Home Search Collections Journals About Contact us My IOPscience

Electromechanical finite element modelling for dynamic analysis of a cantilevered

piezoelectric energy harvester with tip mass offset under base excitations

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2014 Smart Mater. Struct. 23 095037

(http://iopscience.iop.org/0964-1726/23/9/095037)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 134.7.248.130
This content was downloaded on 13/10/2014 at 07:31

Please note that terms and conditions apply.

 Smart Materials and Structures

Smart Mater. Struct. 23 (2014) 095037 (24pp) doi:10.1088/0964-1726/23/9/095037

Electromechanical nite element modelling

for dynamic analysis of a cantilevered
piezoelectric energy harvester with tip mass
offset under base excitations
M F Lumentut and I M Howard
Laboratory for Dynamic Systems of Smart Structure and Vibration, Theoretical and Applied Mechanics,
Department of Mechanical Engineering, Curtin University, Australia

E-mail: m.lumentut@exchange.curtin.edu.au

Received 25 April 2014

Accepted for publication 13 June 2014
Published 15 August 2014

Abstract
A new electromechanical nite element modelling of a vibration power harvester and its
validation with experimental studies are presented in this paper. The new contributions for
modelling the electromechanical nite element piezoelectric unimorph beam with tip mass offset
under base excitation encompass ve major solution techniques. These include the
electromechanical discretization, kinematic equations, coupled eld equations, Lagrangian
electromechanical dynamic equations and orthonormalized global matrix and scalar forms of
electromechanical nite element dynamic equations. Such techniques have not been rigorously
modelled previously by other researchers. There are also benets to presenting the numerical
techniques proposed in this paper. First, the proposed numerical techniques can be used for
applications in many different geometrical models, including micro-electro-mechanical system
power harvesting devices. Second, applying tip mass offset located after the end of the
piezoelectric beam length can result in a very practical design, which avoids direct contact with
piezoelectric material because of its brittle nature. Since the surfaces of actual piezoelectric
material are covered evenly with thin conducting electrodes for generating single voltage, we
introduce the new electromechanical discretization, consisting of the mechanical and electrical
discretized elements. Moreover, the reduced electromechanical nite element dynamic equations
can be further formulated to obtain the series form of new multimode electromechanical
frequency response functions of the displacement, velocity, voltage, current and power,
including optimal power harvesting. The normalized numerical strain node and eigenmode
shapes are also further formulated using numerical discretization. Finally, the parametric
numerical case studies of the piezoelectric unimorph beam under a resistive shunt circuit show
good agreement with the experimental studies.

Keywords: electromechanical discretisation, nite element, MEMS, power harvesting, vibration

(Some gures may appear in colour only in the online journal)

1. Introduction monitoring of intelligent infrastructure, aircraft and environ-

Recently, piezoelectric power harvesting devices have shown
their viability for converting mechanical energy into useful
electrical energy that can be used to power remote, smart
wireless sensor systems. Their applications include
mental monitoring systems and biosensors for the human
body. As a profound alternative energy source, a few pub-
lished literature reviews [17] have discussed how the
piezoelectric vibration-based power harvester applications
from macro-to-micro levels benet researchers who are

0964-1726/14/095037+24$33.00 1 2014 IOP Publishing Ltd Printed in the UK

Smart Mater. Struct. 23 (2014) 095037
investigating further novel techniques. Designing self-char-
ging and lifespan energy harvesting devices requires technical
knowledge and understanding of mechanical and electrical
tuning techniques, material or physical properties, mechanical
energy sources, power conditioning electronic circuits, sensor
systems, and miniature fabrication. However, before con-
ducting a manufacturing process for these smart devices in
such conditions, initial mathematical models using analytical
and numerical techniques become an important tool for ana-
lyzing the electromechanical piezoelectric vibration power
harvesting response behaviour. Consequently, the develop-
ment of new concepts has attracted attention from some
researchers. Starting with the simplied analytical lumped
parameter models of piezoelectric structures [810], the
electrical equivalent systems representing the electro-
mechanical piezoelectric component and standard circuit
interface were developed. Considering the required accuracy
and the necessity of modelling the cantilevered piezoelectric
power harvester with tip mass under base excitation, the
development of analytical approaches using RayleighRitz
methods were explored in an effort to analyze various para-
metric case studies for vibration power harvesters [1113].
Kim et al [14] investigated the vibrational piezoelectric
bimorph power harvester with the effects of tip mass offset
using the RayleighRitz method. Their interest and focus
were to model the system responses of short- and open-circuit
resonance frequency with variable load resistances using tip
mass conguration. Moreover, Erturk [15] further developed
comprehensive analytical approaches using assumed-mode
methods for EulerBernoulli, Rayleigh and Timoshenko
piezoelectric beams with axial deformations. It was proved
that by increasing a certain number of modes in the choice of
admissible trial functions, the approximation solution con-
verges to the analytical solution, especially for the funda-
mental resonance frequency response. Erturk and Inman [16]
have focused on researching the closed-form distributed
parameter system of the EulerBernoulli bimorph beam for
investigating frequency response functions (FRFs) under
variable load resistance. Wickenheiser and Garcia [17]
further developed the analytical modelling of piezoelectric
bimorph beams with magnetic tip mass located near the xed
ferromagnetic structures in order to magnify the driving fre-
quency below the fundamental frequency of the system.
Moreover, Lumentut and Howard [1820] explored the
development of the piezoelectric bimorph beam with tip mass
under both multidirectional excitation and the strain eld
effects of the transverse and initial longitudinal forms, in an
effort to investigate a broad range of case studies using
mathematical investigations of the orthonormalized electro-
mechanical weak-form technique and closed-form boundary
value method. In scrutinizing the system response of the
single piezoelectric beam in light of the existing literature, the
typical optimal power output mostly depends on the rst
mode. For this case, Lumentut et al [21] investigated the
strategy of increasing power output and widening the multi-
frequency band using the orthonormalized multi-
electromechanical dynamic equations reduced from closed-
form boundary value methods. Various parametric case
2
M F Lumentut and I M Howard
studies have been developed using new electromechanical
multimode FRFs of multiple piezoelectric bimorph beams
connected in series, parallel and mixed seriesparallel
connections.
Some nite element literature concerns the use of
piezoelectric components as smart material systems and has
shown a variety of engineering applications, the most notable
of which have been devoted to ultrasonic transducers and
active control systems, either with feedback gain control or
sensing and actuation response systems. However, for the
current research, these piezoelectric nite element models
give the relevancy and basis that allow the development of
new numerical techniques for modelling the electro-
mechanical power harvester, as presented in this paper. In the
ultrasonic transducer application, Allik and Hughes [22] and
Allik et al [23] rst introduced mathematical models of the
linear constitutive equations of the three-dimensional (3D)
nite element piezoelectric vibration model using variational
methods. Later, comprehensive mechanical discretized ele-
ments for analyzing an ultrasonic piezoelectric rod resonator
[24] and the phenomenological electro-acoustic modelling for
ultrasonic probes [25] were developed. Further application of
the piezoelectric nite element has been developed for the
active dynamic control system using feedback gain control.
Tzou and Tseng [26] developed the active dynamic control
matrix equations of the distributed piezoelectric components
onto plate structures using the Hamiltonian principle and the
Guyan reduction technique. Other published research works
that analyze the controlled transient response [2729] and
transfer function [30] behaviour have been further investi-
gated using the negative velocity feedback control system.
For sensing and actuation response systems, the actuation
system response, using applied voltage into the piezoelectric
actuator, activated the piezoelectric sensor to create voltage
output and deformation of the laminated composite structures
[3133]. More recently, Thomas et al [34] formulated the
shunted electrical circuit connected to the piezoelectric pat-
ches using the nite element control system of the beam
structure. Since the actual piezoelectric components are cov-
ered with electrode layers, the piezoelectric nite element
literature ignores the electrode effect by assuming that the
electrical voltages for each element and node are independent.
Only a few of the published papers related to control nite
element modelling have taken the electrode effect into con-
sideration by assuming equal voltage distribution on each
piezoelectric element [29, 31, 34]. However, the previous
works are only based on mechanical discretization, without
considering the electrical elements that are discretized due to
the effect of electrode layers. Ignoring the electrical dis-
cretized elements from the electrode layer can affect the
electromechanical FRF of the piezoelectric power harvester
using the closed-circuit system. Also, there are no other
previous research works that investigate the electro-
mechanical nite element frequency responses using the
electromechanical discretized elements under base excitation.
In earlier studies of the electromechanical nite element
response of piezoelectric power harvesting, Lumentut et al
[35] initially developed an energy harvesting plate structure
Smart Mater. Struct. 23 (2014) 095037
with the segmented piezoelectric element based on the
LoveKirchhoff plate theory for investigating frequency and
transient response behaviour. However, at an earlier stage,
this model only considers the voltage output from the
piezoelectric element without including the electrode effect
with complete derivation of the electromechanical dis-
cretization related to the system response. DeMarqui et al [36]
further presented the cantilevered plate with distributed pie-
zoelectric layers by including the electrode effect, as validated
by Erturk and Inman [16] in the existing literature of the
piezoelectric bimorph with tip mass. However, their proposed
nite element model does not take into account the important
physical issues of the electromechanical orthonormalized
global matrix and scalar forms, and the mathematical tech-
niques of electromechanical discretization. Aladwani et al
[37] further investigated the dynamic magnier using a spring
system attached to the base structure of the cantilevered
piezoelectric bimorph beam with tip mass. The system was
used to model the optimized power harvesting frequency
response behaviour. However, their proposed work neglects
the electrode effect of the piezoelectric component for for-
mulating the electromechanical discretized elements, and also
ignores the offset effect of the tip mass measured between the
tip mass centroid and the end of the bimorph. Further research
investigating piezoelectric power harvesters has been con-
ducted using solid nite element-based ANSYS multiphysics
software. Zhu et al [38] investigated the prediction of power
output of FRFs under the resistive shunt circuit. Moreover,
Yang and Tang [39] developed the electrical equivalent cir-
cuit model using SPICE software, whose circuit parameters
were adopted from the ANSYS simulation for predicting the
power output. However, although ANSYS software provides
capabilities of 3D, coupled eld solid, and plane elements, it
is not able to model the laminated beam elements. For the
standard level of the energy harvesters, mesh generation using
laminated beam elements is sufcient to model the laminated
piezoelectric beam structures (e.g., unimorph, bimorph and
multimorph).
In this paper, new numerical techniques of the electro-
mechanical nite element vibration power harvester using the
EulerBernoulli unimorph beam model are presented and vali-
dated with experimental studies. The numerical models proposed
here provide new technical contributions for analyzing the
interrelationship between coupled eld equations, electro-
mechanical discretization, kinematic motions, Lagrangian elec-
tromechanical dynamic equations and electromechanical nite
element equations. These technical contributions provide clear
insight into the new electromechanical nite element equations
for power harvesting applications, which can be highlighted in
the following points:
In considering the effect of electrode layers, we introduce
the new electromechanical discretization, consisting of
mechanical and electrical discretized elements, to model
the electromechanical nite element piezoelectric struc-
ture. The electromechanical discretized elements provide
an important basis for formulating the FRFs using new
electromechanical nite element dynamic equations.
3
M F Lumentut and I M Howard
The elemental system dynamics of the piezoelectric
unimorph beam, including tip mass offset, are reduced
from the kinematic equations. It is found that the effect of
the offset distance, measured from the tip mass centroid
and the end of the unimorph beam, can result in extra
terms in the mass matrix and the generalized input
dynamic force vector. The benet of applying the tip mass
offset, located after the end of the piezoelectric length, is
that it can provide a very practical MEMS power harvester
design for any level of scalability. It can also avoid the use
of material around the end of the piezoelectric length,
which can readily be damaged because of its brittle nature.
The established, new, electromechanical nite element
approach can be used conveniently to model different levels
of scalability, including the segmented piezoelectric element
coupled with substructure and micro-electro-mechanical
system (MEMS) devices.
The proposed numerical techniques also provide a
complement to study the laminated piezoelectric beam
with electrode layers in different applications.
Lagrangian electromechanical dynamic equations are
developed to formulate the constitutive elemental matrices
of the electromechanical nite element equations. The
orthonormalized global matrices and global scalar forms
can be further reduced to formulate new multimode
electromechanical FRFs of the displacement, velocity,
voltage, current and power harvesting, including the
optimal power harvester. The normalized strain node and
eigenmode shapes are formulated using discretized
elements for convergence studies. Finally, the parametric
case studies with resistive shunt circuits, based on the
proposed novel electromechanical nite element and
experimental validations, are also presented and discussed.
2. Coupled eld equations of the piezoelectric
energy harvester
Constitutive linear piezoelectric beam equations can be for-
mulated using tensor electrical enthalpy concepts, which are
based on the continuum thermodynamics that can be con-
densed using Voigts notation and Einsteins summation
convention [19, 4042]. The electrical enthalpy of typical
smart materials with lead zirconate titanate (PZT) can be
formulated under adiabatic and isothermal processes as
1 1 S
H ( S1, E3 ) = c11E S1 S1 e31 E3 S1 33 E3 E3 . (1)
2 2
Notations of the piezoelectric parameters used here are sta-
ted in accordance with IEEE standards [43]. Here, the parameters
S
c11E , e31, 33 , E3, T1 and S1 represent the piezoelectric elastic
stiffness at constant electric eld, piezoelectric coefcient, per-
mittivity under constant strain, electric eld, stress and strain,
respectively. The parameter S
33 T
= 33 e31 d 31 or
S T 2 E T
33 = 33 d 31 c11 is given where c11 = 1/s11E , 33
E
is the per-
mittivity at constant stress and s11E is the elastic compliance at
constant electric eld. Moreover, the common piezoelectric
Smart Mater. Struct. 23 (2014) 095037 M F Lumentut and I M Howard

