtech s.a.) is presented
together with the volume elements implementation. Some particular electrical boundary condi-
tions are illustrated, and a method to extract a state space model of a piezoelectric input/output
system from the modal nite element analysis is presented.
Chapter 5 (Applications) contains the description of some applications of the developed tools:
actuation, sensing, vibroacoustics, control. The importance of the in-plane components in the
open-loop frequency response functions of nearly colocated control systems is illustrated.
viii
Finite Element Modelling of Piezoelectric Active Structures ix
Contents
1 Introduction 1
1.1 History of piezoelectricity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Piezoelectric materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.1 Origins of electric polarization . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.2 Bi-ionic crystal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.3 Ferroelectric ceramics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.4 Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 Other Smart materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.1 Shape memory alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.2 Electro / magneto-strictive materials . . . . . . . . . . . . . . . . . . . . . 15
1.3.3 Electro / magneto-rheological uids . . . . . . . . . . . . . . . . . . . . . 17
2 Constitutive equations 19
2.1 Unidimensional . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.1.1 Polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.1.2 Piezoelectricity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.2 General constitutive equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.2.1 Thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.2.2 Linear piezoelectricity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3 Piezoelectric laminate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.3.1 Single layer in plane stress . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.3.2 Multilayer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3 Actuation and sensing 35
3.1 Piezoelectric modes of actuation . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
x Contents
3.2 Laminar design modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2.1 Historical overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2.2 Pure bending beam model . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2.3 Isotropic shell model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.2.4 Why is the beam theory not enough to model colocated systems ? . . . . 46
3.3 Spatial lters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.3.1 Modal lters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.3.2 Volume velocity sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4 Finite element approach 51
4.1 History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.2 Variational principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.3 Finite element formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.4 Kirchho element . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.5 Mindlin element . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.6 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.6.1 Shell implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.6.2 Transverse piezoelectric shear mode in shell elements . . . . . . . . . . . . 65
4.6.3 Volume implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.7 Particular electrical boundary conditions . . . . . . . . . . . . . . . . . . . . . . . 68
4.7.1 Voltage driven electrodes . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.7.2 Charge driven electrodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.7.3 Electrodes connected via a passive network . . . . . . . . . . . . . . . . . 68
4.8 From nite element model to state space model . . . . . . . . . . . . . . . . . . . 69
5 Applications 73
5.1 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.1.1 Inuence of the electromechanical coupling on the natural frequencies . . 73
5.1.2 Cantilever piezoelectric plate . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.1.3 Shear bender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.2 Actuation and sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.2.1 Bimorph beam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.2.2 C-blocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
Finite Element Modelling of Piezoelectric Active Structures xi
5.2.3 Membrane eects on nearly colocated control systems . . . . . . . . . . . 84
5.3 Vibroacoustics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.3.1 ASAC plate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.3.2 Alternative cross-ply design of the ASAC plate . . . . . . . . . . . . . . . 91
5.3.3 Array sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.4 Design against fatigue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6 Conclusions 97
A Transverse shear in beams and plates 99
A.1 Beams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
A.1.1 Geometry and kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
A.1.2 Constitutive equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
A.2 Plates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
A.2.1 Constitutive equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
A.2.2 Transverse shear stiness approximation . . . . . . . . . . . . . . . . . . 106
B Longitudinal vibration of a piezoelectric bar 109
C Some data les 115
C.1 Example of a dynamic analysis using shell elements: Inuence of the electrome-
chanical coupling on the natural frequencies . . . . . . . . . . . . . . . . . . . . . 115
C.2 Cantilever piezoelectric plate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
D Matlab toolbox 119
D.1 From nite element to state space . . . . . . . . . . . . . . . . . . . . . . . . . . 119
D.2 Visualization tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
D.3 Other tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
xii Contents
Finite Element Modelling of Piezoelectric Active Structures xiii
List of Figures
1 Laboratory demonstration model of the CFIE experiment. . . . . . . . . . . . . . v
2 Cantilever plate with nearly colocated actuator/sensor piezoceramics: experiment vi
1.1 Origins of polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Structure of the Cd S crystal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Strain change associated with the polarization reorientation . . . . . . . . . . . . 6
1.4 PZT: electric eld / strains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5 Perovskite structure in PZT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.6 Eect of doping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.7 Unidirectional and bidirectional PVDF . . . . . . . . . . . . . . . . . . . . . . . . 10
1.8 The martensitic transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.9 Shape memory eect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.10 Pseudo-elasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.11 Preloaded magnetostrictive material . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.12 Strain / magnetic eld curves for Tb
0.3
Dy
0.7
Fe
1.93
(Moett et al., 1991) . . . . . 16
1.13 Piezoelectricity (PZT ceramics) - Electrostriction (PMN ceramics) . . . . . . . . 17
1.14 Particles alignment due to an applied eld . . . . . . . . . . . . . . . . . . . . . . 17
2.1 Charges in a dielectric-lled plane capacitor . . . . . . . . . . . . . . . . . . . . . 20
2.2 Thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.3 Symmetries in the piezoelectric coupling matrix for all crystal classes - Part I
(Nye, 1957, chap. 7) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.4 Symmetries in the piezoelectric coupling matrix for all crystal classes - Part II
(Nye, 1957, chap. 7) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.5 Composite shell with embedded piezoelectric patch . . . . . . . . . . . . . . . . . 29
2.6 Lamina and laminate coordinate systems . . . . . . . . . . . . . . . . . . . . . . 30
xiv LIST OF FIGURES
2.7 Resultant eorts (Kirchho) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.8 Multilayered material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.1 Common piezoelectric actuator designs . . . . . . . . . . . . . . . . . . . . . . . . 36
3.2 Shear actuation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.3 Bimorph actuation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.4 Piezoelectric lm actuator bonded on a beam structure. . . . . . . . . . . . . . . 39
3.5 Piezoelectric lm actuator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.6 Piezoelectric lm sensor connections . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.7 Pure bending beam model: actuation/sensing duality . . . . . . . . . . . . . . . . 42
3.8 Piezoelectric load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.9 Sensor conguration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.10 Contribution to the output of the piezoelectric isotropic sensor (e
31
= e
32
) . . . . 46
3.11 Equivalent piezoelectric loads for a rectangular piezoelectric patch on a beam . . 46
3.12 Frequency response function of an active cantilever beam with colocated actuator
and sensor. Comparison between the beam theory and the shell theory. . . . . . 47
3.13 QWSIS sensor (Rex & Elliott, 1992) . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.1 Resultant eorts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.2 3-D volume elements from Samcef library . . . . . . . . . . . . . . . . . . . . . . 67
4.3 Electrodes connected to an external impedance . . . . . . . . . . . . . . . . . . . 69
5.1 Simply supported steel plate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.2 FE mesh: 30% grounded . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
5.3 Stiening eect resulting from the piezoelectric coupling . . . . . . . . . . . . . . 74
5.4 FE modal analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.5 Open loop frequency response function . . . . . . . . . . . . . . . . . . . . . . . . 75
5.6 Shear bender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.7 FE mesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.8 Static deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.9 FE mesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.10 Resulting deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.11 Tip deection vs actuator position . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.12 Bimorph pointer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
Finite Element Modelling of Piezoelectric Active Structures xv
5.13 Static deection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.14 Bimorph sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.15 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.16 Comparison between shell and volume solutions . . . . . . . . . . . . . . . . . . . 81
5.17 C-block actuator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.18 FE mesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.19 C-block results: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.20 Cantilever plate with piezoceramics: experimental setup . . . . . . . . . . . . . . 84
5.21 Experimental results: frequency response functions between p
1
and p
i
. . . . . . 84
5.22 Simulation results: frequency response functions between p
1
and p
i
. . . . . . . . 85
5.23 FE meshes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.24 Discretisation of the active plate into narrow quadratic devices . . . . . . . . . . 86
5.25 Piezoelectric beam device . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.26 ASAC experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.27 FE mesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.28 ASAC plate: Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.29 Open loop frequency response functions of the ASAC plate . . . . . . . . . . . . 89
5.30 Eect of the residual mode (in addition to the 20 modelled modes) on the pre-
dicted open-loop frequency response functions . . . . . . . . . . . . . . . . . . . . 90
5.31 FE mesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
5.32 Open-loop frequency response functions . . . . . . . . . . . . . . . . . . . . . . . 91
5.33 Principle of the volume displacement sensor . . . . . . . . . . . . . . . . . . . . . 92
5.34 Array sensor: experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.35 FE mesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.36 Frequency response functions /Shaker #1 . . . . . . . . . . . . . . . . . . . . . . 94
5.37 Piezoelectric elements location and rst vibration modes . . . . . . . . . . . . . . 95
5.38 Damage map (Von Mises) with and without control (Log Scale) . . . . . . . . . . 96
5.39 PSD of the stress components in the center of the plate with and without control 96
A.1 Beam geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
A.2 Displacement of two points p and q . . . . . . . . . . . . . . . . . . . . . . . . . . 101
A.3 Transverse shear stress distribution . . . . . . . . . . . . . . . . . . . . . . . . . . 102
A.4 Resultant eorts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
xvi LIST OF FIGURES
A.5 Displacement of two points p and q . . . . . . . . . . . . . . . . . . . . . . . . . . 104
B.1 Piezoelectric unidimensional bar (cross section: ) . . . . . . . . . . . . . . . . . 110
Finite Element Modelling of Piezoelectric Active Structures xvii
List of Tables
1.1 PZT and PVDF typical properties . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.2 Some eects in materials (o-diagonal coupling smart materials) . . . . . . . . 12
1.3 Shape memory alloys typical properties . . . . . . . . . . . . . . . . . . . . . . . 13
1.4 Comparison of properties of typical ER and MR uids. (Carlson et al., 1995) . . 18
2.1 Matrix notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.1 Comparison between actuation designs . . . . . . . . . . . . . . . . . . . . . . . . 37
4.1 Electromechanical analogy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.1 Characteristics of the materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
5.2 Material properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
5.3 PVDF properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.4 Materials properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.5 Materials properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.6 Inuence of the mesh on the separation between the poles and the zeros for the
conguration (p
1
/p
3
) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.7 Material properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.8 Material properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
xviii LIST OF TABLES
Finite Element Modelling of Piezoelectric Active Structures xix
List of Symbols
[ ] matrix
[ ]
T
matrix transposed
vector

T
vector transposed

,

rst and second time derivatives of

rst and second spatial derivatives of

gradient operator

x
spatial derivation operator
/x spatial derivative of with respect to x

,x
short notation for the spatial derivative of with respect to x
specic heat

i
weights of the adaptive linear combiner
dielectric constant (
1
) (IEEE std)
bre rotation angles

ij
Kronecker index
variation operator
variation (of length), distance, stroke
xx LIST OF SYMBOLS
shear strain (standard engineering notation)
permittivity (IEEE std)

0
permittivity of vacuum ( 8.854 10
12
F/m)

r
relative permittivity (/
0
)
strain (standard engineering notation)
, voltage, electrical potential

i
nodal voltages, electrical potentials
Power Spectral Density
curvature
thermal expansion coecient

k
modal mass of mode k
Poisson ratio

k
natural frequency of mode k
frequency
[] eigen frequencies matrix
surface, area

k
orientation angle between material axes of layer k and the global axes
,
0
temperature, reference temperature
small temperature change
mass density
surfacic charge density
stress (standard engineering notation)
entropy
Finite Element Modelling of Piezoelectric Active Structures xxi
shear stress (standard engineering notation)

k
modal classical damping ratio of mode k
piezoelectric anisotropy ratio
a arbitrary vector
a, b lengths, distances, dimensions
[A] extension stiness matrix
A
s
austenite start temperature (shape memory eect)
A
f
austenite nish temperature (shape memory eect)
[B] extension/bending coupling stiness matrix
[B] shape functions derivatives
c mechanical stiness (IEEE std)
C
0
, C = electrical capacitances
C
st
constant term
[C] structural damping matrix
( contour
d piezoelectric coupling (electrical eld / strain) (IEEE std)
d

electrostrictive coupling (electrical eld / stress)

[D] bending stiness matrix
D electric displacement eld (IEEE std)
[T] derivation operator
[D
HF
] residual mode: static contribution of the unmodelled high frequency modes
to the frequency response function
e piezoelectric coupling (electrical eld / stress) (IEEE std)
xxii LIST OF SYMBOLS
e

electrostrictive coupling (electrical eld / stress)

E electrical eld (IEEE std)
c
k
membrane, bending piezoelectric coupling written in global axes
f
i
element nodal (mechanical) applied resultant forces
F assembled nodal (mechanical) applied resultant forces
F
P
point force
F
V
body force
F

surface force
F
0
uniform load
g piezoelectric coupling (electrical displacement / strain) (IEEE std)
g
i
element nodal (electrical) applied resultant charges
g
i
Gauss weights
G assembled nodal (electrical) applied resultant charges
G thermoelectric Gibbs state function
G, G
12
shear modulus
[(()] frequency response function
H electrical enthalpy density
H magnetic eld
[K] transverse shear stiness matrix
h, h
k
thickness, thickness of layer k
h piezoelectric coupling (electrical displacement / stress) (IEEE std)
I electrical current
I
n
identity matrix of order n
Finite Element Modelling of Piezoelectric Active Structures xxiii
J kinetic energy
k
1
, k
2
ideal spring stiness
[K
uu
] element mechanical stiness matrix
[K
u
] = [K
u
]
T
element piezoelectric coupling matrix
[K

] element capacitance matrix

[K
UU
] assembled mechanical stiness matrix
[K
U
] = [K
U
]
T
assembled piezoelectric coupling matrix
[K

] assembled capacitance matrix

[K
elt
] complemented element stiness matrix
L length
[L] localisation matrix
L Lagrangian
m lineic mass
M
s
martensite start temperature (shape memory eect)
M
f
martensite nish temperature (shape memory eect)
M resultant bending eort
[/] element mass matrix, assembled mass matrix
n vector normal to a contour
N resultant membrane eort
[^] interpolation (shape) functions
p pyroelectric coecient
P
0
, P free polarization, polarization
q elementary electrical charge
xxiv LIST OF SYMBOLS
Q total electrical charge
_
Q

k
stiness matrix of layer k written in the global axes
R electrical resistance
[R
T
] rotation matrix relating the stress written in the orthotropy axes of the
piezo to the stress written in the global axes
[R
S
] rotation matrix relating the strain written in the orthotropy axes of the
piezo to the strain written in the global axes
s mechanical compliance (IEEE std)
S strain (IEEE std)
T stress (IEEE std)
T resultant transverse shear eort
u
0
mid-plane membrane displacement vector
u, v, w displacements relatives to x, y, z respectively
u displacement eld vector
u
i
, U nodal displacements
U stored energy density
V volume
V
vol
volume velocity
J virtual work of external electrical and mechanical forces
x, y, z spatial coordinates
x
k
modal amplitude of mode k
A
k
transverse shear piezoelectric coupling written in global axes
Y Youngs modulus
z
k
coordinate of layer k relative to the mid-plane
Finite Element Modelling of Piezoelectric Active Structures xxv
Z impedance
Z
k
modal shape of mode k
xxvi LIST OF SYMBOLS
Finite Element Modelling of Piezoelectric Active Structures 1
Chapter 1
Introduction
This chapter is divided is three parts. In a rst section, a short history of piezoelectricity is
presented. The second section presents the piezoelectric, pyroelectric and ferroelectric materials.
The third section is a brief overview of dierent types of smart materials.
1.1 History of piezoelectricity
Centuries ago, natives from Ceylan and India already noticed a peculiar property of tourmaline
crystals. Thrown in hot ashes, these crystals rst attract them to reject them a few moments af-
terwards. This experiment came into Europe with the import of tourmaline by dutch tradesmen
at the beginning of the XV III
th
century. The tourmaline was called the Ceylan magnet.
In 1756, the electrical origin of that behaviour was demonstrated by the german physicist Aepi-
nus (electrical capacitance inventor)
1
. That behaviour was named pyroelectricity by the scottish
physicist D.Brewster in 1824. The pyroelectric eect can be dened as the induction of polar-
ization by thermal energy absorption; the induced polarization is proportional to the resulting
temperature variation. The inverse property, of much less amplitude, is called the electrocaloric
eect.
The piezoelectric eect was rst mentioned in 1817 by the french mineralogist Rene Just Ha uy
2
.
It was rst demonstrated by Pierre and Jacques Curie in 1880. Their experiments led them
to elaborate the early theory of piezoelectricity. This theory was complemented by the further
work of G.Lippman
3
, W.G.Hankel
4
, Lord Kelvin and W.Voigt (beginning of XXth century).
Until the beginning of the century, the piezoelectricity did not leave the laboratories. The rst
applications appeared during the rst world war with the sonar in which piezoelectric quartz
are used to produce ultrasonic waves (P.Langevin) and as sensors. In the twenties, the use of
quartz to control the resonance frequency of oscillators was proposed by an american physicist:
W. G. Cady. It is during the period following the rst world war that most of the piezoelectric
1
First observation of the electrical polarization of the tourmaline crystal when its temperature changes
2
First observation of the presence of electric charges on the surface of a stressed tourmaline crystal
3
Mathematical deduction of the inverse piezoelectric eect, conrmed experimentally by P. & J. Curie: 1881
4
W.G.Hankel introduced the term piezoelectricity
2 1. Introduction
applications we are now familiar with (microphones, accelerometers, ultrasonic transducers,
benders, ...) were conceived. However, the materials available at the time often limited device
performance. The development of electronics, specially during the second world war, and the
discovery of ferroelectric ceramics increased the use of piezoelectric materials.
The direct piezoelectric eect consists of the ability of certain crystalline materials (i.e. ceram-
ics) to generate an electrical charge in proportion of an externally applied force. The direct
piezoelectric eect has been widely used in transducers design (accelerometers, force and pres-
sure transducers, ...). According to the inverse piezoelectric eect, an electric eld induces a
deformation of the piezoelectric material. The inverse piezoelectric eect has been applied in
actuators design.
The use of piezoelectric materials as actuators and sensors for noise and vibration control has
been demonstrated extensively over the past few years (e.g. Forward, 1981; Crawley & de Luis,
1987). There are two classes of piezoelectric materials used in vibration control: ceramics and
polymers. The best known piezoceramic is the Lead Zirconate Titanate (PZT); it has a recov-
erable strain of 0.1% and is widely used as actuator and sensor for a wide range of frequencies,
including ultrasonic applications; it is well suited for high precision as well. Piezopolymers are
mainly used as sensors; the best known is the Polyvinylidene Fluoride (PVDF). The PVDF were
rst studied by Kawai (end of the 60s) and were made commercially available in the early 80s.
1.2 Piezoelectric materials
The piezoelectric eects can be seen as transfers between electrical and mechanical energy. Such
transfers can only occur if the material is composed of charged particles and can be polarized.
For a material to exhibit an anisotropic property such as piezoelectricity, its crystal structure
must have no centre of symmetry (See e.g. van Randeraat & Setterington, 1974). 21 crystal
structures out of 32 are non-centrosymmetric. A crystal having no center of symmetry possesses
one or more crystallographically unique directional axes. All 21 non-centrosymmetric crystal
classes, except 1, show piezoelectric eect along the directional axes. Out of the 20 piezoelectric
classes, 10 have only one unique direction axis. Such crystals are called polar crystals as they
show spontaneous polarization. The value of the spontaneous polarization depends on the
temperature. This is called the pyroelectric eect. The pyroelectric crystals for which the
magnitude and direction of the spontaneous polarization can be reversed by an external electric
eld are said to show ferroelectric behavior.
Most of the piezoelectric materials are crystalline solids. They can be single crystals, either
formed naturally or by synthetic processes, or polycrystalline materials like ferroelectric ceramics
which can be rendered piezoelectric and given, on a macroscopic scale, a single crystal symmetry
by the process of poling (by subjecting to a high electric eld not far below the Curie temperature
- see 1.2.3). The piezoelectric eect can also appear in crystals composed of only one type of
element (in this case, the polarization is due to a distortion in the electronic distribution).
Certain polymers can also be made piezoelectric by stretching under an electrical eld.
The origins of electric polarization are rst presented. The molecular nature of piezoelectricity
and the main coupling coecients are introduced next using a simple example: the bi-ionic
monocrystal Cd S (Brissaud, 1986). The more complex mechanisms involved in ferroelectric
Finite Element Modelling of Piezoelectric Active Structures 3
ceramics are described last.
1.2.1 Origins of electric polarization
A material that can be polarized under an electrical eld is called a dielectric. Three origins for
the macroscopic polarization of a dielectric material can be distinguished (Fig.1.1).
electronic polarization: When an electric eld is applied, the electronic clouds deform cre-
ating a macroscopic dipole.
ionic polarization: When an electrical eld is applied, anions are attracted by the anode
and cations by the cathode, creating dipoles
dipole reorientation: A macroscopically non polarized material can be composed of many
polarized domains. An applied electric eld will cause the domains to reorient, inducing a
global polarization
Figure 1.1: Origins of polarization
1.2.2 Bi-ionic crystal
An intuitive explanation of the piezoelectricity can be given by a simple unidimensional example.
Let us consider a simple bi-ionic crystal.
The Cd S crystal structure can be seen as an array of identical rows composed of sub-structures
such as described in Fig.1.2 where cations Cd (charge +q) and anions S (charge q) alternate.
The interatomic cohesion forces are modelled by ideal springs of dierent stiness k
1
and k
2
,
the distance between consecutive pairs of atoms being dierent.
If the crystal is composed of n basic elements of length a, as shown on Fig.1.2, including 2
electrical dipoles
q
2
(a b) and
q
2
b, the global free polarization will be given by
P
0
= n
q
2
(a 2b) (1.1)
4 1. Introduction
Figure 1.2: Structure of the Cd S crystal
For a constrained crystal (along the axis of alignement), the interatomic distances change and
a polarization P is induced; this is the direct piezoelectric eect:
P = P
0
= n
q
2
(a 2b) (1.2)
Conversely, when an electric eld E is applied along the same axis, the ions move and a global
deformation is induced; this is the inverse piezoelectric eect. The static equilibrium relation
for each ion can be written:
qE +k
1
b k
2
(a b) = 0 (1.3)
The induced polarization is related to the electrical eld and to the atomic displacement by:
P = n
q
2
_
2q
k
1
+k
2
E +
k
1
k
2
k
1
+k
2
a
_
(1.4)
which can be written
P =
ion
E +eS (1.5)
where

ion
= n
q
2
2q
k
1
+k
2
(1.6)
is called the ionic polarizability of the crystal,
S =
a
a
(1.7)
is the strain, and
e = na
q
2
k
1
k
2
k
1
+k
2
(1.8)
is the piezoelectric constant.
The electric displacement (induced polarization) can be written
D = eS +
S
E (1.9)
Finite Element Modelling of Piezoelectric Active Structures 5
where the permittivity at constant strain
S
is given by

S
=
0
+
ion
(1.10)
with
0
is the permittivity of the vacuum
The stress induced in a unitary section perpendicular to the considered axis, assuming a total
of N rows on a unitary section and assuming an equal number of springs of each type k
1
and
k
2
in such a section, is given by:
T =
N
2
k
1
b +k
2
N
2
(a b) (1.11)
Considering that we have N = na and using Equ.(1.3) and (1.7), one gets:
T = na
2
k
2
k
1
k
2
k
1
+k
2
S na
q
2
k
1
k
2
k
1
+k
2
E (1.12)
which can be written
T = c
E
S eE (1.13)
where
c
E
= na
2
k
2
k
1
k
2
k
1
+k
2
(1.14)
is the mechanical stiness at a constant electric eld, and e is the piezoelectric constant given by
Equ.(1.8). It is worth noticing that c
E
is related to the stiness of the two springs k
1
and k
2
in
series and that the piezoelectric constant e is zero if the springs are of equal stiness. Equations
(1.9) and (1.13) are the piezoelectric constitutive equations.
However, ions are never connected by ideal springs. In most cases, these springs are anharmonic
(F

= k+k

2
). A fraction of the stress induced is independent of the direction of the applied
eld.
T = c
E
S eE e

E
2
(1.15)
This eect is called the electrostrictive eect, e

is the electrostrictive coecient. (See 1.3.2)