Figure 1. Kinematic motions of the beam and arbitrary tip mass offset.

constant produced from the manufacturing company is in the 3. Kinematic equations for elemental beam and
form d 31, where this can be modied by multiplying the plane arbitrary tip mass offset
stress-based elastic stiffness at constant electric eld to give
e31 = d 31 c11E . The electric displacement of the piezoelectric In this section, the kinematic motions for a continuous geome-
element, D3, can be obtained by differentiating equation (1) with trical beam, as shown in gure 1, are derived from the different
respect to E3. This electric displacement of piezoelectricity can be
congurations at a series of points, resulting in the deformations
formulated as and velocity of the body. However, the kinematics for the tip
mass offset, whose motions are solely affected by the tip of
H S
the beams transverse rectilinear [w abs
tip
(L, t ) ] and angular
D3 = = e31 S1 + 33 E3 . (2) [ (L, t ) ] velocities, is a rigid-body motion with different
E3
congurations that result in extra terms in the kinematic
Note that for the purpose of convenience in the forth- equations. In reviewing previous studies, the most recently
coming notations, the superscripts 1 and 2 that are used in published papers on MEMS power harvesters [4550] included
the parameters of stress, strain, elastic stiffness, density and the tip mass offset without scrutinizing its effect on the kine-
cross sectional area refer to the substructure and piezo- matic equations. The effect of the offset distance between the tip
electric layers, respectively. Since the piezoelectric mass centroid and the end of the beam can effectively contribute
unimorph consists of active and inactive layers, the plane to the mass matrix and the generalized input dynamic forces, as
stress eld from the inactive or substructure layer can also shown in section 4.3. It is important to note that the aim of
formulating the kinematic equations is to formulate the
be formulated as
Lagrangian functional energy forms that consist of the kinetic,
(1) (1) (1) potential and electrical energies of the piezoelectric unimorph,
T1 = c11 S1 . (3)
with the tip mass as given in the forthcoming section.
In gure 1, the undeformed beam under base vector wbase (t )
The piezoelectric stress eld can also be formulated by
moves from the xed reference frame of oXZ to the initial
differentiating equation (1) with respect to S1, to give
reference frame of oX Z . Since the base vector, as the input
excitation, moves from point o to o at the designated reference
H
T1(2) = =
c (2) (2)
S
11 1 e E
31 3 . (4) frame, the position of point p also moves to point p with the
S1(2) same magnitude. As a result, the base vector creates the relative
transverse deformation, w (x, t ), measured from point p moving
Note that alternative derivations of equations (2) and (4) to the nal point, p. The absolute displacement, wabs (x, t ), can
can also be obtained by using the enthalpy equation of state in also be obtained from the xed reference frame of oXZ to the
terms of continuum thermopiezoelectricity, Maxwells rela- nal position. Note that the difference between the absolute
tions and Legendres transformation [41, 44]. displacement and the base vector denes the relative

4
Smart Mater. Struct. 23 (2014) 095037 M F Lumentut and I M Howard

deformation. Moreover, the beam also carries an arbitrary tip Again, the linear system for the small angle,
mass offset whose kinematic motions must be considered. The (L, t ) = w (L, t )/x , can be made using Taylors theorem
position vectors of the tip mass start at point d to the nal dif- to give sin (L, t ) (L, t ) and cos (L, t ) 1. In this
ferential element at point m, where the resultant vectors at points case, the velocity of the elemental tip mass can be obtained
od gm can be dened. The point of attachment, d by differentiating equation (9) with respect to time, to
(deformation point at the end of the beam), to the tip mass yield
centroid at point g can be viewed as the offset of the tip mass.
The origin of tip mass centroid can be determined from the mm (L, z , t ) = R
od + R dg + R gm R od R
dg R
gm
R
differential element, using the extra position vector from points
g to m. As mentioned previously, since the tip mass was base (t ) + w
= (w (L, t ) ) e3
assumed to be a rigid body, its relative motions depend on the tip
(L, t ) e2 ( z c e3 + xc e1 )
of the beams motion at point d. (L, t ) e2 ( z m e3 + xm e1 ). (11)
The position vector, R pp , for the elemental beam can be
dened as The position vector, R pp, can be specied as the relative
pp op
R (x , z , t ) = R R , op
(5) displacement with respect to the moving base support from
reference frame oXZ to oX Z , as
where the position vectors, Rop and Rop , can be written as
R op = (x z sin (x , t ) ) e1 R p p (x , z , t ) = R pp R pp
+ ( wabs (x , t ) + z cos (x , t ) ) e3 , = z(x , t ) e1 + w (x , t ) e3 . (12)
R op = xe1 + ze3 . (6)
The elemental beam strain can be obtained by differ-
In simplication of the linear assumption, the small angle, entiating R pp with respect to x, giving the typical Euler
(x, t ) = w(x, t )/x , can be obtained after applying Taylors Bernoulli beam theory as
theorem to give sin (x, t ) (x, t ) and cos (x, t ) 1. As
the initial position vector of the elemental beam, Rop , is
(R p p e1 ) 2w ( x , t )
dened, the position vector, R pp , from equation (5) can be S1 = =z . (13)
differentiated with respect to time, which gives x x 2
pp (x , z , t ) = R
R op R op Since the piezoelectric unimorph beam is viewed as the
= (x , t ) e2 ze3 asymmetric structure with different material properties and
geometrical dimensions, the asymmetric neutral axis [51]
base (t ) + w
+( w (x , t ) ) e3 . (7)
must be determined as presented in appendix A.
Note that the absolute velocity vectors for the elemental
beam and tip mass can be written as
wabs (x , t ) = ( w 4. Electromechanical nite element vibration
base (t ) + w
(x , t ) ) e3
modelling of the piezoelectric energy harvester
abs
w tip
base (t ) + w
(L , t ) = ( w (L, t ) ) e3 . (8)
In common practice, the surfaces of actual or physical
For the elemental tip mass, the position vector, R mm, can
piezoelectric material currently on the market are covered
be formulated as
evenly with thin conducting electrodes, resulting in single
R mm (L, z , t ) = R od + R dg + R gm R od R dg R gm , (9) voltage output during dynamic motions. As shown in
gure 2(a), we chose the physical structure of the
where each position vector can be formulated as unimorph beam with an arbitrary tip mass offset, which
R od = Le1 , consists of a piezoelectric material with a conducting elec-
trode layer and a substructure. However, in gure 2(b), the
R dg = xc e1 + z c e3 ,
electromechanical discretization, consisting of mechanical
R gm = xm e1 + z m e3 , and electrical discretized nite elements, is completely dif-
R od = Le1 + ( wbase (t ) + w (L, t ) ) e3 ferent from the physical structure. In this situation, the elec-
tromechanical nite element discretization of the piezoelectric
R dg = ( xc cos (L, t ) z c sin (L, t ) ) e1 structure should be analyzed using the generalized multi-
+ ( z c cos (L, t ) + xc sin (L, t ) ) e3 output electrical current from each element connected with an
electrical parallel connection. Since the polarization is pro-
R gm = ( xm cos (L, t ) z m sin (L, t ) ) e1 portional to the stress eld and the stress eld is also pro-
portional to the strain eld [40, 44, 52], the polarization, P,
+ ( z m cos (L, t ) + xm sin (L, t ) ) e3 . (10)
and electric eld, E, directions are chosen in the three-
direction (thickness) along the z-axis.

5
Smart Mater. Struct. 23 (2014) 095037 M F Lumentut and I M Howard

Figure 2. Electromechanical model of the unimorph beam with arbitrary tip mass offset under input base excitation, based on (a) a physical
system connected electrically with the load resistance and (b) a new electromechanical nite element discretization.

Figure 3. First-order Hermite polynomials for a local unimorph element with an arbitrary tip mass at arbitrary nodes n1 and n. Based on
gure 2, by taking the example for n = 2.

4.1. Geometrical analysis of an electromechanical element The rst-order Hermite interpolation in terms of the
polynomial, (1) (x ), and the unknown nodal displacement,
The rst-order Hermite interpolation of the cubic displace-
u i(t ), for the elemental beam can be formulated as
ment function is chosen here to formulate the unknown four
2 1
degrees of freedom for the two-node unimorph beam element
[5355]. The cubic displacement function must satisfy the
w (x , t ) = ki(1) (x ) ui(k )(t ). (14)
i=1 k=0
beam theory continuity conditions for the translation and
rotational displacements for the general location of the ele- The property of Hermite polynomials for modelling the
ments. It is important to note here that the kinematic translation and rotation at the nodes can be formulated in
terms of the l-derivatives to give
equations given in section 3 can be used to formulate the local
element near the tip mass, as shown in gure 3, since the dl ki(1)
translation and angular velocities of the tip masss rigid body ( x n) = in kl for i , n = 1, 2 and , (15)
dx l
depend on the end of the unimorph beam of length, giving
k , l = 0,1
particular insight to formulating the complete electro-
mechanical nite element equations. where xn is the location at the nth node and pq is the

6
Smart Mater. Struct. 23 (2014) 095037 M F Lumentut and I M Howard

Kronecker delta with the property function, , for each node as

0 if p q u1(t ) = u1(0) (t ) = w1(t ),
pq = . (16)
1 if p = q u 2 (t ) = u1(1) (t ) = 1(t ),
Corresponding to equations (15) and (16), the lth deri- u 3(t ) = u2(0) (t ) = w2(t ),
vatives of w(x, t) can be written as u 4(t ) = u2(1) (t ) = 2(t )
2 1 (1)
dl wn dl ki(1) 1(x ) = 01 (x ),
(x , t ) = ( x n ) ui (k ) (t ), (17) (1)
dx l i=1 k=0
dx l 2(x ) = 11 (x ),
(1)
3(x ) = 02 (x ),
or they can be formulated in terms of the unknown nodal
(1)
displacement forms as 4(x ) = 12 (x ). (23)

dl wn 2 1 Corresponding to equation (13), the straindisplacement

(t ) = in kl ui (k ) (t ) = u n(l) (t ).
dx l i=1 k=0
Note that equation (18) clearly shows certain physical
meaning after applying equations (15) and (16), giving the
nodal displacement forms as a function of time. Corre-
sponding to index notation from equation (15), equation (14)
can be modied by applying equation (18) to give
(18) relationship, in terms of the differential form of the shape
function of the piezoelectric unimorph beam, can be for-
mulated by substituting equation (21) to give
S(x , t ) = z e(x ) ue (t ), (24)
where the differential form of the shape function of the strain
displacement relationship can be formulated as

2 1 d2 e (x )
dk wi e (x ) =
w (x , t ) = ki(1) (x ) (t ). (19) dx 2
dx k
i=1 k=0
= 1 (x ) 2 (x ) 3 (x ) 4 (x ) ,
(25)
To nd four terms of rst-order Hermite polynomials, the where
property given from equation (15) can be used to identify four
elemental boundary conditions for each degree of freedom at 6 12 ( x xe )
1 (x ) = 2
+ ,
the nodes. In such conditions, the cubic equations with four Le L e3
unknown coefcients for each degree of freedom at the nodes 6 ( x xe ) 4
can be used, and the results of the Hermite polynomials can 2 (x ) = 2

Le L e
be formulated after simplifying to give
12 ( x xe ) 6
3 (x ) = + 2,
x x e 2 x x e 3 3
Le Le
(1)
01 (x ) = 13 + 2 ,
Le Le 6 ( x xe ) 2
4 (x ) = . (26)
x xe
2 L e2 Le
(1)
11 (x ) = ( x x e ) 1
Le The elemental discretized electric eld, E, induced by the
x xe x xe
2 3 strain eld due to input ambient vibration, can create a
(1)
02 (x ) = 3 2 , polarization in the piezoelectric material in the z-direction
Le Le along its thickness, generating the electrical voltages. The
electrical eld is a function of the electrical potential, with a
x xe
2
x xe
12 (x ) = ( x xe )
(1)
negative gradient operator given by
L e L e
E3 = e (z , t ), (27)
L e = x e+ 1 xe. (20)
where e (z, t ) = e(z ) ve(t ) is the electrical potential with
Corresponding to equation (14), the simplied rst-order linear assumption, the parameter e(z ) = (z z n + h p )/h p is
Hermite interpolation function of the unimorph beam can be the shape function over the interval z n h p z z n , and zn
modied into a matrix form to give indicates the distance from the asymmetric neutral axis, as
w (x , t ) = e (x ) ue (t ), (21) given in appendix A. The symbol is a gradient operator for
the rst derivative of the shape function with respect to the
where thickness direction, giving (e) (z ) = de(z )/dz = 1/h p . In
this case, the discretized electric eld can be reformulated as
e (x ) = [ 1 (x ) 2 (x ) 3 (x ) 4 (x ) ]
E3 = e (z) v(e) (t ). (28)
ue (t ) = [ u1 u 2 u 3 u 4 ]T . (22)
The expression of the stress eld in partial differential
Each parameter from equation (22) can be written in shape function form can be formulated for both piezoelectric
terms of the elemental displacement vector, u, and shape and substructure elements by substituting equations (24) and

7
Smart Mater. Struct. 23 (2014) 095037 M F Lumentut and I M Howard

(28) into equations (3) and (4), to give The electrical energy term for the piezoelectric layer can
be reduced as
T1(2) = zc11
(2)
e (x ) ue (t ) + e31 e (z) ve (t )
1 x e +1
T1(1) = zc11
(1)
e (x ) ue (t ). (29) WE =
2
x e

A (2)
E3T D3 dA (2) dx . (34)

The electric displacement vector of the piezoelectric The nonconservative work on the system, due to the
component can be formulated by substituting equations (24) input base excitation and electrical charge output, can be
and (28) into equation (2) to give written as
S x e +1
D3 = ze31 e (x ) ue (t ) 33 e (z) ve (t ). (30)
WF = x e
A (1)
(1) w (x , t ) T dA (1) dx w base (t )
x e +1
4.2. Lagrangian electromechanical finite element equation x e
A (2)
(2) w (x , t ) T dA (2) dx w base (t )

Constitutive electromechanical nite element dynamic I0tip xc ( x e + 1, t )T wbase (t )

equations are developed for modelling the elemental piezo- I0tip w ( x e + 1, t )T wbase (t ) + q (e) (t ) v(e) (t ). (35)
electric unimorph with the tip mass offset under input base
transverse excitation. The functional energy forms using the The reduced form of equation (35) due to the input base
extended Lagrangian formulation consist of kinetic (KE), excitation can be found in appendix C. The expressions of the
potential (PE) and electrical (WE) energies, and non- functional energy forms, including external work, can be
conservative work (WF). The extended Lagrangian equation modied in terms of the discretized mechanical and electrical
can be written as shape functions. The extended kinetic energy form can be
formulated by substituting equation (21) into equation (32),
d
giving
Fs = 0,

(31)
s= 1
dt s s
1 x e +1 d e (x ) T
KE =
2
x A (1)
z 2 (1) ue (t ) T
dx
where = KE PE + WE , s {u (t ), u (t ), v (t ) } and e

(t ), q (t ) } or {s = 1Fs } = { uWF
Fs {wbase WF
e (t ) + ve (t ) }.
d (x ) e e
u (t )dA(1) dx
The kinetic energy can be formulated based on the mass dx
density of the piezoelectric unimorph and the density of tip 1 x e +1 d e (x ) T
mass, as +
2 xe
A (2)

z 2 (2) ue (t ) T
dx
e
1 x e +1 d (x ) e
KE = x z 2 (1) (x , t ) T (x , t ) dA (1) dx u (t )dA(2) dx
2 e A (1) dx
1 x e +1
+
1
x
x e +1
z 2 (2) (x , t ) T (x , t ) dA (2) dx +
2 xe
A (1)

(1) ue (t ) T e (x ) T (e) (x )u (e) (t )dA (1) dx
2 e A (2)
1 x e +1
+
1
x
x e +1
(x , t ) T w
(1) w (x , t ) dA (1) dx +
2 xe
A (2)

(2) ue (t ) T e (x ) T e (x )ue (t )dA (2) dx
2 e A (1)

1 x e +1 d e ( x e + 1 )
+ I0tip xcu (t )T e ( x e + 1 )T u (t )
+
2
x e
A (2)
(x , t ) T w
(2) w (x , t ) dA (2) dx dx
1
( x e + 1, t )T ( x e + 1, t )
+ I0tip xc w + I0tipue (t )T e ( x e + 1 )T e ( x e + 1 ) ue (t )
2
1
+ I0tip w ( x e + 1, t )T w ( x e + 1, t ) 1 d e ( x e + 1 ) T e T d e ( x e + 1 ) e
2 + I2tip u (t ) u (t ). (36)
1 2 dx dx
+ I2tip ( x e + 1, t )T ( x e + 1, t ). (32)
2 Corresponding to equations (24) and (29), the extended
potential energy form can be reduced as
The derivation of equation (32) can be seen in
appendix B. The potential energy for the piezoelectric 1 x e +1

unimorph can be formulated as

PE =
2
x e
A (1)
z 2ue (t ) T e (x ) T c11(1) e (x ) ue (t )dA (1) dx

1 x e +1
1 x e +1
(1) T
+ x z 2ue (t ) T e (x ) T c11(2) e (x ) ue (t )dA (2) dx
PE =
2
x e
A (1)
S1 T1(1) dA (1) dx 2 e
x e +1
A (2)

1
1 x e +1 T
x ze31 ue (t ) T e (x ) T e (z ) ve (t )dA (2) dx . (37)
+
2
x e
A (2)
S1(2) T1(2) dA (2) dx . (33) 2 e A (2)