1.2.3 Ferroelectric ceramics
Domains reorientation - hysteresis
To gain an understanding of the piezoelectric eect in ferroelectric materials (See e.g. van Ran-
deraat & Setterington, 1974; Uchino, 2000), we must rst consider the behaviour of the material
on a microscopic scale. Above a certain temperature, called the Curie temperature (
c
), the
crystal structure of a ferroelectric material does have a centre of symmetry and has therefore
no electric dipole moment. Below this temperature, it undergoes a phase change to a more
6 1. Introduction
complex structure which is non-centrosymmetric. In this phase, the crystal presents a natu-
ral electric dipole (pyroelectricity) which may be reversed (ferroelectricity) and also switched in
certain allowed directions by the application of a suciently high electric eld.
Ferroelectric crystals possess regions with uniform polarization called ferroelectric domains.
Within a domain, all the electric dipoles are aligned in the same direction. There may be
many domains in a crystal separated by interfaces called domain walls. A ferroelectric single
crystal, when grown, has multiple ferroelectric domains in each of which the electric dipole is
aligned in a specic allowed direction. As each of the allowed direction has the same probability
to appear, the net electric dipole summed over the whole crystal is zero.
Nevertheless, when the crystal is cooled down in the presence of an electric eld, the domains
tend to align in the allowed direction nearest to the electric eld. The crystal as a whole presents
an electric dipole.
If this crystal is subjected to stress, the lattice will be distorted and the stress will also cause
some domains to grow at the expense of others. This results in a change in the total dipole
moment of the crystal. Within a certain range of stress, this variation of dipole moment with
the stress is approximately linear and reversible.
Ferroelectric components may be made piezoelectric in any chosen polar direction by the poling
treatment which involves exposing the material to a high electric eld at a temperature not far
below the Curie point.
Because of the random orientation of the domains and the fact that only certain dipole directions
are allowed within the crystal, it is not possible to get the perfect dipole alignment with the eld.
However, there are several allowed directions within every domain and so a reasonable degree of
alignment with the eld is possible. After cooling of the product and removal of the poling eld,
the dipoles cannot easily return to their original positions, and we have now what is known as
Figure 1.3: Strain change associated with the polarization reorientation
Finite Element Modelling of Piezoelectric Active Structures 7
remanent polarization of the material. The body has become permanently piezoelectric and can
convert mechanical energy into electrical energy, and vice versa.
Fig.1.3 shows schematically the domain reorientation in a multidomain ferroelectric piezoce-
ramic. The material is initially poled along the negative direction (1) and an electric eld is
applied along the positive direction. The crystal will rst shrink with the increase of the eld as
the eld is opposite in direction to the polarization. The strain reaches a minimum at a certain
eld (coercive eld E
c
), where the polarization starts to reverse in each grain (2). Above E
c
,
the crystal expands until E
max
(3) as the eld has now the same direction as the polarization.
Near E
max
, all the reversible polarization have been reversed. As the eld is reduced, the strain
decreases monotonically as no polarization reversal occurs. The situation for a zero electric eld
(4) is similar to the starting situation except that the polarization is reversed; the material is
now poled along the positive direction.
Fig.1.4 shows a typical electric eld / strain curves (directions x
3
and x
1
, parallel and perpendic-
ular to the eld) for a PZT based ferroelectric piezoceramic. In a cycle with a small maximum
electric eld, the eld-induced strain curve is almost linear (a). The curve becomes distorted
as the electrical eld increases and shows a larger hysteresis (b, c, d) and nally transforms into
a symmetric buttery shape when the electric eld exceeds the coercive eld; this is caused by
the polarization due to dipole reorientation.
Many piezoelectric (including ferroelectric) ceramics such as Barium Titanate (Ba Ti O
3
), Lead
Titanate (Pb Ti O
3
), Lead Zirconate Titanate (PZT) and Lead Lanthanum Zirconate Titanate
Figure 1.4: PZT: electric eld / strains
8 1. Introduction
Figure 1.5: Perovskite structure in PZT
(PLZT) have a perovskite type structure (Fig.1.5). Perovskite is the family name of a group of
materials having the same structure as the calcium titanate (Ca Ti O
3
)
5
.
Doping eect in PZT
The Lead Zirconate Titanate (PZT) is a binary solid solution of Pb Zr O
3
and Pb Ti O
3
. The
PZT (Pb(Zr
y
Ti
1y
)O
3
) is composed of a lattice of basic cells having the perovskite structure
(Fig.1.5). The electrical properties of ferroelectric ceramics can be modied by substituting ions
of dierent valence in the lattice.
By doping the crystal with impurities (replacing some Zr
4+
or Ti
4+
ions), a local deciency is
introduced (Fig.1.6).
Figure 1.6: Eect of doping
Acceptor ions (i.e. Fe
3+
) introduce oxygen deciencies ().
Pb(Zr
y
Ti
1yx
Fe
x
)(O
3x/2

x/2
)
Acceptor doping induces a higher mobility of the oxygen ions; They can easily jump to ll in
the adjacent deciency. This implies an easy reorientation of the deciency related dipoles.
The hysteresis becomes more important while the strain/electric eld ratio is lowered. Such
a piezoeceramics is called hard piezoceramic. Hard PZTs usually have lower permittivities,
5
Perovskite is the mineral name of calcium titanate
Finite Element Modelling of Piezoelectric Active Structures 9
smaller electrical losses and lower piezoelectric coecients. These are more dicult to pole and
depole, thus making them ideal for rugged applications such as ultrasonic motors.
Donor ions (i.e. Nb
5+
) introduce Pb deciencies ().
(Pb
1x/2

x/2
)(Zr
y
Ti
1yx
Nb
x
)O
3
Pb ions cannot easily hop to the nearest deciencies due to the oxygen surroundings. Dipoles
cannot easily reorient; The hysteresis is lowered and the strain/electric eld ratio is increased.
We obtain a so called soft piezoceramic. The soft PZTs have a higher permittivity, larger losses,
higher piezoelectric coecient and are easy to pole and depole. They can be used for applications
requiring very high piezoelectric properties such as actuation and sensing applications.
PLZT is a transparent ferroelectric ceramic formed by replacing Pb
2+
ions by La
3+
ions in PZT.
The transparent nature of PLZT has led to its use in electro-optic applications.
1.2.4 Polymers
Piezoelectricity can be obtained by orientating the molecular dipoles of polar polymers such
as Polyvinylidene Fluoride (PVDF or PVF
2
) in the same direction. The PVDF can be made
piezoelectric because uorine is much more electronegative than carbon. The uorine atoms
will attract electrons from the carbon atoms to which they are attached. The CF
2
groups
in the chain will be very polar so when they are placed in an electrical eld, they will align.
Conversely, when the piezopolymer deforms, a macroscopic dipole appears.
This can be obtained by submitting the lm of polymer to a suciently high electric eld after
a mechanical stretching. The obtained polarization is mainly due to the spatial rearrangement
of polar segments of the macromolecular chains.
By machining the material in one or two perpendicular directions prior to the polarization pro-
cess, dierent piezoelectric behaviours can be obtained (Fig.1.7). A uni-axial stretching will
induce nearly unidirectional piezoelectric properties. A bi-axial stretching will induce piezoelec-
tric properties that are isotropic in the plane.
Table 1.1 presents typical PVDF characteristics (Piezotech Documentation, 1997) compared to
PZT ones.
10 1. Introduction
Mechanical stretching Piezoelectric activity
(a) Uni-axial lm
(b) Bi-axial lm
Figure 1.7: Unidirectional and bidirectional PVDF
Material properties PZT PVDF
uni-axial bi-axial
Piezoelectric constants
d
33
(10
12
Cb/N or m/V) 300 -25 -25
d
31
(10
12
Cb/N or m/V) -150 15 3
d
32
(10
12
Cb/N or m/V) -150 3 3
Relative permittivity /
0
1800 12 12
(
0
= 8.854 10
12
F/m)
Youngs modulus (GPa)
Y
1
50 3 3
Y
2
50 1 3
Y
3
50 10 10
Maximum stress in traction (MPa)
direction 1 80 200 200
direction 2 80 40 200
Max. operating Temp. (C) 140 90 90
Max. Electric eld (V/m) 1 10
6
500 10
6
500 10
6
Density (kg/m
3
) 7600 1800 1800
Table 1.1: PZT and PVDF typical properties
Finite Element Modelling of Piezoelectric Active Structures 11
1.3 Other Smart materials
The best known among the smart materials are certainly the piezoelectric materials and more
precisely the piezoceramics. However, several others are available. Table 1.2 summarizes the
dierent eects and couplings existing in materials considering the dierent conjugated physical
elds classically involved.
What characterize a smart or active material ? Couplings that are o-diagonal in Table 1.2 are
typically responsible for the so called smart behaviour of materials. A material can be said smart
if one of its smart coupling is (or can be made) suciently important to be used in active devices.
The few next paragraphs presents two types of smart materials (shape memory alloys and
magneto/electro-strictive materials) and a semi-smart one (magneto/electro-rheological uids).
1.3.1 Shape memory alloys
The Shape Memory Eect is the eect by which a material, apparently plastically deformed,
recovers its original undeformed shape after heating. It initiates in the solid phase change
occuring in metals and, more precisely, in the martensitic transformation. Two types of alloy
exhibit a strong shape memory eect: Copper Alloys (Cu-Zn-Al and Cu-Al-Ni) and Nitinols
(Ni-Ti). Their main typical characteristics are listed in Table 1.3. The main advantage of Shape
Memory Alloys is the possibility to achieve complex movements with very few elements, and
this, with a small temperature change despite the hysteresis (See e.g. Perkins, 1975; Liang, 1990;
Piefort, 1991).
Thermally induced martensitic transformation
Some alloys have interesting properties after quenching. Let us consider an unstressed binary
alloy. At high temperature, this alloy exhibit a structure . After a rapid cooling (quenching
process), it will keep its structure in a metastable equilibrium with a structure

. Let us call,
by analogy with iron, and

, respectively austenitic and martensitic structures.

Figure 1.8: The martensitic transformation
The martensitic transformation between austenitic and martensitic structures in shape mem-
ory alloys occurs without atomic diusion, by small successive displacements. This process
12 1. Introduction
O
u
t
p
u
t
S
t
r
a
i
n
C
h
a
r
g
e
M
a
g
n
e
t
i
c
T
e
m
p
e
r
a
t
u
r
e
L
i
g
h
t
I
n
p
u
t

e
l
d
C
u
r
r
e
n
t

e
l
d
S
t
r
e
s
s
E
l
a
s
t
i
c
i
t
y
P
i
e
z
o
-
I
n
v
e
r
s
e
P
h
o
t
o
-
e
l
e
c
t
r
i
c
i
t
y
m
a
g
n
e
t
o
s
t
r
i
c
t
i
o
n
e
l
a
s
t
i
c
i
t
y
E
l
e
c
t
r
i
c
a
l
I
n
v
e
r
s
e
P
i
e
z
o
-
P
e
r
m
i
t
t
i
v
i
t
y
E
l
e
c
t
r
o
-
E
l
e
c
t
r
o
-
E
l
e
c
t
r
o
-

e
l
d
e
l
e
c
t
r
i
c
i
t
y
C
o
n
d
u
c
t
i
v
i
t
y
m
a
g
n
e
t
i
c
e

e
c
t
c
a
l
o
r
i
c
e

e
c
t
o
p
t
i
c
e

e
c
t
I
n
d
u
c
t
i
o
n
M
a
g
n
e
t
o
-
M
a
g
n
e
t
o
-
P
e
r
m
e
a
b
i
l
i
t
y
M
a
g
n
e
t
o
-
M
a
g
n
e
t
o
-
s
t
r
i
c
t
i
o
n
e
l
e
c
t
r
i
c
e

e
c
t
c
a
l
o
r
i
c
e

e
c
t
o
p
t
i
c
e

e
c
t
H
e
a
t
T
h
e
r
m
a
l
e
x
p
a
n
s
i
o
n
P
y
r
o
-
S
p
e
c
i

c
S
h
a
p
e
m
e
m
o
r
y
e
l
e
c
t
r
i
c
i
t
y
h
e
a
t
L
i
g
h
t
P
h
o
t
o
-
P
h
o
t
o
v
o
l
t
a
i
c
R
e
f
r
a
c
t
i
v
e
s
t
r
i
c
t
i
o
n
e

e
c
t
i
n
d
e
x
A
c
t
u
a
t
i
o
n
S
e
n
s
i
n
g
T
a
b
l
e
1
.
2
:
S
o
m
e
e

e
c
t
s
i
n
m
a
t
e
r
i
a
l
s
(
o

-
d
i
a
g
o
n
a
l
c
o
u
p
l
i
n
g

s
m
a
r
t
m
a
t
e
r
i
a
l
s
)
Finite Element Modelling of Piezoelectric Active Structures 13
Properties Units Cu-Zn-Al Cu-Al-Ni Ni-Ti
Young modulus GPa 70 to 100 80 to 100 80 to 90
Transformation temperature C -200 to 120 -200 to 170 -220 to 110
Hysteresis (A
s
A
f
) C 10 to 20 20 to 30 30 to 50
Recoverable strain %
Simple SME 5 6 8
Double SME
for 10
2
cycles 1 1.2 6
for 10
5
cycles 0.8 0.8 4
Pseudo-elastic strain % 10 10 10
Table 1.3: Shape memory alloys typical properties
induces changes in the elementary crystal shape. Nevertheless, no macroscopic change is ob-
served due to the existence of 24 variants of martensite (only 1 for austenite) so the shape
changes compensate inside each grain of material. When a sample is cooled down, its crys-
tals transform from -austenitic structure into

-martensitic structures without macroscopic

deformation; this transformation is characterized by two temperatures: M
s
(Martensite Start)
at which the transformation begins, and M
f
(Martensite Finish) at which all the austenite has
been transformed into martensite. When reheated, all the

-martensitic variants retransform

into the same -austenitic structure; this transformation is characterized by two temperatures:
A
s
(austenite Start) at which the transformation begins, and A
f
(Austenite Finish) at which all
the martensite has been transformed back into austenite. The four characteristic temperatures
of the martensitic thermal transformation M
s
, M
f
, A
s
, and A
f
are represented on Fig.1.8.
Stress induced martensitic transformation
When a shape memory alloy at a temperature above M
s
(pure austenite) is submitted to a stress
eld, the martensitic transformation occurs at higher temperature. This is the stress induced
martensitic transformation. The most important characteristic of the stress induced martensite
is that the dierent variants have no longer the same chance to appear. The variants giving the
minimum stress state will be favoured: the stress induced martensite is oriented.
Associated eects
Rubber Eect: When a shape memory alloy at a temperature below M
f
(consisting in
dierent variants of martensite) is submitted to a stress eld, the best oriented variants
will grow at the expense of the others (

transformation). After removing the

stress at a temperature below M
f
, the stress induced martensite is stable. The material
recovers its elastic deformation and a fraction of the martensite reorients, giving an extra
strain recovery; this is the rubber eect (1 2% )
Memory eect: After raising the temperature of a plastically deformed SMA above A
f
, all
the martensite transform back into austenite and the material recovers its warm shape. If
cooled back below M
f
, no more shape change is observed as all variants of martensite can
14 1. Introduction
develop equivalently; this is the simple memory eect (Fig.1.9).
Figure 1.9: Shape memory eect
Education: If, during the cooling phase, the material is forced into a cold shape, and the
process is repeated a few times, dislocations and internal residual stress state appear,
favouring the formation of certain variants of martensite: the material has gained the
double memory eect and takes spontaneously its cold or warm shape depending on the
temperature: the shape memory alloy is educated
Pseudoelasticity The stress induced martensite at a temperature above A
f
retransformS
into austenite when unstressed, giving the superelastic behaviour to shape memory alloys
(Fig.1.10). A number of products have been brought to market that use the pseudoelastic
(or superelastic) property. (i.e. Eyeglass frames that use superelastic Ni Ti alloy to absorb
large deformations without damaging the frames are now marketed).
Figure 1.10: Pseudo-elasticity
Finite Element Modelling of Piezoelectric Active Structures 15
1.3.2 Electro / magneto-strictive materials
Magnetostriction is the process by which a ferromagnetic material transforms from one shape to
another in the presence of a magnetic eld
6
. Most ferromagnetic materials exhibit some measur-
able magnetostriction. Conversely, if an external force produces a strain in a magnetostrictive
material, the materials magnetic state will change
7
. This bi-directional magnetomechanical
coupling of a magnetostrictive material provides a transduction capability that is used for both
actuation and sensing devices.
This solid state phenomenon is a result of the rotation of small magnetic domains causing internal
strains in the material inducing an expansion in the eld direction. As the eld is increased,
more domains tend to align with the magnetic eld until magnetic saturation is achieved. If the
eld is reversed, the direction of the domains is also reversed but the strains still result in an
expansion. Since the magnetostriction has a molecular origin, the response is very fast (a few
s).
Figure 1.11: Preloaded magnetostrictive material
Magnetostrictive materials are typically mechanically biased in normal operation. A compressive
load is applied to the material, which, due to the magneto-elastic coupling, forces the domain
structure to orient perpendicular to the applied force. Then, as a magnetic eld is introduced,
the domain structure rotates producing the maximum possible strain in the material (Fig.1.11).
The highest room temperature magnetostriction of a pure element does not exceed 60 microstrain
(Cobalt). By alloying elements, one can achieve giant magnetostriction under relatively small
elds. Terfenol-D
8
(alloys of the form Tb
x
Dy
1x
Fe
2
) exhibits the greatest magnetostrictive
eects of any commercially available material. Operated under a mechanical-bias, it strains
to about 2000 microstrain in a magnetic eld of 2 kOe at room temperature (1 A-turn/m=
4 10
3
Oe).
6
The discovery of magnetostriction is attributed to J. Joule - XIXth century
7
This reciprocal eect is known as the Villari eect
8
The name Terfenol-D is derived from Terbium (Tb); Fe; the Naval Ordnance Laboratory (it was rst discov-
ered in the 70s by a research group led by Clark at the Naval Surface Warfare Center in Silver Spring, Md); and
Dysprosium (Dy).
16 1. Introduction
Figure 1.12: Strain / magnetic eld curves for Tb
0.3
Dy
0.7
Fe
1.93
(Moett et al., 1991)
The magnetostrictive properties of Terfenol-D are strongly dependent on the magnetic and
mechanical bias conditions (Moett et al., 1991). Figure 1.12 shows extension strain / magnetic
eld curves for dierent prestress conditions.
As the the compressive prestress is increased, larger values of eld bias as well as larger drive
eld are required. The coupling factor decreases with increasing magnetic and mechanical bias.
At high drive levels, saturation occurs and the coupling factor becomes independent of the bias
conditions.
Electrostriction is the dependance of the state of strain of a ferroelectric material to the even
powers of the applied electric eld (mainly the second order)(Anderson et al., 1990). It is a
universal property of all dielectrics but it can only be observed in materials with high dielectric
coecients such as PMN (Lead Magnesium Niobate) which exhibits an electrostrictive strain up
to 0.1% (Fig.1.13).
D = dT +
T
E
S = sT +dE +d

E
2
(1.16)
where D is the electrical displacement, E the electric eld, S the strains, T the stresses, d the
piezoelectric coupling, s the compliance and d

the electrostrictive coupling (the dE term is

usually neglected in this case).
Finite Element Modelling of Piezoelectric Active Structures 17
Figure 1.13: Piezoelectricity (PZT ceramics) - Electrostriction (PMN ceramics)
1.3.3 Electro / magneto-rheological uids
Magnetorheological (MR) uids and electro-rheological (ER) uids are materials that respond
to an applied magnetic or electric eld with a important change in rheological behaviour. This
is more a semi-smart behaviour in the sense that the application of a third party eld (here
electric or magnetic) will act on a classical coupling (viscosity) and that there is no reciprocal
eect.
No eld Applied eld
Figure 1.14: Particles alignment due to an applied eld
These uids are non-colloidal suspension of polarisable small particles (a few m). Their essen-
tial characteristic is their ability to reversibly change from a free-owing, linear viscous liquid
to a semi-solid with a controllable yield strength in milliseconds when exposed to a magnetic or
electric eld. In the absence of an applied eld, controllable uids are reasonably well approxi-
18 1. Introduction
mated as Newtonian liquids (Carlson & Sproston, 2000).
The magneto-rheological response results from the polarization induced in the suspended par-
ticles by application of an external eld. The interaction between the resulting induced dipoles
causes particles to align and form chain-like structures parallel to the applied eld (Fig.1.14).
These chain-like structures restrict the motion of the uid, thereby increasing its apparent vis-
cosity. The mechanical energy needed to break these chain-like structures increases with the
applied eld. For most engineering applications a simple Bingham (Eq.1.17) plastic model is
sucient for describing the essential, eld-dependent uid characteristics. In this model, the
total yield stress is given by:
= (H)sign( ) +
p
(1.17)
where (H) is the yield stress caused by the applied magnetic or electric eld H, is the shear
rate and
p
is the eld independent plastic viscosity dened as the slope of the measured shear
stress versus shear strain rate relationship.
Interest in controllable uids stems from their ability to provide simple, quiet, rapid-response
interfaces between electronic controls and mechanical systems. The ability of controllable uids
to be directly used as fast-acting, uid valves with no moving parts in semi-active vibration
control systems has been one of the main motivating factors for the development of such uids.
Typical characteristics for magneto and electro - rheological uids are listed in Table 1.4. It
must be noted that the maximum shear stress obtainable using MR uids is about 20 times
bigger than the maximum shear stress obtainable using ER uids.
Property ER uid MR uid
Yield Strenght (eld) 2 5 kPa (3 5 kV/mm) 50 100 kPa (150 250 kA/m)
Viscosity (no eld) 0.2 0.3 Pa.s at 25 C 0.2 0.3 Pa.s at 25 C
Operating temperature 25 to +125 C 40 to +150 C
Response time ms ms
Density 1 2 g/cm
3
3 4 g/cm
3
Table 1.4: Comparison of properties of typical ER and MR uids. (Carlson et al., 1995)
Finite Element Modelling of Piezoelectric Active Structures 19
Chapter 2
Constitutive equations
In a rst section, the unidimensional constitutive equations are established starting from the
electrostatics point of view. The general thermopiezoelectric constitutive equations are derived
from the laws of thermodynamics in the second section. Linear constitutive equations of piezo-
electricity and further simplications of the coupling matrices due to crystal symmetries are
presented in the third section. The particular case of plane stress state is considered in the last
section of this chapter.
2.1 Unidimensional
2.1.1 Polarization
The electric charge Q appearing on 2 parallel conductive plates separated by vacuum is related
to the applied voltage
0
by the capacitance C
0
= Q/
0
of the system. Substituting the
vacuum by an insulating material and suppressing the electrical source at the same time, a
voltage <
0
is observed (Faradays experiment, 1837). The total electric charge on the
plates being unchanged, this implies that the capacitance of the system has been increased:
C = Q/ > C
0
= Q/
0
.
The capacitance is related to the surface of the plates and to distance d between them by
C
0
=
0
/d for the vacuum dening
0
= (3610
9
)
1
F/m, the permittivity of the vacuum.
Similarily, we have C = /d for an ideal dielectric, dening =
r

0
, the absolute permittivity
of the dielectric and
r
, its relative one.
To explain Faradays experiment, the following hypothesis has been made: In an ideal dielectric
material, the polarization is induced only by the applied electric eld.
Applying a voltage
0
induces a uniform electric eld E, the charge appearing is Q = (
being the charge density on the surface).
When the dielectric is inserted, it gets polarized under the electrostatic eld. Two kinds of
charges can be distinguished (Fig.2.1): free charges Q
1
and bound charges Q
2
. The bound
charges Q
2
being locally compensated by the polarization charges inside the dielectric, only the
20 2. Constitutive equations
free charges Q
1
contributes to the voltage .
=
Q
C
=
Q

r
C
0
=
Q
1
C
0
(2.1)
The total charge is given by:
Q = Q
1
+Q
2
=
Q

r
+Q
_
1
1

r
_
(2.2)
and the charge density
=

r
+
_
1
1

r
_
(2.3)
Figure 2.1: Charges in a dielectric-lled plane capacitor
Let us dene:
D = displacement eld (2.4)
P =
_
1
1

r
_
polarization (2.5)
E =

d
=
Q

r
C
0
1
d
=

0
=

electric eld (2.6)

Which leads to
D = E =
0
E +P (2.7)
and
P = (
0
) E (2.8)
2.1.2 Piezoelectricity
In an unstressed one-dimensional dielectric medium, the dielectric displacement D (charge per
unit area, expressed in Cb/m
2
) is related to the electric eld E (V/m) and the polarization P
(Cb/m
2
) by Equ.(2.7).
Finite Element Modelling of Piezoelectric Active Structures 21
Similarly, in a one-dimensional elastic body placed in a zero electric eld, the stress T (N/m
2
)
and the strain S are related by
T = cS (2.9)
where c is the stiness of the material (Youngs modulus).
For a piezoelectric material, the electrical and mechanical constitutive equations are coupled. A
strain S in the material induces a polarization eS by the direct piezoelectric eect. The total
induced polarization is given by
P = (
0
)E +eS (2.10)
Conversely, an applied electric eld E tends to align the internal dipoles, inducing stresses e E
in the material by the inverse piezoelectric eect. The coupled equations nally become:
T = c
E
S e E
(2.11)
D = e S +
S
E (2.12)
In equation (2.11), the piezoelectric constant e relates the stress to the electric eld E in the
absence of mechanical strain and c
E
refers to the stiness when the electric eld is constant. In
equation (2.12), e relates the electric charge per unit area D to the strain under a zero electric
eld (short-circuited electrodes); e is expressed in NV
1
m
1
or Cb/m
2
.
S
is the permittivity
under constant strain.
Equation (2.11) is the starting point for the formulation of the equation of a piezoelectric actu-
ator, while equation (2.12) is that for a sensor.
In the next section, the general constitutive equations for a thermopiezoelectric continuum
are rst established starting from the basic thermodynamic principles. The linear constitutive
equations for a piezoelectric material are written next, introducing standard engineering tensorial
notations.
2.2 General constitutive equations
2.2.1 Thermodynamics
The conservation of energy for the linear piezoelectric continuum results in the rst law of
thermodynamics considering three contributions to the variation of the stored internal energy:
the work of external forces, the work of applied electric elds and the thermal energy brought
to the system.
dU = T
ij
dS
ij
+E
i
dD
i
+d (2.13)
where T
ij
and S
ij
are respectively the stress and the strain tensors, E
i
and D
i
are respectively
the electric eld and the electric displacement vectors, =
0
+ , the temperature (
0
,
22 2. Constitutive equations
the reference temperature and
0
, a small temperature change), the entropy and U is
the stored energy density for the piezoelectric continuum. The summation convention for all
repeated indices is assumed.
Let us write the constitutive equations using S, E and as independant variables. We have
dT
ij
=
T
ij
S
kl
dS
kl
+
T
ij
E
m
dE
m
+
T
ij

d (2.14)
dD
i
=
D
i
S
kl
dS
kl
+
D
i
E
m
dE
m
+
D
i

d (2.15)
d =

S
kl
dS
kl
+

E
m
dE
m
+

d (2.16)
dening (See Fig.2.2)
T
ij
S
kl
elasticity
T
ij
E
m
inverse piezoelectricity
T
ij

thermal stress
D
i
S
kl
direct piezoelectricity
D
i
E
m
permittivity
D
i

pyroelectricity

S
kl
piezocaloric eect

E
m
electrocaloric eect

heat capacity
To describe the behaviour of the thermopiezoelectric continuum as a thermodynamic system,
the thermoelectric Gibbs state function (free energy) G is introduced:
G = U E
i
D
i
(2.17)
from equations (2.13) and (2.17), there results
dG = T
ij
dS
ij
D
i
dE
i
d (2.18)
and therefore,
T
ij
=
G
S
ij
(2.19)
D
i
=
G
E
i
(2.20)
=
G

(2.21)
by deriving a second time, one gets:
T
ij
E
m
=
D
m
S
ij
(2.22)
Finite Element Modelling of Piezoelectric Active Structures 23
T
ij

=

S
ij
(2.23)
D
i

=

E
i
(2.24)
establishing that the coecients relative to direct and inverse eects are equals. Let us dene
T
ij
S
kl
= c
ijkl
T
ij
E
m
= e
ijm
T
ij

=
ij
D
i
S
kl
= e
kli
D
i
E
m
=
im
D
i

= p
i

S
kl
=
kl

E
m
= p
m

=
Considering a linear behaviour, G can be written
G =
1
2
c
E
ijkl
S
ij
S
kl
e
kij
E
k
S
ij