The extended electrical energy expression can be

reduced by substituting equations (28) and (30) into

8
Smart Mater. Struct. 23 (2014) 095037 M F Lumentut and I M Howard

equation (34), as d e (L e ) T T
I0tip xc wbase (t ) I0tip e ( L e ) wbase (t ). (40)
dx
1 x e +1
WE =
2 xe

A (2)

ze31 ve (t ) T e (z) T e (x ) ue (t )dA (2) dx
The second electromechanical dynamic equation due to
1 x e +1 the electrical form can be expressed as
+
2 xe
A (2)
S e
33
v (t ) T e (z) T e (z) ve (t )dA (2) dx . (38)
1 x e +1

The extended virtual work, in terms of the expression of

2
x e
A (2)
ze31 e (z ) T e (x ) ue (t ) dA (2) dx
nonconservative energy, can be formulated by substituting 1 x e +1
equation (21) into (35) to give
2
x e
A (2)
ze31 e (z ) T e (x ) ue (t ) dA (2) dx

x e +1 x e +1
WF = x (1) e (x ) T ue (t ) dA (1) dxwbase (t ) x e
A (2)
S
33 (z ) T (z ) v (t ) dA (2) dx q e (t ) = 0. (41)
e A (1)
x e +1
x e
A (2)
(2) e (x ) T ue (t ) dA (2) dxwbase (t ) Equation (41) can be modied by differentiating with
respect to time, to give
d e ( x e + 1 ) T e
I0tip xc u (t ) wbase (t ) x e +1
dx 1

2
x A (2)
z (z ) T e31 (x ) u (t ) dA (2) dx
I0tip e ( x e + 1 ) ue (t ) wbase (t ) + q e (t ) ve (t ). (39) e

1 x e +1

The expressions given from equations (36)(39) can be

2
x e
A (2)
z (z ) T e31 (x ) u (t ) dA (2) dx
substituted into equation (31) to give two electromechanical x e +1

dynamic equations. After simplifying, the rst electro-

x e
A (2)
S
33 (z ) T (z )v (t ) dA (2) dx i pe (t ) = 0. (42)
mechanical dynamic equation due to the transverse bending
form can be expressed as
x e +1 d e (x ) T d e (x ) e
x A (1)
z2
dx dx
u (t ) dA (1) dx 4.3. Element matrices of electromechanical dynamic
e equations
x e +1 d e (x ) T d e (x ) e
+ x e
A (2)
z2
dx dx
u (t ) dA (2) dx Constitutive electromechanical nonhomogeneous differential
dynamic equations, in terms of equations (40) and (42), can
x e +1
+ x A (1)
(1) e (x ) T e (x ) ue (t ) dA (1) dx be arranged into matrix form by including Rayleigh damping
e to give
x e +1
+ x e
A (2)
(2) e (x ) T (e) (x ) u (e) (t ) dA (2) dx
M e 0 ue (t ) C e 0 ue (t )
+
d ( x e + 1 ) e
0 0 ve (t ) PeT PDe ve (t )
+ 2I0tip xc e ( x e + 1 )T u (t )
dx K e Pe ue (t ) F e
+ I0tip e ( x e + 1 )T e ( x e + 1 ) ue (t ) + = e , (43)
0 0 v e (t ) i p
d e ( x e + 1 ) T d e ( x e + 1 ) e
+ I2tip u (t )
dx dx where
x e +1
+ x e
A (1)
z 2 c11(1) e (x ) e (x ) T ue (t ) dA (1) dx
x e +1 d e (x ) T d e (x )
x e +1 Me = x z 2 (nl ) dA (nl ) dx
+ x z 2 c11(2) e (x ) e (x ) T ue (t ) dA (2) dx nl = 1 e A (nl) dx dx
A (2)
e
x e +1

1
x
x e +1
e T e e
ze31 (x ) (z) v (t ) dA dx (2)
+

x e
A (nl)
(nl ) e (x ) T e (x ) dA (nl ) dx

nl = 1
2 e A (2)

1 x e +1 d e ( x e + 1 )
+ 2I0tip xc e ( x e + 1 )T

2
x e
A (2)
e T e e
ze31 (x ) (z) v (t ) dA dx (2)
dx
x e +1 + I0tip e ( x e + 1 )T e ( x e + 1)
= x A (1)
(1) e (x ) T dA (1) dxwbase (t )
d e ( x e + 1 ) T d e ( x e + 1 )
+ I2tip
e
x e +1
dx dx
x e
A (2)
(2) e (x ) T dA (2) dxwbase (t )

9
Smart Mater. Struct. 23 (2014) 095037 M F Lumentut and I M Howard

x e +1 corresponding to global displacement vectors at nodes (nodal

Ke = x A (nl)
z 2 c11(nl ) e (x ) T e (x ) dA (nl ) dx ,
degrees of freedom) and global electrical voltage vectors at
nl = 1 e
elements (electrical degrees of freedom), respectively. For
C e = M e + K e
assembling the global matrices, the connectivity matrices
x e +1 whose discretized elements share the same node should meet
Pe =
xe
A (np)
ze31(np) e (x ) T e (z) dA (np) dx ,
the requirements of compatibility. As the unimorph beam is a
np = 1

x e +1 cantilevered structure under input base excitation, the xed

T
Pe =

x A (np)
ze31(np) e (z) T e (x ) dA (np) dx

boundary condition at the base structure must be employed in
np = 1 e
the global matrices. For this case, the derivations for the
x e +1 normalized electromechanical dynamic equations and FRFs
PDe =

x e
A (np)
33(np) e (z) T e (z) dA (np) dx ,
can be solved and given in the next stage. The use of the
np = 1
MATLAB computer program not only provides the best
e e
F = Q wbase (t ) method for solving this case, but also gives an efcient
x e +1 computation technique.
Q (e ) =
xe
A (nl)
(nl ) e (x ) T dA (nl ) dx
The solution form of time-dependent displacement
nl = 1
can be stated in terms of the normalized modal vector
T
d e ( x e + 1 ) and the time-dependent displacement generalized coordi-
+ I0tip xc + I0tip e ( x e + 1 )T . (44)
dx nate as

It is important to note here that equation (43) consists of u (t ) = 1 a1 (t ) + 2 a 2 (t ) + ...

mass, stiffness, electromechanical coupling and piezoelectric + m 1 a m 1 (t ) + m am (t ) = a (t ), (46)
capacitance matrices and dynamic force vectors. These
where the normalized modal matrix can be formulated as,
parameters are modelled at the local electromechanical ele-
ment, which is located between the end of the unimorph beam 12

length and the tip mass. The contribution of the offset tip
(
= U U T MU ) . (47)

mass can be seen in the mass matrices and dynamic force It should be noted that parameters and U represent a
vectors, which most of the existing literature has ignored. As set of the normalized modal matrix and generalized eigen-
the unimorph-based Euler-Bernoulli beam assumption is vector, respectively. To obtain the eigenvector and the
considered, the rst term from the mass matrix (rotary eigenvalue solution, the undamped mechanical dynamic
inertia of the unimorph) can be discarded due to the con- equation from the stiffness and mass matrices should be
tribution of the Rayleigh beam assumption. Also, the con- nonsingular, since the eigenvalue equation,
tribution of input base excitation to the cantilever (K 2M U = 0 , can be used to nd the nontrivial
)
unimorph beam mathematically affects the distributed input eigenvector. Note that the normalized modal matrix is the
dynamic forces into all the elements. Note that when other matrix whose columns consist of the normalized eigenvec-
elements of the unimorph beam are considered in the tors [56]. The elemental strain eld of the unimorph beam
numerical simulation, the parameter for the tip mass should be given in equation (24) can be formulated in terms of
discarded from the mass matrices and input dynamic force equations (24) and (46) to give
vectors. S (x , t ) = z e (x ) ue (t ) = z e(x ) e a(t )
(14) (4 r ) (r 1)

= ze (x )a(t ). (48)
(1 r ) (r 1)
4.4. Normalized global element matrices of electromechanical
dynamic equations It should be noted that the elemental strain node para-
meter, e (x ), based on the normalized eigenvector, can be
Constitutive electromechanical nite element dynamic determined at certain elements and nodes of the system over
equations in global matrix form reduced from the interval, xe x x e + 1. Note that collective data points
the global transformation of local elements can be formulated for strain nodes along the unimorph beam can only be
as identied for each discretized element using the differential
straindisplacement function, e (x ), and elemental nor-
M u(t ) + C u(t ) + K u(t ) + P v(t )
malized eigenvector, . The normalized eigenvector at
( n m n m )( n m 1) ( n m n m )( n m 1) ( n m n m )( n m 1) ( n m n e )( n e 1)
certain elements can only be determined by separating the
= F PT u(t ) + P v(t ) = i p(t ).
D (45) cell vector array from the normalized modal matrix in terms
( n m 1)( n e n m )( n m 1) ( n e n e )( n e 1) ( n e 1) of the elemental degree of freedom and the multimode
system.
Note that the script terms nm and ne showing at each Equation (47) can be used to diagonalize the mass,
parameter in equation (45) indicate the global matrices stiffness and damping matrices to simplify the computation

10
Smart Mater. Struct. 23 (2014) 095037 M F Lumentut and I M Howard

technique. Substituting equation (46) into equation (45) and the coupled equation for the global coordinate system
premultiplying the result by T gives should be based on the electromechanical discretization, as
shown in gure 2(b). The internal parallel connection, in
T Ma (t ) + T Ca (t ) + T Ka (t ) + T P v (t ) = T F ,
terms of the Kirchhoff's voltage law and Kirchhoff's current
PT a (t ) + PDv (t ) = i p (t ), (49) law, can be used to formulate the electrical discretized
elements as
or simplifying equation (49) becomes
v1 (t ) = v2 (t ) = .... = vs (t ) = v (t )
a (t ) + 2a (t ) + 2 a (t ) + P v (t ) = Q wbase (t )
T
i P1 (t ) + i P2 (t ) + .... + i Ps (t ) = i P (t ). (54)
P a (t ) + PD v (t ) = i p (t ). (50)
For voltage output, the external load resistance can
Parameter Qw (t ) from the rst line of equation (50)
base
be used as an external circuit to connect all electrical
indicates the nodal dynamic forces of the discretized discretized piezoelectric elements in a parallel connection,
unimorph harvester model due to the input base excitation.
as
Other normalized parameters from equation (50) can be
stated as v (t ) = i P (t ) Rload . (55)
T T 2
M = I , K = ,
Multimode FRFs of the distributed piezoelectric
( ) (
T C = T M + T K = I + 2 = 2 ) unimorph can be formulated in several ways. First, one can
P = T P ,
T
P = PT , Q = T Q . (51) modify the rst term of equation (53), which algebraically
corresponds to the number of the normalized piezoelectric
It should be noted that the rst two lines of equation (51) elements. Second, one can employ equations (54) and (55)
represent the orthornormality property of mechanical dynamic in the equations obtained from the rst step. Third, one can
equations, the results of which indicate diagonal matrices. apply equation (54) in equation (52). Fourth, the results
obtained from the second and third steps can be algebrai-
4.5. Normalized global scalar form of the electromechanical cally solved using Laplace transforms, giving the result in
dynamic equations matrix form. The superposition of equations for voltage
multimode FRFs can be formulated after simplifying, as
In simplication, the normalized global element matrices of
electromechanical transverse equations from equation (50) v (j)
can be reduced into scalar forms in order to formulate the
2 wbase e jt
multimode frequency analysis of the distributed piezoelectric
NELP
component, as shown in gure 2(a). The electromechanical
Q r j Psr
equations consist of the coupled eld normalized differential NDOF
s= 1
equations for each elemental global coordinate component. In
r2 2 + j 2r r
this case, the rst form of the discretized electromechanical = r= 1
NELP NELP
. (56)
piezoelectric dynamic equation can be formulated for the j Psr Prs
NELP NDOF
multidegree of freedom (multimode) system, 1 s= 1 s= 1
r = 1, 2, 3, , NDOF , in terms of the number of normalized
jPDs 2 2
s= 1
Rload r = 1 r + j 2r r
piezoelectric elements, s = 1, 2, 3, , NELP , as
NELP The multimode FRF of the electric current output
r (t ) + 2r r a r (t) + r2 ar (t ) + Prs vs (t ) related to the input base transverse acceleration can be
s= 1 stated as
= Q r wbase (t ),
r = 1, 2, , NDOF . (52) i p (j)
2 wbase e jt
The second form of the discretized electromechanical NELP
piezoelectric dynamic equation can be formulated as Q r j Psr
NDOF
1 s= 1
NDOF
Psr ar (t ) + PDs vs (t ) = i Ps (t ), Rload r= 1 r2 2 + j 2r r
= NELP NELP
(57)
r= 1
s = 1, 2, 3, , NELP . (53) NELP NDOF
j Psr Prs
1 s= 1 s= 1
jPDs 2 2
.
Note that since the piezoelectric surface with the elec- s= 1
Rload r = 1 r + j 2r r
trode layers is distributed evenly on the substructure layer,

11
Smart Mater. Struct. 23 (2014) 095037 M F Lumentut and I M Howard

The power harvesting multimode FRF, related to the reformulated in terms of the natural normal mode multiplied
input transverse acceleration, can be formulated as by the time-dependent transverse displacement, using
equations (21) and (46) to give
Pp (j)
2 w (x , t ) = e(x ) e a(t )
(w 2
base e j t ) (14) (4 r ) (r 1)

NELP

2 = e (x ) a (t )
Q r j Psr (1 r ) (r 1)

1
NDOF
s= 1

NDOF

2 2

= er (x ) ar (t ). (62)

Rload r= 1 r + j 2r r r=1



=
. (58) Moreover, the characteristic motions for tip mass offset
NELP NELP
over the interval L x L + L tip can be formulated using
NELP
j Psr Prs
NDOF the characteristic motions of the tip of the beam. As men-
jP 1 s= 1 s= 1

tioned previously, note that since the tip mass offset is
Rload r= 1 r + j 2r r
Ds 2 2
s= 1
assumed to undergo rigid body motion, its motion depends

solely upon the tip of the beams transverse rectilinear and
rotation or slope motions. To this point, the transverse dis-
To formulate the optimal multimode FRF of power har- placement response of the tip mass offset can be formulated
vesting, the reduced optimal load resistance needs to be for- as
mulated rst. For this case, equation (58) can be differentiated
d w (L , t )
with respect to load resistance, and the differentiable power
function can be set to zero to give the optimal load resistance
( )
w L + L tip, t = w (L, t ) + L tip
dx
as
= e(L ) + L tip e a ( t )
d e (L )
X () 2 + Y () 2 dx (4 r ) (r 1)

(14)
opt
R load = , (59) (14)
X () 2 + Y () 2 NDOF

where
(
= e L + L tip a(t ) = ) (r 1)
er ( L + L tip ) ar (t ). (63)
(1 r ) r= 1
NELP NELP
Corresponding to equations (61) and (62), the FRF that
NELP NDOF
Psr Prs ( r2 2) relates the multimode transverse displacement to the input
s= 1 s= 1
X () = PDs 2
, base acceleration of the unimorph beam can be formulated as
s= 1 r= 1 ( r2 2 ) + ( 2r r )
2

NDOF
NELP NELP w (x , j) e r (x )
Psr Prs ( 2r r ) =
NDOF 2 wbase e j t r=1 r 2 + j 2 r r
2
s= 1 s= 1
Y () = 2
. (60)
r= 1 ( 2
r 2
) + ( 2r r )
2


NELP
jPDs + R L Q r


s=1
It should be noted that the optimal load resistance can be . (64)
NELP NELP
substituted back into equation (58) to give optimal power j sr Prs
P
harvesting. Moreover, the multimode FRF representing the NELP 1
NDOF

jPDs 2 s = 12 s=1

transverse displacement relative to the input transverse s=1 R load r=1 r + j 2r r
acceleration can be obtained as
For tip mass offset, the multimode transverse displace-
ar (j) 1
= 2 ment FRFs can also be formulated in terms of equations (61)
2 j t 2
wbase e r + j 2r r and (63) to give

(
w L + L tip, j ) = ( L + L )
NDOF e
r tip
NELP 2
wbase e j t 2 + j 2 2
r=1 r r
jPDs + R L Q r
r

s= 1
NELP NELP
. (61) NELP
sr Prs jPDs + R L Q r
NELP NDOF
j P s=1
1 . (65)
jPDs 2 s= 12 s= 1 NELP NELP
s= 1 Rload j Psr Prs
r = 1 r + j 2r r
NELP
1
NDOF

jPDs 2 s = 12 s=1

Before equation (61) is modied into the transverse s=1 R load r = 1 r + j 2r r

displacement FRF with respect to the input transverse accel-
eration at any position along the unimorph beam (x), the For experimental study, a laser Doppler vibrometer was
characteristic transverse motion of the unimorph beam can be used to capture dynamic responses from the small beam and

12
Smart Mater. Struct. 23 (2014) 095037 M F Lumentut and I M Howard

tip mass structures under input base motion, thus giving the mass offset with lt, ht and b (width) were set to 15 mm,
absolute motion. For this case, equations (64) and (65) need 10 mm and 6 mm, respectively.
to be modied in terms of the absolute motion. Since As shown in the experimental setup in gure 5, the
the combination of the base and relative motions denes cantilevered piezoelectric unimorph beam with the tip mass
the absolute motion, the multimode FRFs of absolute offset was clamped to the base structure under input
transverse displacement and velocity at any position along dynamic excitation. The input acceleration from the
the unimorph beam and the tip mass offset can be for- base structure was measured using a B&K accelerometer.
mulated as The B&K exciter connected to the power amplier was used
as the input dynamic excitation. The wave function gen-
wabs (x , j) wbase e jt + w (x , j) erator connected to the power amplier was used to control
=
2
wbase e j t
2 wbase e jt the input excitation using the sine sweep from the fre-
1 w (x , j) quencies of 5 Hz to 1000 Hz, with frequency increments of
= 2 + 0.3125 Hz and 40 averages controlled from the B&K Fast
2 wbase e jt
Fourier Transform (FFT) Pulse analyzer software, in order
d
wbase e + w (x , j) to obtain a very smooth frequency response and to avoid
j t
(x , j)
wabs d t aliasing components. Dynamic velocity of the tip mass
=
2 wbase e jt 2 wbase e jt offset was measured using a Polytec laser vibrometer with a
1 w (x , j) measurement range full scale of 500 mm s1 and a low pass
= + j lter setting of 22 kHz. Before operating the laser, a small
j 2 wbase e jt
reecting lm was attached to the top of the tip mass.
(
wabs L + L tip, j )=w base (
e jt + w L + L tip, j ) Moreover, all signal measurements from the accelerometer,
2 j t 2 j t
wbase e wbase e piezoelectric unimorph, and vibrometer were connected to
the sophisticated B&K FFT Analyzer 3560B data acquisi-
1
= +
(
w L + L tip, j ) tion device, which has ve channels with a maximum fre-
2
2 wbase e jt quency span up to 25.6 kHz. All processing signals through
d the analyzer displayed the measurement results using the

wabs (L + L tip, j d)t
j t
(
wbase e + w L + L tip, j ) FFT pulse software. Note that all sensitivities for scaling
=
2 wbase e j t 2 wbase e jt factors given from the B&K accelerometer of
9.633 mV m s2, and from the laser vibrometer of
=
1
+ j
(
w L + L tip, j
.
) (66) 125 mm s1 V (8 V m s1), were inserted into the FFT pulse
j 2 wbase e jt in order to obtain the correct results. Since the acceleration
is given in per-unit acceleration in m s2, all results of the
FRFs given from the experiment and the numerical analysis
in the next stage are given in per-unit acceleration in m s2.
5. Experimental validation: results and discussion Note that the B&K pulse software also provides the facility
(a unit organizer) for modifying the units of amplitude of
In this section, the multimode FRFs for the velocity, electrical FRF; the amplitudes from the laser measurement can be
voltage, current and power outputs using the new electro- stated in the velocity or displacement per-unit acceleration.
mechanical nite element equations are investigated and The rst four modes of the unimorph beam, as shown in
validated through experimental studies with variable load gure 6, can be used to identify the smoothness of mode
resistance (resistive shunt circuit). The input base transverse shapes using the different sizes of a number of discretized
acceleration onto unimorph structures was chosen to be elements. Although the resonant frequencies with discretiza-
1 m s2. The piezoelectric unimorph structure was made from tion using a lower number of elements seem to be quite close
PZT PSI-5A4E material (Piezo Systems Inc., Woburn, MA), to each other, the mode shapes seem to be quite coarse. These
and the rigid body of the offset tip mass was subsequently results show that the proposed 50 electromechanical dis-
attached. The geometrical structures of the piezoelectric cretized elements appear to be sufcient to provide accurate
unimorph with tip mass offset can be seen in gure 4, and results that can give smooth mode shapes, including the
details of the tip mass geometry, including formulas, can be natural frequencies. With the 50 electromechanical discretized
seen in appendix D. Note that, since the tip mass was elements, the strain node along the unimorph beam, as shown
glued on the top surface of the substructure, the extra in gure 7, can be determined using the inection point of the
substructure length, lt, was assumed to be a rigid body second order differential shape function of the straindis-
contributing to the tip mass offset. The complete properties placement relationship, as shown in equation (48). As the
of the unimorph are given in table 1. In gure 4, the surfaces of the piezoelectric material are covered with the
unimorph length, L, and width, b, with piezoelectric electrode layers, the inection point can be used to identify
thickness, hp, and substructure (brass) thickness, hs, were the location of strain nodes along the beam, indicating the
set to 60 mm, 6 mm, 0.127 mm and 0.5 mm, respectively. change of strains (tension and compressive strains) between
The tip mass congurations were calculated according to two segments of the piezoelectric structure. This can avoid the
the geometry and material properties. The dimensions of tip reduction of the power output at the second or higher modes.