1
2

S
ij
E
i
E
j

1
2

ij
S
ij
p
i
E
i
(2.25)
c
ijkl
, e
kij
,
ij
,
ij
, p
i
and are respectively the elastic, the piezoelectric, the thermal expansion
coecient, the dielectric and pyroelectric constants and the specic heat. From Equ.(2.25), and
Equ.(2.19) to Equ.(2.21), there results the thermopiezoelectric constitutive equations:
T
ij
= c
E
ijkl
S
kl
e
kij
E
k

ij
(2.26)
D
m
= e
mkl
S
kl
+
S
mk
E
k
+p
i
(2.27)
=
kl
S
kl
+p
S
k
E
k
+ (2.28)
2.2.2 Linear piezoelectricity
An important characteristic of piezoeletric materials compared to other smart materials is its
linear behavior within a certain range.
In this section, we use the same notations as in (IEEE std) and only the piezoelectric coupling
is considered (the thermoelectric terms are neglected). In linear piezoelectricity, the equations
of linear elasticity are coupled to the charge equation of electrostatics by the means of the
piezoelectric constants. The quasielectrostatic approach is adequate because the phase velocities
of acoustic waves are several order of magnitude less than the velocities of electromagnetic waves.
Taking advantage of the symmetries of the mechanical tensors, a compressed matrix notation
is introduced in place of the tensor notation. This matrix notation consists of replacing indices
ij or kl by p or q according to Table 2.1 (the non diagonal terms of the deformation tensor are
24 2. Constitutive equations
doubled). The following identications are made:
c
ijkl
c
pq
(2.29)
e
ikl
e
iq
(2.30)
T
ij
T
p
(2.31)
S
ij
S
p
when i = j (2.32)
2S
ij
S
p
when i ,= j (2.33)
ij or kl p or q
11 1
22 2
33 3
23 or 32 4
13 or 31 5
12 or 21 6
Table 2.1: Matrix notation
The constitutive equations read
T = [c
E
]S [e]
T
E
D = [e]S + [
S
]E
(2.34)
or (alternate forms using alternative choices of independant variables)
S = [s
E
]T + [d]
T
E
D = [d]T + [
T
]E
(2.35)
S = [s
D
]T + [g]
T
D
E = [g]T + [
T
]D
(2.36)
T = [c
D
]S [h]
T
D
E = [h]S + [
S
]D
(2.37)
with
T =
_
T
11
T
22
T
33
T
23
T
13
T
12
_
T
is the stress vector
S, the deformation vector
E, the electric eld vector
D, the electric displacement vector
[c] and [s], the elasticity constants matrices
Finite Element Modelling of Piezoelectric Active Structures 25
[] and [], the dielectric constants matrix
[d], [e], [g] and [h], the piezoelectric constants matrix
and supercripts D, E, S and T indicate values at D, E, S and T constant respectively
The element d
ij
of [d] represents the coupling between the electric eld in the direction i (if a
poling occured, its direction is taken as direction 3) and the strain in the j direction. S
j
= d
ij
E
i
.
The following relations between dielectric, elastic, and piezoelectric constants are veried
_
c
E
_
s
E

=
_
c
D
_
s
D

= I
6
(2.38)
_

S
_

=
_

T
_

= I
3
(2.39)
_
c
D

=
_
c
E

+[e]
T
[h] (2.40)
_
s
D

=
_
c
D

[d]
T
[g] (2.41)
_

=
_

+[d]
T
[e] (2.42)
_

=
_

[g]
T
[h] (2.43)
[e] = [d]
_
c
E

(2.44)
[d] =
_

[g] (2.45)
[g] = [h]
_
s
D

(2.46)
[h] =
_

[e] (2.47)
Due to crystal symmetries, the piezoelectric coupling matrices [d] and [e] may have only few non
zero elements (Cady, 1946; Nye, 1957). Symmetries in the piezoelectric coupling matrix for all
crystal classes are described on Fig.2.3 and Fig.2.4.
Examples of piezoelectric coupling matrices:
Quartz [d]
Quartz
=
_
_
2.3 2.3 0 0.67 0 0
0 0 0 0 0.67 4.6
0 0 0 0 0 0
_
_
10
12
m/V
PZT [d]
PXE5
=
_
_
0 0 0 0 515 0
0 0 0 515 0 0
175 175 362 0 0 0
_
_
10
12
m/V
PVF
2
[d]
Solef
=
_
_
0 0 0 0 0 0
0 0 0 0 0 0
23 3 33 0 0 0
_
_
10
12
m/V
(Solef Mono-oriented 130 m)
26 2. Constitutive equations
Figure 2.2: Thermodynamics
Finite Element Modelling of Piezoelectric Active Structures 27
Figure 2.3: Symmetries in the piezoelectric coupling matrix for all crystal classes - Part I (Nye,
1957, chap. 7)
28 2. Constitutive equations
Figure 2.4: Symmetries in the piezoelectric coupling matrix for all crystal classes - Part II (Nye,
1957, chap. 7)
Finite Element Modelling of Piezoelectric Active Structures 29
2.3 Piezoelectric laminate
Figure 2.5: Composite shell with embedded piezoelectric patch
We consider a shell structure with embedded piezoelectric patches covered with electrodes. The
poling direction and the electric and displacement elds direction are parallel, normal to the
patches (Fig.2.5). The piezoelectric patches are parallel to the mid-plane and orthotropic in
their plane.
The following hypothesis are made:
One dimension is signicantly less than the others
Plane stress hypothesis:
zz
= 0, using classical engineering notations
Kirchho assumption: A ber normal to the mid plane remains normal to the mid plane
after deformation
S = S
0
+z (2.48)
where S
0
is the mid plane strain and , the mid plane curvature. The strain and
stress vectors written in axes (xy) are given in classical engineering notation respectively
by S =
x

y

xy

T
and T =
x

y

xy

T
. The Kirchho hypothesis does not
account for any transverse shear (
yz
=
xz
= 0).
Uniform electric eld and displacement across the thickness and aligned on the normal
to the mid-plane (direction 3).
E =
_
_
_
0
0
E
_
_
_
(2.49)
D =
_
_
_
0
0
D
_
_
_
(2.50)
Linear piezoelectricity for each piezoelectric layer k, (we assume that the piezoelectric
principal axes are parallel the structural orthotropy axes and that the poling direction is
30 2. Constitutive equations
direction 3). The constitutive equations for the k
th
layer become (e
36
is assumed to be
zero, which is the case for most commonly used piezomaterials in laminar design: PZT
and PVDF - see 2.2.2)
T = [c]
k
S
_
_
_
e
31
e
32
0
_
_
_
k
E
k
(2.51)
D
k
=
_
e
31
e
32
0
_
k
S +
k
E
k
(2.52)
where [c]
k
is the elastic coecients matrix of the k
th
layer (stiness matrix in the principal
material axes).
2.3.1 Single layer in plane stress
Figure 2.6: Lamina and laminate coordinate systems
Let us rst consider a single layer at the purely mechanical point of vue (stress/strain relation-
ship). An orientation angle
k
between the material axes LT of layer k and the structural axes
xy is assumed (Fig.2.6).
k
is taken positive from x to L. The transformation matrices [R
T
]
k
and [R
S
]
k
relate the stresses and strains written in the material axes (LT) to the stresses and
strains written in the composite axes (xy). We have:
_
_
_

LT
_
_
_
= [R
T
]
k
_
_
_

xy
_
_
_
(2.53)
_
_
_

T
1
2

LT
_
_
_
= [R
T
]
k
_
_
_

y
1
2

xy
_
_
_
(2.54)
_
_
_

LT
_
_
_
= [R
S
]
k
_
_
_

xy
_
_
_
(2.55)
Finite Element Modelling of Piezoelectric Active Structures 31
with the transformation matrices
[R
T
]
k
=
_
_
cos
2

k
sin
2

k
2 sin
k
cos
k
sin
2

k
cos
2

k
2 sin
k
cos
k
sin
k
cos
k
sin
k
cos
k
cos
2

k
sin
2

k
_
_
(2.56)
[R
S
]
k
=
_
_
cos
2

k
sin
2

k
sin
k
cos
k
sin
2

k
cos
2

k
sin
k
cos
k
2 sin
k
cos
k
2 sin
k
cos
k
cos
2

k
sin
2

k
_
_
(2.57)
As we can write
_
_
_

xy
_
_
_
= [R
S
]
1
k
_
_
_

LT
_
_
_
(2.58)
It follows that [R(
k
)]
1
= [R(
k
)]
The stiness matrix
_
Q

k
of the k
th
layer in the composite axes is dened by
_
_
_

xy
_
_
_
=
_
Q

k
_
_
_

xy
_
_
_
(2.59)
Since
_
_
_

xy
_
_
_
= [R
T
]
1
k
_
_
_

LT
_
_
_
= [R
T
]
1
k
[c]
k
_
_
_

LT
_
_
_
= [R
T
]
1
k
[c]
k
[R
S
]
k
_
_
_

xy
_
_
_
(2.60)
It is related to the elastic coecients matrix [c]
k
by:
_
Q

k
= [R
T
]
1
k
[c]
k
[R
S
]
k
(2.61)
As we have [R
T
]
1
k
= [R
S
]
T
k
and [c]
k
symmetric,
_
Q

k
is also symmetric.
Let us now introduce the piezoelectric coupling terms. The electric eld is related to the voltage

k
across the thickness h
k
of the layer k:
E
k
=
k
/h
k
(2.62)
Starting from the constitutive equation (2.51) for the piezoelectric layer k in the material prin-
cipal axes (LT), we get, in composite axes (xy):
T
(xy)
= [R
T
]
1
k
T
(LT)
= [R
T
]
1
k
[c]
k
[R
S
]
k
S
(xy)
[R
T
]
1
k
_
_
_
e
31
e
32
0
_
_
_
k
E
k
=
_
Q

k
S
(xy)
+[R
T
]
1
k
_
_
_
e
31
e
32
0
_
_
_
k

k
h
k
(2.63)
32 2. Constitutive equations
2.3.2 Multilayer
A laminate is formed from two or more layers bonded together to act as a single layer mate-
rial. The bond between two layers is assumed to be perfect, so that the displacements remain
continuous across the bond. The classical theory for multilayered materials is followed and
complemented with electromechanical coupling terms.
Figure 2.7: Resultant eorts (Kirchho)
Figure 2.8: Multilayered material
In-plane eorts N and bending moments M (Fig.2.7) can be determined by integrating the
stresses over the thickness of the multilayered material (Agarwal & Broutman, 1990, Chap.6).
N =
_
+
h
2

h
2
T dz
M =
_
+
h
2

h
2
Tz dz
Finite Element Modelling of Piezoelectric Active Structures 33
Integrating equation (2.63) over the thickness, one gets the resultant normal eort and bending
moment:
_
N
M
_
=
_
A B
B D
_ _
S
0

_
+
n

k=1
_
z
k
z
k1
_
I
3
z I
3
_
[R
T
]
1
k
_
_
_
e
31
e
32
0
_
_
_
k

k
h
k
dz (2.64)
where n is the number of piezoelectric layers and z
k
the coordinate of layer k relative to the
mid-plane as dened on Fig.2.8.
The extension, bending and coupling matrices A, B, and D are given by the classical relationships
for multilayered materials:
[A] =
n

k=1
_
Q

k
(z
k
z
k1
) extension (2.65)
[B] =
1
2
n

k=1
_
Q

k
(z
2
k
z
2
k1
) coupling (2.66)
[D] =
1
3
n

k=1
_
Q

k
(z
3
k
z
3
k1
) bending (2.67)
Putting all the constant terms out of the integral sign, the second term of Equ.(2.64) becomes:
. . . +
n

k=1
1
h
k
_
_
z
k
z
k1
_
I
3
z I
3
_
dz
_
[R
T
]
1
k
_
_
_
e
31
e
32
0
_
_
_
k

k
= . . . +
n

k=1
_
I
3
z
mk
I
3
_
[R
T
]
1
k
_
_
_
e
31
e
32
0
_
_
_
k

k
(2.68)
where
z
mk
=
z
k1
+z
k
2
(2.69)
is the distance from the mid-plane of layer k to the mid-plane of the composite (Fig.2.8).
The second constitutive equation (2.52), giving the electric displacement for the k
th
layer, reads
in the composite axes (xy)
D
k
=
_
e
31
e
32
0
_
k
[R
S
]
k
S
(xy)

k
h
k
=
_
e
31
e
32
0
_
k
[R
S
]
k
_
I
3
z I
3

_
S
0

k
h
k
(2.70)
By integrating Equ.(2.70) over the thickness, one gets
_
z
k
z
k1
D
k
dz =
_
z
k
z
k1
_
e
31
e
32
0
_
k
[R
S
]
k
_
I
3
z I
3

_
S
0

_
dz
_
z
k
z
k1

k
h
k
dz (2.71)
34 2. Constitutive equations
As we assumed D uniform across the thickness, it can be averaged.
D
k
=
_
e
31
e
32
0
_
k
[R
S
]
k
_
I
3
z
mk
I
3

_
S
0

k
h
k

k
(2.72)
In summary, the constitutive equations integrated over the thickness of a multilayered piezo-
electric shell (Kirchho assumption) read:
_
N
M
_
=
_
A B
B D
_ _
S
0

_
+
n

k=1
_
I
3
z
mk
I
3
_
[R
T
]
1
k
_
_
_
e
31
e
32
0
_
_
_
k

k
(2.73)
D
k
=
_
e
31
e
32
0
_
k
[R
S
]
k
_
I
3
z
mk
I
3

_
S
0

k
h
k

k
(2.74)
In the next chapter, the use of piezoelectric materials for actuation and sensing (mainly laminar
design) is discussed. Equation (2.73) is the starting point for the formulation of the equation
of a laminar piezoelectric actuator, while Equ.(2.74) is that for a laminar sensor. It is worth
noticing that, as we have [R
T
]
1
k
= [R
S
]
T
k
, the system of equations above is symmetric.
Finite Element Modelling of Piezoelectric Active Structures 35
Chapter 3
Actuation and sensing
The use of piezoelectric materials for actuation and sensing has been demonstrated extensively
over the years. In a rst section, dierent piezoelectric modes of actuation available are pre-
sented. Models neglecting the electromechanical coupling for the particular case of the laminar
design are presented in the second section; the duality between actuation and sensing is stressed.
Examples of the use of distributed actuation/sensing piezoelectric layers to realize targeted con-
trol devices are presented in the last section.
3.1 Piezoelectric modes of actuation
Due to crystal symmetries, piezoelectric coupling matrices have few non-zero elements (See
Chap.2). The design of actuation and sensing devices is dictated by the available coupling modes
(non-zero elements in the piezoelectric coupling matrix - See typical constitutive equations, of
form (2.35), below).
actuation:
_

_
S
11
S
22
S
33
2S
23
2S
13
2S
12
_

_
=
_

_
s
11
s
12
s
13
0 0 0
s
12
s
22
s
23
0 0 0
s
13
s
23
s
33
0 0 0
0 0 0 s
44
0 0
0 0 0 0 s
55
0
0 0 0 0 0 s
66
_

_
. .
compliance
_

_
T
11
T
22
T
33
T
23
T
13
T
12
_

_
+
_

_
0 0 d
31
0 0 d
32
0 0 d
33
0 d
24
0
d
15
0 0
0 0 0
_

_
. .
coupling
_
_
_
E
1
E
2
E
3
_
_
_
sensing:
_
_
_
D
1
D
2
D
3
_
_
_
=
_
_
0 0 0 0 d
15
0
0 0 0 d
24
0 0
d
31
d
32
d
33
0 0 0
_
_
. .
coupling
_

_
T
11
T
22
T
33
T
23
T
13
T
12
_

_
+
_
_

11
0 0
0
22
0
0 0
33
_
_
. .
permittivity
_
_
_
E
1
E
2
E
3
_
_
_
36 3. Actuation and sensing
The most commonly used piezo materials for actuation and sensing are the piezoceramics (PZT)
and the piezopolymers (PVDF). The available modes of actuation for a piezoelectric material
are determined by its coupling matrix. The coupling matrices of PZT and PVDF look like:
[d]
PZT
=
_
_
0 0 0 0 d
15
0
0 0 0 d
15
0 0
d
31
d
31
d
33
0 0 0
_
_
(d
32
= d
31
, d
24
= d
15
)
[d]
PVDF
=
_
_
0 0 0 0 0 0
0 0 0 0 0 0
d
31
d
32
d
33
0 0 0
_
_
We can therefore anticipate that three main modes of actuation/sensing are available: the in-
plane mode (d
31
, d
32
), the thickness mode (d
33
) and the shear mode (d
15
, d
24
).
The rst two congurations mentioned above are the most commonly used (Fig.3.1). A linear
(a) stacked design (d
33
coupling)
= d
33
n
(b) laminar design (d
31
coupling)
= d
31

h
Figure 3.1: Common piezoelectric actuator designs
actuator consists of a stack of thin ceramic disks separated by electrodes. The material is such
that the coecient d
33
dominates the other piezoelectric constants in the constitutive equations.
As a result, the direction of expansion coincides with that of the electric eld. The stacked
design is often referred to as d
33
design. When no external load is applied, the change of length
is related to the voltage applied by the approximate relationship = d
33
n where n is the
number of disks in the stack and is the applied voltage. This design is mainly used for precision
position control.
Finite Element Modelling of Piezoelectric Active Structures 37
In the laminar design, centre of interest of this thesis, thin piezoelectric lms are bonded on
the structure to obtain a bending action. The geometrical arrangement is such that d
31
(d
32
)
dominates the design and the useful direction of expansion is normal to that of the electric eld.
The piezoelectric constant d
31
(d
32
), however, is only about one half of d
33
in the stacked design.
In the laminar design, the piezoelectric material may consist of ceramics like lead-zirconate
titanate (PZT), generally covered with silver electrodes, or polymers like polyvinylidene uoride
(PVDF or PVF
2
). As mentioned earlier, PZT materials are piezo-isotropic in the plane normal to
the poling direction (d
32
= d
31
) while PVDF can be strongly anisotropic (See Table 1.1, 1.2.4).
As it will be stressed throughout this chapter, it is the shape of the electrodes which determines
the eective part of the material. This property is widely used for distributed actuation and
sensing.
A third actuator concept is proposed by Benjeddou et al. (1997, 1998): the shear actuation.
The proposed conguration is such that, this time, the d
15
(d
24
) coupling coecient dictates the
design. The electric eld is applied perpendicularly to the poling direction, inducing a transverse
shear strain. A shear actuated piezo layer can be included in the core of a sandwich plate as in
Fig.3.2. The core should be softer than the faces and thick enough to produce shear stresses.
= Ld
15

h
Figure 3.2: Shear actuation
Table 3.1 compares the three types of actuation considering PZT piezoceramics and similar
dimensions (2 mm thick, 1 cm long) and electrical eld (50 kV/m). Two laminar design are
considered: standard (Fig.3.1(b)) and bimorph (Fig.3.3). This latter consists in two piezoelectric
plates with opposite polarisation stacked together to obtain a bending actuator. A stroke
from 1 to 65 m can be obtained, depending on the chosen design.
Design Coupling coecient (PZT) Displacement
stacked d
33
362 10
12
Cb/N 1.81 m
laminar standard d
31
175 10
12
Cb/N 0.9 m
bimorph 65.63 m
shear d
15
515 10
12
Cb/N 2.58 m
Table 3.1: Comparison between actuation designs
38 3. Actuation and sensing
=
3
2
d
31

h
2
L
2
Figure 3.3: Bimorph actuation
3.2 Laminar design modelling
3.2.1 Historical overview
The modelling of piezoelectric material used as laminar actuator or sensor has been addressed
by many authors.
Crawley & de Luis (1987) and Crawley & Anderson (1989) proposed an analytical model for
segmented piezoelectric actuators. The model consists in a Bernoulli-Euler beam with piezo-
electric actuators bonded to the surface or embedded in a laminate. A piezoelectric actuator
can be replaced by an equivalent localized tensile force and bending moment. The equivalent
piezoelectric force and moment equations, so called equivalent actuator equations, are devel-
oped. An experimental validation with a cantilever beam actuated with a given voltage across
the piezoactuator is presented.
Crawley & Lazarus (1991) extended that model to the induced strain actuation of both isotropic
and anisotropic plates. Equivalent normal forces and bending moments for piezoactuation of
plates are derived using the Kirchho-Love plate assumptions. A Ritz formulation for approx-
imate solutions is developed and applied to sandwich experimental models; aluminium and
composite cantilever plates are presented.
Dimitriadis et al. (1991) did a similar work to model bidimensional patches bonded to the surface
of a structure, deriving the equivalent actuator equations and applied them to the vibration of
a rectangular plate for various actuator congurations.
Similarly to the equivalent actuator equations, an equivalent sensor equation can be written
taking into account the contributions from extension and curvature of the sensor to the electrical
charge appearing on the piezoelectric media. Lee (1990) established the general formulation for
an anisotropic piezoelectric laminate using the Kirchho-love hypothesis. Sensor and actuator
equivalent equations are derived, boundary conditions are discussed and the reciprocity between
actuation and sensing is pointed out.
Park & Chopra (1996) modelled the piezoactuation of beams in torsion. A one-dimensional
beam model is used to determine the coupled extension/bending/torsion response to an applied
Finite Element Modelling of Piezoelectric Active Structures 39
voltage across the piezoactuator. It uses the principle of virtual work and takes into account the
cross sectional warping. Detailed results are derived for a thin isotropic beam (Bernoulli-Euler)
with a surface bonded piezoceramic actuator and compared to experiments.
A simplied approach neglecting the coupling between the equations of electrostatics and elasto-
dynamics is described in the next paragraphs. This theory aims to approximate the behaviour
of active structures with only a small fraction of laminar piezoelectric material embedded. A
pure bending beam model for surface bonded piezoelectric actuator/sensor is rst presented; it
demonstrates the eect of the shape of the electrodes, which is illustrated in the last section with
some examples. Next, a shell model of embedded piezo actuation/sensing taking the membrane
strains into account is derived; the importance of membrane coupling is stressed. The duality
between actuation and sensing is pointed out throughout the rest of this chapter.
3.2.2 Pure bending beam model
Actuator
Figure 3.4: Piezoelectric lm actuator bonded on a beam structure.
Consider the beam of Fig.3.4 with a thin layer of piezoelectric material bonded on the surface.
It is assumed that the thickness of the piezoelectric strip h
p
is small compared to that of the
beam h. The piezoelectric strip is used as actuator by controlling the voltage applied to the
electrodes, creating a constant electric eld E = /h
p
. It is also assumed that the electrodes
have a variable width b
p
(x) (this can be achieved either by etching the original surface electrode,
or by cutting the piezoelectric layer). Using structural engineering notations, the constitutive
equation (2.11) reads, within the piezoelectric layer,

x
= Y
p

x
e
31

h
p
(3.1)
Where Y
p
is the Youngs modulus of the piezoelectric material, e
31
its piezoelectric constant and
h
p
, its thickness, over which the voltage is applied. The following usual relation applies to the
beam

x
= Y
b

x
(3.2)
40 3. Actuation and sensing
The equation of motion for the beam reads
m w =

2
M
x
2
= M

(3.3)
where M is the bending moment, m the lineic mass density and w(x) is the displacement of
the beam in direction z. According to the Euler-Bernoulli assumption (plane sections normal
to the neutral axis remain plane and normal to the neutral axis), the axial deformation and the
curvature are related by

x
= z w

(3.4)
where z is the distance to the neutral axis. Introducing this into Equ.(3.1) and (3.2) and
integrating over the cross-section, one gets
M =
_
A

x
z dA = Y I w

+e
31

h
p
h
p
b
p
(x) h (3.5)
or
M = Y I w

+e
31
b
p
(x) h (3.6)
where a constant moment arm h across the thickness of the piezo is assumed. The bending
stiness Y I refers to the supporting structure and the piezoelectric lm together. Substituting
into Equ.(3.3), one gets
m w +(Y I w

= e
31
b

p
(x) h (3.7)
It can be seen that when the width of the electrodes varies along the beam, the piezoelectric
actuator produces a distributed load proportional to the second derivative of b
p
(x). Similarly,
for an electrode of constant width, the distributed actuator is equivalent to adding concentrated
moments M
p
at the boundaries of the actuator (Fig.3.5(a)) while a sudden change in the rst
derivative b

p
(x) at some location x
0
produces a point force P at the discontinuity (Fig.3.5(b))
From Equ.(3.6), the equivalent piezoelectric moment is given by
M
p
= e
31
b
p
h (3.8)
And, by integrating the right hand side of Equ.(3.7) over a small interval about x
0
and taking
the limit when the size of the interval goes to zero, the equivalent piezoelectric point force is
given by
F = e
31

_
b

p
(x
+
0
) b

p
(x

0
)

h (3.9)
Sensor
The piezoelectric strip is used as sensor by measuring the electric charge appearing on the
electrodes. The electrodes are short-circuited so that a zero electric eld is enforced (E = 0).
Finite Element Modelling of Piezoelectric Active Structures 41
(a) Constant width b
p
(b) Sudden change of the derivative b

p
(x) at x
0
Figure 3.5: Piezoelectric lm actuator
According to Equ.(2.12), the amount of charge per unit area is related to the strain by D = eS
or, with structural engineering notations,
D = e
31

x
(3.10)
Combining with Equ.(3.4) and assuming a constant strain over the thickness of the strip,
Equ.(3.10) reads
D = e
31
hw