13
Smart Mater. Struct. 23 (2014) 095037
Figure 4. Geometry of unimorph beam with tip mass offset.
Table 1. Properties of the piezoelectric unimorph system.
Material properties Piezoelectric Brass
Youngs modulus, c11 (GPa) 66 105
Density, (kg m3) 7800 9000
Piezoelectric constant, d31 (pm V1) 190
T
Permittivity, 33 (F m1) 1800 o
Permittivity of free space, o (pF m1) 8.854
Further details of the parametric strain node cases associated
with the experimental study, in terms of electrical connection
patterns, can be discussed in future investigations. Our main
purpose in this numerical study is to present the new elec-
tromechanical nite element vibration techniques and show
the validation with experimental studies.
In gure 8(a), the absolute tip velocity FRFs and
experimental results for the rst mode show very good
agreement under the variable load resistance. As we can see,
higher amplitudes can be achieved when load resistances and
frequencies shift from short circuits to open circuits. By
viewing gure 8(b), the maximum tip absolute velocity
responses at the short circuit and open circuit resonance fre-
quencies of 18.5 Hz and 18.9 Hz were reached for load
resistances approaching the lower and the higher values (from
short circuits to open circuit load resistances), respectively.
This behaviour can also be seen more obviously in
gure 8(c). In many cases of vibration, the maximum
dynamic displacement is generally avoided for most struc-
tures. It should be noted here that the power harvesting
response can be shown to give the best results without
having the maximum displacement, as we will discuss
further in the next section. Note that the resonance fre-
quency and amplitude response can shift due to resistive
shunt damping from the variable load resistance, where this
affects the electromechanical behaviour of the piezoelectric
element. In addition, the physical behaviour of the piezo-
electric unimorph, involving piezoelectric couplings and
internal capacitance, also adds electromechanical damping
and electromechanical stiffness. At this point, we realized
that the damping effects consist of the mechanical and
electrical components, due to the electromechanical beha-
viour of the dimensional structure, external load resistance
and material properties of the piezoelectric structure and
14
M F Lumentut and I M Howard
substructure. Identication of mechanical damping with
very low load resistance can be obtained by matching the
amplitude of the experimental and theoretical tip absolute
velocity and voltage FRFs. The mechanical damping ratio
of the rst mode, 1 = 0.0162, under a very low load
resistance of 562 was obtained (approaching the short
circuit resistance) because using the actual short circuit load
resistance (Rload = 0) will lead the theoretical voltage FRF to
be zero, where the tip absolute displacement cannot be
identied. This situation cannot be used to identify the
voltage and displacement FRF behaviour of the electro-
mechanical system.
With variation in the load resistance, the rst mode of
the voltage FRFs also varies while increasing load resis-
tance, as shown in gure 9(a). In this case, the resonance
frequency shifts to the higher value from short circuit to
open circuit load resistances, followed by an increase in the
amplitude. Note that when the voltage FRF with load
resistance is close to open circuit, the amplitude remains
constant.
Figure 9(b) shows the trend of voltage behaviour under
variable load resistance giving good agreement between
numerical and experimental results. In general form, the
voltage under short and open circuit resonances indicates a
slight increase with increasing load resistance. However, the
maximum open circuit resonance amplitude gives a higher
value compared to the short circuit resonance when load
resistance passes over the transitional or critical point of the
short circuit and open circuit amplitudes. Note that the critical
point of the amplitude response between short circuit and
open circuit resonance frequencies shows the same amplitude.
The change of frequency due to the variable load resistance
can be seen more clearly in gure 9(c). For electrical current
FRFs, as shown in gure 10(a), the trend shows the opposite
behaviour of that of the voltage FRFs. The electrical current
shows the highest amplitude when the load resistance
approaches the open circuit. Again, in the opposite trend to
the voltage response, the variation of electric current with
variable load resistance under short circuit and open circuit
resonance frequencies, as shown in gure 10(b), reaches the
maximum amplitude with decreasing load resistance. This
trend can be seen more obviously in the 3D current FRF
shown in gure 10(c). Another important aspect related to the
FRFs is the relationship between dynamic velocity, electrical
voltage and current. The absolute velocity amplitude with the
Smart Mater. Struct. 23 (2014) 095037
Figure 5. (a) Experimental setup and (b) piezoelectric unimorph beam with tip mass offset clamped on the base structure.
load resistance approaching the short circuit seems to
increase, whereas the voltage amplitude decreases with
increasing electrical current. Conversely, the velocity ampli-
tude with the load resistance approaching open circuit tends to
increase, whereas the voltage amplitude increases with
decreasing electrical current.
However, the power harvester FRFs, especially near the
rst mode response, as shown in gure 11(a), seem to behave
with different trends where the short circuit and open circuit
load resistances tend to decrease the amplitude. It can be said
that the highest velocity amplitudes at the short circuit and
open circuit load resistances do not show the highest power
amplitudes. In different congurations with velocity, voltage
and current, the variations in power harvesting under short
circuit and open circuit resonance frequencies, as seen in
gure 11(b), show two different optimum load resistances,
giving the optimum amplitudes. In the 3D graph shown in
gure 11(c), the optimum power harvesting amplitudes can
also be seen clearly at the regions away from the short circuit
and open circuit load resistances.
Optimal power harvesting, based on the optimal load
resistance, can also be seen in gure 12(a), where the absolute
maximum power amplitude point coincides with the load
resistance of 227 k at the resonance frequency of 18.70 Hz.
However, the optimal load resistance in the off-resonance
regions seem to approach and overlap with the load resis-
tances of 155 k and 332 k , where the experimental results
are also very close to the off-resonance regions, except for the
driving frequency below 8 Hz due to unwanted noise. By
scrutinizing the optimal power harvesting response, the
numerical modelling result seen in gure 12(b) shows that
ignoring the rotary inertia of the tip mass results in the shift of
the resonance frequency to a higher value. The resonance
difference is also more pronounced if the model ignores the
offset distance of the tip mass measured from its centroid to
the end of the unimorph beam length. Neglecting both the
offset and rotary inertia of the tip mass results in a larger
15
M F Lumentut and I M Howard
modelling error. It can be seen that when designing either a
micro- or mesoscale power harvesting device, the mathema-
tical model should include the effect of the offset and rotary
inertia of the tip mass.
6. Conclusions
The new numerical study proposed in this paper addresses
ve important points.
The electromechanical discretization was numerically
developed, since the actual piezoelectric material covered
by thin electrode layers only generates a single voltage
across the electrical load. The electromechanical dis-
cretization introduced here consists of the mechanical
discretized elements for elemental displacement elds
and the parallel electrical discretized elements for multi-
output electrical current. The previous published papers
related to the piezoelectric nite element have not
considered this in their mathematical modelling, although
many of them showed different applications.
Instead of developing the kinematic equations of the
unimorph beam, the tip mass offset was formulated, since
it contributes signicantly to the mass matrices and the
input dynamic force vector. The benet of locating the tip
mass with its centroid away from the end of the
piezoelectric beam is that it allows for the modelling of
various applications, especially MEMS power harvesters.
The improved designs avoid contact between the brittle
piezoelectric material and the tip mass. The previous
published papers related to MEMS have ignored this in
their mathematical modelling.
The numerical techniques provide physical interrelation-
ship between the electrode effect, coupled eld
equations, kinematic equations, electromechanical dis-
cretization and electromechanical dynamic equations.
Smart Mater. Struct. 23 (2014) 095037 M F Lumentut and I M Howard

Figure 6. Normalized displacement mode shape with (a) ve elements, (b) 10 elements, (c) 30 elements and (d) 50 elements.

Figure 7. Normalized strain mode shape with 50 elements.

The reduced constitutive electromechanical nite element multimode electromechanical FRFs. Unlike analytical
equations were further derived into global normalized scalar techniques, the proposed new numerical solution techniques
form of the electromechanical dynamic equations, in order provided the benet of analyzing the structure with different
to formulate the strain mode shape, eigenmode shape and scalabilities, including MEMS devices.

16
Smart Mater. Struct. 23 (2014) 095037
Figure 8. Tip absolute velocity FRF with (a) numerical (solid lines) and experimental results (round dots), (b) variable load resistances under
the short circuit and open circuit resonance frequencies, and (c) 3D analysis.
The comparisons between the numerical solution techni-
ques and experimental studies have been discussed in
terms of the variable load resistance, giving good
agreement.
Overall, the experimental verication and parametric
numerical case studies have been explored and have shown
close agreement. For this paper, the proposed new electro-
mechanical nite element modelling has proved that the
velocity, current and power FRFs have shown very similar
trends with the previous published analytical literature.
These numerical results were also validated with experi-
mental studies. Moreover, since the trend of optimal power
harvesting FRF using different cases of the tip mass was
presented, the new method showed that neglecting the
rotary inertia of the tip mass resulted in a resonance fre-
quency error. Notably, modelling errors occurred from
ignoring the tip mass offset and ignoring both tip mass
offset and rotary inertia. The other new aspect shown in
17
M F Lumentut and I M Howard
using these numerical techniques is that the system models
congured normalized strain node and eigenmode shapes
using discretized elements for convergence studies. Further
details of the frequency analysis using the strain node effect
in terms of the electrical circuit connection patterns will be
a focus of future studies.
The new numerical techniques can be a very useful tool
for analyzing FRFs, displacement mode shape and strain node
forms. These techniques can be applied to model different
geometrical aspects for laminated structures and MEMS
power harvester devices. In this case, since our main concern
is presenting the new contribution of numerical techniques, it
is important to validate the results obtained from numerical
FRFs using experimental studies where the trends of the
results were also shown to be very similar with facts estab-
lished in previous analytical literature. Further applications
for new aspects of power harvesting using the proposed
numerical techniques can be demonstrated in future research
studies.
Smart Mater. Struct. 23 (2014) 095037 M F Lumentut and I M Howard

Figure 9. Voltage FRF with (a) numerical (solid lines) and experimental results (round dots), (b) variable load resistances under the short
circuit and open circuit resonance frequencies and (c) 3D analysis.

Appendix A. Determining the asymmetric elastic stiffness and piezoelectric elastic stiffness at a constant
neutral axis electric eld, respectively.

In gure A1, the location of the asymmetric neutral axis,

measured from the y-axis to the top surface of the piezo- Appendix B. Determining the kinetic energy reduced
electric layer, can be determined using the resultant force from the kinematic equations
balance in the cross section of the unimorph structure to give
Corresponding to equations (7) and (11), the total kinetic
zn h p b
energy of the system consisting of the elemental
F (i )
= h + (z
s n hp )
0 c11(1) z dy dz
unimorph and tip mass offset, as shown in gure 3, can be
i=1
zn b formulated as
+ z n hp
0 c11(2) zdydz
KE = KE unimorph + KE tip, (B.1)
= 0. (A.1) where

Manipulating equation (A1), the location of the asym- 1 pp (x , z , t ) R

pp (x , z , t ) dm (1)
metric neutral axis can be reduced as
KE unimorph =
2
m (1)
R

1 pp (x , z , t ) R
pp (x , z , t ) dm (2)
c11(1) h s2 + c11(2) h p2 + 2c11(1) h s h p
+
2
m (2)
R
zn = , (A.2)
1
(
2 c11(1) h s + c11(2) h p ) KE tip =
2
mtip
mm ( x e + 1, z , t ) R
R mm ( x e + 1, z , t ) dm tip . (B.2)

(1)
where coefcients c11 and c11E = c11
(2)
represent substructure

18
Smart Mater. Struct. 23 (2014) 095037 M F Lumentut and I M Howard

Figure 10. Electrical current FRF with (a) numerical (solid lines) and experimental results (round dots), (b) variable load resistances under the
short circuit and open circuit resonance frequencies, and (c) 3D analysis.

As mentioned previously, the translation and angular section, as given in detail in appendix C. For the tip mass
velocities of tip mass at the end of the unimorph beam length, kinetic energy, the second part of equation (B.2) can be
L, as shown in gure 1, can be transformed into the local expanded as
position of the end of the element, x e + 1, as shown in gure 3.
The rst part of equation (B.2) can be expanded as 1
KE tip = (
2 mtip
wbase (t ) wbase (t ) + 2xm ( x e + 1, t ) wbase (t )
KE unimorph + 2xm ( x e + 1, t ) w ( x e + 1, t ) + 2xc xm ( x e + 1, t ) ( x e + 1, t )
1
= ( wbase (t ) wbase
(t ) + 2wbase (t ) w (x , t ) + 2z c z m ( x e + 1, t ) ( x e + 1, t ) + 2xc wbase
( t ) ( x e + 1, t )
2 m (1)
(t ) w ( x e + 1, t ) + w ( x e + 1, t ) w ( x e + 1, t )
+ 2wbase
)
+ z 2 (x , t ) (x , t ) + w (x , t ) w (x , t ) dm (1)
+ 2xc ( x e + 1, t ) w ( x e + 1, t ) + x c2 ( x e + 1, t ) ( x e + 1, t )
1
+
2 m (2)
( wbase (t ) wbase (t ) + 2wbase
(t ) w (x , t ) + z c2 ( x e + 1, t ) ( x e + 1, t ) + z m2 ( x e + 1, t ) ( x e + 1, t )
)
+ z 2 (x , t ) (x , t ) + w (x , t ) w (x , t ) dm (2) (B.3) )
+ x m2 ( x e + 1, t ) ( x e + 1, t ) dm tip . (B.4)

Note that the rst terms for each parenthesis from
equation (B.3) must be zero, due to the Lagrangian
functional form of the system only depending upon the
relative translation and velocity of the unimorph, as
expressed in section 4.2. The second terms for each par-
enthesis contribute to the input dynamic forces, and these
terms can be moved into the nonconservative work
Again, the rst term of equation (B.4) must be zero.
Moreover, the second to fth terms must be zero because they
contain the rst mass moment of inertias for the tip mass body
about the centroid at point g, as shown in gure 3, since the
centroid of the tip mass is located on that centroid itself [57].
Again, the sixth and seventh terms can be moved into the
nonconservative work section.

19
Smart Mater. Struct. 23 (2014) 095037 M F Lumentut and I M Howard

Figure 11. Power FRF with (a) numerical (solid lines) and experimental results (round dots), (b) variable load resistances under the short
circuit and open circuit resonance frequencies and (c) 3D analysis.

Equations (B.3) and (B.4) can simply be rewritten in The zeroth and second mass moment of inertias of the tip
matrix form as mass offset can be stated as

KE unimorph =
1
x
x e +1
z 2 (1) (x , t ) T (x , t ) dA (1) dx
I0tip = V tip
tip dV tip,
2 e A (1)

+
1
x
x e +1
z 2 (2) (x , t ) T (x , t ) dA (2) dx
I2tip =
V tip
( ) (
tip z m2 + x m2 dV tip+ z c2 + x c2 m tip )
2 A (2)

1
e
x e +1 (
= Icgtip + m tip z c2 + x c2 , ) (B.6)
+
2
x A (1)
(1) w (x , t ) T w (x , t ) dA (1) dx
e where dV tip = dxm dym dz m and denite integral forms of f(y, z)
1 x e +1
implied from equation (B.5) can be dened as
+
2
x e
A (2)
(2) w (x , t ) T w (x , t ) dA (2) dx
zn h p b

KE tip
= I0tip xc w ( x e + 1, t ) ( x e + 1, t )
T A (1)
f (y , z) dA (1) = h + (z s n

h ) 0
p
f (y , z ) d y d z ,
1 zn b
+ I0tip w ( x e + 1, t )T w ( x e + 1, t )
2 A (2)
f (y , z) dA (2) = z n hp
0 f (y , z ) d y d z ,
1
+ I2tip ( x e + 1, t )T ( x e + 1, t ). (B.5)
2 {
f (y , z) 1, z , z 2 . } (B.7)

20
Smart Mater. Struct. 23 (2014) 095037 M F Lumentut and I M Howard

Figure 12. Identication of the optimal power FRF with (a) numerical (dashed lines) and experimental results (round dots), based on the
optimal load resistance and related to other load resistances and (b) different cases of tip mass offset.

Note that equation (B.7) can be used for any operational

integrations, not only in the kinetic energy derivation, but also
for other Lagrangian energy forms given from
equations (36)(39).

Appendix C. Determining the nonconservative work

from the input base motion

As mentioned in appendix B, the contributions of non-

conservative forces due to the input base excitation on the
system were reduced from the kinetic energy. As a result, the
Figure A1. Cross-section of piezoelectric unimorph beam. virtual work done by nonconservative forces can simply be

21
Smart Mater. Struct. 23 (2014) 095037 M F Lumentut and I M Howard

formulated in terms of their time-dependent response, using

the Hamiltonian principle as

t2
WF base = t 1
F . r dt . (C.1)

Note that from a mathematical standpoint, the functional

form of Lagrangians principle is reduced from the Hamil-
tonian principle using the variational method. Therefore,
corresponding to equation (39), the virtual nonconservative
works can be expanded by using the second terms for each
parenthesis from equation (B.3) and the sixth and seventh
terms from equation (B.4), as
Figure D1. Geometry of tip mass offset.
1 t2
WF base =
2
t 1 m (1)
2wbase (t ) w (x , t ) dm (1) dt reduced to give
l2 + h2
+ t
t2
base (t ) w
2w (x , t ) dm (2) dt I2tip = tip blt h t
( t t ht
+
) 2
h p zc + z n
m (2) 12 2

1
t2
+ t mtip
base (t ) ( x e + 1, t ) dm tipdt
2x c w l2 + h2
( hs )
2
1
(1)
+ blt h s t s
+ + h p z n + zc
t2 12 2
+ t base (t ) w
2w ( x e + 1, t ) dm tipdt . (C.2)
1 m tip

(
+ I0tip zc2 + xc2 , ) (D.2)
Equation (C.2) needs to be further formulated by
where the offset distances measured from the tip mass cen-
applying partial integration, and the result can be stated in the
troid to the point d in the x- and z-axes can be formulated,
matrix form as respectively, as
t2 x e +1
tip blt2 h t + (1) blt2 h s
WF base = t x A (1)
(1) wbase (t ) w (x , t ) T dA (1) dx dt x c = (D.3)
1
t2
e
x e +1
(
2 tip blt h t + (1) blt h s )
t x
1 e
A (2)
(2) wbase (t ) w (x , t ) T dA (2) dx dt
z c=
t2 ht
t I0tip xc wbase (t ) ( x e + 1, t ) T dt tip blt h t
2
+ z n h p + (1) blt z n h p
( )
1

t2 zn h p hs + h p z n
t I0tip wbase (t ) w ( x e + 1, t ) T dt . (C.3) (
(1) blt h s + h p z n )
1
2 2
.(D.4)
( ) (
tip blt h t + (1) blt z n h p + (1) blt h s + h p z n )
Alternatively, and with the same result, the second mass
moment of inertia of tip mass offset at the point d, located at
Appendix D. Determining the geometry parameters the end of the unimorph beam, can straightforwardly be for-
of tip mass offset mulated as
l2 + h2
Corresponding to the parameters given by equation (B.6), the (
I2tip = tip bl t h t
t t )
+ zn hp +
h t 2 l t 2
+
mass moment of inertias of the tip mass offset, as shown in 12 2 2

gure 4, can be formulated. Note that as mentioned pre- l2 + h2
viously, the extra unimorph beam length, as shown in +(1) bl t h s
( t s )
l t 2 h s 2
z n h p . (D.5)
gure D1, also contributed to the tip mass offset. In this case, 12
+ +
2 2 ( )

the zeroth mass moment of inertia can be stated as
References
I0tip = tip blt h t + (1) blt h s , (D.1)
[1] Zuo L and Tang X 2013 Large-scale vibration energy
and the second mass moment of inertia of tip mass offset at harvesting J. Intell. Mater. Syst. Struct. 24 140530
[2] Sue C-Y and Tsai N-C 2012 Human powered MEMS-based
the point d, located at the end of the unimorph beam, can be
energy harvest devices Appl. Energy 93 390403