(3.11)
(a) Current amplier
(b) Charge amplier
Figure 3.6: Piezoelectric lm sensor connections
The corresponding electric charge is equal to the integral of the electric displacement over the
42 3. Actuation and sensing
electrode area,
Q =
_
b
a
Db
p
(x) dx = h
_
b
a
e
31
w

b
p
(x) dx (3.12)
The electrodes can be connected either to a charge amplier or a current amplier (Fig.3.6).
Assuming a uniform polarization prole (e
31
constant), the output voltage of a current amplier
(Fig.3.6(a)) is given

out
(t) = R
f
i
s
(t) = R
f

Q = R
f
e
31
h
_
b
a
b
p
(x) w

dx (3.13)
R
f
is the constant of the amplier. Thus, the output signal is proportional to the integral of
the time derivative of the curvature weighted by the width b
p
(x) of the electrode.
If a charge amplier is used instead (Fig.3.6(b)), the output voltage is proportional to the electric
charge in the electrode; it is given by

out
(t) =
Q
C
r
=
e
31
h
C
r
_
b
a
b
p
(x) w

dx (3.14)
C
r
is the capacitance of the charge amplier. If the width of the electrode, b
p
, is constant, the
output voltage becomes:

out
(t) =
e
31
hb
p
C
r
_
w

(b) w

(a)
_
(3.15)
The perfect duality between Equ.(3.8) for actuation and (3.15) for sensing should be noted; it
is illustrated on Fig.3.7. The reciprocity between actuation and sensing can be summarized as:
a piezoelectric bending actuator can also be used as slope sensor.
Actuator
Piezoelectric Torque M = e
31
b
p
h
Sensor
Output Charge Q = e
31
b
p
h
Figure 3.7: Pure bending beam model: actuation/sensing duality
Finite Element Modelling of Piezoelectric Active Structures 43
3.2.3 Isotropic shell model
In the previous section, the eect of a piezoelectric strip bonded to a beam when it is used
as actuator or sensor has been considered. If the assumptions of the pure bending beam the-
ory become questionable, more elaborate modelling techniques may be required. In general,
when the actuator and sensor are close to each other (nearly colocated), the frequency response
function becomes more sensitive to the detailed modelling of the coupled system (structure-
actuator-sensor) and the transmission path from the actuator to the sensor includes in-plane
(membrane) deformations as well as bending ones; the in-plane deformations become relatively
more important when the thickness of the structure becomes comparable to that of the strips.
For nearly colocated systems, a theory which accounts for the in-plane deformations should be
used. (Preumont, 1997; Loix et al., 1998)
Actuation
Assuming the Kirchho hypothesis for the strain distribution across the thickness, uniform
electric eld and displacement and a linear piezoelectric behaviour, Equ.(2.73) shows that a
voltage applied between the electrodes of a piezoelectric patch produces equivalent loads and
moments:
_
N
M
_
=
_
I
3
z
m
I
3
_
[R
T
]
1
_
_
_
e
31
e
32
0
_
_
_
(3.16)
If the piezoelectric properties are isotropic in the plane (e
31
= e
32
), we have
e
31
[R
T
]
1
_
_
_
1
1
0
_
_
_
= e
31
_
_
_
1
1
0
_
_
_
(3.17)
Where [R
T
] is the transformation matrix given by Equ.(2.56). It follows that
N =
_
_
_
N
x
N
y
N
xy
_
_
_
= e
31

_
_
_
1
1
0
_
_
_
(3.18)
M =
_
_
_
M
x
M
y
M
xy
_
_
_
= e
31
z
m

_
_
_
1
1
0
_
_
_
(3.19)
It can be seen that the in-plane forces and the bending moments are both hydrostatic; they are
independant of the orientation of the facet. It can therefore be concluded that the piezoelectric
loads result in a uniform in-plane load N
p
and bending moment M
p
acting normally to the
contour of the electrode as indicated on Fig.3.8.
N
p
= e
31
, M
p
= e
31
z
m
(3.20)
where z
m
is the distance from the mid-plane of the piezoelectric patch to the mid-plane of the
plate.
44 3. Actuation and sensing
N
p
= e
31

M
p
= e
31
z
m

Figure 3.8: Piezoelectric load

Sensing
Consider a piezoelectric patch connected to a charge amplier as in Fig.3.9. The charge amplier
imposes a zero voltage between the electrodes; the output voltage is proportional to the electric
charge, obtained by integrating the electric displacement D over the surface of the electrode.

out
=
Q
C
r
=
1
C
r
_

Dd (3.21)
Where C
r
is the capacitance of the charge amplier and D is given by Equ.(2.74) in which the
voltage is set to zero.
D =
_
e
31
e
32
0
_
[R
S
]
_
I
3
z
m
I
3

_
S
0

_
(3.22)
In this equation, [R
S
] is the rotation matrix relating the orthotropy axes of the piezo to the
global axes (Equ.(2.57)).
Figure 3.9: Sensor conguration.
Finite Element Modelling of Piezoelectric Active Structures 45
If the piezoelectric properties are isotropic in the plane (e
31
= e
32
), we have
e
31
_
1 1 0
_
[R
S
] = e
31
_
1 1 0
_
(3.23)
Equ.(3.22) becomes
D = e
31
_

0
x
+
0
y
_
+e
31
z
m
(
x
+
y
) (3.24)
and Equ.(3.21) reads

out
=
e
31
C
r
__

0
x
+
0
y
_
d +z
m
_

(
x
+
y
) d
_
(3.25)
The rst integral represents the contribution of the average membrane strains over the electrode
and the second, the contribution of the average curvature. The integrals extend only over the
electrode (the part of the piezo not covered by the electrode does not contribute to the signal).
Recalling that (Kirchho hypothesis)

0
x
=
u
x
,
0
y
=
v
y
(3.26)

x
=

2
w
x
2
,
y
=

2
w
y
2
(3.27)
where u, v and w are the mid-plane displacements in direction x, y and z respectively, Equ.(3.25)
can be written

out
=
e
31
C
r
__

.u
0
d +z
m
_

.(w) d
_
(3.28)
where is the gradient operator and u
0
u v
T
is the mid-plane membrane displacement
vector.
By using the Green integral
_

.ad =
_
C
a.ndl (3.29)
the foregoing result can be transformed into

out
=
e
31
C
r
__
C
u
0
.ndl +z
m
_
C
w
n
dl
_
(3.30)
where the integrals extend to the contour of the electrode. The rst term is the mid-plane
displacement normal to the contour while the second is the slope of the mid-plane normal to
the contour (Fig.3.10).
It is worth insisting that for both the actuator and the sensor, it is not the shape of the piezoelec-
tric patch that matters, but rather the shape of the electrodes. The strong reciprocity existing
between actuation and sensing relationships (Equ.(3.20) and (3.30)) has been rst demonstrated
by Lee (1990): if certain layers of the piezoelectric laminae serve as a particular type of actu-
ator, the laminae can also be used as a sensor of the corresponding type and vice versa. More
precisely, the resulting equivalent load produced by a piezoelectric device used as actuator is
conjugated to the dimension measured by the same piezoelectric device used as sensor (e.g. if a
bending actuator is used as sensor, the slope or the bending angular velocity can be measured).
46 3. Actuation and sensing
Figure 3.10: Contribution to the output of the piezoelectric isotropic sensor (e
31
= e
32
)
3.2.4 Why is the beam theory not enough to model colocated systems ?
Comparing the equivalent loads (Equ.(3.20)) of the shell theory with the equivalent bending
moment (Equ.(3.8)) of the beam theory, we see that the beam theory accounts only for the
component of the bending moment normal to the beam axis and neglects totally the in-plane
force (Fig.3.11). Conversely, the sensor signal given by the beam theory accounts only for the
component of the rotation along the contour normal to the beam axis.
(a) beam theory
(b) shell theory
Figure 3.11: Equivalent piezoelectric loads for a rectangular piezoelectric patch on a beam
Taking the example of an active cantilever beam (Fig.3.12), it turns out that, when the thickness
of the beam is small and the actuator and sensor are located on opposite sides of the beam
(colocated), the membrane strains contribute signicantly to the frequency response function.
The discrepancy between the pure bending beam theory and the shell theory becomes large,
particularly near the zeros of the frequency response function which are strongly aected by the
feedthrough from the strain actuator to the strain sensor (Preumont, 1997, Chap.3).
The simplied approach developed above is well suited for a classical nite element implemen-
tation (Loix, 1998). The structure (including the piezoelectric components) is modelled with
classical nite elements and the piezoelectric loads are regarded as external loads (Equ.(3.20))
Finite Element Modelling of Piezoelectric Active Structures 47
Figure 3.12: Frequency response function of an active cantilever beam with colocated actuator
and sensor. Comparison between the beam theory and the shell theory.
and the sensor signal is computed from Equ.(3.25).
However, this approach neglects the coupling between the equations of electrostatics and elas-
todynamics; it is sucient if the piezoelectric material represents only a small fraction of the
entire structure.
A nite element approach accounting for the piezomechanical coupling is presented in Chap.4
and its implementation into the commercial nite element package Samcef is discussed. Some
applications are presented in Chap.5.
Next section illustrates the realization of dedicated sensor devices by using distributed piezoelec-
tric elements with shaped electrodes; two examples are considered: modal ltering and acoustic
radiation sensing.
3.3 Spatial lters
3.3.1 Modal lters
Discrete point sensors/actuators to control vibrations of exible structures often suer from
spillover (destabilization of exible modes outside the bandwidth of the controller) leading to
instabilities (Preumont, 1997; Loix, 1998). The solution of modal ltering requires heavy real
time computation (integration in the space domain). The main idea leading to the concept of
inherent modal ltering is that, by simply suitably shaping the distributed actuator/sensor, the
spatial integration is embedded into the sensor itself and consequent real time computation can
be saved.
48 3. Actuation and sensing
Burke & Hubbard Jr (1987) developed a formulation for the control of thin beams subject to
most combination of free, clamped, or pinned boundary conditions in which the active elements
were spatially varying piezoelectric transducer layers.
Lee et al. (1989, 1991) and Lee & Moon (1990) proposed the use of distributed actuators/sensors
to reduce the computational load by taking advantage from the directional properties of piezo-
electric materials. By suitably shaping the actuators/sensors, the inherent modal ltering is
realized; sensors for mode 1 & 2 of a one-dimensional cantilever plate were constructed by
shaping the electrodes to achieve the desired characteristics and tested.
Callahan & Baruh (1996) presented a control law for modal control. Miller et al. (1996a,b, 1998)
proposed a selective modal control design methodology by introducing the concept of Selective
Modal Transducer (SMT). SMTs are capable of sensing and exciting any set of vibrational modes
in a selectively weighted fashion. The transduction of selected modal subset is accomplished
through combining the eects of multiple piezolaminae whose piezoelectric eld distribution
vary spatially. Simulation results are presented. Some experimental validation (anisotropic
PVDF layers with dierent orientations) can be found in (Miller et al., 1997).
Ryou et al. (1998) presented a design method of the electrode shape based on genetic algorithms.
The optimization of the electrode pattern to realize the concept of modal transducer for a two-
dimensional structure is sought. The electrode pattern is determined by deciding on or o each
electrode tile. The optimization criteria chosen for the actuator design is to minimize the energy
in the control mode. The one chosen for the sensor design is the minimization of the spillover.
The performance of the design is demonstrated by experiments.
The theory lying behind modal ltering by electrode shaping can be illustrated with the pure
bending beam model above (Euler-Bernoulli assumption) (Lee & Moon, 1990; Loix, 1998).
A beam of length L, covered with piezoelectric material such as depicted on Fig.3.4 is consid-
ered; its dynamics follows Equ.(3.7). The transverse displacement w can be decomposed into a
weighted sum of its mode shapes (e.g. Preumont, 1997, Chap. 2).
w(x, t) =

k=1
x
k
(t)Z
k
(x) (3.31)
where Z
k
(x) is the mode shape and x
k
(t), the modal amplitude for mode k.
Substituting this expression into Equ.(3.7), assuming there is no rigid body modes and using
the following orthogonality properties of the mode shapes
_
L
0
mZ
i
(x)Z
j
(x)dx =
i

ij
(3.32)
_
L
0
Y IZ

i
(x)Z

j
(x)dx =
i

2
i

ij
(3.33)
where
i
are the eigen frequencies,
i
the modal masses and
ij
is the Kronecker index (
ij
= 1
for i = j, 0 otherwise), we get the equation for the amplitude of mode k

k
( x
k
+
2
k
x
k
) = e
31

_
L
0
Z
k
b

p
(x)dx (3.34)
Finite Element Modelling of Piezoelectric Active Structures 49
This equation shows that, by shaping the electrode in such a way that
b

p
(x) = C
st
mZ
k
(x) (3.35)
the authority of the actuator can be limited to the single mode k.
Conversely, the output voltage of a charge amplier connected to the electrodes (Equ.(3.14))
can be decomposed in the sum of the contribution of each mode.

out
(t) =
Q
C
r
=
e
31
h
C
r

k=1
__
L
0
b
p
(x) Z

k
(x) dx
_
x
k
(t) (3.36)
and, again, by shaping the electrode in such a way that
b
p
(x) = C
st
Y IZ

k
(x) (3.37)
the sensitivity of the sensor can be limited to the single mode k.
3.3.2 Volume velocity sensor
In recent years, various possibilities have been explored to measure the sound power radiation by
structure borne sensors. These are preferable to microphones in vibroacoustic control because
they do not introduce time delays in the control loop. It can be shown that, at low frequency,
there is a strong correlation between the sound power radiated by a baed plate and its volume
velocity.
V
vol
=
_

wd (3.38)
This allows to replace the sound power sensor by a volume velocity (or displacement) sensor
directly mounted on the structure. For a supported plate, the volume sensor can be based on
an array of strain sensors.
The QWSIS (Quadratically Weighted Strain Integrating Sensor- Rex & Elliott, 1992) consists
of an array of narrow PVDF strips with parabolic shape layed on the surface of the plate
(Fig.3.13). It is based on the discretization of the plate sensor into narrow bands of width ;
each strip being considered as a beam covered with piezoelectric material with quadratically
shaped electrodes. The parabolic shape performs an inherent spatial integration, reconstructing
the volume velocity; the charge signals from all strips are summed to obtain the volume velocity
of the entire plate. The piezoelectric material used for the sensor is a highly anisotropic PVDF
(e
32
e
31
).
For a beam, the output signal of a current amplier (Fig.3.6(a)) connected to an electrode
of variable width b
p
(x) is proportional to the integral of the time derivative of the curvature
weighted by the width (See Equ.(3.13)).

out
(t) = e
31
hR
f
_
L
0
b
p
(x) w

(x) dx (3.39)
50 3. Actuation and sensing
Figure 3.13: QWSIS sensor (Rex & Elliott, 1992)
Considering an electrode of parabolic shape
b
p
(x) = 4
x
L
_
1
x
L
_
(3.40)
and integrating Equ.(3.39) twice, by parts, one gets

out
(t) = e
31
hR
f
_
_
b
p
(x) w

(x)

L
0
+
_
b

p
(x) w(x)

L
0
+
_
L
0
b

p
(x) w(x) dx
_
(3.41)
of which the rst term vanishes, b
p
(x) being zero for x = 0 and x = L, and the second, the plate
being supported ( w(x) = 0 for x = 0 and x = L).
The second derivative of b
p
being the constant 8/L
2
, the output signal is directly related to
the volume velocity of the beam.

out
(t) = 8e
31
R
f
h
L
2

_
L
0
w(x) dx V
vol
(3.42)
For the plate, it can be shown that the output signal generated by the sensor is proportional to
(Preumont et al., 1999)
D = e
31
n

i=1
_
L
0
dx
_
y
i
+

2
y
i

2
(

2
w
x
2
+

2
w
y
2
)dy (3.43)
where n is the number of strips and is the piezoelectric anisotropy ratio of the PVDF (e
32
=
e
31
). It contains an additional term due to the coupling e
32
in the transverse direction.
Finite Element Modelling of Piezoelectric Active Structures 51
Chapter 4
Finite element approach
The study of physical systems frequently results in partial dierential equations which either
cannot be solved analytically or lack an exact analytic solution due to the complexity of the
boundary conditions or domain. For a realistic and detailed study, a numerical method must be
used to solve the problem. The nite element method is often found the most adequate.
Over the years, with the development of modern computers, the nite element method has
become one of the most important analysis tool in engineering. It has penetrated successfully
many areas such as heat transfer, uid mechanics, electromagnetism, acoustics and fracture
mechanics. Finite element packages are now widely available on personal workstations.
Basically, the nite element method (Zienkiewicz, 1971; Reddy, 1984; Hughes, 1987) consists
in a piecewise application of classical variational methods to smaller and simpler subdomains
called nite elements connected to each other in a nite number of points called nodes.
The fundamental principles of the nite element method (displacement) are:
The continuum is divided in a nite number of elements of geometrically simple shape
These elements are connected in a nite number of nodes
The unknowns are the displacements of these nodes
Polynomial interpolation functions are chosen to describe the unknown displacement eld
at each point of the elements related to the corresponding eld values at the nodes.
The forces applied to the structure are replaced by an equivalent system of forces applied
to the nodes.
As mentioned in Chap.3, the frequency response functions between the inputs and the outputs
of a control system involving embedded distributed piezoelectric actuators and sensors in a shell
structure are not easy to determine numerically. The situation where they are nearly colocated
is particularly critical, because the zeros of the frequency response function are dominated by
local eects (See Preumont (1997); Loix et al. (1998) and 3.2.4). These can easily be accounted
for by nite elements (Piefort & Henrioulle, 2000).
52 4. Finite element approach
A nite element formulation accounting for the coupling between the equations of electrostatics
and elastodynamics becomes necessary when the piezoelectric material represents a non negli-
gible fraction of the entire structure.
The present chapter is divided in six sections. The rst one is an historical overview of the nite
element modelling of structures embedding piezoelectric material and, more precisely, laminar
piezoelectric designs. In the second section, a variational principle is built starting from the linear
piezoelectric constitutive equations presented in Chap.2 and the Hamilton principle (generalized
virtual work principle). A nite element formulation is derived in the third section and the
particular case of piezoelectric shells is presented in the next section. The actual implementation
in the commercial nite element package Samcef is described in the fth section. The sixth
section is devoted to particular electric boundary conditions and how they can be considered
in the nite element model. Finally, a method to extract a state space model describing the
dynamics of a structure embedding piezoelectric actuators and sensors from a nite element
analysis is described in the last section.
4.1 History
Following the early work of Eer Nisse (1967) and Tiersten (1967) who have established varia-
tional principles for piezoelectric media, the nite element modelling of structures with embed-
ded piezoelectrics has known important developements in recent years. Allik & Hughes (1970)
proposed a tetrahedral volumic element accounting for the piezoelectricity. Starting from Hamil-
tons principle, and the constitutive equations for piezoelectric media, a simple volumic element
(tetrahedron) taking into account the piezoelectric coupling is presented. The element has 4
nodes and 4 degrees of freedom per node (3 translations and the electric potential); it uses
linear shape functions for both displacement and electrical elds.
Lerch (1990) developed a general formulation accounting for the piezoelectric coupling for two-
and three- dimensional nite element modelling of piezoelectric devices. Comparisons between
numerical simulation and experiment are presented for the vibration of piezoceramic parallelip-
ipedic bars.
A higher order tetrahedral element is proposed by Moetakef et al. (1995). Interpolation functions
of higher order are used; tetrahedron of 10 (linear strain element) and 20 nodes (quadratic strain
element) are presented; brick elements are obtained by assembling tetrahedrons using a Guyan
condensation of the resulting internal nodes to reduce the number of degrees of freedom; these
elements are used to model a bimorph pointer (actuation), for which the results are compared to
an analytical solution, and a cube under uniform pressure (sensing). An experiment consisting
in generating an elastic wave in a cantilever beam is described in Moetakef et al. (1996); results
from the model are compared to experimental ones and shown good qualitative agreement.
With the increasing number of shell structures, the need for a nite element modelling tool for
plates and shells with embedded distributed piezoelectric actuators and sensors has become more
and more evident. Dierent approach for modelling thin and thick shells have been proposed.
Tzou & Tseng (1990, 1991) derived a thin brick element for distributed dynamic measurement
and active vibration control of a rectangular plate; the element consists in a thin solid piezo-
Finite Element Modelling of Piezoelectric Active Structures 53
electric brick having 8 structural nodes with 4 degrees of freedom per node (3 translations and
the electric potential) and 3 internal nodes (condensed using a Guyans reduction) to dissipate
the excessive shear energy due to the small dimension in the thickness direction. A classic con-
guration for an intelligent structure is composed of a master structure sandwiched between 2
piezoelectric thin layers acting as the distibuted sensor and actuator. Both bonded and embed-
ded piezoelectric sensors and actuators result in a laminate; the multilayer structure is modelled
by stacking the thin brick elements together and connecting the corresponding nodes. The model
is applied to the vibration control of a simply supported square plate. Mode shape and modal
voltage distribution are obtained thanks to the model.
Ha et al. (1992) used a similar brick element where the multilayer structure is taken into account;
the element matrices are integrated over the thickness of each layer and summed. That element
is used to model the cantilever plate described in (Crawley & Lazarus, 1991) (static case), to
determine the step-response of a cantilever beam and to design the active damping of the rst
mode of sensor/actuator composite cantilever plate. The results are compared to the results
found in (Crawley & Lazarus, 1991) and shown good agreement with experiments.
Rao & Sunar (1993) established a nite element formulation of thermopiezoelectric problems
starting from the linear thermopiezoelectric constitutive equations established by Mindlin (1974)
and the Hamiltons principle. In (Sunar & Rao, 1996, 1997), they used the quasistatic equations
of thermopiezoelectricity to develop heat, sensor and actuator equations; a nite element for-
mulation is presented. A distributed control system consisting in a cantilever beam sandwiched
between a piezoelectric sensor/actuator pair is used to evaluate the proposed nite element
approach on the static and dynamic behaviour.
Tzou & Ye (1996) derived a 12-nodes triangular thin solid plane element with 4 degrees of free-
dom per node; it uses shape functions quadratic in the two in-plane directions and linear in the
transverse direction with the assumption of a layerwise constant shear angle (Mindlin hypothe-
sis). A laminate is composed of laminae which could be either elastic material or piezoelectric
material; the laminated structure is obtained by stacking elements together and connecting the
corresponding nodes; this element is validated by modelling the actuation of a bimorph pointer.
To stress the inuence of the piezoelectric coupling on the vibration characteristics, a semi-
circular ring shell has been modelled using 60 triangular shell elements (20 for each layer and 10
element meshes along the length); the evolution of its eigenfrequencies with a growing number
of short-circuited electrodes is examined. The number of short-circuited electrodes varies from
1 to 10 (fully short-circuited). This element has been extended later by Koppe et al. (1998)
to isoparametric curved triangular and quadrangular elements with shape functions of dier-
ent polynomial degree for each layer; the model is applied to a rectangular plate of composite
material with surface bonded piezo patches under static voltage load, simply supported on two
edges.
The modelling of shells using solid elements results in an excessive shear strain energy in the
thickness direction. By reducing one dimension compared to the others, the transverse shear
stiness term becomes excessively important (linked to the linear interpolation of the strain)
leading to what is usually called the shear locking phenomenon. A commonly used solution to
overcome this diculty consists in adding internal degrees of freedom resulting in large problem
size requiring techniques such as the Guyans reduction (also known as static condensation) to
reduce the number of degrees of freedom.
54 4. Finite element approach
Lee & Saravanos (1996) derived a thermopiezoelectric multilayer beam element; it uses shape
functions linear along the beam and linear through the thickness of each layer (layerwise linear).
A reduced integration scheme for the transverse shear stiness is used; the element takes into
account the eect of constant thermal load (constant gradient of temperature); A cantilever
beam under thermal load is modelled. Heyliger et al. (1996) extended the layerwise linear
formulation to a piezoelectric shell element and applied it to static and dynamic modelling of a
simply supported plate and a cylindrical shell. Later, Saravanos (1997) presented a multilayer
piezoelectric thin plate using the Kirchho-Love assumption (linear displacement eld through
the thickness) and bilinear shape functions; it has 1 electrical degree of freedom per piezoelectric
layer per node, assuming a constant electric eld through the thickness for each layer (layerwise
linear transverse shape function for the electric potential). That shell element has been applied
to the modelling of a simply supported plate and shown good agreement with exact solutions
for moderately thin plates (a/h 50), an actuated cylindrical panel to study the eect of the
actuator placement through the thickness which exactly matches a Ritz solution and a cantilever
cylindrical shell to show the eect on actuation and sensing of the dierence between continuous
and discontinuous transducers (4 across the length) and the eect of the curvature on the tip
displacement and sensing. It was later used to evaluate the passive damping of piezoelectric
shells with integrated electrical network and compared with experiments (Saravanos, 2000).
A pure bending (Kirchho assumption) plane rectangular plate element is proposed in (Hwang
& Park, 1993); the main idea is the use of a multilayered plate element with a single electrical
degree of freedom per piezoelectric layer, the voltage across the thickness of the layer, uniform
on the element surface. This multilayer element has 4 nodes with 3 degrees of freedom per
node (1 translation and 2 rotations) and 1 electrical degree of freedom for each piezoelectric
layer (voltage across the layer). This element neglects the transverse shear and is therefore not
suitable to model thick shells; it does not account for the extension, modelling only the bending
behaviour. The bimorph pointer is modelled numerically and the results were compared to an
analytical solution and shown a good agreement. Chen et al. (1997) used an isoparametric pure
bending element to model a bimorph beam and for vibration control design.
Samanta et al. (1996) used a cubic displacement eld with a 8 node quadratic rectangular
multilayer plate with 2 electrical degrees of freedom (constant voltage over the element across the
only two piezoelectric outer layer) and 11 mechanical degrees of freedom per node (3 translations,
3 slopes and 5 higher order rotations). A simply supported plate is modelled; the fundamental
natural frequencies and forced response were computed and shown good agreement with exact
solutions.
Suleman & Venkayya (1995a) and Suleman & Goncalv`es (1995) proposed a 4 node plate element
using bilinear shape functions and the Mindlin assumption (constant shear angle) to accomodate
thick as well as thin shells; each node has 5 degrees of freedom (3 translations and 2 rotations),
the element has one additional electrical degree of freedom per piezoelectric layer (voltage across
the thickness). It uses a reduced integration scheme for the transverse shear stiness to avoid the
transverse shear locking phenomenon. This element is demonstrated using the plate described
by Crawley & Lazarus (1991), a bimorph pointer and panel utter control; comparison with the
work of Ha et al. (1992) is made.
Chattopadhyay et al. (1999) developed a quasi-static coupled thermopiezoelectric model for a
smart composite plate structure with surface bonded piezoelectric materials using a variational
Finite Element Modelling of Piezoelectric Active Structures 55
approach; linear piezoelectricity is assumed, a higher order transverse shear strain distribu-
tion is used (third order). A simply supported unidirectional graphite/epoxy laminate plate is
modelled; the inuence of transverse shear and couplings are discussed. A shell nite element
formulation was derived (Zhou et al., 2000) with the same transverse shear strain distribution
and a higher order thermal eld. A rectangular ber-reinforced laminated plate with surface
bonded piezoelectric patches is modelled; the inuence of the couplings on the dynamics of piezo-
and thermo- actuated structures is discussed.
When the thickness becomes small, the behaviour of the elements accounting for the transverse
shear strain (e.g. Mindlin elements) is driven by transverse shear stiness while the transverse
shear strain should be negligible. This also leads to the shear locking phenomenon. Solutions
to overcome this problem can be to use a reduced integration scheme for the transverse shear
stiness (Suleman & Venkayya, 1995b; Hong & Chopra, 1999) or to use dierent interpolation
functions for the transverse shear strain. The latter is the solution adopted by the element used
in the present study: Mindlin shell element from the commercial nite element package Samcef
(Samtech s.a.).
4.2 Variational principle
The piezoelectric constitutive equations can be written under the form (2.34)
T = [c
E
]S [e]
T
E (4.1)
D = [e]S +[
S
]E (4.2)
where T represents the stress vector, S, the strain vector, E, the electric eld, D, the
electric displacement, [c
E
], the elastic coecients at constant E, [
S
], the dielectric coecients
at constant S, and [e], the piezoelectric coupling coecients
The dynamic equations of a piezoelectric continuum can be derived from the Hamilton principle,
in which the Lagrangian and the virtual work are properly adapted to include the electrical
contributions as well as the mechanical ones (Allik & Hughes, 1970; Lerch, 1990; Tzou & Tseng,
1990). The potential energy density of a piezoelectric material includes contributions from the
strain energy and from the electrostatic energy (Tiersten, 1967).
The dynamic equations of a piezoelectric continuum can be derived using Hamiltons principle:

_
t
2
t
1
(L +J)dt = 0 (4.3)
where t
1
and t
2
dene the time interval (all variations must vanish at t = t
1
and t = t
2
), L is
the Lagrangian and J is the virtual work of external mechanical and electrical forces.
The Lagrangian L is dened by the sum of kinetic energy J and electrical enthalpy H (linear
piezoelectricity) (The Gibbs free energy G = H -Equ.(2.17)- would have been used to also
take into account the pyroelectric coupling - see 2.2.2). The electrical enthalpy density H is
dened by
H = U E
i
D
i
(4.4)
56 4. Finite element approach
The conservation of energy for the linear piezoelectric continuum results in the rst law of
thermodynamics (Equ.(2.13)):
dU = T
ij
dS
ij
+E
i
dD
i
(4.5)
from equations (4.5) and (4.4), there results
dH = T
ij
dS
ij
D
i
dE
i
(4.6)
and thus
T
ij
=
H
S
ij
(4.7)
D
i
=
H
E
i
(4.8)
In linear piezoelectric theory, the form taken by H is (IEEE std)
H =
1
2
c
E
ijkl
S
ij
S
kl
e
kij
E
k
S
ij

1
2

S
ij
E
i
E
j
(4.9)
where c
ijkl
, e
kij
and
ij
are respectively the elastic, the piezoelectric and the dielectric constants.
J =
1
2
u
T
u (4.10)
H =
1
2
_
S
T
T E
T
D

(4.11)
L =
_
V
(J H)dV =
_
V
_
1
2
u
T
u
1
2
_
S
T
T E
T
D

_
dV (4.12)
where u is the velocity eld.
The essential boundary conditions are: prescribed displacement eld on
3
(u = u) and
electric potential on
4
( = ).
The virtual work done by the external mechanical forces and the applied electric charges for
an arbitrary variation of the displacement eld u and of the electrical potential both
compatible with the essential boundary conditions (i.e. u = 0 on
3
and = 0 on
4
) is
J =
_
V
u
T
F
V
dV +
_

1
u
T
F

d +u
T
F
P

_

2
d Q (4.13)
where F
V
, are the body applied forces, F

, the surface applied forces (dened on

1
), F
P
,
the point loads, , the electric potential, , the surface charge brought on
2
, Q, the applied
concentrated electric charges, and , the mass density.
From Equ.(4.11) and (4.13), the analogy between electrical and mechanical variables can be de-
duced (Table 4.1). The electrical tensors are one degree lower than the corresponding mechanical
ones.
Finite Element Modelling of Piezoelectric Active Structures 57
Mechanical Electrical
Force F
Displacement u
Stress [T]
Strain [S]
Charge
Voltage
D Electric Displacement
E Electric Field
Table 4.1: Electromechanical analogy
Integrating the variation of kinetic energy term

u
T
u by part over the time interval, one
gets
_
t
2
t
1

u
T
udt =
_
u
T
u

t
2
t
1

_
t
2
t
1
u
T
udt (4.14)
of which the rst term vanishes, u being equal to zero in t = t
1
and t = t
2
.
Taking into account the constitutive equations (4.1) and (4.2) and substituting the Lagrangian
and virtual work into Hamiltons principle (4.3) yields
0 =
_
V
_
u
T
u S
T
[c
E
]S +S
T
[e]
T
E +E
T
[e]S
+ E
T
[
S
]E +u
T
F
V

dV +
_

1
u
T
F

d
+u
T
F
P

_

2
d Q (4.15)
4.3 Finite element formulation
The displacement eld u and the electric potential over an element are related to the
corresponding node values u
i
and
i
by the mean of the shape functions [^
u
], [^

]
u = [^
u
]u
i
(4.16)
= [^

]
i
(4.17)
And therefore, the strain eld S and the electric eld E are related to the nodal displace-
ments and potential by the shape functions derivatives [B
u
] and [B

] dened by
S = [T][^
u
]u
i
= [B
u
]u
i
(4.18)
E = [^

]
i
= [B

]
i
(4.19)
where is the gradient operator and [T] is the derivation operator dened such as S = [T]u
[T] =
_

x
0 0
0
y
0
0 0
z
0
z

y

z
0
x

y

x
0
_

_
(4.20)
58 4. Finite element approach
Substituting expressions (4.16) to (4.20) into the variational principle (4.15) yields
0 = u
i

T
_
V
[^
u
]
T
[^
u
]dV u
i
u
i

T
_
V
[B
u
]
T
[c
E
][B
u
]dV u
i

u
i

T
_
V
[B
u
]
T
[e][B

]dV
i

i

T
_
V
[B

]
T
[e]
T
[B
u
]dV u
i

+
i

T
_
V
[B

]
T
[
S
][B

]dV
i
+ u
i

T
_
V
[^
u
]
T
F
V
dV
+ u
i

T
_

1
[^
u
]
T
F

d + u
i

T
[^
u
]
T
F
P

T
_

2
[^

]
T
d
i

T
[^

]
T
Q (4.21)
which must be veried for any arbitrary variation of the displacements u
i
and electrical
potentials
i
compatible with the essential boundary conditions.
For an element, Equ.(4.21) can be written under the form
[/] u
i
+[K
uu
] u
i
+[K
u
]
i
= f
i
(4.22)
[K
u
] u
i
+[K

]
i
= g
i
(4.23)
with
[/] =
_
V
[^
u
]
T
[^
u
]dV (4.24)
[K
uu
] =
_
V
[B
u
]
T
[c
E
][B
u
]dV (4.25)
[K
u
] =
_
V
[B
u
]
T
[e]
T
[B

]dV (4.26)
[K

] =
_
V
[B

]
T
[][B

]dV (4.27)
[K
u
] = [K
u
]
T
(4.28)
respectively the element mass, stiness, piezoelectric coupling and capacitance matrix and
f
i
=
_
V
[^
u
]
T
P
b
dV +
_

1
[^
u
]
T
P
S
d + [^
u
]
T
P
c
(4.29)
g
i
=
_

2
[^

]
T
d [^

]
T
Q (4.30)
the external mechanical force and electric charge.
Each element k of the mesh is connected to its neighbouring elements at the global nodes and
the displacement is continuous from one element to the next. The element degrees of freedom
(dof) (u
i

(k)
,
i

(k)
) are related to the global dof (U, ) by the mean of the localization
matrices [L
u
]
(k)
and [L

]
(k)
:
u
i

(k)
= [L
u
]
(k)
U (4.31)

(k)
= [L

]
(k)
(4.32)
Finite Element Modelling of Piezoelectric Active Structures 59
The element ij of [L
u
]
(k)
is equal to 1 if the i
th
mechanical dof of the nite element k corresponds
to the j
th
global dof and is zero otherwise. The element ij of [L

]
(k)
is equal to 1 if the i
th
electric
dof of the nite element k is connected to the j
th
global electric dof and is zero otherwise.
The Hamiltons principle (4.3) must be veried for the whole structure, which results in (by
summation of the contribution from each nite element):
0 = U
T
__

k
[L
u
]
(k)
T
[/]
(k)
[L
u
]
(k)
_

U +
_

k
[L
u
]
(k)
T
[K
uu
]
(k)
[L
u
]
(k)
_
U
+
_

k
[L
u
]
(k)
T
[K
u
]
(k)
[L

]
(k)
_

k
[L
u
]
(k)
T
[f
k
]
_
+
T
__

k
[L

]
(k)
T
[K
u
]
(k)
[L
u
]
(k)
_
U +
_

k
[L

]
(k)
T
[K

]
(k)
[L

]
(k)
_

k
[L

]
(k)
T
[g
k
]
_
(4.33)
again for any arbitrary variation of the displacements U and electrical potentials veri-
fying the essential boundary conditions.
Equ.(4.33) can be written under the form
[/]

U +[K
UU
]U +[K
U
] = F (4.34)
[K
U
] U +[K

] = G (4.35)
where the assembled matrices are given by:
[/] =

i
[L
ui
]
T
_
/
(i)
_
[L
ui
] (4.36)
[K
UU
] =

i
[L
ui
]
T
_
K
(i)
uu
_
[L
ui
] (4.37)
[K
U
] =

i
[L
ui
]
T
_
K
(i)
u
_
[L
i
] (4.38)
[K
U
] =

i
[L
i
]
T
_
K
(i)
u
_
[L
ui
] (4.39)
[K

] =

i
[L
i
]
T
_
K
(i)

_
[L
i
] (4.40)
F =

i
[L
ui
]
T
[f
i
] (4.41)
G =

i
[L
i
]
T
[g
i
] (4.42)
Equ.(4.34) and (4.35) couple the mechanical variables U and the electrical potentials ;
F represents the external forces applied to the structure and G the electric charges brought
to the electrodes.
4.4 Kirchho element
It is assumed that the electric eld and displacement are uniform across the thickness and aligned
with the poling direction on the normal to the mid-plane. The electrical degrees of freedom are
60 4. Finite element approach
the voltages
k
across the piezoelectric layers; it is assumed that the potential is constant over
each element (this implies that the nite element mesh follows the shape of the electrodes). One
electrical dof per piezoelectric layer is dened.
From these assumptions, it can be concluded that, over an element
=
k
(4.43)
E =
_

k
h
k
_
(4.44)
where h
k
is the thickness of layer k, involving that [^

] = I and [B

] = diag(1/h
k
).
Starting from the constitutive equations for a Kirchho piezoelectric shell (Equ.(2.73) and
(2.74)), the element mass, stiness, coupling and piezoelectric capacitance matrices as dened
by Equ.(4.24) to (4.27) read ([^] and [B] are used for [^
u
] and [B
u
] respectively).
[/] =
_

m[^]
T
[^]d (4.45)
[K
uu
] =
_

[B]
T
_
A B
B D
_
[B]d (4.46)
[K
u
] =
_

[B]
T
_
... c
T
k
...
... c
T
k
z
mk
...
_
d (4.47)
[K

] =
_

_
.
.
.
0

k
/h
k
0
.
.
.
_

_
(4.48)
[K
u
] = [K
u
]
T
(4.49)
with m is the surfacic mass density.
We have introduced
c
k
=
_
e
31
e
32
0
_
k
[R
S
]
k
(4.50)
and we have used the fact that
[R
T
]
1
k
_
e
31
e
32
0
_
T
k
= c
T
k
(4.51)
The external applied mechanical forces and the electric charges brought to the electrodes of the
system read
f =
_

[^]
T
P
S
d +P
c
(4.52)
g =
_
. . .
k
. . .
_
T
(4.53)
where [P
S
] and [P
c
] are respectively the external distributed forces and concentrated forces and

k
the external charges brought to the electrodes.
Finite Element Modelling of Piezoelectric Active Structures 61
4.5 Mindlin element
The classical Kirchho theory neglects the transverse shear strains. Alternative theories which
accomodate the transverse shear strains have been developed and have been found more accurate
for thick shells (Hughes, 1987).
We consider a shell structure with embedded piezoelectric patches covered with electrodes. The
poling direction and the electric and displacement elds direction are parallel, normal to the
patches. The piezoelectric patches are parallel to the mid-plane and orthotropic in their plane.
The following hypothesis are made (See also Appendix A):
One dimension is signicantly less than the others
Plane stress hypothesis:
zz
= 0, using classical engineering notations
Mindlin assumption: a ber normal to the mid-plane remains straight but no longer or-
thogonal to the mid-plane; a transverse shear strain is introduced, constant through
the thickness and equal to the angle dierence between the normal to the mi-plane and
the material ber.
S = S
0
+z (4.54)
= +w (4.55)
where S
0
is the mid plane strain, the actual curvature (spatial derivatives of the ber
rotation angles ), w is the transverse mid-plane displacement and is the gradient
operator (w is the mid-plane slope). The in-plane strain and stress vectors written in
axes (xy) are given in classical engineering notation respectively by S =
x

y

xy

T
and
T =
x

y

xy

T
and the transverse shear strain and stress vectors by =
yz

xz

T
and =
yz

xz

T
.
There is no coupling between extension/bending and transverse shear
Uniform electric eld and displacement across the thickness and aligned on the normal
to the mid-plane (direction 3).
E =
_
_
_
0
0
E
_
_
_
, D =
_
_
_
0
0
D
_
_
_
(4.56)
Linear piezoelectricity for each piezoelectric layer k; it is assumed that the piezoelectric
principal axes are parallel the structural orthotropy axes and that the poling direction
is direction 3. It is also assumed that no shear strain is induced by a transverse electric
eld (i.e. e
34
= e
35
= e
36
= 0), which is the case for most commonly used piezoelectric
materials in laminar design: PZT and PVDF (see 2.2.2).
62 4. Finite element approach
The constitutive equations for the k
th
layer become
T = [c]
k
S
_
_
_
e
31
e
32
0
_
_
_
k
E
k
(4.57)
D
k
=
_
e
31
e
32
0
_
k
S +
k
E
k
(4.58)
= [c
t
]
k
(4.59)
where [c]
k
and [c
t
]
k
are the in-plane and transverse shear elastic coecients matrix of the
k
th
layer (stiness matrices in the principal material axes).
Figure 4.1: Resultant eorts
In-plane eorts N, bending moments M and transverse shear loads T (Fig.4.1) can be
determined by integrating the stresses over the thickness of the multilayered material
N =
_
+
h
2

h
2
T dz
M =
_
+
h
2

h
2
Tz dz
T =
_
+
h
2

h
2
dz
Upon introducing the transverse shear strains =
xz

yz

T
and the transverse shear loads
T = T
xz
T
yz

T
, the constitutive equations for a Mindlin piezoelectric shell can be derived
from the ones for a Kirchho piezoelectric shell (Equ.(2.73) and (2.74)). The global constitutive
equations of the piezoelectric Mindlin shell nally read
Finite Element Modelling of Piezoelectric Active Structures 63
_
_
_
N
M
T
_
_
_
=
_
_
A B 0
B D 0
0 0 K
_
_
_
_
_

_
_
_
+
n

k=1
_
_
I
3
z
mk
I
3
0
_
_
[R
T
]
1
k
_
_
_
e
31
e
32
0
_
_
_
k

k
(4.60)
D
k
= e
31
e
32
0
k
[R
S
]
k
[I
3
z
mk
I
3
0]
_
_
_

_
_
_

k
h
k

k
(4.61)
where the load vector now includes the in-plane loads N and moments M and the transverse
shear loads T . Similarly, the strain vector includes the mid-plane membrane strains
0
, the
curvatures and the transverse shear strains . In Equ.(4.61), the stiness matrices A, B
and D are given by the classical relationships for a multilayered material (Equ.(2.65) to (2.67)).
The transverse shear stiness matrix K is obtained following a method similar to that described
in Appendix A (See also Jetteur, 1991).
It follows that the element mass, stiness, piezoelectric coupling and capacitance matrices for a
piezoelectric Mindlin shell element read
[/] =
_

m[^]
T
[^]d (4.62)
[K
uu
] =
_

[B]
T
_
_
A B 0
B D 0
0 0 K
_
_
[B]d (4.63)
[K
u
] =
_

[B]
T
_
_
. . . c
T
k
. . .
. . . c
T
k
z
mk
. . .
. . . 0 . . .
_
_
d (4.64)
[K

] =
_

_
.
.
.
0

k
/h
k
0
.
.
.
_

_
(4.65)
[K
u
] = [K
u
]
T
(4.66)
where we have introduced
c
k
=
_
e
31
e
32
0
_
k
[R
S
]
k
(4.67)
and we have used the fact that
[R
T
]
1
k
_
e
31
e
32
0
_
T
k
= c
T
k
(4.68)
64 4. Finite element approach
4.6 Implementation
Referring to the electromechanical analogy mentioned in Table 4.1 (4.2), the electrical charge
brought to the system can be considered similarly to the applied forces, and the voltages similarly
to the nodal displacements. To implement the piezoelectricity into the commercial nite element
package Samcef, the strategy adopted was to modify existing elements in the Samcef library.
More precisely, the piezoelectric coupling has been added at the element level by complementing
the integrated stiness matrix with the piezoelectric coupling and capacitive contributions. The
voltages being similar to nodal displacement and the electric charges brought to the system
similar to external forces, the assembly and resolution algorithms remain. However, some bad
numerical conditioning is likely to appear due to the several order of magnitude dierence
between the numerical values of the elds involved. If displacements are written in meters
(m), forces in Newtons (N), voltages in Volts (V), electrical charges in Coulomb (Cb), then
elastic coecients are written in Pascals (N/m
2
or Pa), capacitances in Farads (Cb/V or F)
and piezoelectric coupling coecients in (Cb/m
2
). Taking the example of PVDF with typical
dimensions ( 1 mm thick, with elements of 1 cm
2
), the orders of magnitude are: 10
3
to
10
4
for mechanical stiness terms, 10
5
to 10
2
for the piezoelectric coupling terms and 10
9
for the capacitance terms; there are 13 orders of magnitude dierence between terms of the
complemented stiness matrix. A suitable choice of units reduces this dierence signicantly
(e.g. voltages in kV and electrical charges in mCb would lead to a 6 orders of magnitude
dierence).
With the cooperation of Samtech s.a., electrical dof were added (1 per piezoelectric layer for the
shell elements, 1 per node for the volume elements) and a new material of type piezoelectric has
been added.
Both rectangular and triangular Mindlin shell elements are available, together with several
volume elements: prismatic, tetrahedric and parallelipipedic.
4.6.1 Shell implementation
As many electrical dof as piezoelectric layers have been added to the Mindlin shell elements. The
element stiness matrix is integrated over the surface of the element using a Gauss integration,
that is, a weighted sum of the values to integrate, taken in g particular points called Gauss
points (e.g. Hughes, 1987).
As we start from an existing element, the value of the derivatives of the shape function ([B])
taken at the Gauss points are already available. The piezoelectric coupling and the capacitance
contributions to the element stiness matrix [K
elt
] at each Gauss point is computed. The
generalized stiness matrix is complemented as well as the matrix of shape function derivatives.
Finite Element Modelling of Piezoelectric Active Structures 65
[K
elt
] =

g
i=1
g
i
_
_
_
_
_
_
_
_
_
_
_
[B]
T
(i)
0
0 I
n
_
_

_
A B 0 c
T
k
B D 0 . . . c
T
k
z
mk
. . .
0 0 K 0
.
.
.
.
.
.
0
c
k
c
k
z
mk
0
k
/h
k
.
.
. 0
.
.
.
_

_
_
[B]
(i)
0
0 I
n
_
_
_
_
_
_
_
_
_
_
_
(4.69)
Where g is number of Gauss points and g
i
represent the Gauss weights (i = 1, . . . g) and n is the
number of piezoelectric layers (k = 1, . . . n). The assembly takes into account the equipotentiality
condition of the electrodes; this reduces the number of electric variables to the number of
electrodes.
A voltage is specied across a piezoelectric layer by imposing a mechanical displacement to the
corresponding node. An electrical charge is brought to a piezoelectric layer by applying a force
to the corresponding node. A voltage appears as a mechanical displacement and an electrical
charge as a reaction force.
4.6.2 Transverse piezoelectric shear mode in shell elements
Implementing the multi-layered piezoelectric elements, we did not consider the piezoelectric shear
mode (dictated by coecient e
15
or e
24
); the electric and displacement elds were assumed to be
parallel to the poling direction. However, the use of a Mindlin formulation (accounting for the
transverse shear) allows us to consider transverse shear piezoelectric coupling and therefore, a
poling direction no longer normal to the plate. As the poling direction is not anymore direction
3, the piezoelectric coecients to be considered are not anymore necessarily e
31
and e
32
. The
piezoelectric coecients considered next are e
1
, e
2
, e
6
= 0, e
4
and e
5
where is put for the
material direction aligned with the direction normal to the plate.
As the transverse shear is entirely decoupled, Equ.(4.60) can be decomposed as follow:
_
N
M
_
=
_
A B
B D
_ _

0

_
+
n

k=1
_
I
3
z
mk
I
3
_
[R
T
]
1
k
_
_
_
e
1
e
2
0
_
_
_
k

k
(4.70)
T = [K] +
n

k=1
[]
1
k
_
e
4
e
5
_
k

k
(4.71)
D
k
= e
1
e
2
0
k
[R
S
]
k
[I
3
z
mk
I
3
]
_

0

_
+e
4
e
5

k
[]
k


k
h
k

k
(4.72)
The transformation matrix []
k
relates the transverse shear stresses and strains written in the
material axes (LT) to the transverse shear stresses and strains written in the composite axes
(xy). We have
_

Tz

Lz
_
= []
k
_

yz

xz
_
(4.73)
66 4. Finite element approach
_

Tz

Lz
_
= []
k
_

yz

xz
_
(4.74)
The transformation matrix is given by
[]
k
=
_
cos
k
sin
k
sin
k
cos
k
_
(4.75)
If we introduce
A
k
= e
4
e
5
[]
k
(4.76)
We have also
A
T
k
= []
1
k
e
4
e
5

T
(4.77)
and the generalized stiness matrix (4.69) can be complemented to account for the transverse
shear piezoelectric coupling contribution
[K
elt
] =

g
i=1
g
i
_
_
_
_
_
_
_
_
_
_
_
[B]
T
(i)
0
0 I
n
_
_

_
A B 0 c
T
k
B D 0 . . . c
T
k
z
mk
. . .
0 0 K A
T
k
.
.
.
.
.
.
0
c
k
c
k
z
mk
A
k

k
/h
k
.
.
. 0
.
.
.
_

_
_
[B]
(i)
0
0 I
n
_
_
_
_
_
_
_
_
_
_
_
(4.78)
By suitably choosing the coupling coecients corresponding to the application considered, this
implementation allows to consider the piezoelectric shear mode as well as the extension mode.
extension
e
1
= e
31
; e
2
= e
32
e
4
= e
5
= 0
shear
e
1
= e
2
= 0
e
4
, e
5
= e
24
, e
15
depending on the coupling characteristics and the poling direction.
This was also implemented for the sake of completeness. However, the hypothesis of a uniform
transverse shear strain distribution through the thickness is not satisfactory (See e.g. the shear
bender of Fig.5.8, 5.1.3); a more elaborate shell element would be necessary (higher order
transverse shear strain distribution). As the shear actuation requires a relatively important
thickness, a volumic approach can also be used (strategy used in 5.1.3).
Finite Element Modelling of Piezoelectric Active Structures 67
(a) Brick (T008) (b) Prism (T046) (c) Tetrahedron (T047)
Figure 4.2: 3-D volume elements from Samcef library
4.6.3 Volume implementation
The variety of 3-D volume elements available in the Samcef library (Fig.4.2) permits a good
representation of geometrically complex structures.
These elements are available at degree one (linear shape functions), degree two (quadratic shape
functions) or mixed. If each element edge is degree one, internal modes improve exural be-
haviour. These internal modes are selected by default, but it is possible to remove them. Degree
two can be as well explicitely dened by nodes, or implicitely (with automatic creation of inter-
face degrees of freedom on the edges). When one of the dimensions becomes small with respect
to the others, the shape of the element tends to be a shell. In this case, a particular constitutive
law, uncoupling the stress along the thickness, can be adopted so that the element can better
simulate shell behaviours (SAMTECH, 2000).
The idea is again to complement existing elements with electrical degrees of freedom. One
electrical degree of freedom per node has been added: the electric potential. The elements
stiness matrice is augmented accordingly with the piezoelectric coupling and capacitive terms.
The derivatives of the shape functions ([B
u
]
(i)
, [B

]
(i)
) taken at the Gauss point i are already
available. The piezoelectric coupling and the capacitance contributions to the element stiness
matrix [K
elt
] in each Gauss point are computed and left and right multiplied by the shape
functions derivatives.
[K
elt
] =

g
i=1
g
i
_
[B
u
]
(i)
0
0 [B

]
(i)
_
T
_
[K
uu
] [K
u
]
[K
u
]
T
[K

]
__
[B
u
]
(i)
0
0 [B

]
(i)
_
(4.79)
Where g is the number of Gauss points and g
i
represent the Gauss weights (i = 1, . . . g). We
have
[K
elt
] =

g
i=1
g
i
_
_
[B
u
]
T
(i)
[K
uu
][B
u
]
(i)
[B
u
]
T
(i)
[K
u
][B

]
(i)
[B

]
T
(i)
[K
u
]
T
[B
u
]
(i)
[B

]
T
(i)
[K

][B

]
(i)
_
_
(4.80)
68 4. Finite element approach
4.7 Particular electrical boundary conditions
4.7.1 Voltage driven electrodes
If the electric potential is controlled, the governing equations become
[/]

U +[K
UU
]U = F [K
U
] (4.81)
where the second term in the right hand side represents the equivalent piezoelectric loads.
Once the mechanical displacements have been computed, the electric charges appearing on the
electrodes can be computed from Equ.(4.35). From Equ.(4.81), we see that the eigenvalues
problem of the system with short-circuited electrodes ( = 0) is:
_
[K
UU
]
2
[/]
_
U = 0 (4.82)
It can be seen from Equ.(4.82) that the natural frequencies and modes shape are the same as if
there was no piezoelectric electromechanical coupling.
4.7.2 Charge driven electrodes
Conversely, open electrodes correspond to a charge condition G = 0. In this case, Equ.(4.35)
becomes
= [K

]
1
[K
U
]U (4.83)
and, upon substituting into Equ.(4.34), it becomes
[/]

U +
_
[K
UU
] [K
U
][K

]
1
[K
U
]
_
U = F (4.84)
which shows that the stiness matrix depends on the electrical boundary conditions. The piezo-
electric electromechanical coupling increases the overall stiness of the system if the electrodes
are left open (the terms of [K

] being negative); the natural frequencies are greater than if the

coupling is neglected. (See Appendix B). This phenomenon becomes more and more important
as the amount of piezoelectric material embedded increases.
4.7.3 Electrodes connected via a passive network
If the electrodes are connected via a passive electrical network of impedance matrix [Z],
Equ.(4.34) and (4.35) must be complemented by the network equation (Fig.4.3). Since the
electrical current is given by the time derivative of the electrical charge I =

G, the con-
nected electrical network equation reads, in Laplace notations:
= [Z]I = [Z(s)]sG (4.85)
Finite Element Modelling of Piezoelectric Active Structures 69
Figure 4.3: Electrodes connected to an external impedance
4.8 From nite element model to state space model
The idea behind modelling structures embedding piezoelectric actuators and sensors using nite
elements is indeed to gather the necessary information to design a good control strategy. It is
therefore necessary to interface the structural analysis software (nite element package) with
a control design software. A method to extract a state space representation of the structure
dynamics with its piezoelectric inputs and outputs from a nite element analysis is presented
hereafter.
Equ.(4.34) can be complemented with a damping term [C]

U to obtain the full equation of

dynamics and the sensor equation:
0 = [/]

U + [C]

U + [K
UU
]U + [K
U
] (4.86)
G = [K
U
] U + [K

] (4.87)
where U represents the mechanical dof, the electric potential dof, [/] the inertial matrix,
[C] the damping matrix, [K
UU
] the mechanical stiness matrix, [K
U
] = [K
U
]
T
the electrome-
chanical coupling matrix and [K

] the electric capacitance matrix.