22
Smart Mater. Struct. 23 (2014) 095037
[3] Kim H U, Lee W H, Dias H V R and Priya S 2009
Piezoelectric microgenerators-current status and challenges
IEEE Trans. Ultrason. Ferroelectr. Freq. Control 56
155568
[4] Priya S, Ryu J, Park C-S, Oliver J, Choi J-J and Park D-S 2009
Piezoelectric and magnetoelectric thick lms for fabricating
power sources in wireless sensor nodes Sensors 9 636284
[5] Hande A, Bridgelall R and Zoghi B 2010 Vibration energy
harvesting for disaster asset monitoring using active RFID
tags Proc. IEEE 9 16208
[6] Chao P C-P 2011 Energy harvesting electronics for vibratory
devices in self-powered sensors IEEE Sensor J. 11 310621
[7] Kim H S, Kim J-H and Kim J 2011 A review of piezoelectric
energy harvesting based on vibration Int. J. Prec. Eng.
Manufact. 12 112941
[8] Roundy S and Wright P K 2004 A piezoelectric vibration
based generator for wireless electronics Smart Mater. Struct.
18 113142
[9] Takeda H, Mihara K, Yoshimura T, Hoshina T and Tsurumi T
2011 Effect of material constants on power output in
piezoelectric vibration-based generators IEEE Trans.
Ultrason. Ferroelectr. Freq. Control 58 18529
[10] Liang J and Liao W-H 2012 Impedance modeling and analysis
for piezoelectric energy harvesting systems IEEE/ASME
Trans. Mechatronics 17 114557
[11] duToit N E, Wardle B L and Kim S 2005 Design
considerations for MEMS scale piezoelectric mechanical
vibration energy harvesters Int. J. Integr. Ferroelectr. 71
12160
[12] Liao Y and Sodano H 2008 Model of a single mode energy
harvester and properties for optimal power generation Smart
Mater. Struct. 17 065026
[13] Goldschmidtboeing F and Woias P 2008 Characterization of
different beam shapes for piezoelectric energy harvesting
J. Micromech. Microeng. 18 104013
[14] Kim M, Hoegen M, Dugundji J and Wardle B L 2010
Modeling and experimental verication of proof mass
effects on vibration energy harvester performance Smart
Mater. Struct. 19 045023
[15] Erturk A 2012 Assumed-modes modeling of piezoelectric
energy harvesters: EulerBernoulli, Rayleigh, and
Timoshenko models with axial deformations Comp. Struct.
106-107 21427
[16] Erturk A and Inman D J 2009 An experimentally validated
bimorph cantilever model for piezoelectric energy harvesting
from base excitations Smart Mater. Struct. 18 025009
[17] Wickenheiser A M and Garcia E 2010 Broadband vibration-
based energy harvesting improvement through frequency
up-conversion by magnetic excitation Smart Mater. Struct.
19 065020
[18] Lumentut M F and Howard I M 2013 Analytical and
experimental comparisons of electromechanical vibration
response of a piezoelectric bimorph beam for power
harvesting Mech. Syst. Signal Proc. 36 6686
[19] Lumentut M F and Howard I M 2011 Analytical modeling of
self-powered electromechanical piezoelectric
bimorph beams with multidirectional excitation Int. J. Smart
Nano Mater. 2 13475
[20] Lumentut M F and Howard I M 2013 Electromechanical
piezoelectric power harvester frequency response modelling
using closed-form boundary value methods IEEE/ASME
Trans. Mechatronics 19 3244
[21] Lumentut M F, Francis L A and Howard I M 2012 Analytical
techniques for broadband multielectromechanical
piezoelectric bimorph beams with multifrequency power
harvesting IEEE Trans. Ultrason. Ferroelectr. Freq. Control
59 155568
[22] Allik H and Hughes T J R 1970 Finite element method for
piezoelectric vibration Int. J. Numer. Methods Eng. 2 1517
23
M F Lumentut and I M Howard
[23] Allik H, Webman K M and Hunt J T 1974 Vibrational
response of sonar transducers using piezoelectric nite
elements J. Acoust. Soc. Am. 56 178291
[24] Kagawa Y and Yamabuchi T 1976 Finite element approach for
a piezoelectric circular rod IEEE Trans. Sonics Ultrason. 23
37984
[25] Naillon M, Coursant R H and Besner F 1983 Analysis of
piezoelectric structures by a nite element method Acta
Electron. 25 34162
[26] Tzou H S and Tzeng C I 1983 Distributed piezoelectric sensor/
actuator design for dynamic measurement/control of
distributed parameter system: a piezoelectric nite element
approach J. Sound Vib. 138 1734
[27] Saravanos D A, Heyliger P R and Hopkins D A 1996
Layerwise mechanics and nite element for the dynamics
analysis of piezoelectric composite plate Int. J. Solids Struct.
34 35978
[28] Moita J M, Correia I F P and Soares C M M 2004 Active control
of adaptive laminated structures with bounded piezoelectric
sensors and actuators Comp. Struct. 82 134958
[29] Lam K Y, Peng X Q, Liu G R and Reddy J N 1997 A nite-
element model for piezoelectric composite laminates Smart
Mater. Struct. 6 58391
[30] Gaudenzi P, Carbonaro R and Benzi E 2000 Control of beam
vibrations by means of piezoelectric devices: theory and
experiments Compos. Struct. 50 3739
[31] Detwiler D T, Shen H M-H and Venkayya V B 1995 Finite
element analysis of laminated composite structures
containing distributed piezoelectric actuators and sensors
Fin. Elem. Anal. Des. 20 87100
[32] Fernandes A and Pouget J 2003 Analytical and numerical
approach to piezoelectric bimorph Int. J. Solids Struct. 40
433152
[33] Wang S Y 2004 A nite element model for the static and
dynamic analysis of a piezoelectric bimorph Int. J. Solids
Struct. 41 407596
[34] Thomas O, De J-F and Ducarne J 2009 Vibrations of an
elastic structure with shunted piezoelectric patches: efcient
nite element formulation and electromechanical coupling
coefcients Int. J. Numer. Methods Eng. 8 23568
[35] Lumentut M F, Teh K K and Howard I M 2008 Computational
FEA model of a coupled piezoelectric sensor and plate
structure for energy harvesting Aust. J. Mech. Eng. 5
199208
[36] DeMarqui C Jr, Erturk A and Inman D J 2009 An
electromechanical nite element model for piezoelectric
energy harvester plates J. Sound Vib. 327 925
[37] Aladwani A, Arafa M, Aldraihem O and Baz A 2012
Cantilevered piezoelectric energy harvester with a dynamic
magnier ASME J. Vib. Acoust. 134 031004
[38] Zhu M, Worthington E and Njuguna J 2009 Analyses of power
output of piezoelectric energy-harvesting devices directly
connected to a load resistor using a coupled piezoelectric-
circuit nite element method IEEE Trans. Ultrason.
Ferroelectr. Freq. Control 56 130918
[39] Yang Y and Tang L 2009 Equivalent circuit modeling of
piezoelectric energy harvesters J. Intell. Mater. Syst. Struct.
20 222335
[40] Tiersten H F 1969 Linear Piezoelectric Plate Vibrations (New
York: Plenum)
[41] Nye J F 1984 Physical Properties of Crystals: Their
Representation By Tensors and Matrices (Oxford:
Clarendon)
[42] Tanaka H 1993 Generalized basic equations for bending
motions of piezoelectric bars formulated from Hamiltons
principle J. Acoustic. Soc. Am. 4 176472
[43] Standards Committee of the IEEE Ultrasonics, Ferroelectrics,
and Frequency Control Society 1987 IEEE Standard on
Piezoelectricity (New York: IEEE)
Smart Mater. Struct. 23 (2014) 095037
[44] Lumentut M F 2011 Mathematical dynamics of
electromechanical piezoelectric energy harvester PhD Thesis
Curtin University (Australia) http://espace.library.curtin.edu. [50]
au/R/?func=dbin-jump-full&object_id=186549&local_
base=gen01-era02
[45] Jeon Y B, Sood R, Jeong J-H and Kim S-G 2005 MEMS
power generator with transverse mode thin lm PZT Sens. [51]
Actuators A 122 1622
[46] Fang H B, Liu J Q, Xu Z Y, Dong L, Wang L, Chen D, [52]
Cai B C and Liu Y 2006 Fabrication and performance of
MEMS-based piezoelectric power generator for vibration [53]
energy harvester Microelectron. J. 37 12804
[47] Elfrink R, Kamel T M, Goedbloed M, Matova S, Hohlfeld D,
van Andel Y and van Schaijk R 2009 Vibration energy [54]
harvesting with aluminum nitride-based piezoelectric
devices J. Micromech. Microeng. 19 094005 [55]
[48] Andosca R, McDonald T G, Genova V, Rosenberg S, Keating J,
Benedixen C and Wu J 2012 Experimental and theoretical [56]
studies on MEMS piezoelectric vibrational energy harvesters
with mass loading Sens. Actuators A 178 7687 [57]
[49] Miller L M, Halvorsen E, Dong T and Wright P K 2011
Modeling and experimental verication of low-frequency
24
M F Lumentut and I M Howard
MEMS energy harvesting from ambient vibrations
J. Micromech. Microeng. 21 045029
Park J C, Park J Y and Lee Y-P 2010 Modeling and
characterization of piezoelectric d33-mode MEMS energy
harvester IEEE/ASME J. Microelectromech. Syst. 19
121522
Dankert H and Dankert J 1995 Technische mechanik vol 2
(Stuttgart: Teubner)
Ikeda T 1990 Fundamentals of Piezoelectricity (New York:
Oxford University Press)
Zienkiewicz O C and Taylor R L 2005 The Finite Element
Method for Solid and Structural Mechanics 6th edn (UK:
Butterworth-Heinemann)
Weaver W W Jr and Johnston P R 1984 Finite Elements for
Structural Analysis (US: Prentice-Hall)
Reddy J N 1984 An Introduction to the Finite Element Method
(New York: McGraw-Hill)
Meirovitch L 1967 Analytical Methods in Vibrations (US:
Prentice Hall)
Beer F P and Johnston E R Jr 1977 Vector Mechanics for
Engineers: Statics and Dynamics (New York: McGraw-Hill)