Voltage actuation and charge sensing are considered. Actuation is done by imposing a voltage
on the actuators and sensing by imposing a zero voltage ( = 0) and measuring the
reaction (electric) charge G appearing on the sensors.
What do we want to do ?
We want to reduce the electromechanical system to its actuation inputs and sensor outputs,
keeping its dynamic behaviour; we want to get a state space model of the system to be used in
a control design process.
What is the strategy ?
We have basically three categories of electrodes:
1. unused (for control) electrodes (with various boundary conditions):
(u)

2. actuator electrodes (imposed voltage):

(i)

70 4. Finite element approach

3. sensor electrodes (zero voltage), the reaction (electric) charge G
(o)
is sought
The unused electrodes are considered as part of the mechanical system; the related dof are
condensed according to the boundary conditions (See 4.7) and the mass, damping, stiness,
piezoelectric coupling and capacitance matrices are adapted in consequence (the notation re-
mains for the sake of clarity).
0 = [/]

U + [C]

U + [K
UU
]U + [K
(i)
U
]
(i)
(4.88)
G
(o)
= [K
(o)
U
]U + [K
(oi)

]
(i)
(4.89)
The strategy is to use a truncated modal decomposition. A solution of the form U =

k
Z
k
x
k
(t) = [Z]x(t) is considered, it consists in a linear combination of the mode shapes
Z
k
. The mode shapes Z
k
and the corresponding eigenfrequencies
k
are solutions of the
classical eigenproblem with closed-circuit electrical boundary conditions on both actuator and
sensor electrodes.
([K
UU
]
k
[/]) Z
k
= 0 (4.90)
and x
k
(t) represents the modal amplitude of mode k.
Finite element problems have usually a large number of degrees of freedom, especially if the
geometry is complicated, because of the diculty of accurately representing the stiness of the
structure. This number of degrees of freedom is unnecessarily large to represent the structural
response in a limited bandwidth. If a structure is excited by a band-limited excitation, its
response is dominated by the n normal modes whose eigenfrequencies belong to the bandwidth of
the excitation. The solution of Equ.(4.90) can often be restricted to these n modes. The number
of degrees of freedom contributing eectively to the response is therefore reduced drastically in
modal coordinates.
Equ.(4.88) and (4.89) become
0 = [/][Z] x + [C][Z] x + [K
UU
][Z]x + [K
(i)
U
]
(i)
(4.91)
G
(o)
= [K
(o)
U
][Z]x + [K
(oi)

]
(i)
(4.92)
and left-multiplying Equ.(4.93) by [Z]
T
:
0 = [Z]
T
[/][Z] x + [Z]
T
[C][Z] x + [Z]
T
[K
UU
][Z]x + [Z]
T
[K
(i)
U
]
(i)
(4.93)
G
(o)
= [K
(o)
U
][Z]x + [K
(oi)

]
(i)
(4.94)
Using the orthogonality properties of the mode shapes (e.g. Preumont, 1997, Chap. 2).
[Z]
T
[/] [Z] = diag(
k
) (4.95)
[Z]
T
[K] [Z] = diag(
k

2
k
) (4.96)
and a classical damping
[Z]
T
[C][Z] = diag(2
k

k
) (4.97)
Finite Element Modelling of Piezoelectric Active Structures 71
the dynamic equations of the system in the state space representation nally read:
_
x
x
_
=
_
0 I

2
2
_ _
x
x
_

_
0

1
Z
T
K
(i)
U
_
_

(i)
_
(4.98)
_
G
(o)
_
=
_
K
(o)
U
T
Z 0
_
_
x
x
_
+ [D
HF
]
_

(i)
_
(4.99)
where
the modes shape [Z],
the eigenfrequencies [] = diag(
k
),
the modal masses [] = diag(
k
),
the modal electric charge on the sensors [K
(o)
U
][Z], and
the modal electric charge on the actuators, transposed (by reciprocity) [Z]
T
[K
(i)
U
] , rep-
resenting the participation factor of the actuators to each mode,
are obtained from a dynamic nite element analysis.
[] = diag(
k
) are the modal classical damping ratios of the considered structure
and
[D
HF
] is the feedthrough component of the frequency response function; part of the output
is proportional to the input. It is calculated as follow:
As distinct actuator and sensor electrodes are considered, there is no cross-capacitance terms;
the capacitance matrix [K
(oi)

] matrix is zero and will be omitted next. If we consider the steady

state response to an harmonic input voltage of frequency , the response will also be harmonic
with the same frequency. Equ.(4.93) and (4.94) are decoupled for each mode and can be written:

k
_

2
x
k
+2
k

k
x
k
+
2
k
x
k
_
= [Z
k
]
T
[K
(i)
U
]
(i)
for each mode k (4.100)
G
(o)
=

k
[K
(o)
U
]
T
[Z
k
]x
k
(4.101)
and the frequency response function [(()] dened by G
(o)
= [(()]
(i)
reads
[(()] =

k
_
[K
(o)
U
]
T
[Z
k
]
__
[Z
k
]
T
[K
(i)
U
]
_

k
_

2
+2
k

k
+
2
k
_ (4.102)
The frequency response function [(()] can be written in the sum of the response of the modes
inside the bandwidth (that respond dynamically) and outside the bandwidth (that respond
statically).
[(()] =

n
k=1
_
[K
(o)
U
]
T
[Z
k
]
__
[Z
k
]
T
[K
(i)
U
]
_

k
_

2
+2
k

k
+
2
k
_ +
72 4. Finite element approach

...
k=n+1
_
[K
(o)
U
]
T
[Z
k
]
__
[Z
k
]
T
[K
(i)
U
]
_

2
k
(4.103)
Considering that the stationary part of frequency response function ( = 0)
[((0)] =

k
_
[K
(o)
U
]
T
[Z
k
]
__
[Z
k
]
T
[K
(i)
U
]
_

2
k
(4.104)
can also be given by (from Eq.4.88 and Eq.4.89: static response)
[((0)] = [K
(o)
U
]
T
[K
1
UU
][K
(i)
U
] (4.105)
One can write that:
[(()] =

n
k=1
_
[K
(o)
U
]
T
[Z
k
]
__
[Z
k
]
T
[K
(i)
U
]
_

k
_

2
+2
k

k
+
2
k
_ +
_
[K
(o)
U
]
T
[K
1
UU
][K
(i)
U
]
_

n
k=1
_
[K
(o)
U
]
T
[Z
k
]
__
[Z
k
]
T
[K
(i)
U
]
_

2
k
(4.106)
The rst term is the dynamic response of the modelled modes, the two last terms form the
static contribution [D
HF
] of the unmodelled modes (outside the bandwidth) to the frequency
response function; it is often called the residual mode (Preumont, 1997; Loix, 1998; Loix et al.,
1998). It is independent of the frequency and introduces an important feedthrough component
in the frequency response function. The second term can easily be obtained from a static nite
element analysis (Its columns are the charge appearing on the sensors when a unit voltage is
applied on each actuator) while the third is the static response of the modelled modes.
[D
HF
] =
_
[K
(o)
U
]
T
[K
1
UU
][K
(i)
U
]
_

n
k=1
_
[K
(o)
U
]
T
[Z
k
]
__
[Z
k
]
T
[K
(i)
U
]
_

2
k
(4.107)
Truncating the modal expansion of a frequency response function without introducing a residual
mode can lead to substantial errors in the prediction of the open-loop zeros. This will be
illustrated in Chapter 5 (See Fig.5.30, 5.3.1).
Such a state space representation is easily implemented in a control oriented software allowing
the designer to extract the various frequency response functions and to use the available control
design tools.
Finite Element Modelling of Piezoelectric Active Structures 73
Chapter 5
Applications
This chapter is devoted to the test cases, benchmarks and experimental validations of the mod-
elling tools presented.
5.1 Examples
5.1.1 Inuence of the electromechanical coupling on the natural frequencies
The electromechanical piezoelectric coupling is responsible for an inuence of the electrical
boundary conditions on the mechanical natural frequencies (See Appendix B).
Figure 5.1: Simply supported steel plate
To illustrate this phenomenon, a steel plate (100 cm 50 cm, thickness: 0.5 mm) entirely
covered with piezoceramics (PZT, thickness: 0.25 mm) as represented on Fig.5.1 is considered.
The plate is simply supported on two opposite sides (AA

and BB

). The grounded steel plate

acts as an electrode and 10 electrodes are regularly distributed along the length (AB A

).
Initially not connected, they are gradually grounded. A dynamic analysis is done for each
additional grounded electrode and a record of the natural frequencies is kept.
The materials characteristics are listed in Table 5.1. The FE mesh used is represented on Fig.5.2.
74 5. Applications
Figure 5.2: FE mesh: 30% grounded
The evolution of the natural frequencies with respect to the percentage of grounded electrodes
is presented on Fig.5.3. It should be noted that the results obtained with the uncoupled model
correspond to 100% of grounded electrodes; a zero voltage imposed on every piezoceramics
involves no stiening of the structure (See 4.7.1): if all the electrodes are short-circuited, the
natural frequencies and modes are as if there was no piezoelectric coupling.
Y
steel
210 (GPa)

steel
0.3

steel
7800 (kg/m
3
)
E
piezo
65 (GPa)

piezo
0.3

piezo
7800 (kg/m
3
)

r
2600
d
31
= d
32
190.10
12
(m/V)
Table 5.1: Characteristics of the materials
The data le is listed in Appendix C.1 to illustrate the way the piezoelectric characteristics and
electrical boundary conditions are specied compared to mechanical ones.
Figure 5.3: Stiening eect resulting from the piezoelectric coupling
Finite Element Modelling of Piezoelectric Active Structures 75
5.1.2 Cantilever piezoelectric plate
A cantilever steel plate 455 mm long and 50 mm wide is considered; the steel plate is 0.5 mm
thick and two piezoceramic strips 250 m thick, 55 mm long and 25 mm wide are bounded
symmetrically 15 mm from the clamp. One is used as actuator and the other as sensor. The
nite element mesh used and the three rst eigenmodes are represented on Fig.5.4.
Figure 5.4: FE modal analysis
For the description of the les and commands used to generate this example, see Appendix C.2.
The state space model generated can be used with all the tools provided with Matlab-Simulink
and the various toolboxes available. The frequency response function of the system is shown on
Fig.5.5. The dedicated Matlab macros are succintly described in Appendix D.
Figure 5.5: Open loop frequency response function
76 5. Applications
5.1.3 Shear bender
Figure 5.6: Shear bender
This is an example of a static analysis using piezoelectric volume elements. The use of the
shear mode of piezoelectric materials has been investigated by Benjeddou et al. (1997, 1998).
The proposed architecture consists in a sandwiched beam for which part of the core has been
replaced by piezoelectric material. The proposed conguration is such that, this time, the d
15
coupling coecient dictates the design (See 3.1). The electric eld is applied perpendicularly to
the poling direction, inducing a transverse shear strain. A nite element solution using sandwich
beam has been proposed by Benjeddou et al. and compared to analytical results.
The bender of Fig.5.6 is considered; it consists in a cantilever beam 100 mm long formed of a
2 mm rigid foam core sandwiched by two 8 mm thick aluminium skins. The core is partially
replaced by PZT piezoceramics to form an actuator of length a at a distance d
c
from the clamp.
A 20 V voltage is applied between top and bottom surfaces of the piezoelectric layer. The
material properties are summarized in Table 5.2.
As a rst test case, the core is totally replaced by the piezoactuator (there is no rigid foam).
The mesh is shown on Fig.5.7 and the static deformation on Fig.5.8. The comparison with the
FE and analytical results from Benjeddou et al. (1997) shows a good agreement (tip deection:
1.18 10
7
m (This study), 1.19 10
7
m (Benjeddou et al., 1997))
Figure 5.7: FE mesh Figure 5.8: Static deformation
Finite Element Modelling of Piezoelectric Active Structures 77
Aluminium
2690 (kg/m
3
)
Y 70.3 (GPa)
0.345
Foam
32 (kg/m
3
)
Y 35.3 (MPa)
0.383
PZT-5H
7730 (kg/m
3
)
[c]
_

_
126 79.5 84.1 0 0 0
79.5 126 84.1 0 0 0
84.1 84.1 126 0 0 0
0 0 0 23.3 0 0
0 0 0 0 23.0 0
0 0 0 0 0 23.0
_

_
(GPa)
[]
_
_
1.503 0 0
0 1.503 0
0 0 1.3
_
_
10
8
(F/m)
[e]
_
_
0 0 0 0 0 17
0 0 0 0 17 0
6.5 6.5 23.3 0 0 0
_
_
(Cb/m)
Table 5.2: Material properties
As a second case, an actuator of length a = 10 mm replace part of the core. Its position is set
to vary between 10 mm and 90 mm. The mesh is shown on Fig.5.9 and resulting deformations
for dierent locations of the actuator are shown on Fig.5.10. Tip deection vs actuator position
is compared to results found in (Benjeddou et al., 1997) on Fig.5.11.
Figure 5.9: FE mesh Figure 5.10: Resulting deformation
78 5. Applications
Figure 5.11: Tip deection vs actuator position
5.2 Actuation and sensing
5.2.1 Bimorph beam
The piezoelectric bimorph pointer (Fig.5.12) is a beam made of two uniaxial piezoelectric layers
with opposite polarity stacked together to obtain a bending actuator/sensor. This device can
be used for micro-actuation or strain sensing.
Figure 5.12: Bimorph pointer
When an external voltage is applied across the thickness, the induced strain generates moments
that bend the bimorph beam. The calculated static deection of the beam (Mindlin shell nite
element Samcef) is compared with the analytical solution described hereafter and with the nite
element and experimental solutions found by Hwang & Park (1993) and Tzou & Ye (1996). The
bimorph pointer is rst considered as an actuator: a unit voltage ( = 1 V ) is applied across
the thickness of a PVDF beam (length L = 100 mm, width b = 5 mm, thickness h = 1 mm).
The material properties of the considered mono-oriented PVDF are shown in table 5.3.
Finite Element Modelling of Piezoelectric Active Structures 79
1800 (kg/m
3
)
Y
1
2.0 (GPa)
Y
2
2.0 (GPa)
G
12
0.775 (GPa)
0.29
1.062 10
10
(F/m)
d
31
2.2 10
11
(Cb/N)
d
32
0 (Cb/N)
Table 5.3: PVDF properties
Actuation: analytical solution
Using the classical theory of beams (Bernoulli-Euler) and the linear piezoelectric constitutive
equations (Equ.2.34), the bending moment M reads:
M =
bh
2
4
e
31
E
=
bh
2
4
(Y d
31
)

h
=
bh
4
d
31
Y (5.1)
and the deection:

2
z
x
2
=
M
Y I
(5.2)
z (x) =
M
Y I
x
2
2
=
3
2
d
31

h
2
x
2
(5.3)
Actuation: nite element solution
The bimorph pointer was meshed with 10 identical rectangular shell elements along the length
and clamped at one end. One should note that the use of shell elements will induce unwanted
stress concentration near the clamping points due to the Poisson eect. Figure 5.13 shows the
comparison between the analytical, nite element, and experimental results found by Tzou &
Ye (1996) and the analytical and nite element results of the present study.
Sensing: nite element solutions
Conversely, the piezoelectric bimorph beam can be used as a sensing device: a 10 mm tip
deection is imposed and the output voltage appearing over the electrodes is calculated (the
electrical circuit being left open).
80 5. Applications
Figure 5.13: Static deection
a. (Tzou & Ye, 1996)
b. Present study
Two cases for the electrical boundary conditions are considered (two sensor electrodes distribu-
tions). The rst case consists in two electrodes sandwiching the whole beam resulting in one
unique sensor averaging the strain over the length of the beam. The second case consists in ve
identical pairs of electrodes sandwiching the beam and regularly distributed along the length
(Fig.5.14) resulting in ve sensors discretizing the beam. The results are compared to those
found by Hwang & Park (1993) using a quadrangular pure bending plate element neglecting the
transverse shear, in Fig.5.15.
Figure 5.14: Bimorph sensor
Finite Element Modelling of Piezoelectric Active Structures 81
Figure 5.15: Results
This case has also been used to benchmark the volume elements. The result given by a 2 25
volume elements mesh is compared to the 5 shell elements mesh result on Fig.5.16.
Figure 5.16: Comparison between shell and volume solutions
82 5. Applications
5.2.2 C-blocks
Commonly used piezoelectric actuators produce either high forces coupled with small deections
(stacks) or large deections coupled with low forces (bimorph). To ll the gap between these
two types of actuators, a curved piezoelectric bender actuator has been recently proposed:
the C-block (Moskalik & Brei, 1997). It consists in a semi-circular ring shell clamped at the
Figure 5.17: C-block actuator
Figure 5.18: FE mesh
bottom edge (Fig.5.17). Three congurations are considered: one piezoelectric layer (PVDF lm,
d
31
= 2.3 10
11
m/V) bonded to a substrate layer, two piezoelectric layers and four piezoelectric
layers bonded together. The curved shells are 22 mm wide and have a radius to the neutral axis
R
n
respectively equal to 14.3 mm, 13.6 mm and 14.9 mm. The properties of the materials are
summarized in Table 5.4.
Layers Width Thickness Young
(mm) (m) Modulus (MPa)
PVDF 22 52 2900
Electrode 17.5 6.5 700
Substrate 20 25 6500
Bonding 22 25 1900
Table 5.4: Materials properties
In this study, they are modelled using a 4 20 mesh (Fig.5.18) of quadrangular multilayer
piezoelectric elements. The comparison between FE prediction results and the analytical and
experimental results from Moskalik & Brei for the three congurations is shown on Fig.5.19. The
load consists of a static voltage applied across the electrodes of each piezoelectric layer (400 V
to +400 V). The tip deection of the actuator with respect to the applied voltage is considered.
The theoretical model used by Moskalik & Brei is based on a Bernoulli-Euler beam formulation
and the Hamiltons principle and neglects the membrane load term (small compared to the
bending moment term).
Finite Element Modelling of Piezoelectric Active Structures 83
Figure 5.19: C-block results:
84 5. Applications
5.2.3 Membrane eects on nearly colocated control systems
Figure 5.20: Cantilever plate with piezoceramics: experimental setup
In order to resolve the technological issues related to the practical realisation of a nearly colocated
actuator/sensor system using bonded piezoceramics, a simple cantilever plate was used. Impor-
tant dierences in the experimental frequency response functions for similar actuator/sensor
positions appeared. Consider the cantilever plate represented on Fig.5.20; the steel plate is
0.5 mm thick and four piezoceramic strips of 250 m thickness are bounded symmetrically as
indicated in the gure, 15 mm from the clamp. The material properties are summarized in Table
5.5. The size of the piezos is respectively 55 mm25 mm for p
1
and p
3
, and 55 mm12.5 mm
for p
2
and p
4
. p
1
is used as actuator while the sensor is taken successively as p
2
, p
3
and p
4
.
The experimental frequency response functions between a voltage applied to p
1
and the electric
charge appearing successively on p
2
, p
3
and p
4
when they are connected to a charge amplier
are shown on Fig.5.21.
Figure 5.21: Experimental results: frequency response functions between p
1
and p
i
We note that the frequency response functions, particularly the location of the zeros, vary sub-
stantially from one conguration to the other. This is because the frequency response functions
of nearly colocated control systems are very much dependent on local eects, in particular the
Finite Element Modelling of Piezoelectric Active Structures 85
membrane strain in the thin steel plate between the piezo patches. This can only be accounted
for by nite elements.
E
steel
210 (GPa)

steel
0.3

steel
7800 (kg/m
3
)
E
piezo
65 (GPa)

piezo
0.3

piezo
7800 (kg/m
3
)

r
2600

0
8.854.10
12
(F/m)
d
31
= d
32
205.10
12
(m/V)
Table 5.5: Materials properties
Figure 5.22: Simulation results: frequency response functions between p
1
and p
i
Figure 5.22 shows the numerical results corresponding to the three sensor congurations (a modal
damping ratio of = 0.5% was assumed in the numerical simulations); they agree reasonably
well with the experiments, in particular for what concerns the location of the zeros. The study
was conducted with the three meshes of Figure 5.23. Surprisingly, no signicant dierence
appears in the frequency response functions which are almost identical; this can be further
assessed from Table 5.6 which compares the frequency dierence between the poles and zeros of
the conguration (p
1
/p
3
) for the three meshes.

1
/
1

2
/
2
Mesh 1 0.0340 0.0618
Mesh 2 0.0340 0.0610
Mesh 3 0.0339 0.0614
Experiment 0.0455 0.0674
Table 5.6: Inuence of the mesh on the separation between the poles and the zeros for the
conguration (p
1
/p
3
)
86 5. Applications
Figure 5.23: FE meshes
5.3 Vibroacoustics
This section illustrates the situation of shell structures with embedded piezoelectric actuators
and sensors where they are nearly colocated. It, once again, stresses the importance of membrane
components on the zeros of the frequency response function and suggests means of improving
the performance when anisotropic piezoelectric material is used. The use of array sensors for
modal ltering and volume velocity sensing is also considered.
5.3.1 ASAC plate
The ASAC plate
1
(Active Structural Acoustical Control) is a volume velocity control device
based on the principle of the QWSIS sensor (See 3.3.2) . The QWSIS sensor is based on the
discretization of the plate sensor into narrow strips of width (Fig.5.24); each strip is considered
as a beam covered with piezoelectric material with quadratically shaped electrodes.
Figure 5.24: Discretisation of the active plate into narrow quadratic devices
1
The ASAC panel was developed and built under the Research Project DAFNOR
Finite Element Modelling of Piezoelectric Active Structures 87
The ASAC plate is covered on both side with piezoelectric material; the quadratic shaping of the
electrodes does not only provide a volume velocity sensor, but also a uniform pressure actuator
as a consequence of the duality of actuation and sensing properties of piezoelectric devices.
(a) Piezoelectric lm on a beam structure (b) Current amplier
Figure 5.25: Piezoelectric beam device
For a piezoelectric beam sensor of quadratic shape, the output signal of a current amplier
(Fig.5.25(b)) is proportional to the volume velocity of the beam (Eq.(3.42), 3.3.2):

out
(t) = 8e
31
R
f
h
L
2
_
L
0
w(x) dx V
vol
Similarly, a beam actuator with an electrode of quadratic shape is equivalent to a uniform
distributed load acting along the beam (See Eq.(3.7)).
F
0
= e
31
b

p
(x) h
Figure 5.26: ASAC experimental setup Figure 5.27: FE mesh
The main idea behind the ASAC panel is to realize a perfectly colocated actuator/sensor pair
to control the volume velocity. The control system consists in a clamped 1 mm thick plate of
88 5. Applications
aluminium (420 mm320 mm) covered on both sides with 0.5 mm thick piezoelectric PVDF
lms (400 mm300 mm). For the actual laboratory model (Fig.5.26) the actuation and sensing
layer electrodes consist of 24 strips. The material properties are summarized in Table 5.7. The
direction of smaller piezoelectric coupling coecient (e
32
) is perpendicular to the strips. The
nite element mesh is shown on Fig.5.27.
Aluminium
Y 71 (GPa)
0.3
2800 (kg/m
3
)
PVDF
Y 2.7 (GPa)
0.29
1800 (kg/m
3
)
d
31
1.8 10
11
(Cb/N)
d
32
0.3 10
11
(Cb/N)

r
2600
Table 5.7: Material properties
Figure 5.28: ASAC plate: Sensing
Figure 5.28 shows the comparison between the experiment
2
and the FE analysis for the plate
used as a volume velocity sensor; the frequency response function between an incident sound
pressure (provided by a loudspeaker in the experiment and assumed to be a uniform pressure in
2
Experimental datas kindly provided by Kris Henrioulle (KUL-PMA)
Finite Element Modelling of Piezoelectric Active Structures 89
the FE analysis) and the volume velocity sensor signal is represented.
Figure 5.29: Open loop frequency response functions of the ASAC plate
The panel was next considered as a control device; the open-loop frequency response function
between actuator and sensor was determined experimentally
2
and is shown on Fig.5.29 (dotted
lines). Since the performance of the control system is to a large extend related to the distance
between the poles and the zeros of the open-loop frequency response function, these results were
considered as disappointing, contrary to simplied analytical predictions which indicated far
better performances (Gardonio et al., 1999). At rst, this lack of performance was attributed
to imperfect alignment of top and bottom layers (non colocated actuator/sensor pair) or to an
electrical coupling due to the wiring.
In fact, the nite element based simulations have shown that this lack of controllability is actually
due to local membrane eects (Piefort & Henrioulle, 2000), neglected in the rst analytical
models together with the static contribution of the unmodelled high frequency modes (also
called residual mode) (See 4.8 and Preumont (1997), Chap. 2).
In a rst attempt to model the open-loop frequency response function of the ASAC panel using
nite elements, the agreement of results with the experiment was rather unsatisfactory (Fig.5.29,
FE #1). It appeared soon that the boundary conditions were not those of a clamped plate: in
the actual experiment, the plate was almost free to move in its plane. The in-plane movement of
the plate results in an even stronger inuence of the membrane components and, therefore, in a
stronger in-plane mechanical coupling between actuator and sensor. This induces an important
feedthrough term in the frequency response function: a substantial part of the strain induced
by the actuator induces directly membrane strain in the sensor, without contributing to the
transverse displacement which produces the volume velocity (useful control). The frequency
90 5. Applications
response function of (Fig.5.29, FE #2) was obtained by freeing the in-plane movement of the
plate in the nite element model; it shows a very good agreement with the experimental result.
The inuence of the residual mode on the open-loop frequency response function is illustrated on
Fig.5.30 for the two considered boundary conditions above (perfect clamp and in-plane free). The
rst twenty eigenmodes were extracted from the dynamic analysis and taken into account for the
state space representation. The inuence of the residual mode is independent of the frequency
and introduces an important feedthrough component in the frequency response function. By not
taking it into account, the in-plane coupling is almost completely washed out (because in-plane
vibration modes are quite higher in frequency, outside the bandwidth, and therefore unmodelled)
leading to an incorrect prediction of the control system performances.
Figure 5.30: Eect of the residual mode (in addition to the 20 modelled modes) on the predicted
open-loop frequency response functions
Note that, in the current design, the in-plane coupling is particularly strong because the direction
of higher piezoelectric eect (e
31
e
32
) for the sensor and the actuator are parallel; the most
important strain is induced in a direction parallel to the direction of the strips and the sensor has
the highest sensitivity to the strain in the direction of the strips. The actuation and sensing strips
being layed in the same direction for the ASAC plate, it is in the worst possible conguration.
However, for the actuator and the sensor taken separately, the direction of the strips has no
inuence on their characteristics. From this observation, the idea raised that the feedthrough
component could be substantially reduced by using sensor and actuator strips perpendicular to
each other.
Finite Element Modelling of Piezoelectric Active Structures 91
5.3.2 Alternative cross-ply design of the ASAC plate
Figure 5.31: FE mesh
In the current design, the in-plane strain induced by a voltage in the direction of the highest
sensitivity of the sensor (e
31
) is directly related to e
31
while the in-plane strain induced in
the direction of the lowest sensitivity of the sensor (e
32
) is directly related to e
32
; neglecting
the Poisson eect, we have a feedthrough factor related to e
2
31
+ e
2
32
. In a cross-ply design,
this feedthrough factor would be related to 2e
31
e
32
; assuming a piezoelectric anisotropy ratio
(e
32
= e
31
), an in-plane feedthrough term reduction of about 2/(1 +
2
) can be expected,
which may be substantial for small ( = 0.2 is a common typical value).
Figure 5.32: Open-loop frequency response functions
By using sensor strips perpendicular to the actuator strips, the control device would then exhibit
a cross-ply actuator/sensor architecture and the in-plane feedthrough term would be greatly
92 5. Applications
reduced. The FE-based tools allow to modelize such architectures quite easily and to extract
the corresponding frequency response functions to verify if this alternative is any better. The
mesh used is represented on Fig.5.31; the sensor electrode forms a right angle with the actuator
electrode. The comparison of the frequency response functions between the voltage applied to
the actuator layer and the charge measured on the sensor layer for the parallel and cross-ply
architectures for two piezoelectric anisotropy ratios are represented on Fig.5.32.
Indeed, the distance between the poles and zeros of the frequency response function is much
larger for the cross-ply conguration, as compared to the parallel conguration, and the dis-
tance increases when the piezoelectric anisotropy ratio of the material decreases. As a result,
improved closed-loop performances may be expected from the cross-ply design.
5.3.3 Array sensor
Figure 5.33: Principle of the volume displacement sensor
A noise radiation sensor consisting of an array of independent piezoelectric patches connected
to an adaptive linear combiner was proposed (Preumont et al., 1999); the piezoelectric patches
are located at the nodes of a rectangular mesh. The electric charges Q
i
induced on the various
patches by the plate vibration are the independent inputs of a multiple input adaptive linear
combiner (Fig.5.33). The coecients
i
of the linear combiner are adapted in such a way that
the mean-square error between the reconstructed volume displacement (or velocity) and either
numerical or experimental data is minimized. It must be noted that the same array sensor can
also be used as modal lter by suitably adapting the coecients
i
of the linear combiner.
This strategy can be used for reconstructing the volume displacement of a baed plate with
arbitrary boundary conditions. If the piezoelectric patches are connected to current ampliers
instead of charge ampliers, the output signal becomes the volume velocity instead of the volume
displacement.
Finite Element Modelling of Piezoelectric Active Structures 93
Figure 5.34: Array sensor: experimental setup Figure 5.35: FE mesh
The laboratory demonstration model (Fig.5.34) consists in a glass plate (54 cm124 cm, 4 mm
thick) mounted in a standard window tting and covered with an array of 4 by 8 piezoelectric
patches (PZT piezoceramics - 13.75 mm25 mm, 0.25 mm thick). The materials properties are
summarized in Table 5.8.
Glass plate
Y 72 (GPa)
0.22
2500 (kg/m
3
)
PZT
Y 69 (GPa)
0.3
7800 (kg/m
3
)
d
31
205 10
12
(Cb/N)
d
32
205 10
12
(Cb/N)

r
2600
Table 5.8: Material properties
The nite element mesh used for the numerical analysis is shown on Fig.5.35. The rst thirty
eigenmodes were extracted from the dynamic analysis and taken into account for the state space
representation. We used the opportunity given by the scanner laser vibrometer to measure the
velocity of an array of points over the glass plate to deduce the volume velocity. The excitation
used was provided by two shakers actuating the plate directly. Figure 5.36 shows the comparison
between the frequency response functions between the excitation of Shaker #1 (in the center of
the glass plate) and, respectively, sensors #7, #14 and the volume velocity obtained by nite
94 5. Applications
element analysis and experimentally.
Figure 5.36: Frequency response functions /Shaker #1
Finite Element Modelling of Piezoelectric Active Structures 95
5.4 Design against fatigue
Fatigue damage is one of the most frequent form of failure of metallic structures. With the
increasing demand for high performance structures, the fatigue damage has become more and
more important, in particular for metallic structures subjected to complex multiaxial loads due
to random vibrations.
Figure 5.37: Piezoelectric elements location and rst vibration modes
The active damping can be considered as a solution to reduce the vibrations level and at the same
time, reduce the weight and increase the fatigue lifetime of such structures. Simulation tools
allowing the designer to include directly piezoelectric actuators and sensors in a nite-element
model have been developed as well as a multiaxial random fatigue matlab toolbox (Pitoiset et al.,
1998; Pitoiset & Preumont, 2000; Pitoiset, 2001) allowing the direct assessment of the fatigue
lifetime of a structure from a spectral analysis.
The quantication of the damage reduction that can be achieved thanks to an active damping
can then be performed very easily, directly from a modal and a spectral analysis with nite-
elements. This has been done on a simple example: A simply supported rectangular steel
plate (15.24 cm30.48 cm0.813 mm) is considered. It is subjected to a band limited white
noise random pressure eld with perfect spatial coherence; Its PSD (Power Spectral Density)
is
pp
() = 400 Pa
2
/rad/s between 0 Hz and 1000 Hz. The rst three vibration modes are
within the bandwidth of excitation. A piezoelectric actuator and a piezoelectric sensor (PZT
15.24 cm30.48 mm0.2 mm) are bonded to the surface as shown on Fig.5.37.
The multiaxial random fatigue matlab toolbox was used after a FE spectral analysis to evaluate
the fatigue damage ratio (per second) over the plate. The state space form of the system is built
under Matlab directly from the results of a FE modal analysis. A controller is designed using
Matlab functionalities, in this case: a PPF tuned on the rst mode (Positive Position Feedback).
The new modal damping ratios and eigen frequencies can be used in the multiaxial random
fatigue matlab toolbox to evaluate the fatigue damage ratio over the controlled plate. The
96 5. Applications
comparison between damage maps for the plate without and with active damping is shown on
Fig.5.38. The same comparison for the PSD of the stress components in an element in the center
of the plate are shown on Fig.5.39.
Figure 5.38: Damage map (Von Mises) with and without control (Log Scale)
Figure 5.39: PSD of the stress components in the center of the plate with and without control
Finite Element Modelling of Piezoelectric Active Structures 97
Chapter 6
Conclusions
After an historical overview of piezoelectricity, some general considerations about common piezo-
electric materials and other smart materials have been presented in Chapter 1.
In Chapter 2, the piezoelectric constitutive equations were rst introduced with a unidimensional
example using an electromagnetic approach. The general thermopiezoelectric constitutive equa-
tions were then established starting from the thermodynamic principles and simplied for a
Kirchho laminate embedding piezoelectric layers.
Various piezoelectric actuation and sensing mechanisms have been investigated in Chapter 3.
The pure bending beam model has been generalized to the piezo-isotropic shell. The reciprocity
between actuation and sensing has been stressed.
After a short review of nite element modelling of structures embedding piezoelectric devices,
a nite element formulation for an electromechanically coupled piezoelectric structure start-
ing from the constitutive equations has been presented in Chapter 4. The particular case of
piezolaminated shells has been adressed. General multilayered piezoelectric shell elements have
been succesfully integrated into the nite element package Samcef (Samtech s.a.). Volume ele-
ments have also been implemented. A method to extract a state space model of a piezoelectric
input/output system from the modal nite element analysis of the structure has been presented.
Numerous applications of the developed tools in structure embedded actuation and sensing
and in vibration and vibroacoustic control have been described in Chapter 5 and shown that
good performances are achieved. The importance of the in-plane components in the open-
loop frequency response functions has been illustrated. More complex modes of piezoelectric
actuation/sensing can be modelled using volume elements; it is illustrated with the modelling
of a shear bender.
Original aspects of this study:
a full implementation of a library of piezoelectric elements (triangular and rectangular
shells, brick, prism and tetrahedron volume elements) in a commercial nite element pack-
age is now available (Samcef V8.1-4, October 2000) and is suitable for industrial problems
resolution.
the equivalent loads approach has been extended to electrodes of arbitrary contour; the
98 6. Conclusions
strong duality between actuation and sensing properties of piezoelectric devices has been
further assessed.
the inuence of the membrane component on the location of the zeros of the frequency
response functions of colocated piezoelectric control systems has been extensively illus-
trated.
the tools developed allowed an accurate modelling of the ASAC panel, a better under-
standing of the dierent physical phenomenons involved has been gained; this allowed
to explain the poor performances of the setup and to propose and test numerically an
alternative design.
the developed software tool has been applied to several ongoing research projects at the
Active Structures Laboratory, namely the design and simulation of array vibroacoustics
sensors and the reduction of fatigue damage through active damping.
Finite Element Modelling of Piezoelectric Active Structures 99
Appendix A
Transverse shear in beams and plates
A.1 Beams
In this section, a theory for plane bending of beams is presented (Batoz & Dhatt, 1990). The
inuence of transverse shear strain is taken into account. This model, based on the plane
sections hypothesis, is often referred to as Timoshenkos model. It is the generalization of more
classical models neglecting transverse shear strain and generally referred to as Bernoulli-Euler
models. It can also be seen as the unidimensional version of the plates and shells model known
as Reissner-Mindlin model presented in section A.2. It is a rst order theory, the displacement
being, at most, linear over the thickness.
A.1.1 Geometry and kinematics
A beam is a solid of which two dimensions (directions y, z) are small compared to the other
(direction x, taken as reference) and the smallest curvature radius (see Fig.A.1). In this section,
the following hypothesis are made:
Reference axis x is straight
the beam deforms in the xz plane (membrane, bending and shear)
Plane stress state (
zz
= 0)
Eorts transmitted: normal force N
x
(x), shear force T
z
(x) and bending moment M
y
(x).
They will be referred to as N, T and M.
A.1.2 Constitutive equations
The stress-strain relationships are
(x, z) = Y (x, z) (A.1)
(x, z) = G(x, z) (A.2)
100 A. Transverse shear in beams and plates
Figure A.1: Beam geometry
where Y and G are respectively the Youngs modulus and the shear modulus and and are
the axial and shear strain given respectively by
=
u(x, z)
x
= u
,x
(A.3)
=
u(x, z)
z
+
w(x, z)
x
= u
,z
+w
,x
(A.4)
Considering the plane section hypothesis (a plane section remains plane after deformation) also
known as Timoshenko hypothesis, one gets the displacement eld
u(x, z) = u(x) +z(x) (A.5)
w(x, z) = w(x) (A.6)
where u, w, are the small displacements between the initial and deformed congurations.
Combining equations (A.3) to (A.6), one gets
=
0
+z (A.7)
= +w
,x
(A.8)
where

0
= u
,x
is the membrane strain,
Finite Element Modelling of Piezoelectric Active Structures 101
the transverse shear strain,
=
,x
the curvature,
(x) can be seen as the rotation of section pq (gure A.2), z represents the bending strain, and
w
,x
the rotation of the reference surface (z = 0)
Figure A.2: Displacement of two points p and q
Upon integrating equations (A.1) and (A.2) over the cross-section, the resultant eorts N, M,
and T read
N =
_

d =
_

(Y
0
+zY ) d = A
0
+B (A.9)
M =
_

z d =
_

_
zY
0
+z
2
Y
_
d = B
0
+D (A.10)
T =
_

d = K (A.11)
where A, B and D, respectively the membrane, coupling and bending stinesses, are given by
A =
_

Y (x, z)d (A.12)

B =
_

zY (x, z)d (A.13)

D =
_

z
2
Y (x, z)d (A.14)
and K is the transverse shear stiness
K =
_

G(x, z)d (A.15)

The transverse shear strain dened by (A.8) is constant over the thickness. Therefore, the trans-
verse shear stress dened by (A.2) will not verify the boundary conditions on lower and upper
skins (transverse shear stress must be zero on the external surfaces) and the local equilibrium
102 A. Transverse shear in beams and plates
Isotropic Rectangular Section Theoretical

xz
=
_
1
4z
2
h
2
_
3T
h
Mindlin (no correction)

xz
=
T
h
Figure A.3: Transverse shear stress distribution
relations in x. To illustrate this, a shear stress distribution in an isotropic rectangular section
is represented on Fig.A.3.
Several propositions were made to determine a shear stiness correction factor k such that
K = kK (A.16)
where K is the shear stiness dened by equation (A.15).
Timoshenko proposed a correction factor based on the assumption that the value in the
middle of a rectangular section should be correct k = 2/3
Reissner proposed a correction factor based on the equality of the strain energy k = 5/6
Mindlin, using a dynamic approach based on the natural frequencies equality, proposed
k =
2
/12
A method to approximate the transverse shear stiness for multilayered materials will be pre-
sented in A.2.2.
A.2 Plates
In this section, the main concepts of a rst order plate theory are presented (Hughes, 1987;
Batoz & Dhatt, 1990). This theory, usually referred to as Reissner-Mindlin theory, is based
on the plane section hypothesis and accounts for the transverse shear strain. It can be seen
as a generalization of the more classical model neglecting the transverse shear strain based on
the normal conservation hypothesis generally referred to as the Kirchho model of plates. The
plane anisotropy is assumed: there is no coupling between membrane/bending deformations and
transverse shear deformation.
Finite Element Modelling of Piezoelectric Active Structures 103
A plate is a solid dened by a plane reference (xy) and a thickness h(x, y), small compared to
the other dimensions. The following hypothesis will be considered:
plane sections remain plane after deformation
plane stress state (
zz
= 0)
there is no coupling between membrane/bending deformations and transverse shear defor-
mation
the eorts transmitted are (gure A.4) the normal eort N = N
x
N
y
N
xy

T
, the
transverse shear T = T
xz
T
yz

T
, the bending moment M = M
x
M
y
M
xy

T
Figure A.4: Resultant eorts
A.2.1 Constitutive equations
Using standard engineering notations, the constitutive equations read
_

_
=
_
[H
m
] 0
0 [H
t
]
_ _

_
(A.17)
where
[H
m
] and [H
t
] are respectively the membrane/bending modulus matrix and the transverse
shear modulus matrix
104 A. Transverse shear in beams and plates
=
_
_
_

xx

yy

xy
_
_
_
, =
_

xz

yz
_
(A.18)
=
_
_
_

xx

yy
2
xy
_
_
_
, =
_
2
xz
2
yz
_
(A.19)

xx
=
u(x, y, z)
x
= u
,x
(A.20)

yy
=
v(x, y, z)
y
= u
,y
(A.21)
2
xy
=
xy
=
u(x, y, z)
y
+
v(x, y, z)
x
= u
,y
+v
,x
(A.22)
2
xz
=
xz
=
u(x, y, z)
z
+
w(x, y, z)
x
= u
,z
+w
,x
(A.23)
2
yz
=
yz
=
v(x, y, z)
z
+
w(x, y, z)
y
= v
,z
+w
,y
(A.24)
Figure A.5: Displacement of two points p and q
Considering the plane section hypothesis (Mindlin), the displacement eld is given by
u(x, y, z) = u(x, y) +z
x
(x, y) (A.25)
v(x, y, z) = v(x, y) +z
y
(x, y) (A.26)
w(x, y, z) = w(x, y) (A.27)
Finite Element Modelling of Piezoelectric Active Structures 105
where u, v, w,
x
,
y
are the small displacements between the initial and deformed congurations.
Linear strains read
=
0
+z (A.28)
=
_

xz

yz
_
=
_

x

y
_
+
_
w
,x
w
,y
_
(A.29)
where

0
=
_
_
_
u
,x
v
,y
u
,y
+v
,x
_
_
_
(A.30)
is the mid-plane membrane strain,
=
_
_
_

x,x

y,y

x,y
+
y,x
_
_
_
(A.31)
is the actual curvature, z represents the bending strain and w
,x
, w
,y
the rotations of the
reference surface (mid-plane z = 0).
Developping the resultant eorts N and M, and T one gets
N =
_
+h/2
h/2
dz = [A]
0
+[B] (A.32)
M =
_
+h/2
h/2
z dz = [B]
0
+[D] (A.33)
T =
_
+h/2
h/2
dz = [K] (A.34)
where [A], [B] and [D], respectively, the membrane, coupling and bending stiness matrix are
given by
[A] =
_
h/2
h/2
[H
m
(x, z)] dz (A.35)
[B] =
_
h/2
h/2
z [H
m
(x, z)] dz (A.36)
[D] =
_
h/2
h/2
z
2
[H
m
(x, z)] dz (A.37)
and [K] is the transverse shear stiness matrix, which is discussed below
106 A. Transverse shear in beams and plates
A.2.2 Transverse shear stiness approximation
This method (Lardeur, 1990; Batoz & Dhatt, 1990; Jetteur, 1991), based on strain energy
equivalence, matches the Reissner theory for homogeneous plates (shear stiness correction
factor k = 5/6). In the present theory, the following hypothesis are made:
there is no membrane/bending coupling: [B] = 0 (as encountered in a symmetric multi-
layered material for example)
the transverse shear stiness matrix [K] will be dened in such a way that the strain
energy density U
1
obtained with an exact distribution of shear stresses (according to the
tridimensional equilibrium equations) is equivalent to the the strain energy density U
2
based on the Reissner-Mindlin hypothesis.
a) Strain energy U
1
U
1
=
1
2
_
h/2
h/2

T
[H
t
]
1
dz (A.38)
b) Strain energy U
2
U
2
=
1
2
T
T
[K]
1
T =
1
2

T
[K] (A.39)
using Equ.(A.34).
To establish the expression for U
1
, the distribution of the transverse shear stresses is required.
This will be obtained by integrating the following tridimensional equilibrium equations (A.40)
and (A.41) over the thickness of the plate.

xx,x
+
xy,y
+
xz,z
= 0 (A.40)

xy,x
+
yy,y
+
yz,z
= 0 (A.41)
One gets

xz
=
_
z
h/2
(
xx,x
+
xy,y
) dz (A.42)

yz
=
_
z
h/2
(
xy,x
+
yy,y
) dz (A.43)
with
xz
=
yz
= 0 in z = h/2 (free surface boundary conditions).
In-plane stresses are written in relation to resultant eorts (equations (A.32) to (A.34)) assuming
pure shear/bending deformation and no coupling between membrane deformation and bending
([B] = 0). The displacement eld reads
u(x, y, z) = z
x
(x, y) (A.44)
v(x, y, z) = z
y
(x, y) (A.45)
w(x, y, z) = w(x, y) (A.46)
Finite Element Modelling of Piezoelectric Active Structures 107
so that in-plane strain and stresses become
= z (A.47)
= z [H
m
] = z [H
m
] [D]
1
M (A.48)
(A.49)
= z [/] M (A.50)
where
[/] = [H
m
] [D]
1
(A.51)
Using the tridimensional equilibrium equations (A.40) and (A.41), the equilibrium of in-plane
moments can be written:
_
h/2
h/2
z (
xx,x
+
xy,y
+
xz,z
) dz = 0 (A.52)
_
h/2
h/2
z (
xy,x
+
yy,y
+
yz,z
) dz = 0 (A.53)

_
h/2
h/2
z
xz,z
dz = M
xx,x
+M
xy,y
(A.54)

_
h/2
h/2
z
yz,z
dz = M
xy,x
+M
yy,y
(A.55)
Integrating the rst member of equations (A.54) and (A.55) by parts and assuming that
xz
=

yz
= 0 in z = h/2 (free surface boundary conditions), one gets the global equilibrium relations
for a plate:
T =
_
T
xz
T
yz
_
=
_
M
xx,x
+M
xy,y
M
xy,x
+M
yy,y
_
(A.56)
Using relation (A.50) in equations (A.42) and (A.43), the transverse shear stresses read

xz
=
_
z
h/2
z (/
11
M
xx,x
+/
12
M
yy,x
+/
13
M
xy,x
+ /
31
M
xx,y
+/
32
M
yy,y
+/
33
M
xy,y
) dz (A.57)

yz
=
_
z
h/2
z (/
31
M
xx,x
+/
32
M
yy,x
+/
33
M
xy,x
+ /
11
M
xx,y
+/
12
M
yy,y
+/
13
M
xy,y
) dz (A.58)
which can be rewritten in matrix form and using relation (A.56):
= [T
1
] T +[T
2
] (A.59)
108 A. Transverse shear in beams and plates
where
=
_
M
xx,x
M
xy,y
M
xy,x
M
yy,y
M
yy,x
M
xx,y
_
T
(A.60)
[T
1
] =
_
z
h/2
z
2
_
/
11
+/
33
/
13
+/
32
/
31
+/
23
/
22
+/
33
_
dz (A.61)
[T
2
] =
_
z
h/2
z
2
_
/
11
/
33
/
13
/
32
2/
12
2/
31
/
31
/
23
/
22
/
33
2/
32
2/
21
_
dz (A.62)
Using equation (A.59) into expression (A.38), the strain energy U
1
can be written
U
1
=
1
2
_
T

_
T
_
[(
11
] [(
12
]
[(
12
]
T
[(
22
]
_ _
T

_
(A.63)
where
[(
11
] =
_
h/2
h/2
[T
1
]
T
[H
t
]
1
[T
1
] (A.64)
[(
22
] =
_
h/2
h/2
[T
2
]
T
[H
t
]
1
[T
2
] (A.65)
[(
12
] =
_
h/2
h/2
[T
1
]
T
[H
t
]
1
[T
2
] (A.66)
The equivalence between the two expressions of the transverse shear strain energy can be written
1
2
T
T
[K]
1
T =
1
2
_
T

_
T
_
[(
11
] [(
12
]
[(
12
]
T
[(
22
]
_ _
T

_
(A.67)
which leads to use
[K] = [(
11
]
1
(A.68)
as an approximation of the transverse shear stiness matrix
Finite Element Modelling of Piezoelectric Active Structures 109
Appendix B
Longitudinal vibration of a
piezoelectric bar
In order to emphasize the electromechanical coupling phenomenon, the natural frequencies (me-
chanical characteristics) of a unidirectional piezoelectric bar with dierent electrical boundary
conditions is considered
Resonance frequency, = 0 (closed circuit)
Antiresonance frequency, I = 0 (open circuit)
The equivalent form (2.37) of the constitutive equations (2.11) and (2.12) is
T = cS hD (B.1)
E = hS + D (B.2)
Using the notations of Fig.B.1, and considering an harmonic displacement in the z direction
W(z)e
it
, the stress reads T = and the strain S = W/z.
The system of equations to be solved is formed by
the constitutive equations (B.1) and (B.2)
= c
W
z
hD (B.3)
E = h
W
z
+D (B.4)
the equation of motion (axial equilibrium of the bar)

z
=
d
2
W
dt
2
=
2
W (B.5)
110 B. Longitudinal vibration of a piezoelectric bar
Figure B.1: Piezoelectric unidimensional bar (cross section: )
the Poisson equation
D
z
= 0 (B.6)
and the electrostatic equations
I =

Q = i(D) (B.7)
=
_
L
0
E dz (B.8)
With the following mechanical boundary conditions

W
0
= iW(0) (B.9)

W
L
= iW(L) (B.10)
N
0
= (0) (B.11)
N
L
= (L) (B.12)
Upon dierentiating (B.3) with respect to the spatial coordinate z, one gets the expression

z
= c

2
W
z
2
h
D
z
(B.13)
Replacing (B.13) in (B.5) and taking into account equation (B.6), one gets the equation of
vibration in extension of the bar

2
W +c

2
W
z
2
= 0 (B.14)
of which the solution can be written
Finite Element Modelling of Piezoelectric Active Structures 111
W =
1
cos(z) +
2
sin(z) (B.15)
with

2
=

2
c
(B.16)
The constants
1
and
2
can be determined from the boundary conditions

1
=

W
0
i
(B.17)

2
=

W
L

W
0
cos(L)
i sin(L)
(B.18)
Developing (B.3) and (B.4) with (B.15), one gets
=
c
i sin(L)
_

W
0
cos ((L z))

W
L
cos(z)
_
hD (B.19)
E =
h
i
_

W
0
cos ((L z))

W
L
cos(z)
_
+D (B.20)
To write the inuence of the electrical boundary conditions on the natural frequencies, a method
based on the electromechanical impedance will be used. The equations relating explicitely all
the boundary conditions can be found developing (B.11), (B.12) and (B.8). We obtain
_
_
_
N
0
N
L

_
_
_
= Z()
_
_
_

W
0

W
L
I
_
_
_
(B.21)
Where Z is the electromechanical impedance matrix
Z() =
_

_
Z

()

h
i

h
i
h
i

h
i
1
iC
_

_
(B.22)
where C =

L
is the electrical capacitance under constant strain and Z

, the mechanical
impedance matrix, given by
Z

() =
c
i
_
tan(L)
1
sin(L)
1
sin(L)
1
tan(L)
1
_
(B.23)
112 B. Longitudinal vibration of a piezoelectric bar
Z depends only of the vibration frequency (pulsation ). The non-triviality condition for equa-
tion (B.21) under particular boundary conditions will result in an eigenvalue equation for the
frequency.
In the particular case of a free-free bar (N
0
= N
L
= 0), the characteristic equation becomes
_
_
_
0
0

_
_
_
= Z()
_
_
_

I
_
_
_
(B.24)
resonance (f
r
) = 0 (closed circuit) det(Z) = 0
antiresonance (f
a
) I = 0 (open circuit) det(Z

) = 0
If the impedance matrix Z is written
Z() =
_
_



_
_
(B.25)
Antiresonance and resonance conditions are given by
det(Z

) = ( )( +) = 0 (B.26)
det(Z) = ( )(2
2
( +)) = 0 (B.27)
= 0 cos(L) = 1 L = 2n
These conditions on results in the frequencies
f
2n
a
= f
2n
r
= n
_
c
L
2
(B.28)
It can be shown that this condition is equivalent to impose simultaneously = 0 and I = 0.
The displacement is of the form W
2n
ar
= W cos(2nz/L); The speed of the extemities are
equal in amplitude and phase.
+ = 0 cos(L) = 1 L = (2n +1).
These conditions on results in the antiresonance frequencies
f
2n+1
a
=
2n +1
2
_
c
L
2
(B.29)
The speed of the extemities are equal in amplitude and in opposition of phase.
2
2
( +) = 0, which results in
k
2

L
2
cotan
_
L
2
_
= 0 (B.30)
with the coupling factor k
2
=
h
2
c
(See e.g. IEEE std; Rogacheva, 1999).
Finite Element Modelling of Piezoelectric Active Structures 113
this condition gives the resonance frequencies. If we assume that the piezoelectric eect
brings only small changes to the natural frequencies, we can use a perturbation method,
allowing us to write L = (2n+1) +; upon developing (B.30) to the rst order, we nd

4k
2
(2n +1)
(B.31)
and the frequencies
f
2n+1
r
=
2n +1
2
_
1
4k
2
[(2n +1)]
2
__
c
L
2
(B.32)
In summary
f
2n
a
= f
2n
r
= n
_
c
L
2
(B.33)
f
2n+1
a
=
2n +1
2
_
c
L
2
(B.34)
f
2n+1
r
=
2n +1
2
_
1
4k
2
[(2n +1)]
2
__
c
L
2
(B.35)
Which denotes the inuence of the electromechanical coupling induced by the piezoelectricity.
Comparing equations (B.33) to (B.35), we see that the ratio between the resonance and the
antiresonance frequencies depends only on the coupling factor k
2
. This constitutes a practical
way of determining k
2
experimentally.
114 B. Longitudinal vibration of a piezoelectric bar
Finite Element Modelling of Piezoelectric Active Structures 115
Appendix C
Some data les
C.1 Example of a dynamic analysis using shell elements: In-
uence of the electromechanical coupling on the natural
frequencies
The Samcef data le (input for preprocessor BACON) used to generate the nite element model
of example in 5.1.1 is listed below.
! abbreviations
!===============
abre /h 2.5e-4 ! thickness
abre /Lo 1. ! length
abre /La 0.5 ! width
abre /NEl 10 ! total number of electrodes
abre /Ncov 3 ! number of grounded electrodes
!
! topology
!==========
! 1. nodes
!
.noe i 1
i 1 x (/La:/NEl) r /NEl
i (/NEl+1) y (/Lo:/NEl) q /NEl a (/NEl+1)
!
! 2. elements
!
.mai i 1 n 1 2 (/Nel+3) (/Nel+2) at 1
i 1 m 1 r (/NEl-1)
i /NEl m (/NEl+1) q (/Ncov-1) a /NEl
i (/NEl*/Ncov+1) n ((/NEl+1)*/Ncov+1) ((/NEl+1)*/Ncov+2) $
((/NEl+1)*/Ncov+/NEl+3) ((/NEl+1)*/Ncov+/NEl+2) at 2
116 C. Some data les
i 1 m 1 r (/NEl-1)
i /NEl m (/NEl+1) q (/NEl-/Ncov-1) a /NEl
.sel groupe 1 maille attr 1
groupe 1 maille attr 2
!
! 3. electrical nodes
!
.noe i 1001 z .0001
i 1002 z .0002
!
! electrodes
!
.ael attr 1 en 1001
attr 2 en 1002
!
! mechanical boundary conditions
!================================
!
.fix i 1 j ((/NEl+1)*(/NEl)+1) k (/NEl+1) c 1 2 3 6
i (/NEl+1) j ((/NEl+1)*(/NEl+1)) k (/NEl+1) c 1 2 3 6
!
! electrical boundary conditions
!================================
!
.fix i 1002 c 1
!
! Physical properties
!=====================
! 1. materials definition
!
.mat i 1 YT 0.65e11 G 0.25e11 $ ! piezo
NT 0.3 M 7800 a 0 0 0 $
PZEE (8.854e-12*2600) $ ! * permittivity
PZUE 190e-12 190e-12 ! * coupling (d31, d32)
i 2 YT 2.1e11 G 0.8e11 $ ! steel
NT 0.3 M 7800 a 0 0 0
I 3 YT 1.00e00 G 1.00e00 $ ! none
NT 0.3 M 0000 a 0 0 0
!
! 2. laminate
!
.pli mat 1 t (/h) an (0) pli 1
mat 2 t (2*/h) an (0) pli 2
mat 3 t (/h) an (0) pli 3
.lam lam 1 pli 1 2 3
.eta gr 1 lam 1 maille 1 dir 1 2 angle (0)
Finite Element Modelling of Piezoelectric Active Structures 117
gen
!
! resolution parameters
!=======================
!
.hyp mindlin
.sam nop5 -1 nop6 -1 nalg 4 nval 10
C.2 Cantilever piezoelectric plate
Four les are generated from the nite element modal analysis. lame.inp and lame.out contain
the modal electric charge, respectivey for the piezoactuator and the piezosensor. lame.mui and
lame.wi contain the modal masses and the eigenfrequecies. lame.fth comes from a static analysis
and contains the static response of the piezosensor for a unit voltage applied to the piezoactuator.
The Matlab commands listed below generate the state space model corresponding to the in-
put/output system dened above and plot the corresponding frequency response function shown
on Fig.5.4. This state space model can be used with all the tools provided with Matlab-Simulink
and the various toolboxes available.
clear all;
Name=lame
kk=1e-2;%modal damping
%dynamic properties actuator - sensor
in=readmat([Name .inp]);
out=readmat([Name .out]);
mui=readmat([Name .mui]);
wi=readmat([Name .wi]);
% static response sensor/actuator
fth=readmat([Name .fth]);
% extraction of the state space system
[ap,bp,cp,dp]=sys(mui,wi,in,out,fth,kk);
% plot of the transfer function
F=logspace(1,3,1000)/2/pi;
[Gp,Pp,Fp]=ptft(ap,bp,cp,dp,1,F);
118 C. Some data les
Finite Element Modelling of Piezoelectric Active Structures 119
Appendix D
Matlab toolbox
This appendix contains a succint description of the Matlab scripts (organized as a toolbox) used
to generate the various examples of Chap.5.
D.1 From nite element to state space
* sys.m
function [a1,b1,c1,d1]=sys(mui,wi,in,out,fth,kk);
build the system piezo-piezo with a modal damping of kk using
mass matrix: mui
frequencies: wi
input matrix: in
output matrix: out
static response: fth
D.2 Visualization tools
* ptft.m
function [G,P,F]=ptft(a1,b1,c1,d1,IU,FF);
plots the transfer functions of the system defined by the state
space matrices a1, b1, c1, d1 for the input specified by IU and using
the frequency vector FF
outputs the corresponding gain, phase and frequency vectors
* plotmesh.m
function [MX,MY,MZ,CZ]=plotmesh(elt,nod,C);
plot the mesh defined by elt (containing the number, type and the list of
120 D. Matlab toolbox
nodes for each element) and nod (node coordinates vector) and C
defining the coloration
D.3 Other tools
* readmat.m
function [V]=readmat(Name);
reads the files generated by the finite element analysis
* syst_pr.m
function [a1,b1,c1,d1]=syst_pr(out,mui,wi,phi,nco,cha,pfth,nele,kk);
builds the system pressure-pressure (direction (3)) with a modal
damping of kk using
modal mass matrix: mui
frequencies: wi
input matrices: cha, nco
output matrix: out
static response: pfth
* vol_vel.m
function vvol=vol_vel(elt,nod,phi,noun,nco,nele);
compute the volume velocity in the direction (3) from the
topology (elt,nod), the modal shape (phi, noun, nco)
amplitude dof nr
node
assuming the (nele) electrical dof are at the end !
Finite Element Modelling of Piezoelectric Active Structures 121
Bibliography
Agarwal, B. D. & Broutman, L. J., 1990, Analysis and Performance of Fiber Composites, John
Wiley, 2nd edn.
Allik, H. & Hughes, T. J. R., 1970, Finite element method for piezoelectric vibration, Interna-
tional Journal for Numerical Methods in Engineering, 2:151157.
Anderson, E. H., Moore, D. M., Fanson, J. L. & Ealey, M. A., 1990, Development of an active
member using piezoelectric and electrostrictive actuation for control of precision structures,
AIAA-90-1085-CP.
Batoz, J.-L. & Dhatt, G., 1990, Modelisation des structures par elements nis, Volume 2: poutres
et plaques, Herm`es, Paris, France.
Benjeddou, A., Trindade, M. A. & Ohayon, R., 1997, A unied beam nite element model
for extension and shear piezoelectric actuation mechanisms, Journal of Intelligent Material
Systems and Structures, 8.
Benjeddou, A., Trindade, M. A. & Ohayon, R., 1998, A new shear actuated smart structure
beam element, AIAA 98-1922.
Brissaud, M., 1986, Caracterisation Tensorielle et Modelisation des Ceramiques Piezoelectriques
Utilisees en Imagerie Ultrasonore, Ph.D. thesis, INSA-Lyon and Universite Claude Bernard -
Lyon I, Lyon, France.
Burke, S. E. & Hubbard Jr, J. E., 1987, Active vibration control of a simply supported beam
using a spatially distributed actuator, IEEE Control Systems Magazine, pp. 2530.
Cady, W. G., 1946, Piezoelectricity; an Introduction to the Theory and Applications of Elec-
tromechanical Phenomena in Crystals, Mc Graw-Hill, New York, London.
Callahan, J. & Baruh, H., 1996, Active control of exible structures by use of segmented
piezoelectric elements, Journal of Guidance, Control and Dynamics, 19(4):808815.
Carlson, J. D., Catanzarite, D. M. & St Clair, K. A., 1995, Commercial magneto-rheological
uids devices, 5th International Conference on Electro-Rheological , Magneto-Rheological
Suspensions and Associated Technology, Sheeld.
Carlson, J. D. & Sproston, J. L., 2000, Controllable uids in 2000 - status of ER and MR uid
technology, 7th International Conference on New Actuators, Bremen, Germany.
122 BIBLIOGRAPHY
Chattopadhyay, A., Li, J. & Haozhong, G., 1999, Coupled thermo-piezoelectric-mechanical
model for smart composite laminate, AIAA Journal , 37(12).
Chen, S.-H., Wang, Z.-D. & Liu, X.-H., 1997, Active vibration control and suppression for
intelligent structures, Journal of Sound and Vibration, 200(2):167177.
Clark, A. E., 1980, Magnetostrictive rare earth-Fe
2
compounds, in Ferromagnetic Materials
(Wolfhart, E. P., editor), vol. 1, chap. 7, pp. 531589, North-Holland, Amsterdam.
Crawley, E. F. & Anderson, E. H., 1989, Detailed models of piezoceramic actuation of beams,
AIAA 89-1388-CP.
Crawley, E. F. & de Luis, J., 1987, Use of piezoelectric actuators as elements of intelligent
structures, AIAA Journal , 25(10):13731385.
Crawley, E. F. & Lazarus, K. B., 1991, Induced strain actuation of isotropic and anisotropic
plates, AIAA Journal , 29(6):944951.
Dimitriadis, E. K., Fuller, C. R. & Rogers, C. A., 1991, Piezoelectric actuators for distributed
vibration excitation of thin plates, Transactions of the ASME, Journal of Vibrations and
Acoustics, 113:100107.
Eer Nisse, E. P., 1967, Variational method for electroelastic vibration analysis, IEEE Transac-
tions On Sonics And Ultrasonics, 14(4):153160.
Forward, R., 1981, Electronic damping of orthogonal bending modes in a cylindrical mast
experiment, AIAA Journal of Spacecraft, 18(1):1117.
Gardonio, P., Lee, Y., Elliott, S. & Debost, S., 1999, Active control of sound transmission
through a panel with a matched PVDF sensor and actuator pair, Active 99, Fort Lauderdale,
Florida, USA.
Ha, S. K., Keilers, C. & Chang, F. K., 1992, Finite element analysis of composite structures
containing distributed piezoceramic sensors and actuators, AIAA Journal , 30(3):772780.
Heyliger, P., Pei, K. C. & Saravanos, D., 1996, Layerwise mechanics and nite element model
for laminated piezoelectric shells, AIAA Journal , 34(11):23532360.
Hong, C.-H. & Chopra, I., 1999, Modeling and validation of induced strain actuation of com-
posite coupled plates, AIAA Journal , 37(3).
Hughes, T. J. R., 1987, The Finite Element Method, Prentice-Hall International Editions.
Hwang, W. S. & Park, H. C., 1993, Finite element modeling of piezoelectric sensors and actu-
ators, AIAA Journal , 31(5):930937.
IEEE std, 1988, IEEE standard on piezoelectricity, ANSI/IEEE Std 176-1987.
Jetteur, P., 1991, Validation des elements 28 et 29, Tech. Rep. 39, Samtech s.a., Li`ege, Belgium.
Finite Element Modelling of Piezoelectric Active Structures 123
Koppe, H., Gabbert, U. & Tzou, H. S., 1998, On three-dimensional layered shell elements
for the simulation of adaptive structures, in Proceedings of the Euromech 373 Colloquium:
Modeling and Control of Adaptive Mechanical Structures - Preprint, ADAMES, University of
Magdeburg.
Lardeur, P., 1990, Developpement et Evaluation de deux Nouveaux Elements Finis de Plaques
et Coques Composites avec Inuence du Cisaillement Transversal , Ph.D. thesis, Universite de
Compi`egne, Compi`egne, France.
Lee, C.-K., 1990, Theory of laminated piezoelectric plates for the design of distributed sen-
sors/actuators. Part I: Governing equations and reciprocal relationships, Journal of the
Acoustical Society of America, 87(3):11441158.
Lee, C.-K., Chiang, W.-W. & OSullivan, T. C., 1989, Piezoelectric modal sensors and actuators
achieving critical active damping on a cantilever plate, AIAA 89-1390-CP.
Lee, C.-K., Chiang, W.-W. & OSullivan, T. C., 1991, Piezoelectric modal sensor/actuator
pairs for critical active damping vibration control, Journal of Acoustical Society of America,
90(1):374384.
Lee, C.-K. & Moon, F. C., 1990, Modal sensors/actuators, Journal of Applied Mechanics,
57:434441.
Lee, H.-J. & Saravanos, D. A., 1996, Coupled layerwise analysis of thermopiezoelectric compos-
ite beams, AIAA Journal , 34(6):12311237.
Lerch, R., 1990, Simulation of piezoelectric devices by two- and three-dimensional nite ele-
ments, IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control , 37(3):233
247.
Liang, C., 1990, The Constitutive Modeling of Shape Memory Alloys, Ph.D. thesis, Virginia
Polytechnic Institute and State University, Blacksburg, Virginia.
Loix, N., 1998, Amortissement Actif de Structures Flexibles, Ph.D. thesis, Universite Libre de
Bruxelles, Brussels, Belgium.
Loix, N., Piefort, V. & Preumont, A., 1998, Modeling aspects of active structures with collo-
cated piezoelectric actuators and sensors, Benelux Quarterly Journal on Automatic Control
(Journal a), 39(1).
Loix, N. & Preumont, A., 1995, Remarks on the modelling of active structures with colocated
piezoelectric actuators and sensors, in ASME Design Engineering Technical Conference: Vi-
bration Control, Analysis, and Identication, vol. 3, pp. 335339, Boston.
Miller, S. E., Abramovich, H. & Oshman, Y., 1997, Selective modal transducers for anisotropic
rectangular plates: Experimental validation, AIAA Journal , 35(10):16211629.
Miller, S. E., Oshman, Y. & Abramovich, H., 1996a, Modal control of piezolaminated anisotropic
rectangular plates - Part I: Modal transducer theory, AIAA Journal , 34(9):18681875.
124 BIBLIOGRAPHY
Miller, S. E., Oshman, Y. & Abramovich, H., 1996b, Modal control of piezolaminated
anisotropic rectangular plates - Part II: Control theory, AIAA Journal , 34(9):18761884.
Miller, S. E., Oshman, Y. & Abramovich, H., 1998, Modal transducers for piezolaminated
anisotropic zero-gaussian curvature shell systems, AIAA 98-1953.
Mindlin, R. D., 1974, Equations of high frequency vibrations of thermopiezoelectric crystal
plates, International Journal of Solids Structure, 10:625637.
Moetakef, M. A., Joshi, S. P. & Lawrence, K. L., 1996, Elastic wave generation by piezoceramic
patches, AIAA Journal , 34(10):21102117.
Moetakef, M. A., Lawrence, K. L., Joshi, S. P. & Shiakolas, P. S., 1995, Closed-form expressions
for higher order electroelastic tetrahedral elements, AIAA Journal , 33(1):136142.
Moett, M. B., Clark, A. E., Wun Fogle, M., Linberg, J., Teter, J. P. & Mc Laughlin, E. A., 1991,
Characterization of Terfenol-D for magnetostrictive transducers, Journal of the Acoustical
Society of America, 89(3):14481455.
Moskalik, A. J. & Brei, D., 1997, Deection-voltage model and experimental results for poly-
meric piezoelectric c-block actuators, AIAA Journal , 35(9):15561558.
Nye, J. F., 1957, Physical Properties of Crystals: Their Representation by Tensors and Matrices,
Clarendon Press, Oxford.
Park, C. & Chopra, I., 1996, Modeling piezoceramic actuation of beams in torsion, AIAA
Journal , 34(12):25822589.
Perkins, J., editor, 1975, Shape Memory Eects in Alloys, Plenum Press, New-York.
Piefort, V., 1991, Applications des Alliages `a Memoire de Forme, Masters thesis, Universite
Libre de Bruxelles, Brussels, Belgium.
Piefort, V., 1997, Finite element modeling of piezoelectric structures - Final report, ULB -
Active Structures Laboratory; Convention 95/2792 - Region Wallonne.
Piefort, V., 2000, Modeling of the piezoelectric coupling in active structures - Final report,
ULB - Active Structures Laboratory; Convention 9713549 - Region Wallonne.
Piefort, V. & Henrioulle, K., 2000, Modelling of smart structures with colocated piezoelectric
Actuator/Sensor pairs: Inuence of the in-plane components, 5th International Conference
on Computational Structures Technology, Leuven, Belgium.
Piefort, V., Loix, N. & Preumont, A., 1998, Modeling of piezolaminated composite shells for
vibration control, European Conference on Spacecraft Structures, Materials and Mechanical
Testing, Braunschweig.
Piefort, V. & Preumont, A., 1998, Finite element modeling of piezolaminated composite shells,
Samtech Users Conference, Li`ege, Belgium.
Finite Element Modelling of Piezoelectric Active Structures 125
Piefort, V. & Preumont, A., 2000a, Finite element modelling of smart piezoelectric shell struc-
tures, 5th National Congress on Theoretical and Applied Mechanics, Louvain-La-Neuve, Bel-
gium.
Piefort, V. & Preumont, A., 2000b, Modeling of active shell structures with embedded piezoelec-
tric actuators and sensors, Materials Week, International Congress on Advanced Materials,
their Processes and Applications, M unchen, Germany.
Piefort, V. & Preumont, A., 2000c, Modeling of smart piezoelectric structures with nite ele-
ments, ISMA25, 25th International Conference on Noise and Vibration Engineering, Leuven,
Belgium.
Piefort, V. & Preumont, A., 2001, Finite element modeling of piezoelectric structures, Samtech
Users Conference, Paris, France.
Piezotech Documentation, 1997, Films piezo- et pyro-electriques, Document DT-96/a.
Pitoiset, X., 2001, Methodes Spectrales pour une Analyse en Fatigue des Structures Metalliques
sous Chargements Aleatoires Multiaxiaux, Ph.D. thesis, Universite Libre de Bruxelles, Brus-
sels, Belgium.
Pitoiset, X. & Preumont, A., 2000, Spectral methods for multiaxial random fatigue analysis of
metallic structures, International Journal of Fatigue, 22(7):541550.
Pitoiset, X., Preumont, A. & Kernilis, A., 1998, Tools for a multiaxial fatigue analysis of
structures submitted to random vibration, European Conference on Spacecraft Structures,
Materials and Mechanical Testing, Braunschweig.
Preumont, A., 1997, Vibration Control of Active Structures - An Introduction, Kluwer Academic
Publishers, Dordrecht, The Netherlands.
Preumont, A., Francois, A. & Dubru, S., 1999, Piezoelectric array sensing for real-time, broad-
band sound radiation measurement, Journal of Vibration and Acoustics, 121.
Rao, S. S. & Sunar, M., 1993, Analysis of distributed thermopiezoelectric sensors and actuators
in advanced intelligent structures, AIAA Journal , 31(7):12801286.
Reddy, J. N., 1984, Energy and Variational Methods in Applied Mechanics, John Wiley & sons.
Rex, J. & Elliott, S. J., 1992, The QWSIS - a new sensor for structural radiation control,
MOVIC-1, Yokohama.
Rogacheva, N. N., 1999, Analysis and optimization of energy conversion eciency for piezoelec-
tric transducers, in Smart Structures - Requirements and Potential Applications in Mechan-
ical and Civil Engineering (Holnicki-Szulc, J. & Rodellar, J., editors), pp. 295302, NATO
Sciences Series, Kluwer Academic Press, Dordrecht, The Netherlands.
Ryou, J.-K., Park, K.-Y. & Kim, S.-J., 1998, Electrode pattern design of piezoelectric sensors
and actuators using genetic algorithms, AIAA Journal , 36(2):227233.
Samanta, B., Ray, M. C. & Bhattacharyya, R., 1996, Finite element model for active control of
intelligent structures, AIAA Journal , 34(9):18851893.
126 BIBLIOGRAPHY
SAMTECH, 2000, SAMCEF Users Manual V8.1, Vol.3: Elements.
Saravanos, D. A., 1997, Mixed laminate theory and nite element for smart piezoelectric shell
structure, AIAA Journal , 35(8):13271333.
Saravanos, D. A., 2000, Passively damped laminated piezoelectric shell structures with inte-
grated electric networks, AIAA Journal , 38(7).
Suleman, A. & Goncalv`es, M. A., 1995, Optimization issues in application of piezoelectric
actuators in panel utter control, IDMEC-Instituto Superior Tecnico, Departamento de En-
genharia Mecanica, 1096 Codex, Portugal.
Suleman, A. & Venkayya, V. B., 1995a, Flutter control of an adaptive laminated composite panel
with piezoelectric layers, IDMEC-Instituto Superior Tecnico, Departamento de Engenharia
Mecanica, 1096 Codex, Portugal.
Suleman, A. & Venkayya, V. B., 1995b, A simple nite element formulation for a laminated
composite plate with piezoelectric layers, Journal of Intelligent Material Systems and Struc-
tures, 6:776782.
Sunar, M. & Rao, S. S., 1996, Distributed modeling and actuator location for piezoelectric
control systems, AIAA Journal , 34(10):22092211.
Sunar, M. & Rao, S. S., 1997, Thermopiezoelectric control design and actuator placement,
AIAA Journal , 35(3):534539.
Tiersten, H. F., 1967, Hamiltons principle for linear piezoelectric media, in Proceedings of the
IEEE, pp. 15231524.
Tzou, H. S. & Tseng, C. I., 1990, Distributed piezoelectric sensor/actuator design for dy-
namic measurement/control of distributed parameter systems: a piezoelectric nite element
approach, Journal of Sound and Vibration, 138(1):1734.
Tzou, H. S. & Tseng, C. I., 1991, Distributed modal identication and vibration control of
continua: Piezoelectric nite element formulation and analysis, Transactions of the ASME,
Journal of Dynamic Systems, Measurement and Control , 113:500505.
Tzou, H. S. & Ye, R., 1996, Analysis of piezoelastic structures with laminated piezoelectric
triangle shell elements, AIAA Journal , 34(1):110115.
Uchino, K., 2000, Ferroelectric Devices, Marcel Dekker inc., New-York.
van Randeraat, J. & Setterington, R., editors, 1974, Piezoelectric Ceramics, Mullard Ltd, Lon-
don, UK, 2nd edn.
Zhou, X., Chattopadhyay, A. & Haozhong, G., 2000, Dynamic responses of smart composites
using a coupled thermo-piezoelectric-mechanical model, AIAA Journal , 38(10).
Zienkiewicz, O. C., 1971, The Finite Element Method in Engineering Science, Mc Graw-Hill,
Maidenhead.